Annalisa Coliva - Mind, Meaning, and Knowledge - Themes From The Philosophy of Crispin Wright (2012, Oxford University Press) PDF

Mind, Meaning, and Knowledge
This page intentionally left blank

Mind, Meaning, and
Knowledge
Themes from the Philosophy of
Crispin Wright
Edited by Annalisa Coliva
1
3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
# the several contributors 2012
The moral rights of the authors have been asserted
First Edition published in 2012
Impression: 1
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Data available
ISBN 978–0–19–927805–3
Printed in Great Britain by
MPG Books Group, Bodmin and King’s Lynn
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Contents
Notes on the Contributors vii
Introduction 1
Bio-bibliographical Note 15
Section I. Rule-Following and the Normativity of Meaning
1. Blind Rule-Following 27
Paul A. Boghossian
2. Understanding and Rule-Following 49
Christopher Peacocke
3. Regularities, Rules, Meanings, Truth-Conditions,
and Epistemic Norms 77
Paul Horwich
4. Why Meaning Intentions are Degenerate 96
Akeel Bilgrami
Section II. Knowledge of Our Own Minds and Meanings

5. The Publicity of Meaning and the Interiority of Mind 127
Barry C. Smith
6. Expression, Truth, and Reality: Some Variations
on Themes from Wright 162
Dorit Bar-On
Section III. Truth, Objectivity, and Relativism

7. Some Remarks about Minimalism 195
Simon Blackburn
8. Objectivity, Explanation, and Cognitive Shortfall 211
Stewart Shapiro
9. How to Formulate Relativism 238
Carol Rovane
vi CONTENTS
Section IV. Warrant, Transmission Failure, and Scepticism

10. When Warrant Transmits 269
James Pryor
11. Wright on Moore 304
José L. Zalabardo
12. Moore’s Proof, Liberals, and Conservatives – Is
There a (Wittgensteinian) Third Way? 323
Annalisa Coliva
13. Wright Against the Sceptics 352
Michael Williams
Replies
Crispin Wright
Foreword 377
Part I: The Rule-Following Considerations and the Normativity
of Meaning 379
Part II: Knowledge of Our Own Minds and Meanings 402
Part III: Truth, Objectivity, Realism, and Relativism 418
Part IV: Warrant Transmission and Entitlement 451
Index Nominum 487

Index 489
Notes on the Contributors
Dorit Bar-On is Professor of Philosophy at the University of North Carolina at

Chapel Hill. She is the author of Speaking My Mind: Expression and Self-Knowledge
(OUP 2004). She is presently working on a book manuscript, tentatively titled
Expression, Action, and Meaning as part of a large-scale collaborative project on
Expression, Communication and the Origins of Meaning, as well as on another
book manuscript (with Keith Simmons) If Truth be Told.
Akeel Bilgrami is Johnsonian Professor of Philosophy and Director of the Heyman
Center for the Humanities at Columbia University. He is the author of Belief and
Meaning (Blackwell 1992), of Self-Knowledge and Resentment (Harvard 2006), and of
Politics and the Moral Psychology of Identity (Harvard forthcoming). He is presently
working on a long project on the relationship between value, reason, and agency.
Simon Blackburn is Professor of Philosophy at the University of North Carolina at
Chapel Hill and Fellow of Trinity College Philosophy at the University of Cambridge.
His recent books include Lust (OUP 2004), Truth: A Guide (OUP 2005), and Plato’s
Republic: A Biography (Atlantic Monthly Press 2006).
Paul A. Boghossian is Silver Professor of Philosophy at New York University and
Director of the New York Institute of Philosophy. He is the author of Fear of Knowledge:
Against Relativism and Constructivism (OUP 2006). A collection of his essays appeared
under the title Content and Justification: Philosophical Essays (OUP 2008); he is editor
(with Christopher Peacocke) of New Essays on the A Priori (OUP 2000).
Annalisa Coliva is Associate Professor at the University of Modena and Reggio
Emilia and associate director of COGITO Research Centre in Philosophy (Bologna).
Her recent publications include Moore and Wittgenstein: Scepticism, Certainty and
Common Sense (Palgrave Macmillan 2010) and, as editor, The Self and Self-Knowledge
(OUP 2012). She is presently working on a book tentatively titled Within a System:
A Plea for Moderatism and (with Danièle Moyal-Sharrock) on a collected volume titled
Hinge Epistemology: Basic Beliefs After Moore and Wittgenstein.
Paul Horwich is Professor of Philosophy at New York University. His most recent
books include From a Deflationary Point of View (OUP 2004), Reflections on Meaning
(OUP 2005), Truth–Meaning–Reality (OUP 2010), and Wittgenstein’s Meta-Philosophy
(OUP 2012).
Christopher Peacocke is Professor of Philosophy at Columbia University and holds
the Richard Wollheim Chair of Philosophy at University College London. His most
recent books include Being Known (OUP 1999), The Realm of Reasons (OUP 2003), and
viii NOTES ON THE CONTRIBUTORS
Truly Understood (OUP 2008). He is currently completing a book on the first person,
the self, and self-representation.
James Pryor is Associate Professor at New York University. He is the author of several
papers in epistemology, philosophy of language, and philosophy of mind.
Carol Rovane is Professor of Philosophy at Columbia University. She is the author of
The Bounds of Agency: An Essay in Revisionary Metaphysics (Princeton 1998) and of For
and Against Relativism (forthcoming).
Stewart Shapiro is O’Donnell Professor of Philosophy at the Ohio State University
and Arché Professorial Fellow at the University of St Andrews. He is the author of
Foundations without Foundationalism: A Case for Second-Order Logic (OUP 1991),
Philosophy of Mathematics: Structure and Ontology (OUP 1997), Thinking about
Mathematics: The Philosophy of Mathematics (OUP 2000), and Vagueness in Context
(OUP 2006).
Barry C. Smith is Professor of Philosophy at Birkbeck College London and Director
of the Institute of Philosophy, School of Advanced Study, at University of London. He
is the editor (with Ernest Lepore) of The Oxford Handbook of Philosophy of Language
(OUP 2006), and (with Cynthia MacDonald and Crispin Wright) of Knowing Our Own
Minds (1998 Clarendon), and of Questions of Taste (OUP 2007).
Michael Williams is Krieger-Eisenhower Professor and Chair of the Department of
Philosophy at Johns Hopkins University at Baltimore. In addition to numerous articles,
he is the author of Groundless Belief (Princeton 1997), Unnatural Doubts (Princeton
1992), and Problems of Knowledge (OUP 2001).
José L. Zalabardo is Reader in Philosophy at University College London. In addition
to many articles, his main publications include Introduction to the Theory of Logic
(Westview 2000), Scepticism and Reliable Belief (OUP forthcoming), and, as editor,
Wittgenstein’s Early Philosophy (OUP forthcoming).
Introduction
Anyone who has been fortunate enough to benefit from philosophical interaction
with Crispin Wright knows his incredible talent for engaging with the deepest aspects
of the subject and his generosity—especially with colleagues and students—in the
time and energy he devotes to the discussion of their views. Indeed, this latter aspect
of his approach to the discipline has given rise, in the last decade or so, to the creation
of two research centres—Arché at St Andrews and the Northern Institute of Philosophy at
Aberdeen—that have had vast impact for the practice of analytic philosophy in
Europe. None of this would have been possible, however, if not for Wright’s
enormous intellectual influence, due to his seminal contributions in a number of
areas as diverse as the philosophy of language and mind, the philosophy of mathe-
matics and logic, metaphysics and epistemology, and the study of Frege and Witt-
genstein. It is indeed a mark of the exceptional breadth of his philosophical interests,
and the impact that his ideas have had in all these fields of philosophical inquiry, that
Oxford University Press felt it fitting to publish not one but two volumes of original
papers in his honour. Alex Miller’s edition of the sequel to this volume, devoted to
discussion of Wright’s work on the philosophies of logic, language, and mathematics,
will be keenly anticipated.
This, the first volume, covers four other principal areas of concentration in
Wright’s philosophy: first, his contribution to the launch of and development of
the late twentieth-century debates concerning rule-following, with their conse-
quences for the notion of meaning; second the issue, left open by his treatment of
the rule-following problem, of how knowledge of our own mental states—in
particular of our intentions—is possible; third, Wright’s distinctive approach to the
metaphysics of realism, comprising minimalist treatments of truth and assertoric
content, and its significance for the notion of objectivity and the problem of relativ-
ism; and finally, his contributions to epistemology, especially with respect to the
assessment of Moore’s “Proof” and to the problem of responding to the traditional
sceptical challenges concerning our putative knowledge of the existence of a material
world. Accordingly, the essays to follow have been divided into four corresponding
sections.
2 M I N D, M E A N I N G , A N D K N OW L E D G E
Section I: Rule-Following and the

Normativity of Meaning
Since his Wittgenstein on the Foundations of Mathematics (1980),1 the so-called “rule-
following considerations” in Wittgenstein’s Remarks on the Foundations of Mathematics2
and Philosophical Investigations,3 have been a central concern of Wright’s reflection. For,
in his view, that problem lies at the heart of any aspect of human life in which there is a
need to distinguish between correct and incorrect judgment and practice. In the
book, which predates Saul Kripke’s enormously influential Wittgenstein on Rules and
Private Language,4 Wright was concerned to connect the ideas about rules with the
undermining of Platonism and a repudiation of investigation-independent truth-
values. Later on, in a number of papers published afterwards and now collected in
his Rails to Infinity (2001),5 he maintained an Intention view of rule-following. While
agreeing with Kripke’s Wittgenstein’s rejection of dispositional accounts of rule-
following, Wright maintained that the Intention view could be successfully embraced.
Accordingly, to follow a rule would consist in forming an intention that has that rule as
its content and in behaving in accordance with it. The real issue would then be to
explain how we can have immediate and authoritative access to our intentions. But
that problem—the problem of how at least that kind of self-knowledge is so much as
possible—has to do with the epistemology of mind (if indeed there is one, as we shall
see in Section II) rather than with issues related to linguistic and mental content.
So much for a statement of Wright’s original account of rule-following. However, it
must be registered that in recent years he has gone back to this issue to express at least
partial dissatisfaction with the Intention view.6 The main problem with it is that its
application conforms to the “modus ponens” model, according to which one must be
able to entertain a general formulation of the rule, couched in terms of a conditional,
recognize that the conditions for the application of the rule, expressed in the anteced-
ent, obtain and then go on to apply the rule. According to Wright’s more recent
position, that model cannot always be applied, because it already requires a conceptual
repertoire—viz. the one which is needed to grasp the content of the conditional and to
recognize the satisfaction of its antecedent. This, however, from the standpoint of the
Philosophical Investigations, is a serious mistake. For grasp of concepts cannot in general
be anterior to the ability to give them linguistic expression but rather resides in that
very ability. So it seems it must be recognized either that there is a form of basic
rule-governed behavior which is “blind” because it doesn’t depend on the
1
Cambridge, Mass., Harvard University Press.
2
Oxford, Blackwell, 1956.
3
Oxford, Blackwell, 1953.
4
Cambridge, Mass., Harvard University Press, 1982.
5
6
See C. Wright, 2007 “Rule-following without Reasons: Wittgenstein’s Quietism and the Constitutive
Question”, Ratio 20/4, pp. 481–502.
I N T RO D U C T I O N 3
appreciation of facts about what rules require, or that basic judgment is not a matter
of rule-following.
In his essay “Blind rule-following”, Paul Boghossian challenges the Intention view
at its heart. He contends that even if a number of criticisms usually raised against it were
successfully rebuffed—that it may not be naturalistically acceptable, that it may be
partial insofar as not all cases of rule following involve the formation of general
intentions, and that it may be caught in a vicious regress, since intentions must have
content and content requires intentions—there is a more fundamental and fatal
objection that must be raised against it. Again, the problem presents itself in the form
of a regress, but it targets the way in which rules, conceived under the aegis of the
Intention view, should be applied. In particular, application of a rule of the form “If C,
do A!” requires sensitivity to the fact that C obtains, which together with acceptance of
that rule, requires one to do A. This, however, is an application of modus ponens,
which is, presumably, in turn yet another rule. Hence, if all rule-following had to
conform to the Intention model, one should then form an intention which has modus
ponens as its content, and which, in turn, requires reasoning in accordance with that
scheme of inference in order to be applied, and so on ad infinitum. Hence, according to
Boghossian, the Intention view is doomed. At bottom, rule-following must be
“blind”—that is, it must be done without the intentional encoding of the rule’s
requirement.
Christopher Peacocke agrees with Boghossian’s criticism of the Intention view. But
in his contribution—”Understanding and rule-following”—he is mostly concerned
with offering a positive account of rule-following which can steer a mid-way between
linguistic idealism, on the one hand, and rampant Platonism, on the other. The former
is what, according to Peacocke, Wright’s early view is committed to and amounts to
the idea that in all domains not only sense, but also reference is determined by human
reactions and judgmental propensities. The latter, in contrast, is what, according to
Wright’s Wittgenstein, an objectivist account of rule-following commits one to, with
the unacceptable consequence of placing rules in a domain which makes them beyond
our reach. Peacocke’s positive account agrees with the intuition that rules are inde-
pendent of human judgment, but it aims to avoid the Platonist commitment to their
inaccessibility. At the heart of his proposal lies the intuition that there are fundamental
rules of reference, characteristic of the concepts employed in a judgment, which
determine in all possible cases and independently of a community’s agreement what
is to count as their correct application. In Peacocke’s view, a subject’s grasp of the
concepts used in his judgment consists in tacit knowledge of such fundamental rules of
reference. Tacit knowledge, in its turn, is to be explained by reference to subpersonal
computational contentful states. Finally, the issue of which fundamental rules of
reference a subject grasps will have to be decided, on Peacocke’s view, on the basis
of an inference to the best explanation given his patterns of observable behavior.
In his “Regularities, rules, meanings, truth conditions and epistemic norms”, Paul
Horwich argues that, at bottom, there is implicit rule-following. That is to say, a kind of
rule-following where the formulation of the rule isn’t operative either consciously or
unconsciously. According to Horwich, a person implicitly follows rule R if and only if,
given a certain pattern of regularities in his behavior, two further conditions obtain.
First, his activity is governed by the ideal rule R and, second, there is a tendency for the
subject to correct himself when he has an inclination to behave contrary to R. On such
a view, rule-following turns out to be a theoretical phenomenon, whose identification
requires empirical investigation. Furthermore, Horwich thinks that the rule-following
problem is not identical to the problem of meaning. For, on the one hand, there are
many more instances of rule-informed behavior than those involving language. On
the other, the second requirement placed upon rule-following—viz. that a subject be
able to correct himself on occasion—presupposes his ability to enjoy contentful
intentional states and hence his ability to entertain meanings. Horwich then concludes
by showing, first, how his proposal establishes the explanatory priority of meaning over
the determination of truth-conditions. Secondly, he shows how his proposal allows for
certain forms of normativity—most notably, those concerned with truth, epistemic
justification and with the fact that a speaker is part of a linguistic community where
there is room for expertise regarding the use of certain terms—while it entails the falsity
of the view that meaning facts are constitutively normative.
A somewhat similar conclusion is reached by Akeel Bilgrami in his “Why meaning
intentions are degenerate”. In particular, Bilgrami argues for a qualified version of that
thesis, viz. that if the intrinsic normativity of meaning is supposed to follow from the
normativity of meaning intentions, then meaning isn’t and can’t be normative after all.
While, on Bilgrami’s view, it is an important lesson to be learnt from Wittgenstein’s
Philosophical Investigations that intentions are normative—that intending to do some-
thing and fulfilling that intention can come apart—a similar mismatch can’t occur
when meaning intentions are at stake. That, in turn, suffices to show that meaning
intentions are “degenerate”—that they fail of normativity, contrary to other kinds of
intentions. More specifically, Bilgrami argues that the correct understanding of mean-
ing intentions should take the following form: one utters the words “a is F” with the
intention of saying something which is true if and only if a is F. If one then
misperceives a and says it is F when it is G, that doesn’t mean that one hasn’t lived
up to one’s initial meaning intention. Nor does one fail to live up to it if one utters
sentences that in one’s linguistic community are usually associated with different truth-
conditions, like in Tyler Burge’s example with “arthritis”. For, in that case—Bilgrami
argues—the relevant truth-conditions will have to be determined individualistically.
Finally, and by the same token, one won’t fail to live up to such an intention in cases of
“mis-speaking” or of slips of the tongue where words that either usually fail to
contribute to the determination of truth-conditions, or that typically contribute to
different ones are used. Bilgrami then connects his claim about how to determine the
relevant truth-conditions in the last two cases with an original reading of Frege’s
argument about the cognitive value of identity statements. In his view, the notion of
Sinn is introduced to make sense of the fundamental distinction between being
ignorant of a factual truth—the identity of a and b—and being irrational. The latter
requires, on Bilgrami’s interpretation of Frege, the transparency of Sinn. Meaning
intentions, therefore, cannot fail to be fulfilled and hence exert no normativity.
Section II: Knowledge of Our Own

Minds and Meanings
As anticipated, in Wright’s original treatment the solution of the rule-following
problem connects with the issue of self-knowledge—with the puzzling problem of
accounting for how we can know in a characteristically immediate and authoritative
way what we believe, desire and intend. Since his “Wittgenstein’s rule-following
considerations and the central project of theoretical linguistics” (1989), and then in a
number of papers as well as in his Whitehead Lectures (1996), now all collected in Rails
to Infinity, Wright has proposed a form of Constitutive account of self-knowledge.
On such a view, there is no real cognitive achievement that underlies first-person
psychological self-ascriptions of propositional attitudes. Rather, there is a presumption,
under suitably specified conditions, that if a subject makes an intentional avowal—
“I believe/desire/intend that p”—he does so believe, desire or intend. There is no
question of a subject’s “tracking” a pre-existing first-order mental state. Rather, his
readiness to accept the psychological self-ascription will, at least most of the times, be
constitutive of his being in the relevant mental state.
In “The publicity of meaning and the interiority of mind”, Barry C. Smith raises a
challenge for this account when it is applied to our knowledge of meanings—that is to
say, to our knowledge of what we mean by the words we use—once meanings are
thought of in a Wittgensteinian perspective as constituted by patterns of use of
linguistic items within a given community. The problem is one of reconciliation, as
Smith sees it. Namely, if meaning is use and use is public, it becomes difficult to see
how individual speakers can have immediate and authoritative knowledge of what
they mean by the words they employ. According to Smith, the problem with Wright’s
account of self-knowledge, when applied to the case of one’s meaning intentions, is
that it doesn’t really explain this form of first-personal knowledge, but at most only why
subjects’ avowals of their intentions are true, and it does so by the lights of a putative
audience that, at least in central cases, cannot but take such avowals at face-value. Smith
then goes on to canvass a possible solution to the reconciliation problem, which
partially draws on Wright’s most recent work in epistemology and his development
of the notion of epistemic entitlement.7 According to Smith, through a process of joint
attention towards a salient object, involving teachers and learners of language, children
endow linguistic items with meaning. They then have a natural tendency to assume
others mean the same as they do by the words they use; and they are strategically
7
We will come back to this issue in Section IV.
entitled to do so because otherwise the communicative practice couldn’t get started.

That, in turn, explains the typical phenomenology of immediate understanding of
others’ speech (at least when dealing with members of the same linguistic community),
since that is just the immediacy of interpreting other people’s utterances as if they were
speaking in one’s own idiolect.
In her “Expression, truth and reality: some variations on themes from Wright” Dorit
Bar-On agrees with Wright’s defence of the view that first-person, present-tense
psychological ascriptions—or “avowals”—enjoy a special kind of epistemic security
that distinguishes them from corresponding third-personal ascriptions. But she also
raises a challenge for Wright’s view. For Bar-On argues that on the Constitutive
account, it is a mystery why subjects are supposed to be better placed than others in
passing truth-determining judgments about at least some of their own mental
states. Moreover, it remains unexplained why only some of their psychological self-
ascriptions are constitutive of the corresponding first-order mental states, while others
aren’t. Furthermore, if the Constitutivist maintains that this is simply how things are
and that there is no need for an explanation, that would be on a par with the
“unphilosophical turning of the back”8 that Wright himself finds disappointing in
Wittgenstein’s later treatment of the issue. Bar-On then goes on to expound her neo-
expressivist account of self-knowledge. Its key element is the contention that some
psychological, present tense self-ascriptions are just expressions of mental states subjects
find themselves in, rather than descriptions of them. In her view, this is also what
accounts for their distinctive epistemic security. Finally, drawing on some of Wright’s
ideas in Truth and Objectivity, Bar-On maintains that her neo-expressivism can allow for
the fact that the content of avowals is truth-evaluable. She concludes with some
suggestive remarks about how the neo-expressivist paradigm developed to account
for self-knowledge could be applied in the case of ethical discourse.
Section III: Truth, Objectivity and Relativism

In Truth and Objectivity (1992),9 Wright turned to the problem of addressing and
partially redefining the rules of the debate between so-called “realists” and “anti-
realists”. He therefore proposed a new form of minimalism about truth and assertoric
content, which could rehabilitate anti-realist views such as expressivism, by allowing
them to solve both the Frege–Geach problem—viz. how sentences containing pre-
dicates that, on an expressivist interpretation, do not refer to a property, but merely
express a subject’s reaction or attitude, could meaningfully figure as the antecedents of
conditionals, or be prefixed by negation—and the problem of how such sentences could
genuinely be said to be true or false. Wright also proposed further criteria—such as
8
C. Wright, 2001, Rails to Infinity: Essays on Themes from Wittgenstein’s Philosophical Investigations,
Cambridge, Mass., Harvard University Press, p. 369.
9
“cognitive command”, “width of cosmological role”, and the “order of determination

test”10—for establishing whether a discourse, which would qualify as minimally truth-apt
and as minimally representational, was to be thought of along realist lines or, rather, in an
anti-realist fashion, as about properties ultimately dependent on human reaction and
opinion. Connectedly, these criteria would determine whether a discourse is subject to
a “robust” kind of truth or to a different notion of truth, more intimately tied to human
judgment, which Wright called “superassertibility”. The take-home message of what has
proved to be an extremely influential book was, then, that under the apparent syntactic
uniformity of our language, which needs to be respected, there is a variety of functions it
can serve—going from being robustly representational to subjectively expressivist—as
well as a variety of properties which may correspond to our truth-predicate.
In his “Some remarks about minimalism”, Simon Blackburn doesn’t take issue with
any of Wright’s views as developed in Truth and Objectivity. Rather, he explores other
minimalist positions about truth—in particular Paul Horwich’s and Hartry Field’s—
arguing that their minimalism is in fact unstable. For, in order for it to be sensibly
maintained, they need to inflate other notions, such as that of proposition, or
of “sentence as understood”. Field—Blackburn contends—must acknowledge the
distinction between truth-conditions and mere “indication relations”—i.e. reliable
dispositions to assert or believe that P—and thus end up assigning a central role to
the former in a theory of mind, contrary to his deflationist propensities. By contrast,
according to Blackburn, Horwich’s proposal about reference amounts to a hardly
plausible form of particularism—viz. the view that there is nothing general to be said
about how apparently uniform classes of terms manage to refer to their corresponding
properties, which may in fact harbour, at least at times, a commitment to inflated
notions, such as that of disposition explained in terms of causal relations. Blackburn
concludes by agreeing with Wright that while different discourses may vary from being
genuinely representational to purely expressivist, that doesn’t stand in the way of
the possibility of embedding sentences characteristic of either domain within truth-
disquotational schemas.
Stewart Shapiro, in his “Objectivity, explanation and cognitive shortfall”, takes issue
with two criteria Wright proposed for demarcating objective areas of discourse from
non-objective ones, namely what Wright labelled “width of cosmological role” and
“cognitive command”. Roughly, the former consists in holding that a given area of
discourse is objective just in case the situations it describes can feature in several
explanations of contingencies in different areas. The latter, in contrast, amounts to
the idea that a discourse is objective just in case it is a priori that disagreement among
parties engaging in it depends on at least one of them displaying some cognitive
10
There will be more on the first two notions in the following. The order of determination test has to do
with the Euthyphro’s problem of which side of certain biconditional statements has explanatory priority over
the other. Hence, the test should help adjudicate the issue of whether a certain property is exemplified
because human beings judge that to be the case, or else it is exemplified independently of their judgment,
which, in turn, reflects a mind-independent reality.
shortcoming, varying from ignorance or error, to inattention and distraction, up to

prejudice and dogmatism. According to Shapiro, both these criteria are problematic.
For, he claims, the former depends on subtle issues having to do with the notion of
explanation. Once they are more closely examined, it turns out that width of cosmolog-
ical role either presupposes objectivity and cannot therefore be used to explain it; or else
it becomes irrelevant for determining whether an area of discourse is objective because,
to such an end, it is sufficient that its propositions figure in some explanations, rather than
in many. By contrast, cognitive command is problematic because we can envisage a case
where even science seemingly fails to pass the test. If such a case is taken at face value and
one doesn’t want to be anti-realist about science, it constitutes a counterexample to
Wright’s criterion for objectivity. If, in contrast, one wants to explain the case away, a
number of very subtle issues both in epistemology and with respect to what counts as
“cognitive” need to be addressed. Hence, cognitive command cannot be straightfor-
wardly employed to tell objective and non-objective areas of discourse apart.
Since at least “On being in a quandary: relativism, vagueness and logical revision-
ism” (2001),11 as well as in a number of subsequent papers, Wright has explored the
possible relativist implications of the anti-realist stance developed in Truth and Objectiv-
ity, as well as various proposals, most notably about truth-relativism, that have appeared
in the literature in recent years.12 One characteristic feature of some areas of discourse,
according to Wright, is that disputes in the relevant areas fail at “cognitive command”
and exhibit “faultless disagreement”, whereby opposite parties hold incompatible
views, yet neither need be wrong or be making any mistake. Still, he is careful to
avoid the implication that they must, therefore, both be right. For that would entail a
denial of the Principle of Non-contradiction that Wright aims to avoid. In order to
avoid that implication, he then elaborates a philosophical notion of being in a quandary,
which comes down to a form of thorough and irresolvable uncertainty as to whether
either P or not-P is the case, yet doesn’t involve a commitment to the truth of both.
In “How to formulate relativism” Carol Rovane, who agrees with Wright in
rejecting contemporary forms of truth-relativism as suitable formulations of that
doctrine, takes issue with his characterization of relativist disputes as manifesting
some irresoluble conflict and with his account of the latter in terms of facing a
quandary. For, in her view, the fact that we don’t know how to resolve a given
conflict doesn’t entail that it can’t be resolved in principle. But it is only if it were so
irresolvable—and hence if both parties could actually both be right—that it could
signal a threat of relativism. Hence, she thinks Wright faces the dilemma that afflicts any
11
Now in C. Wright, 2003 Saving the Differences. Essays on Themes from Truth and Objectivity, Cambridge,
Mass., Harvard University Press.
12
See C. Wright, 2006 “Intuitionism, Relativism and Rhubarb”, in P. Greenough and M. Lynch. (eds.)
Truth and Realism, Oxford, Oxford University Press, pp. 38–60; C. Wright, 2008 “Relativism about Truth
Itself: Haphazard Thoughts about the Very Idea”, in M. Garcia-Carpintero and M. Kölbel (eds.) Relative
Truth, Oxford, Oxford University Press, pp. 157–85; C. Wright, 2007 “New Age Relativism and Epistemic
Possibility: The Question of Evidence”, Philosophical Issues 17/1, pp. 262–83.
attempt to formulate relativism in terms of irresoluble conflict. Either both parties to

the conflict can be right, in which case there is a violation of the Principle of Non-
contradiction, or it is not the case that both parties can be right, in which case they face
an ordinary disagreement of the sort that opponents of relativism would be prepared to
countenance.
Rovane proposes to capture the distinctive metaphysical commitment of relativism
by appealing to the idea of alternatives, which she defines as truths that cannot be
embraced together. But she does not suppose that the reason why they cannot
be embraced together is that this would be a violation of the Principle of Non-
contradiction. Rather, in her view, such truths fail to stand in logical relations at all.
As she puts it, they are “normatively insulated from one another”. This means that
although they are not inconsistent with one another, they are not consistent either and
so cannot be conjoined. She couples this idea with a metaphysical stance according to
which there can be, literally, multiple worlds—that is to say, multiple incomplete and
insulated bodies of truths. She then draws out the epistemic and ethical consequences
of that “multimundial” view: a principled and necessary indifference to others’ beliefs
and a profound and unbridgeable sense of otherness. Finally, she argues that the very
reasons for thinking that the framework of anti-realism might have given rise to
irresoluble conflicts in which both parties are right invite her formulation of relativism
as multimundialism.
Section IV: Warrant, Transmission Failure

and Scepticism
Since “Facts and certainty” (1985),13 as well as in a number of papers appeared later,14
Wright has been concerned with the characterization of the phenomenon that he dubs
“failure of warrant transmission” and with an account of its difference from another
familiar alleged phenomenon—viz. the possibility, which Wright rejects, of violations
of the Principle of Closure of warrant across known entailment, originally argued for
by Fred Dretske and Robert Nozick.15 Moreover, he has explored the bearing of
transmission failure on the assessment of a number of well-known philosophical
arguments, such as Moore’s Proof of an external world, Putnam’s argument against
13
Henriette Hertz Philosophical Lecture for the British Academy, December 1985, in Proceedings of the
British Academy LXXI, pp. 429–72.
14
C. Wright, 2000 “Cogency and Question-Begging: Some Reflections on McKinsey’s Paradox and
Putnam’s Proof ”, Philosophical Issues 10, pp. 140–63. C. Wright, 2002 “Anti-sceptics Simple and Subtle:
Moore and McDowell”, Philosophy and Phenomenological Research 65, pp. 330–48. C. Wright, 2003 “Some
Reflections on the Acquisition of Warrant by Inference”, in S. Nuccetelli (ed.) New Essays on Semantic
Externalism, Scepticism and Self-Knowledge, Boston, Mass., MIT Press, pp. 57–78.
15
F. Dretske, 1970 “Epistemic Operators”, Journal of Philosophy 67, pp. 107–23. R. Nozick, 1981
Philosophical Explanations, Cambridge, Mass., Harvard University Press, ch. 3. Notice that Dretske and Nozick
have actually argued for the failure of the Principle of Closure with respect to knowledge, rather than
warrant.
the hypothesis that we might be brains in a vat, and McKinsey’s argument for the
incompatibility between self-knowledge and semantic externalism.
It is in connection with whether Moore’s Proof of an external world exhibits a
failure of transmission of warrant that a serial debate has arisen in recent years between
Wright and Jim Pryor.16 As is by now familiar, Pryor holds that Moore’s Proof is
epistemically sound: that in the conditions in which Moore formed the belief that there
was a hand in front of him, he did acquire a warrant and that warrant did transmit to the
conclusion of the argument—viz. “There is an external world”. It is only when
presented against a sceptic with respect to the existence of an external world that the
argument is dialectically ineffective, because such a sceptic, given his doubts about the
existence of an external world, won’t concede that Moore’s experience as of a hand in
front of him gives him warrant for the premises of his argument in the first place.
In his contribution to the present volume—“When warrant transmits”—Pryor
defends his analysis of Moore’s Proof from some objections that have appeared in
the rapidly flourishing literature on the topic, and which have been partly fostered by
Wright himself.17 He then clarifies the idea that the Proof is dialectically ineffective
when presented against a sceptic. However, his main concern is to defend the view that
there is at most only one kind of transmission failure—the one made familiar by
Wright’s writings—that he endeavours to clarify by going through different formula-
tions of it which can be found in the latter’s papers. Accordingly, transmission failure is
deemed to be due to the fact that prior warrant for the conclusion of an argument is
needed in order to sustain the warrant for its premises in the first place. By contrast,
Pryor argues that other possible forms of transmission failure are all defective. These
include a notion of transmission failure according to which warrant for the premises
depends on the obtaining of the conclusion, or on taking it for granted (when it entails the
falsity of a possible underminer of one’s warrant for the premises). Finally, Pryor rejects
the idea that transmission failure would arise when warrant for the conclusion of an
argument is required in order to be warranted in believing that one has warrant for the
premises. Hence, if Pryor is right, in order to make its point, dogmatism, when applied
to Moore’s argument, needs only to show that the Proof doesn’t fail to transmit
warrant in Wright’s sense.
José L. Zalabardo, in his “Wright on Moore”, aims to offer a slightly different
characterization of the phenomenon of transmission failure from the one originally
presented by Wright. According to Zalabardo, the initial understanding of transmission
failure has it that an argument fails to transmit warrant from the premises to the
conclusion if certain “limitation clauses” obtain. For instance, if one already needs
independent warrant for the conclusion in order to have warrant for the premises in the
16
J. Pryor, 2000 “The Skeptic and the Dogmatist”, Noûs 34, pp. 517–49. J. Pryor, 2004 “What’s Wrong
with Moore’s Argument?”, Philosophical Issues 14, pp. 349–78.
17
C. Wright, 2007 “The Perils of Dogmatism”, in S. Nuccetelli and G. Seay (eds.) Themes From
G. E. Moore: New Essays in Epistemology and Ethics, Oxford, Oxford University Press, pp. 25–48.
first place. But, on Zalabardo’s view, such a limitation clause would yield the wrong
results if coupled with a reliabilist account of warrant—that he finds more promising
than any internalist alternative—according to which one can have warrant for a
proposition such as “Here is a hand” without having prior and independent warrant
for “There is an external world”. He therefore proposes a different characterization,
according to which, granted the Principle of Closure, warrant will fail to transmit from
the premises to the conclusion of a logically valid argument any time it can be shown
that, prior to recognizing the validity of an inference, a subject already possesses a
warrant for the proposition that the premise entails the conclusion. On Zalabardo’s
view, this is precisely the case with Moore’s Proof, where anyone who has the concepts
necessary to entertain the premise “Here’s a hand” and the conclusion “There is an
external world” will also have a warrant for the entailment between them. Zalabardo
then goes on to argue against Pryor that Moore’s Proof can give one evidence for its
conclusion. He points out that having a piece of evidence that makes a proposition
such as “Here’s a hand” more likely for one to be true doesn’t entail that that piece of
evidence makes also a proposition known to be entailed by the former, such as “There
is an external world”, more likely to be true.
Wright’s reflections on transmission failure and Moore’s Proof, however, belong to
a larger and more ambitious project—viz. that of characterizing various forms of
scepticism, specially about the existence of an external world, and of finding a satisfac-
tory response to them. That project, initiated in “Facts and certainty” with the
individuation of what Wright considered an instance of Humean scepticism, and
further pursued in “Scepticism and dreaming: imploding the demon” (1991),18 with
the identification of the main features of Cartesian scepticism, has been taken up once
more in recent years, and has been marked by a considerable change of mind. For in his
earlier work Wright had been rather close to the letter of Wittgenstein’s remarks in On
Certainty,19 in arguing that certain propositions, such as “There is an external world”,
or “I’m not now dreaming”, can’t sensibly be doubted because they are rules—of
evidential significance, rather than of meaning—devoid of factual content, whose
acceptance doesn’t depend on evidence. In his later work, in contrast, Wright aims
to defend those propositions from sceptical doubts, while recognizing their factual
content, and he does so by developing and deepening his understanding of their
epistemological status. In particular, in “Warrant for nothing (and foundations for
free)?” (2004),20 Wright clearly brings out the difference in structure between Carte-
sian and Humean scepticism: while the former involves appeal to a sceptical scenario
whose obtaining we aren’t in a position to rule out on evidential grounds, the latter
doesn’t make play with such a sceptical hypothesis, but points out a circularity in
arguments, such as Moore’s, which are designed to produce evidential warrant for the
belief in the existence of an external world. However, Wright argues for a “unified”
18 19
Mind 100, pp. 87–116. Oxford, Blackwell, 1969.
20
The Aristotelian Society Supplementary Volume LXXVIII, pp. 167–212.
strategy of response to both forms of scepticism.21 Such a strategy grants to sceptics of

both stripes that no evidential warrant can be provided either to discard the Cartesian
sceptical scenario or to embrace the belief in the existence of an external world. Yet, it
claims that there are non-evidential warrants—called “entitlements”—which allow us
rationally to discard the hypothesis that we might be dreaming right now, as well as
to take it for granted that there is an external world.
Annalisa Coliva’s essay “Moore’s Proof, liberals and conservatives—is there a
(Wittgensteinian) third way?”—originates from her dissatisfaction with, on the one
hand, Wright’s notion of entitlement and its application to our acceptance of the
existence of an external world; and, on the other, from her discontent with respect
to Pryor’s liberal account of the architecture of empirical warrants. In such a
predicament, she finds relief in the idea that there may be a “third way” according
to which while it isn’t enough merely to have a certain course of experience in order
to have warrant for “Here is a hand”, no independent warrant for “There is an
external world” is needed in order to have warrant for the former either. Rather, it
suffices to have a certain course of experience while simply assuming or trusting in
the existence of an external world, with no warrant for it—evidential or otherwise.
She then shows how the third way has the resources to provide an answer to
Humean scepticism. She agrees with Wright that such a sceptic is guilty of wrongly
inferring that, since no evidential warrants for “There is an external world” are
possible, that belief isn’t rationally held. But she disagrees with Wright in thinking
that that calls for a wider conception of warrant, in which entitlements are
countenanced. Rather, on her view, that calls for a wider—in fact the right—notion
of epistemic rationality, according to which the presuppositions that are constitutive of
it, such as that there is an external world, and that can’t be independently war-
ranted—evidentially or otherwise—are part of it. She then explores the bearing of
the third way on the assessment of Moore’s Proof and shows how the latter turns out
to be defective because of a different kind of transmission failure than Wight’s.
Coliva concludes by highlighting the relationship between the third way and
Wittgenstein’s remarks in On Certainty.
In “Wright against the sceptics” Michael Williams criticizes Wright’s recent views
against scepticism on several counts. First, he thinks that Wright’s notion of entitlement
makes only for the pragmatic rationality of assuming certain cornerstone propositions,
but doesn’t vindicate their epistemic rationality—that is to say, it doesn’t speak to their
(likely) truth. Second, he argues that if this is Wright’s position, then it is a response
merely against a sceptic who holds that our trust in certain “hinges” isn’t rational at all.
But, arguably, the best sceptical challenge concerns only the fact that we can’t redeem
21
Another element of novelty is the fact that Wright has become more and more interested in Cartesian
and Humean scepticism construed as bearing not so much on the possession of warrant for given beliefs but
on the possibility of claiming such a warrant. See also C. Wright, 2008 “Internal-external: Doxastic Norms
and the Defusing of Sceptical Paradox”, Journal of Philosophy 105/9, pp. 501–17.
the epistemic rationality of our most basic acceptances and Wright’s defence has nothing
to say against it. Third, Williams objects that Wright’s entitlements aren’t really
unearned warrants. True, they aren’t earned through the acquisition of evidence. Still,
they are obtained by showing “how they subserve epistemic interests”. However, in
order to avoid requiring subjects to be able to appreciate how that comes about—a
condition which would make empirical warrants beyond most people’s reach—
Williams argues that Wright must adopt an externalist stance on his own entitlements.
That is to say, they must be thought of as obtaining independently of anyone’s grasp of
them. This—fourthly—seems to create a tension: arguably, if externalism is allowed,
then scepticism about the possibility of warrant (and knowledge) with respect to
ordinary empirical beliefs may be silenced. Yet, Wright seems to think that more is
needed to meet the sceptical challenge and, furthermore, he characterizes his strategy as
internalist. On Williams’ understanding, this tension can be relieved only if it is
recognized that Wright is more concerned with “scepticism about philosophy” than
with philosophical scepticism. Accordingly, scepticism may not threaten our warrants
but it may preclude our being in a position to claim them and it is this sort of
epistemological self-understanding we are after when we do philosophy. Wright’s
entitlements are ultimately brought into the picture to respond to this latter sceptical
challenge. Yet—fifthly—they come in too late as they leave open the possibility of
redeeming the rationality of different cornerstones that, intuitively, we wouldn’t be
prepared to recognize as rational; or else, they avoid such an outcome only by
presupposing the existence of an external world all along. Finally, Williams argues
that Wright’s principal mistake, which he in fact shares with sceptics, is to hold on to
a form of foundationalism according to which some cornerstones are held to be
beyond doubt and revision. In contrast, Williams thinks that Wittgenstein’s deep
lesson in On Certainty is that all cornerstones are revisable at least in principle
and hence that foundationalism should be abandoned in favor of epistemological
contextualism.
The preceding should have made clear the extent in which Wright’s philosophical
agenda has been set by his reading of Wittgenstein’s post-Tractarian works, from the
Remarks on the Foundations of Mathematics, to the Philosophical Investigations, up to On
Certainty. That is why, for a while, the publisher and I thought of signalling this by
including “Wittgenstein” in the title of the present volume. But two sorts of consider-
ation militated against this choice. The first is the length of the prospective title. The
second is the fact that Wright’s work is not so much exegetical of as inspired by
Wittgenstein’s writings. I personally think it is a great merit of Wright’s own contri-
bution to philosophy that, in his own distinctive and sometimes critical way—or
perhaps because of such an attitude—he has ensured that Wittgensteinian themes and
philosophical preoccupations have remained part of our discipline as it is practised
today, even when his writing looks, at least superficially, remote from Wittgenstein’s
ideas both in content and style.
In closing I would like to thank all authors for their patience in dealing with a
seemingly never-ending editorial process; but, most of all, for their extremely thought-
provoking contributions. And last, my thanks go to Crispin for his characteristically
thorough and insightful replies which, I am sure, will help both contributors and the
readers of this volume to get clearer about the intricacies of the issues here discussed, as
well as about our own ideas, of which, in many cases, he has been a strong inspirational
source.
Bio-bibliographical Note
Crispin J. G. Wright, born in 1942 in Surrey, was educated at Birkenhead School and
at Trinity College, Cambridge, graduating in Moral Sciences in 1964 and taking a PhD
in 1968. He took an Oxford BPhil in 1969 and was elected Prize Fellow and then
Research Fellow at All Souls College, Oxford, where he worked until 1978. He then
moved to the University of St Andrews, where he became the first Bishop Wardlaw
University Professor in 1997. From autumn 2008, he has been professor at New York
University. He has also taught at the University of Michigan, Oxford University,
Columbia University and Princeton University. He is a Fellow of the British Academy
and of the Royal Society of Edinburgh. He is the founder of Arché – Philosophical
Research Centre for Logic, Language, Metaphysics and Epistemology, at St Andrews,
which he left in September 2009 to take up leadership of the Northern Institute of
Philosophy at the University of Aberdeen.
Books
1980 Wittgenstein on the Foundations of Mathematics, London, Duckworth.

1983 Frege’s Conception of Numbers as Objects, Aberdeen, Aberdeen University Press.
1986 Realism, Meaning and Truth, Oxford, Blackwell.
1992
(i) Truth and Objectivity, Cambridge, Mass., Harvard University Press.
(ii) Expanded paperback 2nd edition of Realism, Meaning and Truth, Oxford, Blackwell.
(iii) Limited edition re-issue of Wittgenstein on the Foundations of Mathematics in hard-
cover by Gregg Revivals.
2001
(i) The Reason’s Proper Study, co-authored with Bob Hale, Oxford, Oxford Univer-
sity Press.
(ii) German translation of Truth and Objectivity published as Wahrheit und Objectivität,
Frankfurt, Suhrkamp.
(iii) Rails to Infinity: Essays on Themes from Wittgenstein’s Philosophical Investigations,
2003 Saving the Differences: Essays on Themes from Truth and Objectivity. Cambridge,
Wright, C. (ed.) 1984 Frege: Tradition and Influence, Oxford, Blackwell.
Edited Volumes
1984
(i) Synthese 58/3, special number entitled Essays on Wittgenstein’s Later Philosophy.
(ii) Philosophical Quarterly 34/136, special number on Frege, paperbacked as Frege:
Tradition and Influence by Blackwell.
1986–7 With Graham Macdonald, Fact, Science and Morality, Oxford, Blacwell.
1988 With Peter Clark, Mind, Psychoanalysis and Science, Oxford, Blackwell.
1993 With John Haldane, Reality: Representation and Projection (Oxford), proceedings of
the conference on Realism and Reason held at St Andrews in March 1988.
1986–94 With John McDowell and Philip Pettit, editor of Philosophical Theory,
Oxford, Blackwell.
1997 With Bob Hale, A Companion to the Philosophy of Language, Oxford, Blackwell.
1998 With Cynthia Macdonald and Barry Smith, Knowing Our Own Minds, Oxford,
Oxford University Press.
2002 With Alex Miller, Rule-Following and Meaning, Durham, Acumen.
Articles and Critical Studies

Epistemology
1983 “Keeping Track of Nozick”, Analysis 43, pp. 134–40.
1984 “Comment on Lowe”, Analysis 44, pp. 183–5.
1985 “Facts and Certainty”, Proceedings of the British Academy 71, pp. 429–72.
1991 “Skepticism and Dreaming: Imploding the Demon”, Mind 100, pp. 87–115.
1991 “On Putnam’s Proof that we are not Brains-in-a-Vat”, Proceedings of the Aristote-
lian Society XCII, pp. 67–94.
1994 Extended version of “On Putnam’s Proof that we are not Brains-in-a-Vat”, in
P. Clark and R. Hale (eds.) Reading Putnam, Oxford, Blackwell, pp. 216–41.
1998 “McDowell’s Oscillation”, Philosophy and Phenomenological Research 58/2, pp. 395–402.
2000 “Cogency and Question-Begging: Some Reflections on McKinsey’s Paradox
and Putnam’s Proof”, Philosophical Issues 10, pp. 140–63.
2000 “Replies”, Philosophical Issues 10, pp. 201–19.
2002 “(Anti-)Sceptics Simple and Subtle: G. E. Moore and John McDowell”, Philoso-
phy and Phenomenological Research 65, pp. 330–48.
2003 “Some Reflections on the Acquisition of Warrant by Inference”, in S. Nuccetelli
(ed.) New Essays on Semantic Externalism and Self-Knowledge, Cambridge, Mass., MIT
Press, pp. 57–77.
2004 “Scepticism, Certainty, Moore and Wittgenstein”, in M. Kölbel and B. Weiss
(eds.) Wittgenstein’s Lasting Significance, London, Routledge, pp. 228–48.
2004 “Scepticism, Certainty, Moore and Wittgenstein”, in A. Coliva and E. Picardi
(eds.) Wittgenstein Today, Padova, Il Poligrafo, pp. 371–96 (abridged version of
“Wittgensteinian Certainties”).
BIO-BIBLIOGRAPHICAL NOTE 17
2004 “On Epistemic Entitlement: Warrant for Nothing (And Foundations for Free)?”,
Proceedings of the Aristotelian Society, Supp. Vol. LXXVIII, pp. 167–212.
2004 “Wittgensteinian Certainties”, in D. McManus (ed.) Wittgenstein and Scepticism,
London, Routledge, pp. 22–55.
2004 “Hinge Propositions and the Serenity Prayer”, in Knowledge and Belief, in
W. Loffler and P. Weingartner (eds.) Proceedings of the 26th International Wittgenstein
Symposium, Vienna, Holder-Pickler-Tempsky, pp. 287–306.
2007 “The Perils of Dogmatism”, in S. Nuccetelli and G. Seay (eds.) Themes from
G. E. Moore: New Essays in Epistemology and Ethics, Oxford, Oxford University Press,
pp. 25–48.
2008 Comment on John McDowell’s “The Disjunctive Conception of Experience as
Material for a Transcendental Argument”, in A. Haddock and F. MacPherson (eds.)
Disjunctivism: Perception, Action, Knowledge, Oxford, Oxford University Press,
pp. 390–404.
2008 “Internal—External: Doxastic Norms and the Defusing of Sceptical Paradox”,
Journal of Philosophy 105/9, pp. 501–17.
2011 “Frictional Coherentism? A Comment on Chapter 10 of Ernest Sosa’s Reflective
Knowledge”, Philosophical Studies 153/1, pp. 29–41.
2011 “McKinsey One More Time”, in A. Hatzimoysis (ed.) Self-Knowledge, Oxford,
Oxford University Press, pp. 80–104.
Forthcoming “Welfare State Epistemology (Entitlement II)”, in D. Dodd and
E. Zardini (eds.) Contemporary Perspectives on Scepticism and Perceptual Justification,
Oxford, Oxford University Press.
Ethics
1984 “The Moral Organism”, in A. Manser and G. Stock (eds.) The Philosophy of
F. H. Bradley, Oxford, Oxford University Press, pp. 77–97.
1988 “Moral Values, Projection and Secondary Qualities”, Proceedings of the Aristotelian
Society, Suppl. Vol. LXII, pp. 1–26.
1995 “Truth in Ethics”, Ratio (New Series) 8, pp. 209–26.
Language
1975 “On the Coherence of Vague Predicates”, Synthese 30, pp. 325–65.
1976 “Language-Mastery and the Sorites Paradox”, in G. Evans and J. McDowell
(eds.) Truth and Meaning, Oxford, Oxford University Press, pp. 223–47, reprinted in
R. Keefe and P. Smith (eds.) 1996 Vagueness: A Reader, Cambridge, Mass., Brad-
ford/MIT, pp. 151–73.
1982 “Anti-realist Semantics: The Role of Criteria”, in G. Vesely (ed.) Idealism: Past and
Present, Cambridge, Cambridge University Press, pp. 225–48.
1984 “Second Thoughts about Criteria”, Synthese 58, pp. 383–405.
1985 With Stephen Read, “Hairier than Putnam Thought”, Analysis 45, pp. 56–8.
1986 “How Can the Theory of Meaning be a Philosophical Project?”, Mind and
Language 1, pp. 31–44.
1986 “Theories of Meaning and Speakers’ Knowledge”, in S. Shanker (ed.), Philosophy
in Britain Today, London and Syndey, Croom Helm, pp. 267–307.
1986 “The Sorites Paradox”, Quaderni di Semantica 7/2, pp. 277–91.
1987 “Further Reflections on the Sorites Paradox”, in Philosophical Topics, 15/1,
pp. 227–290, reprinted in R. Keefe and P. Smith (eds.) 1996 Vagueness: A Reader,
Cambridge, Mass., Bradford/MIT, pp. 204–50.
1989 “The Sorites Paradox and its Significance for the Interpretation of Semantic
Theory”, in Recherches sur la Philosophie et le Langage, Grenoble, Imprimerie de la
Bibliothèque Interuniversitaire et des Sciences Sociales de Grenoble, pp. 119–36,
reprinted with a new appendix on higher-order vagueness in N. Copper and
P. Engel (eds.) 1991 New Inquiries into Meaning and Truth, London, Simon &
Schuster, pp. 135–62.
1992 “Is Higher Order Vagueness Coherent”, Analysis 52, pp. 129–39.
1995 “The Epistemic Conception of Vagueness”, The Southern Journal of Philosophy 33,
pp. 133–59.
1997 “The Indeterminacy of Translation”, in B. Hale and C. Wright (eds.) A Compan-
ion to the Philosophy of Language, Oxford, Blackwell, pp. 397–426.
1997 With Bob Hale “Putnam’s Model-theoretic Argument Against Metaphysical
Realism”, in B. Hale and C. Wright (eds.) A Companion to the Philosophy of Language,
Oxford, Blackwell, pp. 427–57.
1998 “Why Frege Does Not Deserve His Grain of Salt: A Note on the Paradox of ‘The
Concept Horse’ and the Ascription of Bedeutungen to Predicates”, in J. Brandl and
P. Sullivan (eds.) Grazer Philosophische Studien 55, New Essays on the Philosophy of
Michael Dummett, Vienna, Rodopi, pp. 239–63.
2005 “Vagueness, Response Dependence and Rule-Following—Some Reflections”, in
S. Moruzzi and A. Sereni (eds.), Issues on Vagueness, Padova, Il Poligrafo, pp. 51–74.
2006 “Vagueness-related Partial Belief and the Constitution of Borderline Cases”,
Philosophy and Phenomenological Research 73/1, pp. 225–32.
2007 “ ‘Wang’s Paradox’ ”, in The Philosophy of Michael Dummett, Library of Living
Philosophers, Chicago, Open Court, pp. 415–44.
2010 “The Illusion of Higher-order Vagueness”, in R. Dietz and S. Morruzzi (eds.) Cuts and
Clouds: Vagueness, Its Nature, and Its Logic, Oxford, Oxford University Press, pp. 523–49.
2010 With Bob Hale “Assertibilist Truth and Objective Content: Still Inexplicit?”, in
B. Weiss and J. Wanderer (eds.) Reading Brandom: Making It Explicit, London,
Routledge, pp. 276–93.
Forthcoming “On the Characterisation of Borderline Cases”, in G. Ostertag (ed.)
Meanings and Other Things: Essays on Stephen Schiffer, Cambridge, Mass., MIT Press.
With Bob Hale “Horse Sense” in R. May and C. Parsons (eds.) special number of the
Journal of Philosophy on Frege’s Philosophy.
Metaphysics
1976 “Truth-Conditions and Criteria”, Proceedings of the Aristotelian Society, Supp. Vol.
50. pp. 217–45.
1979 “Strawson on Anti-Realism”, Synthese 40, pp. 283–99.
1980 “Realism, Truth-value Links, Other Minds and the Past”, in J. Dancy (ed.) Papers
on Language and Logic, Keele, Keele University Press, pp. 192–226, reprinted in Ratio
22, pp. 112–32.
1986 “Scientific Realism, Observation and the Verification Principle”, in
G. MacDonald and C. Wright (eds.) Fact, Science and Morality, Oxford, Blackwell
(US edition 1987), pp. 247–74.
1988 “Realism, Anti-Realism, Irrealism, Quasi-Realism”, in H. Wettstein, T. Uehling
and P. French (eds.) Realism and Anti-Realism, Midwest Studies in Philosophy, vol. XII,
Minneapolis, University of Minnesota Press, pp. 25–50, reprinted in J. Kim and
E. Sosa (eds.) 1999 Metaphysics: An Anthology, Oxford, Blackwell, pp. 649–65.
1989 “Misunderstandings Made Manifest, a Response to Simon Blackburn”, in
H. Wettstein, T. Uheling and P. French (eds.) Contemporary Perspectives in the
Philosophy of Language II, Midwest Studies in Philosophy, vol. XIV, Notre Dame,
Ind., University of Notre Dame Press, pp. 48–67.
1989 “The Verification Principle: Another Puncture, Another Patch”, Mind 98,
pp. 611–22.
1992 “Wohin fuhrt die aktuelle Realismusdebatte?”, in W. R. Kohler (ed.) Realismus und
Antirealismus, Suhrkamp Taschenbuch Wissenschaft 976, Frankfurt/M, Suhrkamp,
pp. 300–34.
1993 “On an Argument concerning Classical Negation”, Mind 102, pp. 123–31.
1993 “Anti-realism: The Contemporary Debate—Whither Now?”, in J. Haldane and
C. Wright (eds.) Reality: Representation and Projection, Oxford, Oxford University
Press, pp. 63–84.
1993 “Scientific Realism and Observation Statements”, in D. Bell and W. Vossenkuhl
(eds.) Science and Subjectivity: The Vienna Circle and Twentieth Century Philosophy, Berlin,
Akademie Verlag, pp. 21–46, reprinted in the International Journal of Philosophical Studies
1 (1993), pp. 231–54. Slightly shortened version reprinted in Italian as “Il realismo
scientifico e gli asserti osservativi”, in A. Pagnini (ed.) 1995 Realismo/Antirealismo:
Aspetti del Dibattito Epistemologico Contemporaneo, Firenze, La Nuova Italia, pp. 205–35.
1994 “Response to Jackson”, Philosophical Books 35, pp. 169–75.
1994 “Realism, Pure and Simple?”, International Journal of Philosophical Studies 2,
pp. 147–61.
1996 Book Précis and Response to Commentators (Horgan, Horwich, Pettit, Sainsbury,
van Cleve, Williamson) for a Book Symposium on Truth and Objectivity in Philosophy
and Phenomenological Research 56/4 (1996), pp. 863–8 (précis) and pp. 911–41 (responses).
1998 “Comrades against Quietism: Reply to Simon Blackburn on Truth and Objectivity”,
Mind 107, pp. 183–203.
1998 “Euthyphronism and the Physicality of Colour”, in R. Casati and C. Tappolet

(eds.) European Review of Philosophy 3, pp. 15–30.
1998 “Truth: A Traditional Debate Reviewed”, in C. Misak (ed.) Suppl. Vol. Canadian
Journal of Philosophy 24, pp. 31–74, reprinted in S. Blackburn and K. Simmons (eds.)
1999 Truth, Oxford, Oxford University Press, pp. 203–38; German version in
M. Vogel and L. Wingert (eds.) 1999 Unsere Welt gegeben oder gemacht? Menschliches
Erkennen zwischen Entdeckung und Konstruktion, Frankfurt/M, Suhrkamp, pp. 55–106.
2000 “Truth as Sort of Epistemic: Putnam’s Peregrinations”, Journal of Philosophy 97/6,
pp. 335–64.
2001 “Miminalism, Deflationism, Pragmatism, Pluralism”, in M. P. Lynch (ed.) The Nature
of Truth: Classic and Contemporary Perspectives, Cambridge, Mass., MIT Press, pp. 751–87.
2002 “The Conceivability of Naturalism”, in T. Gendler and J. O’Leary-Hawthorne
(eds.) Imagination, Conceivability and Possibility, Oxford, Clarendon Press, pp. 401–39.
2002 “Relativism and Classical Logic”, in A. O’Hear (ed.) Logic, Language ad Thought,
Cambridge, Cambridge University Press, pp. 95–118.
2002 “What Could Anti-Realism about Ordinary Psychology Possibly Be?”, The
Philosophical Review 111, pp. 205–33.
2006 “Intuitionism, Realism, Relativism and Rhubarb,” in P. Greenough and
M. P. Lynch (eds.) Truth and Realism, Oxford, Oxford University Press, pp. 77–99.
2007 “New Age Relativism and Epistemic Possibility: The Question of Evidence”,
Philosophical Issues 17/1, pp. 262–83.
2008 “Relativism about Truth Itself: Haphazard Thoughts about the Very Idea”, in
M. Garcia-Carpintero and M. Kölbel (eds.) Relative Truth, Oxford, Oxford Univer-
sity Press, pp. 157–86.
2008 “Fear of Relativism”, Philosophical Studies 141/3, pp. 379–90.
2009 With Sebastiano Moruzzi “Trumping Assessments and the Aristotelian Future”,
in B. Brogaard (ed.) Synthese 166/2, pp. 309–31.
2011 “The Pain of Rejection, the Sweetness of Revenge”, Philosophical Studies,
contribution to a book symposium on Mark Richard’s When Truth Gives Out
(Oxford 2008), Philosophical Studies, Online first, August 30 2011.
2012 “Meaning and Assertibility: Source Reflections ou Paolo Casabeguo’s ‘The
Problem of Nau-Conclusiveness’”, Dialectica 66/2, Special number on the philoso-
phy of Paolo Casaleguo, E. Paganini (ed.), pp. 249–66.
“A Plurality of Pluralisms”, in N. Pedersen and C. Wright (eds.) Truth Pluralism:
Current Debates, Oxford, Oxford University Press.
Philosophy of Logic
1977 With Aidan Sudbury, “The Paradox of the Unexpected Examination”, Australian
Journal of Philosophy 55/1, pp. 41–58, reprinted in The Philosophers Annual, Vol.
I (1978), pp. 186–208.
1985 “In Defence of the Conventional Wisdom”, in I. Hacking (ed.) Exercises in

Analysis, Cambridge, Cambridge University Press, pp. 171–97.
1986 “Inventing Logical Necessity”, in J. Butterfield (ed.) Language, Mind and Logic,
1989 “Necessity, Caution and Scepticism”, Proceedings of the Aristotelian Society, Suppl.
Vol. LXIII, pp. 203–38.
2001 “On Being in a Quandary: Relativism, Vagueness, Logical Revisionism”,
Mind 110, pp. 45–98, reprinted in The Philosopher’s Annual, Vol. 24 (2001),
pp. 273–325.
2001 “On Basic Logical Knowledge”, Philosophical Studies, 106, pp. 41–85, reprinted in
J. Bermudez and A. Millar (eds.) Reason and Nature, Oxford, Clarendon Press, pp. 49–84.
2001 “On Knowing What is Necessary: Three Limitations of Peacocke’s Account”,
Philosophy and Phenomenological Research 64, pp. 656–63.
2003 “Rosenkranz on Quandary, Vagueness and Intuitionism”, Mind 112, pp. 465–7.
2003 “Vagueness: A Fifth Column Approach”, in J. C. Beall and M. Glanzberg (eds.)
Liars & Heaps: New Essays on Paradox, Oxford, Oxford University Press, pp. 84–105.
2004 “Intuition, Entitlement and the Epistemology of Logical Laws”, Dialectica 58,
pp. 155–75.
2012 Comment on Paul Boghossian, “The Nature of What is an Inference”, in
Philosophical Studies, online first. March 22 DOI ID 1007/511098-012-9892-9.
Philosophy of Mathematics
1982 “Strict Finitism”, Synthese 51, pp. 203–82.
1985 “Skolem and the Sceptic”, Proceedings of the Aristotelian Society, Suppl. Vol. LIX,
pp. 117–37.
1988 “Why Numbers Can Believably Be: A Reply to Hartry Field”, in P. Kitcher (ed.)
Revue Internationale de Philosophie 42, pp. 425–73.
1990 “Field and Fregean Platonism”, in A. Irvine (ed.) Physicalism in Mathematics,
Dordrecht, Kluwer Academic Publishers, pp. 73–93, reprinted in French translation
in M. Marion and A. Voizard (eds.) Frege: Logique et Philosophie, Montreal, Harmat-
tan 1998, pp. 85–108.
1991 “Wittgenstein on Mathematical Proof”, in A. Phillips Griffiths (ed.) Wittgenstein
Centenary Essays, Royal Institute of Philosophy Supplement 28, Cambridge, Cam-
bridge University Press, pp. 79–99.
1992 With Bob Hale “Nominalism and the Contingency of Abstract Objects”, Journal
of Philosophy 89, pp. 111–35.
1994 With Bob Hale “A Reductio ad Surdum? Field on the Contingency of Mathe-
matical Objects”, Mind 103, pp. 169–84.
1994 “About ‘The Philosophical Significance of Godel’s Theorem’: Some Issues”, in
B. F. McGuinness and G. Oliveri (eds.) The Philosophy of Michael Dummett, Synthese
Library vol. 239, Dordecht, Kluwer, pp. 167–202.
1995 “Intuitionists are not (Turing) Machines”, Philosophia Mathematica 3, pp. 86–102.
1997 “On the Philosophical Significance of Frege’s Theorem”, in R. Heck (ed.)
Language, Thought and Logic: Essays in Honour of Michael Dummett, Oxford, Oxford
University Press, pp. 201–44.
1998 “On the (Harmless) Impredicativity of Hume’s Principle”, in M. Schirn (ed.)
Philosophy of Mathematics Today, Oxford, Oxford University Press, pp. 339–68.
1998 “Response to Dummett”, in M. Schirn (ed.) Philosophy of Mathematics Today,
Oxford, Oxford University Press, pp. 389–405.
1999 “Is Hume’s Principle Analytic?”, Notre Dame Journal of Formal Logic 40/1, pp. 6–30.
2000 With Bob Hale “Implicit Definition and the A Priori”, in P. Boghossian
and C. Peacocke (eds.) New Essays on the A Priori, Oxford, Clarendon Press,
pp. 286–319.
2000 “Neo-Fregean Foundations for Real Analysis: Some Reflections on Frege’s
Constraint”, Notre Dame Journal of Formal Logic 41, pp. 317–34.
2001 With Bob Hale “To Bury Caesar . . .”, in B. Hale and C. Wright, C. The Reason’s
Proper Study: Essays towards a Neo-Fregean Philosophy of Mathematics, Oxford, Clar-
endon Press, pp. 335–96.
2002 With Bob Hale “Benacerraf’s Dilemma Revisited”, European Journal of Philosophy
10/1, pp. 101–29.
2003 With Bob Hale “Responses to Commentators”—book symposium on The
Reason’s Proper Study, Philosophical Books 44/3, pp. 245–63.
2005 With Bob Hale “Logicism in the 21st Century”, in S. Shapiro (ed.) The Oxford
Handbook in the Philosophy of Mathematics and Logic, Oxford, Oxford University Press,
pp. 166–202.
2006 With Stewart Shapiro “All Things Indefinitely Extensible”, in A. Rayo and
G. Uzquiano (eds.) Absolute Generality, Oxford, Oxford University Press, pp. 255–304.
2007 “On Quantifying into Predicate Position: Steps towards a New(tralist) Perspective”,
in M. Potter (ed.) Mathematical Knowledge, Oxford, Oxford University Press, pp. 150–74.
2008 With Bob Hale “Abstraction and Additional Nature”, Philosophia Mathematica
16/2, pp. 182–200.
2008 Contribution to V. Hendriks and H. Hannes Leitgeb (eds.) Philosophy of Mathe-
matics: 5 Questions, Denmark, Automatic Press/VIP, pp. 301–11.
2009 “The Metaphysics and Epistemology of Abstraction”, in A. Hieke and H. Leitgeb
(eds.) Reduction–Abstraction–Analysis, Heusenstamm bei Frankfurt, Ontos Verlag,
pp. 195–216.
2009 With Bob Hale “Focus Restored: Comment on John Macfarlane’s ‘Double
Vision’”, in O. Linnebo (ed.) Synthese 170/3, pp. 457–82.
2009 With Bob Hale “The Metaontology of Abstraction”, in D. J. Chalmers,
D. Manley and R. Wasserman (eds.) Metametaphysics: New Essays on the Foundations
of Ontology, Oxford, Oxford University Press, pp. 178–212.
Forthcoming “An Entitlement to Hume’s Principle?”, in P. Ebert and M. Rossberg

(eds.) The Philosophy of Abstraction, Oxford, Oxford University Press.
“Whence the Paradox? Axiom V and Indefinite Extensibility”, in M. Frauchiger
and W. K. Essler (eds.), volume to celebrate the award of the 2010 Lauener prize
to Sir Michael Dummett, to appear in the series, The Lauener Library of Analytical
Philosophy.
“Frege and Benacerraf’s Problem”, in R. DiSalle, M. Frappier and D. Brown (eds.)
Analysis and Interpretation in the Exact Sciences: Essays in Honour of William Demopoulos,
Springer.
Wittgenstein, Mind, Rules and Privacy
1981 “Rule-following, Objectivity and the Theory of Meaning”, in C. Leich and

S. Holtzman (eds.) Wittgenstein: To Follow a Rule, London, Routledge, pp. 99–117.
1984 “Kripke’s Account of the Argument Against Private Language”, Journal of
Philosophy 81, pp. 759–78.
1986 “Rule-following and Constructivism”, in C. Travis (ed.) Meaning and Interpreta-
tion, Oxford, Blackwell, pp. 271–97.
1986 “Does Philosophical Investigations 258–60 Suggest a Cogent Argument Against
Private Language”, in P. Pettit and J. McDowell (eds.) Subject, Thought and
Context, Oxford, Oxford University Press, pp. 209–66, reprinted in Wright
(2001), pp. 223–90.
1987 “On Making Up One’s Mind: Wittgenstein on Intention”, in P. Weingartner
and G. Schurz (eds.) Logic, Philosophy of Science and Epistemology, Proceedings of the
XIth International Wittgenstein Symposium, Vienna, Holder-Pickler-Tempsky,
pp. 391–404.
1989 “Wittgenstein’s Rule-following Considerations and the Central Project of The-
oretical Linguistics”, in A. George (ed.) Reflections on Chomsky, Oxford, Blackwell,
pp. 233–64, reprinted in Wright (2001), pp. 170–214.
1989 Critical Study of Colin McGinn’s Wittgenstein on Meaning, Mind 98, pp. 289–305.
1989 “Wittgenstein’s Later Philosophy of Mind: Sensation, Privacy and Intention”,
Journal of Philosophy 76/11, pp. 622–34, extended version in K. Puhl (ed.) 1991
Meaning Scepticism, Berlin, de Gruyter, pp. 126–47.
1993 “Eliminative Materialism: Going Concern or Passing Fancy?” in Mind and
Language 8, pp. 316–26.
1995 “Can there be a Rationally Compelling Argument for Anti-realism about
Ordinary (‘Folk’) Psychology?”, Philosophical Issues 6, pp. 197–221.
1998 “Self-knowledge: The Wittgensteinian Legacy” (shortened version) in
A. O’Hear (ed.) Current Issues in Philosophy of Mind, Royal Institute of Philosophy
Supplement 43, Cambridge, Cambridge University Press, pp. 101–22.
1998 “Self-knowledge: The Wittgensteinian Legacy”, in C. Wright, B. Smith and

C. Macdonald (eds.) Knowing Our Own Minds, Oxford, Oxford University Press,
pp. 15–45.
2007 “Rule-following without Reasons: Wittgenstein’s Quietism and the Constitu-
tive Question”, in J. Preston (ed.) Wittgenstein and Reason, Ratio 20/4, pp. 481–502.
2012 “Reflections on François Recanati’s Immunity to Error through Misidentification:
What It Is and Where It Comes From”, in S. Prosser and F. Recanati (eds.) Immunity to
Error Through Misidentification: New Essays, Cambridge, Cambridge University Press,
pp. 247–80.
SECTION I
Rule-Following and the
1
Blind Rule-Following
Paul A. Boghossian
1. Introduction
It is a great pleasure to be able to contribute to this Festschrift in honor of Crispin Wright,
with whom I have enjoyed countless stimulating conversations about a host of funda-
mental philosophical issues over the past twenty years. It is especially appropriate that my
contribution to this Festschrift concern the topic of rule-following for this topic has been
central to both of our preoccupations and has dominated our discussions.
As anyone with the slightest familiarity with this subject will know, Wright has
written several important and illuminating papers on the phenomenon of rule-following,
papers that draw their inspiration from Wittgenstein’s seminal discussion of that notion.
In those papers, Wright lays out an original view both about what the rule-following
problem is, or should be taken to be, and about its correct resolution. In both respects, his
view differs from that of Saul Kripke’s influential account.1
Kripke’s well-known view is that there is an enormous skeptical problem seeing
how rule-following is so much as possible.2 Wright maintains that this problem is
misguided, that Kripke’s skeptical challenge can receive a relatively straightforward
solution. According to Wright, we can follow rules by forming intentions to uphold
certain patterns in our thought or behavior and by acting on those intentions. Call this
the Intention View of rule-following.
The real problem posed by Wittgenstein’s discussion, Wright continues, is not to
explain the very possibility of rule-following but, rather, to explain how we can have
the sort of privileged access to the contents of our intentions—and of our intentional
states more generally—that we normally credit ourselves with.
I shall argue for three claims. First, Wright is correct to think that it is possible to
respond to the specific challenge that Kripke poses with the Intention View. For
1
Wright’s papers are collected in Wright 2001. For Kripke’s view, see Kripke 1982. Wright’s view of
rule-following in Rails is substantially different from the view he presented in Wright 1980. In more recent
work, Wright 2007, Wright independently explores ideas that are much closer in spirit to the conclusions of
this essay than his earlier views.
2
Caveat: Kripke presents this as an exposition of Wittgenstein’s view rather than his own.
28 PAU L A . B O G H O S S I A N
reasons that will emerge, this is not quite as straightforward as Wright seems to suppose,
but it is ultimately defensible.
Second, there is little doubt that Wright’s problem about self-knowledge of inten-
tional states is a real problem, although I believe that his own proposed solution to that
problem is unlikely to succeed.
Finally, though, I shall argue that while the Intention View may constitute an
adequate response to the specific considerations adduced by Kripke, it cannot in the
end be regarded a correct account of rule-following. In its most fundamental incarna-
tion, I shall argue, rule-following cannot consist in some intentional fact.
In and of itself this result might be considered significant but not necessarily
dispiriting. Matters begin to look significantly worse, however, when we combine
this result with the powerful arguments to be found, in Kripke and elsewhere, that
threaten to show that rule-following cannot consist in some non-intentional fact either.
And the full depth of our problem emerges when we realize that skepticism about
following a rule is not ultimately a coherent option.
If all of this is right, then we face what might be called, following Kant, an antinomy
of pure reason: we both must—and cannot—make sense of someone’s following a
rule. That is what I propose to argue for in this essay.
2. What is a Rule? What is Rule-Following?

We should begin with a very basic question: What are we talking about when we talk
about someone following a rule? I don’t mean: What is the right theory of rule-
following? I mean to be asking in an intuitive and pre-theoretic manner: what
phenomenon is at issue? It is surprising how often writers will launch into a discussion
of this topic without pausing long enough to give it an intuitive characterization,
especially since the answer turns out to be anything but obvious.
The question—What is it to follow a rule?—naturally breaks up into two. What is a
rule? And what is to follow one?
The first question is surprisingly tricky, though limitations of space prevent me from
discussing it in the requisite detail.3
Clearly, when we talk about people following rules we mean that they are somehow
or other observing (or attempting to observe) certain general principles or standards. But
there are at least two importantly different ways of conceiving such principles.
On the one hand, we can think of rules as “directions” or “instructions”—i.e.
contents that are expressed by imperatives of the form: If C, do A! On this conception,
which appears to be Kripke’s, rules are contents that prescribe certain patterns of
behavior under certain conditions. This is certainly a very common way of under-
standing the notion of a rule.
3
For more discussion see Boghossian 2008.
B L I N D RU L E - F O L L OW I N G 29
However, not everything that we call a rule in ordinary language, or in the course of
philosophical theorizing, conforms to this characterization. For example, we talk about
the “rules of chess.” One of these rules is:
(Castle) If the configuration is C, you may castle.

This is not an imperative but, rather, what I would call a normative proposition. It is a
norm of permission. It cannot be expressed by the imperative
If configuration is C, castle!
because that would suggest that whenever the configuration is C, you must castle,
whereas the rule for castling merely permits castling and does not require it. Indeed,
I would argue, but won’t do so here, that there is no good way to express a norm of
permission in imperatival terms.4
So we have a distinction between an imperatival content and a normative proposi-
tion and we need to decide whether, when we talk about following rules, we are
talking about following the one sort of content or the other.
For a variety of reasons, I am inclined to think that the more fundamental notion is
that of a normative proposition and not that of a prescription or direction. But I won’t
argue for it here.5 For the purposes of the present essay, I will take a rule to be either a
general normative proposition or a general imperatival content.
Let me turn instead to asking about the more pressing question: What is it to follow a rule?
In answering this question, we should distinguish between a personal-level notion of
rule-following and a subpersonal notion. We should not assume, at the outset, that our
talk of a person’s following a rule comes to exactly the same thing as our talk of, say, his
brain’s following a rule, or of his calculator’s computing a function.
I propose to start with attempting to understand the personal-level notion, returning
to the subpersonal notion later. My view will be that there is a core concept that is
common to both notions, but that the personal-level notion is richer in a particular
respect that I shall describe below. Once we have a handle on the personal-level notion
it will be easy to indicate the weakening that gets us the subpersonal notion.
Apropos of the personal-level notion, we certainly know this much: to say that S is
following rule R is not the same as saying that S’s behavior conforms to R. Conforming
to R is neither necessary nor sufficient for following it.
It is not necessary because S may be following R even while he fails to conform to it.
This can happen in one of two ways. Say that R is the instruction ‘If C, do A!’ S may
fail to recognize that he is in circumstance C, and so fail to do A; yet it may still be true
that S is following R. Or, he may correctly recognize that he is in C, but, as a result of a
performance error, fail to do A, even though he tries.
4 5
For further discussion, see ibid. Again, see ibid.
Conformity to R is not sufficient for S’s following R because, for any behavior that
S displays, there will be a rule—indeed, infinitely many rules—to which his behavior
will conform. Yet it would be absurd to say that S is following all the rules to which his
behavior conforms.
There is another possible gloss on our notion that we need to warn against. There is
a persistent tendency in the literature to suggest that the claim that S is following rule
R means something roughly like: R may correctly be used to evaluate S’s behavior.
I am not confident that this construal can be pinned on Wright; but it is suggested by
the remarks with which he tends to introduce the topic of rule-following, for example:
The principal philosophical issues to do with rule-following impinge on every normatively
constrained area of human thought and activity: on every institution where there is right and
wrong opinion, correct and incorrect practice.6
Whether or not this can be attributed to Wright, it is worthwhile seeing what is

wrong with it. Intuitively, and without the help of controversial assumptions, it looks
as though there are many thoughts that S can have, and many activities that he can
engage in, that are subject to assessment in terms of rule R even if there is no intuitive
sense in which they involve S’s following rule R.
Consider Nora playing roulette. She has a “hunch” that the next number will be
‘36’ and she goes with it: she bets all her money on it. We need not suppose that, in
going with her hunch, she was following any rule—perhaps this was just a one-time
event. Still, it looks as though we can normatively criticize her belief as irrational since it
was based on no good evidence.
Or consider Peter who has just tossed the UNICEF envelope in the trash without
opening it. Once more, we need not suppose that Peter has a standing policy of tossing
out charity envelopes without opening them and considering their merits. However,
even if no rule was involved it can still be true that Peter’s behavior was subject to
normative assessment, that there are norms covering his behavior.
In both of these cases, then, a norm or rule applies to some thought or behavior even
though there is no intuitive sense in which the agent in question was attempting to
observe that norm or rule.
Of course, some philosophers—like Kripke’s Wittgenstein—think that wherever
there is intentional content there must be rule-following, since meaning itself is a matter
of following rules. But that is not a suitably pre-theoretic fact about rule-following; and
what we are after at the moment is just some intuitive characterization of the phenom-
enon. We will come back to the question whether meaning is a matter of following
rules.
When we say that S is following a rule R in doing A, we mean neither that
S conforms to R nor simply that R may be used to assess S’s behavior, ruling it correct
if he conforms and incorrect if he doesn’t. What, then, do we mean?
6
Wright 2001, p. 1.
Let us take a clear case of personal-level rule-following. Suppose I receive an email

and that I answer it immediately. When would we say that this behavior was a case of
following the:
(Email Rule) Answer any email that calls for an answer immediately upon receipt!
as opposed to just being something that I did that happened to be in conformity with
that rule?
I think it is clear that it would be correct to say that I was following the Email Rule
in replying to the email, rather than just coincidentally conforming to it, when it is
somehow or other because of the Email Rule that I reply immediately.
Equally clearly, the because here is not any old causal relation: if a malicious scientist
(or an enterprising colleague) had programmed my brain to answer any email upon
receipt (in some zombie-like way) because he accepted the rule that I should answer
any email upon receipt, that would not count as my following the Email Rule.
(It might count as my brain following the rule.) Rather, for me to be following the
rule, the ‘because’ must be that of rational action explanation: I follow the Email Rule
when that rule serves as my reason for replying immediately, when that rule rationalizes
my behavior.
I want to suggest, then, that the minimal content of saying that person S follows rule
R in doing A is that R serves as S’s reason for doing A.
Now, R is just a content of some sort, either an imperative or a normative
proposition, as previously discussed. How is it possible for a content to serve as S’s
reason for doing something? Obviously, by being accepted or internalized by him.
I shall typically refer to this as S’s acceptance or internalization of the rule, though,
clearly, it will be very important to understand this as neutrally as possible for now. 7
However exactly it is understood, what is important is that, in any given case of rule-
following, we have something with the following structure: a state that can play the
role of rule-acceptance; and some non-deviant casual chain leading from that state to a
piece of behavior that would allow us to say that the accepted rule explains and (in the
personal-level case) rationalizes the behavior in question.
Occasionally, I will also describe the matter in terms of the language of commit-
ment: In rule-following, I will say, there is, on the one hand, a commitment, on the part
of the thinker, to uphold a certain pattern in his thought or behavior; and, on the other,
some behavior that expresses that commitment, that is explained and rationalized by it.
I will leave it to the reader to discern whether I have construed these notions in a
way that is illicit or question begging. For the moment, let me just note that this
characterization coincides well with the way Kripke seems to be thinking about the
phenomenon of rule-following. As he says apropos of following the rule for addition:
7
“Internalization” is Kripke’s preferred word, as we shall see below; it is probably more neutral than
“acceptance.”
I learned—and internalized instructions for—a rule, which determines how addition is to be

continued . . . This set of directions, I may suppose, I explicitly gave myself at some earlier
time . . . It is this set of directions . . . that justifies and determines my present response.8
I think it was a mistake on Kripke’s part to use the word “justify” in this passage,
rather than the word “rationalize.” In talking about rule-following, it is important to
bear in mind that we might be following bad rules. The problem of rule-following
arises no less for Modus Ponens than it does for Affirming the Consequent or
Gambler’s Fallacy. If I am following Gamblers’ Fallacy, my betting big on black after
a long string of reds at the roulette wheel wouldn’t be justified; but it would be
rationalized by the rule that I am following. Given that I am committed to the fallacious
rule, it makes sense that I would bet big on black at that point.
We may summarize our characterization of personal-level rule-following by the
following four theses:
(Acceptance) If S is following rule R (‘If C, do A’), then S has somehow accepted R.

(Correctness) If S is following rule R, then S acts correctly relative to his acceptance if
it is the case that C and he does A.
(Explanation) If S is following rule R by doing A, then S’s acceptance of R explains
S’s doing A.
(Rationalization) If S is following rule R by doing A, then S’s acceptance of
R rationalizes S’s doing A.
With this characterization of the personal-level notion in place, it is possible, I think,
to see the subpersonal notion of following a rule as involving the first three elements
but not the fourth.
If I say of a calculator that it is adding, then I am saying that its ‘internalization’ of the
rule for addition (via programming) explains why it gives the answers that it gives. But
I am obviously not saying that the addition rule rationalizes the calculator’s answers.
The calculator doesn’t act for reasons, much less general ones.9
3. Acceptance, Intention, and Wright’s Problem

With these important preliminaries behind us, let us turn to asking why there is
supposed to be a problem about following a rule. What, in particular, does Kripke
find so mystifying about it?
Kripke’s problem is focused on the personal-level notion and on the Acceptance
condition for it. He is struck by two facts. First, by the fact that most of the rules we
follow are rules with infinitary contents; and, second, by the fact that our rules are
supposed to rationalize (he says “justify”) the behavior that constitutes following them.
8
Kripke 1982, p. 16.
9
This account summarizes a large number of considerations that I do not have the space to describe in
detail here.
And he is mystified by how it might be possible for finite minds like ours to instantiate
the sort of state that would have both features. What could rule acceptance be such that
it could have both of these requisite features?
Kripke illustrates his problem by considering the case of the symbol ‘+’. In using this
symbol, he supposes, I may be taken to be following the rule for addition. In what does
this fact consist? There look to be two serious candidates: either in some fact about my
dispositions to use the symbol, or in some sort of intentional fact about me. Kripke
finds both candidates wanting.
Now, Wright agrees with Kripke’s rejection of dispositionalism. He maintains,
however, that the intentional suggestion emerges unscathed from Kripke’s skeptical
considerations. As he puts it:
so far from finding any mystery in the matter, we habitually assign just these characteristics [the
characteristics constitutive of the acceptance of a rule] to the ordinary notion of intention . . .
intentions may be general, and so may possess, in the intuitively relevant sense, potentially
infinite content.10
This is the position that I earlier called the Intention View.11

The Intention View, of course, is just a special version of a more general class of
views according to which rule acceptance consists in some intentional state or other,
even if it is not identified specifically with an intention. Call this more general view the
Intentional View of rule acceptance. Although I will follow Wright in focusing on the
Intention View, most everything I say will apply to the less committal Intentional
View.
Wright considers the Intention View to be a perfectly adequate response to Kripke’s
question, as far as it goes. He believes the real difficulty lies not in explaining how
it is possible for there to be such infinitary commitments, but rather in explaining
how it is possible for a subject to have the sort of authoritative self-knowledge about
them that we seem to have. For Wright takes it to be characteristic of the intuitive
notion of intention, that
it is a state of mind, alongside mood, thought, desire, sensation, etc., for which, in at least a very
large class of cases, subjects have special authority and whose epistemology is first/third person
asymmetric.12
And he takes the difficult question to be how states with such epistemologies are
possible.
10
Wright 2001, pp. 125–6.
11
See also Pettit 2002, p. 27: “The notion of following a rule, as it is conceived here, involves an
important element over and beyond that of conforming to a rule. The conformity must be intentional, being
something that is achieved at least in part, on the basis of belief and desire. To follow a rule is to conform to it,
but the act of conforming, or at least the act of trying to conform—if that is distinct—must be intentional. It
must be explicable, in the appropriate way, by the agent’s beliefs and desires.”
12
Wright 2001, p. 125.
I think there is no doubt about two points. First, that there are many instances of
rule-following that are well captured by the Intention View, the email example
outlined above being one of them. One adopts the email rule by forming a general
intention to answer any email upon receipt and by having this intention subsequently
inform and control one’s behavior.
The second point on which we can all agree is that Wright’s worry is a genuine one,
in as much as it is very difficult to explain how we are able to have authoritative first-
personal knowledge of our intentional states. There are at least two problems here
(externalist conceptions of content would add a third).
First, intentional states typically lack an individuative phenomenology. How, then,
are we able to introspect them so reliably? And why are they first-person/third-person
asymmetric?
Second, intentional attributions often behave as though what is getting attributed is a
capacity or a disposition to do something, and not some state of mind which could be
wholly present to the subject’s consciousness at a given moment in time.13 Thus, we
allow that Doron can authoritatively avow at time t that he wants to play chess, even if
he did not explicitly think at t of all the rules that chess consists in. But this avowal will
be defeated if Doron’s subsequent performance falls short of the standard implicit in the
attribution—if for example, he goes on to show that he doesn’t understand the rules of
chess. How can we reconcile these two facets of the attribution of an intentional state?
By way of solving these problems, Wright experiments with treating avowals as fact-
constituting rather than fact-detecting judgments. Given the difficulty of accounting
for these judgments in classically Cartesian terms, this is clearly a line of research that is
worth pursuing.
But it is hard to see how to pull it off: it’s hard to see how it could in general be true
that facts about mental content are constituted by our judgments, since, it would
appear that, by the terms of the theory itself, facts about the contents of the putatively
fact-constituting judgments would have to be constituted independently of such
judgments.14 Still, I agree with Wright that this consideration doesn’t decisively refute
such views and that there are many interesting questions about them that remain to be
investigated.
4. Problems for the Intention View

But what concerns me in this essay is whether it is true that we can solve Kripke’s
problem as easily as all that? Can we really just appeal to intentions with infinitary
contents to explain how the Acceptance condition on rule-following gets fulfilled?
I can think of three reasons that may be found in the literature why someone might
resist Wright’s Intention View.
13
Wright 2001, pp. 134 ff.
14
For more discussion, see Boghossian 1989, pp. 544–8.
The first reason would be provided by the assumption of Naturalism. A Naturalist

would insist that intentions be shown to be naturalistically reducible before they could
legitimately be appealed to in solving Kripke’s problem. However, it is none too clear
how such a reduction is to be pulled off and Kripke’s book provides a battery of
arguments against its feasibility (more on this below).
Second, and even if we were to put the Naturalist Assumption to one side, there look to
be straightforward counterexamples to the Intention View: not everything that we would
intuitively count as rule-following is intentional action in the sense that it specifies.
To be sure, and as I have already emphasized, there are many cases of rule-following
that are captured very well by the Intention View, the email rule discussed above being
one of them. But there also seem to be several significant cases of rule-following in
which it would be implausible to say that there was a general intention to conform to a
pattern involved.
The trouble comes in part (and ironically) from a feature of intentions that Wright
himself emphasizes—namely, that their contents are typically thought of as highly
accessible to their owners. The problem is that we appear to follow many rules that we
aren’t able to specify at all precisely.
Take, for example, the rules by which we regulate our moral conduct. Are we
Kantians or consequentialists? It is, of course, a highly controversial normative matter
which rules are the morally correct ones. But even just as a matter of descriptive fact, it
is controversial whether we employ deontological principles, consequentialist ones, or
some confused mixture of the two. If these moral rules were the contents of intentions
of ours, wouldn’t we expect to know what they are with a much higher degree of
precision and clarity? A similar point could be made using principles of etiquette or the
epistemic rules by which we update our beliefs.
A third type of consideration against the Intention View is provided by an assump-
tion that is crucial to Kripke’s thinking about rule-following. Kripke sets up the rule-
following problem by asking what determines whether I am using the ‘+’ sign
according to the rule of addition as opposed to the rule for quaddition, where
quaddition is a function just like addition, except that it diverges from it for numbers
larger than we are able to compute. He considers saying that what determines that rule-
following fact is some general intention I formed to use the symbol according to the
one rule rather than the other:
What was the rule? Well, say, to take it in its most primitive form: suppose we wish to add x and
y. Take a huge bunch of marbles. First count out x marbles in one heap. Then count out y
marbles in another. Put the two heaps together and count out the number of marbles in the
union thus formed. The result is x+y. This set of directions, I may suppose, I explicitly gave
myself at some earlier time. It is engraved on my mind as on a slate. It is incompatible with the
hypothesis that I meant quus. It is this set of directions, not the finite list of particular additions
I performed in the past that justifies and determines my present response.
Kripke continues:
Despite the initial plausibility of this objection, the sceptic’s response is all too obvious: True, if
‘count’ as I used the word in the past, referred to the act of counting (and my other words are
correctly interpreted in the standard way) then ‘plus’ must have stood for addition. But I applied
‘count’ like ‘plus’ to only finitely many past cases. Thus the sceptic can question my present
interpretation of my past usage of ‘count’ as he did with ‘plus’.15
How should we understand this passage? On one way of reading it, Kripke would be
assuming that the contents of mental states are derived from the contents of linguistic
expressions. But if that’s the assumption, it is vulnerable: most philosophers think that
the relation between mind and language is in fact the other way round, that linguistic
meaning derives from mental content.
On another way of reading it, Kripke would be assuming not some controversial
view of the relation between mind and language, but rather just the familiar ‘language
of thought’ picture of thought—that thoughts themselves involve the tokenings of
expressions (of mentalese)—and claiming that those expressions, too, get their meaning
by our following rules in respect of them.
I think this latter assumption is clearly what Kripke had in mind. Let’s call it Kripke’s
Meaning Assumption and let’s go along with it for now.
Now, it should be obvious that combining the Meaning Assumption with the
Intention View will lead rather quickly to the conclusion that rule-following, and
with it mental content, are metaphysically impossible. For given the two assumptions,
we would be able to reason as follows. In order to follow rules, we would antecedently
have to have intentions. To have intentions, the expressions of our language of thought
would have to have meaning. For those expressions to have meaning, we would have
to use them according to rules. For us to use them according to rules, we would
antecedently have to have intentions. And so neither content nor rule-following
would be able to get off the ground.
Since Kripke regards the Meaning Assumption as non-optional, he rejects the
Intention View. The problem then becomes to find a way in which someone could
be said to have committed himself to a certain pattern of use for a symbol without this
being the result of his forming an intention (or other intentional state) to uphold that
pattern. And that is why so much attention is focused on the dispositional view.
5. Are there Solutions to these Problems

for the Intention View?
Let us take stock. If the second of the three considerations we have outlined is correct,
then there must be a species of rule-following that is non-intentional.
And, if either the first or the third of our three considerations is correct, then not
only must there be a species of rule-following that is non-intentional, all rule-following
15
Kripke 1982, p. 16.
must be at bottom non-intentional, because even intentional forms of rule-following

will presuppose the non-intentional kind.
Now, since we know that it is going to be extremely difficult to make sense of rule-
following in purely non-intentional, dispositional terms, we should ask whether there
is any way around these considerations. Can they be rebutted?
To the first objection, one might respond by saying that Naturalism is not obviously
correct and so can hardly be used to constrain the acceptability of an otherwise intuitively
compelling theory of rule-following. After all, it continues to prove difficult to account
for consciousness within a naturalistic setting. Kripke tried to argue that there would be
something irredeemably queer about an intentional state that is simply taken unreduced,
without hope of reconstruction in terms of materials that would be naturalistically
acceptable. But I agree with Wright that his argument here is not successful and doesn’t
rise above begging the question against the anti-reductionist suggestion.
To the objection based on cases of alleged non-intentional rule-following, one
could try responding by introducing the notion of a tacit intention, an intention to do
something that is not explicitly articulated in someone’s consciousness but which the
thinker could be said to have implicitly or tacitly. Specifying such a notion in a
satisfactory way would obviously be a huge undertaking and has not yet been shown
to be feasible. But it is not clearly hopeless.
However, even if the foregoing responses were accepted, I hope it is clear that we
would still be stuck with a huge problem for the Intention View, if Kripke’s Meaning
Assumption is left in place. The problem, of course, is that even unreduced, tacit
intentions are contentful states and so it would still not be possible to combine the
Intention View with the Meaning Assumption. But can the Meaning Assumption be
plausibly discarded?
Let’s distinguish between the question whether public language expressions get their
meaning through rule-following and the question whether the expressions of the language
of thought do.
Is the Meaning Assumption correct at least when it comes to the words of public
language? Is it right to say that the words of English, for example, get their meaning as a
result of our following rules in respect of them?
Well, a word is just an inscription, a mark on paper. Something has got to be done to
it by its user for it to get a meaning. That much is clear.
It is also clear that meaningful words have conditions of correct application. Thus, the
word ‘tiger’ is correctly applied only to tigers and the word ‘red’ only to red things.
But it doesn’t follow from these obvious truths that the way the word ‘tiger’ comes
to mean what it does for a given speaker S—the way it comes to have the correctness
conditions that it has in S’s idiolect—is by S committing himself to using it according to
the rule: Apply the word ‘tiger’ only to tigers!16
16
Jerry Fodor may have been the first to appreciate this clearly; see Fodor 1990, pp. 135–6. I don’t believe
that any of the main arguments of Boghossian 1989 would be affected by paying greater heed to this
distinction, although I am sure I wasn’t as clear about it in that paper as I should have been.
For meaning to be a matter of rule-following in the sense presupposed by the

Meaning Assumption, it must be true not only that words have satisfaction conditions
but that they get their satisfaction conditions by their users committing themselves to
using them according to certain patterns.
Still, it does look as though one can make a strong case for the Meaning Assumption
as applied to public language expressions.17 When I apply the word ‘tiger’ to a newly
encountered animal, it is very natural to think that my application of the word is guided
and rationalized by my understanding of its meaning, an understanding that is general
and which determines what the word does and does not apply to.
However one may feel about the relation between public language and rules,
though, there looks to be very little prospect that the Meaning Assumption applies at
the level of mental expressions.
Since we are dealing with a personal-level notion of rule-following, it makes very
little sense to say that we follow rules in respect of our mental expressions, expressions
that we have no access to and which, for all that the ordinary person knows, may not
even exist. (Whether we should regard their meaning as generated by subpersonal rule-
following is a question that I shall come back to.)
So, here, then, is the problem for Kripke’s Meaning Assumption that I alluded to
at the beginning. Kripke is clearly working with a personal-level notion of rule-
following. That is why he can confidently claim that when someone is following a
rule that rule justifies (rationalizes) his behavior. But it can hardly be true that all
meaning is a matter of rule-following in this sense. In particular, it can hardly be true
that the expressions of mentalese get their meaning by our following rules in respect
of them in this sense.
So, it looks as though we are free to reject Kripke’s Meaning Assumption, at least as
it applies to mental expressions. And with that we seem to have answered the third of
the three objections we had posed for the Intention View.
If we reject the Meaning Assumption, we give up on the claim that mental
expressions get their meaning by our following rules in respect of them. How, then,
do they get their meaning?
Kripke’s discussion may be seen as containing a battery of effective arguments
against reductive accounts of meaning facts. But as I have already mentioned above,
and argued at length elsewhere, his arguments against anti-reductionist accounts are
rebuttable.18
If we adopt such an anti-reductionist conception of mental content, doesn’t that
mean that we are now free to adopt the Intention View of rule-following?
17
I have gone back and forth about the plausibility of the Meaning Assumption as applied to public
language expressions. In my NYU seminar of Spring 2006, I defended it, but in an earlier version of this essay,
I retreated to saying that it was not settled. I thank Christopher Peacocke for rightly insisting to me that it met
my characterization of person-level rule-following.
18
See Boghossian 1989.
6. Can the Intention View be Saved?

Not quite. For what I now want to argue is that even if all of these responses were to
pan out, that still wouldn’t suffice to salvage the Intention View. The Intention View
suffers from a further and seemingly fatal flaw. It concerns not, as on Kripke’s view,
Acceptance, but rather, Explanation or Rationalization.
To see what it is, let us waive Naturalism; let us ignore the examples of putatively
non-intentional forms of rule-following; and let us reject the Meaning Assumption.
And let us simply help ourselves to an anti-reductionist view of mental content.
Once such contentful thoughts are available, they can be used to frame intentions—
and so, it would seem, to account for our acceptance of rules. If something like this
picture could be sustained, would that imply that there is nothing left of the rule-
following problem?
In a passage whose import I believe many commentators have missed, Wittgenstein
seems to indicate the answer is No—even if we could simply help ourselves to the full
use of intentional resources, Wittgenstein appears to be saying, there would still be a
problem about how rule-following is possible.
The passage I have in mind is at Philosophical Investigations 219. In it Wittgenstein
considers the temptation to say that when we commit ourselves to some rule, that rule
determines how we are to act in indefinitely many future cases:
“All the steps are really already taken,” means: I no longer have any choice. The rule, once
stamped with a particular meaning, traces the lines along which it is to be followed through the
whole of space.
If we were reading this with Kripke’s eyes, what would we expect Wittgenstein to
say in reply? Something along the following lines (with absolutely no aspiration to
capturing Wittgenstein’s literary tone):
And how did you get to stamp the rule with a particular meaning so that it traces the lines along
which it is to be followed through the whole of space? To do that you would need to be able to
think, to frame intentions. But that assumes that we have figured out how we manage to follow
rules in respect of mental expressions. And that is something that we have not yet done.
But what Wittgenstein says in reply is rather this:

But if something of this sort really were the case, how would it help?
Even if we were to grant that we could somehow imbue the rule with a meaning
that would determine how it applies in indefinitely many cases in the future, Wittgen-
stein seems to be saying, it would still not help us understand the phenomenon of rule-
following.
How mystifying this must seem from a Kripkean point of view. How would it help?
How could it not help? We wanted an answer to the question: By virtue of what is it
true that I use the ‘+’ sign according to the rule for addition and not some other rule?
According to the picture currently under consideration, one of our options is to say
that it is by virtue of the fact that I use the ‘+’ sign with the intention that its use
conform to the rule for addition and where it is understood that the availability of such
intentions is not itself a function of our following rules in respect of them. Under the
terms of the picture in place, what would be left over?
How should we understand what Wittgenstein is saying here? It is, of course, always
hard to be confident of any particular interpretation of this philosopher’s cryptic
remarks; but here is a suggestion that seems of independent philosophical interest.
Let us revert to our email example. Suppose I have adopted the rule: Answer any
email (that calls for an answer) immediately upon receipt. And let us construe my
adoption of this rule as involving an explicit intention on my part to conform to the
instruction:
Intention: For all x, if x is an email and you have just received x, answer it
immediately!
Now, how should we imagine my following this rule? How should we imagine its
guiding, or explaining, the conduct that constitutes my following it?
To act on this intention, it would seem, I am going to have to think, even if very
fleetingly and not very consciously, that its antecedent is satisfied. The rule itself, after all,
has a conditional content. It doesn’t call on me to just do something, but to always
perform some action, if I am in a particular kind of circumstance. And it is very hard to
see how such a conditional intention could guide my action without my coming to have
the belief that its antecedent is satisfied. So, let us imagine, then, that I think to myself:
Premise: This is an email that I have just received.

in order to draw the
Conclusion: Answer it immediately!

At least in this case, then, rule-following, on the Intention model, requires inference:
it requires the rule-follower to infer what the rule calls for in the circumstances in
which he finds himself.
At least in this case, then, rule-following, on the Intention model, requires some sort
of inference.
In this regard, though, the email case is hardly special. Since any rule has general
content, if our acceptance of a rule is pictured as involving its representation by a
mental state of ours, an inference will always be required to determine what action the
rule calls for in any particular circumstance. On the Intention View, then, applying a
rule will always involve inference.
Inference, however, is an example of rule-following par excellence. In the email
case, in moving from the intention, via the premise about the antecedent, to the
conclusion, I am relying on a general rule that says that from any such premises I am
entitled to draw such-and-so conclusion. Since, as I have set up the example, I have
construed the email rule as an imperative, this isn’t quite Modus Ponens, of course, but
it is something very similar:
(MP*) From ‘If C, do A’ and C, conclude ‘do A’!

But now: If on the Intention View, rule-following always requires inference; and if
inference is itself always a form of rule-following, then the Intention View would look
to be hopeless: under its terms, following any rule requires embarking upon a vicious
infinite regress in which we succeed in following no rule.
To see this explicitly, let us go back to the email case. On the Intention View,
applying the Email Rule requires, as we have seen, having an intention with the rule as
its content and inferring from it a certain course of action. However, inference, we
have said, involves following a rule, in this case, MP*. Now, if the Intention View is
correct, then following the rule MP* itself requires having an intention with MP* as its
content and inferring from it a certain course of action. And now we would be off on a
vicious regress: inference rules whose operation cannot be captured by the intention-
based model are presupposed by that model itself.19
This argument bears an obvious similarity to Lewis Carroll’s famous argument in
“What the Tortoise Said to Achilles.”20 The Carrollian argument, however, is meant
to raise a problem for the justification of our rules of inference—how can we justify our
belief, for example, that Modus Ponens is a good rule of inference?
The argument I am putting forward, though, raises an even more basic problem for
how it is possible to follow an inference rule of any kind, good or bad, justified or
unjustified. Even if it were Affirming the Consequent that was at issue, the problem
I am pointing to would still arise.
It would seem, then, that there would still be a problem with the Intention View
even if we somehow managed to resolve all the other difficulties that we outlined for it.
The mere combination of the Intention View and a Rulish of inference are sufficient
for generating a problem.
7. Intentions and Intentional States

How should we proceed? I have been talking about the Intention View, but, of course,
everything I’ve been saying will apply to any Intentional View. So let me restate our
problem in full generality exposing as many of our assumptions as possible.
19
This, I believe, is the correct interpretation of Wittgenstein’s remarks about needing a rule to interpret a
rule. In the Kripkean framework, this is read as supposing that a rule can only be given to you as an inert sign
whose meaning you would then have to divine. And this sets off an infinite regress of interpretations.
However, a different way of reading Wittgenstein here is to see him as concerned not with the question:
“How could an inert sign guide us, if not through the use of further rules?” But rather with the question:
“How could a general content guide us, if not through the use of further rules?”
20
See Carroll 1895. There is also a similarity to Quine’s arguments in Quine 1936.
The claim is that the following five propositions form an inconsistent set.
1. Rule-following is possible.
2. Following a rule consists in acting on one’s acceptance (or internalization) of
a rule.
3. Accepting a rule consists in an intentional state with general (prescriptive or
normative) content.
4. Acting under particular circumstances on an intentional state with general
(prescriptive or normative) content involves some sort of inference to what the
content calls for under the circumstances.
5. Inference involves following a rule.
If my argument is correct, then one of these claims has to go.21 The question is which
one.
Giving up (1) would give us rule-following skepticism. (2) seems to be the minimal
content of saying that someone is following a rule. (3) is the Intentional View. (4)
seems virtually platitudinous. (How could a general conditional content of the form
‘Whenever C, do A’ serve as your reason for doing A, unless you inferred that doing
A was called for from the belief that the circumstances are C?) (5) seems analytic of the
very idea of deductive inference (more on this below).
When we review our options, the only plausible non-skeptical option would appear
to be to give up (3), the Intentional View. To rescue the possibility of rule-following, it
seems, we must find a way of understanding the notion of accepting or internalizing a rule
that does not consist in our having some intentional state in which that rule’s require-
ments are explicitly represented. Wittgenstein can be read as having arrived at the same
conclusion.
The full passage from Investigations 219 reads as follows:
“All the steps are really already taken,” means: I no longer have any choice. The rule once
stamped with a particular meaning, traces the lines along which it is to be followed through the
whole of space. But if something of this sort really were the case, how would it help?
No; my description only made if it was understood symbolically.—I should have said: This is
how it strikes me.
When I obey a rule, I do not choose.
I obey the rule blindly.
The drift of the considerations I have been presenting seems to capture the intended
point behind this passage.
21
Notice that this argument is not only neutral on whether what is at issue are intentions as opposed to
other sorts of intentional state, but also on whether what is at issue are personal-level intentional states as
opposed to subpersonal content-bearing states. So long as you think that the acceptance of a rule consists in
some sort of intentional state with general content and that, as a result, inference will be required to act on
that state, there will be a problem—it doesn’t matter whether this is thought of as occurring at the personal or
the subpersonal level—more on this below.
Even without assuming Naturalism as an a priori constraint on the acceptability of a

solution to the rule-following problem, and without assuming that mental content
itself must be engendered by rule-following, it would seem that we have shown that, in
its most fundamental incarnation, rule acceptance cannot consist in the formation of a
propositional attitude in which the requirements of the rule are explicitly encoded.
Such a picture would be one according to which rule-following is always fully
sighted, always fully informed by some recognition of the requirements of the rule
being followed. And the point that Wittgenstein seems to be making is that, in its most
fundamental incarnation, not all rule-following can be like that—some rule-following
must simply be blind. The argument I have presented supports this conclusion.
8. Rule-Following without Intentionality: Dispositions

The question is how rule-following could be blind. How can someone commit himself
to a certain pattern in his thought or behavior, which can then rationalize what he
does, without this consisting in the formation of some appropriate kind of intentional
state?
The only option that seems to be available to us is the one that Kripke considers at
length, that we should somehow succeed in understanding what it is for someone to
accept a given rule just by invoking his or her dispositions to conform to the rule. If we
were able to do that, we could explain how it is possible to act on a rule without
inference because the relation between a disposition and its exercise is, of course, non-
inferential.
Now, Kripke, as we all know, gives an extended critique of the dispositional view.
However, that critique is not generally thought to be very effective; many writers have
rejected it.22 So perhaps there is hope for rule-following after all along the lines of a
dispositional understanding.
My own view, by contrast, is that Kripke’s critique is extremely effective, although
even I underestimated the force of what I now take to be its most telling strand. And so
I think that it can’t offer us any refuge, if we abandon the Intentional View.
The core idea of a dispositional account is that what it is for me to accept the rule
Modus Ponens is, roughly, for me to be disposed, for any p and q, upon believing both
p and ‘if p, then q,’ to conclude q.
Kripke pointed out that any such dispositional view runs into two problems. First, a
person’s dispositions to apply a rule are bound to contain performance errors; so one
can’t simply read off his dispositions which rule is at work.
Second, the rule Modus Ponens, for example, is defined over an infinite number of
pairs of propositions. For any p and q, p and ‘if p, then q,’ entail q. However, a person’s
dispositions are finite: it is not true that I have a disposition to answer q when asked
22
See, for example, Soames 1998, Horwich 1998.
what follows from any two propositions of the form p and ‘if p, then q,’ no matter how
large.
To get around these problems, the dispositionalist would have to specify ideal
conditions under which (a) I would not be capable of any performance errors and
(b) I would in fact be disposed to infer q from any two propositions of the form p and ‘if
p, then q.’
But it is very hard to see that there are conditions under which I would be
metaphysically incapable of performance errors.
And whatever one thinks about that, it’s certainly very hard to see that there are ideal
conditions under which I would in fact be disposed to infer q from any two proposi-
tions of the form p and ‘if p, then q’ no matter how long or complex. As Kripke says,
for most propositions, it would be more correct to say that my disposition is to die
before I am even able to grasp which propositions are at issue.
Along with many other commentators, I used to underestimate the force of this
point. The following response to it seemed compelling. A glass can have infinitary
dispositions. Thus, a glass can be disposed to break when struck here, or when struck
there; when struck at this angle or at that one, when struck at this location, or at that
one. And so on. But if a mere glass can have infinitary dispositions, why couldn’t a
human being?23
There is an important difference between the two cases. In the case of the glass, the
existence of the infinite number of inputs—the different places, angles, and loca-
tions—just follows from the nature of the glass qua physical object. No idealization is
required.
But a capacity to grasp infinitely long propositions—the inputs in the rule-
following case—does not follow from our nature as thinking beings, and certainly
not from our nature as physical beings. In fact, it seems pretty clear that we do not
have that capacity and could not have it, no matter how liberally we apply the notion
of idealization.
These, then, are Kripke’s central arguments against a dispositional account of rule-
following, and although it would take much more elaboration to completely nail these
arguments down, I believe that such an elaboration can be given.
But both before and after he gives those arguments, Kripke several times suggests
that the whole exercise is pointless, that it should simply be obvious that the dispositional
account is no good. Thus, he says:
To a good extent this [dispositional] reply ought to appear to be misdirected, off target. For the
skeptic created an air of puzzlement as to my justification for responding ‘125’ rather than ‘5’ to
the addition problem . . . he thinks my response is no better than a stab in the dark. Does the
suggested reply advance matters? How does it justify my choice of ‘125’? What it says is “ ‘125’ is
the response you are disposed to give . . . ” Well and good, I know that ‘125’ is the response I am
23
See the discussion in Boghossian 1989.
disposed to give (I am actually giving it!) . . . How does any of this indicate that . . . ‘125’ was an
answer justified in terms of instructions I gave myself, rather than a mere jack-in-the-box
unjustified response?
This passage can seem puzzling and unconvincing when it is read, as Kripke seems to
have intended it, as directed against dispositional accounts of mental content. After all,
one of the most influential views of mental content nowadays is that expressions of
mentalese get their meaning by virtue of their having a certain causal role in reasoning.
Could it really be that this view is so obviously false that it’s not worth discussing, as
Kripke suggests? And is it really plausible that the facts by virtue of which my mentalese
symbol ‘+’ means what it does have to justify me when I use it one way rather than
another?
But if we see the passage as directed not at dispositional accounts of mental content
but rather at dispositional accounts of personal-level rule-following, and if we substi-
tute “rationalize” for “justify,” then its points seem correct. It should be puzzling that
anyone was inclined to take a dispositional account of rule-following seriously. We can
see why in two stages (this is a different argument than the one Kripke gives).
First, and as I have been emphasizing, if I am following the rule Modus Ponens, then
my following that rule explains and rationalizes my concluding q from p and ‘if p, then
q,’ ( just as it would be true that, if I were following the rule of Affirming the
Consequent, then my following that rule would explain and rationalize my inferring
q from p and ‘if q, then p’).
Second, if I am following the rule Modus Ponens, then not only is my actually
inferring q explained and rationalized by my accepting that rule, but so, too, is my
being disposed to infer q. Suppose I consider a particular MP inference, find myself
disposed to draw the conclusion, but, for whatever reason, fail to do so. That disposi-
tion to draw the conclusion would itself be explained and rationalized by my accep-
tance of the MP rule.
However, it is, I take it, independently plausible that something can neither be
explained by itself, nor rationalized by itself. So, following rule R and being disposed to
conform to it cannot be the same thing.
Here we see, once again, how Kripke’s Meaning Assumption gets in the way of his
argument: a good point about rule-following comes out looking false when it is
extended to mental content.
9. Is Going Subpersonal the Solution?

It might be thought a crucial assumption of the preceding argument is that all rule-
following is personal-level rule-following and that its moral, therefore, should be that
at least some rule-following is subpersonal. In particular, would our problem about
rule-following disappear if we construed the inferences that mediate between inten-
tions and actions in some subpersonal way?
This suggestion resonates with a number of discussions of the Intentional View that
may be found in the literature. These discussions tend to accuse the Intentional View of
being ‘overly intellectualized’ and recommend substituting a subpersonal notion in its
place.24 It isn’t very often made clear exactly what that is supposed to amount to. The
preceding discussion should help us see that this is not a particularly useful suggestion.
In the present context, going subpersonal presumably means identifying rule
acceptance or internalization not with some person-level state, such as an intention,
but with some subpersonal state. Such a state will either be an intentional state or some
non-intentional state.
Let us say that it is some intentional state in which the rule’s requirements are
explicitly represented. Then, once again, it would appear that some inference (now,
subpersonal) will be required to figure out what the rule calls for under the circum-
stances. And at this point the regress problem will recur. (That is what I meant by
saying earlier that the structure of the regress problem seems to be indifferent as to
whether the states of rule acceptance are personal or subpersonal.)
On the other hand, we could try identifying rule internalization with some non-
intentional state. Indeed, even if the state of rule internalization is initially identified
with a subpersonal intentional state, it will ultimately, I take it, have to be identified
with some sort of non-intentional state.
But then what we would have on our hands would be some version or other of a
dispositional view (with the dispositions now understood subpersonally). And although
we would no longer face the rationalization problem—because subpersonal mechan-
isms are not called upon to rationalize their outputs—we would still face the enormous
problems posed by the error and finitude objections.
In consequence, I don’t believe that going subpersonal offers any sort of panacea for
our problems.
10. Conclusion
In my 1989 paper, “The Rule-Following Considerations,” I was concerned to expli-
cate and assess Kripke’s arguments for his rule-following skepticism. However, since at
that time something like Wright’s Intention View about rule-following struck me as
correct, I thought that the interesting issues had to center on the notion of mental
content rather than on rule-following in particular. Rule-following, like any other
intentional process, I thought, came along for the ride.
Viewed in that light, Kripke’s arguments were most effectively seen as arguments
against various attempts to naturalize intentional content. And I believed then, and
continue to believe now, that his arguments are very effective against that particular
target.
24
See, for example, Cruz and Pollock 1999, chapter 5.
The upshot, however, was that the so-called “rule-following considerations” had
very little to do with rule-following per se and a great deal to do with meaning and
content—as I noted at the time.
I now believe that this was a mistake, induced by the illusory plausibility of the
Intention View. In its most fundamental incarnation, rule acceptance cannot consist in
an intentional state. For if it did, rule-following would have to be inference; and we
know that, whatever else it may be, it cannot be that.
In its most fundamental incarnation, rule-following is something that must be done
blindly, without the benefit of some intentional encoding of the rule’s requirements.
The question is whether that is so much as possible.
The only way to make sense of it, it would seem, is if rule acceptance could be
understood dispositionally. But there seem to be powerful objections to any such
understanding.
As a result, it is hard to explain how rule-following is so much as possible, and this
difficulty arises even without our having to assume (in the way that Kripke effectively
does) that intentional states need to be given a naturalistic reduction.
What are we to do?
Perhaps we should embrace rule-skepticism, denying that our reasoning is under the
influence of general rules?
The trouble is that this seems not only false about reasoning in general, but also
unintelligible in connection with deductive inference. It is of the essence of deductive
inference that the reasons I have for moving from certain premises to certain conclu-
sions are general ones.
So what we are contemplating, when we contemplate giving up on the Rulish
picture of deductive inference, is not so much giving up on a Rulish construal of
deductive inference as giving up on deductive inference itself.
But that is surely not stable a resting point—didn’t we arrive at the present
conclusion through the application of several instances of deductive inference?
Hence we have what I have somewhat grandly called an “antinomy of pure reason”:
we both must—and cannot—make sense of the notion of someone’s following a rule.
The only non-skeptical option that seems open to us is to try taking the notion of
following—or applying—a rule as primitive, effectively a rejection of proposition (4)
above.
Notice that this goes well beyond the sort of anti-reductionist response to Kripke’s
arguments that I was already inclined to favor—an anti-reductionism about mental
content.
It would involve a primitivism about rule-following or rule-application itself:
we would have to take as primitive a general (often conditional) content serving as the reason
for which one believes something, without this being mediated by inference of any kind.
It is not obvious that we can make sense of this, but the matter clearly deserves greater
consideration.
Note
This essay draws heavily on my 2008. A very early version of some of its arguments appeared in
Boghossian 2005, as part of a symposium on Philip Pettit’s 2002 book. I have benefited greatly
from feedback over the intervening years from various audiences—at various seminars at NYU in
2001, 2003, and 2006, the Graduate Conference at the University of Warwick, the Workshop
on Epistemic Normativity at Chapel Hill, UCLA, Stony Brook, Princeton, and the Transcen-
dental Philosophy Network Workshop in London, and the Workshop on Wittgenstein in
Tokyo, to name just those that come to mind. I am also grateful to, Sinan Dogramaci, Paul
Horwich, Matthew Kotzen, Christopher Peacocke, Josh Schechter, and Stephen Schiffer for
valuable comments on earlier drafts.
Bibliography
Boghossian, P. A. 1989 “The Rule-Following Considerations,” Mind 98/392, pp. 507–49.
Boghossian, P. A. 2005 “Meaning, Rules and Intention,” Philosophical Studies 124/2, pp. 185–97.
Boghossian, P. A. 2008 “Epistemic Rules,” Journal of Philosophy CV, 9, pp. 472–500.
Carroll, L. 1895 “What the Tortoise Said to Achilles,” Mind 4/14, pp. 278–80.
Cruz, J. and Pollock, J. 1999 Contemporary Theories of Knowledge, Oxford, Rowman and Little-
field.
Fodor, J. 1990 A Theory of Content and Other Essays, Cambridge, Mass., MIT Press.
Horwich, P. 1998 Meaning, Oxford, Oxford University Press.
Kripke, S. 1982 Wittgenstein on Rules and Private Language, Cambridge, Mass., Harvard University
Press.
Pettit, P. 2002 Rules, Reasons and Norms, Oxford, Oxford University Press.
Quine, W. V. O. 1936 “Truth by Convention,” in W. V. O. Quine 1976 The Ways of Paradox
and Other Essays, Cambridge, Mass., Harvard University Press, pp. 77–106.
Soames, S. 1998 “Skepticism About Meaning: Indeterminacy, Normativity and the Rule
Following Paradox,” Canadian Journal of Philosophy 23, pp. 211–50.
Wittgenstein, L. 1953 Philosophical Investigations, ed. G. E. M. Anscombe and R. Rhees, trans.
G. E. M. Anscombe, Oxford, Blackwell.
Wright, C. 1980 Wittgenstein on the Foundations of Mathematics, London, Duckworth.
Wright, C. 2001 Rails to Infinity: Essays on Themes from Wittgenstein’s Philosophical Investigations,
Wright, C. 2007 “Rule-Following Without Reasons: Wittgenstein’s Quietism and the Consti-
tutive Question,” in J. Preston (ed.) Wittgenstein and Reason, Ratio 20/4, pp. 481–502.
2
Understanding and
Rule-Following1
Christopher Peacocke
Crispin Wright and I were colleagues for two years, back in All Souls thirty years ago
now, and then and ever since I have learned from his energetic contributions to
discussions. A paper or chapter I was working on was never the same after it had
been exposed to his comments. It was not only on my efforts that he had these effects.
In rereading Crispin’s early writing, I was struck by how much philosophical work in
Oxford in the 1970s grew in part from our reaction to Crispin’s reading of Wittgen-
stein, and its highly sceptical stance on issues of concern then and now. This applies to
such varied matters as the justification of logic, the nature of norms and concept-
possession, the proper characterization of what is at stake between realism and
verificationism, the metaphysics of modality, and the nature of structured linguistic
understanding. This same effect on the work of others was then repeated on a larger
scale in the influence throughout the Anglophone world of the formulations in his
book Truth and Objectivity. It was not necessary to agree with Crispin’s views to
appreciate their interest, and to acknowledge the mark they have left in sharpening
and deepening our understanding of a wide range of philosophical issues. I wish Crispin
(and us) many more years of his contributions at the centre of our subject.
A proper account of rule-following has ramifications throughout philosophy. Issues
of objectivity, justification, meaning, truth, and large tracts of metaphysics and episte-
mology look entirely different according as one takes one stand or another on basic
issues about rule-following. It is to Wittgenstein that we owe the full identification of
the importance of rule-following, and an appreciation of its significance throughout
our subject, whatever one may think of his positive account. It is to Wright that we
owe the most extended elaboration of one understanding of Wittgenstein’s position on
1
I thank Paul Boghossian, Allan Gibbard, and an anonymous referee for valuable comments on an earlier
version of this essay. Readers of the present essay and Boghossian’s contribution to this volume will find that
we agree, though for different reasons, in rejecting accounts of meaning based on intention (see }3 below),
and more generally on which avenues are closed off in making progress on these issues.
50 C H R I S TO P H E R P E AC O C K E
these issues, and an exposition of its significance. My aim in this essay is to propose a
positive account of rule-following very different from the one Wright attributes to
Wittgenstein. This positive account draws on resources that preserve our pre-theoretical
conceptions of objectivity, truth, and meaning.
I begin by discussing some features of basic cases of concept-application that are
highly intuitive, but whose full consequences are not always appreciated (}1). These
features are regarded as illusory under the conception of rule-following Wright
attributes to Wittgenstein. A rival account of rule-following and concept-possession
respects these intuitive features (}2). This rival account provides the resources for
replying to Wright’s arguments that these features are illusory. It makes possible an
account of the first-person phenomenon of knowing what one means (}3). It allows us
to steer a middle course between the two very unattractive positions of extreme
Platonism on the one hand and linguistic idealism on the other (}4). It leaves room
for substantive philosophy of particular concepts and particular types of concepts (}5).
The account also raises questions of its own. In a final section I consider some
objections, and the prospects for elaborating the notions on which the account relies.
1. The Intuitive Conception

We have an intuitive, pre-theoretical, conception of cases of concept-application
meeting four conditions: a thinker applies some concept he possesses by making a
judgment whose content contains that concept; the thinker’s judgment is rational; the
judgment is made on the basis of his grasp of the concept; and it is some relation of the
judgment, and the circumstances as they strike him, to his grasp of the concept that
makes his application rational. We have a similar intuitive conception of a kind of case
for linguistic expressions that are understood. A thinker applies some word by making
an assertion containing it; the assertion is rationally made; the assertion is made on the
basis of the thinker’s understanding of the word; and it is some relation of the assertion,
and the circumstances as they strike him, to his understanding that makes the assertion
rational. It is very plausible that for any concept a thinker possesses, there are, or could
easily be, at least some such cases of understanding-based rational application of a
concept. This seems to be a requirement for possessing the concept. A corresponding
point applies at the linguistic level for the understanding of expressions. Understanding-
based application is a phenomenon which has rightly been central in discussions of rule-
following, and it requires a philosophical elucidation for that and many other reasons.
Understanding-based rational application is a complex thing which unfolds into
many properties. I take the rationality of the application of a concept first.
One way in which a concept can be rationally employed in judgment is the case in
which a judgment containing the concept is properly reached by valid inference from
premises already accepted. It is far from the only way. Even when a concept is applied
not as a result of making an inference from other propositions, the application can still be
rational. A thinker’s application of the concept red can be made rationally as a response to
U N D E R S TA N D I N G A N D RU L E - F O L L OW I N G 51
an experience which represents a surface has having a shade that is in fact a shade of red.
A thinker’s self-ascription of a movement or a bodily posture may be made rationally
because the thinker has an action-awareness of so moving or positioning himself.
A thinker may judge that 57+5 is 62 because counting up 5 from 57 takes him to 62.
In none of these cases does it seem right to describe the thinker’s application of the
concept as ‘blind’, as Wittgenstein does at some points.2
The judgments in these cases are not leaps in the dark, or mere yieldings to an
inclination to judge that has no further basis. There are states of the thinker that make
the judgment reasonable. Even in the most subjective of cases, such as a thinker’s
judgment of whether a particular food he is eating has a pleasing taste, the affect
produced by the taste in the thinker gives a reason, his reason, for making the judgment
that it is pleasing, or not, as the case may be.
As these cases illustrate, a state or event can be a thinker’s reason for making a
judgment without the thinker having a reflective grasp of a general notion of reasons.
Arguably in some of these examples the thinker does not even have to have the
capacity to think about the states that make his judgment rational.
At this stage in the discussion, I just note the apparent existence of such rational
judgments. What is involved in such rationality, what relation has to obtain between a
state and a judgment for the former to make the latter rational—all this is a matter for
further theory.
Equally intuitive is the description of these cases as ones in which the thinker makes
the judgment he does because he has enough of a grasp of the concept to appreciate
that the judgment is correct in the circumstances in which he is making it. The thinker
makes the judgment he does as a result of his grasp or understanding on the one hand,
and the circumstances and states he finds himself in on the other. Such dispositions to
judge as the thinker has are an upshot of this understanding and his apparent circum-
stances. His understanding is rationally and explanatorily prior to any such dispositions.
The thinker’s understanding determines a correctness-condition for his judgment
which, taken together with his apparent circumstances, contribute to the explanation
of (some of ) his dispositions to judge the content, when the thinker is functioning well.
The thinker’s judgments are rationally answerable to the correctness-condition that
is fixed by his understanding. The thinker’s understanding, together with his circum-
stances, states, and possibly other background information, make a particular judgment
rational for the thinker. It also seems to be part of our intuitive conception that the
thinker makes the judgment because it is thus made rational. This rational dependence
on the thinker’s understanding seems to be a respect in which judgments so made are
not made ‘blindly’.
2
“I obey the rule blindly”, Philosophical Investigations }219; the German ‘blind’ is italicized in the German
text also. The immediately preceding sentence, in Anscombe’s translation, is “When I obey, I do not
choose.” If rationality requires one judgment rather than another, you will indeed have no choice if your
judgment is to be rational. But this does not mean you do not see what you are doing.
Again, this is all at the level of description, not the level of philosophical explanation.
It is a hard task, philosophically, both to specify the nature of such understanding, and
to say how it can have this rational explanatory role. But whatever the correct theory of
these matters, an account that does not acknowledge the existence of rational, under-
standing-based judgments will count as revisionary of basic elements of our self-
conception as rational thinkers.
Wright’s Wittgenstein holds a position which seemingly contradicts this intuitive
conception. First, of course, it seems to be part of the position Wright attributes to
Wittgenstein that for the individual thinker, application of a concept is blind in just the
way we have been disputing. For Wright’s Wittgenstein, there is no such thing as
recognition of what judgments or assertions understanding requires in given circum-
stances. “The real objection is that nothing can be done to defend the description that
the trainee recognizes how, if his conjectural understanding is right, we will treat the
new case, against the more austere account that he simply expects this application to be
acceptable to us.”3 An expectation is something falling short of understanding-based
rational judgment. Second, it is not at all apparent under the intuitive view that the
distinction between the case in which a thinker is employing one concept rather than
another, or is employing a genuine concept as opposed to being under the illusion that
he is doing so, has to be elucidated by reference to the agreement of judgments of
members of an actual or possible community of thinkers. That the community has to
be brought in to elucidate such distinctions is precisely the positive claim of Wright’s
Wittgenstein. “None of us unilaterally can make sense of the idea of correct employ-
ment of language save by reference to the authority of securable communal assent on
the matter; and for the community itself there is no authority, so no standard to meet.”4
Perhaps the intuitive conception meeting the four conditions is just some kind of
illusion, and the only way at all of elaborating versions of distinctions of the sort it
endorses is by making use of the materials of Wright’s Wittgenstein—in short, that
only some form of sceptical solution to the rule-following considerations is possible, in
the familiar post-Kripkean terminology.5 There is certainly an onus at this point in the
debate on the person who thinks that the intuitive features of understanding-based
rationality are not to be given some sceptical treatment. His task is to give some positive
philosophical account that acknowledges those features and respects their explanatory
role, but does so in a non-sceptical fashion. That is what I will now attempt.
2. A Positive Account of Understanding-Based

Application
Judging is a mental action, something that can be done for reasons. It is internal to the
nature of judgment that only true contents should be judged. (Arguably it is internal to
3 4 5
Wright 1980, p. 217. Ibid., p. 220. Kripke 1982.
judgment that only known contents should be judged, but the issue of truth is quite
enough for us to bite off at this stage.) Proper judgment aims to be on the true side of
the true/false distinction for the content that is judged. The classification of a content as
true or as false depends on the way the world is. The condition for truth of a complete
content depends in turn on the references of its several conceptual components, and
the way in which they are combined.
There is equally a form of the correct/incorrect distinction for concepts, the con-
stituents of complete conceptual contents that are judged. We can draw a distinction
between those entities (of the appropriate category) which meet the condition for
being the semantic value of the concept, and those which do not. This is a distinction
that is necessarily of normative significance. Semantic values contribute to the deter-
mination of truth or falsity of the complex complete contents of which they are
constituents. Meeting the goal of truth that is internal to judgment unfolds into the
goal of judging only what meets a condition for correctness. Since that correctness-
condition is in turn determined by the references of the conceptual constituents of the
complete content in question, the distinction between those entities that meet the
condition for being a concept’s semantic value and those that do not cannot fail to be of
normative significance.6
This means that a positive account of concept-application that respects the intuitive
characterization has to address at least these two major questions:
(1) How is the understanding, which a thinker draws upon in making rational
judgments, related to the reference-determined truth-condition of the content
judged? This is a question about the nature of the understanding.
(2) How does this understanding, as elucidated in answering Question (1), enable
the thinker to be rationally sensitive to what determines the truth of the judged
content? This is a question about the application of understanding.
The answer to the question about the nature of the understanding, Question (1),
that I offer is that grasping a given concept consists in having tacit knowledge of its
fundamental reference rule. The fundamental reference rule for a concept states the
fundamental condition for something to be the concept’s semantic value. Partial
understanding of the concept consists in having tacit partial knowledge of its funda-
mental reference rule.
In earlier work, I argued that the fundamental reference rule for a concept con-
tributes to the explanation of norms or reasons distinctive of that concept.7 This claim
about the relations between reference and reasons is an indispensable lemma in the next
6
It is the role of a notion of correctness-condition that unifies concept-possession and meaning with the
other cases classified as rule-following in discussion of these matters. Understanding is not a matter of
knowing a rule that one should do such-and-such, or is permitted to do such-and-such, in specified
circumstances. But such rules, and concept-possession and understanding, do all involve correctness-
conditions of one sort or another.
7
Peacocke 2008, Chapter 2.
part of the account I offer here. The answer I propose to Question (2), about the
application of understanding, is that by drawing on his tacit knowledge of the
fundamental reference rule for the concepts involved, possibly in combination with
further information, a thinker is able to come to appreciate that certain states or
propositions give reasons for thinking that particular contents containing the concepts
in question are true. In short, the theory involves these two claims:
Fundamental reference rules for concepts contribute to the explanation of truth-
supporting reasons or norms involving those concepts.
A thinker’s tacit knowledge of those fundamental reference rules thereby enables
him to appreciate that certain states or propositions are reasons in favor of contents
containing those concepts.
This appreciation is appreciation of normative facts as such. Tacit knowledge in general
is not a normative notion, but tacit knowledge of fundamental reference rules is tacit
knowledge of something that has normative consequences, for correctness-conditions.
How does this theory operate in particular cases? Here are some plausible examples
of fundamental reference rules:
The fundamental reference rule for the observational concept round is that some-
thing falls under it (is mapped to the True by it, as Frege would say) iff it has the same
shape as things are represented as having when they are experienced as round.8
The fundamental reference rule for the concept here is that for any thinker x at any time
t and for any location l, here as employed by x at t refers to l iff l is the location of x at t.
The fundamental reference rule for the perceptual demonstrative that F, where the
apparently perceived object is given in way W in perception, is that for any thinker x at
any time t, and for any object x, that F (as given in way W) refers to x iff x is the F
perceived by x at t in way W.
The fundamental reference rule for the concept plus is that it denotes the unique
function f such that for all numbers n, f(n,0) = n and f(n,m0 ) = (f(n,m))0 , where0 is the
successor operation.
We can trace out the consequences of the answers I am proposing to Questions
(1) and (2) about the nature and application of understanding for a couple of particular
sample concepts.
Since it has been so salient in the discussion since Kripke’s treatment, I take the
concept plus first. Tacit knowledge of the above fundamental reference rule is an
essential element in the explanation of the thinker’s appreciation that certain calcula-
tions give good reasons for judging that 57 plus 68 equals 125. These calculations
include in particular certain kinds of counting up in working out the sum of two
8
The experience will have a non-conceptual representational content on my own views: see for instance
Peacocke 2001. But this fundamental reference rule could also be used in a more thoroughly non-reductive
account by someone who holds McDowell’s view that the representational content of perception is always
fully conceptual: see McDowell 1994, Lectures I–III. What is important for the fundamental reference rule
for observational concepts is that the content of observational judgments be individuated by their relations to
perceptual experiences that are not judgments.
numbers. They also include certain kinds of ‘carrying’ operation used with Arabic
numerals in the computation procedures for Arabic representations involving multiple
digits. These carrying procedures are ultimately ratified as correct by their derivability
from the above fundamental reference rule. On the present account, a thinker’s
grasping plus rather than some other concept consists in his having tacit knowledge
of the above fundamental reference rule.
Nothing here implies that a thinker’s having tacit knowledge of this reference rule
for plus and his knowledge of its normative consequences can be reduced to a body of
dispositions to judge. Having an appreciation of the norms does not at all mean that
you succeed in judging in accordance with those norms. A thinker’s having tacit
knowledge of the fundamental reference rule for plus is entirely consistent with the
thinker’s making arithmetical mistakes in his judgments containing that concept.
How are mistakes possible if thinker already has tacit knowledge which, when
combined with other information he possess, entails specifications of what are good
reasons for judging a given content? Why does the thinker go against those entailed
good reasons? A wide variety of explanations is possible. Here are three varieties.
First, we should recall that mistakes can be made in other areas of which it would be
widely agreed that tacit knowledge should be invoked to explain our judgments and
knowledge. Garden-path sentences such as “The horse raced past the barn fell” are an
obvious example. A competent speaker-hearer of English, with tacit knowledge of the
syntactic and semantic rules of English, may hear an utterance of this sentence, and may
still perceive it as ungrammatical. The relevant knowledge is not operative, not causally
involved, in the production of his perceptions, and the thinker relies on some other only
roughly correct heuristic that gives the wrong answer in this case. Tacit knowledge of a
particular fundamental reference rule may equally not be operative in a particular case,
and other information or misinformation may influence the thinker’s judgment.
A second possible intelligible of cause of error, consistent with having the relevant
tacit knowledge, is reliance on an incorrect apparent memory (e.g. of a line of the
multiplication tables). In this example, if the thinker were to test his answer to the
arithmetical question against what is entailed by the fundamental reference rule that he
tacitly knows, he would realize his error. In fact however the thinker does not engage
in such testing, and so gives a wrong answer.
In a third sort of case, the relevant tacit knowledge is causally involved, but the
thinker makes inferential or computational mistakes in moving from this knowledge,
owing to fatigue, distraction, confusion, or myriad other human frailties.
To give a non-arithmetical illustration of how fundamental reference rules can
contribute to the explanation of reasons, I take the above treatment of the observa-
tional concept round. A thinker’s tacit knowledge of this rule explains his taking a
perceptual experience of something given in perception as round when he experiences
it as round, in circumstances in which he is entitled to take experience at face value.
That is a causal explanation. The content of the tacitly known reference rule itself also
contributes to an explanation, non-causal, of why the experience of an object as round
56 C H R IS TO P H ER P EAC O C K E
is a good reason for the judgment that the perceptually given object is round, when the
thinker is entitled to take experience at face value. The causal explanation is not a
matter of conscious inference or thought. There is a subpersonal transition from the
content of the tacit knowledge to the appreciation that a certain kind of perceptual
experience makes a perceptual application of the concept round rational in the given
circumstances.9
Tacit knowledge of this particular fundamental reference rule can also explain the
thinker’s taking something as evidence that an unperceived object is round, in the
presence of information that this evidence is evidence that the unperceived object has
the same shape as things perceived to be round are represented as having. It is an
empirical matter which evidence has this property. What is not empirical is that it has to
be evidence that the object has the same shape as things are represented to be in
perceptual experiences of things as round, if it is to be evidence that the object falls
under the observational concept round. This application to the non-perceptual case
shows, incidentally, that it would be a mistake to think that tacit knowledge of the
fundamental reference rule for observational concepts comes to no more than the
practical ability to respond rationally to perceptual states in making one’s judgments.
The ‘same shape as . . . ’ part of the fundamental reference rule generates norms of
correctness for being round for objects arbitrarily far away, objects too small to see, and
many other cases where judgments of shape must be made partly on theoretical and
abductive grounds.
In both the numerical and the observational illustrations, normative facts are ex-
plained in part by the fundamental reference rule. It is correspondingly an appreciation
of reasons that is explained by tacit knowledge of the fundamental reference rule. It is a
further, and apparently orthogonal, issue whether the normative can be explained in or
reduced to non-normative terms. This account of understanding and its relations to the
appreciation of reasons does not require any such reduction (and it may not in itself
preclude it).
Even in these simplest observational examples, we can see how distant this account
of rational observational judgment is from a blind disposition to judge. The thinker’s
perceptual experience makes rational the observational judgment, and does so on the
basis of the thinker’s understanding as characterized by his tacit knowledge of the
fundamental reference rule for the observational concept round.
More generally, the pair of answers I have offered to the questions (1) and (2) suggest
a corresponding elucidation of what it is for a thinker to judge something on the basis
of his understanding of the concepts involved. It is for the thinker to make the
9
A note on my use of the words ‘rational transition’ as applied to the transition from a state or event to
judgment of a particular content: I use this for all cases in which it is rational for the thinker to move from the
state or event to the judgment, even if it is not rational for the thinker to be in the state, and even if the event
is not a mental action, so that (as in the case of visual experience), the question of rationality does not even
arise.
judgment rationally on the basis of a reason-giving state that is recognized as such in

part by suitable derivation from the fundamental reference rules for the concepts in
question, possibly in combination with other information. As before, the derivation
will be subpersonal. This judging on the basis of one’s understanding is a substantial
condition.
Wittgenstein, after elaborating what he calls a ‘paradox’ about rule-following, wrote
that “what this shews is that there is a way of grasping a rule which is not an
interpretation, but which is exhibited in what we call ‘obeying the rule’ and ‘going
against it’ in actual cases” (Investigations }201). The account of understanding-based
application I have offered is an account of concept-possession that does not involve
“interpretation” in the sense Wittgenstein intends. Explanations of understanding-
based application of a concept that appeal to computation from a tacitly known
fundamental reference rule do not involve applying an interpretation to some other-
wise uninterpreted string, sequence, or image that can be interpreted in arbitrarily
many different ways. That is how Wittgenstein’s paradox is answered in the treatment
I am offering.
It is essential to the adequacy of this answer that subpersonal computation is regarded
as content-involving, and the facts explained by it also be regarded as content-
involving. If subpersonal computation were conceived of purely syntactically, then
computation from a subpersonal state realizing tacit knowledge of a fundamental
reference rule would indeed simply be an operation on an uninterpreted string of
subpersonal symbols or representations. Wittgenstein’s ‘paradox’ would apply again,
and he would be entirely correct to insist that such purely syntactic subpersonal
operations could never explain the content-involving facts of concept application
exhibited in actual cases. When computation is conceived of as content-involving, as
explaining content-involving states and events from other content-involving states and
events, the arguments for the paradox do not get a grip.
The notion of what would be accepted by a thinker in ideal circumstances
sometimes enters discussions of concept-possession and rule-following. The idea is
that the identity of a concept employed by a thinker depends constitutively on what
judgments involving the concept the thinker would be willing to make in certain
ideal circumstances. What I have already said implies that this approach can work
only in the presence of very substantial restrictions. What are the ideal circumstances
for judging of something too small to see, or too large to see, that it is round?
(The concept round certainly applies to some such objects.) A plausible specification
of the ideal circumstances for this case would that they are circumstances which
provide evidence that the tiny, or gigantic, object is of the same shape as things that
are observably round. But this specification is evidently reliant on the fundamental
reference rule for the concept round. It relies on some prior substantial individuation
of that concept. On the other hand, if we use specifications that rely on empirical
facts about what would be evidence of shape in the very small or very large, we will
be citing conditions which are not essential to the individuation of the concept round.
What makes the concept the concept it is would still hold even if those empirical facts
were different.
Nevertheless, there is a different application of the notion of ideal conditions that
does get a grip in the context of the answers we have given to Questions (1) and (2),
and the account of understanding-based application that goes along with them. We can
consider the case of a thinker who always judges in accordance with the norms
derivable, in combination with other information, from the fundamental reference
rules for the concepts in his judgments. If this thinker’s background presuppositions
(such as that he is perceiving properly) are correct, and any background beliefs on
which he is relying are true, then the judgments reached by applying these norms will
also be true. They will not go beyond what, in the circumstances, is made rational by
the fundamental rules of reference themselves. This will be, in one legitimate sense of
the adjective, an ideal thinker. But the notion of an ideal case is applicable here only
because there are substantive accounts being offered of what it is to possess a given
concept, and of what it is to judge in accordance with one’s understanding of that
concept. The notion of an ideal case here is constitutively and explanatorily posterior
to the substantive explication of understanding and meaning. It is not being used here
to give that substantive explication.
Does this leave any work for a notion of ideal circumstance to perform in the
theory of concepts and meaning? Here we need to distinguish at least two notions of
ideal circumstances (there may be further notions too). Under the first notion, ideal
circumstances are those in which some parameter in a specified theory is given a
certain value. Examples are the ideal plane in classical mechanics, in which friction is
set to zero, or the ideal case of agents with perfect information about prices, costs,
and about each other, in a neo-classical model of an economy. This notion of an
ideal case applies only when there is a background theory in which to set parameters
to a certain value. Any psychology of a thinker would surely have to take into
account what concepts he possesses and is employing. Such a theory would already
take for granted that the thinker has certain concepts, and would need to say
something about what is to be expected of a thinker with those concepts. It
would already involve a theory of concept-possession, rather than supplying the
materials for constructing that theory. But there is a different notion of ideal
circumstance, as Allan Gibbard pointed out to me. When we speak of the ideal
circumstances in which Hooke’s Law holds, we are not setting a parameter in a
theory to a certain value: we are instead simply speaking of the cases in which the
law does hold. Some suitably limited version of ideal circumstances in this second
sense is by no means excluded by what I have said. In fact tacit knowledge of a
fundamental reference rule is arguably itself a state individuated by what would be
proper computation to and from such a state. But incorrect computations also occur,
and the circumstances in which they do are plausibly non-ideal in this second sense.
Ideality in this second sense seems to be a legitimate resource in the theory of
concepts and meaning.
3. A Comparison with Wright’s Wittgenstein

I now compare this account of understanding and concept-application with the
conception of rule-following attributed to Wittgenstein by Wright in Wittgenstein on
the Foundations of Mathematics. The comparison can be developed by taking a series of
issues at stake in discussions of rule-following, and contrasting how the two accounts
treat them.
(i) Investigation-Independence
We think intuitively that there is an investigation-independent fact of the matter as to
whether one of a thinker’s concepts applies to an uninvestigated case. Intuitively, there
are the facts of the matter, and there is the verdict on a conceptual content required by
those facts, given the identity of the concepts composing the content. For Wright’s
Wittgenstein, it is an illusion that there is such a thing as the verdict required by a
thinker’s individual grasp of a concept, when considered independently of any actual
or possible community of thinkers. It is one of the virtues of Wright’s discussion
that it makes so clear that the relatively pre-theoretical conception of investigation-
independence has no application on the conception of rule-following Wright ascribes
to Wittgenstein.
By contrast, on the view I have proposed, the correctness-condition for application
to a new case is determined by the content of the fundamental reference rules that the
thinker tacitly knows. Whether a given concept applies to some uninvestigated case
depends only on the two factors of how the world is, and whether the uninvestigated
object meets the correctness-condition determined by those fundamental reference
rules. It is not even a matter of how the thinker would judge were he to investigate
(and were the facts to remain the same). As we noted, the thinker may be subject to all
kinds of psychological factors that lead him not judge in accordance with what his
understanding and the facts require. To say it once again, the nature of his understand-
ing and what it requires is not to be explained in terms of dispositions to judge. The
correctness-condition required by the thinker’s understanding is given by the content
of his tacitly known reference rules. It is not constitutively dependent on the actual or
possible reactions of other thinkers in a community.
(ii) Patterns and Rails to Infinity

A frequent argument of Wright’s Wittgenstein is that we can give no sense to the idea
of an individual being faithful to the pattern of use he has so far established. If “pattern
of use” or “pattern of application” picks out just extensionally what the thinker has so
far judged to fall under a concept, no one should disagree. Everyone should accept that
a given proper segment of the extension of a concept is equally the extension of
arbitrarily many distinct concepts. Many different rails to infinity will overlap with a
given partial segment of an extension. Faithfulness to the thinker’s concept has to do
not with which things have been previously judged to fall under the concept (after all,
60 C H R IS TO P H E R P E AC O C K E
he may have made mistakes about that too). It has rather to do with judging in
accordance with the norms of correctness explained by the fundamental reference
rule that the thinker tacitly knows.
(iii) Individual Understanding and its Objectivity

Is it true that we cannot explain the distinction between its merely seeming to a thinker
that he means something by an expression, and his really meaning something by it,
unless we mention the possibility of communal assent in the case in which he really
does mean something by it? Prima facie, the distinction between someone who is
under an illusion that he means something by an expression, and someone who by
contrast has tacit knowledge of a specific fundamental reference rule for that expres-
sion, is in good standing without mentioning communal assent at all. If others have
tacit knowledge of the same fundamental reference rule for the expression, they will be
likely to agree in judgments with our individual thinker, other things equal. But that is
consequential upon the nature of understanding, rather than being constitutive of it.
It may be complained that we have not given sufficient substance to the notion of a
thinker’s having tacit knowledge of a given concept’s, or expression’s, fundamental
reference rule. When we look at the actual ascriptions of understanding that we make,
it may be objected, we actually rely substantially on agreement within a community of
thinkers in our ascriptions of understanding. There are two issues here we should take
separately. There does not seem to me to be anything objectionably non-empirical or
indeterminate in the account of what it is for a thinker to have tacit knowledge of one
fundamental reference rule rather than another. Different fundamental rules will
contribute to the explanation of different norms. Correspondingly, different rules
will entail different rationality requirements on judgments involving different con-
cepts. Evidence that a thinker has tacit knowledge of one particular content rather than
another will be evidence about the best explanation of various phenomena, including
the thinker’s appreciation of various norms involving the concept he is employing. In
being known by others by means of its explanatory consequences, this basis for
attribution of tacit knowledge is like any other case of tacit knowledge.
It is undeniable that in attributing concepts and understanding to our conspecifics,
we do in fact assume that others are like ourselves. We do take for granted similarities
that we do not, and should not, presume in members of other species. In this respect,
semantics resembles syntax, where the phenomenon of bold subpersonal generalization
from a slender body of evidence has been emphasized in Chomsky’s writing on the
acquisition of syntax.10 These facts are entirely consistent with the constitutive account
of what it is for a thinker to be, individually, meaning one thing rather than another, or
hearing a sentence under one syntactic description rather than another, not needing to
mention potential agreement within a community. These facts about conspecifics and
10
Summarized in Chomsky 1965, Chapter 1.
about attributions to others are facilitative of our making correct attributions of

understanding, rather than constitutive of their being correct.
(iv) Knowing Your Own Meaning: the Reasons-Awareness Account

If meaning green by the word ‘vert’ is having tacit knowledge of the appropriate
fundamental reference rule for ‘vert’, what is it to know, at the personal and conscious
level, that you yourself mean green by ‘vert’?
It is not plausible to identify knowledge of your own meaning with knowing that
you have tacit knowledge that ‘vert’ has a certain fundamental reference rule. The
identification of meaning with tacit knowledge of a fundamental reference rule is a
controversial philosophical claim. It is certainly not something known to everyone
who knows the meaning of his own words. Even if you have such a piece of
philosophical knowledge (if such it be), this philosophical knowledge is not as closely
connected to meaning green by ‘vert’ as is your knowledge of your own meaning.
Knowing that by ‘vert’ you mean green can long precede any philosophical knowledge
about the nature of understanding. There must be a way of knowing what you mean
that entails that you do in fact have the tacit knowledge involved in understanding, a
way of knowing that does not need to involve your conceptualizing your grasp or
understanding as a state of tacit knowledge.
I have been arguing that the tacit knowledge involved in understanding contributes
to the explanation of a thinker’s appreciation of (epistemic) reasons for making judg-
ments involving the concept in question, or for accepting sentences containing the
word in question. For someone who knows what he means, this appreciation of
reasons is conscious appreciation. The thinker is aware that such-and-such reasons
are (epistemic) reasons for making judgments or accepting sentences. This point seems
to apply to an arbitrary state or set of propositions that he takes as a reason for applying a
concept when this appreciation is explained in part by his tacit knowledge of the
fundamental reference rule for the concept or expression in question. The natural
hypothesis to propose is then this:
to know that by ‘vert’ you mean green is for your pattern of awareness of what are
(epistemic) reasons, in given informational contexts, for accepting sentences containing
the word ‘vert’, to be explained in an appropriate way by your tacit knowledge of the
fundamental reference rule involved in meaning green by ‘vert’.
I label the generalization of this account to arbitrary concepts ‘the Reasons-Aware-
ness’ account of knowing what one means. The Reasons-Awareness account meets the
desideratum that, under it, a thinker can know what he means by an expression
without having philosophical knowledge of the nature of understanding. Under the
Reasons-Awareness account, a thinker’s awareness of reasons for applying a concept
has to stand in a certain relation to the tacit knowledge in which his understanding
consists. But the thinker himself does not have to conceptualize his understanding as
tacit knowledge, nor does he have to think about the relations in which his awareness
has to stand to that tacit knowledge. The account does, however, require that the
thinker have that tacit knowledge.
There is an intuitive philosophical case for the Reasons-Awareness account. Not
knowing what is meant by an expression involves, in some central cases, not being
aware of what, in a given informational context, would be a good reason for applying
the expression. Similarly, if you were to mean something different by ‘vert’ than green,
and to know this different meaning, then there would be some reasons of which you
would be aware, if the issue arises, that would then be reasons for accepting a sentence
containing ‘vert’ (in a given informational context) that would not be reasons if you
were to mean green by it, or vice versa.
A further, more philosophical, consideration in support of the Reasons-Awareness
account is that it accords with other cases in which we want to say that meaning involves
tacit knowledge, and in which there is also knowledge of meaning. To use the old familiar
case, I have tacit, not explicit, knowledge of the definition of the concept chair. I may not
even know that I have tacit knowledge in meaning chair by ‘chair’; yet I do know what
I mean by ‘chair’. This knowledge of what I mean involves my awareness that various
types of presented example provide reasons for applying the concept chair, and involves
this awareness being explained by my tacit knowledge of the definition of the concept.
The Reasons-Awareness account of knowing what one means has certain features
that locate it in relation to other accounts and theses about knowledge of meaning that
have been proposed.
(i) The Reasons-Awareness account specifies a way for each person to know what he
means by an expression, a way that is available only for what he means, and not for
what others mean. Only his own tacit knowledge of fundamental reference rules can
suitably explain his awareness of reasons for applying a concept. Others’ tacit knowl-
edge of fundamental reference rules can explain their appreciation of reasons for
applying the concepts they possess. For each thinker, the Reasons-Awareness account
describes a way of coming to appreciate reasons for applying a concept that is available
only for the concepts he himself possesses. In this respect the account gives a distinctive
treatment of an arbitrary thinker’s knowledge of his own meanings that distinguishes it
from his knowledge of others’ meanings. This does not by itself of course have any
implications about incommunicability.
(ii) Thinking one means so-and-so by an expression is not, on the Reasons-Awareness
account, even partially constitutive of meaning so-and-so by the expression. The correct-
ness of such thinking is answerable to an independent condition, that of the thinker’s
really meaning so-and-so rather than something else by the expression in question. That
in turn is to be elucidated in terms of tacit knowledge of a certain fundamental reference
rule. The Reasons-Awareness account is, then, distinct from judgment-dependent ac-
counts of what it is for a thinker to be meaning one thing rather than another. Under the
Reasons-Awareness account, one can explain the phenomena that motivate judgment-
dependent accounts without any commitment to judgment-dependence.
(iii) The account I have offered of knowledge of what one means does not reduce to
knowledge of the triviality that a truth is expressed by ‘An object satisfies “green” iff it
is green’, or by ‘Un objet satisfait “vert” si et seulement si il est vert’. You can know
these sentences express truths without being aware of what would be reasons for
applying the words, in given circumstances. By contrast, such an awareness of reasons
is required for knowledge of meaning under the Reasons-Awareness account.
(iv) For the record, I note that this treatment is not committed to a ‘thought-first’
view of intentional content. The view is neutral on the relative priority, if any, of
thought and language. The fundamental reference rule may concern an expression
rather than a concept, and some theorists may hold that expressions are philosophically
more fundamental than concepts. The Reasons-Awareness account of knowledge of
your own meaning can still be built on top of such a view that language is, in the
philosophical order, explanatorily prior to thought. As far as I can see, the positions
developed in this essay are neutral on that issue.
The Reasons-Awareness account of knowledge of your own meaning is not
committed to a conceptual-role account of meaning. The reasons of which a thinker
is aware, when he knows the meaning of one of his own expressions, may be reasons
derivative from the fundamental reference rule for the expression.
(v) In writings on these issues subsequent to his first book on Wittgenstein, Wright
proposed that individual meaning is a matter of intention. Wright said, in discussing the
sceptical challenge raised in Kripke’s discussion, “ . . . to rebut the sceptical argument, it
would have sufficed, at the point where the sceptic challenged you to adduce some
recalled mental fact in order to discount the grue-interpretations, to recall precisely
your former intention with respect to the use of ‘green’. ”11 I disagree. An account of
the intention to mean so-and-so by a particular word cannot, it seems to me, be a full
account of individual meaning, because we also need an account of what it is for this
intention to be fulfilled. Intending to mean something by a word does not make it so,
not even for the subject’s own individual meaning. The subject may forget his
intention, or be unable to carry it through owing to some psychological blockage or
other. This need for an account of what it is for the intention to be fulfilled remains if
we replace ‘intention to mean’ by ‘intention to use the word in question so that it is
fundamentally true of just the so-and-so’s’. The intention on the one hand, and the
condition that it is met once it is executed on the other, are always different things.
There is also a converse problem with the appeal to intention in attempts to explicate
individual meaning. It is plausible that you can mean so-and-so by a particular word
without ever having had any intentions about the meaning that word should have on
your own lips. Yet without having had any such intentions, you still mean something,
and know what you mean by that word. This implies that your knowledge of meaning is
something distinct from your knowledge of your intentions with respect to the word.
11
Wright 1984, pp. 777–8.
On the Reasons-Awareness account of your knowledge that ‘vert’ means green, your
understanding explains your awareness of various patterns of reasons, in suitable
circumstances. Under the Reasons-Awareness account, knowledge of meaning, so
explicated, can indeed be present without your having any particular intentions about
what the word should mean individually for you. Your awareness of reasons for
applying the concept is properly explained by your understanding, be your intentions
as they may.
4. Between Linguistic Idealism and Extreme Platonism

Linguistic idealism holds that what it is for a Thought expressed in language to be
true—what it is for its truth-condition to hold—depends in part on human reactions
and practices. Linguistic idealism is a radical doctrine, to be distinguished from the
trivial and uncontroversial claim that what words mean depends on human reactions
and practices. The subject-matter of linguistic idealism is the intentional contents or
Thoughts expressed in meaningful sentences and the doctrine states, of those contents,
that what it is for them to be true depends in part on human reactions and practices.
The doctrine is entirely general. It claims to apply to all contents, not just to a proper
subset thereof. In endorsing this formulation of linguistic idealism, I am in step with
what Elizabeth Anscombe gives as her ‘test question’ for linguistic idealism: “Is this
truth, this existence, the product of human linguistic practice?”12
Rejecting investigation-independence does not by itself involve a commitment to
linguistic idealism. We can conceive of a theorist who insists that a certain kind of
content is not true unless investigated, but who also insists that the condition for the
correctness of the content does not concern human reaction or judgments, but rather
concerns something in the world. Some interpretations of quantum mechanics seem to
occupy this intermediate position. Such an intermediate position is not, however,
available to Wright’s Wittgenstein. The possibility of such an intermediate position is
dependent upon there being a condition, given by the meaning of the sentence in
question, whose holding, when investigation takes place, is not a matter of the thinker’s
judgments or reactions. That there is no such condition seems precisely the implication
of the position on rule-following held by Wright’s Wittgenstein. If there were such a
meaning-determined condition, there would be an account of going on in the right
(meaning-determined) way independent of the thinker’s judgments or reactions. The
application of the concepts in question would not be blind. But Wright’s Wittgenstein
holds a position on which insofar as we can make sense of going on in the correct way,
that is something to be explained in terms of human judgments and reactions. The
particular fashion in which Wright’s Wittgenstein denies investigation-independence
does seem to involve a commitment to linguistic idealism.
12
Anscombe 1981, p. 121.
It may be tempting to hold that the rule-following considerations, under a Wright-

like interpretation of them, imply only that human reactions and practices are involved
in the individuation of sense, and do not impinge upon the level of reference, proper-
ties, relations, and truth. The hope would be that, while a dependence of the nature of
sense on human reactions and practices goes beyond trivial doctrines about association
of particular words with senses, it does not go as far as an idealism about truth. But
I doubt that there is any real room here for a position to occupy. The rejection of
investigation-independence cannot be so easily confined. To each instance of correct
rule-following in the case of concepts, there corresponds a truth, something that is true,
under the conception in question, because the rule is being followed correctly. To the
fact that 1002, and not 1004, is the correct answer to the question asking for the sum of
1000 with 2, there corresponds the consequential truth that 1000 + 2 = 1002. If the
answer’s being correct depends constitutively on human reactions and practices, then,
under Wittgenstein as read by Wright, so too does the truth that 1000 + 2 = 1002.
I pause the argument at this point to note that there is a real question about
Wittgenstein’s own views on these matters. In }241 of the Investigations, he seems at
first glance to be saying precisely that truth is independent of human agreement, at
least:
“So you are saying that human agreement decides what is true and what is false?”—
It is what human beings say that is true and false; and they agree in the language they use.
That is not agreement in opinions but in form of life.
There are several possible reactions to this passage. One is that Wittgenstein has
simply misunderstood the consequences of his own views. A second reaction is that
such a simple misunderstanding is very unlikely, and }241 is therefore evidence that
Wright’s interpretation is wrong. A third reaction, which is neutral between these,
Wright’s, and other interpretations, is that Wittgenstein’s conception of ‘agreement in
opinions’ here is simply de facto agreement, and one part of his point here is that de facto
agreement never by itself amounts to correctness, of the fulfillment of normative
conditions. I won’t enter the exegetical issues here, but do note one non-exegetical
philosophical point. If the normative is indeed entering Wittgenstein’s discussion, that
still would not block the argument from one conception of rule-following to a
dependence of truth on human reactions and practices. For the crucial question is
whether the norms get a grip—whether they fix the application of a correct/incorrect
distinction—in a way that is constitutively independent of human judgments and
reactions. If they do, then it seems we have an account of content and rule-following
that is not of the sort endorsed by Wright’s Wittgenstein. If the norms do not get a grip
independently of human judgments and reactions, then even if the proper characteri-
zation of human reactions and practices involves norms (as it surely has to), we do
not have a conception of truth of a content that applies independently of human
reaction and practices. In either case, the connection between rule-following as
conceived by Wright’s Wittgenstein and the threat of linguistic idealism as character-
ized remains tight.
Few contemporary philosophers would endorse an unrestricted linguistic idealism.

Many would regard it as an extreme doctrine, a sign that something has gone seriously
wrong in any theory that entails it. Yet there is little contemporary agreement on
the correct reaction to rule-following arguments of the sort described by Wright.
(A present-day Kant might describe the situation as a ‘scandal of philosophy’.13) The
hypothesis I have offered, that understanding consists in tacit knowledge of a funda-
mental reference rule, allows us to reject a blanket linguistic idealism, whilst still having
a positive account of rule-following. On this account, whether a concept or a word
applies to some object depends only upon whether the object meets the condition
specified in the fundamental reference rule that is tacitly known by those who grasp the
concept. For observational shape concepts, this will be a matter of the object being of
the same shape as things are presented as being in certain specified types of experience.
For numerical concepts such as plus it will involve a triple of numbers satisfying a
certain recursive definition; and so forth. The fundamental reference rule need not
have anything to do with human reactions or practices in the application of the
concept.
Sometimes the content of the reference rule may involve relatively anthropocentric
notions. Colour concepts are arguably one such case. For such concepts, we may
legitimately say that the concepts themselves are anthropocentric. The legitimacy of
such a description turns on the details of the case, the details of the content of the
fundamental reference rule. On the present account, there is no universally applicable
argument, available for all concepts, to the conclusion that the truth of a content
containing the concept depends on human reactions and practices.
On David Lewis’s account, we ascribe by default to a rule-follower the content that
picks out the most natural property that best fits his judgmental practices.14 Natural
properties may be mentioned in the content of a fundamental reference rule, and for
concepts of shape, size, and other physical magnitudes, we may expect them to be so
mentioned. For observational concepts whose fundamental reference rules mention
perceptions that are instance-individuated in the sense of The Realm of Reason, it is in
the nature of the case that that is so.15 The perceptions have the contents they do
because in certain normal circumstances they are caused by what they are as of. So the
properties they represent must be capable of featuring in causal explanation of spatial
and temporal perceptions. But selection pressures may lead humans to form concepts
whose fundamental reference rules mention properties that are only locally important
for their survival, comfort, reproduction, and satisfactory emotional interactions with
others. Again, cases vary. A natural property may or may not be the best candidate to
attribute in the content of a thinker’s tacit knowledge. Whether it is so or not turns, as
13
Kant’s phrase in the Critique of Pure Reason (B xxxix note 144), about having to take “on faith the
existence of things outside us”.
14
See the final section, “The Content of Language and Thought”, in Lewis 1999.
15
On instance-individuation, see Peacocke 2004, pp. 67–73.
always, on the explanatory power of that attribution in explaining the thinker’s

appreciation of reasons for making judgments containing the concept in question.
Sometimes the reasons appreciated may make it more explanatory to attribute an
anthropocentric property as the reference of the concept; and sometimes not.
If we reject an unrestricted linguistic idealism, are we forced into an equally unaccept-
able Platonism about rules and concepts? Wright gives this description of a Platonistic
conception of rules, a conception both he and McDowell, whom he is discussing, reject:
The misunderstanding conceives of the requirements of a rule as, if they are to be
genuinely objective, matters whose detail has to be settled, case by case, in full
independence of any human propensity for reaction or judgment, so that the direction
that the rule takes might in principle be appreciable from a perspective wholly external
to ours—appreciable by a radically alien cognitive subject. This sublimation of rules
leaves us with a pair of hopeless tasks: first, that of providing at least some sort of
account of what, wholly independent of human propensity, determines the real
direction followed by a rule; and second, that of saying something useful about how
and why we might be presumed to be in cognitive touch with this “real direction”.
We have to learn the hard lesson that objectivity cannot be that.16
My account accepts some parts of this characterization of Platonism whilst denying
that these objectionable consequences follow from it. If the fundamental reference rule
for an observational concept such as round is as given earlier, then it really does follow
that what Wright calls the ‘requirements’ of the rule, the fundamental condition for
something to fall under the concept, viz. that the object be of the same shape as things
are represented as being in certain experiences, is in one clear sense something fully
independent of “any human propensity for reaction or judgment”. The required
condition of sameness of shape does not involve human propensities for reaction or
judgment. It involves a sameness of shape with a specified, given shape. This physical
condition is as independent of human propensities for reaction and judgment as any
other physical condition. Nonetheless, the condition is one that we humans can
sometimes know is fulfilled. That is enough to show that we can still, on the Platonistic
conception, be, as Wright puts it, in cognitive touch with what the rule requires in a
particular case. The fact that we can know, concerning what are the requirements of
the rule, that they are fulfilled is consistent with those requirements not consisting in
anything to do with human judgments or reactions.
That the rule which is grasped does require one thing rather than another—that one
content rather than another is the content of a thinker’s tacit knowledge in possessing
the concept—is of course not entirely independent of human propensities. One
content rather than another has that property because it is part of the best explanation
of the thinker’s appreciation of what are good reasons for judging contents containing
the concept in question.
16
Both passages are from Wright 1992, p. 206.
It also does not follow on the Platonistic conception that the direction the rule takes
might in principle be appreciable from a perspective wholly external to ours. Only a
being who knows something about what experiences and their representational con-
tents are will be in a position to know what shape is required by the correctness-
condition for something to fall under the concept round. The shape or shape-property
referred to by the concept round may of course also be referred to and thought about by
minds with very different perspectives on the world from our own, but not thought
about in the particular way involved in the concept we employ.
In short, on the hopeless tasks that Wright says Platonism faces, I would say: the real
direction followed by the rule is built into the content of the rule; we are in cognitive
contact with what it requires because the fundamental reference rule determines
reasons for applying the concept in new cases, reasons of which we can be aware and
whose requirements we can sometimes know to obtain; and what is involved in our
grasping one rule rather than another is a matter of best explanation of facts about our
application of concepts. What is objectionable is Platonism surrounded by ancillary
doctrines that would make the requirements of concepts and rules inaccessible to us,
and make an account of grasp correspondingly problematic. What is objectionable in
this package is the alleged inaccessibility, not the Platonism.
Rule-following is thus yet another area for which we must provide a theory that
meets the challenge of integrating its metaphysics and its epistemology.17 Epistemo-
logical concerns about accessibility may seem to push us in the direction of endorsing
some kind of judgment-dependence. Metaphysical concerns about objective condi-
tions for correctness of application may seem to push as towards an extreme Platonism.
Such seemings are both illusory on the account I have been offering. The treatment of
concept-possession as tacit knowledge of fundamental reference rules steers a middle
course, accepting a moderate Platonism whilst endorsing the judgment-independence
of truth in the general case. If the approach is right, the Integration Challenge for rule-
following can be met without endorsing inaccessible metaphysical distinctions, and
without abandoning the idea that in general, the condition for correctness of a
judgment is something independent of our making the judgment, that is, without
abandoning the idea of objectivity.
5. The Possibility of Substantive Philosophy

of Specific Concepts
Does the account of rule-following as involving tacit knowledge of a fundamental
reference rule undermine, or does it instead underwrite, an idea found in Richard
Rorty and others: that a proper appreciation of the rule-following considerations
undermines many classical and current disputes in metaphysics, and shows them to
17
See Peacocke 1999, Chapter 1.
be based on illusions? Rorty argues that once we have appreciated the full force of
Wittgenstein’s rule-following considerations there is no point to drawing a distinction
between a domain, such as shape concepts, for which a realistic conception of truth is
appropriate and one, such as notions of humour, for which some thinner conception is
appropriate.18 In this, he takes a position described, but not endorsed, by Wright late in
Truth and Objectivity (pp. 229–30). Does Rorty’s point apply to the account of rule-
following I have offered?
Rorty’s reason for his view involves the idea that what is represented is relative to a
convention (33), or at least to an ‘agreement’ (ibid.). In more detail, Rorty thinks that
the ‘input’ to a machine can always be described in various different ways, according
to varying ‘points of view’ (35). He connects this point with Wittgenstein’s discus-
sion of ‘plus’ (36). It is indeed true that a finite sample of triples of numbers can be
described as instances of the addition function, or as instances of any one of infinitely
many other functions that agree on that finite sample but diverge elsewhere. It is not
true, however, that the account of rule-following I have offered is tacitly relying on
some description of the input that is favored by agreement or convention. The
account I have offered does not favor, and does not need to favor, one rather than
another description of this ‘input’. These multiple descriptions of the input are
consistent with there being a fact of the matter as to which fundamental reference
rule is tacitly known by the rule-follower. Which rule he is following is fixed not by
any favoritism, based on convention or agreement, about the description of the
examples. The rule is fixed instead by the content of the tacit knowledge that
contributes to the thinker’s appreciation of reasons for making certain judgments.
Which fundamental reference rule it is that is tacitly known is evidenced by its
explanatory powers. As I emphasized, different contents for the tacit knowledge in
question have different consequences in some circumstance or other. There is, on my
view, nothing conventional or agreement-involving in the thinker’s having one
piece of tacit knowledge rather than another.
It is a different question, however, whether on the view of rule-following Wright
attributes to Wittgenstein, Rorty’s objection does get a grip. Wright insists, fairly
enough, that whatever one’s view on rule-following, the sort of distinctions he
draws between domains in which his criterion of Cognitive Command is fulfilled,
and domains in which it does not, remains applicable.19 It is true that that sort
of metaphysical distinction, in the terms in which Wright formulates it, does not
18
Rorty 1998, p. 37. Subsequent references to Rorty are to this reprinting.
19
Wright’s initial, ‘extremely rough’, formulation of Cognitive Command is that a discourse exhibits it “if
and only if it is a priori that differences of opinion arising within it can be satisfactorily explained only in terms
of ‘divergent input’, that is, the disputants’ working on the basis of different information (and hence guilty of
ignorance or error, depending on the status of that information), or ‘unsuitable conditions’ (resulting in
inattention or distraction and so in inferential error, or oversight of data and so on), or ‘malfunction’ (for
example, prejudicial assessment of data, upwards or downwards, or dogma, or failings in categories already
listed)”—Wright 1992, pp. 92–3.
obviously imply any particular stance on rule-following. The problem is rather that
when we look at the details of what is involved in Cognitive Command and its
contrary, a quite different model of concept-possession and rule-following is suggested.
This different model is one that, at least to first appearances, does not seem to invoke
human reactions or the community in the way in which the rule-following considera-
tions invoke them on Wright’s reading. Mention of an underlying state of tacit
knowledge that explains a thinker’s sensitivity in observational judgments of shape to
his perceptions of a shape property is the obvious way of developing an account of
shape concepts when we hold that Cognitive Command applies in spatial discourse.20
That account does not, or not without a lot of further argument, make correct rule-
following, for such concepts, constitutively a matter of human reactions, or a relation
to a community. Even in the case of humour, although the non-judgmental reaction of
finding something funny is certainly itself a human reaction, making judgments of
funniness that are suitably sensitive to the occurrence of this reaction is a further matter
involving rationality rather than a primitive human reaction. The more closely we look
at the nature of particular concepts, and the better the fine-grained theories of them
that we come up with, the less plausible it is that there is some entirely general element
of human reaction present in all such cases simply in virtue of their being cases of
concept-application.
The best explanation of why discourses differ in respect of whether Cognitive
Command holds of them involves a treatment of concepts that involves a commitment
to investigation-independence, and to a rejection of the rule-following considerations
in the form Wright ascribes to Wittgenstein. It is hard to think of alternative explana-
tions of differences between discourses in respect of Cognitive Command that do not
appeal to such treatments of concepts. The view of concepts I am offering, then,
diverges from Rorty in holding that there is plenty of room for substantive metaphysi-
cal and epistemological theory about the differences between various types of concept.
Good accounts of concept-possession do not undermine, but instead open the door to
the possibility of, substantive metaphysical theories of different domains. But my
position does agree with Rorty that the best explanation of these differences is possible
only under a treatment of concepts that also endorses the intelligibility of investigation-
independent truth.
6. Are These Explanations in Terms of Tacit

Knowledge Unsatisfactory?
There are many possible doubts and good questions about the legitimacy of explana-
tions of understanding and the rational application of concepts in terms of tacit
knowledge of reference rules.
20
Such an account is further elaborated in Chapter 1 of Peacocke 2008.
A first doubt may be that computational explanation by states of tacit knowledge is

still explanation by content-involving states, so that the approach has not solved the
problem I set out to solve. Part of this doubt involves attribution to me of an aim I do
not have. The aim was to give some philosophical account of understanding-based
rational application of concepts. The subpersonal content-involving explanations that
involve tacit knowledge are not explanations that involve transitions made for reasons,
or involve personal-level rationality. Similarly, computational explanations of the
content of visual perception in no way involve reasons for being in visual states.
(There are no such reasons, since having a visual experience is not a mental action.)
Nor do computational explanations of the content of visual experience involve
rationally drawing conclusions from the postulated subpersonal states with content.
But it was making a rational transition, rational on the basis of one’s understanding,
that was to be explained. Tacit knowledge of a fundamental reference rule can
contribute both to the philosophical elucidation and to the empirical explanation
of this phenomenon.
If the objection is instead that content-involving explanation should be eliminated
altogether, I protest. Content is essential in content-involving computational explana-
tion of the occurrence of mental states and events with specified contents. What is to be
explained is a content-involving, and commonly partially environmentally individu-
ated event—that someone is judging the content for a certain reason, in the presence of
other information. This is not just a mere de facto statement of a property or relation of
the event whose occurrence is to be explained. It is the very property or relation that
the explanation is to explain. Explanation at the neurophysiological level, or some
other level below that of content, would not be sufficiently general, and would not
support the counterfactuals that content-involving explanation supports. This is simply
the application to the case of computational explanations of rational judgment of points
that were made some years ago for content-involving computational explanation in
other areas.21
A second complaint might be that the alleged states of tacit knowledge are simply
projections from what they are alleged to explain, and so lack explanatory power. As
against this, it is a substantive claim that tacit knowledge with one content rather than
another explains why a thinker makes the rational judgments he does. Different
attributions of tacit knowledge have different predictive consequences. For observa-
tional concepts, for example, there is a real divergence in explanatory consequences
between the realistic reference conditions of the sort illustrated in }2 above and the
predictions of evidential and counterfactual accounts of observational concepts. That
was the case I tried to make in Chapter 1 of Truly Understood. In short, that case is that
neither evidential nor counterfactual conditions are ever necessary for the truth of a
predication of an observational concept. They can be made necessary only by turning
21
See Peacocke 1994; the point generalizes one made forcefully in the writings of Hilary Putnam.
the account into something of a much more realistic stripe. Our ordinary, non-
philosophical evaluation of predications of observational concepts in various circum-
stances reflects this fact. To keep this essay to a reasonable length, I can only refer the
reader to Chapter 1 of Truly Understood for detailed support of these claims.
If the tacit knowledge attributed were merely that in understanding round the
thinker tacitly knows that an object satisfies it iff it is round, that would indeed be
unexplanatory. The computational explanation would be replicating too much of
what was to be explained, viz. specific transitions to and from judgments and states
involving the conceptual content round. But none of the fundamental reference rules
I have proposed as tacitly known are of this unexplanatory form.
There are also other explanatory and counterfactual consequences of tacit knowl-
edge of the fundamental reference rules. Consider two observational shape concepts,
round and square, say. The fundamental reference rules for these two different concepts
both plausibly employ a notion of sameness of shape. If they do, then there will, other
things equal, be some potential patterns of pathology in which, because the brain has
lost the ability to represent sameness of shape between objects at least one of which is
unperceived by the subject, the thinker loses full grasp of both concepts—and any
other grasp of which involves the notion of sameness of shape.
A very different objection is that there is nothing in need of explanation here. The
suggestion would be that nothing explains why we make judgments for reasons in
certain kinds of circumstances. But why should we hold that? Would anyone say the
same about why we have the visual experience of stereopsis in certain circumstances,
and not others, or about why we find certain sequences of expressions grammatical and
others not? The need for genuine psychological explanation does not stop when we
move from mental states we passively enjoy to those that involve mental action. We
have an open-ended capacity, for all concepts we possess, to appreciate that certain
judgments of contents involving them are rational in certain circumstances and not in
other circumstances. The question of how we are able to do this is as much in need of
explanation as the question of how we are able to perceive depth stereoscopically, or of
how we are able to perceive the syntactic and semantic structure of sentences. We
similarly need an explanation of how we are able to appreciate, on the basis of our
understanding, that certain judgments are rational in certain circumstances. What is in
question at this point is not the explanation of what makes certain judgments rational.
That is an equally important and legitimate philosophical question; but it is a different
one from the phenomenon being discussed here, which has to do not with timeless
issues of rationality, but with the psychological phenomenon of the appreciation of
rationality.
A more reasonable concern is that these explanations in terms of tacit knowledge
have too large an a priori component to be genuine empirical psychological computa-
tional explanations. The objection would be that the computational explanation of
stereopsis certainly is not similarly a priori. It needs empirical investigation which
environmental arrays produce the distinctive experience of stereopsis, and which do
not. We have no good reason to accept a particular attribution of a computational

mechanism underlying stereopsis without relying on such wholly empirical data. Yet
the attribution of particular contents to tacit knowledge of fundamental reference rules
seems, by contrast, much more a priori.
I agree that the appearances are so. But the explanation of the difference is that in
working out, by inference to the best explanation of the data, the content of the tacit
knowledge underlying possession of a concept, we already know the range of data to
be explained, because we already know what would be reasons for making judgments
containing the concept in various circumstances. This knowledge is made available by
the awareness I discussed in the motivating the Reasons-Awareness account of knowl-
edge of meaning and intentional content. It is a relatively a priori inference that such-
and-such tacit knowledge, as underlying possession of the concept F, would explain a
range of examples of a thinker’s appreciating so-and-so as reasons for judging . . . F . . . in
the presence of such-and-such background information. So the explananda which
could only be obtained by empirical investigation in the case of stereopsis have
analogues in the case of concepts and understanding that are available without similar
empirical investigation.
The situation is comparable to that of a grammarian investigating the tacit knowl-
edge underlying mastery of a language she herself already understands. She knows from
her own case, without empirical research, which strings are found to be grammatical by
speaker-hearers, and which are found to be ungrammatical. This can be the basis for
her reasonable attribution of tacit knowledge of one set of grammatical rules rather
than another. We would not and should not conclude in the grammatical case that our
linguist is not making empirical claims about the psychological explanation of gram-
matical capacities possessed by speaker-hearers. No more should we conclude that
properly reached hypotheses about the tacit knowledge underlying possession of a
concept, reached by abductive inference from information about thinkers’ apprecia-
tion of various good reasons in various circumstances, are not equally empirical in their
import and explanatory power.
A more serious set of questions concerns the use of a notion of content at the
subpersonal level. Some of the sceptical issues raised in discussions of rule-following
question whether meaning and intentional content is possible at all. If the correct
answer to some of these questions involves, as I have argued, appeal to subpersonal
states with their own content, it follows, in the nature of the case, that we cannot give
the same answer over again in replying to sceptical challenges about subpersonal
content. There is no relevant subsubpersonal level to which to appeal. Even if there
were, it would only postpone the issue, and not remove it. There must be some
different and more direct way of addressing challenges about content that is available at
the subpersonal level. What answer can we give to the objector who says that, even
though it is agreed that we are now below the level of reason-involving states and
reason-involving explanation, we must still say what is involved in a subpersonal state
having one content rather than another? Certainly we must be able to say something
74 C H R I S TO P H ER P EAC O C K E
about what is involved in a subpersonal state’s having successor in its content rather than
some variant notion, having distance rather than some variant, and so forth.
In the case of some of the subpersonal computational states involved in early vision,
the notion of causal explanation plays a part in answering the question of what it is for
them to have one content rather than another. A subpersonal representation of the
depth of an edge (say) in early vision has the content it does in part because representa-
tions of its relevant type are, predominantly and normally, causally explained by the
depth of the edge from the perceiver. It would be quite wrong to say such representa-
tions could equally be described as causally explained by any other property that agrees
extensionally on edges perceived hitherto, but diverges on edges as yet unperceived.
‘Causally explains’ does not work that way. The operator ‘causally explains’ is more
discriminating; and the correct attribution of content to subpersonal events relies in
part on this additional discrimination. I would say roughly the same about cases in
which causal explanation runs in the opposite direction. The internal state of a Turing
machine that represents the relation of being one step forward on the machine’s tape
when it is issuing commands has the property that, predominantly and normally, its
instances explain the print-head moving one step forward.
These causal considerations are not the only factor in the determination of content.
In the case of depth perception, the account must discriminate the relation of depth
from relations nomologically coextensive with it in ordinary circumstances. It is
plausible that such further discrimination will also have to rely on facts about the
effects of the representation further downstream in the system.
What these points do suggest is a programme for answering the general query about
subpersonal content. The proposal is that we can marshal facts about causal explana-
tion, both of the subpersonal states and by the subpersonal states, to ground legitimate
attributions of content to them, attributions that favor the contents needed, rather than
their sceptical variants. Though no one could claim that this programme has already
been carried out, it seems to me that it has reasonable prospects. There are some
examples for which the programme is plausible; and it is not clear that it faces sceptical
obstacles of principle. If the programme can be carried through, then there is a level of
states with content available for explicating the tacit knowledge that is involved in
understanding and the rational application of concepts.
The programme of explaining appreciation of reasons for accepting contents in
terms of tacit knowledge of reference rules for the concepts in the contents may still
seem unduly speculative and very distant from our ordinary explanations of events in
personal-level mental life. I conclude by suggesting that it should not seem so distant.
There is a parallel between the explanation of someone’s hearing an utterance as having
a certain meaning, and explaining someone’s appreciation of something as an (episte-
mic) reason for making a judgment with certain content containing specific concepts.
Part of our ordinary conception of persons in the linguistic case is that the person’s
understanding of the whole utterance is explained by his understanding of the individ-
ual words uttered, and his understanding of their mode of composition. We have this
ordinary conception even though we do not in our everyday folk knowledge have any
articulated idea of how this explanation works, or even of what it is to understand an
individual word. A similar combination holds for appreciating something as an episte-
mic reason for judging a given content. It is part of our personal conception that this
appreciation is explained by the thinker’s grasp of the concepts that compose the
content, but how this explanation works, or even what it is to grasp the concept, is
not something present in folk knowledge or conceptions.
There are, however, constraints on this explanation, and the computations involved
in it. These constraints are analogous to those on correct computational explanations of
early vision. The fact that vision is generally reliable—we largely see the world around
us correctly—imposes constraints on the computations. The computations must, by
and large, involve correct computations of magnitudes, properties, and relations from
the input on which they operate. Similarly, the fact that thinkers are, by and large,
roughly right about what constitutes a good reason for making a judgment involving a
given concept imposes the constraint on subpersonal computations from fundamental
reference rules that they be largely sound. I suggest that an explanation of knowledge
of reasons should take its place alongside explanations of linguistic understanding and
visual experience as something that involves a combination of pre-theoretical con-
straints, empirical mechanisms, and an essentially content-involving conception of
computation.
Bibliography
Anscombe, G. E. M. 1981 ‘The Question of Linguistic Idealism’, reprinted in The Collected
Philosophical Papers of G. E. M. Anscombe. Volume One: From Parmenides to Wittgenstein,
Minneapolis, University of Minnesota Press, 1981, pp. 112–33.
Chomsky, N. 1965 Aspects of the Theory of Syntax, Cambridge, Mass., MIT Press.
Press.
Lewis, D. 1981 “New Work for a Theory of Universals”, reprinted in Papers in Metaphysics and
Epistemology, Cambridge, Cambridge University Press, vol. 2, pp. 8–55.
McDowell, J. 1994 Mind and World, Cambridge, Mass., Harvard University Press.
Peacocke, C. 1994 “Content, Computation and Externalism”, Mind and Language 9,
pp. 303–35.
Peacocke, C. 1999 Being Known, Oxford, Oxford University Press.
Peacocke, C. 2001 “Does Perception have a Nonconceptual Content?”, Journal of Philosophy 98,
pp. 239–64.
Peacocke, C. 2004 The Realm of Reason, Oxford, Oxford University Press.
Peacocke, C. 2008 Truly Understood, Oxford, Oxford University Press.
Rorty, R. 1998 “Is Truth a Goal of Inquiry? Donald Davidson versus Crispin Wright”, in
Truth and Progress: Philosophical Papers, Volume 3, Cambridge, Cambridge University Press,
pp. 19–42.

Wright, C. 1980 Wittgenstein on the Foundations of Mathematics, London, Duckworth.
Wright, C. 1984 “Kripke’s Account of the Argument Against Private Language”, Journal of
Wright, C. 1992 Truth and Objectivity, Cambridge, Mass., Harvard University Press.
3
Regularities, Rules, Meanings,
Truth-Conditions, and Epistemic
Norms
Paul Horwich
1
My plan is to trace an explanatory route through the above-mentioned phenomena.
I guess it’s uncontroversial that they are in some way intimately related. The question
is, how. At one extreme there are philosophers who suggest that we’re faced with a
holistic hodge-podge: the interconnections are messy, and none of these things is
objectively more fundamental than any of the rest. At the other extreme there are those
who think that there is a correct order of grounding amongst them—a definite
hierarchy whereby the more superficial ones reduce to the more basic ones. My
own sympathies are more in line with this second point of view. However, it’s not
that I hold for some sort of general metaphysical reason that matters simply must be like
that. It’s rather that such a story would be particularly simple; simplicity is good if you
can get it—a significant explanatory virtue; and in this case it seems to me that you can
get it.
More specifically, what I’ll be suggesting is that the basic facts here are law-like
regularities of word use (characterized in non-semantic, non-normative terms); that such
regularities help engender (i.e. are the primary reductive basis of ) facts about which
rules of use we are implicitly following; that these facts suffice to fix what we mean by
our words and hence sentences; and that the meanings of our sentences (given
contextual factors) determine their truth-conditions—which we ought to desire to be
the conditions in which we accept them.
This picture is based on Wittgenstein’s ideas about meaning and rule-following.
Nonetheless it is at odds with Saul Kripke’s treatment of those concepts, which is billed
as Wittgensteinian. And it also conflicts with the important body of work on them
elaborated over the last thirty years by Crispin Wright—another influential exponent
of Wittgenstein. So a fair part of the defense of my own position will consist in
78 PAU L H O RW I C H
responses to their arguments.1 At stake are both the true nature of these phenomena
and the most fruitful way of reading Wittgenstein’s discussion of them.
2
Normal rules are explicit instructions. Typically, we hear or read formulations of them,
or think them to ourselves;2 we understand these formulations—that is, we attach
semantic contents to them; we then decide to do what is dictated by those contents;
and so we conform more or less successfully. The rules of games such as chess and
football are like this, as are the rules of a club, the rules of parliamentary procedure, the
laws of certain countries, the rule that I impose on myself for how many times a week
to go to the gym, and so on.
However, we are sometimes inclined to invoke rule-following when the immediate
sources of action involve no such formulation of the rule, hence no understanding and
no deliberate decision to abide. For example, we speak of there being rules of English
grammar. We say that the utterance, “John seems sleeping,” goes against these rules,
whereas “John seems to be sleeping” doesn’t. But not even linguists—let alone
ordinary speakers—are aware of the rules or able to state them. Nor is there reason
to think that we articulate them to ourselves unconsciously—that they are ‘written
down’ in some deep recess of the mind. Similarly, consider how far apart two people
stand from one another when they are having a conversation. We have a sense that
some distances are too close for comfort and some are too great. I’ve heard that feelings
and customs on this matter vary somewhat from society to society. Again it is
somewhat tempting to speak of rules—even in absence of articulations of them and
decisions to do what they say.
Let me use the terms “explicit” and “implicit” to signal these two kinds of rule-
following. Explicit rule-following is relatively familiar: the rule is spelled out, and
that formulation, in virtue of its meaning, is allowed to guide our activity. Implicit
rule-following is more of a theoretical posit and more murky. It’s not the same as
unconscious rule-following—because that could be a matter of unknowingly, but
explicitly, following a rule that is formulated in one’s language of thought. The
essence of implicit rule-following is that no formulation of the rule is directly
operative.
1
See Kripke 1982 and Wright 2001, especially his “Wittgenstein’s Rule-following Considerations and
the Central Project of Linguistics.” See also a more recent essay of Wright’s— “What is Wittgenstein’s Point
in the Rule-following Discussion?”—presented to the NYU Language and Mind Seminar in April 2002, but
as yet unpublished.
2
Note that even when someone’s rule is not overtly articulated—even when there is merely the active
intention to conform with a given regularity—this state may nonetheless be explicit, i.e. it may involve a
mental formulation, a mental ‘conveyor’ of its propositional content.
R E G U L A R I T I E S , RU L E S , M E A N I N G S . .. 79
3
When we philosophers speak of ‘the problem of rule-following’ we tend to have in
mind the implicit kind. Granted, one might raise the question of what are the facts in
virtue of which someone explicitly follows the rule, R!—the rule dictating conformity
with regularity R. What makes it the case that an act is, or is not, in accord with a
person’s explicit rule? But the initial answer to both questions is obvious: it’s the
meaning of the rule-formulation that’s responsible for these things; a written-down rule
is the rule it is in virtue of its particular imperatival content; and if its content is to
perform an action of type A in circumstances C, then an act (in C) is in accord with the rule
just in case it is an act of type A. No doubt one then can press the issue by asking how it
comes to be that a given rule-formulation means what it does. And no doubt this is a
tough problem. But an answer to it will yield a full account of explicit rule-following.
Implicit rule-following, on the other hand, presents peculiar difficulties. For, with no
formulation to help us, it is particularly hard to see what the facts might be in virtue of
which someone would be implicitly following a given rule (rather than a different one,
or no rule at all). We don’t see what could possibly mark a particular act as a violation of
the rule. Indeed, the difficulty of these questions might even persuade us to resist our
initial inclination to recognize ‘implicit rule-following’ as a genuine phenomenon.
4
However, such a retreat would be unwarranted. It’s not, as some would say, that the
hard questions are bad questions, vitiated by irrational reductionist presuppositions, and
that we should be content with primitive, ungrounded, implicit rule-following.3 It’s
rather that, as it seems to me, an adequate explication of the notion can be provided.
Consider the following suggestive extract from paragraph 54 of the Philosophical
Investigations. After first mentioning explicitly formulated rules of a game, Wittgenstein
continues:
Or a rule is employed neither in the teaching nor of the game itself; nor is it set down in a list of
rules. One learns the game by watching how others play. But we say that it is played according to
such-and-such rules because an observer can read these rules off from the practice of the game—
like a natural low governing play. But how does the observer distinguish in this case between
players’ mistakes and correct play?—There are characteristic signs of it in the players’ behavior.
Think of the behavior characteristic of correcting a slip of the tongue. It would be possible to
recognize that someone was doing so even without knowing his language.
Along roughly these lines, I would suggest that person S implicitly follows rule R! if
and only if
3
Anti-reductionists about rule-following include Kripke 1982, McDowell 1984, Wright 2001, and
Brandom 1994. According to these philosophers (but in contrast with what I am about to suggest)
Wittgenstein too would have been unsympathetic to any reductionist account.
(a) S’s activity is governed by the ideal law R

(b) There is some tendency for S to correct instances of non-conformity (i.e. to
react against his initial inclinations)4
Here, the role of condition (a) is to specify which aspect of the agent’s activity fixes
the content of his rule—i.e. what makes it R!, rather than R*!, that is being followed.
And the role of condition (b) is to explain why S may be regarded as following a rule,
rather than as merely obeying a natural law (which is what, for example, the planets do
in orbiting the sun). For a tendency to self-correct manifests sentiments of relative
dissatisfaction and satisfaction, and thereby encourages talk of “goals,” of “what
ought to be done,” and of “violations.”5 Let me stress—and perhaps this diverges
from Wittgenstein’s idea—that I am not relying on S’s practice of self-correction to
identify precisely which actions are to count as violations. I’m relying on it merely to
motivate applying terms such as “violation” to cases of non-conformity—cases that
have been independently identified as deviations from the ideal law.6
5
Now one might worry about the notion of ‘ideal’ that is deployed here. The point of
inserting it is, of course, to register that we are talking about the sort of law that has
exceptions—for we have to accommodate the fact that rules are sometimes inadvertently
disobeyed. But isn’t reliance on such a notion inconsistent with our naturalist/reduc-
tionist aspirations?
I would suggest that there is no such tension, because this particular notion of ‘ideal’
is not normative, and is in no way naturalistically problematic. There is a familiar form
of explanatory theory that consists in postulating a system that is governed, in so-called
‘ideal’ conditions, by certain laws, but which is subject to a variety of factors that
cause deviations from the behavior that would ‘ideally’ occur. For example, only in
certain ‘ideal’ conditions do the orbits of the planets obey Kepler’s Laws; and only
in certain ‘ideal’ circumstances do gases obey Boyle’s Law. Questions of whether some
4
With Wittgenstein, I see no need to insist that the correction-practices involved in implicit rule-
following must stem from other people, and therefore no reason to reject the possibility of individualistic
rule-following. Granted, he says (PI 202): “And hence also ‘obeying a rule’ is a practice. And to think one is
obeying a rule is not to obey a rule. Hence it is not possible to obey a rule ‘privately’: otherwise thinking one
was obeying a rule would be the same thing as obeying it.” But what he means here is that someone’s
following a rule requires more than her subjective impression of doing so: an objective regularity (“practice”)
must be there. He is not backing away from his earlier (above-quoted) remark, which would suggest that
someone’s following a given rule, R!, might be an intrinsic property of that individual.
5
There will be further discussion of the normative import of rule-following in section 9.
6
We can’t simply identify cases in which S deviates from his rule with cases in which he corrects himself
(or has a disposition to do so)—because he may well fail to notice certain cases of non-compliance. Thus S’s
rule cannot be straightforwardly read off his practice of self-correction. Nonetheless, that practice is an
important part of the empirical evidence that can help us (in the way indicated in section 5) to reach plausible
conclusions as to which combination of ideal laws and occasional distorting factors are influencing S’s activity.
such theory is plausible in a given context, and of which ideal laws and potential
distorting factors should be postulated, are settled by the standard canons of scientific
methodology—i.e. by reference to empirical adequacy, simplicity, coherence with
other successful theories, etc. Thus explanatory accounts of this kind may be thor-
oughly empirical and naturalistic.7
6
But how does my proposed account of implicit rule-following differ from the one that
Kripke himself considers and refutes? He assumes—rather as I am assuming—that if
S follows rule R!
can be analyzed at all, it would have to be to something like
in ideal conditions S’s activity would fit regularity R

However, he argues that no such account can in fact be sustained; for it isn’t possible
to articulate, in a satisfactory way, what is meant here by “ideal conditions.” All
attempts to do so either fall into circularity (e.g. ‘conditions in which S would not be
led to violate his rule’), or resort to a list of allegedly ‘perfect’ circumstances (e.g. ‘S lives
forever, has an infinitely large memory capacity, etc.’)—highly remote postulations
whose hypothetical bearing on S’s activity are a matter of empirical speculation.
So what is my defense against this objection? Simply this. Kripke seems to presup-
pose that we are obliged to give an a priori specification of which conditions are ideal.
And I think he’s right that we can’t do that. But I think he’s wrong in assuming that we
need to. Although my explication of rule-following in terms of the notion of ‘ideal
law’ is indeed a priori, one must then deploy standard scientific desiderata (including
simplicity and empirical adequacy) to discover which particular combination of ideal
law and potential distorting factors governs S’s activity. It’s an a posteriori matter.8
7
It might be objected that there can be no determinate fact as to the (unique) ideal law governing a given
phenomenon. For various idealizations—various suppositions of the form, ‘Let’s pretend that such-and-such
complicating factors are absent’—are always possible; and so it is always up to us, in light of the purposes at hand,
to stipulate which such simplifying suppositions we are making on a given occasion. But what I have in mind
here by “the ideal law” is the law that obtains in the scientifically best idealization—the one that best combines
simplicity, empirical adequacy, and the satisfaction of our other methodological desiderata. Such ‘ideal laws’ are
also known as ‘ceteris paribus laws.’ Outside fundamental physics, nearly all scientific laws are of this kind.
Granted, there may sometimes be more than one ‘equally best’ idealization. And in such a case—provided
the ‘correction condition (b)’ on implicit rule-following is satisfied—we would have to say that there is a
certain degree of indeterminacy as to which rule is being implicitly followed. In particular there may well turn
out to be such indeterminacy as to which particular rules implicitly govern our use of certain words.
However—as we shall see in section 13, and contrary to what one might be tempted to think—this
would not engender any indeterminacy in the meanings or extensions of those words. I would like to thank
Paul Boghossian and Allan Gibbard for pressing me on the points in this footnote. See Gibbard 2008.
8
One might suspect that the problem of specifying ‘ideal conditions’ cannot be so easily disposed of.
For in the case of certain rules—e.g. ‘Apply word w to all and only electrons’ and ‘Accept “x plus y = z” iff x
7
This suggestion may seem tantamount to another of the proposals that Kripke rightly
dismisses: namely, that one ground
S implicitly follows the rule R!

in something along the lines of
that roughly conform with S’s activity R is the simplest of the simple regularities
Kripke’s criticism is that this proposal is off-target, that it misconstrues his skeptical
challenge.9 For simplicity is to be invoked only when we are confronted with an
epistemological question as to which of two well-understood hypotheses is true; but
the present problem is to articulate what it means to say that that a certain rule is being
followed, to articulate the contents of competing hypotheses of that sort.
Now, I am inclined to agree with Kripke that the account he criticizes does indeed
wrongly confuse what might arguably be part of the evidence for a rule-following thesis
with the underlying fact that would make the thesis true.10 But my own proposal
certainly respects that distinction. For what I suggested was that ‘following rule, R!’
consists (partly) in ‘being governed by ideal law, R.’ And if pressed on what this
underlying fact comes to, my reply is that there is no deeper analysis of it, but that one
can say, invoking considerations of simplicity, what kinds of data would give us reason
to postulate it.
plus y = z’—it is plausible that the difficulty in coming up with conditions necessary and sufficient for
obedience is not merely that no such (non-semantically specifiable) conditions can be discovered a priori, but
that no such conditions c! And I would concede this point. But I’d argue that we can happily embrace what it
implies: namely, that not all rules of the particular form—‘Apply w to all and only the fs’—can be implicitly
followed.
This will not in turn imply that certain predicate-meanings are irreducible. For, as we’ll see in section 12,
it’s a mistake to assume that the implicit rule-following that engenders ‘meaning F ’ is always of the truth-
oriented rule, ‘Apply w to all and only fs.’ Rather, an unprejudiced empirical investigation of which factors
are governing overall word usage will turn up ideal laws (hence, rules) of a variety of very different forms.
9
See Kripke 1982, p. 38.
10
Kripke remarks that the operative confusion results perhaps from “the influence of too much philoso-
phy of science.” It’s not clear exactly what he has in mind here. But maybe he is expressing disagreement with
the Ramsey–Lewis account of what it is to be ‘a basic law of nature’ (roughly, ‘One of the axioms in the
simplest of the simple axiomatizations whose theorems characterize all (and only) the facts’). And maybe he
thinks that this piece of philosophy of science is an especially prominent example of confusing evidential
conditions with constituting conditions.
Of course, the evidential import of simplicity is no bar to the existence of concepts that are best explicated
in terms of the notion of simplicity, and hence no bar to the possibility of hypotheses whose contents are to be
characterized by reference to that notion. So in giving an analysis of rule-following in terms of simplicity
someone needn’t be failing to see that a metaphysical (constitution) thesis is called for. Therefore, Kripke’s
criticism is best articulated as follows: that the proposed analysis is incorrect, and that its air of plausibility stems
from moving too quickly from a theory of evidence to a theory of constitution.
8
But this involves an idea that Crispin Wright has found objectionable—the idea (which
he attributes to Chomsky) that instances of ‘following rule R!’ must be regarded as
theoretical phenomena whose identification requires scientific investigation. For he claims,
on the contrary, that it is via some privileged introspective route that we know which
rules we are following. As he puts it
Kripke himself objects to the dispositional response that it cannot account for the normativity of
understanding an expression in a particular way, intending to follow a particular rule, and so on.
The reason for dissatisfaction which I have in mind, however, is not this. It is rather that
Chomsky’s suggestion, that the identity of followed rules is a strictly theoretical question,
threatens, like the dispositional account, to make a total mystery of the phenomenon of non-
inferential, first-personal knowledge of past and present meanings, rules and intentions.11
We’ll be addressing Kripke’s ‘normativity objection’ in the next section. But, as for
Wright’s point, I’d say that it’s far from obvious that each of us knows, independently
of any sophisticated empirical research, which rules we are implicitly following. Con-
scious, deliberate, explicit rule-following is a very different matter. There, perhaps, we
are able to tell simply by introspection, what we are up to. But in the implicit case, the
most we can be expected to detect through introspection are occasional inclinations
and disinclinations to do this or that particular thing. The underlying rule is available to
us only via objective methods of inquiry that are equally open to everyone else.12
9
This brings us to what is the perhaps the most widely repeated of the skeptical
objections to broadly ‘dispositional’ accounts of rule-following. It’s implicit in Kripke,
as Wright says, and it’s echoed in a great deal of the work provoked by Kripke’s
discussion. The allegation is that ‘naturalistic regularity’ proposals such as mine could
not do justice to the normative import of rule-following, to the fact (roughly speaking)
that
S follows rule R! !
S’s activity ought (other things being equal) to be consistent with regularity R
11
Wright 2001, see especially pp. 174–7. Kripke sketches a similar argument in Kripke 1982, pp. 39–40
(including footnote 25).
12
Both Wright and Kripke also maintain, in the same vein, that a word’s meaning what it does cannot be
constituted by theoretical facts (for we surely know directly what we mean). However, the only fact as to what
(say) “plus” means that ordinary speakers explicitly know is that it means PLUS and that it is true of the plus-
triples—and, granted, that knowledge is immediate and trivial. But such facts, although perhaps a priori, are
nonetheless contingent, and may perfectly well be empirically reducible to facts that are not ‘directly’
accessible. Compare the relationship between “Our rain, rivers, lakes and seas consist mainly of water”
(which is more-or-less a priori) and “Our rain, etc. consists mainly of H2O” (which is a scientific discovery).
But I would dispute this allegation. Granted, no ‘ought’ flows merely from S’s being
governed by a certain natural law—not even if it’s an ideal law. However, once we
bring in the further condition, (b), that I’ve argued must be satisfied in order for the
operation of that law to qualify as rule-following—namely, S’s tendency to correct
some of his own initial impulses to deviate—we are in a position to see why deviations
are naturally considered to be “wrong.” For the practice of occasional self-correction
manifests selective dissatisfaction, and hence orientation towards a goal; and there is
surely some positive value in the achievement of goals.
This is not to claim that rule-following, under my analysis of it, logically entails
something normative. Why should it? After all, ‘S wants to do X’ does not—even in
the case of explicit desire—logically (or conceptually) necessitate ‘S has reason to do X.’
Yet, for all that, we do accept that S has reason to do what he wants to do; we do give
normative import to a person’s explicit desires. And all I am claiming for my proposed
analysis is that it enables us to see why it is unsurprising for us also to give that sort of
significance to implicit rule-following.
Granted, we haven’t explained that significance. We are treating the normative
import of rule-following—like that of explicit desire—as explanatorily fundamental.
Still, we are able to understand why we accept that if
R is an ideal law governing S’s activity (in a context of occasional self-correction)

then
S has some reason to conform his activity with R

and we are thereby explaining our attribution of normative import to rule-following.13
10
Let me turn now to facts about meaning. And let me begin with a point that, I think, is
not sufficiently recognized and emphasized in the post-Kripkean discussion: namely,
that meaning and rule-following are distinct phenomena and that the philosophical
problems surrounding them are somewhat separate. This bears emphasis because
Kripke writes as though it’s all the same thing. He moves from one issue to the
other in the same breath, as if there were a single topic here.
As far as explicit rule-following is concerned, since understanding is essentially
involved, any skeptical problems concerning meaning will indeed infect it. But such
problems will equally infect all other forms of linguistic activity. There is no reason to
single out rule-following.
13
It may be, strictly speaking, that one has reason to conform with one’s rule only if it isn’t unreasonable to be
following it. But presumably this insertion of normative content into the antecedent of the conditional at issue
could only make it easier for a regularist to accommodate its normative consequent.
And as far as implicit rule-following is concerned, its relationship to meaning is

arguably (but not uncontroversially—see section 14) that meaning a given thing by a
word is a matter of implicitly following particular rules for its use. If that is right, then
any problems surrounding rule-following will be inherited by meaning. But it is worth
stressing the obvious point that rules can govern activities that have nothing to do with
with language. Therefore, issues that relate specifically to meaning (e.g. the question of
which particular rule is involved in engendering a given meaning) need correspond to
no general problems about rule-following.
Thus I don’t think we can go along with Wright’s remark that the rule-following
problem “in its most general form” is the problem of how it is determined that a person
ought (or ought not) to assert a given sentence of his language in given circumstances—
i.e. how it is determined that the rules governing its use dictate assent (or not).14 For, as
just noted, there are rules that do not bear upon the use of words. Moreover the role of
rules in language is highly atypical. For they operate at several different levels and
correspond to different kinds of normative constraint. In particular, we can distinguish
(1) meaning-determining rules of use, (2) the rule instructing us to mean the same by a
word as others do within our community, (3) rules of epistemic justification, and (4)
the rule dictating that we try to accept only those sentences that are true. Therefore, to
treat language-use as a basic paradigm of rule-following is to court confusion.15 It is
preferable to proceed by first trying to achieve (as in sections 2–9) a truly general
account of what it is to implicitly follow rule R!; then considering (as we will in
sections 11–14) questions of whether and how a word’s meaning might derive from
following rules for its use; and lastly (see section 15) examining the important norma-
tive features of meaning, including (i) what we ought to mean, (ii) which sentences we
ought to accept (in light of their meanings and the available evidence), (iii) and which
we should want to accept (in relation to what is true).
11
If meaning a given thing by a word is a matter of following rules for its use, it must be at
least sometimes a matter of implicit rule-following. Otherwise we’d have a vicious
regress. For any explicit formulation of such a rule would be in a language the
understanding of which would rest upon deeper explicit rules, the understanding of
which would rest upon yet deeper explicit rules, . . . , ad infinitum. To put it another
14
See Wright 2001, p. 4.
15
For example, consider the following passage: “Is not the image of the syntactic-semantic engine simply
a graphic expression of the rule-governed nature of linguistic practice? And if we reject it, are we not, in
effect, rejecting the idea that proper linguistic practice is rule-governed—and thereby rejecting the notion of
strictly correct linguistic practice, and with it the notion of meaning?”—Wright 2001, p. 180.
From a perspective in which the various above-mentioned rules and norms are kept distinct from one
another, there will be little temptation to think that if word-use were not a matter of following rules then the
notions of correctness (i.e. truth) and meaning could not enter the picture.
way: explicit rule-following is explanatorily dependent on facts of meaning; and

explanatory dependence is a transitive and irreflexive relation; so it cannot be that
each word’s meaning what it does is explanatorily dependent on our explicit following
of rules for its use.16
This sort of concern admittedly leaves it open that only some terms—our most basic
ones—are understood in virtue of implicit rules, that certain further words are under-
stood in virtue of explicit rules formulated using these most basic terms, that yet further
words are understood in virtue of explicit rules using terms from the first two groups,
and so on.
However, in the absence of any persuasive considerations to the contrary, it is not
plausible that there be two kinds of meaning-property: those partly grounded in
explicit rule-following and those grounded—very differently—entirely in implicit
rule-following (i.e. in ideal laws and self-correction). Therefore, if rule-following is
involved at all in the constitution of meaning, then—given the need to avoid regress—
it is plausibly always a matter of implicit rule-following.
Of course, one might well wonder why we should think that meaning is rule-
following of any kind. I’ll come back to this issue in section 14.
12
But if we assume for the time being that meaning is a matter of implicit rule-following,
the central questions become: (1) how are we to find out which rule (or rules) are the
ones the following of which will constitute a word’s having any given meaning—for
example, meaning water, or meaning true, or meaning electron; and (2) what
specifically will the rules turn out to be, say, in these three particular cases?17
A view that has captivated many philosophers—including Kripke, I think, but he is
frustratingly inexplicit—is that, for the case of predicates
S’s meaning F by word w (i.e. S’s meaning by w what “f ” means in English) is

constituted by S’s following the truth-oriented rule: apply w to all and only the fs18
A primary cause of the temptation to say this, I suspect, is—for example—that if
S means dog by w then (probably) S would ideally like it to transpire that he applies
w to all and only the dogs. Given this meaning, that’s what he will aim to do. And vice
versa: he will regulate his use of w in such a way only if he means dog by w.
16
I am taking the view that a person’s explicitly intending to conform with R depends on there being a
certain combination of mental elements (within his medium of thought) that possesses the meaning,
conform with r. So—given the present hypothesis that meaning is rule-following—this content would
have to derive from those elements themselves being rule-governed.
17
I employ the convention of writing expressions in small capital letters in order to refer to their
meanings. Thus “water” is to name the meaning of our (English) word, “water,” etc.
18
Arguably, the rule here should instead be something along the lines of: “Apply w, when queried (or
when considering the matter), to any f and only to an f.” But, for ease of exposition, I’ll continue to work
with the simpler formulation.
But I think we can see that this rationale is defective and that the view it is supposed
to motivate is implausible.
For it is a mistake to conflate, on the one hand, explicitly desiring that one apply w to all
and only the dogs, and, on the other hand, implicitly following the rule: to apply w to those
things. The former (explicit) state does indeed correlate with meaning dog by w. But it is
explained by the meaning-fact rather than the other way around.19 Moreover, this
correlation gives no reason to expect that the latter (implicit) state will even so much as
correlate with meaning dog, let alone constitute it.
In addition, the implicit state clearly cannot be a purely a priori (conceptual) ground of
‘meaning dog’. For although we know a priori that our word “dog” means dog, we
can’t tell, independently of empirical research, which ideal law explains our overall use of
it and hence which rules for its use we are implicitly following. In particular, we can’t
know a priori that we are implicitly following the rule, ‘Apply “dog” to all and only dogs.’
And, as for a posteriori grounding (of the sort exemplified by ‘salt’ and ‘sodium
chloride’), attention to the normal methodology for settling such questions also points
us away from the assumption that meaning-constituting rules are typically truth-
oriented. In general, a relatively superficial property, S, is constituted by a relatively
basic property, U, just in case (i) they are co-extensive, and (ii) this co-extensivity
explains why the symptoms of S’s presence are what they are. For example, the
property of ‘being a sample of salt’ is constituted by the property of ‘being a sample
of sodium chloride’ because the symptoms of salt (e.g. its characteristic interactions
with other substances) are explained on the assumption that it coincides with sodium
chloride. But the main symptoms of a word’s meaning are its various ways of being
used—i.e. the various facts concerning the sentences containing it that are accepted and
the circumstances in which this is done. So the meaning-constituting property will be
one that explains all such acceptance phenomena; it will be the word’s basic rule of use.
So far this is a priori. But it’s an empirical discovery, often quite hard to make, which
particular rule is the explanatorily basic one for the use of any given word. And it does
not seem likely that the rule governing our use of “f ” will typically (or even ever) turn
out to be: apply it to all and only fs. For the basic rule will surely be something more
readily obeyable, and will relate rather to what we do for the sake of trying to achieve
that result (i.e. for the sake of trying to believe what is true). Thus, more plausibly, our
use of “electron” is guided by our acceptance of a certain physical-theory-formulation;
our use of “true” is governed by our deployment of an equivalence schema; and our
basic use of “water” stems from some rule along the lines of, ‘apply it to whatever has
the same underlying structure as the stuff in our rain and lakes.’20
19
Here’s the explanation. Suppose S understands “#”. In that case he (probably) explicitly desires the
truth of the proposition that he would express by “(x)[I apply ‘#’ to x $ #(x)].” (I’m pretending that “I” and
“apply” are in S’s language.) Now assume that “#” means dog—i.e. that “#” means the same as “dog.”
Then, the content of his desire is that he apply “#” to all and only dogs.
20
For further discussion of the question of how to determine which particular non-semantic property of a
word provides it meaning, see Horwich 2005, Chapter 2.
13
It is widely assumed that an account of how meaning is constituted can be adequate
only if one can explain, given some alleged meaning-constituting property, why any
word that possesses that property comes to have the particular extension that it does.
For example, if it is in virtue of a word’s possession of property, U, that it means dog
then, according to this common assumption, it must be possible to show why it is that
any word with U is true of exactly the dogs.21 And this requirement helps motivate the
presumption, just criticized, that, in general, “f ” ’s meaning-constituting rule will be
something like: Apply w to all and only fs. For such uniformly truth-oriented rules will
satisfy that requirement (as we see in footnote 23). But the more complex and more
varied meaning-constituting rules, of the kind that I have been urging, will not. How,
for example, can we get from the rule, ‘Accept “That is f ” when something seems red,’
to “f ” ’s extension being the set of red things? What explanatory deduction could take
us from the rule, ‘Tentatively accept instances of “<p> is f $ p” ’ to “f ” ’s extension
being the set of truths? On the other hand, we have just (in section 12) encountered
solid reasons for thinking that complex rules like these are more likely to be the
right ones!
How is this dilemma to be resolved? It seems to me that we should resolve it by
appreciating that the alleged adequacy condition is mistaken. One should not expect to
be able to derive a word’s extension directly from its meaning-constituting property. In
particular, one should not expect to be able to explain, on the basis of a word’s
meaning-constituting rule of use, why it is true of the things it is (rather than, say, a
slightly different set of things).
Why not? Because of the plausibility of the deflationary view of the truth-theoretic
notions, ‘true,’ ‘refers,’ and ‘true of.’ Applied to the last of these concepts—which is the
pertinent one here—deflationism is the view that “true of ” is implicitly defined
(relative to a prior understanding of instances of “f ” and of “w means F ”) by our
acceptance of the schema
w means F ! (x)(w is true of x $ x is f )22
21
Closely related to this ‘explanation requirement’ is the requirement (see Kripke 1982, p. 26) that one be
able to ‘read off ’ from each alleged meaning-constituting property what the extension would be of any word
possessing it. Insofar as this ‘reading off ’ would be made possible by a general theory that reduces ‘meaning
F ’ to ‘bearing naturalistic relation R to fs,’ then these two requirements will coincide. (For one might
explain, by reference to that general theory, why, for example, w means DOG if and only if w bears R to
dogs.) But suppose we come to suspect that meaning-constituting properties are all of a certain kind—that
they are all —-properties—where these have no such relational form. In that case, if we are given the
information that w has the very —-property that is possessed by our predicate “f,” we could nonetheless infer
(‘read off ’) that w is true of all and only fs. But we would not have explained why it is that if a word has the
same —-property as our “f ” then it is true of fs. My hunch is that the more fundamental intuition here is the
‘explanation requirement,’ and that the ‘reading off requirement’ is an imperfect expression of that intuition.
22
For the case in favor of deflationism, see Horwich 1998a.
Assuming that this is correct, then the explanatory route leading from w’s meaning-
constituting property (e.g. that “dog” has non-semantic property, U) to w’s truth-
theoretic property (e.g. that “dog” is true of exactly the dogs) proceeds via the
intermediate fact that w means what it does (e.g. that “dog” means dog). In other
words, we had no right to insist that a good analysis of (say) ‘w means dog’ provide a
direct explanation of “w is true of the dogs’ (i.e. an explanation that does not presuppose
the correctness of that analysis).
Any such direct explanation would require (i) that the non-semantic constitutor of
‘w means dog’ take the form ‘(x)(wCx $ x is a dog),’ and (ii) that C-ness either
constitute ‘being true of,’ or otherwise bring it about. Therefore, an implication of
deflationism is that we have no right to expect any such relational theory of meaning-
constitution.23
The deflationist-friendly alternative is not to deny that a word’s meaning-constitut-
ing property fixes its extension, but to recognize that it does so only because it first fixes
the word’s meaning. Thus, “dog” ’s being governed by a certain basic rule may, in
virtue of the ability of that fact to explain the word’s overall deployment, constitute its
meaning what it does—i.e. its meaning dog. And we can then invoke the deflationist’s
trivial and explanatorily-fundamental logical relation between a word’s meaning and its
extension, i.e.
w means DOG ! (x)(w is true of x $ x is a dog)

to infer (by transitivity) that the word is true of exactly the dogs.24
14
These observations on how to determine what constitutes a given meaning-property
put us in a position to address the question left pending in section 11: “Why suppose in
the first place that meaning is a matter of rule-following?”
23
Note that if we were to take the view, criticized above, that ‘S’s w means F ’ reduces to ‘S follows
“Apply w to all and only fs,’ ” then we would have a relational theory, and so the requirement in question
could be satisfied. For, supposing that S’s w is true of x is engendered by (∃g)[S follows ‘Apply w to all and only gs’
and g(x)], we can see how the proposed analysis of ‘S’s w means F ’ will entail, and hence directly explain, why
S’s w is true of all and only fs.
24
For further discussion of the explanation and reading-off requirements, and their affiliation with
inflationism about truth, see Horwich 2005, Chapter 3. See also Horwich 2010, Chapter 6.
Note that it is perhaps consistent with deflationism for us, first, to discover (via the proper methodology)
which properties constitute the meanings of a variety of words; second, to notice that, no matter what
w means, the non-semantic ground of ‘w means F ’ always entails ‘(x)(wCx $ fx)’ (where ‘C’ is constant);
and third, in light of that observation, to reduce ‘w is true of x’ to ‘wCx.’ What deflationism clearly precludes
is the advance assumption that any adequate reductive account of meaning must have that character. Thus it is
one thing for us to find to our surprise, in light of the meaning-constituting proposals that we have discovered
to be correct, that ‘explanations of extensions’ and ‘reading off ’ are possible. But it is another thing to impose
that possibility as a requirement in order to filter out ‘bad’ proposals. That would be tantamount to our
assuming a priori that ‘w is true of x’ had some reductive analysis or other.
A tempting answer might go like this. Given the above discussion, what someone
means by a word stems from whatever explains his overall use of it; and surely a good
candidate for that explainer is his following certain basic rules of use. Moreover,
meanings have distinctive normative implications (e.g. if a word means dog one had
better not apply it to cats). And it may seem that this phenomenon will fall out nicely if
we assume that meaning facts are intrinsically regulative.
But, on reflection, one can see that neither of these motivations is valid. Regarding
the explanatory consideration, if my above-suggested analysis (in section 4) of rule-
following is roughly correct then a person’s following a certain rule is grounded in two
factors: (a) that he is governed by a certain ideal law; and (b) that this takes place within
a context of occasional self-correction. But only the first of these components is needed
to explain his overall use of the word. The second component is not doing any of that
explanatory work, and so is not essential to the word’s meaning what it does.
Therefore, although we might say, speaking somewhat loosely, that someone means
what he does ‘in virtue of ’ following certain rules—for it is true that the rule-following
suffices for the meaning—a more accurate statement would be that the meaning is
constituted merely by the ideal law of use.
And, as for the second consideration, the normative import of meaning would not in
fact flow straightforwardly from the assumption that meaning is rule-following. One
problem is a certain mismatch between the pertinent rules and norms. As we have seen,
even if ‘meaning F by w’ were constituted by implicitly following a certain rule, there’s
good reason to think that the rule would not be ‘Apply w to all and only fs’; so it
remains unclear how the obligation to apply w only to fs is to be accounted for.
Another problem is that the normative import of rule-following is rather minimal;
it’s merely the value of satisfying the desire to conform with a certain regularity. But
there is an enormous gulf between this value and the significant instrumental and
epistemological value of true belief, of applying words to the members so their
extensions. So, regarding meaning as a species of rule-following does not, on reflec-
tion, help us to accommodate its most obvious form of normative import. (And in
section 15, we’ll see that there is a perfectly good alternative way of accommodating it.)
In addition, it’s worth noting that, insofar as we assume (in the absence of any better
account) that the contents of thoughts derive from the meanings of mental terms, and
insofar as we allow that these terms may turn out to be innate and universal (i.e.
symbols of Mentalese), then the idea that meaning is rule-following becomes highly
conjectural. For although our deployment of Mentalese symbols is likely to be law-
governed, the operation of these ideal laws may well owe nothing to any practices of
correction. Indeed, it may well be that ‘self-correction’ cannot occur at such a
cognitively deep level.
This latest worry about identifying meaning with rule-following goes hand-in-hand
with another. Correction is an intentional state, since it involves feeling that certain
responses are OK, disliking others, etc. Now, if these states have to be mentally
articulated, then the meanings of the terms in which they are articulated cannot, on
pain of circularity, be engendered by rule-following. So perhaps one can instead take

the states to be states of implicit desire. But it would then be necessary to give a plausible
analysis of such states in non-intentional terms. However—pace Wittgenstein’s above-
quoted PI 54—it is far from obvious that there is such a thing as ‘unmistakable self-
correction behavior.’ If there is, then the present objection to identifying meaning with
rule-following may be defective (although the others remain in force). But if not, then
we have a final reason for denying that a grounding of meaning in rule-following could
qualify as a fully non-semantic reduction.
The clear moral of these considerations is that a word’s meaning what it does is a
matter of its being governed by a certain ideal law—the one that explains its overall use.
Often, such a law will operate in a context of self-correction—in which case we can
suppose that a rule (dictating conformity with the law) is being implicitly followed. But
the operation of a word’s law of use need not be a case of ‘rule-following’ in order for it
to engender meaning.
15
What about the normativity of meaning? Can this be reconciled with what I have been
suggesting about the reduction of meaning-properties to ideal laws of use? Well, the
answer is not simply “yes” or “no”; for it depends on what one has in mind by “the
normativity of meaning.”
If all someone has in mind is that meaning-facts have certain normative implications,
then the present story is not in jeopardy. For it easily accommodates three distinct kinds
of normative implication.
The first of these concerns truth. Insofar as we think—as I believe we should—that
true belief is desirable, then we will hold, in particular, that
S means dog by w ! S ought to desire that (x)(S applies w to x ! x is a dog)25

And this will put us in a position to infer, given our view of how the meaning-fact is
constituted, that
S’s use of w is governed by ideal law R!

S ought to desire that (x)(S applies w to x $ x is a dog)
Granted, we are not able to explain the first conditional (i.e. the value of truth) by
means of the second one, since our explanatory direction is the other way around. But
I see nothing untoward about this ordering. Consider, by analogy, the relationship
between the fact that happiness is good and the fact that H-fiber firings are good
25
Nothing hinges here on whether this is really the best precise way of articulating the idea that truth is
valuable. For discussion see Horwich forthcoming.
(supposing that H-fiber firings are the neurological basis of happy feelings). Presumably
the former normative fact is what accounts for the latter. So it would be illegitimate to
object to the constitution supposition on the grounds that it does not put us in a
position to explain the value of happiness.26
A second sort of meaning-related norm concerns epistemic justification. For example:
given what we mean by “if ” and “not” we are epistemically obliged never to
simultaneously accept instances of “p,” “if p, q,” and “not q.” But—as we have just
seen—one should not think that our account of meaning (or any other account either)
must put us in a position to explain this conditional. Rather, the basic epistemically
normative implications of meaning-facts appear to be explanatorily fundamental.27
A third form of normative import that meaning possesses hinges on ‘social external-
ism.’ The ideal law of use governing a single word typically varies somewhat from person
to person within a linguistic community. Perhaps most people (or the experts within the
relevant area) are governed by R(w); but many are governed by slightly different laws.
Nonetheless, our ordinary practice of communal meaning attribution is such that all
these people are said to mean the same thing by w (they all mean F, say). Now, these
divergences in basic usage are somewhat undesirable; communication would be
smoother if they didn’t exist, if everyone’s basic law for the use of a given word coincided
with that of the majority (or that of the experts). Therefore we have the conditional
S means F by w ! S ought to be governed by R(w)

Thus, a regularist account of meaning is able to accommodate its having three
distinct kinds of normative import.
Suppose, however, that what someone has in mind by “the normativity of meaning” is
not merely that meaning-facts have normative implications, but that they are constitutionally
normative—that they are somehow grounded in normative facts. In that case my response
is to concede that this idea has not been respected here—but to insist that it needn’t be,
since we should not think that ‘meaning is normative’ in this very strong sense.28
26
I owe this example to Anna-Sara Malmgren.
Note that when a constitution thesis is advanced as a conceptual analysis then the situation is very different.
For when predicate “S” merely re-articulates the meaning of “U,” then we can indeed insist, if we accept “S
is good,” that our acceptance of it be explained by our acceptance of “U is good.” Thus, it was incumbent on
me (in section 9) to explain our attribution of normative import to rule-following by first showing that the
suggested conceptual analysans of rule-following would be given that normative import. But my proposed
reductions of meaning-properties to specific basic laws of use are to be established empirically.
If we focus instead on what we are taking to suffice, a priori, for w’s meaning F —i.e. w’s being governed
by the very law that governs our use of “f ”—then we can thereby understand the normative import of
meaning: we can see why it is that if S’s word “g” means F, then S should desire to apply “g” to all and only
fs. For suppose S’s law for “g” is the same as his law for “f.” Then he will accept “The fs are the gs.”
Therefore, since his desire-box should contain “I apply ‘g’ to all and only gs,” it should also contain (via
practical inference) “I apply ‘g’ to all and only fs.”
27
For supporting arguments see Horwich 2010, Chapter 10.
28
Philosophers who have urged the intrinsically normative nature of meaning include Boghossian 2003,
Brandom 1994, Davidson 1984, Gibbard 1994, Lance and O’Leary-Hawthorne 1997, McDowell 1984.
Certainly it would be a blunder to simply infer the constitutive normativity of

meaning from its having normative import. Many things—e.g. killing—have norma-
tive import without being constitutively normative.
Another mistake would be to try to get there by combining the assumption that
meaning is rule-following with the assumption that rule-following is constitutively
normative. For, as we argued in section 14, it is an over-simplification to suppose that
meaning is a matter of rule-following. More accurately, meaning reduces to just one
component of rule-following, namely, being governed by a certain ideal law; and this
obviously isn’t constitutively normative.29
Finally, there is a route to the constitutive normativity of meaning that has been
suggested by Paul Boghossian, amongst others. The rough idea (expressed in my own
terms) is that meaning-constituting laws of use concern the circumstances in which
sentences are accepted; but—so the argument goes—this notion of ‘acceptance’ is
constitutively normative, since a condition of someone’s possessing it is that he
endorses the norm, ‘We ought to accept only what is true.’
I don’t have space here to do full justice to this line of thought. But, in a nutshell, my
response is that there is a psychological relation we can characterize non-normatively and
non-intentionally, purely in virtue of its functional role—namely, the relation of relying
on a sentence in theoretical and practical inference. And we can proceed to explain on the basis
of that characterization why we take such ‘reliance’ to have a certain normative import,
why we take this relation to a given sentence to be one that we should wish to bear to that
sentence only if it is true. Such a demonstration provides reason to think that the ‘reliance
relation’ we have functionally defined is nothing less than acceptance (i.e. inclusion within
a person’s ‘belief box’) and that our recognition of the truth-norm applicable to it is a
mere consequence of that definition rather than integral to it.30
In sum, it seems to me that the respects in which meaning is genuinely normative—
its having normative implications relative to the value of true belief, relative to the
value of epistemic justification, and relative to the value of smooth communication—
are phenomena that the present model is able to handle perfectly well. Therefore we
have no reason at all to regard meaning as constitutionally normative.
16
I hope to have given some initial plausibility to the following explanatory picture:
(i) At the base, metaphysically speaking, are facts to the effect that a given ideal law
explains a given speaker’s overall use of a given (mental) word.
29
Moreover, the same can be said of the other component as well: namely, the practice of occasional self-
correction. This, I have argued, does have normative import; but is, in itself, a purely factual phenomenon.
30
See Boghossian 2003. See also Shah and Velleman 2005. For further critical discussion of this line of
thought, see Horwich 2005, Chapter 5. For a sketch of how ‘acceptance’ might be functionally defined in
purely factual, non-semantic terms, see Horwich 1998b, pp. 94–6.
(ii) In light of practices of self-correction, his being governed by that law consti-
tutes the fact that he is implicitly following the rule dictating conformity to it.
(iii) So his overall use of the word stems from his following that rule. Therefore,
what he means by the word (in his idiolect) is determined by that fact. But not
everything that goes into the rule-following is involved. In particular, the
practice of self-correction (needed to allow operation of the ideal law to qualify
as a case of rule-following) plays no essential role in the explanation of the
word’s overall use, and therefore no role in constituting its meaning. Thus the
meaning reduces merely to the fact that the ideal law is explanatorily sufficient.
(iv) Word meanings (so constituted) engender sentences meanings—since, when
combinatorial procedure P is applied to words, <w1, w2, . . . , wN>, the
meaning-property of the resulting complex expression is constituted by the
property: ‘__ is the result of applying P to words whose meanings are <Meaning of w1,
Meaning of w2, . . . , Meaning of wN>’31
(v) Assuming context-insensitivity (for the sake of simplicity), sentence-meanings
engender truth-conditions via the explanatorily fundamental triviality:
u means what our “p” means ! (u expresses a truth $ p)
(vi) But true belief is valuable. In other words, one should hope that the circum-
stances in which we accept a given sentence of ours include its just-explained
truth-condition. Thus the normative correctness of an assertive utterance depends
in part on the regularities of use that govern its component words.32
Bibliography
Boghossian, P. A. 2003 “The Normativity of Content,” Philosophical Issues 13/1, pp. 31–45.
Brandom, R. 1994 Making It Explicit, Cambridge, Mass., Harvard University Press.
Davidson, D. 1984 Truth and Interpretation, Oxford, Oxford University Press.
Gibbard, A. 1994 “Meaning and Normativity,” in E. Villanueva (ed.) Philosophical Issues 5: Truth
and Rationality, Atascadero, Ridgeview Publishing Company, pp. 95–115.
Gibbard, A. 2008 “Horwich on Meaning,” Mind 117/465, pp. 141–66.
Horwich, P. 1998a Truth, 2nd edn., Oxford, Oxford University Press.
Horwich, P. 1998b Meaning, Oxford, Oxford University Press.
Horwich, P. 2005 Reflections on Meaning, Oxford, Oxford University Press.
Horwich, P. 2006 “The Value of Truth,” Nous 40/2, pp. 347–60.
Horwich, P. 2010 Truth, Meaning and Reality, Oxford, Oxford University Press.
31
For details of this approach to compositionality, see Horwich 2005, Chapter 8.
32
The first draft of this essay was written for a workshop on Knowledge, Rules, and Reason, held on
November 4–6, 2005 at the Wye River Conference Center, under the auspices of the Department of
Philosophy of Johns Hopkins University. A subsequent draft was presented at a conference entitled, “Is
There Anything Wrong With Wittgenstein?,” held at the University of Reggio Emilia (Italy) in June 2006.
I would like to thank all those who participated in these discussions for their helpful questions and comments.
Horwich, P. forthcoming “Belief-Truth Norms,” in T. Chan (ed.) The Aim of Belief, Oxford,
Press.
Lance, M. and O’Leary-Hawthorne J. 1997 The Grammar of Meaning: Normativity and Semantic
Content, Cambridge, Cambridge University Press.
McDowell, J. 1984 “Wittgenstein on Following a Rule,” Synthese 58/3, pp. 325–63.
Shah, N. and Velleman, D. 2005 “Doxastic Deliberation,” The Philosophical Review 114,
pp. 497–534.
4
Why Meaning Intentions
are Degenerate
Akeel Bilgrami
1. Introductory Remarks
The relations between intentions, linguistic meaning, and normativity have been
explored with subtlety and analytical power by Crispin Wright in a number of essays
that have focused on Wittgenstein’s and Kripke’s discussion of the nature of rule-
following.
This essay will present an argument—an essentially Fregean argument—to put into
doubt a fairly widespread assumption about the normative nature of linguistic meaning
by looking at the relation that linguistic meaning bears to an agent’s linguistic inten-
tions.
I believe that there are elements in Wright’s thinking about self-knowledge of
intentionality and meaning that, to some extent, support my skepticism. But since he
has never taken an explicit position resisting the assumption of the normative nature of
linguistic meaning, I would be very curious to know where he stands on the matter and
on the particular argument owing to Fregean considerations offered here.
In several passages in his mature work where Wittgenstein discusses the nature of
intentional phenomena, focusing most particularly on intentions (as well as expecta-
tions), he is keen to distinguish viewing them as mental processes and experiences from
viewing them in terms of the intentions’ (or expectations’) fulfillment. This latter is the
idea of elements in the world (including our own actions) that are in accord with these
intentional states. Thus, just as Crispin Wright’s walking in through my front door is a
fulfillment of a certain expectation that I have (the expectation that he will come to a
reading group we have arranged to have at my place on a Friday morning), so is my act
of taking an umbrella a fulfillment of my intention to do so on a rainy morning. Both
are described as being in ‘accord’ with the intentional states in question.
The terms ‘fulfillment’ and ‘accord’ convey something that is describable as ‘norma-
tive’ in a very broad sense of the term. Things are ‘right’ in some sense when there is
accord and fulfillment of this kind, wrong if there is not. Such is the minimal
W H Y M E A N I N G I N T E N T I O N S A R E D E G E N E R AT E 97
normativity of intentional states. Sticking with ‘intentions,’ which will be the particular
intentional state that is the focus of my essay, if I were to intend to take an umbrella but
took a walking stick instead of an umbrella by mistake, then it would be, well,
‘a mistake’ by these broadly conceived normative lights. So Wittgenstein’s view (not
explicitly made in these terms, but implicitly very much part of his picture of inten-
tionality in his mature work) is that the very idea of intention is such that it generates an
ideal or norm of correctness, something by the lights of which one can assess one’s
actions for being correct or wrong, depending on whether they are or fail to be in
accord with the intention.
What is the philosophical force behind such talk of the normativity of intentional
states? Its force is contrastive: not merely a contrast with the apparently processual and
experiential aspects of mentality just mentioned, but also with what Kripke brought to
center stage in his book on Wittgenstein, the dispositional character of mental states.
Put most generally, the contrasts are asserted with anti-psychologistic ends in mind: the
normative is set against the psychologism of process and of inner experiences as well as
of mental tendencies and propensities. Since these contrasts are well known in the
discussion of these topics, I will not labor them here beyond saying that normativity, so
conceived, is said to be constitutive of intentional states, and if that is so, it puts into
doubt that the processual, the inner experiential, and the dispositional, can really be
what is primary in our philosophical understanding of intentionality.
There is no gainsaying the centrality of such a normative element in the very idea of
intentions, in particular, and intentionality, in general. What I want to question is
whether what is true as a general point is true in the case of linguistic intentions, in
particular the intentions that speakers have regarding the meanings of their words.
Might these not be a very special kind of exception to the generality of this truth,
providing a sort of limiting or degenerate case of intention and intentionality?
Here is how I have allowed myself to think of it.
2. Getting Meaning Intentions Right

What are the intentions one has when one says things or means things with one’s words
(restricting ourselves to assertoric statements for the sake of simplicity and conve-
nience)? Since Grice’s analysis1 (I should say ‘analyses’ since he fortified his initial
analysis in subsequent work2) of meaning, which linked meaning with intention
explicitly and elaborately, is so canonical, let us take that as a point of departure.
The initial part of his analysis points out that when we say things we have certain
nested intentions to have some effect on hearers. In the assertoric case, the intention is
to get them to acquire certain beliefs—say, about the world in one’s near vicinity. Thus
for instance, someone says “That is a snake” with the intention to get someone else to
1 2
Grice 1957. See for instance, Grice 1969.
98 AKEEL BILGRAMI
believe that there is a snake in one’s path. (In Grice’s analysis this intention, as I said,
nests with two others—at least3—whose point is to ensure that the case is a case of
getting someone to believe something by telling someone something rather than
merely getting it across to them, something that could not be ensured with just that
one intention. What prompts these other two nesting intentions that go into the first
part of the analysis are not relevant to the concerns of this essay.)
But, in Grice, this analysis invoking these three nested intentions is supposed to be just
the beginning of an analysis of meaning. One has to add various things to move from an
account of speaker’s meaning, which this analysis provides, to an account of sentence
meaning. The speaker’s meaning of the words uttered is analyzed in terms of the specific
purpose or intention that the speaker has on that occasion (in the canonical assertoric cases,
to get someone to believe something). The sentence meaning is the meaning of the words
that the speaker takes his words to have—in Grice’s rhetoric—‘timelessly.’ This contrast
between what the analysis provides in this first stage with the three nested intentions
(i.e. speaker’s meaning) and sentence meaning is most explicitly visible or audible when
they fail to coincide even on the surface, as for instance in metaphors or in indirect speech
acts. In a metaphor, one might say some words, such as the familiar “Man is a wolf ” with
the intention of getting someone to believe that “Human beings are competitive”; in
indirect speech acts one might say some words, such as “The train is about to leave,” with
the intention to get someone to believe that they ought to walk faster and get on the train.
The three intentions of Grice’s analysis do not provide the analysis of the sentence
meaning, only of what the speaker meant to convey to the hearer on that occasion
with the utterance of those words. The speaker does not take the respective sentences to
mean that human beings are competitive or that someone ought to walk faster. He does
intend to get the hearer to believe that human beings are competitive in the one case and
that he ought to walk faster in the other, but that is merely speaker’s meaning; what he
takes the sentences he utters to mean is something quite else.
Grice gave additional analysis of the sentence meaning that the utterance possesses
and initially seemed to have some hope that one could build up to sentence meaning
on the basis of the intentions that go into the analysis of speaker’s meaning, with as few
extraneous elements as possible. Thus, for instance, one might think that sentence
meaning might be built up out of speaker’s meaning by saying that it is given in terms of
the intentions that speakers usually have on given occasions of utterance of that
sentence. Later there was some suggestion (by Jonathan Bennett, for instance)4 that
Lewis’s work5 on convention might need to be brought in to go from speaker’s meaning
3
In subsequent commentary on Grice as a result of a point made in a seminal article by P. F. Strawson in
his “Intention and Convention in Speech Acts,” in Strawson 1971. Further intentions began to be added to
Grice’s analysis to patch up what were seen as counter-examples. Schiffer 1972 is a thorough and careful
elaboration of the Gricean project on meaning and takes Strawson’s patching-up operation towards what
seems like a definitive conclusion.
4
Bennett 1973.
5
Lewis 1969.
to sentence meaning since the statistical ideal of ‘usual’ was too contingent and
unprincipled to achieve the analysis. I will not pause to expound here these various
struggles among Griceans towards such further analysis since it is not Gricean exegesis
that I am primarily interested in. Suffice it to say that in a careful commentary on Grice,
Stephen Schiffer6 came to what was then widely considered to be a sensible conclu-
sion: that Grice needs to bring in something like a truth-conditional analysis of the
sentence meaning—‘timeless meaning’—that the speaker takes his words to have, over
and above what he means on that occasion with the utterance of that sentence. Since
truth-conditional analyses of sentence meaning are very familiar by now, let me for the
sake of convenience assume that it is they rather than some other analysis that will be
the best account of sentence meaning. (If someone were to doubt Schiffer’s claim and
give some other analysis of sentence meaning, that should not spoil what I want to say
here, since all I want to say, is that even in Grice there is a distinction between speaker’s
meaning given in his initial analysis with those three nested intentions, and sentence
meaning. Which analysis best accounts for the latter makes no difference to my
purposes.) The chief point that needs to be made is that in my examples, the truth-
conditions of the sentences by no means coincide with what the initial Gricean analysis
of the speaker’s meaning, yields.
This point is well known; still, it is worth being explicit about it. It would be quite
wrong to say that the speaker has in mind that “ ‘Man is a wolf ’ is true if and only if
human beings are competitive” or “ ‘The train is about to leave’ is true if and only if the
hearer ought to walk faster and get on the train.” Rather, he takes it to be the case that
“‘Man is a wolf ’ is true if and only if man is a wolf ” and “ ‘The train is about to leave’ is
true if and only if the train is about to leave.” These are his sentence meanings and they
depart on the visible surface, in these examples, from the speaker’s meaning. And the
important point remains that even in cases where there is no visible departure of this
obvious kind as there is in metaphors or indirect speech acts, one should nevertheless
acknowledge the difference between speaker’s meaning and sentence meaning.
If someone were to say “Human beings are competitive” with the intention to get
someone to believe that human beings are competitive that would still leave intact the
distinction between speaker’s meaning and sentence meaning since the latter would be
given by the truth-conditions of the sentence, not the intention to get someone to
believe something that happens to coincide (in this but not other cases) with what is
specified by the truth-conditions of the sentence.
Though, as I said, that point is well known, there is a source of possible confusion
here against which we should protect ourselves because it is crucial to a point that will
come later. I, following Grice and others, have said that when a speaker says something,
the sentence meaning is something (relatively) independent of the intentions he has
which are emphasized in Grice’s initial analysis, because the initial analysis is only of
6
Schiffer, 1972, Chapter 6.
100 AKEEL BILGRAMI
speaker’s meaning, of what he means on that occasion. This may give the quite wrong
impression that sentence meaning is not to be thought of as something that he means at
all, that it attaches to the words he utters but are not meant by him, in any sense. But it is
indeed he (the speaker) who also takes the sentence he utters to have a sentence meaning
over and above what he intends someone to believe with the utterance of that
sentence . . . The speaker is not left out of the picture in sentence meaning. Just because
sentence meaning is contrasted with speaker’s meaning, it doesn’t follow that it is not
speakers who take their utterances to have sentence meaning. It is not as if the sentences
uttered by speakers possess meaning in some ulterior way and the speakers who speak
them don’t take them to have that meaning. (Grice’s rhetoric of ‘timeless’ as opposed
to ‘occasional’ meaning—clearly echoing the ‘sentence’/‘speaker’ meaning distinc-
tion—may also mislead in the same way and that too should be guarded against. Just
because so-called ‘timeless’ meaning is contrasted with what a speaker means on an
occasion, it doesn’t mean that it is not the speaker on that occasion who takes it to have
that timeless meaning.)
Let us now return to the question of the normativity of intentional states as laid out
in Wittgenstein’s characterization of them, in particular his normative characterization
of intentions. Our question, as I said, is the relation between the normative nature of
intentions and the normative nature of meaning. More specifically, if, as Grice shows,
intentions are deeply involved in meaning, what I want to explore is the extent to
which the normative nature of intentions imparts, or is of a piece with, the alleged
normativity of meaning.
What is often said in the literature on this subject is this. Our terms (at any rate many
of them) apply to things, and to misapply them is to make a mistake with our terms;
and the very possibility of such mistakes amounts to the normativity built into the meanings
of our terms. Thus we are right when we apply the term ‘snake’ to snakes but not to
any other thing. When related to our intentional utterances of sentences with these
terms, such a view of the normativity of meaning amounts, then, to something like
this. We intend to say things with the words we utter. Thus—staying, as we have, with
assertoric utterances—one might utter, “That is a snake” with the intention of applying
those words to a snake in one’s path. Now, should it turn out that what is in front of us
is, say, a rope and not a snake, we would have gotten things wrong; and that possibility
of getting things wrong redeems in the particular case of meaning things with one’s
words, Wittgenstein’s general idea (true of all intentions whether they are in play in
meaning or in anything else) that intentions are, in their essence, normative. Such
intentions as the one just mentioned with which one utters words such as the ones just
cited, are just examples of intentions targeting specifically, not actions such as taking
one’s umbrella but rather linguistic actions. Just as one might make a mistake and not
take one’s umbrella (taking a walking stick instead), so also one might make a mistake
and say “That is a snake” in the presence of a rope. In both cases, it is the possibility of
such mistakes that reveals the intrinsic normativity of intentions, but in the second case
in particular this normativity of intentions captures for meaning (a notion, we have
acknowledged to be undeniably tied to intentions) via the intentions with which

words are uttered, a special case of this same normativity.
Thus the normativity of the meaning of terms that comes from the idea of the
correct or incorrect application of our terms passes over into the normativity of the
intentions with which we use our terms in utterances. We act in accord with these intentions
to use words, the intention, say, to use the words “That is a snake” to apply to a snake,
only when we do so in the presence of snakes, not in the presence of anything else.
That it should pass over in this way might be said to be a very important claim in
Wittgenstein because unlike Platonic conceptions of the normativity of meaning,
shunned by him, this sort of normativity does not owe to abstractions such as Plato’s
‘forms’ or ‘ideas’ but merely to the intentions with which words are used by linguistic
agents. Misapplication of a term is not the violation of a norm because it falls foul of
some abstracted ideal (the CONCEPT snake) but because terms are used in utterances
with intentions and we can act in ways that fail to be in accord with those intentions.
That is what is often said in the philosophical literature.7 But there is very good
reason to doubt that this picture of the normativity of meaning gets things right. Even a
cursory look at what we have been saying about Grice should reveal what the grounds
of doubt are, but before I relate it to Grice, let me say what seems obviously wrong
with such a view of the normativity of meaning. What it gets wrong is the intention
that is relevant to meaning. The intention it identifies as being relevant to meaning is in
fact relevant to something quite else. The intention relevant to meaning, when one
makes assertoric utterances such as the one in the example we are discussing, is not
(1) “One utters the words ‘That is a snake’ with the intention of applying them to a
snake in one’s path.” Rather it is (2) “One utters the words ‘That is a snake’ with the
intention of saying something which is true if and only if that is a snake—or true if and
only if there is a snake in one’s path.” (Once again, I mention truth-conditions in
(2) for the sake of mere convenience since it is the most widely held analysis of sentence
meaning. If someone had another view of sentence meaning than a truth-conditional
one—say, one invoking verification or assertibility conditions—one could reformulate
(2) accordingly.)
Returning now to normativity: we have said, surely uncontroversially, that the
possibility of getting it wrong is a necessary condition for normativity, in this (or any other)
matter. And in linguistic examples (of assertoric utterances in particular) that possibility
was supposed to be actualized in cases of the misapplication of terms, cases such as when
one says “That is a snake” in the presence of, say, a rope. So let us suppose that one does
say those words when there is no snake but a rope in front of one. If one assumes that it
is intentions of the form (1) which are relevant to meaning in assertoric utterances, then
one is indeed making a mistake. But if one assumes that it is intentions of the form (2)
which are relevant to meaning, then no mistake is being made (about meaning) at all
7
For example this idea of normativity via misapplication is represented as the standard view on the matter
in an article intended as a survey of the recent literature on the subject of rule-following in Boghossian 1989.
102 AKEEL BILGRAMI
when one utters those words in those circumstances. Even if a rope rather than a snake
is present, one’s intention to say something with certain truth-conditions (something
which is true if and only if there is a snake there) is an intention that is impeccably met
in these circumstances. The fact that there is a rope and not a snake, which is present in
the vicinity, does not affect in the slightest the aptness of that intention about meaning.8
Thus it is only by misidentifying the intention relevant to meaning that one is led to
think that examples such as these are revealing of the normativity of meaning because
they have revealed the built-in possibility of mistakes. These are not mistakes of
meaning. They are quite other sorts of mistakes, mistakes due to misperceptions, as
in this particular example—in other examples they may be due to quite other causes.
It won’t help to say that the idea of mistakes of meaning has to do with the
misapplication of terms, so one must find intentions targeting the terms in the sentence
uttered and show that those too are fulfilled when there are ropes rather than snakes
present. It is true that I have only focused on intentions that target the utterances of
whole sentences, but the analysis carries over perfectly to intentions that target the
terms that compose uttered sentences as well, assuming for the moment that we do
have such intentions. Suppose when I utter, “That is a snake,” I have a meaning
intention that targets, just the word ‘snake.’ What shall we identify as the meaning
intention for that term? Should it be, “I intend to apply ‘snake’ to a snake” or should it
be “I intend to utter a term, a predicate, that is satisfied if and only if there is a snake
there.” I claim that it is the latter intention that is properly thought of as a meaning
intention. And one has acted in accord with it, even if there is a rope in front of one. It
is only the former intention that one has failed to act in accord with, in that circum-
stance. Misapplication of terms is beside the point (or beside the primary point) as far as
meaning intentions are concerned, whether the intentions target the utterance of
whole sentences or the terms that compose those sentences.
What I have said about getting the intentions relevant to meaning correctly identi-
fied can be related in detail to the exposition of Grice I presented earlier. In order to
keep the main line of argument of the entire essay uncluttered and undistracted by
detail, however, I will elaborate these relations to Grice in Appendix I, which picks up
from just this point where I leave the matter now.
If the argument so far is convincing, the deep issue then becomes: now that we have
properly identified the intentions relevant to meaning, what follows about the norma-
tivity of meaning? I had said, surely uncontroversially, that it is the possibility of mistakes
(in the case of meaning, it would be, what I called, ‘failures of accord’ with one’s
meaning intentions) in which the general idea of normativity is revealed. We must
then ask: if the possibility of such things as mistaking ropes for snakes does not amount
to the relevant kind of normativity-revealing failures of accord with one’s meaning
intentions for assertoric utterances such as “That is a snake,” what sort of thing does
8
I had developed this point at length in a paper written quite some years ago in Bilgrami 1993.
amount to it? I pose this question in just this way in order to invite the following
suspicion: can anything amount to a failure to act in accord with the intention we have
now properly identified as being relevant to the meaning of utterances of that kind?
That is, is there so much as a possibility of being mistaken about meaning? If, as the
suspicion is supposed to suggest, the answer to this question is “No,” then one puts into
doubt the idea that meaning is normative, at least to the extent that such normativity is
supposed to derive from the (undeniable) normativity of intentionality, in general, and
of intentions, in particular. I think it is arguable that the answer to this question is ‘No.’
3. Can One Fail to Fulfill Meaning Intentions?

Once properly identified, we have learnt that the intention relevant to meaning targets
the truth-conditions of one’s words. Hence the failure to fulfill that intention would
presumably occur only if one failed to get right what their truth-conditions are—as
opposed to occurring when the truth-conditions, which one gets right, fail to hold (in
our example, when there is no snake but a rope in front of one).
How, then, might one fail to get right what the truth-conditions of the sentences
one utters, are? One clear sense in which it might be supposed that one can fail to get
them right—or better, one clear source for one’s failing to get them right—is if one does
not know what they are. (There is another supposed source, which I will take up a little
later.)
The idea here will have to be that there are cases in which, because I don’t know
what the truth-conditions of my words are, when I intend that they have certain truth-
conditions, they are not the correct truth-conditions of those words. So a question
arises: why should the truth-conditions of one’s words not always be what one intends
them to be? We will return to this question at the end. But first let’s ask: how exactly is
it that one can intend our words to possess truth-conditions they don’t in fact possess as
a result of one not knowing what the truth-conditions are? Let’s take an example, a
familiar one from the philosophical literature, of such an occurrence.9 A medical
ignoramus intends to say some words that are true if and only if he has a disease either
of the joints or ligaments in his thigh. And he says, “I have arthritis in my thigh.” He
doesn’t know that arthritis is a disease of the joints only. So he has said something with
9
The example is from Burge 1978. Burge’s is a highly fortified and carefully presented and interesting
version of a position that was first argued by Michael Dummett in Dummett 1978. The position stresses the
social constitution of an individual’s meanings and therefore of his thoughts. It is the position that I passingly
refer to in this paper as a ‘secularized’ version of Platonistic ideals of meaning. My own opposition to this
position is much influenced by Donald Davidson’s later views on meaning, though the considerations
presented here via the Fregean puzzle about identity are intended to provide an argument for this opposition,
an argument that is not found in Davidson, partly at least because he too was highly suspicious of Fregean
notions of sense for reasons that are very similar to the ones I partially attribute to Kripke in the text. I also
believe—as I argue both in Bilgrami 1992 and more decisively in Bilgrami 2000—that Davidson’s views on
the nature of truth are quite incompatible with those aspects of his thinking that have influenced me and it is a
real question whether Davidson really understood the implications of some of his later views on meaning.
104 AKEEL BILGRAMI
truth-conditions other than the truth-conditions he intended. He has failed to act in

accord with his intention. This looks like an example of how, when one does not
know what the words one intends to utter mean, one can say something that fails to
live up to an intention that (unlike the intention in the example about snakes and
ropes) is properly identified as being relevant to meaning.
The crucial task now is to assess this claim that one may not know the meanings of
the words one intends to speak. Here I do not think the primary question to be asked
is: what theoretical account of meaning allows it to be the case that a speaker does not
know what he means? It is easy to devise a number of such accounts and they have
been devised ever since Plato’s highly objectivized notions of meaning understood as
given in a heavenly world of ‘forms’ or ‘ideas,’ with contemporary versions bringing
Plato’s heaven down to earth and calling it ‘society’ or ‘community.’ Any assessment of
the claim needs instead to step back and ask the prior question whether we can tolerate
any theoretical account of meaning in which we breezily allow speakers to fail to know
the meanings or truth-conditions of their own intended words, and that, in turn,
means stepping even further back to ask: by what criteria shall we decide what is and is
not tolerable in a theoretical account of meaning thought of in terms of truth-
conditions?
Responsible stepping back of this sort requires one to at least notice the historical fact
that the idea that the meaning of a sentence is given by its truth-conditions was first
explicitly formulated in Frege’s doctrine of sense and so it is perhaps in the notion of
sense that we should seek the criteria by which we can assess what seems tolerable or
not in an account of meaning.10 What we will or will not tolerate will depend,
therefore, on stating what the notion of sense was introduced to do and see whether
it will do what it was introduced to do, if we allow that one may not know the senses or
meanings of one’s words. So let’s ask: what is a notion of sense (or meaning) for?
In Frege, as we know, the notion is introduced initially to solve a puzzle about
identity. Though that is the occasion in Frege for theoretically motivating the notion of
sense, he had in mind very large issues in raising the puzzle about identity—the puzzle
is a mere surface reflection of one of the most deep and fundamental issues about the
relations between language and mind. In fact, Frege’s own formulations of the puzzle
and his solution to the puzzle don’t always make explicit just how deep and funda-
mental the issue at stake is. One way of putting the point at stake is to describe it as
follows: to be misinformed or uninformed is not to be irrational. No account of the
mind can confuse these two ways in which a mind can go wrong. Failures of empirical
knowledge and logical error are not to be conflated. The puzzle arises precisely because
10
If one were to replace truth-conditions with assertibility-conditions in this sentence, that would not
spoil the Fregean provenance that I am stressing. After all Michael Dummett precisely was elaborating
Fregean sense in his many works, when he argued that we ought to replace a theory of meaning in terms of
truth-conditions with a theory of meaning in terms of assertibility-conditions. See Dummett 1976. For Frege
himself, see Frege 1884, 1892.
the examples discussed by Frege (and by Kripke, who raises a slightly different version
of it) threaten to conflate them.11 The protagonist in the puzzle who, ex hypothesi,
merely lacks worldly knowledge of the identity of a planet (or in Kripke’s version, a
city) is threatened with a charge of irrationality by precisely such an elementary
conflation. And it is Frege’s view (though not Kripke’s) that introducing the notion
of sense will provide the best solution to the puzzle. It is the notion of sense or meaning
which makes it clear that no irrationality, no inconsistency, is entailed by someone
saying, for example, that “Hesperus is bright” and “Phosphorus is not bright” or
“Londres et jolie” and “London is not pretty” . . . So the puzzle lays down a crucial
desideratum: we know the protagonist in the puzzle to be someone who merely lacks
knowledge of an a posteriori identity, so we must find a way to characterize his
mentality (or this fragment of his mentality) as representing a completely consistent
state of affairs. Since it is the positing of senses to his words (over and above their
reference) which, according to Frege, helps us achieve such a representation, nothing
should be tolerated in our understanding of senses that prevents them from decisively
carrying out this task. In other words, nothing should be tolerated in the understanding
of the notion of sense or meaning, which will prevent senses from doing what they are
supposed to do: solve the puzzle and, by doing so, maintain the most fundamental of
philosophical distinctions—that between logical error or irrationality and lack of
empirical knowledge.
The fact is that senses will not decisively solve the Frege style puzzles if it is allowed
that we fail to know our own senses or meanings. A failure of transparency in sense will
leave it entirely possible that the puzzles about identity can arise one level up and so the
puzzles will not be satisfactorily solved; or better, they will not once and for all be
arrested. Let me explain.
If someone does not know his own senses, he may be in a position to be just as
confused as Frege’s protagonist in the puzzle, thinking that there are two senses rather
than one. Suppose someone wonders, in his ignorance of astronomy: “I wonder if
Hesperus is Phosphorus.” To make such a wondering so much as intelligible, a Fregean
posits senses. But if the wonderer doesn’t know his own senses, he may similarly
wonder, one step up, if the sense of ‘Hesperus’ is the same as the sense of ‘Phosphorus’
(or as Benson Mates pointed out in an ever so slightly different context of discussion, he
may wonder whether—or doubt that—the sense of ‘bachelor’ is the sense of ‘unmar-
ried man’).12 Thus, there is no genuine arrest of the Frege puzzle (and no eventual
solution to it, therefore) if it is not laid down as a basic criterion of senses that they be
11
I have said Frege himself leaves inexplicit this way of raising the puzzle because he tends to raise it via
considerations of cognitive content rather than rationality, but it is obvious to the reader that issues of
rationality are not far from the surface. It is Kripke who makes things more explicit when he raises the puzzle as
a puzzle about belief and not merely about identity and reference. See Kripke 1976.
12
Mates 1952. Mates, in allowing this kind of wondering or doubt, clearly takes senses to be the sort of
thing that one can get confused about in the way I am denying. He was an early figure in shaping a view of
senses that is very different from what I say about them in the next paragraph.
106 AKEEL BILGRAMI
transparent, i.e. known to their possessors. Without this being laid down, the puzzle
can always arise one step up, with protagonists as confused about the identity of their
senses as they are about planets and cities.
One implication of this—and a very deep one—is that it amounts to something like
a proof that senses are not the sorts of things that we can have multiple perspectives on
such that one can get their identities confused in the way that we can with planets and
cities. Whatever senses are, then, they are not the kind of things that planets and cities
are. They are not like any thing which allows multiple perspectives on itself and which
therefore allows such confusion to be possible. Things on which we can have more
than one perspective are by their nature not transparent, even if they are often in fact
known. I suspect that it is, at least partly, because Kripke doesn’t quite see this point
about the sort of thing senses are that he doesn’t follow Frege in invoking senses when
he addresses the puzzle.
Those, then, are the considerations that make it intolerable for meanings to not be
known by those who speak meaningful words: we will not be guaranteed to solve the
Fregean puzzles, at least not in a way that arrests them once and for all; and that, in turn,
amounts to meanings failing to do the very thing that meanings and senses were
introduced to do, viz., allowing one to preserve a fundamental distinction of philoso-
phy between logical error and lack of empirical knowledge. We should therefore
regard with suspicion the many accounts of meaning from Plato’s down to the secular
and mundane versions of Plato in our own time, which allow such an intolerable
outcome as a result of prising apart our meanings from our knowledge of them.
The medically ignorant man who says “I have arthritis in my thigh,” therefore,
though he certainly makes a mistake, makes a mistake about how the term is used in
the social linguistic practice, especially among the medically informed experts. His own
linguistic practice is not grooving with theirs. That is his only mistake, apart from the,
ex hypothesi, medical ignorance. He makes no mistake of failing to act (speak) in accord with his
meaning intentions. The words on his lips are intended by him to mean something that is
true if and only if he has a disease of the joints or ligaments in his thigh, he says and
thinks something that is both self-known to him and something that is perfectly true.
After I set up more conceptual apparatus, I will say more about how to represent this
idea of his own individual, idiosyncratic practice (see particularly the long footnote 26).
But until then, le me briefly address two immediate false impressions that the very idea
of such idiosyncrasy in meaning often prompts.
First: it may seem that allowing such idiosyncrasy in our understanding of linguistic
meaning would be in some way to undermine the stability of linguistic practice, the
regularity of usage that makes interpretation of one another possible. It is these
regularities that are captured in the norms of meaning, which the views I am opposing
insist on (presumably norms such as ‘arthritis’ ought to be used to refer to a disease of the
joints only, etc.), norms which someone may fail to know and therefore fail to know
his own meanings when he uses a word like ‘arthritis,’ as in the example above.
Without such norms, there would be no stability in linguistic practice, it will be said,
so I have only managed to say that meaning intentions are fulfilled in these cases (and
meanings and truth-conditions are self-known by speakers in such cases) by a theoreti-
cal move that destabilizes linguistic practice. I will say more about the point and
rationale for norms of meaning towards the end of the essay—for now notice a quite
straightforward confusion in this anxiety. It is undoubtedly true that there would have
to be a fair amount of regularity in the use of words, for speakers to be intelligible to
each other and interpretable by one another. But regularity in usage does not amount
to norms. Norms say such things as “Use ‘arthritis’ in such and such a way.” One may
violate any such so-called norm of meaning attaching to any word and be understood by
others, even if, when such a violation occurs with any such word, it puts others to a bit
of strain before they understand one. This happens constantly in communication and
the utterance of the medical ignoramus just discussed is only one such example. That,
in itself, is quite sufficient to show that understanding does not require norms of
meaning. So if there are norms of meaning, they don’t have anything to do with any
notion of understanding. They are not constitutive of any notion of meaning that is
captured in such expressions as “I understand the meaning of what you just said.”
Perhaps they are constitutive of some other notion of meaning. Whatever that notion
of meaning is, it does not seem necessary to capture the stability of linguistic practice
and the intelligibility and interpretability of one speaker by another. Neither, therefore,
are norms necessary to such stability. It is true that if there was widespread idiosyncrasy
in usage and no or virtually no regularity, there would be no communication and
mutual understanding possible, but to say that is to say nothing about there being
norms of meaning. All that it says is that for understanding each other’s utterances to be
possible, there must be some general and background regularity in the way we use words.
Given this general, background regularity, any particular word may be idiosyncratically
used and be understood.
Second: there is a common impression that an admission of idiosyncrasy in meaning
fails to keep faith with facts about how individuals show deference to the social,
especially the experts in their society. By allowing for one’s meaning intentions to be
fulfilled in the sorts of cases that I am considering, it will be said that I disregard the
ubiquitous fact of deference to the expert’s in one’s society, as soon as one realizes the
shortcomings in one’s knowledge and therefore sometimes in one’s speech. Such
deference suggests that our meanings are in the first place constituted by what experts
think and how they speak and if so, then in these cases the right thing to say is that the
protagonist in the example above does not know what his own meanings are, in his
pre-deference speech, thereby allowing the source of difficulty for the exceptionless
fulfillment of one’s meaning intentions that I am trying to block. There is no need for
the view I am presenting to deny that this medically ignorant protagonist, as he
becomes more knowledgeable, will most likely13 defer to the experts’ linguistic
13
I say ‘most likely’ quite deliberately. There can be contexts in which someone may not change his use
of words if he thought that people around him had become used to his idiosyncratic usage and that it would
108 AKEEL BILGRAMI
usage and wishes to defer to them. Deference is perfectly compatible with the view. All
that deference amounts to on this view is that he will change his linguistic behavior and
adopt theirs. He will start speaking as they do. So understood, deference need not ever
be seen as evidence for the claim that he came to know more about what he himself means
and thinks. He always knew what he meant and thought. (This is not to suggest that
people can never fail to know what they think. But it is to suggest that they when they
fail to know what they think, it is generally for psychological reasons—roughly of the sort
Freud studied having to do with repression, self-deception, etc.—and not because
certain Professors of Philosophy have devised secularized Platonist theories about the
social constitution of reference and meaning.)14 He has only learnt something about
what his fellows, especially the experts, think and how they use words and because he
wishes to defer to the experts he will now start using words as they do. Of course,
people learn more than something about the use of words, they also become more (in
this case, medically) knowledgeable about the world (about diseases, etc.) To gain some
medical knowledge that one hitherto lacked is one thing. To gain knowledge of one’s
own meanings which one hitherto lacked is another. There is never any reason to say
that the former kind of acquisition of knowledge implies that the latter kind of
knowledge is acquired. That is because there is never any reason to say that lack of
the former kind of knowledge implies the lack of the latter sort of knowledge. When
one gains medical knowledge one learns things about the world (of diseases, etc.) but
one also partly learns how one’s fellows, in particular the experts, the doctors, use
certain words. That doesn’t mean that one has to say that one also learns more about
the words one oneself used in the past. (It is in general quite odd, a violation of
common sense, to think that someone, our protagonist, say, has to gain knowledge of
medicine in order to gain self-knowledge.) If hitherto, in one’s ignorance one had used
certain words idiosyncratically, one then, on gaining the knowledge, usually defers to
others in one’s own subsequent use of those words. Deference may therefore be
now put them to strain if he changed his usage. If this happens, it is proof that convenience and utilitarian
considerations of that kind are more important in governing usage than norms of meaning are. That is to say,
we defer in order to be understood without strain by others, and in unusual contexts of the sort I have just
mentioned, we may also not defer if convenience and utility dictates that one shouldn’t. The common
element to two opposing responses (both deference and non-deference), then, is utility. If so, deference does
not support the idea of normativity, except in the low-class sense of utility. Certainly not in any constitutive
or intrinsic sense of normativity. I say something more about this at the very end of the essay and in footnote
18 below.
14
What I say in the parenthesis should make it clear that it would be a crude and elementary misreading of
my insistence that theories of reference and meaning should not allow one to say that one might not know
the truth-conditions of one’s words (and therefore the contents of one’s thoughts) to say that it is a Cartesian
position of infallibility. See Travis 1995 for an example of such a crude and elementary misunderstanding.
Nothing in my insistence denies that there might be the most ordinary forms of self-deception or the more
elaborate forms of repressions and self-censor mechanisms that Freud studied which thwart self-knowledge of
our states of mind. I say more about this subject in my Bilgrami 2006 but before that too in Bilgrami 2003 and
Bilgrami 1998. In the latter article, there is also much discussion of the issues in this essay, in particular on the
question of whether Fodor’s efforts to deal with the puzzle about identity can escape some of the issues
presented in this essay.
admitted, without providing any reason to deny that our protagonist knew what he
said and meant when he said “I have arthritis in my thigh.”
All of these points about transparency, meaning, norm, deference, etc., follow as
straightforward theoretical implications of meeting the desideratum that we must have
a decisive solution to the Frege-style puzzles about identity, where by ‘decisive’ I mean
a solution that arrests these puzzles. I have given this Fregean argument for the
transparency of meaning or sense to block one alleged source for the (normativity-
revealing) possibility of failure to act in accord with the intention to say something with
certain truth-conditions. It had been claimed that we can fail to act on such an
intention if we do not know what the truth-conditions of our words are and it is
this last claim that the considerations about the Frege puzzle have shown is intolerable
for any account of meaning to allow.
But I had also said that that it is not the only supposed source.
If one has knowledge of the right truth-conditions for one’s intended words, can one
still get the truth-conditions of one’s spoken words wrong? How might one intend to
say something with certain (correct) truth-conditions but say something with some
other (incorrect) truth-conditions or with no truth-conditions? This can happen only if
one misspeaks, if the sounds one produces do not amount to the words one intends to
utter—as for instance in slips of the tongue. So, suppose I say, “I am going towndown”
with the intention of saying something that has the truth-conditions (something that is
true if and only if ) I am going downtown. The sounds I make do not, it might be said,
amount to words that in fact have those truth-conditions. (In this particular example,
they don’t seem to have any truth-conditions.) Misspeaking, then, is the second alleged
source for failing to live up to our intentions that target truth-conditions and thereby
falling foul of the alleged normativity of meaning.
Is this the best way to analyze such cases of misspeaking—to see them as giving rise to
such failures? The issues here are, at bottom, not really different from those already
discussed in the cases where the apparent source of the difficulty was an apparent failure
to know the meanings of the words one intends to speak. In the present cases, one
knows the meanings or truth-conditions of the words one intends to speak but not of
the words one actually ends up (mis)speaking. But the question is why should the
words we actually speak fail to have the truth-conditions we intend them to have, even
in these cases of misspeaking?
Once again: is an account of meaning which allows such a failure tolerable? We
would only allow it if we were in thrall to accounts of meaning that allow for the
possibility of the meanings of our words, the words we speak on given occasions with
intentions, to be such that we are not aware of what they mean, at the time we utter
them. It is only if it is said that I am not aware of the meanings of the words I misspeak
that my misspeaking could be seen as a sign that I have uttered something with
different truth-conditions than the one I intended or, in the particular example
I mentioned (“I am going towndown”), something with no truth-conditions at all.
But why shouldn’t misspeakings of this kind get a quite different theoretical treatment,
110 AKEEL BILGRAMI
one in which they have just the truth-conditions I intend for them, in which case I am
perfectly aware of what my words mean? On this view, the misspeaking is not a case of
meaning something one doesn’t intend, only one of producing sounds one didn’t intend
to produce. But those sounds mean just what I intended to mean by the sounds that
I intended to make (but didn’t); so I can’t fail to know them. One would have thought
this theoretical account keeps better faith with the idea of ‘misspeaking’ because there is
nothing amiss with the meaning at all. The alternative view, which I am opposing and
which has it that I ended up meaning something I didn’t intend might be capturing
something that is better termed ‘mismeaning’. But it is misspeaking we want to capture,
not mismeaning.
There will be a protest: You unfairly foist on the alternative view an attribution of
meaning to the misspeaker’s utterance that is his meaning. But it is not his meaning, it is
what the words he utters mean. This move does not help matters at all. The protest, if it
is correct, only reveals the true nature of the view I am opposing. It is with such a
protest in mind that I stressed earlier, when expounding Grice, that the contrast
between speaker’s meaning and sentence meaning does not in any way put into
doubt that it is the speaker who takes a sentence he utters to have a certain sentence
meaning (that is, intends it to have certain truth-conditions). The speaker is not left out
of the idea of sentence meaning just because sentence meaning has a relative indepen-
dence from speaker’s meaning. By contrast, the opposing view, which prompts this
protest, prises apart what one’s words mean, what truth-conditions they have, from the
meaning or truth-conditions one intends them to have. Such a prising apart has
disastrous consequences of a kind that I have already discussed. The reason to see the
phenomenon of misspeaking as I am suggesting we should, where the intention to
mean something and the meaning of the words one utters are inextricably linked (that
is to say, not prisable apart), is quite simply that if they were not so linked, Frege style
puzzles would get no solution that arrests them. Fregean puzzles are puzzles that can
only be solved if there is no gap between the meanings of the words one utters and the
intentions with which we utter them and, therefore, no threat to our self-knowledge
of their meaning. By creating a gap between the truth-conditions or meaning of the
words a speaker utters and what the speaker means (by way of the truth-conditions
he intends for his words), the alternative understanding of misspeaking threatens to
make it possible for us to be unaware of the meanings of the words we utter (even as it
allows us to be aware of the meanings we intend to utter). What truth-conditions or
meanings our words have may now turn on factors that one may be entirely unaware
of; and we have already seen that if we allow for speakers to lack self-knowledge of the
meaning of what they say, we will have no satisfactory solution to the Fregean puzzles
about identity, at any rate no decisive solution which arrests those puzzles and prevents
them from arising one step up.
The important point is that the puzzles arise in a particular conceptual place—at the
site of the meanings of the words someone utters (“Londres est jolie”/“London is not
pretty,” “Hesperus is bright”/“Phosphorus is not bright”) and they need a solution at
that site. Thus it is no good to say that a speaker must know what he intends his words
to mean, but he needn’t know what his words in fact mean. He needs to know what
his words in fact mean, if the puzzle is to get a satisfactory solution because the puzzle is
located in the meanings of words, that is, in the sentence meaning. And they won’t get this
solution if what his words mean are prised apart from what he intends them to mean,
because that prising apart is what is responsible for the non-transparency of the
meaning or senses of his words that thwarts decisive solutions to the puzzle. Indeed,
the problem is worse than that. Quite apart from failing to decisively solve the puzzles, if
we take the view that prises these two things apart, we will not even understand what
makes the Fregean style puzzles go as deep as they do, a depth that Kripke understood
very well when he saw that a puzzle about identity is a puzzle about belief. Again, in
order not to distract from the main line of argument, I will spell out some of this depth
in Appendix II, where I say more about the detailed relations between the theme of
misspeaking and these puzzles. The interested reader should read that appendix directly
from where this issue is left here. The point for now remains: the reason to find empty
and dry the second source for saying that we might fail to fulfill our meaning intentions
is the very same reason for finding the first source fruitless. This protest we are
considering reveals the true nature of the mistaken view of misspeaking and it shows
how it too falls foul of the Fregean requirement that meaning cannot fail to be
transparent without failing to do what it was introduced by Frege to do, when he
raised the puzzles about identity.
There is hereabouts a final diagnostic clarification worth making quickly. There is a
very natural (though unreflective) tendency to resist this view of misspeaking that
I have promoted because misspeakings, at least on the surface, are very similar to
metaphors and indirect speech acts in the following respect. In all such phenomena,
it seems natural to think that something is meant by the speaker and needs to be
uncovered by artful interpretation that gets past the conventional sentence meaning
of what is uttered. If so, the interpretation that the speaker will go downtown (of the
spoken words “I am going towndown”) gets assimilated to the interpretation that
human beings are competitive (of the spoken words “Man is a wolf ”) and the
interpretation that I better hurry up and get on the train (of the spoken words “The
train is about to leave”). In other words, the interpretation that I am going downtown
of the misspoken words gets assimilated to the interpretation of speaker’s meaning and
not sentence meaning. With all this seeming similarity of the three cases, one is landed
with the idea that “I am going towndown” does not mean (is not true if and only if ) the
speaker is going downtown—anymore than “ ‘Man is a wolf ’ is true if and only if
human beings are competitive.” If this is right, then in the case of misspeaking (though
not in the case of metaphor and indirect speech acts), the misspeaker does not know his
own sentence meaning because it is not something about which he had any intention
whatever since he has uttered it by accident, it being a slip of the tongue, a misspeaking,
after all. So if it is right, we are landed once again with the disastrous consequence
I have been discussing—the inability to arrest the Fregean puzzles. It is, however,
112 AKEEL BILGRAMI
superficial and misleading to think that misspeakings are similar in this way to meta-
phors and indirect speech acts.
In the latter phenomena, speaker’s meaning, as I pointed out earlier, comes visibly
apart from the sentence meaning. In fact speakers exploit something in the sentence
meaning in order to convey something else to hearers. They convey that ‘something else’
by deploying the sentence meaning of the utterances they make. But in the case of
misspeaking, there is only a false impression of speaker’s meaning and sentence
meaning being visibly apart in the same sense. The idea that the speaker says what
she in fact merely happens to say (“I am going towndown”) in order to deliberately
convey something quite else—that she is going downtown—is completely inappropri-
ate in the analysis of misspeaking. The description ‘quite else’ is entirely out of place as a
description of what the speaker is deliberately up to, while it is perfectly correct in
describing what the speaker is deliberately up to in metaphors and indirect speech acts.
The fact that misspeaking turns on a visible difference between what is uttered and
what one is intending to get someone to believe (respectively, “I am going town-
down” and “I am going downtown”) should not confuse anyone into thinking that
the case is similar to the cases of metaphor and indirect speech acts, which also turn on a
visible difference between what is uttered and what one is intending to get someone to
believe (respectively “Man is a wolf ” and “Human beings are competitive” or “The
train is about to leave” and “Walk faster to the train”). In the case of misspeaking (the
utterance of “I am going towndown”), the speaker has the following two intentions: to
say something that is true if and only if the speaker is going downtown and to get across
to the hearer the belief that he is going downtown. There is thus coincidence—rather
than departure—of what the speaker intends to get across with what he intends his
words to mean whereas in the case of metaphors and indirect speech acts, there is
departure rather than coincidence. Metaphors and indirect speech acts are manifestly
pragmatic phenomena because our interest in them as speakers and hearers is not in the
deliberately intended sentence meaning of what is uttered but by what else is deliberately
conveyed by their utterance. By contrast, our interest as hearers in misspeaking (as well
as the misspeaker’s interest) is to get (and get across) the truth-conditions or sentence
meaning deliberately intended by the utterance—neither the misspeaker nor the hearer
have any interest in what else is accidentally conveyed by the uttered sounds.
In sum, then, my claim has been that if we see cases of misspeaking, such as slips of the
tongue, as having a literal, sentence meaning, a meaning that is different from what it
sounds like—a meaning that is imparted by the semantic intentions of the misspeaker—
then we can block misspeaking from becoming a second source for thinking that speakers
do not have their own meanings or truth-conditions right. And so, just as with the first
source discussed earlier, without the possibility of being wrong about our own meanings,
we lose our grip on the very idea of a notion of norm that holds of meaning since no one
would want to say that talk of normativity is apt when there is no possibility of being
wrong or mistaken. More specifically, to put in terms that we began with owing to
Wittgenstein: we cannot fail to act in accord with the intention relevant to meaning.
I argued (in Section 2) that that intention is not the intention to apply particular
words to particular things rather than others, but the intention to say particular words
with particular truth-conditions and satisfaction-conditions. The subsequent discussion
(in the present section) of the phenomenon of misspeaking helps me to stress a point
that I have tried to be careful about throughout the essay whenever I have formulated
this intention relevant to meaning. The intention relevant to meaning is best formu-
lated by saying that a speaker intends with an (assertoric) utterance to say something
which has particular truth-conditions. The word ‘something’ in this formulation has
the right generality. Sometimes speakers do not produce the exact sounds they intend
to produce, as when they misspeak. Thus when the intention is described with the
right generality, such cases will not spoil the efficacy of that intention. Our protagonist
who utters “I am going towndown” does indeed intend to say something that is true if
and only if he is going downtown. That intention, formulated with that generality, is
perfectly well fulfilled when he misspeaks, even if another intention formulated
without that generality (to utter “I am going downtown” in particular) is not. And it
is the former more generally formulated intention that is his semantic intention which
targets the sentence meaning of his utterance.
What I will concede is that when the intention relevant to meaning gets such a
general description as I am proposing (“I intend to say something which is true if and
only if . . . ”), we may sometimes find that what truth-conditions an utterance gets may
be rather idiosyncratic, given what he actually utters. A slip of the tongue displays this
vividly. If, by a slip of the tongue, I utter “Sara’s writing is eligible” when what
I intended to utter was “Sara’s writing is illegible,” then, even though on my view
I have intended that I utter something that is true if and only if Sara’s writing is illegible
and fulfilled that intention quite nicely (and even been understood to have intended
that, though perhaps with some initial strain), what I happened to utter (“Sara’s writing
is eligible”), gets a very idiosyncratic truth-condition. It gets a very idiosyncratic
sentence meaning.
Is there any shortcoming to allowing this sort of idiosyncrasy into literal, sentence
meaning, and not restricting such idiosyncrasy to non-literal, deliberatively creative
phenomena like metaphors and other figures of speech? The answer to this is, “yes.”
Idiosyncratic semantic intentions for one’s words put our hearers to some strain in
interpreting the words correctly. Unlike metaphors, which at least in poetry and other
creative writing, are intended to strain and surprise the reader (in a pleasurable way), the
routine utterances that may be the product of misspeaking, presumably are not so
intended. What this means is that even if a speaker cannot fail to act in accord with his
semantic intentions, as I am insisting, there may be another intention that a speaker has
that he does fail to act in accord with, which is the intention to say something that will
be easily understood by others, understood without strain or without surprise. One
assumes that speakers have such an intention in their ordinary speech most of the time,
and misspoken utterances or utterances made in medical (or other forms of ) ignorance
would be actions that fail to be in accord with such an intention. So I am not denying
114 AKEEL BILGRAMI
that various intentions, such as the one to be easily understood, are not always fulfilled
in the theoretical treatment of meaning I am proposing. But these are not the failures of
fulfillment that reflect any intrinsic normativity of meaning. “Speak so as to avoid
hearers strain in understanding what you have to say” is not an intention towards one’s
speech that reflects a norm that is intrinsic to language. It is a purely utilitarian norm.15
(See below footnote 15 and footnote 18 for more on this point.) It is not a norm of the
sort that squares with the deep assumption of normativity that philosophers have made
so central to meaning. That assumption of normativity is said by those philosophers to
be intrinsic to meaning rather than merely utilitarian.
I have first argued that if we make such an assumption of intrinsic normativity, then
that assumed normativity would have to reside in intentions which, when properly
identified, are intentions that target the sentence meaning or truth-conditions of one’s
words and which, when formulated correctly, have an appropriate generality. But if
this essay’s further Fregean argument is convincing, intentions, so identified and so
formulated, cannot and must not fail to get fulfilled in the speech of those who possess
them. And if they cannot fail to get fulfilled, they cannot reflect any genuine norma-
tivity, which requires the possibility at least, of the failure to get fulfilled. In other words,
we have a reductio ad absurdum of the assumption of normativity.
4. Concluding Remarks
Meaning intentions, then, are exceptions to Wittgenstein’s insight about the nature of
intention. They are not refutations of his insight and it would be a misunderstanding of
the argument of this essay to think it was presented with a view to providing a counter-
example to his claim about the normativity built into the very idea of intention. As a
generality, it is indeed true that intentions do have the normativity built into them that
Wittgenstein first profoundly brought to our notice. The point rather is that intentions
regarding meanings are a degenerate species of intentions and the deepest reasons for
this, which I cannot explore here in what is already a long essay, have to do with the
fact that meaning something is a rather unique kind of thing in that intending a meaning
and living up to that intention are—to put it flamboyantly and perhaps a little per-
versely—more like one thing rather than two, and so failures are not really possible.
This remarkable fact is one that I have pursued only briefly elsewhere and intend to
explore at much greater length in the future. The present essay’s conclusion is,
accordingly, relatively modest. Without fully spelling out this unique nature of mean-
ing intentions, it has given an argument to show why they must be viewed, in a very
specific sense, as degenerate. It is also modest in another sense. It has not argued that
15
In my Bilgrami 1992, Chapter 3, I allow for utilitarian and non-constitutive norms of meaning and
discuss at some length the difference between them and the sort of normative element one finds pervasively
assumed in the literature.
meaning is not normative in any interesting sense,16 though I believe that to be true
and have argued it elsewhere.17 The conclusion is merely skeptical about the norma-
tivity of meaning owing to the normativity of intentions. It argues only that meaning is
not normative because, despite its intimate link with intention, it does not inherit the
normativity that intentions possess; and the argument is that the normativity that
intentions possess lapse when intentions target meanings.
Should it be a cause for concern that normativity of this kind goes missing when it
comes to meaning? In the passage of this essay’s argument, we have seen the extent to
which there would have to be a loss of self-knowledge of meaning in order for
meaning to be normative and I have hinted at the extent to which that loss would
itself owe to a location of meaning in the social realm or in the objective realm of
scientific essences. I described these as the mundane versions of Plato’s more meta-
physically abstract locations. I reckon, then, that any concern we feel at such an absence
of norms reflects a residual, secularized yearning for Platonist forms of objectivity,
something that Wittgenstein would have seen as calling for therapy, not philosophical
satisfaction.18
16
By interesting I mean in some intrinsic and categorical sense normative, as opposed to merely utilitarian
norms. The norm or imperative “Do x if you want to be easily understood” does not have the same
philosophical interest (though it may have a lot of practical interest) as “Do x if you want to be understood.”
Because ‘being understood’ (as opposed to being understood easily) has internal dialectical links with meaning
itself, the norm is more intrinsic and categorical, rather than conditional. The fact is that when one says “I am
going towndown” and “I have arthritis in my thigh” there is no problem of one’s individualistically
conceived meanings being understood. Clearly, the idea that meanings are to be thought of individualistically
should not be confused with the idea that they are private and therefore unavailable for others to understand.
They are perfectly public phenomena. They merely put the hearer to strain in trying to understand one and
therefore violate the utilitarian norm about being understood easily.
17
See Chapter 3 of Bilgrami 1992.
18
Philosophers devoted to the mature philosophy of Wittgenstein may, however, be made anxious by
other aspects of this essay’s claims and arguments. First, I have assumed with many other philosophers that
there is a basic distinction between sentence meaning and speaker’s meaning, and devotees of the mature
Wittgenstein have often said that he never did subscribe to the idea of sentence meaning, leave alone to a
truth-conditional theory or an assertibility conditional theory or indeed any other kind of theory of sentence
meaning. I must confess to finding this implausible as a reading of Wittgenstein. I have argued in Bilgrami
1992 for a view of meaning that denies that there can be a theory of meaning understood in terms of certain
notions of truth-condition that rely on standard notions of reference, and I think, in doing so, I have
presented a view of meaning quite congenial to Wittgenstein, but which make makes no concessions to the
implausible idea that there should be no distinction between sentence meaning and speaker’s meaning. In
fact, I would have thought that Wittgenstein’s well-known remarks hinting at a ‘redundancy’ conception of
truth and its relation to language and meaning would suggest an almost explicit (to the extent that anything
of this sort could be explicit in Wittgenstein) commitment to sentence meaning. For an interesting discussion
of these matters, see Dummett 1981 in which he explicitly argues against those who think Wittgenstein
denied the very idea of what I have been calling ‘sentence meaning’. In general, not merely in that work but
in works too numerous to be cited, Dummett’s understanding of Wittgenstein exploits some of his ideas to
construct an argument for an assertibility rather than a truth-conditional conception of sentence meaning, that
he describes in terms of ‘sense’ which stands in contrast to the quite different phenomenon of ‘force,’
something that, though it is not to be confused with speaker’s meaning, stands like speaker’s meaning in
contrast with sentence meaning, which falls on the side of the phenomenon of ‘sense.’ In this respect,
Dummett has spent a philosophical lifetime trying to marry Wittgenstein and Frege, informing each of these
philosopher’s positions with the other’s in an attractive mutual elaboration of each. One does not have to
116 AKEEL BILGRAMI
Appendix I: Relating Meaning Intentions

to Grice’s Analysis of Meaning
Grice’s initial analysis captures what it is that the speaker tries to get across to a hearer in a
communicative episode. In doing so, it gives the speaker’s meaning on that occasion. The
analysis captures speaker’s meaning by citing three nested intentions of the speaker. These
intentions themselves have as their target an effect on the hearer. Essential to the speaker hitting
this target, I have said, is that the words on the speaker’s lips have a sentence meaning as well. And
this kind of meaning needs a further analysis. Such analysis is often given by notions such as truth-
conditions (or assertibility-conditions . . . ). I have said that the speaker must also have intentions
regarding this sentence meaning since he takes the words he utters to have it. It is not as if the
sentence meaning is outside the orbit of his mentality. It is right perhaps to describe all this in
terms of instrumentality. One tries to19 get something across to someone, get him or her to have
a certain belief, by uttering something, some words, with a certain sentence meaning. Both
(trying to get someone to believe something and saying something with a certain sentence
meaning) are therefore intentional acts—they are done with intentions. No doubt there are
other actions involved in such episodes as well, which have further instrumentality: one says
something with a certain sentence meaning by moving one’s throat, etc. This too is an intentional
act; it too is done with an intention. Some, perhaps all, of these intentions are highly implicit;
they are not in the forefront of the speaker’s mind. But that does not cast doubt on their being
bona fide intentions. It would be quite wrong to think that one’s moving one’s throat to speak in
agree with the details of Dummett’s verificationist tendencies regarding meaning, which he often tries to
draw from passages in Wittgenstein, to see the attractiveness of this general strategy of interpretation of these
two philosophers in mutually influential ways. A second aspect of my essay and argument that would make
scholars of Wittgenstein anxious is my manner of introducing neologisms to deal with cases of misspeaking as
well as with cases that are formulated by Frege and Kripke and Burge in their various puzzles and
philosophical thought experiments. Neologisms, it might be said, are willful intrusions into ordinary language
made with theoretical purposes in mind. Is this not anathema to the mature Wittgenstein? Since such
devotion as I have to the remarkable philosophical insights in Wittgenstein does not drive me to make every
last conviction of my own line up with every detail of his views, I am not myself made anxious at the thought
that it would be anathema. Even so, it is a real question as to how much these neologisms do fail to line up
with his views. If someone says “I am going towndown” and I refuse to take his words at face value and
interpret him as saying that he is going downtown instead, it is to put aside what he has uttered in order to
capture what he has thought and intended to express. Is that a remote theoretical intrusion or something that
common sense demands? And the same common-sense aspiration—putting aside what is uttered to capture
what the speaker thinks and wishes to express with the words he utters—is in play when one interprets a
medically ignorant person’s utterance of “I have arthritis in my thigh” along the lines I mention above in my
discussion of that case. In all these cases, one is not imposing some theoretical reformulation on the speaker’s
thoughts as they are expressed in his words, as for instance is found in efforts to provide the logical form of a
person’s utterances, something often considered to be anathema by some devotees of Wittgenstein. I share no
such distaste for those efforts. But I am registering a difference between what I have suggested in the cases
I am concerned with and those efforts. In my cases, as I present them, one is merely saying what a speaker’s
thoughts are and one is insisting that what his words are taken to mean must keep faith with what his thoughts
are rather than keep faith with what I have called a secular and mundane version of Plato’s ulterior,
objectivist, normative ideals for meaning. By contrast, when one reformulates someone’s utterances into
logical form, one is suggesting that his thoughts have a deeper structure than they have as they surface in his
speech; that is a more theoretically motivated enterprise than anything I have suggested and defended in my
discussion of my cases.
19
I say ‘tries to’ to stress that success or uptake on the part of the hearer is not required here.
normal forms of linguistic communication is not an intentional act and not an act done with an
intention merely because of its implicitness, in this sense. And the same holds of the intentional
act of saying something with a certain sentence meaning. A good schematic description of the
instrumentalities of this speech episode, then, is that our protagonist tries to (1) get someone to
believe that there is a snake in front of him by (2) uttering words which are true if and only if
there is a snake in front of him, which he, in turn, does by (3) moving his throat and producing
the sounds . . . Each specifies an intentional act, acts done with intentions. This can happen in one of
two ways. Say that R is the instruction ‘If C, do A!’ S may fail to recognize that he is in circumstance
C, and so fail to do A; yet it may still be true that S is following R. Or, he may correctly recognize
that he is in C, but, as a result of a performance error, fail to do A, even though he tries.
And it has been my point that it is only right to describe the second and intermediate step here
as the one that involves the notion of meaning. It is the notion of meaning as it is lodged in
(2) that philosophers have long made great and sophisticated efforts to provide some analysis and
theory of. Grice was no exception to this, but his is a particularly interesting case because he began
his analysis by focusing on the sorts of intentions with which (1) is done. He hoped to build on this
basis and move further to the eventual analysis of meaning as it is found in (2). He called this latter
‘timeless’ meaning (while I am invoking the more standard vocabulary of ‘sentence meaning’)
and it was his eventual goal to give an account of it. Whether he was able to do so on that basis is
not a question I am going to discuss here at all. But it is easy to be misled by something in Grice.
In the initial analysis in terms of the intentions that go into the intentional act (1), he used the
term ‘meaning’ (sometimes called by him ‘occasional’ meaning to contrast it with ‘timeless’) in
describing what he was analyzing in this initial phase. This may give the impression that when we
talk of linguistic meaning, (1) is our focus. But it is not. As I said, the subject of meaning is the
subject that resides in (2). It is what philosophers (including Grice) interested in meaning or
semantics have made the object of their theoretical efforts when they study those aspects of
language that are not purely syntactic nor purely pragmatic.20
That Grice himself was clear about this is obvious from the fact that in subsequent years he
provided a detailed supplementary account21 of certain elements found in (1) by introducing a
notion of conversational implicature, and explicitly viewed that as being in the domain of
pragmatics, not semantics. The details of that account are not essential here since what I want
to really stress is that the supplement is a supplement to Grice’s analysis of the phenomenon in (1)
and once one notices that something like conversational implicature is relevant to this phenom-
enon, it becomes explicit that it is to be distinguished from meaning, even for Grice. The idea is
that getting something across to someone by telling them things, getting them to form certain
beliefs and so on, can be (even if it often is not) achieved by means that may be quite at odds with
sentence meanings—as the case of metaphors obviously makes clear. That someone should get
someone to believe that human beings are competitive (by saying something manifestly false, i.e.
saying words that are true if and only if men are wolves which, they manifestly are not) falls
squarely within the domain of (1) above. What Grice does in his work on pragmatics and
conversational implicature is to give principles by which one may identify departure of speaker’s
meaning from meaning (sentence meaning), principles such as “Don’t state things that are
20
Chomsky in a different categorization sees no point to this three-fold characterization for reasons that
are not relevant here. For a sympathetic report on what motivates Chomsky, see Bilgrami 2002.
21
Grice 1975.
118 AKEEL BILGRAMI
obvious falsehoods.” Violations of such principles are the signs that something like metaphor, for
instance, is in play. A speaker intends (in the sense of (1)) to get someone to believe that human
beings are competitive by asserting an obvious (zoological) falsehood, thus getting the hearer
alerted to the fact that some interesting departure from the sentence meaning must be in play.
Thus the theory of conversational implicature or pragmatics is concerned primarily with the
initial three-fold intentional analysis in (1), making clear that (1) is not the site of semantics or
meaning, even though Grice (in effect misleadingly, we can say in hindsight) used the term
‘meaning’ to describe what was being analyzed in (1). Meaning is located in (2) and since Grice
(invoking Frege, sometimes using Tarski-style axiomatizations of the insight in Frege) it has
come to be explicitly understood in terms of truth-conditions (or assertibility-conditions or . . . ).
As a result, in the case of metaphors it is very clear that when the world is manifestly uncoopera-
tive, when the truth-conditions have very obviously not obtained, then a principle of conversa-
tional implicature (Don’t speak obvious falsehoods) has been violated. This alerts the hearer to
the fact that the speaker (not likely to be making such an obvious mistake or uttering such a
blatant and easily discoverable falsehood) intends (in the sense of (1)) something that departs
dramatically from what is identified as the meaning or the truth-conditions of the words, intends
to get across that human beings, far from actually being wolves, are instead merely competitive.
And the crucial point remains that the fact that there is often no departure (dramatic or otherwise)
from (2) in the beliefs that one conveys with one’s intentions (in the sense of (1)) should not
confuse us into thinking that (1) is where semantics or meaning and not pragmatics is in play.
I’ve spent some time on Grice partly because he was the first philosopher since Wittgenstein to
study with some systematic care the relations between intentions and linguistic meaning, which is
my particular way of approaching the subject of this essay—the question of the normativity of
meaning—but partly also to diagnose why philosophers in their claims for the normativity of
meaning have misidentified the intentions relevant to linguistic meaning, in the way I mentioned
earlier in the main body of the text. If my diagnosis is right, one can, if one is not careful enough,
be misled by Grice’s use of the term ‘meaning’ to describe the phenomenon in (1), to think that it
is there rather than in (2) where the intention-based analysis of meaning is to be entirely located.
This would account for why one might conclude that someone’s use of the words “That is a
snake,” mistaking a rope for a snake, amounts to a mistake about meaning. How exactly would it
account for it?
If one thought that the intention relevant to meaning is that one says something with the
intention of conveying something to a hearer or, as Grice puts it in (1), with the intention of
getting the hearer to believe something, then perhaps we would be able to establish the
normativity of meaning along the lines that philosophers have assumed. If someone said “That
is a snake” with the intention to get a hearer to believe that there is a snake in front of him, then
pointing to a rope rather than a snake might well produce the requisite sort of error in the matter
of meaning. One must point to a snake in order to fulfill that particular intention because if the
hearer notices that it is a rope, he will not come to believe that that is a snake. Perhaps the idea is
better put when it is elaborated more explicitly as follows. If one intends to get someone to
believe something which one takes to be true, viz., that there is a snake in front of one, then one
way in which one might do it is to further intend, in one’s literal and sincere assertoric utterance of
“That is a snake,” to apply those words to a snake in front of one. If there is no snake there, one
has misapplied the terms, one has failed to act in accord with that further intention, and so, in turn,
quite possibly may fail to fulfill the original intention of producing a certain effect in the hearer.
Notice that I am being careful to say “quite possibly may,” because the hearer, as I said, has to
notice that I am perceptually wrong in order to fail to come to have the belief that there is a snake
there. Remember Grice’s initial analysis is an analysis appealing to the phenomenon of intending
to have effects on hearers, getting them to have beliefs, and so on. So the hearer would have to fail
to acquire the belief that there is a snake there for the intentions, which provide the analysis, to be
unfulfilled. It is not therefore sufficient for that Gricean intention to be unfulfilled that I say “That
is a snake” when there is a rope present. The hearer, if he is not alert, may not notice that I am
wrong and come to believe that there is a snake there, on hearing my words. In that case the
Gricean intention would be fulfilled. The only intention that immediately fails to get fulfilled in
the presence of a rope is what I referred to as the ‘further’ intention that is tied to what one is
applying one’s words to. This is the intention that is claimed by many to be directly and
immediately tied to the normativity of meaning, when that normativity is viewed in terms of
correct and incorrect application of words.
Even so, the important point, if my diagnosis is right, is that this last sort of intention (what
I called the ‘further’ intention) is squarely in the region of the intentions to convey things to
hearers, the phenomenon that Grice analyzes in his initial analysis. It is not at all in the region of
the semantic intentions of saying things with certain truth- or satisfaction-conditions, which
analyze the quite different phenomenon of ‘timeless’ or sentence meaning that he wants his
analysis to eventually work up to. The ‘further’ intention is caught up in the former region of
intentions because it is directly a part of one’s project of conveying things to people; it is directly in
the service of getting people to believe things, etc. It is not to be collapsed with or even really
deeply associated with sentence meaning and with the latter region of intentions—to say things
with certain truth-conditions. And my point is that once we (a) see this and (b) see no direct
relevance to meaning of any intention in the former region, then whatever else may reveal
the possibility of failures to fulfill the intentions relevant to meaning (intentions in the latter
region), it won’t be such things as mistaking ropes for snakes and, more generally, misapplying
one’s terms.
Appendix II: Getting Disquotation Right So As To See

Why The Puzzles About Identity Go So Deep
The puzzles about identity to be found in Frege and Kripke raise some of the most wide-ranging and
deep issues in philosophy because they are puzzles that show that it is not really possible to do the
Philosophy of Language fruitfully without bringing onto center-stage the Philosophy of Mind. On
the view of misspeaking that I was inveighing against in the main body of the text above, the words
we utter are not inextricably linked to our mentality (as I was suggesting they are). Rather they are
prised apart from our mentality. This diminishes the significance of the puzzles as I see them—and is
plain for anyone to see. What is plain to see is that the puzzles are not just raising points about
meaning and reference, they are puzzles that reach down to (because they threaten) our most
elementary assumptions about the nature of the rationality of speakers of a language (as is evident in
the fact that I have been stressing, viz. that if they are not decisively addressed and arrested, we will
not be able to make a distinction between lack of empirical knowledge and failure of logic or
reasoning). And rationality is a property of a speaker’s mind. So the links between language (or
meaning) and mind have to be closer than the view I am opposing permits, to so much as raise the
120 AKEEL BILGRAMI
puzzles about identity in their full depth and significance. This can happen in one of two ways. Say
that R is the instruction ‘If C, do A!’ S may fail to recognize that he is in circumstance C, and so fail to
do A; yet it may still be true that S is following R. Or, he may correctly recognize that he is in C, but,
as a result of a performance error, fail to do A, even though he tries.
Kripke himself, despite the fact that he is no friend of Frege’s solution (that appeals to senses) to
the puzzles about identity, nevertheless sees the importance for the puzzle of this inextricability
of the link between meaning and mind. Whatever one may think of his skepticism about ‘senses’
and of his own way of addressing the puzzle, his understanding of how the puzzle arises and how
deep it goes, is insightful and instructive. It is precisely because he takes meaning and belief/
intention/ . . . (the intentional aspects of the mind, generally) to be inextricably linked, that his
puzzle ends up being a ‘puzzle about belief,’ as his very title makes clear, and therefore about the
rationality of his protagonist, Pierre.
Having said that, it also must be said that there are complications in the specific way he elaborates
the link between meaning and belief that are of special interest when it comes to misspeaking and
they ought to be discussed and clarified because they are highly instructive. The link is elaborated by
Kripke in what he calls the principle of disquotation. That principle makes the link by saying that
whenever a linguistic agent sincerely utters or assents to a sentence (say, “London is not pretty”) one
may remove the quotation marks and place the sentence in a that-clause specifying the content of
the agent’s belief ( . . . believes that London is not pretty). This principle, along with two other
things—something he calls a ‘principle of translation’ (which essentially asserts that translation
preserves truth-conditions) and Kripke’s own causal, anti-Fregean account of reference—together
give rise to his version of the Fregean puzzle. (The principle of translation is only needed for Kripke’s
own version of the puzzle because it has two languages, French and English.) Suppose, then, that
Pierre, brought up in Paris and influenced by his Anglophile nanny, asserts sincerely “Londres est
jolie,” and then moves to London, without knowing that Londres is London, and lives in some
shabby locality and asserts sincerely in his newly acquired language, English, “London is not pretty.”
With all this supposed, an adherence to the combination of the three things I mentioned (the two
principles—of disquotation and of translation—and an anti-Fregean, Kripkean causal account of
reference), give rise to the puzzle. We have an agent, who is merely uninformed of the identity of
London and Londres, believing two inconsistent things—believing that London is pretty and
believing that London is not pretty. In that combination of three things, Fregeans would readily
adhere to the two principles22 but will reject the causal doctrine about reference, introducing
instead the notion of senses (disallowed by that doctrine) to block this unacceptable implication of
Pierre’s irrationality. Kripke finds this an unsatisfactory solution to the puzzle, partly because of his
prejudices about the notion of sense, some of which quite possibly flow from a failure to understand
that senses are in a very deep way not at all like planets and cities—one cannot have multiple
perspectives on them, and so they have to be the sorts of thing that are self-known to speakers, and
therefore they cannot be the subject and occasion of further puzzles one step up. But here I don’t
want to focus on the dispute between Kripke and the Fregeans. I want to just focus on the principle
of disquotation since that is Kripke’s particular elaboration of the general idea that meaning and
mentality are inextricably linked, the very idea that is essential to my claims about what goes on in
misspeaking.
22
Though see the next few paragraphs for a serious qualification added to the principle of disquotation, as
sensible Fregeans would understand it.
The trouble with this particular elaboration of the general idea is that, at least on the surface, it
seems to actually spoil rather than elaborate the inextricability when it comes to the case of misspeaking.
When someone says, “I am going towndown,” a specific understanding in terms of disquotation of
the general idea of the inextricable link would seem to have it that he believes that he is going
towndown. And that is not a very clear and perspicuous attribution of belief. By contrast, in my gloss
on misspeaking, I have the agent believing that he is going downtown (a perfectly clear and
perspicuous attribution) because I also have his utterance “I am going towndown” mean that he is
going downtown. And it is possible to have it mean that because one takes the speaker to intend that
his words are true if and only if he is going downtown, and one takes the meanings of his words to be
fixed by that semantic intention of his and not by some other factors, often adduced by philosophers,
such as social linguistic practice23 (or the scientific essence of substances, diseases, etc.) that may in
some cases not even by known to the speaker. It is these close links between his meaning and his
mentality (i.e. his semantic intention and therefore also his belief ), which allow me to make the clear
and perspicuous attribution of belief to the speaker. In fact, there is a close circle of links between
two mental and one linguistic phenomena: a speaker believes that he is going downtown, he intends
the words he utters (“I am going towndown,” as it happens) to be true if and only if he is going
downtown, and the words he utters are indeed true if and only (they indeed mean that) he is going
downtown. So, it looks as if something goes amiss when one elaborates the close links in terms of the
principle of disquotation. What the links should attribute in a perfectly clear attribution of a belief, as
I have it, gets mangled into a bizarre attribution of the belief that he is going towndown, when the
links are seen in strictly disquotational terms.
The fault line here is the strictly disquotational elaboration of the close links between meaning
and mind.
Let me raise the issues first not with the case of misspeaking, which is some distance away from
the puzzles about identity, but with a case much closer to the kind of examples discussed in the
puzzle by Frege and Kripke. There is a version of the puzzle that Kripke mentions briefly, where
the two names uttered in the two seemingly inconsistent sentences generated by the puzzle are not
merely in the same language as they are in Frege’s original version of the puzzle (unlike in the
puzzle about Pierre) but are indeed, on the surface, the same name, “Paderewski.” There is in fact
only one person called “Paderewski,” but the protagonist doesn’t know that and thinks that there
are two different persons, one a statesmen, the other a famous musician. Thus the apparent
inconsistency seems to be starker, with the protagonist saying, not merely, “Hesperus is bright”
and “Phosphorus is not bright,” but, let’s say, “Paderewski is musically accomplished” and
“Paderewski is not musically accomplished.” Here, elaborating the close link between meaning
and mentality in terms of a strict version of the principle of disquotation will land us in a midden.
If, as I have been insisting, it is one’s semantic intentions that impart meanings to one’s words and
one’s semantic intentions are to be thought of as having close links with one’s beliefs as mentioned
in the previous paragraph, and the belief itself is identified via a disquotational strategy that links it
with the utterance, one will have to say that the speaker semantically intends his first utterance to
23
None of this implies that social linguistic practice is not a necessary and crucially enabling condition in
the background for one to be able to have the kind of individualistic semantic intentions that fix the meanings
of one’s words. Individualism about meaning does not have to implausibly deny all role to the social
embedding of individual speech in the background. It merely denies its place in the foreground, determining
an individual’s meaning in terms of what experts in a society think, i.e. the secular (mundane) version of the
Platonic idea I mentioned earlier.
122 AKEEL BILGRAMI
be true if and only if Paderewski is musically accomplished and the second utterance to be true if
and only if Paderewski is not musically accomplished. If that happens, senses or meanings, even if
they are imparted by our semantic intentions, do not seem to have solved the puzzle at all.
What is obviously needed here to make disquotation line up with the spirit and the underlying
point of the close links it elaborates between meaning and mind, is to allow that disquotation
comes into play after we make clear what the carriers of the semantics really are. In this example,
the term “Paderewski” in each sentence contains an unpronounced (and unwritten) subscript.
The sounds and inscriptions as they originally stand are incomplete and they don’t properly
reveal the words that comprise the full and proper semantic items. The semantics (the senses) are
carried by neologisms that we will therefore have to introduce, “Paderewski1” and “Pader-
ewski2.” The sounds and inscriptions misleadingly leave the subscripts out. Once we neologize,
disquotation can proceed apace. Now, the semantic intentions with which the speaker utters the
two sentences can be formulated in a way that allows the meanings or senses to remove the
puzzle. Of course, this solution to the puzzle is not Kripke’s solution, as I have already said. It is
Fregean. It appeals to senses, given in two different semantic intentions, which intend for each of
the two (only seemingly inconsistent) sentences, two different truth-conditions. So also, in the
case of misspeaking, disquotation should proceed after we make clear that “towndown” is really a
mispronunciation of “downtown,” just as each utterance of “Paderewski” in the two sentences
of that version of the puzzle about identity is a mispronunciation of “Paderewski1” and
“Paderewski2” respectively.24 Once this is done, we are landed with no unclear and unintelligi-
ble attributions of belief to the person who misspeaks and says “I am going towndown.”
24
With all this spelt out, we are in a position now to go back and say that neologisms are also relevant to
the discussion of the first source of difficulties we were considering earlier in the essay, in the example of the
medically ignorant protagonist. There we said that such a speaker intended his utterance “I have arthritis in
my thigh” as being true if and only if the speaker had a disease of the joints or ligaments in his thigh. We know,
however, that arthritis is a disease of the joints only. So, in my treatment of that example, I was presupposing
that the term ‘arthritis’ on his lips was a mispronunciation of, say (neologizing here), ‘tharthritis.’ Once we
neologize in this way, there too we can proceed apace with disquotation instead of describing the intended
truth-conditions as “ . . . if and only if the speaker has a disease of the joints or ligaments in my thigh” as I did
at that point in my essay, when the apparatus of neologizing had not been introduced. We can now simply
say the intended truth-conditions are to be described as “ . . . if and only if the speaker has tharthritis in this
thigh.” The operative slogan is: “fix the left hand side by neologizing and then disquote on the right.” In my
discussion of this case above, I had said the only mistake the medically ignorant man had made was that he
didn’t speak like others did. He used the term ‘arthritis’ in a way they didn’t, to mean something they didn’t.
But now we have the apparatus to say that his term ‘arthritis’ was just a mispronunciation of ‘tharthritis,’ so it
wasn’t even as if he was using the word differently, but rather merely mispronouncing, as it were. I don’t
deny that this amounts to gerrymandering the terrain of usage here. Which is the right gerrymandering will
be forced on us by the philosophical issues at stake (issues such as needing decisive solutions to the Frege puzzles
and so on) since the data of usage itself leave it underdetermined which way we should gerrymander. The data
(including data about deference, as I have discussed it) don’t determine whether we should say (1) that he fails
to know what his own meanings and thoughts are or instead (2) he knows what his own thoughts and
meanings are but merely doesn’t know what his fellows’ meanings and thoughts are. And I am saying that
which of these two things we should say about him depends on which better squares with the philosophical
issues, in particular the issues about solving something like Frege’s puzzles in a way that preserves basic
distinctions between rationality and empirical error. My claim has been that these philosophical issues decide
in favor of (2). For more on this sort of example and the philosophical questions they raise, see the exchange
between Brian Loar and me (Bilgrami 1985, Loar 1985a, 1985b). I respond to some of the points Loar makes
in his “Reply to Bilgrami” there in Chapter 3 of my Bilgrami 1992. The fact that neologisms are involved in
responding to both sorts of cases shows that the issues in the two sources of difficulty that I am discussing are
very similar. In the first source, one neologizes to keep track of the idiosyncratic meanings and concepts of a
Someone may wish to enter a quarrel here, claiming that this is to abandon the disquotation
principle, not rescue it from a strictly rigid reading. But this dispute is paltering and terminologi-
cal. Whether we say the theoretical move here is ‘disquotational’ or not, hardly matters. What
matters is to register that the move fastens on an insight in Kripke that the puzzle about identity is
a puzzle that reaches all the way down to issues of belief, mind, and rationality, and builds from
that insight a notion of sense based on an individual speaker’s semantic intentions to characterize
the meanings of the words in his speech. Kripke’s insight that the puzzle goes that deep is just the
insight that meaning and mind have inextricably close links, but he then does not see in those close
links the possibility of developing this individualist Fregean25 notion of sense that would
decisively resolve the puzzle. And I am claiming that this combination of insight and failure on
his part is revealingly present in his own principle of disquotation. The principle itself is an
acknowledgement of the close links and is of a piece with (in fact it makes for) the insight that the
puzzle reaches all the way down to issues of rationality. The failure to give the principle the less
strict and more relaxed reading is of a piece with the failure to build on the insight about those
links and provide a plausible version of the Fregean solution. If one sees disquotation in a less
strict way than it seems on the surface to be, one has all the ingredients for a convincing Fregean
solution to the puzzles, one that would arrest them once and for all.
Bibliography
Bennett, J. 1973 “The Meaning-Nominalist Strategy,” Foundations of Language 10, pp. 141–68.
Bilgrami, A. 1985 “Comments on Loar,” in R. H. Grimm and D. D. Merrill (eds.) Contents of
Thought, Tucson, University of Arizona Press, pp. 110–39.
Bilgrami, A. 1992 Belief and Meaning, Oxford, Blackwell.
Bilgrami, A. 1993 “Norms and Meaning,” in R. Stoecker (ed.) Reflecting Davidson, Hawthorne,
De Gruyter, pp. 121–44.
Bilgrami, A. 1998 “Why Holism is Harmless and Necessary,” Nous 32 suppl. 12, pp. 105–26.
Bilgrami, A. 2000 “Is Truth a Goal of Inquiry? Rorty and Davidson on Truth,” in R. Brandom
(ed.) Rorty and his Critics, Malden, Blackwell, pp. 242–61.
Bilgrami, A. 2002 “Chomsky and Philosophy,” Mind and Language 17/3, pp. 290–302.
Bilgrami, A. 2003 “A Trilemma for Redeployment,” Philosophical Issues 13/1, pp. 22–30.
Bilgrami, A. 2006 Self-Knowledge and Resentment, Cambridge, Mass., Harvard University Press.
Bilgrami, A. and May, R. Forthcoming. “Frege and Chomsky on Senses and Perspectives.”
Bilgrami, A. and Rovane, C. 2005 “Mind, Language and the Limits of Inquiry,” in J. McGilvray
(ed.) Cambridge Companion to Chomsky, Cambridge, Cambridge University Press.
Boghossian, P. A. 1989 “The Rule-Following Considerations,” Mind 98/392, pp. 507–49.
medically ignorant person. In the second source one neologizes to keep track of the concepts of a misspeaking
person. In each case one does it to ensure that there is no failure to know what a speaker means.
25
Of course, not all Fregeans are individualist in their conception of sense in the way that I have been.
And it is doubtful that Frege himself should be counted as such. Though in my co-authored essay with
Robert May, tentatively entitled “Frege and Chomsky on Senses and Perspectives,” we excavate elements in
Frege’s texts, which suggest an individualist reading of senses of the sort being presented here. See also
Bilgrami and Rovane 2005.
124 AKEEL BILGRAMI
Burge, T. 1978 “Individualism and the Mental,” in P. French, T. Uehling, and H. Wettstein
(eds.) Midwest Studies in Philosophy 6, Minneapolis, University of Minnesota Press, pp. 73–121.
Dummett, M. 1976 “What is a Theory of Meaning II,” in G. Evans and J. McDowell (eds.) Truth
and Meaning, Oxford, Oxford University Press, pp. 67–137.
Dummett, M. 1978 “The Social Character of Meaning,”in Truth and Other Enigmas, London,
Duckworth, pp. 420–30.
Dummett, M. 1981 “Frege and Wittgenstein,” in I. Block (ed.) Perspectives on the Philosophy of
Wittgenstein, Oxford, Blackwell, pp. 31–42.
Frege, G. 1884 Die Grundlagen der Arithmetik, trans. J. L. Austin as The Foundations of Arithmetic,
Oxford, Blackwell.
Frege, G. 1892 “On Sense and Reference,” in Translations from the Philosophical Writings of Gottlob
Frege, 1952, trans and ed. P. Geach and M. Black, Oxford, Blackwell.
Grice, H. P. 1957 “Meaning,” The Philosophical Review 66/3, pp. 377–88.
Grice, H. P. 1969 “Utterer’s Meaning and Intention,” The Philosophical Review 78/2, pp. 147–77.
Grice, H. P. 1975 “Logic and Conversation,” Syntax and Semantics Vol. 3, in P. Cole and
J. L. Morgan (eds.) Speech Acts, New York, Academic Press, pp. 41–58.
Kripke, S. 1976 “A Puzzle about Belief,” in E. Margalit (ed.) Meaning and Use, Dordrecht,
D. Reidl, pp. 239–83.
Lewis, D. 1969 Convention: A Philosophical Study, Oxford, Oxford University Press.
Loar, B. 1985a “Social Content and Psychological Content,” in R. H. Grimm and D. D. Merrill
(eds.), Contents of Thought, Tucson, University of Arizona Press, pp. 99–110.
Loar, B. 1985b “Reply to Bilgrami,” in R. H. Grimm and D. D. Merrill (eds.), Contents of
Thought, Tucson, University of Arizona Press, pp. 000–000.
Mates, B. 1952 “Synonymity,” in L. Linsky (ed.) Semantics and the Philosophy of Language, Urbana,
Ill., University of Illinois Press, pp. 111–38.
Schiffer, S. 1972 Meaning, Oxford, Oxford University Press.
Strawson, P. 1971 “Intention and Convention in Speech Acts,” in Logico-Linguistic Papers,
London, Methuen, pp. 115–30.
Travis, C. 1995 “Order out of Messes,” Mind 104/413, pp. 133–44.
SECTION II
Knowledge of Our Own
Minds and Meanings
5
The Publicity of Meaning
and the Interiority of Mind1
Barry C. Smith
It is a great pleasure to write a paper in honour of Crispin Wright. Anyone who has
interacted with him will know the sense of urgency and excitement he conveys about
philosophical discussion. The power and intensity of his thinking are immediately
obvious, and one is put on one’s mettle, struck not only by the quality and precision of
his thinking but the impressive depth of his engagement with the issues. More than
most, he gives one a sense of getting at the fundamental concerns, the really haunting
problems of philosophy. Whatever we make of his many constructive suggestions it is
his unerring capacity to highlight philosophical problems that is perhaps his greatest
contribution to the subject. His sharp focus, and clear insight have opened up inquiries
that will be of lasting importance for the discipline. It is an impressive contribution and
we are far richer for it.
1. Two Themes in Wittgenstein

In his writings on the later Wittgenstein, Crispin Wright deals in detail with two
themes he sees as closely related: the nature of our knowledge of our own intentional
states and the nature of rule-following. Each topic has received considerable attention
in the philosophical literature but what sets Wright’s discussion of these topics apart is
the deep connections he sees between the two issues and his attempt to provide a
relevantly similar treatment of both.
1
Versions of this essay were given to audiences at the University of California at Berkeley, the University
of Edinburgh, the Institute Jean Nicod in Paris, the University of Stockholm, and the Department Seminar at
Birkbeck College, University of London. I am grateful to audiences on all those occasions for comments. I am
deeply indebted to Marcia Cavell, Donald Davidson, Guy Longworth, Fraser Macbride, John McDowell,
Peter Pagin, and Crispin Wright for discussion of these issues. I should also like to express my deepest
gratitude to Ophelia Deroy for invaluable help with a previous draft.
128 B A R RY C . S M I T H
2. Following a Rule
A central theme of the Philosophical Investigations is the idea that meaning is use: that all
there is to the meaning of a word is the use speakers make of it. And for a word to have
a meaning there must be something that counts as correct and incorrect uses of it.
Otherwise, any application of the word would be as good as any other, and there
would be nothing to ensure a word meant one thing rather than another. So meaning
may be determined by use but there must be something to fix a standard of correctness
for that use. As we know, Wittgenstein thought that the only thing that would serve to
do so was the rule-governed practice of a community of language users. The use of
words by individual members of the linguistic community will count as meaningful just
so long as they are part of the practice that maintains the linguistic rules of that
community.
However, there are notorious difficulties in giving content to the idea of an
individual’s participating in a communal practice and adhering to the rules governing
use. The idea of rules as enshrined in human practices was offered as an alternative to
the Platonic conception of autonomous rules as a way to ensure an objective standard
for the correct use of a word. The Platonic conception sees rules as extending beyond
any actual applications, stretching into the future like rails laid to infinity.2 However, it
is not enough for an individual to count as a rule-follower simply that there exists some
standard independent of him by reference to which his uses of a word count as correct
or incorrect. The speaker’s merely happening (or failing) to coincide with the objective
requirements of a rule would not show that he was using a word in accordance with its
meaning, adhering to that rule in his use. Some way is needed of showing how a speaker
conforms his use to a rule. And for this we need to understand how speakers take the
requirements of rules into account when using their words.
It is precisely the lack of any credible epistemology for the rule-follower on the
Platonic conception of a rule that Wittgenstein demonstrated so effectively in the
Philosophical Investigations. Nothing the speaker can avail himself of: no intuition,
inclination, consultation with an inner item, or private interpretation of the rule, can
put him in touch with, or establish that he is adhering to, an autonomous standard of
application to which his subsequent use of words is required to conform. For how, on
the basis of inner consultation or interpretation of the rule, could the speaker make out
an intelligible distinction between applications of a word that seem right, and those that
are right by the proposed objective standard? Without such a distinction no sense can
be given to the speaker having a standard of correctness for his use of words.
Some philosophers consider Wittgenstein to have been so successful in demolishing
candidate explanations of what it is to follow a rule that he left nothing in place to
count as a speaker’s following a rule, as his meaning one thing rather than another by
his words. However, the meaning scepticism that threatens in response to the failure of
2
The image is of course Wittgenstein’s, }} 218–19.
THE PUBLICITY OF MEANING AND THE INTERIORITY OF MIND 129
the inner consultation model is vitiated by Wittgenstein’s famous remarks at Investiga-

tions 201 and 202:
What this shews is that there is a way of grasping a rule which is not an interpretation, but which is
exhibited in what we call ‘obeying the rule’ and ‘going against it’ in actual cases.
And hence also ‘obeying a rule’ is a practice. And to think one is obeying a rule is not to obey a
rule. Hence it is not possible to obey a rule ‘privately’; otherwise thinking one was obeying a rule
would be the same as obeying it.3
The appeal to practice is meant to provide the needed contrast between what the
speaker is inclined to do, and his actually using the word correctly; namely, his using it
in accordance with the norms of the practice to which he belongs. Unlike the Platonic
alternative, what now grounds the objectively correct use of a word, and thus confers
meaning on it, is no longer beyond the epistemic reach of the language user. Objective
meanings become publicly determinable: i.e. it is in principle possible to know what
speakers of a linguistic community mean by their words just by observing facts about
their linguistic practice. However, it is worth noting that this is not the way members
of a community know the meanings of their words. They know what they mean
without observing their own or anybody else’s use. As speakers of the language they are
immediately apprised of the meanings of their words from the first-person point of
view. So how should we account for the first-person epistemology of the rule-
follower? The postulation of publicly observable practices in place of Platonic rules
does nothing, it seems, to cast light on this phenomenon. Word meaning may be a
public and objective affair, but speakers know what they mean without first checking
the publicly available facts. Thus, one particularly pressing difficulty is that of showing
what makes possible such immediate and effortless grasp of objective and third-
personally knowable meaning-facts.
3. Wright on Rules
Crispin Wright has done much to stress the epistemological dimension of the rule-
following considerations, insisting that any satisfactory answer to the question of what it
is to follow a rule must at the same time be an account of what the rule requires of us at any
point.4 His constructive proposals may or may not be in line with Wittgenstein’s own
thinking on the matter, but this is not my concern. Nor I think Wright’s; though it is
evident that Wittgenstein was struggling with the problem Wright identifies and that
remains live in contemporary philosophy of language and philosophy of mind. The
3
Wittgenstien 1953, }} 201–2.
4
Some will no doubt demur at Wright’s epistemological interpretation of the rule-following considera-
tions, focusing instead on what they consider, following Kripke, to be more constitutive or metaphysical
questions about which facts, if any, make it the case that a speaker means one thing rather than another by his
words. But here I am with Wright in thinking that the two issues cannot come apart: “truths about meaning
have to be, ultimately, constituted by facts about understanding”—Wright 1986, p. 210.
problem is how to reconcile the existence of public facts about meaning with a
speaker’s first-person comprehension of them. It emerges clearly for Wittgenstein at
Philosophical Investigations }138 where he asks how meaning can be all there at once in
the mind of the speaker if meaning is also the unfolding of use through time.5
Objective meanings for words are secured by publicly observable facts. However,
diffuse social facts about speakers’ uses, extending into the future and reaching back
into the past, are hard to square with the phenomenon of our immediately knowing
the meaning of our words. Speakers do not consult others’ uses, nor observe their own,
in order to understand what their words mean. As Wright points out “we may and
usually do non-inferentially know of our present meanings” and “Phenomenological-
ly, at any rate, construal of a novel utterance is often immediate and spontaneous.”6
The problem, then, is to see how such immediate and effortlessly arrived at intuitions
of linguistic meaning can engage with the publicly discernible meaning-facts to which
they are answerable if they are to count as knowledge.
4. Self-Knowledge
A similar set of issues arises, of course, for the case of judgments about our own mental
states. Other people’s psychological states are interpretable on the basis of their publicly
observable behavior, so how can they be, at the same time, immediately knowable by
their possessors from the first-person point of view without recourse to behavioral
evidence? Conversely, if the facts of mental life are part of the inner world of a subject
how can they also be publicly knowable on the basis of outwardly interpretable
behavior? As Wright points out, the phenomenon is the same in both the linguistic
and the psychological case:
Each of us is, for the most part, effortlessly authoritative, without inference about our past and
present intentions, about the rules ‘we have in mind’, about how we understand or have
understood particular expressions.7
The question is: how do our effortless and unsupported first-personal judgments
amount to genuine claims to knowledge. As Wright puts it:
How can [such] judgments lack a substantial epistemology in this way, and yet still be objective—
still have to answer to something distinct from our actual dispositions of judgment?8
In the linguistic case, the publicity of meaning and the unsustainability of a private
language convince Wright that, “Meaning . . . cannot lie within the province of individual
5
“But we understand the meaning of a word when we hear or say it; we grasp it in a flash, and what we grasp
in this way is surely something different from the ‘use’ which is extended in time!” Wittgenstein 1953, } 138.
6
Wright 2001, pp. 177, 190.
7
Wright 2001, p. 176.
8
Wright 2001, p. 191.
psychology.”9 Meaning is clearly more than just a subjective matter of how things are in
the speaker’s consciousness. Similarly, in the case of knowledge of our mental states, it
may seem as though the Cartesian can capture the peculiar immediacy and intimacy of
our self-knowledge, since the private states in question enjoy no distance from the subject.
But this will not explain how we can make accurate judgments about our own mental
states: being in a psychological state is one thing; knowing which pychological state we are
in is another. Also, the Cartesian cannot explain how such mental states or meanings could
be available or accessible to anyone else, since they have no existence outside of the mind
to which they belong. If we reject subjective meanings and essentially private states of
mind, and endorse the publicity conditions that bring genuine meaning-facts and mental
states into public view we need to leave room for the subject to know, albeit by different
means, the same thing that others who understand or interpret him know. Therefore, we
need to see how the subject’s inner experience or awareness can, pace the Cartesian,
amount to knowledge of objective and third-personally determinable facts about his
thoughts and the significance of his talk. The trouble is that if we stress the public side
of meaning and mind too much it is hard to see how they can “sustain immediate first-
personal avowal”.10 The publicity of meaning certainly makes it possible for other people
to know what we mean. However, fixing meaning though outward, social facts “make[s]
a mystery of the phenomenon of non-inferential first-person knowledge of past or present
meanings”.11 So just what does make the publicly accessible, objective facts about some-
one’s mind and meaning immediately and effortlessly available to that person?
The problem is acute and Wright’s discussion of Wittgenstein does much to bring it
to the fore. Wittgenstein removes the barrier to knowledge of others’ minds and
meanings erected by the Cartesian conception of the mental. However, in doing so
he leaves us with the problem of accounting for the special knowledge we have of our
own minds and meanings. This is the unfinished business of Wittgenstein’s rejection of
Cartesianism, due to a picture of meaning that stresses the outer and the public over the
inner and the private. The objectivity of meaning is not established by appeal to a
Platonic standard of use external to the mind of the speaker, but by appeal to a realm of
publicly recognizable facts. The doctrine of meaning as use makes third-person
knowledge of what speakers mean by their words unproblematic, but it leaves it
mysterious how speakers can have immediate first-person knowledge of the same
thing. Any satisfactory episetmology mind and meanings will have to do justice to
both their first- and third-person dimensions. It will have to explain how meaning can
be independent of speakers’ opinions or mere inclinations to use and yet still have an
immediate first-person epistemology.
The theme here chimes in with a central aspect of Wright’s work: namely, to make
out objective notions of meaning and truth that meet certain epistemic constraints.
The particular case of rule-following echoes a Dummettian constraint that a fully
9 10 11
See Wright 1986, p. 173. Ibid., p. 206. Ibid., p. 175.
satisfactory theory of meaning must be at the same time a theory of understanding.

What a word means is what a speaker who understands it knows. So whatever confers
meaning on a word, the meaning conferred has to be within the mental reach of the
subject, and in reach in a way that makes his effortless understanding of the word
intelligible. For Wright the problem is to reconcile the objectivity of meaning—its
having a life beyond whatever impressions of meaning speakers have of it—with
speakers’ having immediate and authoritative first-person knowledge. But how can
someone can enjoy first-person, authoritative knowledge of an objective and other-
wise third-personally knowable subject-matter? This is the problem at the heart of
the essay.
In the end, I will reject Wright’s solution since his attempt to account for “subjects’
non-inferential knowledge of their own meanings” relies on a resolutely deflationary
reading of their first-person authority. It is surprising, that having done so much to
articulate the issue of immediate and authoritative knowledge, Wright ends up dis-
avowing the first-person perspective that made the problem so gripping in the first
place: “There is no essential inner epistemology of rule-following.”12
First-person knowledge of meaning and psychological self-knowledge are not a
matter of our being epistemically in touch with some essentially inner states. They are
judgments or pronouncements that carry a certain authority lacking in the claims other
people make about us. This authority amounts to a presumption that these pronounce-
ments are usually right.
The goal is to explain the correctness of the avowals or pronouncements speakers are
typically inclined to make about what they mean and think. Wright’s aim is to show
how what people suppose to be the case can fail to track some inner fact about their
minds or meanings and yet still be answerable to something for their correctness. We
know that a word’s having the meaning it does for a speaker is not constituted by just
whatever the speaker thinks it means, so we must find suitable constraints for judg-
ments of meaning to meet so that meeting those constraints makes those judgments
true. Such judgments are usually cast by Wright as avowals, and this shift to securing
correctness for publicly avowed first-personal statements in effect discards any genuine
involvement with the inner.
I am more confident than Wright that a re-habilitated notion of the inner, cleansed
of all Cartesian associations, can still be made out and squared with the knowable facts
about another’s mind and meaning. Ironically enough, the account I will give draws
upon materials to be found in other parts of Wright’s work, although, as far as I know, he
has not brought those materials to bear on the issues in question. Let us now establish the
dimensions of the problem and see what elements have to be re-configured in order to
reach for a satisfactory solution.
12
Wright 2001, p. 188.
5. The Reconciliation Problem

The issue is just one part of a general concern with the public and private; in this case,
how to reconcile the inner and conscious dimension of speech with its outer and public
dimension. For if what people mean by their words involves or consists in what they
have in mind when they speak, then how can what someone has in mind—the
meaning she attaches to her words—also be publicly accessible to others on the basis
of her behavior? Is there a notion of linguistic meaning that can do justice to both the
inner experience of comprehension—the first-person point of view of the language
user—and the outer facts of linguistic practice? Let us call this the reconciliation problem.
The problem is striking, and yet most accounts of meaning simply fail to address it
because they assume that the only problem is to establish what words mean in a shared,
public language, where the meanings are the same for all. But of course this way of
proceeding merely postpones the question of how a word gets a communal meaning,
and how each member of that community comes to have immediate knowledge of
that word’s meaning. Casting facts about expression meanings as social facts fails to
touch the issue of how inner and outer aspects of meaning are reconciled. Alternatively,
treating meaning as part of an individual speaker’s psychology may show us how
immediate, inner awareness of a word’s significance is possible but it fails to explain
how that word’s significance can be knowable by others on the basis of their outwardly
observable behavior. Can an immediately available part of our inner lives be part of the
outer, public world of others? This is what makes the reconciliation problem so acute.
I will begin by looking briefly at each side of the problem.
6. First-personal dimensions of meaning

Each of us is immediately and effortlessly authoritative about the meanings we attach to
our words. We know what we are saying or trying to say in speaking, and we cannot be
alienated from the meaning of our own words. We sometimes choose between two
words, trying to decide which to use on an occasion; consciously weighing the subtle
nuances that distinguish their meanings, and at such times the meaning of words seems
to be—as Wittgenstein says—all there at once in the mind—all there at an instant.
When we encounter a word we either know its meaning or we don’t. We don’t wait
to see if we can use it appropriately. In this way each of us is immediately and
effortlessly apprised of what our words mean.
Things are different when listening to the speech of others. I have no special
authority over the significance you attach to your words. Instead, I discover what
you mean by observing what you say, given the circumstances in which you say it.
Thus, there is an asymmetry between understanding what I mean and understanding
what you mean. Understanding what I mean is based on no grounds at all—I just
know—while I base my claim to know what you mean on evidence about how you
use your words.
Each of us is sure of what we mean by our own words, but how can we be sure what
someone else means by their words? And yet, surely if anything unites the public and
the private aspects of the minds of different individuals it is language. What is more,
understanding what someone else is saying often appears to be just as easy as under-
standing what I am saying, despite the lack of any guaranteed authority over what they
are saying. How can this be?
7. Third-personal dimensions of meaning

By constrast with ordinary speakers, philosophers usually focus on theories of the
behavior of a given linguistic community. They observe language from an external,
third-person point of view. Classic examples of this approach are Quine’s Radical
Translator, and Davidson’s Radical Interpreter, who presuppose no prior knowledge
of the language under study. Although ordinary speakers wield no such theories of
translation or interpretation these accounts are intended to reveal the limitations of
what speakers could know of the meaning of their expressions. As Davidson puts it:
The semantic features of language are public features. What no-one can, in the nature of the case,
figure out from the totality of the relevant evidence cannot be part of meaning.13
What makes this conclusion appear plausible is the assumption that the theorist and
the speakers are (to begin with) in the same predicament: having nothing more to go
on in their attempts to understand one another than the behavioral evidence speakers
produce in observable conditions. But how can understanding a language by means of
a theory help to explain how speakers understand a language without one? What we
need is an account of the speaker’s understanding of the language.
What should such a theory of understanding look like? According to Dummett:
What we want to arrive at is a model of that in which our understanding of our language consists,
a model which will be adequate to explain the entire practice of speaking the language.14
To explain the entire practice of speaking and understanding a language would

require a very complex model, dealing with the syntax, semantics, and pragmatics of
language use.15 But however the various factors fit together, everyone agrees there
must be something that fixes the specific contribution linguistic expressions make to
13
Davidson 1984, p. 235.
14
Dummett 1976, p. 133.
15
Those who place greatest emphasis on semantics aim to provide a systematic account of the way a
sentence’s meaning is determined by the meaning of its semantically relevant syntactic constituents given the
role they play in the logical form of the sentence. Those who place more emphasis on pragmatics will try to
uncover the pragmatic processes or inferences hearers rely on to arrive at interpretations of speakers’
utterances of sentences in a given context.. The precise role that context plays in enabling us to understand
what is said by a speaker in uttering a sentence in a context is highly controversial and I shall not take a stand
on it here, save to say that all approaches rely on some knowledge of word meaning. See Bach 2001, Carston
2002, Recanati 2001, Sperber and Wilson 1986, Stanley 2000, and Travis 1997.
sentence or utterance interpretations,16 and a story will be needed about how speaker-
hearers recognize that contribution. Equally, all accounts will have to acknowledge the
role syntax plays in determining which arrangements of words are sentences and the
role certain syntactic features play in the semantics of complex expressions. Speaker-
hearers can produce and recognize indefinitely many new expressions and sentences
just as easily as they produce and understand those they have already heard or uttered,
which suggests that the same underlying system is at work in their understanding of old
and new cases. To be able to recognize indefinitely many utterances of new sentences
one must rely on knowledge of words and knowledge of grammar (as well as the
phonological knowledge needed to recognize the phonemes that make up words, and
the pragmatics required to understand the utterances of sentences in contexts). Knowl-
edge of word meaning tells us the things our sentences are about, and knowledge of
grammar tells us which arrangements of our words are sentences.
8. Understanding: Knowledge of Word

Meaning and Knowledge of Grammar
There is a considerable difference between the kind of knowledge and the objects of
knowledge in these two cases. In the case of grammar, there is a well-understood
domain of syntactic facts illuminated by a welter of research findings about which there
is considerable theoretical agreement. The controversy is about whether speakers have
knowledge (albeit tacit) of the syntactic properties and generalizations about their
language recorded by generative grammar. By contrast, we have no hesitation in
attributing knowledge of word meaning to speakers: speakers are largely authoritative
about what their words mean. The controversy here is about the nature of the mean-
ings they know. No consensus exists among philosophers or linguists about the nature
of word meanings or how we should theorize about them. Unlike grammar, we don’t
even know what the primitives of a suitable theory of word meaning should look
like.17
What are we to make of this asymmetry? Speakers are presumed to know what they
mean, but the objects of their knowledge remain obscure. The facts of grammar are
relatively clear but the nature of speaker’s knowledge of those facts remains problem-
atic. I am less troubled than some about how we can attribute knowledge of grammar
to a speaker. (I deal with these issues elsewhere and will not repeat them here.18) More
16
Even if, as some believe, expression meanings are modulated in the course of pragmatic processing.
See Recanati 2004 for an account of how word meanings are modulated in context to fix their extensions
and thereby the contribution they make to the truth-conditions of utterances of sentences containing
them.
17
For examples, are word meanings complex or atomic, are they constituted by mental representations in
the minds of speakers, are they conventions for the use of public symbols, or denotation relations between
symbols and objects? There is simply no agreement about where to begin.
18
See Smith 2001, 2006a, 2006b.
troubling, however, is sceptical threat to meaning-facts.19 The way forward, I suggest,

is to see a theory of language as a theory of the speaker’s knowledge of language,20 and
to see facts about speakers’ knowledge of word meaning as determining facts about
word meanings. A full story will need to reflect the systematic connections between
meaning and form, showing how knowledge of word meanings and knowledge of
syntax interact to contribute to our understanding of complex expressions. The
worry, however, is how the facts about knowledge of meaning can determine the
facts about meaning when knowledge should be answerable to the facts it concerns.
Another worry is that if first-person knowledge of meaning consists in facts about an
individual speaker’s psychology, how can the facts it determines be knowable to
others on the basis of the speaker’s publicly observable behavior? Let us begin with
the second worry by looking at the constraints on meaning imposed by its allegedly
public character.
9. Motivating the Publicity of Meaning

A key to understanding the Reconciliation Problem is getting clear about the publicity
of meaning claim. What does it require? The core is that we cannot mean anything by
our words unless others could know the meanings we attach to them. To mean
something requires more than one’s having the subjective impression of significance.
Another way to understand publicity is as a claim about what meaning must be like in
order to be detectable from observation of people’s linguistic behavior. The idea here is
to draw metaphysical conclusions about the nature of meaning from epistemic premises
about how meaning can be knowable third-personally.21 The relevant knowledge
claims are usually based on what can be observed by a theorist. So W. V. Quine, for
example, tells us that meaning can only be as determinate as translation between
languages since meaning is what is preserved by good translation, and the choice of
translation manual for a language is fixed by the totality of behavioral evidence
available to a field linguist with no prior knowledge of the language in question.
Moreover, Quine’s view is that since this is all the evidence available to a child learning
the language, the child and the theorist will be in the same position. Hence, there can
be no more to the meaning of words than the child or theorist could glean from
observing the available evidence:
Language is a social art which we all acquire on the evidence solely of other people’s overt
behavior under publicly recognizable circumstances. Meanings, therefore, those very models of
mental entities, end up as grist for the behaviorists’ mill.22
19
Here I am thinking about Quine’s indeterminacy of translation and Davidson’s inscrutability of
reference theses.
20
Dummett and the Chomsky are in full agreement on this point.
21
I have been much aided in this section by Pagin 2000.
22
Quine 1969, p. 26.
In psychology one may or may not be a behaviorist, but in linguistics one has no choice . . . We
depend strictly on overt behavior in observable circumstances . . . There is nothing in linguistic
meaning beyond what is to be gleaned from overt behavior in observable circumstances.23
Claims about the nature of meaning available to speakers must be tailored to fit the
evidential predicament of the best-placed theorist and learner. And yet, there is another
kind of knowledge: the knowledge of meaning available to speakers from the first-person
perspective. The trouble is that conclusions drawn exclusively from the theorist’s perspec-
tive cannot readily accommodate the first-person character of linguistic understanding.
Publicity and the first-person

Is there an inevitable tension between self-knowledge and the publicity of mind and
meaning? It may appear that the public side to mind and meaning threatens the
possibility of self-knowledge, and that we must either side with Quine or abandon
publicity. But appearances to the contrary, the external dimension ensures that wheth-
er a subject is in a given mental state, or means such and such by his or her words, is an
objective, or, at any rate, inter-subjective, matter, answerable to more than just the
subject’s opinions. Thus the possibility of third-person knowledge serves as a require-
ment on a satisfactory account of self-knowledge and knowledge of meaning. It ensures
that the inner world does not stand alone, removed from the rest of reality. Our mental
states and meanings can be credited with an objective reality beyond our immediate
impressions of them. Even facts about a speaker’s idiolect are publicly determinable
facts studied theoretically by linguists from the third-person point of view. (We should
not confuse a language being the language of an individual with its being private.) The
hard problem is that of explaining how one’s immediate impressions (the way things
seem) can amount to knowledge of objective, empirical facts (the way thing are)—one’s
being in certain publicly determinable states of mind, or expressing publicly determinable
meanings. The existence of the interpreter’s third-person perspective secures some
objectivity for a person’s claims about his psychology or language. What a fully satisfying
account of psychological self-knowledge, linguistic and otherwise, would have to
explain is why opinions about our own psychological states and meanings, not arrived
at by means of interpretation (i.e. not theoretically generated), are nonetheless answer-
able, for their correctness, to how we can be interpreted by others; that is, to the
deliveries of interpretation theory. Let us now return to the publicity claim itself.
Publicity and the exteriorizing move

Meaning something by one’s words cannot be a private affair: meaning must be
knowable by others. Were it not possible to know what others mean by their words,
it would be hard to see how linguistic communication could succeed (though all sorts
of non-linguistic communication may still be possible). And without communicative
23
Quine 1990, pp. 37–8.
success there can be nothing purely private to sustain my meaning anything by my

words. So lack of communicability threatens to undermine the objectivity of meaning.
The rejection of private meanings leads to the following reading of the publicity of
meaning thesis:
(P) All a person can mean is all he or she can be known to mean by others.24
Although (P) provides a fairly minimal constraint on the notion of meaning, it does
rule out purely private meanings. It must be possible for others to know the meanings
one attaches to one’s words, thus ensuring others play an essential role in these words
meaning anything. Notice, however, that the minimal understanding of publicity
requires only that it be possible for others to know what a speaker’s words mean, it does not
speak of meaning being publicly displayed in facts about linguistic behavior, which is a
more metaphysically loaded reading of publicity. How is this stronger reading arrived
at, and why is it so often run together with the minimal reading?
Here is a possible reconstruction of the transition to the stronger reading of publicity.
If the meanings attaching to words were items residing solely in the minds of individuals,
there would be no knowing for sure what anyone meant by their utterances. Working
out what others mean would be, as John McDowell puts it, “a mere matter of guesswork
as to how things are in a private sphere concealed behind their behavior”.25
Such a picture is utterly hopeless. Constraint (P) insists that meaning must be
communicable, and since purely private associations with a word are not communica-
ble, an expression’s meaning cannot “contain anything which is not manifest in the use
made of it, laying solely in the mind of the individual who apprehends that mean-
ing”.26 And if meaning cannot lie solely in the mind of an individual:
the significance of others’ utterances in a language must, in general, lie open to view, in publicly
available facts about linguistic behavior in its circumstances.27
How do Dummett and McDowell reach this position? From the idea of its being
possible to know the meanings of other people’s words, they advance to a reading of
(P) according to which the meaning of people’s words must be located on the surface of
speech. Let us call this the exteriorizing move.28 The question to ask is whether this
move is warranted. Does denying the essential privacy of meaning requires the stronger
reading of publicity? If it does, the reconciliation problem becomes particularly acute:
for how can meaning be all there in the mind of the speaker if it lies open to view on
the surface of linguistic practice?
24
The principle has to be restricted to suitably qualified others, and as we shall see the qualifications
are important.
25
McDowell 1981, p. 225.
26
Dummett 1978, p. 216.
27
McDowell, 1998b, p. 315. Here, at least, McDowell takes himself to be following Dummett.
28
It would of course by much easier to say ‘externalizing’ but externalism already has an established use.
For the difference between these issues and issues of externalism about mind and meaning, see Smith 2006c.
Notice the constraint in (P) says nothing about what makes the meanings of my
words available to others; a fortiori it does not say that it is due to meanings being present
on the surface of speech. Must someone who subscribes to (P) take this further step?
A good reason for endorsing the exteriorizing move would be if the move to make
meanings public by locating them on the surface of speech represented the only escape
from the hopeless picture of Cartesian privacy just canvassed. But so far we have been
given no reason for thinking it does represent the only escape. Must we assume with
Dagfin Follesdal that:
when we want to understand meaning and communication we should not turn inward, towards
mental states, but outward, to what is publicly observable.29
Many take Wittgenstein’s thesis of meaning as use (PI }43), to lead straightforwardly
to the strong construal of publicity. For if the meaning of a word is its use and use is
conceived in terms of public practice then surely we can locate meaning among the
observable features of the practice. Support for this reading comes from Wittgenstein’s
correlative claim at Investigations }89 that understanding is a not a mental process, and
his insistence that nothing is hidden. But as we have seen, for all his insistence that
meaning is use, Wittgenstein struggles at }138 with the tension felt between the two
perspectives we are trying to reconcile.
Quine insists that observable behavior in observable circumstances is all we have
to go on in working out what others mean. And from there he concludes that
meaning-facts must reside among behavioral facts. Meaning must be located in, or
reconstructed from, what we display to others in our observable behavior in those
circumstances. For Quine, it is evident that meaning cannot transcend use as behav-
iorally described. If it did, how could people learn the language? The reconstruction of
linguistic meaning from behavior amounts to an expression’s stimulus-meaning: the set
of conditions in which a speaker’s assent can be elicited in response to an utterance of
the expression. The notion serves well for the behavioristic study of people speaking
but is woefully inadequate to illuminate the speaker’s point of view.30 We don’t just
use words: we understand them. And we do not understand them by checking on how
we are disposed to use them. Besides, our dispositions are not introspectible.31 I simply
29
Follesdal 1990, p. 98.
30
Quine attempts to account for the speaker’s understanding third-personally, and behaviorally, in terms
of use: “understanding a word consists in knowing how to use it in sentences and how to react to such
sentences”—Quine 1990, p. 58.
31
Dummett, of course, comes close in a number of places to saying that the behavior in which meaning is
manifested amounts to what it is to know the meaning of expressions. However, he also claims that speech is a
conscious rational activity and that only regularities consciously selected count as part of the language. The
thought that abilities may be cognitive abilities with “insides” as well as “outsides” is perhaps what Dummett
is after when he talks of linguistic abilities as being more than merely practical abilities and as having an
ineliminable theoretical component that guides a speaker as to which uses to make of his expressions. See
Dummett 1983, p. 112.
know what a word means, and I know this in advance of any use I make of it.32 How
I am disposed to use a word may be a test of understanding, but it cannot constitute
understanding. If we are to capture the first-person perspective of the speaker we must
appeal not just to use, but the first-person character of linguistic understanding.33
Quine supposes that because all that speakers can show to one another are bits of
behavior—noises and gestures—meaning must be found among these strands of behav-
ior. But why suppose this? Quine assumes that all we have to go on in knowing what
someone means are the details of their observable behavior and its circumstances. But this
is highly contestable. By looking outwards at the verbal behavior of the speaker, Quine
overlooks the active role of the listener, neglecting what is in the listener’s mind and what
he cognitively brings to bear on observable behavior in the course of speech perception.
In focusing solely on what the child confronts when learning the language, Quine forgets
the part played by the child. We need to look not just outward at the behavior speakers
display, but inward at what the child does with the input.
Davidson, like Quine, makes the exteriorizing move; and, like Quine, he takes this
to have consequences for the indeterminacy of reference. However, Davidson’s
version of strong publicity differs from Quine’s. The public features of mind and
language are not reducible to behavior, but are displayed in behavior glossed by
interpretation theory. On Davidson’s story, all a person can mean, believe, and intend
is all that person can be known to mean, believe, and intend by a fully informed
interpreter. A speaker’s behavior is re-described in the course of theoretical interpreta-
tion by imposing intentional descriptions on otherwise physical goings on. Interpreta-
tion consists in the application of irreducible intentional and semantical concepts––
concepts of mind and meaning––to the observable facts about a speaker’s behavior.
The semantical and the intentional, although not reducible to, supervene upon,
behavior.
The epistemology of mind and meaning proposed here is resolutely third-personal
and theoretical. However, Davidson does acknowledge that a question remains about
how a subject can know facts about her mind without interpreting herself. The
question arises because of some noticeable asymmetries between self- and other-
ascriptions of the same mental states. Other-ascriptions are based on evidence and
inference: self-ascriptions are not. Subjects are authoritative about what they are
currently thinking; others interpreting them are not. What explains first-person au-
thority, especially when subjects’ self-ascriptions are not based on evidential grounds?
32
In “What Do I Know When I Know a Language?” (in his 1993, pp. 94–105), Dummett offers the
example of the society hostess in a P. G. Wodehouse novel who is asked whether she speaks Spanish
and replies, “I don’t know. I’ve never tried.” I take it that the absurdity (and pretension) of the reply
contrasts sharply with the normal case.
33
It is interesting to note that despite his behaviorism, even Quine is not free of the tension between first-
and third-person perspectives. There is a tension between Quine’s behaviorist and exterior account of
linguistic significance (extracted though radical translation) and the individual’s foundationalist need to
construct a theory of the world––a web of belief––from the sequences of stimulations of his sensory surfaces.
Davidson’s commitment to publicly knowable facts about mind and meaning

brought to light by the theory of interpretation, makes it difficult to reconcile such
facts with a thinker’s having first-person knowledge of them. However, rather than
facing the challenge, Davidson attempts to play down the reconciliation challenge by
offering a deflationary account of first-person knowledge. Deflationary accounts
usually try to honour some mark of the first-person, such as the first-person authority
of a person’s psychological ascriptions. First-person authority is construed by Davidson
as the presumption of correctness attaching to our own but not to other people’s
pronouncements about our psychological states. This is an important feature of a
person’s present-tense psychological self-ascriptions but it falls far short of illuminating
the first-person perspective of that individual.
Davidson’s theory-driven epistemology forces him to explain why ascriptions of
psychological states not arrived at by means of interpretation, are nonetheless answer-
able, for their correctness, to the pronouncements of interpretation theory. Meeting
this condition—not insignificant in itself —will be the most that deflationary accounts
can do to illuminate the first-personal character of meaning and mind. I cannot go into
the details of Davidson’s view here34 save to say that it relies on the idea that subjects
literally speak their minds, and they know what they mean when they say what they
think. Unlike third parties, subjects do not have to work out what they are saying in
order to know what they think; so in knowing what they mean subjects are at an
advantage over others. However, Davidson omits to tell us how subjects know what
they mean without interpreting their own speech. Thus he stops short of a fully
satisfying account of self-knowledge.35
Crispin Wright, too, favors a deflationary treatment of linguistic and psychological
self-knowledge:
knowing of one’s own beliefs, desires and intentions is not really a matter of “access to”—being
in cognitive touch with—a state of affairs at all.36
Nevertheless:
Intentional states are avowable: they are subject to groundless, authoritative self-ascription, and it
belongs to their essence that they are so.37
Thinkers voice their opinions with an authority not due to any special contact they
have with their mental states. First-person authority is the presumption of correctness
34
See Smith 1998.
35
In effect, Davidson opts for a very deflated notion of linguistic knowledge, preferring to replace the notion
of a speaker’s knowing what she means with the notion of her meaning what she says. And since she could not be
counted as meaning what she says unless her speech was interpretable, it is validation from the theorist’s third-
person point of view that underwrites her claim to know what she means, and thus be correct in judgments about
what she thinks.
36
Wright 1989a, pp. 631–2.
37
Wright 1989a, p. 629.
such avowals carry and others’ psychological ascriptions to them lack. The philosophi-
cal task is to explain this presumption of correctness.
Wright’s deflationary answer is to suggest that these self-pronouncements occupy
pride of place among the data we draw upon to interpret people psychologically. Such
pronouncements will be taken as true, by default, just so long as granting people
authority in what they say about their current mental states serves to render them more
intelligible than not doing so, and there is no reason to reject what they say.38 Of
course, the default position can be overridden if there is an interpretation of the subject
that makes better overall sense of his behavior, in rational terms, by overturning or
discounting his self-pronouncements.39 Thus self-pronouncements will typically com-
port with the ascriptions of attitudes and meanings the interpreter is prepared to make
to that person anyway to explain their acts and utterances.40 This essentially third-
personal condition on the truth and epistemic security of a subject’s avowals ensures
first-person authority, but:
The authority standardly granted to a subject’s own beliefs, or expressed avowals about his
intentional states [and meanings] is a constitutive principle: something which is not a conse-
quence of the nature of those states [or meanings] and of an associated epistemologically
privileged relation in which the subject stands to them, but enters primitively into the conditions
for identification of what a subject believes . . . intends [and means].41
we do not cognitively interact with states of affairs which confer truth upon our opinions
concerning our own intentional states [or meanings] rather, we are, as it were, inundated, day
by day, with opinions concerning our own intentional states [or the meanings of our utterances]
for which truth is the default position.42
The problem is that while this story shows how to account for the correctness of a
subject’s self-pronouncements, in no real sense does it show, from the subject’s
perspective, what it is to be self-knowing: that status is conferred on the subject by
an interpreter. Wright’s deflationary story is thus not an account of what it is for a subject
to know what he thinks or what he means, rather it is an account of why certain
pronouncements he makes are true. On this picture, subjects simply voice opinions
about their own mental states for which truth is the default assignment. Such avowals
count as correct by the interpreter’s lights. But nothing here shows the holder of those
opinions to be in the position of knowing his own mind. That his self-opinions are true
38
“ . . . our opinions concerning our own intentional states . . . count as true provided that we hold them
and no good purpose is served, in another’s quest to find us intelligible, by rejecting them”—Wright 2001.,
p. 313.
39
Wright offers two slightly different explanations of why this should be so. On the one hand he talks
about it as a deep contingency that interpretations that respect what a subject says about himself by and large
give better service in making sense of him than interpretations that leave them out or overturn them. On the
other hand, he speaks at times about it being an a priori constraint on interpretation that one take into
account and respect, as far as possible, a subject’s self-conception.
40
See Wright 2001, p. 314, where he identifies “the correctness conditions of ascriptions” with “their
interpretative utility”.
41 42
Wright 2001, p. 312. Ibid., p. 313.
is a necessary condition of being self-knowing but it does nothing to illuminate that

predicament. The deflationary strategy does not solve the reconciliation problem but,
rather, downplays one of the two perspectives we are seeking to reconcile. How
satisfactory is this strategy?
Not only do subjects’ self-pronouncements contribute to determining the mental
states subjects can be said to be in, by contributing to the evidence an interpreter
relies on to make appropriate attributions, but, for Wright, a subject’s judgments
about what he believes, desires, or intends, when made under cognitively ideal
conditions, determine the truth of his claims. His best opinions about his current
mental states provide default sufficient conditions for his being in those states. Thus
there is a standard of correctness for such self-ascriptions: only best opinions count.
However, correctness does not consist in tracking independently constituted states;
other externally validated self-ascriptions enter into the constitution of the mental
states they ascribe. 43 A virtue of subjects’ judgments determining their mental states
could be that it may cast light on how knowledge of meaning could determine
meaning.
But what prompts subjects to issue self-ascriptions? What equips one with the
wherewithal to make them, and how does making them amount to knowing what
one is thinking? To be told that when they are issued under the cognitively ideal—
independently specifiable—conditions these self-pronouncements will count as cor-
rect44 offers us little to go on. Pronouncements simply issue from the subject’s lips with
nothing to guide them but causal forces: “we are ceaselessly, but subcognitively moved
to produce them”.45 We produce them blind: being “inundated, day by day, with
opinions for which truth is the default position”. But how does the subject understand
the opinions he ceaselessly voices? We are not told.
Wright entertains the idea that this apparatus may be used to explain a subject’s
opinions about what his words mean, or which rule he follows. This does not require
conceiving of meaning as a species of intention, “it is enough that the concepts are
relevantly similar—that both sustain authoritative first-person avowals”.46 To know
what I mean by a word:
A sufficient answer need only advert to my present opinion that [such and such] was what
I formerly meant, and still mean, and to the a priori reasonableness of the supposition, failing
evidence to the contrary, that this opinion is best.47
43
Remember a subject’s condition provides at best a default sufficient condition for being in an intentional
state. So how could such states obtain anyway, without their possessor’s judgments that they are in them and
still be partly constituted by a subject’s judgments, when, under ideal conditions, he judges himself to
be in such states? How could the same state be there with and without the constitutive contribution from
the subject’s judgment? I shall not pursue this odd feature of the account.
44
In this case, it is not a priori true that the subject’s opinion is best, but, failing evidence to the contrary, it
will be a priori reasonable to think the subject’s opinion is best and therefore settles the matter of what he is
thinking. See Wright 2001, pp. 204–5.
45 46 47
Ibid., p. 313. Wright 2001, p. 206. Ibid.
And since we are not tracking a Platonic standard for meaning, the speaker’s best
opinion will be all there is to settle the truth about what he means on that occasion.
Unlike subjectivist treatments, meaning is not settled by whatever opinion a subject has:
it is best opinions that settle the facts about meaning, and there are conditions to be met
for opinions to be best so there is room, here, for an is/seems distinction.48 Perhaps such
an account can cast light on what a speaker making self-pronouncements takes himself
to be up to, and thus how he comes to know what he is thinking. However, Wright
holds out less hope for an account on which best-opinion fixes what one’s words mean
since “certain elements of convention, which have no counterpart in the case of
intending in general, sustain my ability to mean anything in particular by a word”49
and it is far from clear how my immediately issued opinions can take in or reflect these
‘conventional (presumably social)’ facts about the meaning of words in the public
language. Wright’s problem here is that he is saddled with an account of word meaning
as belonging to a shared, public, and conventionally governed language—and why
should an individual’s immediate and spontaneous opinion determine these facts?
Without this assumption the way would cleared for a proposal along the lines Wright
suggests. Perhaps he thinks an appeal to social facts about the public language offers us
the only way to avoid Platonism and Cartesianism. However, ditching the commit-
ment to a shared public language in favor of (non-private) idiolects would give
Wright’s strategy more chance of success since shared conventions would play no
role in sustaining my ability to mean something by my words. Wright’s resistance to
such a move is fuelled, I suspect, by doubts about whether there is any non-Cartesian
conception of idiolectical meaning to be had. Thus he signs up for the idea that it is
only fully public and indeed communal facts that could sustain one’s meaning some-
thing by one’s words; a view that comes perilously close to the exteriorizing move and
which makes a solution to the reconciliation problem so remote.
The deflationary treatment of linguistic and psychological self-knowledge offered by
Wright and Davidson explain the presumption of correctness for self-opinions but cannot
accommodate the first-person perspective of the language user. Let us now turn to
accounts that give due weight to the first-person character of linguistic understanding.
10. Interiority and the Exteriorizing Move

John McDowell repeatedly stresses the inner dimension of understanding:
Knowledge of meanings is wholly a matter of how things are in a subject’s mind.50
command of a word’s meaning is a mental capacity . . . the mind [is] the locus of our
manipulations of meanings . . . 51
48
Just so long as the conditions are independently specifiable.
49
Wright 2001, p. 206 fn 40.
50
McDowell 1998b, p. 36.
51
But as we have seen he also embraces the exteriorizing move when granting the
publicity of meaning:
the significance of others’ utterances in a language must, in general, lie open to view, in publicly
available facts about linguistic behavior in its circumstances.52
How, then, can McDowell reconcile (what he would call) the interiority of meaning
with its being publicly available to others in observable facts about language use? The
exteriorizing move seemed to leave us with a dilemma: on the one hand, the bare
behavioral facts cannot sustain genuine attributions of meaning; on the other hand, our
spontaneous understanding of speech is nothing like the bringing to bear of a theory on
the available data. McDowell attempts to steer a middle course between these two
unpalatable options by rejecting the conception of a bare surface to speech—a con-
ception of still to be interpreted facts, common to both horns of the dilemma.
Linguistic understanding is not the upshot of a theory, because we are not confronted
with mere bits of uninterpreted behavior. Quine’s picture of the surface of speech
distorts the facts by describing them in hopelessly impoverished terms. This is not how
they strike us. As McDowell points out, when listening to speakers talk:
Our attention is drawn to . . . something present in the words––something capable of being heard or
seen in the words by those who understand the language.53
Command of a language equips us to know one another’s meaning without needing to arrive
at that knowledge by interpretation, because it equips us to hear someone else’s meaning in his
words.54
Since meanings are immediately accessible to us and yet publicly available also, they
must also lie open to view on the surface of speech. Thus the behavior that confronts us
needs to be characterized richly in content-using terms. In effect, we must presuppose
command of the language in question. This is, for McDowell, the quietist assumption
that we should not expect anything more than a modest theory of meaning, one that
merely reminds us of what we already know about the language. Such a manoeuvre
exteriorizes meaning without reducing it to mere behavioral constructs, physically
described.55 McDowell thinks Quine is right to insist on the public and external nature
of meaning but that he simply mischaracterizes what is, so to speak, out there in
excessively meagre terms.
Does this revised view of publicity meet the reconciliation challenge? Remember,
meaning belongs to the interior dimension of mind. But do we directly perceive
minds, or aspects of minds, as parts of the world? This picture of observable minds
52
53
McDowell 1987, p. 99 italics mine.
54
McDowell 1984, pp. 350–1.
55
Despite other drawbacks, I take it to be a virtue of Wright’s proposals that although he thinks with
McDowell that we cannot reduce meaning to a mere behavioral construct, he does think that attempts should
be made to explain what settles the facts about meaning.
receives some support from the doctrine of semantic externalism. The fact “that a
speaker means what she does by ‘water’ must be constituted at least in part by her
physical and social environment”. This is not just an externalist thesis about meanings:
“Meanings are in the mind but as [Putnam’s] argument establishes, they cannot be in
the head; therefore, we ought to conclude, the mind is not in the head.”56 So
recognizing others’ meanings can be recognizing what they have in mind, and this
need not be hidden from us in some private or internal sphere:
the outward aspect of linguistic behavior is essentially content-involving, so that the mind’s role
in speech . . . is on the surface.57
McDowell’s picture appears to offer us a way of combining the phenomenology of

understanding with the exteriorizing of meaning; the behavioral means of transmitting
meaning with the immediate phenomenological character of understanding. It recon-
ciles inner and outer perspectives, though at the cost of an extravagant metaphysics of
mind. Does it give us everything we want? Not quite, and the reasoning is dubious.
The questionable move is to go from the phenomenological recognition of imme-
diately intelligible speech to the claim that we are directly confronted with an outward
surface replete with meanings. To be sure, we hear certain sounds linguistically, as the
articulation of words and the utterances of sentences, and yet to hear sounds as
meaningful speech does not show that meanings are in the sounds.58 The phenome-
nological observation is spot on: we do not experience speech in a familiar language as
just noise (as we do when listening to an utterly foreign language). Nor does speech
strike us as bare verbal behavior. We cannot but hear certain people’s talk as meaningful
when they utter words we know. But McDowell intends the phenomenological claim
as an epistemological one, taking these experiences at face value: our hearing meanings
as there in people’s speech. We hear meanings out there (where?) on the surface of
speech. We perceive them directly. Detailed questions about where experiences of
speech occur, or what is meant by speech’s surface are never addressed.59 What is it for
meanings to be part of the external world in this way? The problem is to understand
how meaning can be a phenomenological notion and yet public: something directly
56 57
McDowell 1992, p. 276. McDowell 1987, p. 101.
58
Notice that McDowell talks about hearing meaning in words. Words are different from sounds: words
can be written, spoken, or signed. Words are relatively abstract and can only be identified when they carry
their full freight of semantic, syntactic, and phonological features. However, if he wants to say we hear
meaning in words, he must mean we hear them in speech sounds.
59
A moment’s reflection should tell us that it is not so easy to locate words. Are the words people use on
their lips? Where should we locate the sounds that articulate those meaningful words? In the air, in our ears?
Speech sounds don’t seem to travel towards us. Nor do they seem to be just on the speaker’s lips. And words
don’t seem to be straightforwardly present in the behavior we observe. Words have different pronunciations;
the same word can be written, spoken, or signed. Perceiving linguistic behavior as the articulation of certain
words is a tricky business: missing phonemes can be ‘restored’ in the percept the hearer’s mind forms in
response to the somewhat deficient sound. McDowell’s metaphor of a perceptible surface radically under-
estimates the difficulties here. For a more sustained critque of McDowell’s view of speech sounds see Smith
2009.
heard or seen in the linguistic activity of others. Can we directly observe aspects of the
inner lives of others? Are conscious phenomena directly observable features of the
world? Gregory McCullogh found this unproblematic:
we should maintain that meanings, our own as well as others, can figure as integral components
of our conscious life. Moreover, these are indeed public features of the use of language, since the
point is that public matters like speech, interpretation and communication are themselves
fundamentally conscious phenomena. What I think is often what I put into words; and suitable
audiences frequently hear that same thing.60
Meaning figures in phenomenology not only in the first-person case, where one is speaking
consciously, but also in the third-person case when someone is speaking and one is aware of what
they are saying . . . It figures there directly, unmediated.61
Questions remain about how illuminating or cogent these suggestive remarks are. It is a
big step to suppose phenomenological entities can be literally out there on the surface
of behavior.
Notice that for McDowell, unlike Wright and Davidson, there is no asymmetry
between the first- and third-person perspectives on meaning. The word meanings we
know in a specially immediate and authoritative way, are the very same meanings we
hear in the speech of others.62 They are the meanings the words have in the public
language sustained by the communal practice. It is by coming to learn a language that
subjects come to have these meanings in mind and are able to display them to one
another in their speech behavior.
McDowell insists that the conception of meanings as out there in the world need not
seem so strange once we embrace a conception of the world as re-enchanted. However,
anxiety about this seemingly magical picture of external reality replete with extended
minds and observable meanings is not relieved by replacing the word “magical” by the
word “enchanted”. We do not need to follow McDowell here unless the picture he
offers us provides the only way to solve the Reconciliation Problem. But that is not
the case. Rejecting McDowell’s epistemology of meaning need not mean rejecting the
phenomenology. Nor does it require us to settle for psychologism.63 A key aspect of
McDowell’s phenomenological insight is that we hear more in the speech sounds of a
language we understand than in one we don’t. When listening to an utterly foreign
language all we hear are continuous sounds with no perceptible word boundaries. So
what enables us to hear meaningful words and sentences when listening to a language
we understand? McDowell admits that the meanings present in a speaker’s words are
not perceptible by just anyone: they are perceptible only to those who understand the
60 61
McCulloch 1998, p. 265. Ibid.
62
Doubts about this idea can be sown by considering our experience with speech synthesizers, where we
have the same immediate experience of understanding—of hearing sounds as meaningful speech—without
being entitled to say that meanings are out there on the surface of those events.
63
By ‘psychologism’ I mean here a view opposed by Frege and Wittgenstein; the former because it
threatened communicability and the objectivity of meaning; the latter because it flirted with an incoherent
notion of our inner lives as insulated from the outer world.
language in question. This is a vital concession, and the significance of it tends to go

unnoticed. If we need knowledge of the language in order to perceive the relevant
meaning-facts, we are entitled to ask how linguistic knowledge contributes to the
character of linguistic experience. In what way does knowing a language make the
meaning-facts come to be perceptible? What does our knowledge of a language consist
in, how is it acquired and how is it exercised? McDowell is somewhat reticent about
these issues. He treats knowledge of language as simply a perceptual capacity, but it is
surely the same knowledge we exercise in speaking. Knowledge of language features in
both our abilities to produce and comprehend speech, so should not be identified
exclusively with either.
If linguistic facts lie open to view on the surface of a practice, acquiring knowledge
of a language is acquiring a capacity to pick up on the detectable features of language out
there. But what makes such features detectable to the learner in the first place? The best
McDowell can do is to talk of training in a behavioral repertoire that (somehow) brings
into view the meaningful surface of speech. He asks: “How can drilling in a behavioral
repertoire stretch one’s perceptual capacities––cause one to be aware of facts of which
one would not otherwise have been aware?”64 How can the outer world of meanings
come into view by our being inculcated into the behavioral practices of the speech
community? Unsurprisingly, McDowell doesn’t have an answer to this vital question,
though he badly needs one. The suspicion surely is that we won’t get an answer because
we are asking the wrong question. It is not behavior that induces the form and character
of our linguistic experience but the special mental organization of language learners: in
particular, the largely innate faculty of language that enables them to acquire any natural
language from a certain course of experience within a given developmental time period.
Eventually, children come to hear people’s words as meaningful, and are able to
recognize and produce grammatical sentences, but any account of how they do so has
to be compatible with facts about language revealed to us through empirical study—
facts that McDowell simply ignores.65 Facts about the grammatical structure of sen-
tences and even word boundaries cannot be located in the sounds (acoustic signals) that
provide the actual surface of speech, and only certain linguistic properties are phenom-
enologically accessible in our linguistic experience of speech. The key issue is where we
are to locate the linguistic properties responsible for the rich contents of our linguistic
experiences. Do the linguistic percepts we form when listening to speech go beyond
information present in the environment? The best empirical theories we have tell us
they do. The grammatical structures of the sentences heard are not linear but hierar-
chical. The structures contain phonetically null elements, empty categories like PRO
and trace, and bound variables, postulated to explain the generalizations about quanti-
fication, scope, reference, and anaphoric connections that speakers conform to in their
comprehension of speech. Consider the following strings:
64
McDowell 1998b, p. 333.
65
Or that he gets badly wrong on straightforward empirical grounds. See Smith 2009.
(1) Peter promised John to leave

(2) Peter persuaded John to leave
(3) Peter was persuaded to leave
(4) Peter seemed to think John promised to leave
(5) John seemed to think Peter persuaded him to leave
Our intuitive understanding of them reveals that ‘Peter’ is the intended subject of ‘to
leave’ in (1), ‘John’ is the intended subject of ‘to leave’ in (2), and although no overt
item appears in ‘subject’ position in (3), that again it is clear that ‘Peter’ is the intended
subject of ‘to leave’. The syntactic structure of (1) to (5) can be represented as follows:
(1*) Peter1 promised John2 [PRO1 to leave]
(2*) Peter1 persuaded John2 [PRO2 to leave]
(3*) Peter1 was persuaded [PRO1 to leave]
(4*) Peter1 seemed t1 to think John2 promised PRO2 to leave
(5*) John1 seemed t1 to think Peter2 persuaded him1 to leave
Which referential expression or arbitrary element in (3*), controls the empty
category ‘PRO’ depends on the syntax of the construction. We can see that ‘John’ is
not the grammatical subject of ‘to leave’, PRO is, as we can see: if we replace ‘John’
in (2*) by a pronoun which would have to be the accusative ‘him’ and not the
nominative ‘he’, as we see in it (5*). Notice that ‘Peter’ and ‘John’ are not the logical
subjects of ‘seem’ in (4) and (5). They do not do the seeming. Rather, what they think
is in the scope of its seeming to be the case. Thus, the grammatical subjects of (4*) and
(5*) are semantically interpreted as though they were in the position of their traces t,
which mark the extraction sites from which they were plucked to provide grammati-
cal, though not logical, subjects for the sentences containing them.
These elaborate syntactic structures, remote from surface form, do not line up with
the contours we find in the outer world, nor do they appear in the phenomenological
percepts we generate in speech perception. Nonetheless, it is because similar strings
have these subtly different syntactic configurations that speakers respond to them
differently. Notice, speakers can respond differentially to the same (structurally ambig-
uous) string: ‘John saw Peter with the binoculars’, first hearing it one way then another.
Are they perceiving different external but coinciding facts, or is it similar to our
experience with Necker’s cube, where switching occurs internally as a matter of
how we represent the structure of the figure? There is no hope of locating the syntactic
properties needed to characterize natural languages in external features of the environ-
ment,66 nevertheless, these properties do real work in explaining and predicting how
speakers will understand particular word strings. Given that speakers have little or no
conscious knowledge, and not even accurate reflective knowledge, of the syntactic
properties of their language, the only way such properties can shape our linguistic
66
See Smith 2006a and 2006b.
experience is if they are aspects of the subpersonal mind that affect the form and
character of conscious linguistic experience. Notice, that not only do the legitimate
configurations of syntactic elements capture generalizations speakers conform to in
their intuitive linguistic judgments, they also explain what speakers cannot hear strings
as meaning. For example, (6) is two ways ambiguous depending on whether we treat
‘duck’ and ‘swallow’ as verbs or nouns:
(6) I saw her duck and swallow
But why not four ways ambiguous? After all we can theoretically conceive of the
other two readings. But we cannot consciously hear the string in either of those other
ways. The explanation appears to be that the language faculty is organized so that the
structures it assigns to strings satisfy a Coordination Constraint, permitting the con-
junction of noun-phrases with noun-phrases, verb-phrases with verb-phrases, but not
noun-phrases with verb-phrases, or vice versa. That our internal system is so config-
ured explains why no such conscious linguistic experiences are possible. Linguistic
internalists will insist that the demand for the publicity of language creates irresolvable
difficulties in accounting for the experiential linguistic character of perceived speech,
and as such it must be rejected in any adequate scientific study of language.
11. Language as Internal to the Mind

The exteriorizing move fails to locate the essential properties of language, including
those of linguistic structure, in the environment. By contrast, generative linguists
will locate all properties of language—except its phonetics—inside the mind of the
speaker.67 Clearly, the internalist conception, advocated by Noam Chomsky, is at odds
with the folk view of languages much beloved by philosophers. The folk notion sees
languages as out there, external to speakers’ minds, something we avail ourselves of in
order to express and communicate our thoughts to others. But the view makes little
sense in any serious study of language. All there is out there are sounds and marks, as we
readily appreciate when we encounter an utterly foreign language. Sounds are per-
ceived as linguistic items only if speakers’ internal linguistic systems assign phonologi-
cal, syntactic, and semantic properties to them. And which linguistic properties sound
tokens are taken to have—how those tokens are typed—depends on the competence
and performance of the speaker-hearers who produce and respond to them. Without
such inner states, people’s speech would be heard as a series of continuous sound
signals. Sounds and marks have linguistic significance for us because we are able to hear
or read into them the linguistic information we bring to bear on them. Linguistic
significance is not an intrinsic property of speech sounds: there is no external locus of
67
“Language has no existence except for its mental representation [in the mind of speakers]”—Chomsky
1975, p. 169, fn.
linguistic significance. The full set of properties needed to explain it are neither
environmentally nor phenomenologically salient features of speech.
On this internal, mentalist conception of a language, championed by Chomsky, facts
about the meaning and form of expressions are determined by a faculty of language in
the mind of speakers. The sounds and signs people produce are not the subject-matter
of linguistics, but the linguistic forms they impose on those sounds and signs as a result of
their internal states. No remotely plausible case has been made out that linguistics could
concern physical or perceptible properties of the environment. On the internalist view,
the focus of linguistic inquiry shifts from the actual and potential behavior of speakers
to the internal organization of speakers’ minds, as probed by eliciting their linguistic
intuitions. It is “mentalistic, since it is concerned with discovering a mental reality
underlying actual behavior”.68 The linguistic phenomena the listener hears are due to
linguistic entities her internal apparatus supplies in response to speech sounds. To arrive
at the full story of language and knowledge of language, the internalist conception of
syntax has to be reconciled with the minimal publicity claim (P).
One obstacle to this reconciliation is the uncritical adoption of the folk conception
of language. And yet the folk conception fails to explain much that has already been
explained in linguistics. The acquisition of language is shaped by innate principles,
including those like the semantically relevant binding principles69 that govern the
referential dependence of items on one another, constraining the semantic readings we
can give to strings. Such properties are universal features of all human languages. There
is no alternative to accepting these empirical findings.
So if the facts of syntax are not to be found on the behavioral surface of language, in
how it is perceived, why do philosophers persist in supposing that language is public and
external to speakers’ minds? Partly, this is due to Quine’s conviction that the child and the
theorist must be in the same position when first confronting a language, and that therefore
the learner can only have the same external, behavioral evidence to go on as the theorist
studying the language. However, Quine is mistaken about the child and theorist being in
the same position. Chomsky distinguishes the project facing the linguist and that facing the
child acquiring a first language. The linguist is trying to work out what native endowment
the child must be equipped with—what internal structuring mechanisms must be in
play—in order to acquire a natural language on exposure to a particular course of
experience. The child is trying to work out no such thing. The child is simply exposed
to that course of experience and on the basis of its native endowment, maps the experience
it has in listening to the sounds others produce onto an internalized grammar that equips it
to speak and understand. The theorist will try to identify the nature of that native
68
Chomsky 1965, p. 4.
69
The Binding Principles explain facts speakers know such as why in ‘Mary expected to feed herself ’ the
reflexive pronoun is referentially dependent on ‘Mary’, but in ‘I wonder who Mary expected to feed herself ’
it is not. The interrogative pronoun ‘who’ is the displaced element interpreted as if the subject of ‘to feed’ and
so ‘herself ’ is referentially dependent on the trace of ‘who’ occupying that position.
endowment, the factors that determine the choice of resulting grammar, and attempt to
understand the way the child’s resulting grammar grows out of parameterization of the
initial structural properties the language faculty imposes on the primary linguistic data. The
rules of a grammar the theory finitely records specify the potential infinity of knowledge
the speaker has about a language. The theory states what the speaker knows but does not
ascribe knowledge of the theory to the speaker. The linguist’s theory is a model of the
grammar not a specification of the form of its encoding in the speaker. As speakers we
don’t need to work out what structures others are assigning to their expressions, nor to
perform any theoretical inferences from raw speech behavior: our linguistic system works
automatically to enable us to hear certain sound sequences as the utterance of sentences.
The fast and mandatory operation of this system makes it possible immediately to
experience sounds linguistically in the way McDowell suggests. Phenomenologically,
the internalist can agree with McDowell: we simply hear sentences as structured.
Now while this view is undoubtedly right for syntax, where there is no credible
alternative, it is not so obviously right for word meanings, about which Chomsky has
much less to say. However, it is a constraint of a full account of the use and understanding
of language that whatever account of meaning is given, it should be compatible with the
empirically known facts about syntax handled by our subpersonal linguistic system, since
facts about sentence meaning depend systematically on syntactic form. Thus the full
reconciliation story must reconcile first-personal, third-personal, and subpersonal dimen-
sions of linguistic activity. However, it is open to someone, in principle, to propose a
mixed economy, recognizing internal and subpersonal facts about syntax, on the one
hand, and public and external facts about word meaning, on the other. Such a view
would commit us to saying that there is no single locus of linguistic significance. A story
would still be required about how lexical meaning and syntactic structure come together
to determine the meanings of syntactically structured expressions, and how speakers
come to have a unified conscious percept that gave them first-person knowledge of such
expressions’ meanings. A simpler option would be to locate meaning in the conscious
minds of speaker-hearers and try to show how such a view is compatible with the
minimal publicity claim that it be possible for others to know what we mean by our
words. I shall end by providing an outline of such an account.
12. An Alternative Answer to the Reconciliation

Problem
I have stressed the first-personal nature of linguistic understanding and acknowledged
McDowell’s phenomenological insight that we hear words as meaningful.70 But how
70
The first-person nature of linguistic understanding, including its phenomenology, has been emphasized
not only by McDowell 1992 but by Burge 1999; McCulloch 1998; Smith 1992, 1998; Wittgenstein 1953,
}} 138, 185; Wright 1989. Galen Strawson (1996) has even suggested the need to acknowledge the
experience of meaning—not experiences that accompany our grasp of meaning, but the experience of
can we respect that insight and claim that the speaker’s first-person perspective is basic
while: (1) meeting the publicity constraint; (2) resisting the exteriorizing move; and
(3) conceding that objective facts about meaning depend on factors external to the
speaker?
For meaning to depend on external factors and yet be a phenomenological notion—
part of the conscious experience of the speaker—seems to need us to see the mind as
extended, and yet this is mysterious. On the other hand, exclusive focus on the
perspective of the interpreter neglects, at cost, the phenomenology of thought and
meaning. However, restoring the latter requires more than merely adding an extrane-
ous conscious accompaniment to the external manifestation of meaning. The speaker
must be accredited with knowledge of what he means or thinks, knowledge that is
available to him in a way that is not available to others since he knows what he means
or thinks without recourse to the external evidence by reference to which they come
to know what he means. The problem is to explain how he has knowledge of an
objective fact about himself ––a fact publicly discernible by others––in an effortless,
non-evidential way. Admittedly, understanding the speech of others can appear
equally immediate and effortless. But there is an authority we have about what we
mean when we speak which we don’t have when listening to what is another’s
apparently intelligible speech. It remains to be seen whether others do understand
their words in the way we do.
We need to focus on the individual language user’s understanding. The only
meanings individuals can hear their own and others’ uses of words as having, initially
are the meanings they themselves attach to those words. Who else’s meanings could
they rely on? So we need an account of what it is to attach meanings to our words,
and how doing so is compatible with our understanding others, and their under-
standing us.
I take this first-person case—my knowing what I mean by my words—as basic, and
it is this knowledge I rely on to know what you mean. Thus, on the view I’m
espousing, when listening to you speak, I am not the passive recipient of the meanings
I hear in what you say: I supply them. Philosophers have tended to find this suggestion
problematic, but that is largely due to a mistaken conception of the epistemology. I will
show why this picture does not lead to the worries usually leveled at it and why there is
a credible and natural epistemology of understanding that explains the ease with which
we often recognize the meanings of one another’s words.
The speaker’s inner, conscious awareness of meaning constitutes his knowing
what he means and provides the basis for his knowing what others mean. Though
one is authoritative about what one means by one’s utterances, one is not infallible.
meaningful speech itself. Roughly, the experience you have when listening to a language you understand, in
contrast to an utterly foreign language. Even Jerry Fodor (1983) appeals to the phenomenology of under-
standing to argue for the existence of a dedicated language module on the grounds that mandatory processes
operate to ensure we hear words in a familiar language as meaningful and not merely as sounds.
Immediate impressions of an expression’s meaning do not constitute its meaning.

There are illusions of meaning that show failures of judgment about one’s own
competence. As an example consider a sentence like the following:
(7) Many more people have been to Paris than I have.
At first it looks, or better sounds, perfectly reasonable until reflection tells us that there
is nothing that it could intelligibly mean. The individual speaker’s opinion or impres-
sions of understanding will not always triumph. Hence, the facts we lay claim to from
the first-person perspective are not merely constituted by how things seem to us, or
what we take them to be.
Nevertheless, under certain conditions, how the individual takes things to be in his
language is how they are. What are these conditions? In the case of syntax, the
condition is clear. When a speaker’s linguistic intuitions about what he produces or
comprehends are in line with what his underlying linguistic system assigns to a string as
its structure—i.e. he takes the string to be acceptable when it is grammatical, or takes it
to be ambiguous when it can be assigned more than one structure—then he counts on
those occasions as knowing the grammar of the string. He may think he is responding
to grammatical facts about the language external to him, but actually he will be
responding to something in his own breast.71
What are the relevant conditions in the case of word meaning? To answer this
question we need to look at what happens when a language learner acquires knowl-
edge of a word’s meaning. The usual assumption, well expressed by Tyler Burge, sees
word acquisition as a matter of the language learner inheriting the meanings from others
with whom he interacts:
Words are initially acquired from others, who are already applying those words to cases . . . In
many cases, we intentionally take over the applications that others have made. We rely on their
experience to supplement our own. And we accept corrections of our explications from them
because they have better access to the examples which partly determine the nature of our
concepts.72
It is assumed that we somehow take possession of something already there—

something that eluded us at first. But how do we take over the norms governing a
word’s use just by being in contact with another speaker? By which mechanism can this
be explained? All we have is the hugely different case of sophisticated language users
consciously deferring to those from whom they acquired a word. Without that, what
other kind of epistemology of word meaning is in prospect? None is offered. McDo-
well falls back on the ironically unilluminating phrase from Wittgenstein that “Light
dawns gradually over the whole.” However, to make genuine progress we need to
reject the usual picture, and here we are helped by a key insight from Davidson, who,
71 72
See Smith 2001 and 2006b. Burge, 1989, p. 185.
on this occasion, defies the usual assumptions about his overly third-personal account
of meaning and understanding. The idea we need to take advantage of is this:
Someone who is consciously teaching a beginner the use of a word may think of herself as simply
passing on a meaning that already attaches to the word. But from the learner’s point of view, the
word—the sound—is being endowed with a meaning [by the learner].73
Davidson’s insight is that for each speaker it is a matter of him or her investing the
heard item with meaning and not simply recognizing a meaning that is already there,
let alone there on the surface of other people’s speech. Once we have a story about
how sounds come to be imbued with meaning we can use this account as the basis for
explaining how we come to know what other people are saying. The first-personal
account of understanding I propose is based on the following schematic claims:
(A) I take my words to mean such and such.
(B) In hearing you utter those words I take you to mean what I mean by these
words.
How do we account for (A), and why does the assumption in (B) provide me with
an entitlement to knowledge of what you mean? Is there a metaphysics and epistemology
of meaning that makes the first-person case basic in the envisaged way? I think there is
and the account I will sketch here has three parts:
(i) An account of (A) showing what it is to take words to mean something;
(ii) An account of our epistemic entitlement to claim understanding of others on
the basis of our own understanding in (B);
(iii) An explanation of (B) drawing on an analogy with looking and seeing;
13. Taking Words to Mean Something

For a language learner to take a word to mean something requires more than just the
ability to utter the word and the word having the meaning my community give it.
I could do this with an utterly foreign word and not understand it. At the same time, a
word’s meaning something to me is not constituted by my simply taking it to mean
something. But if it is not a personal impression of meaning that gives the word a
meaning for me, why does my knowing that it means such and such seem to require
little more of me than my taking it to mean such and such? The answer is that further
conditions have to be met for my taking the word to mean so and so will coincide with
its having that meaning. Under the right conditions what I take my words to mean is
what they do mean. The further conditions involve links to others with whom, in the
initial stage of language learning, I fix a word’s meaning. These are the basic cases of
early word learning, from which the child will build to more complex cases. Children
73
Davidson 2001, p. 14.
acquire the particular meanings of certain words under conditions of joint attention to
commonly perceived objects. In joint attention there is a coordination of experience
between parent and child, requiring more than both people happening to attend to
the same thing. The child often checks the eye gaze of the parent, tracing the line from
eye to object, and back. Under these conditions, introducing a word to accompany
the perceptual experience of a jointly attended to object (an experience essentially
involving another) can serve as a sound label for the object perceived.74 Saying ‘ball’ in
the presence of a ball, even if the child is looking at it, will not secure the link between
word and object until child and parent are jointly attending to the object. The sound
label adds a phenomenological colouring to the experience of the object, and re-
introduction of the object with that sound will commemorate the original experience,
until eventually, the sound label without the object can re-create the experience—the
experience common to parent and child—of attending to that particular thing. When
we speak, what we are doing, in effect, is trying to shape and direct the experience of
the other so they think the same thing, or in the way we do. We use words to
coordinate our experience with others. And in situations of joint attention, we come
to endow our words with meaning.
The coordination of the minds of speaker and hearer, teacher and learner, is what
connects the minds of each to the other and ensures that the meaning the hearer invests
sounds with will be tightly connected to, and in that sense utterly dependent on, what is
in the mind of the speaker/teacher (and beyond this to others too). Coming to mean
something is not just a matter of individuals acting in isolation and happening, correctly,
to assume others have the same internal significance-investing states they have.
Once a repertoire of simple lexical items has been built up, perhaps involving verbs
to denote types of events, the child will go from the one-word stage to the two-word
stage, combining words in no particular order. Thereafter, at around twenty to twenty-
four months, it will reach the syntax spurt where it will suddenly arrange words into
sentences, At this point the syntactic and semantic dependencies between words opens
up new fields of meaning, previously unavailable.75
But how does it work that taking my words to mean something is enough for me to
suppose you mean the same thing, and why should this be enough (at times) to supply
knowledge of what other speakers do in fact mean?
14. Understanding Others

Once children take words to mean something there is a natural tendency for them to
assume others mean the same thing. But why would simply supposing this give them
74
We know that under these conditions but not others the child rapidly acquires the meaning of new
words for things. See Baldwin 1993, 1995; Baldwin et al. 1996.
75
There is clearly a very complex story to be given about pre- and post-syntax competence but I cannot
go into details here.
knowledge of another’s meaning? The issue here is about how transitions are made
from I to II.
I. I take you to mean M by W (where M is what I take word W to mean)
II. You mean M by W
How can I make a warranted transition from I to II? There would be warrant to
make the transition if I could rely on a general premise:
III. Anyone who uses W means what I mean by it
However, I can only justify III, it seems, when I have sufficiently many instances
of justified transitions from I to II to provide inductive warrant for III. But how
could I secure warrant for the initial transitions to warrant the inductive inference in
the first place? By now one should recognize the form of the problem as the one
which Crispin Wright identifies in Wittgenstein’s On Certainty and deploys to halt
scepticism in his “Warrant for Nothing (and Foundations for Free)”. In all such
cases, we need to rely on general premises like III to entitle us to move from
instances of I to corresponding instances of II. But instead of using transitions from
I to II to justify III, it is the prior acceptance of III that warrants those transitions.
Our entitlement to III is, in Wright’s terms, strategic and not supported by empirical
warrant. The idea being that we must presuppose III in order to reason from I to II,
for without this presupposition (at least initially), our communicative practices
would not get started. Such propositions are hinges (as Wittgenstein puts it), or
cornerstones of a practice (as Wright puts it). Were we not to make a prior assumption
of III, there could be no practice of understanding others. Thus, III has to be
assumed from the outset, and when it is, we are entitled to claim to know what
others mean. With III assumed, we are entitled to move from instances of I to
corresponding instances of II.
But why assume III to begin with? Well, the most natural thing to think is that for
the child it is the default case. Children arrive at it by being oblivious to the distinction
between what they take a word to mean and what it means in the mouths of others. At
first, all their encounters with a sound come with the significance it has for them: the
significance with which they have endowed it. To act as though this is what everyone
means by the word will, by and large, serve them well. So, as long as the learner uses
the word within his own linguistic circle—among the people with whom he fixed a
meaning for that word—he need never entertain the idea that there is any other way to
take the word. And any production of theirs will be heard as having the same meaning,
which it will have if those are the people, or in the same way related to the people, who
helped him fix a meaning for the word initially. To suppose that anyone who uses
W means what I do by it amounts—in the default case—to thinking that’s just what
W means. Let’s return to why the default case should come so naturally to language
learners.
15. An Analogy With Looking and Seeing

Psychological grounds can be given for III being the default assumption. An analogy
with looking and seeing in our assumptions about other people’s perceptions is helpful
here. Where there is a looking there is a seeing. The person looking has an experience of
seeing that others cannot experience. They can only observe him looking. However,
in my own case, I don’t observe my looking, I just have the experience of seeing. But
when I observe you looking, and I want to know what you have seen, I look in the
same direction you are looking and read my experience for yours. I use what I am
seeing to tell me what you have seen. Now I know that you’ve seen that odd looking
person over there. The assumption I make—that I see what you see (when we look in
the same direction)—is so automatic that it is not even consciously entertained. The
habit of moving from observing you looking at something, to my looking and thus
seeing it—to my taking myself to know what you are seeing—is so habitual and
unconsidered as not even to amount to the making of an inference. It is a natural
transition, available to many animals. We just read off our experience for theirs without
further thought. This is the default case. However, in special circumstances, where
I know something of your limitations or mine, the default can be overturned.
Increasingly, through development, children learn to be less egocentric, to realize
that they do not always share the same perspectives as others around them, and cannot
safely assume that everyone sees what they see. Though at first, when in close
proximity with others their egocentricity serves them well.
In the linguistic case there are analogous connections between use and understand-
ing. I don’t observe my use of language; I just understand what I’m saying. Others
cannot have my experience in speaking but they can observe my use. I can observe
other people’s uses of words, and because use goes together with understanding in my
own case—like looking and seeing—a use of those words comes backed with an
understanding: the understanding I have of those words. I hear them with the mean-
ings they would have were I using those words. I don’t separate my use of a word from
my understanding of it, so any use of the word I encounter will be heard, at first, with
that meaning. I will read my understanding as yours since I will, at first, take use and
understanding of the word to go together in the same way for everyone who uses the
word. Use and understanding of a word are wedded together for me, as two sides of the
same coin, and so it is natural, at first, to go from one to the other based on my own
case. When using a word I understand it. When you use a word there must be an
understanding you have and I read my understanding for yours. Again, this is so
automatic and regular that we hardly notice the transition. Uses of words are heard
with the meanings they have for me. It is as if (and I stress the ‘as if ’) this is just what the
word itself means. On this picture, the default case will serve the language user well.
Provided the child talks to people in its immediate linguistic circle who take the word
to have the significance the child takes it to have, the default assumption will cause no
failures of communication with respect to that word. Later, and with greater linguistic
and communicative facility, the child may find out, to its surprise, that not everyone
uses the word in the same way, or that not everyone can understand the word.
Monolingual children’s first encounters with foreign speakers are extremely bemusing
at first.
By using my understanding to add a dimension out other speakers’ uses of words
I can be in a position to know what they mean just by my taking words to mean what
they do for me—so long as the conditions are in place for me to take a word to mean
something in the first place. The epistemology of understanding is secured from a first-
person point of view, and arising as it does from interactions with others and parts of
the world, it is not a solipsistic or semantic individualist point of view. There is no need
for a mysterious and unwieldy metaphysics where the meanings of people’s words are
literally present on the surface of their speech behavior. Publicity of meaning is
respected in spirit if not in letter. It is true that we can only mean what other people
can know we mean: there is no way for an individual to fix meaning in isolation. This is
as it should be. And since I can know what others in my circle mean, and they can
know, in the same way, what I mean, condition (P) is respected. As learners move into
larger circles they will learn to revise the natural assumption of similar understanding.
At that point we go in for interpretation of others—not everyone has the same or
closely overlapping idiolect. But by then enough lexical meanings will have been
secured, which, together with grasp of syntax, will allow speakers to learn the meaning
of more complex linguistic expressions.
The picture I have sketched respects publicity without externalizing meaning. It
respects the inner aspect of comprehension and the phenomenology of understanding
in which certain people’s speech can be immediately intelligible. In effect, this is due to
the immediacy of one’s own understanding. It avoids the metaphysics of McDowell,
by making use of Wright’s epistemology of entitlement to explain how speakers can,
initially, rely on their own understanding to make sense of others’ talk. I suggest a
picture along these lines may offer the best hope of solving the reconciliation problem.
The first-personal character of this account addresses many of the original motivations
Wright provided for recognizing a speaker’s first-person authority. But unlike Wright’s
deflationary treatment of the latter, the account offers a substantial view of the inner life
of a speaker. And Wright’s own epistemology helps us to see how what is thereby
immediately available to the individual speaker can serve, in the default case, in his
understanding of others. What is jettisoned is any attempt to accommodate a speaker’s
knowledge of the linguistic rules of a public language; but the account is none the
worse for that.
Bibliography
Bach, K. 2001 “You Don’t Say”, Synthese 128, pp. 15–44.
Baldwin, D. A. 1993 “Infants’ Ability to Consult the Speaker for Clues to Word Reference”,
Journal of Child Language 20/2, pp. 395–418.
Baldwin, D. A. 1995 “Understanding the Link between Joint Attention and Language”, in
C. Moore and P. J. Dunham (eds.) Joint Attention: Its Origins and Role in Development, Hillsdale,
Lawrence Erlbaum Associates, pp. 131–58.
Baldwin, D. A., Markman, E. M., Bill, B., Desjardins, R. N., and Irwin, J. M. 1996 “Infants’
Reliance on a Social Criterion for Establishing Word–Object Relations”, Child Development
67/6, pp. 3135–53.
Burge, T. 1989 “Languages and Ideolects”, in A. George (ed.) Reflections on Chomsky, Oxford,
Blackwell, pp. 000–000.
Burge, T. 1999 “Comprehension and Interpretation”, in L. E. Hahn (ed.) The Philosophy of
Donald Davidson, Chicago, Open Court, pp. 229–50.
Carston, R. 2002 “Linguistic Meaning, Communicated Meaning and Cognitive Pragmatics”,
Mind and Language 17, pp. 125–48.
Chomsky, N. 1965 Aspects of the Theory of Syntax, Cambridge, Mass., MIT Press.
Chomksy, N. 1975 Language and Mind, New York, Harcourt Brace Jovanovich.
Davidson, D. 1984 Essays on Meaning and Interpretation, Oxford, Oxford University Press.
Davidson, D. 2001 “Externalisms”, in P. Koťátko, P. Pagin, and G. Segal (eds.) Interpreting
Davidson, Stanford, CSLI, pp. 1–16.
Dummett, M. 1976 “What Does the Appeal to Use Do for the Theory of Meaning”, in
A. Margalit (ed.) Meaning and Use, Dortrecht, Spinger, pp. 123–35.
Dummett, M. 1978 “What Do I Know When I Know a Language?” reprinted in M. Dummett,
The Seas of Language, Oxford, Oxford University Press 1993, pp. 94–105.
Dummett, M. 1983 “Language and Truth”, reprinted in M. Dummett, The Seas of Language,
Oxford, Oxford University Press 1993, pp. 147–65.
Fodor, J. A. 1983 The Modularity of Mind, Cambridge, Mass., MIT Press.
Follesdal, D. 1990 “Indeterminacy and Mental States”, in R. and R. Gibson (eds.) Perspectives
on Quine, Oxford: Blackwell, pp. 98–109.
Lepore, E. and Smith, B. C. (eds.) 2006 The Oxford Handbook of Philosophy of Language, Oxford,
McCulloch, G. 1998 “Intentionality and Interpretation”, in A. O’Hear (ed.) Current Issues in
Philosophy of Mind, Cambridge, Cambridge University Press, pp. 253–71.
McDowell, J. 1981 “Anti-Realism and the Epistemology of Understanding”, reprinted in
McDowell 1998b, pp. 314–43.
McDowell, J. 1984 “Wittgenstein on Following a Rule”, Synthese 58, pp. 325–63.
McDowell, J. 1987 “In Defence of Modesty”, reprinted in McDowell 1998b, pp. 87–107.
McDowell, J. 1992 “Putnam on Mind and Meaning”, reprinted in McDowell 1998b, pp. 275–91.
McDowell, J. 1998a Mind, Value, and Reality, Cambridge, Mass., Harvard University Press.
McDowell, J. 1998b Meaning, Knowledge, and Reality, Cambridge, Mass., Harvard University
Press.
Pagin, P. 2000 “Publicness and Indeterminacy”, in A. Orenstein and P. Ko’átko (eds.) Knowl-
edge, Logic and Language: Questions for Quine, Dordrecht: Kluwer, pp. 163–80.
Quine, W. V. O. 1960 Word and Object, Cambridge, Mass., MIT Press.
Quine, W. V. O. 1969 Ontological Relativity and Other Essays, New York, Columbia University
Press.
Quine, W. V. O. 1990 The Pursuit of Truth, Cambridge, Mass., Harvard University Press.
Recanati, F. 2001 “What is Said”, Synthese 128, pp. 75–91.
Recanati, F. 2004 Literal Meaning, Cambridge, Cambridge University Press.

Smith, B. C. 1992 “Understanding Language”, Proceedings of the Aristotelian Society CII, pp. 109–41.
Smith, B. C. 1998 “On Knowing One’s Own Language”, in C. Wright, B. C. Smith, and
C. Macdonald (eds.) Knowing Our Own Minds, Oxford, Oxford University Press, pp. 391–428.
Smith, B. C. 2001 “Idiolects and Understanding”, Mind and Language 16/3, pp. 284–90.
Smith, B. C. 2006a “What We Know When We Know A Language”, in E. Lepore and
B. C. Smith (eds.) The Oxford Handbook of Philosophy of Language, Oxford, Oxford University
Press, pp. 941–82.
Smith, B. C. 2006b “Why We Still Need Knowledge of Language”, Croatian Journal of Philosophy
6, pp. 431–57.
Smith, B. C. 2006c “Publicity, Externalism and Inner States”, in T. Marvan (ed.) What
Determines Content, Cambridge, Cambridge Scholars Press.
Smith, B. C. 2009 “Speech Sounds and the Direct Meeting of Minds”, in M. Nudds and
C. O’Callaghan (eds.) Sounds and Perception: New Philosophical Essays, Oxford, Oxford University
Press, pp. 183–210.
Sperber, D. and Wilson, D. 1986 Relevance: Communication & Cognition, Oxford, Blackwell.
Stanley, J. 2000 “Context and Logical Form”, Linguistics and Philosophy 23, pp. 391–434.
Strawson, G. 1996. Mental Reality, Cambridge, MA, MIT Press.
Travis, C. 1997 “Pragmatics”, in B. Hale and C. Wright (eds.) A Companion to the Philosophy of
Language, Oxford, Blackwell, pp. 87–107.
Wright, C. 1986 “Does Philosophical Investigations 258-260 Suggest a Cogent Argument Against
Private Language?”, reprinted in Wright 2001, pp. 223–90.
Wright, C. 1989 “Wittgenstein’s Rule-Following Considerations and the Central Project of
Theoretical Linguistics”, reprinted in Wright 2001, pp. 170–214.
Wright, C. 2004 “On Epistemic Entitlement: Warrant for Nothing (and Foundations for
Free)?”, Proceedings of the Aristotelian Society, Supp. Vol. LXXVIII, pp. 167–212.
6
Expression, Truth, and Reality:
Some Variations on Themes
from Wright
Dorit Bar-On
1. Introduction: Expressivism as an
‘Anti-Realist Paradigm’
Expressivism, broadly construed, is the view that the function of utterances in a given
area of discourse is to give expression to our sentiments or other (non-cognitive)
mental states or attitudes, rather than report or describe some range of facts. This
view naturally seems an attractive option wherever it is suspected that there may not be
a domain of facts for the given discourse to be describing. Familiarly, to avoid
commitment to ethical facts, the ethical expressivist suggests that ethical utterances (e.g.
“Gratuitous torture is wrong,” “What John did was morally good”) do not serve to
ascribe ethical properties to objects, actions, persons, or states of affairs. Instead, they
simply function to give voice to certain of our sentiments (or ‘pro/con’ attitudes).
Along similar lines, philosophers have entertained versions of expressivism about the
aesthetic, the modal, the mental, what is funny, even about theoretical science and
knowledge ascriptions.1
In several of his writings, Crispin Wright lists expressivism among the chief ‘anti-
realist paradigms’ and not a successful one at that. Insofar as expressivism purports to
allow us to avoid unwanted ontological commitments, its success depends on deny-
ing—implausibly—that the relevant discourses “really deal in truth-evaluable con-
tents.” Yet Wright thinks that these discourses do possess truth-evaluability, since they
exhibit all the relevant ‘assertoric trappings,’ which, on the ‘minimalist’ conception
Wright develops in Truth and Objectivity, suffice for truth-aptitude and truth.2 If it were
indeed true that expressivism, in every case, required denying truth-evaluability, then
1
See Wright 1988, pp. 46–47, endnote 12.
2
See Wright 1992, Chapter 1, and Wright 1996 for a summary.
E X P R E S S I O N , T RU T H , A N D R E A L I T Y 163
Wright would be correct to declare the expressivist move ultimately a ‘faux pas’
(Wright 1988: 34). For, in that case, expressivism would require compromising the
palpable semantic continuity between putatively expressive and ordinary descriptive
discourses. Ethical sentences, for example, behave just like ordinary descriptive sen-
tences syntactically as well as logico-semantically. They can be involved in logically
valid inferences, they admit negation, embed in conditionals and in propositional
attitude contexts, and so on.3 To maintain expressivism as a thesis that directly concerns
the semantics of ethical, aesthetic, psychological, modal, etc. sentences—a thesis that
treats them as effectively synonymous with verbal expressions such as “Good for you!,”
“Boo!,” “Yuk!,” “Ouch!” and so on—is to do serious violence to a wide array of
linguistic appearances and intuitions.
I think that the alliance standardly assumed to exist between expressivism, on the
one hand, and the denial of truth-evaluability as well as anti-realism, on the other, is
unfortunate, though historically understandable. With the proper distinctions in
place, I believe we can decouple the core expressivist idea from the denial of truth-
evaluability and break its apparently necessary link to anti-realism. The result will,
I hope, present us with a viable option in various areas where philosophers have been
drawn to expressivism—an option which I also hope Wright himself will welcome.
I begin by discussing one kind of expressivism, concerning avowals and self-
knowledge—”avowal expressivism,” for short. I present a simple version of avowal
expressivism and some reasons canvassed by Wright for rejecting it (Section 2). Wright,
however, maintains that avowal expressivism fares better than ethical expressivism. In
Section 3, I take up the comparison between these two expressivist views and argue
that, understood as a radical anti-realist view, simple avowal expressivism is in fact
incoherent. In Section 4, I consider an alternative anti-realist view of first-person
authority that Wright himself flirts with in several places under the title “the default
view.”4 I argue that it shares a defect with simple avowal expressivism. I then turn (in
Section 5) to my own neo-expressivist view of avowals, which, I submit, is not only
independently plausible but also superior to the default view. I will not here argue for
the independent plausibility of this neo-expressivist view.5 Instead, in Section 6, I will
explain how the view avoids the incoherence of simple avowal expressivism, and
indicate how it can be generalized to other cases so as to yield a new expressivist
paradigm, one that allows for truth-evaluablity and avoids anti-realist commitments.
3
These are referred to by Wright under “Geach’s point” (Wright 1998, sec. 3, especially p. 31, see also
Geach 1960).
4
Wright 1998, p. 41.
5
For a full treatment, see Bar-On 2004. Though I have so far spoken of avowals as utterances, my account
is intended to apply to avowals made in thought, and not just in speech. (The distinction between avowals
and other kinds of self-ascriptions is relevant to thought as well as speech. A person’s spontaneous thought:
“I’m sick of this!” has a different epistemic status from thoughts she might have about how she felt yesterday,
about the position of her limbs, about others’ mental states, etc.)
164 DORIT BAR-ON
In all this, I take myself to be offering some variations on themes near and dear to
Wright’s heart. I do not expect to be doing justice to the richness of Wright’s own
treatment of these themes, nor will I be attempting a careful exegesis. Nonetheless, I do
hope to be articulating a view that Wright himself may find congenial—or else learn
why he does not.
2. Avowal Expressivism
Consider everyday utterances such as “I have a terrible headache,” “I feel sick,” “I’d
like some tea,” “I’m scared of that dog,” “I’m wondering whether it’s going to rain,”
“I hope that I’m not late.” On their face, such utterances—“avowals,” as they are often
called—resemble other everyday utterances in which we report various occurrences,
such as “There is a red cardinal at the bird-feeder,” “We’ve just run out of dog food,”
or “I have a mosquito bite on my leg.” Like these reports, avowals appear to inform us
of certain contingent states of affairs; they tell us of states of mind that the speaker
happens to be in at a given moment. Yet when compared to other sorts of utterances,
avowals appear to enjoy a special status and to be especially secure. Thus, avowals
contrast with third-person present-tense ascriptions of mental states (e.g. “She is very
tired” said of me by someone else, or “John is very tired” said by me about someone
else), or past-tense mental self-ascriptions (“I was very tired then,” said by me at a later
time). Avowals also contrast with non-observational bodily self-ascriptions (“My legs
are crossed,” “I’m sitting down”) and present-tense self-attributions of psychological
traits or standing dispositions (“I’m a patient person,” “I like going to restaurants”). It is
also interesting to consider that an ordinary avowal of fatigue would seem much more
secure than some present-tense self-ascriptions of occurrent mental states, such as “DB is
very tired,” if it were said by me about myself when I fail to recognize that I am DB, or
“I am very tired,” if it were said by me on the basis of, say, looking in the mirror, or
inference from some test results.
Unlike these other utterances, avowals appear to be, as Wright puts it, ‘groundless’:
they are seemingly made on no epistemic basis, and it would normally be considered
out of place to ask of a person issuing an avowal for reasons or justification.6 Yet
avowals enjoy a special security: except under special circumstances, avowals are not
subject to correction and are taken at face value. Whereas it seems that other pro-
nouncements are only as secure as their epistemic basis, avowals appear remarkably
secure even though they are apparently made on no epistemic basis.7
6
Wright 1998, p. 14.
7
Wright distinguishes between phenomenal and attitudinal avowals in point of their security (Wright
1998, pp. 14–17). The former, he thinks, are ‘strongly authoritative,’ whereas the latter are only ‘weakly
authoritative.’ By contrast, Bilgrami 1998 argues that, properly understood, attitudinal avowals are incorrigi-
ble (Bilgrami sets aside phenomenal avowals altogether in his account of first-person authority). For reasons
I detail in Bar-On 2004, I don’t think much weight should be placed on the phenomenal/attitudinal divide
when trying to understand the distinctive security of avowals.
As the Cartesian tradition would have it, avowals enjoy absolute security: they are
absolutely indubitable, incorrigible, and infallible. Moreover, avowable mental states
are thought to be ‘self-intimating’: one who is in a mental state is guaranteed to know
it. There may be good reasons to depart from the Cartesian tradition on these scores.
But even so, the epistemic asymmetry between avowals and all other ascriptions (includ-
ing, as can be seen from the above list, various kinds of self-ascriptions) calls for an
explanation.8
If we are no longer wedded to absolute incorrigibility or infallibility as the mark of
avowals, it may be tempting to play down the contrast between avowals and other self-
ascriptions. Specifically, we can note that there is a class of bodily predicates, such as
“crossed legs,” or “is sitting down,” whose application exhibits a certain epistemic first-
person/third-person asymmetry. Such predicates are applied on the basis of external
observation in the third-person case, but on the basis of proprioception or kinesthesis in
the first-person case. Likewise, we can think of mental predicates “in pain,” or “is
angry at y,” and so on, as predicates that are applied on different epistemic bases in the
third- and first- person cases. They apply on the basis of observation of behavior, or
inference, or conjecture, in the third-person case, but, in the first-person case, though
they appear to apply on no basis, they in fact apply on the basis of introspection.
Suppose, contra the Cartesian understanding, we think of introspection in the
materialist’s preferred terms, as a faculty whose job is to deliver reliable though not
infallible reports on certain goings-on inside our bodies (more specifically, in our
brains), which reports are articulated by avowals. Might we not then hope to have a
solution to our puzzle regarding epistemic asymmetry without paying the heavy
metaphysical cost incurred by the Cartesian introspectionist?9
The materialist introspectionist view (“MI” for short) identifies a distinct epistemic
‘route’ that we can associate with avowals. But doing so cannot by itself allow us to
understand the perceived contrast in degree of security between avowals and certain
bodily self-reports, such as those arrived at through proprioception. After all, I have a
different way from my observers of learning where my limbs are, or whether I am
sitting down, but my word on these matters is not clearly better than theirs. On a given
occasion, my body could be in a state that my proprioceptive mechanism cannot
distinguish from the state of my legs being crossed. I could also prove to be systematically
unreliable in reporting my limb positions or bodily movements without looking. The
ability to tell my bodily conditions through proprioception or kinesthesis thus seems
alieanable. Furthermore, we can even conceive of someone whose brain was so hooked
up to my limbs as to receive direct information about their position. Such a person
8
Wright does not endorse absolute Cartesian security, though he does include self-intimation (which he
refers to as ‘transparency’) among the key features of ‘avowals’: “in the normal run of cases, the subject’s
ignorance of the truth or falsity of an avowal . . . is not, it seems, an option”—Wright 1998, p. 15.
9
The discussion that follows makes several points of contact with Wright’s rejection of the ‘observational’
model. See, for instance, Wright 1987, pp. 399–402, Wright 1989, sec. IV, and Wright 1998, pp. 22–28.
166 DORIT BAR-ON
could inherit my proprioceptive ability, and would be able to tell where my limbs are
the same way I do. Proprioception (as well as kinesthesis) thus seems also entirely
transferable.10
If the distinctive security of avowals were entirely due to the epistemic security of
introspective mechanisms, we might expect so-called ‘first-person authority’ to be
equally alienable and transferable. Yet I agree with Wright that, by commonsense
lights, it isn’t. It is one thing to show that various mental self-ascriptions we make are
false; it is another to show that we can make sense of the idea of a subject who, though
she has mental states, is ‘chronically unreliable’ in her avowals.11 Even more problem-
atic is the notion that I might be able to transfer my first-person authority to someone
else—someone who was able to ascribe to me reliably and correctly present mental
states without relying on any observation or evidence.12 For one thing, my avowals
would surely constitute a crucial part of the data against which to measure my putative
mindreader’s claim to be able to provide reliable non-evidential reports of my occur-
rent mental states. But more importantly, so long as there was no question about my
linguistic competence or my sincerity, if disagreement broke out between us over what
is now going on in my mind, her consistent past success in reading my mind would not
be sufficient ground for taking her word over mine. It seems, rather, that we would
take the disagreement as signaling the waning of her mind-reading powers.13
There is another sort of error that should be possible on the MI model: brute local error—
an error that is not due to any defect of any of my psychological mechanisms but is simply
due to the fact that my internal introspector has been ‘fooled’ into false detection. On the
commonsense view, though, the possibility of brute local error seems no less problematic
than the possibility of global systematic failure, and much more problematic than MI
would have us expect. It’s not that commonsense does not allow for false avowals. As
Wright notes, false attitudinal avowals—e.g. of beliefs, intentions, hopes, etc.—can be
issued in self-deception or as a result of wishful thinking (Wright 1998, p. 17). I also think
that under certain circumstances, even a sincere phenomenal avowal may be thought
false. For example, as you sit on the dentist’s chair and say “My tooth hurts!” before the
drill reaches your mouth, your dentist may sensibly question whether you really are in
pain, even without questioning your sincerity.14 But although these are cases in which
10
See also the kaleidoscope analogy in Wright 1998, p. 22.
11
Wright says that, even in the case of attitudinal avowals, which he thinks are only ‘weakly author-
iatitve,’ “[w]holesale suspicion about my attitudinal avowals—where it is not a doubt about sincerity or
understanding—jars with conceiving of me as an intentional subject at all”—Wright 1998, p. 17.
12
Wright 1998, p. 24.
13
Similar remarks would apply to the possibility entertained by Rorty of a brain-scanner set up to replace
one’s own self-scanning mechanism and to serve as a mechanical mind-reader. See Rorty 1970.
14
I discuss other examples of sincere but false phenomenal avowals in Bar-On 2004, pp. 329–35 and 394–6.
I thus disagree with Wright that phenomenal avowals differ from attitudinal avowals in being (treated as)
‘strongly authoritative,’ meaning that someone’s making, or being disposed to make, an avowal is “a
guarantee of the truth of what they say” so that “[a] doubt about such a claim has to be a doubt about the
sincerity or the understanding of the one making it”—Wright 1998, p. 14.
we are prepared to question the truth of an avowal, they do not provide examples of brute
error; for these are precisely cases in which the avowal’s falsity is assumed to be due to
some psychological irregularity, failure, or defect on the part of the avower. We don’t
suppose in any of these cases that the subject might have simply been fooled into issuing a
false avowal by an ‘uncooperative’ mental world.15
Indeed, reflection on how best to understand cases of false avowals may help point us
in the direction of an alternative solution of our puzzle. Consider again the above case of
falsely (though sincerely) avowing a toothache at the dentist. At first blush, this kind of
case may seem grist to the materialist introspectionist mill: your tooth doesn’t really
hurt, but your internal introspector mistook what is in fact (say) fear of the approaching
drill for pain, and it is this finding that your false avowal represents. But how plausible is
it to regard your exclamation “My tooth hurts!” as a report of your (false) belief that
your tooth hurts? After all, under the circumstances, you might have equally said:
“Ouch!,” or emitted a yelp, or winced. Should we think of the wince as equally the
upshot of a false belief that you formed about your internal state? Isn’t it much more
plausible to regard the exclamation: “My tooth hurts!” as on a par with the wince—as
something forced out of you, though in this unusual case not by an actual toothache, but
rather by the priming effect of the fear (as well as your previous history at the dentist)?
In this somewhat roundabout way, we have arrived at a well-known counter to
introspectionism often attributed to the later Wittgenstein, namely, avowal expressi-
vism.16 This is the view that avowals are importantly similar to natural expressions in
being bits of behavior that typically serve directly to express our present mental states
rather than to report their presence. The key idea of simple avowal expressivism is that,
although avowals closely resemble in their surface form ordinary descriptive reports,
logically and semantically speaking they are more like natural expressions of sensations,
such as moans and groans, grimaces, giggles, etc.
Now, natural expressions of sensations are spontaneous reactions that are in some
essential way characteristic of the subject’s relevant conditions. A subject in pain naturally
groans, or cries out, or grimaces, or clutches the painful part. A subject being tickled will
typically squirm and giggle. Natural expressions are not limited to sensations. There are
natural expressions characteristic of emotions and feelings such as fear, anger, excitement,
joy, etc. as well as of wants.17 What is crucial about natural expressions of all these kinds is
15
It is for this reason perhaps that full-blown ‘internal world skepticism’ (the analogue of Cartesian
external world skepticism) seems incoherent. What is required to fuel such skepticism is the possibility of
holding fixed all of our present ‘judgments of appearances’ concerning our internal world while question-
ing their veridicality. But plausible cases of false avowals are precisely not cases in which it is reasonable to
suppose that it appears to the subject that some internal state of hers is a toothache (a belief that p, a desire for
y, . . . ) when it isn’t. (I discuss the contrast between external and internal world skepticism in Bar-On,
forthcoming.)
16
See }244 in Wittgenstein 1953.
17
Compare Hacker: “A child who wants a toy reaches for it and tries to get it. The child’s anger is
manifest in striking out, contorted features, and screams or rage . . . if he is frightened, he blanches, cries, and
runs to Mummy”—Hacker 1993, p. 88.
168 DORIT BAR-ON
that they are in no way thought to represent a subject’s judgment or belief about herself.
A subject’s smile is not supposed to represent the subject’s judgment that she is pleased—
rather, it is supposed to give expression to the pleasure itself. Correlatively, we do not
expect the person who smiles, or cries in pain to give reasons for her smile or cry; we do
not query or challenge a gasp of fear, or a sigh of relief, and so on. Insofar as we regard a
form of behavior as a natural expression of a subject’s condition, it seems entirely
inappropriate to impose on it any epistemic assessment.
Simple avowal expressivism can be seen as offering to explain the epistemic asym-
metries that give rise to our puzzle by breaking away from the face-value understand-
ing of avowals’ grammar. As a first step, the account maintains that, despite surface
appearances, avowals of sensations have the same ‘logical grammar’ as natural expres-
sions of sensations. That is to say, ‘grammatically’ speaking, mental terms as they are used
in avowals do not have the semantic function of picking out mental states and ascribing
them to individuals. 18 Avowals serve only to express subjects’ mental states, just like
cries, or winces, or joyful hugs. As a second step we can then realize that, epistemically
speaking, it is misguided to regard the special status of avowals as a consequence of a
special access subjects have to their own states of mind, and there is no need to seek any
secure epistemic basis on which avowals are made. If avowals are not vulnerable to
doubt, and are not open to query, or correction, this is not because of the security of
their epistemic basis, but because they do not serve to tell or describe the avower’s
present condition in the first place. Like natural expressions, they merely serve to
express subject’s mental states, and so, they belong in the wrong ‘grammatical’ category
for any epistemic assessment. A subject who avows “I am in pain” is just like one who
spontaneously lets out a cry: such a one cannot be legitimately asked to give reasons,
questioned, or challenged.19 And this can serve to explain the asymmetries between
avowals and other ascriptions. But with this explanation of the epistemic asymmetries
in place, we can finally agree with materialists that ontologically speaking, there is no need
to invoke Cartesian entities to play the role of especially accessible mental items that are
referred to by sensation (or any other mentalistic) terms.
The ‘no-ascription’ aspect of simple avowal expressivism is invoked to explain the
epistemic asymmetries between avowals and other ascriptions while avoiding the
Cartesian appeal to private objects. However, this very aspect is its downfall. If avowals
do not involve genuine—let alone true—ascription of mental states to the avowing
subjects, then they cannot have truth-conditional equivalents that do involve such
ascription and with which they can be legitimately interchanged in certain contexts;
and they cannot serve as legitimate premises in relevant logical inferences. In short,
simple avowal expressivism requires explaining the epistemic asymmetries by
compromising notable semantic continuities between avowals and other ascriptions.
18
In Bar-On 2004, Chapter 7, I suggest that this move is analogous to Anscombe’s denial that ‘I’ refers.
19
Compare Wright 1998, pp. 34–5. Notably, however, we do sometimes ‘take issue’ with non-verbal
expressive behavior. “You can’t really be tired,” we may say to someone who is yawning and stretching.
Avowals, understood as simply alternative ways of moaning and groaning would be

invulnerable to error and incorrigible—indeed, absolutely so—but only in the degen-
erate sense of belonging to the wrong grammatical category. And they would be
protected from epistemic assessment, but only by being altogether excluded by
‘grammar’ from the realm of linguistic meaning, judgment, and knowledge.
3. Simple Avowal Expressivism vs. Ethical

Expressivism20
In its suggestion that avowals do not involve genuine ascription of (or reference to)
mental states, but only serve to express them, simple avowal expressivism is reminiscent
of traditional ethical expressivism. According to the latter view, ethical claims (e.g. “This
is good/bad,” “Racial discrimination is unjust”) are mere expressions of certain of our
emotions, preferences, or attitudes. They are not genuine, truth-evaluable assertions
either about objective ethical states of affairs (as ethical objectivists would have it) or
about the expressed attitudes (as subjectivists would have it). As is well known,
traditional ethical expressivists, like Ayer and Stevenson,21 were motivated in good
part by wanting to avoid all commitment to an ethical realm of facts populated by
ontologically suspect ethical properties. Rather than taking ethical claims to be asser-
tions about our subjective attitudes, they suggested that ethical terms do not function
to designate any properties but are rather terms of dis/approbation, dis/approval, or
condemnation/commendation (much like “Hooray,” “Yuk,” “Super”).
Of the various complaints made against ethical expressivism over the years, the one
most pertinent to the comparison with simple avowal expressivism is this: ethical
expressivism seems unable to accommodate the semantic continuity of ethical dis-
course with ordinary descriptive discourses. In particular, it cannot accommodate the
fact that ethical terms, just like ordinary descriptive terms, can occur in all sorts of
grammatical contexts. For instance, suppose we were to accept that to say “Helping the
poor is good” is only ever to express a pro-attitude toward helping the poor. Still, it is
difficult to accept that the sentence is used that way when embedded in the context of a
conditional: “If helping the poor is good, then we should tax the rich and give the
money to poor people.” As Geach has famously argued,22 in such a context, it seems as
though the sentence makes a semantic contribution that is not properly captured by the
expressivist analysis.
However, as Wright points out, the ‘Geach point’ can perhaps be handled by the
ethical expressivist.23 After all, take an utterance of a sentence such as “It usually rains
20
Some of the material for this section is taken from a similarly titled section of Chapter 7 of my Bar-On
2004.
21
See Ayer 1946, Chapter 6 and Stevenson 1944.
22
Geach 1960.
23
Wright 1995a, sec. 1, Wright 1998, sec. 3.
170 DORIT BAR-ON
here at this time of the year” which normally has uncontroversially assertoric force. If
we utter the same sentence as part of a conditional or embed it in any other ‘force-
stripping’ context (negation, propositional attitude construction, etc.), it will no longer
be used to make an assertion (though the sentences in which it is embedded may). This
just means that from the possibility of embedding an indicative sentence in a force-
stripping grammatical context we cannot infer anything about the illocutionary force
that is standardly associated with it when it is uttered on its own. Given the familiar
separability of force and surface grammar, this should come as no surprise. (“I’d like to
know what time it is,” for example, is an indicative sentence that is standardly used to
ask a question, rather than to make an assertion about oneself.) But then the ethical
expressivist can maintain that, at most, the Geach point establishes that sentences
containing ethical terms must have whatever semantic features that allow sentences
to function (grammatically and logically) like ordinary indicative sentences. Thus the
fact that ethical sentences can be embedded in force-stripping grammatical contexts
does not tell against the expressivist claim that the standard function of ethical utter-
ances is to express attitudes or sentiments rather than to make assertions about the
world.
Applied to simple avowal expressivism, the Geach point would be that it is impossi-
ble to reconcile the claim that avowals merely serve to express the self-ascriber’s mental
states, and in no way constitute assertoric reports about them, with a host of grammati-
cal facts about avowals. Like ethical sentences, and like ordinary descriptive sentences
(e.g. “I am 50 500 tall), avowals embed in force-stripping constructions (“If I’d like some
ice-cream, then it would be nice for you to get me some,” “Rachel thinks that I’d
like some ice-cream”), they admit of tense-transformation (“Yesterday, I wanted some
ice-cream”), and so on. In these contexts, they do not plausibly serve to express the
mental states apparently named in them. But the above response to the Geach point
offered on behalf of the ethical expressivist can serve the simple avowal expressivist just
as well. The Geach point can show that “I’d like ice-cream” must possess whatever
features are required for a sentence to take indicative form. From this, it does not
follow that the sentence must always be used to make assertions or descriptive reports
about oneself, or that it cannot serve to express rather than report one’s desire outside
force-stripping contexts.
Wright agrees that the Geach point “is powerless to determine that the standard use
of a locution is to assert [truth-evaluable] content.”24 However, he thinks that insofar
as Geach’s observation exposes classes of sentences as capable of making positive
contributions to truth-conditions (in force-stripping contexts), it shows that these
sentences must possess “truth-evaluable contents.” For this reason, Wright seems to
think that Geach’s point ultimately causes more trouble to ethical expressivism than to
avowal expressivism. For the ethical expressivist is committed to the thesis that “there
24
Wright 1998, p. 36.
are no real moral states of affairs; so the occurrence of what are apparently truth-
evaluable contents couched in distinctively moral vocabulary has to be some kind of
illusion” (ibid.).25 In other words, in the ethical case, as in other familiar cases, the
expressivist thesis is put to the service of an anti-realist ontological agenda. But the case
of avowal expressivism, Wright seems to think, is different:
[since] it is no part of the present, allegedly Wittgensteinian expressivist proposal that there is no
such thing as a statement of ordinary psychological fact. No one is questioning that ‘He is in pain’
is an assertion. The expressivist thesis distinctively concerns avowals.26
Wright’s treatment here brings up two issues. First, Wright’s idea that “the Geach
point signals . . . the presence of truth-evaluable content” presupposes that all indicative
sentences that are subject to Geach’s point must possess truth-evaluable content, and
that the only way for a sentence to have such content is by having the kind of truth-
conditions possessed by ordinary descriptive sentences. But suppose one accepts that
indicative form must go hand in hand with truth-evaluability. Still, it seems that the
ethical expressivist could try to explain how ethical sentences can earn indicative
grammatical form, as well make systematic contributions to truth-conditions, even if
ethical terms do not denote ethical properties.27 This raises the general question
whether expressivism must be saddled with the denial of truth-evaluability, even if it
is coupled with anti-realism. Relevant here is the minimalist conception of truth and
truth-aptitude that Wright himself recommends, according to which, roughly, posses-
sion by a discourse of the key features that underwrite the Geach point suffices for
truth-evaluability.28 For, on a minimalist conception, it would seem that one can
be an anti-realist regarding a discourse without denying that its sentences are truth-
evaluable.29 But then it’s hard to see why an expressivist could not embrace the
minimalist understanding and preserve truth-evaluability. Secondly, it is notable
that, in the case of avowals, Wright appears to be allowing an expressivist view that
isn’t anti-realist, thereby departing from his general characterization of expressivism as
25
This claim may seem an odd one for Wright to make, given his minimalism about truth and truth-
evaluability; see discussion below.
26
Wright 1998, p. 36.
27
For the separation between grammatical form and propositional content, on the one hand, and illocu-
tionary force, on the other, see Austin 1961. For attempts by ethical expressivists to explain how ethical claims
can earn propositional form in the absence of ethical reality, see e.g. Simon Blackburn 1984, pp. 181–94, 1993,
essay 10, and Allan Gibbard 1990, Chapter 5, 2003, Chapters 3 and 4. However, both Blackburn and Gibbard’s
‘anti-realist’ proposal shy away from accepting that ethical terms can contribute to truth-conditions. Ridge 2006
defends an ‘ecumentical expressivist’ view that seems closer to the line I have in mind here. Below, I suggest that
Wright’s own minimalism about truth-evaluability can also be roped into this service.
28
Wright 1992, p. 29.
29
See Wright 1996, p. 864. An ethical anti-realist might worry that accepting (even minimal) truth-
evaluability for moral claims commits you to moral beliefs through a platitude according to which only beliefs
can have truth-evaluable content. For discussion of this issue see Blackburn 1998, pp. 158–60. For partial
response see Bar-On and Chrisman 2009.
172 DORIT BAR-ON
an ‘anti-realist paradigm.’30 Moreover, as I shall now explain, in the case of avowals,

one cannot be an expressivist and at the same time be an anti-realist about mental states.
Let me begin by pointing out a difficulty for simple avowal expressivism that does
not beset ethical expressivism. The difficulty has to do with the fact that avowal
expressivism attempts to capture a contrast that arises within a single area of dis-
course—i.e. mentalistic discourse. This contrast concerns the epistemic asymmetries
between avowals, on the one hand, and, most notably, mental ascriptions to others,
as well as past-tense mental and present-tense bodily self-ascriptions, on the other. By
contrast, ethical expressivism, at least in its traditional form, aims to capture an alleged
an ontological difference between ethics and other, straightforwardly ‘factual’ areas of
discourse. The ethical expressivist who is moved by the ‘queerness’ of ethical facts31
invokes the expressivist view in order to fill an apparent gap that opens up once we pull
the metaphysical rug from under our ethical pronouncements—once we deny that
there is a realm of facts for ethical claims to describe truly or falsely. We should
distinguish, then, between
(a) the positive expressivist claim that ethical proclamations regularly serve to express
pro- and con- attitudes, rather than to make assertions about objective states of
affairs,
and
(b) the negative ontological claim that there are no ethical properties for ethical terms
to refer to and no ethical facts for ethical sentences to describe,
as well as
(c) the negative semantic claim that ethical sentences are not truth-evaluable, since
they do not express true or false propositions.32
In the ethical case, if the expressivist can show how, despite the truth of (b), ethical
sentences can earn truth-evaluability (perhaps by endorsing a minimalist conception
thereof ), she will have succeeded in offering a systematic, alternative conception of
ethical discourse on which ethical claims need never be construed as ascribing ethical
properties to some individual, or act, or state of affairs, and on which such claims, when
made on their own, have the distinctive point of expressing certain of our sentiments or
attitudes.33 By contrast, the scope of simple avowal expressivism is limited to a certain
30
It’s possible to take Wright here to be allowing only that third-person mental ascriptions are minimally
true, and thus not requiring the independent reality of mental states. I think the argument that follows suggests
that he must allow something stronger. In any event, he is at least leaving room for a realist construal.
31
For locus classicus see Mackie 1977, Chapter 1.
32
The closely related (but still separate) ethical non-cognitivist claim is that ethical sentences
never express cognitive, truth-evaluable beliefs or judgments.
33
The expressivist claim can be brought in not directly to fill in a perceived gap that opens once the
negative ontological claim is embraced, but rather to capture some distinctive positive features of ethical
discourse. See below.
range of mentalistic utterances: it encompasses (at most) those utterances that involve
apparent ascriptions of mental states to oneself in the present tense.34 The expressivist
proposal here is that such utterances—avowals, as we have been referring to them—are
mere expressions of subjects’ present mental states. They do not represent assertoric
reports about the subject’s present mental states, and they involve no reference to those
states. However, this proposal is at risk of prising mentalistic self-ascriptions too far
apart from other ascriptions within a single (i.e. mentalistic) discourse.
Let me explain. Consider again my avowal: “I’d like some ice-cream.” You could
presumably use my avowal to infer: “She (DB) would like some ice-cream.” And
I could later conclude, remembering my avowal: “Earlier, I wanted some ice-cream.”
Semantically and logically speaking, these three sentences are transformations of each
other that can support valid inferences and chains of reference. However, on simple
avowal expressivism as fashioned after traditional ethical expressivism, we are to
combine the idea that avowals serve only to express subjects’ mental states with the
negative ontological claim that there are no properties (or states/processes) for the
mental terms contained in avowals to refer to the way ordinary descriptive terms do.
But it is very hard to see how to reconcile this negative claim with the logico-semantic
continuities between avowals and other mentalistic ascriptions. In the ethical case, the
negative claim implies that ethical terms do not refer to genuine properties wherever they
occur. Thus if the ethical expressivist can explain how ethical utterances can earn
propositional form and content, even though there are no ethical properties for ethical
terms to refer to, her work is done. The proponent of simple avowal expressivism, on
the other hand, needs to explain how an avowal such as “I have a toothache,” which
allegedly makes no genuine reference to a state of an individual, can still be truth-
conditionally equivalent to non-expressive, assertoric utterances which do involve such
reference.
I have in mind here utterances of three types:
(i) Third-person reports (e.g. “She/DB would like some ice-cream”)
(ii) Tense transformations (e.g. “I wanted some ice-cream earlier”)
(iii) Mentalistic present-tense self-reports (e.g. the admittedly unusual but never-
theless possible “I want some ice-cream” made on the basis of evidence or
inference).
(In his discussion of avowal expressivism, Wright lists the fact that avowals submit to
tense transformation along with the fact that they can be embedded in conditionals/
negation under the heading “the Geach point.” Wright also cites two additional facts:
the fact that avowals can stand in logical relations to other statements (“I am in pain” !
34
I say “at most,” because some present-tense mentalistic self-ascriptions, namely, ones made on the basis
of evidence, inference, testimony, therapy, or reflective self-interpretation do not exhibit the distinctive
features of avowals. They are plausibly to be regarded as descriptive (fallible, corrigible) self-reports. Compare
Wright 1998, pp. 15–16.
174 DORIT BAR-ON
“Someone is in pain”) and the fact that they can be embedded in knowledge
ascriptions, as in “He knows that I am in pain.”35 I think that the case of tense-
transformation ((ii) above), should be grouped together with cases of truth-
conditionally equivalent third-person ascriptions (i) and of present-tense mentalistic
self-reports (iii), both of which are omitted by Wright. The reason is that I take cases
(i)–(iii) to illustrate how, pace Wright, avowal expressivism as presented so far faces a
greater prima facie difficulty than ethical expressivism.)
It may seem that, in response, simple avowal expressivism could follow the ethical
expressivist’s lead and adopt a more thoroughgoing non-referential account of psy-
chological discourse, according to which psychological/mentalistic terms are never
taken to refer to mental states. On this view—call it “psychological expressivism”—
when making psychological ascriptions, all we ever do, whether in the first-person or
third-person, whether in the present or past tense, is express various subjective condi-
tions. So, for instance, saying of another “She is in pain” would be like emitting a
sympathetic cry, thereby expressing one’s pity for her, as when a parent exclaims:
“Ouch!” upon seeing her child hurt. Notice that this non-referential account goes
beyond the (by now familiar) claim that psychological discourse requires special
treatment, because it differs from non-psychological discourse in being subject to
various normative constraints.36 It also goes beyond the no-ascription account we
canvassed at the end of the previous section. For, the account we are considering now
requires maintaining that terms such as “is feeling pain” or “is afraid of the dog” or “is
thinking about Vienna” simply do not serve to pick out states of individuals in any of
their uses, first-person or third-person, present or past. Thus, an utterance such as: “DB
has a headache” would be understood as (only) expressing your sympathetic attitude
toward me and not at all referring to a state of mine. Or, perhaps it could be taken to
express some other more complex attitude on the part of the attributer, having to do
with her readiness to take a special ‘stance’ toward me (à la Dennett).
It should be clear that an account along these radical lines is not sustainable. The
difficulty most relevant to us here can be seen by considering how we can consistently
maintain that psychological terms do not refer, while at the same time holding that
psychological attributions serve to express attributers’ states of mind.37 For, it seems clear
that this claim itself requires quantifying over states of mind. If psychological terms do
not refer in any context, and sentences involving such terms never serve to represent
truth-evaluable judgments about mental states, what are we to make of the expressivist
account of avowals itself? The trouble for the psychological expressivist who opts for a
thoroughgoing no-ascription account is that her alternative account of psychological
35
Wright 1998, p. 35.
36
See, e.g., Davidson, 1973 and 1984, and Dennett 1987, passim.
37
Could the anti-realist avowal expressivist hold that her expressivist thesis itself is non-factual and serves
only to express some of our attitudes (as suggested to me by Matthew Chrisman)? Perhaps. But then it is
unclear to me how we are to assess the apparent disagreement between the psychological anti-realist and the
psychological realist.
discourse must appeal to items of the very same kind she takes to be questionable,
namely, mental states. Claiming that, even when ascribing mental states to others, we
never report or describe those mental states, but only express our own attitudes and
mental states, still requires acknowledging that there are mental states to be expressed.
(This, of course, leaves open various accounts of what the nature of those states might
be.) By contrast, claiming that when we make ethical pronouncements we never
report or describe moral properties of things—people, acts, states of affairs—but only
express our own pro- and con- attitudes toward things, only requires acknowledging
that there are pro- and con- attitudes to be expressed; it does not require acknowl-
edging that there are moral properties.
The foregoing can serve to elaborate on a complaint Wright makes against expres-
sivism “applied to ordinary psychology under the aegis of metaphysical anti-realism.”38
Focusing on what I described earlier as the expressivist’s positive claim, Wright points
out that it requires separating states of mind into different kinds—feelings, as opposed
to beliefs, for example. However the separation proceeds,
[t]he very claim that a discourse is expressive will presuppose an underpinning in facts about
aspects of the characteristic attitudinal psychology of its participants. But facts of that genre are
just what metaphysical anti-realism about psychology is unwilling to countenance. So expressi-
vism can offer it no consistent outlet.39
The problem, in a nutshell, is that the expressivist’s positive claim depends on the
invocation of the very same kind of states that the metaphysical anti-realist concerning
ordinary psychological discourse wants to shun.
However, even if psychological expressivism could somehow avoid ‘teetering into
incoherence’ (as Wright puts it), it should be recognized that it is not an attractive
option in the context of solving the puzzle with which we began—viz., how to explain
the distinctive security of avowals. Such an account, even if it were deemed adequate
to explain the contrast between psychological and non-psychological discourses,
would undermine the expressivist’s attempt to account for the asymmetries between
avowals and other mentalistic ascriptions. As we have construed it, psychological
expressivism attempts to account for all psychological talk, on the assumption that
psychological terms do not refer anywhere, and thus that sentences containing them
cannot be used to describe or report psychological facts. But then what is to account for
the distinctive character of avowals as contrasted with third-person or first-person past-
tense mental reports? As presented earlier, the central idea of simple avowal expressi-
vism was that when I ascribe a mental state to myself (at least when I do so in the
‘avowing mode’), I do something different in kind from what others do when they
ascribe a mental state to me. But this contrast would seem to escape the expressivist
account construed as an anti-realist account that applies to all mentalistic discourse.
38 39
Wright 2002, p. 212. Wright 2002, p. 213; see also Wright 1995b, pp. 204–5.
176 DORIT BAR-ON
The upshot of our discussion so far is this. There is a crucial difference between
expressivism as it has traditionally been advanced in metaethics and expressivism about
avowals. In the ethical case, expressivism is invoked to preserve the cogency of ethical
discourse even when it’s admitted that it differs from ordinary descriptive discourses in
lacking a proper grounding in matters of fact. If we take expressivism about psycho-
logical discourse to have a similar agenda, we court incoherence. Moreover, the puzzle
for whose solution expressivism may be needed pertains to an epistemological contrast
that arises within psychological discourse. And this puzzle is not addressed by psycho-
logical expressivism. Indeed, the puzzle remains even if we set aside ontological qualms
about psychological discourse and take it to be straightforwardly factual.
4. Wright’s Default View

After explaining that avowal expressivism can adequately address the ‘Geach point,’
Wright raises other difficulties for the view, and concludes: “[T]he expressivist proposal
flies rather further than is usually thought. But it is a dead duck all the same.”40 In its
place, Wright considers another (allegedly Wittgensteinian) anti-realist view, which he
dubs “the default view.” In this section, I’d like to argue that the default view suffers a
defect similar to the one identified for psychological expressivism toward the end of the
previous section.
As Wright presents it, the default view, like avowal expressivism, and unlike the
introspectionist view, denies that an avowal can represent a genuine “cognitive
achievement, based on cognitive privilege” that can, in turn, explain so-called first-
person authority.41 On the default view, the first-person/third-person asymmetries
“belong primitively to the ‘grammar’ of the language-game of ordinary psychology.”
The fact that psychological discourse exhibits these features should simply be regarded
as part of what makes it psychological discourse.42 On this account, then, the special
status enjoyed by avowals is not due to the special nature of the states they ascribe (viz.,
mental, as opposed to non-mental), or the special epistemic relation between subjects
and their own mental states. Rather, it is “a constitutive principle” of mentalistic dis-
course, which “enters primitively into the conditions of identification” of the subject’s
mental states.43
In contrast to avowal expressivism, however, the default view does not deny truth-
evaluability to avowals. Rather, it proposes a reconstrual of the truth-conditions of all
psychological ascriptions. The truth-conditions of such ascriptions are governed by a
special constraint, which Wright formulates as follows: “unless you can show how to
make better sense of her by overriding or going beyond it, [a subject’s] active self-
conception, as manifest in what she is willing to avow, must be deferred to” (Wright
1998, p. 41). Instead of denying truth-conditions to avowals, then, the default view
40
Wright 1998, p. 38. I address these difficulties in sections of Chapters 8–10 of Bar-On 2004.
41 42 43
Wright 1987, p. 402. Wright 1998, pp. 41–3. Ibid., p. 42.
maintains that mastery of the truth-conditions of all mentalistic ascriptions requires

recognizing that, unless there are good reasons to the contrary, one must accept what a
subject is willing to avow, and only make ascriptions to her in accordance with what
she herself is willing to avow. The epithet “default view” is apt, since the view presents
“a subject’s opinions about herself ” as “default authoritative and default limitative.”44
The default view takes the practice of psychological ascription to provide a kind of
‘bedrock’ in accounting for asymmetries between avowals and other ascriptions. On
this view, the presumed truth of avowals and the fact that they are not subject to
ordinary epistemic assessment are to be regarded as simply constitutive of the concepts
involved. The default status of avowals is part of what defines our concepts of mental
states; it is not a product of non-conceptual facts about subjects’ mental life or their
relationships to their mental states. Psychological ascriptions are all ultimately answer-
able to the subject’s own avowals because this is how psychological concepts work. This is the
sense in which the default view presents the epistemic asymmetries between avowals
and other ascriptions as a consequence of ‘grammar,’ rather than substantive epistemo-
logical differences. At the same time, however, the default view, unlike the simple
avowal expressivism, can accommodate the semantic continuities between mental
ascriptions to oneself and to others. For there is nothing in the default view as described
so far to stand in the way of taking the surface grammar similarities between these two
types of ascriptions at face value. For example, nothing seems to stand in the way of
taking my self-ascription “I am in pain” to share truth-conditions with your ascription
“She is in pain.”
By making the asymmetries out to be a matter of a conceptual constraint built into
psychological discourse, the default view can offer a clear contrast between psycholog-
ical and non-psychological ascriptions. As Wright intends it, the default view is an anti-
realist and non-factualist view of psychological discourse, which presents our practices of
psychological ascriptions, in contrast to our practices of non-psychological ascriptions,
as ultimately “not accountable to any reality.”45 However, we must recall that the
epistemic asymmetries that give rise to our puzzle include ones that arise within
psychological discourse; they pertain to differences between avowals and other psycho-
logical ascriptions—specifically, ascriptions of mental states to others, or to oneself at
other times, and most notably ‘reportive’ present-tense “I”-ascriptions that are based
on evidence, inference, etc. We still need to hear what distinguishes applications of
psychological concepts in one’s own case, when made in the course of avowing. On the
face of it, it doesn’t look as though an account that builds first-person authority into
the truth-conditions of all psychological ascriptions is equipped to capture those
44
Ibid, p. 41. The default view is also discussed in Wright 1987, pp. 401–402, 1989, sec. IV and 1991,
sec. IV. Note that it is not obvious how the above constraint captures the truth-conditions of avowals, as
opposed to describing our practice of accepting them as true. It is far from obvious how to work out the link
between the fact that the linguistic practices surrounding avowals are governed by the constraint, on the one
hand, with the conditions under which avowals are true. (Thanks to Dylan Sabo here.)
45
Wittgenstein 1974, }133 as quoted in Wright 1998, p. 39.
178 DORIT BAR-ON
asymmetries that arise within psychological discourse, and that seem sensitive not (or at
least not only) to the semantic content (or truth-conditions) of avowals, but to the special
way in which they are issued.
In offering the default view as an alternative to the Cartesian ‘privileged access’ view,
Wright is partially motivated by an analogy to other areas of discourse where it seems
misguided to conceive of successful judgments as tracking a completely independent
reality—e.g. color judgments, and, on some views, ethical judgments. In these areas,
Wright thinks it may be explanatory to suppose that there is judgment-dependence: what
is true about an object’s color, for example, may not be entirely independent of the
color-judgments of a well-placed perceiver.46 In the case at hand—the mental or
psychological realm—the claim would be that whether, for example, the avowal “I feel
awful” is true or not, and whether or not the self-ascriber does feel awful, is not entirely
independent of what she thinks about her state. Since this dependence on the subject’s
own verdict constrains any ascription of a mental state to her, if someone said of me:
“She feels awful,” the truth of her ascription would presumably also not be indepen-
dent of what I take to be the case about my state. If so, it should be no surprise that
there is such a fit between what a subject says (or thinks) about her condition and the
truth of the matter, and ultimately no point in seeking a substantive explanation of the
asymmetries between avowals and other ascriptions; hence, the default view.
However, the mentalistic case seems different from paradigm cases of judgment-
dependence. In the former case, what calls for explanation are (partly) the asymmetries
between a subset of mentalistic “I”-ascriptions (avowals) and other mentalistic ascrip-
tions, including certain mentalistic “I”-ascriptions. Yet the notion of judgment-
dependence by itself does not seem apt to capture these asymmetries. In paradigm
cases (such as that of color), the idea of judgment-dependence is invoked to capture the
intuition that a suitably well-placed judge can presumably not go wrong when making
the relevant judgments. But in the case of avowals the intuition is not simply that any
well-placed judge is authoritative regarding any arbitrarily chosen mental fact. Rather,
the intuition is that subjects may be well-placed only with respect to their own present mental
states and only when avowing, so that being well-placed will systematically depend on
whether the relevant mental state is the subject’s or someone else’s. But then we may
still be under explanatory pressure: why is it that subjects are to be systematically
presumed better placed then their observers to pass ‘truth-determining’ judgments on
certain facts in the mental realm, only some times, and only when avowing?
A proponent of the default view may object that the above complaint misses the
point. The whole idea behind the view is to deny that there is any substance to the idea
of first-person authority, anything ‘behind’ the special status assigned to avowals,
beyond the conventions of mentalistic discourse. If so, the above complaint simply
serves to expose how radical—and ultimately unsatisfactory—the default view really is.
46
For discussion of this idea, see e.g. Wright, 1989: sec. 3 and 1992, Chapter 3, appendix.
To adopt the default view is in effect to deny that “the features of avowals . . . which
seem to betray something remarkable about self-knowledge” reflect any deeper facts
about the subject-matter of avowals, or subjects’ relationship to it (Wright 1998, p. 39).
The view is that there is nothing to explain concerning the special security of avowals.
We can note the security as marking a constitutive feature of psychological discourse;
but it is futile to try to offer a substantive explanation of it. But then, as Wright himself
remarks in criticism of the view, adopting the default view may be nothing more than
“merely an unphilosophical turning of the back” on issues surrounding avowals and
privileged self-knowledge.47
Like the view we labeled “psychological expressivism,” the default view attempts to
solve our puzzle regarding the status of avowals by focusing on conceptual or ontolog-
ical features that separate psychological discourse as such from ordinary descriptive
discourses with unproblematic subject-matters. But while our puzzle does encompass
contrasts between mental and bodily ascriptions, it is in the first place a puzzle that
concerns the contrastive epistemology of avowals, and it also encompasses contrasts
arising within psychological discourse and even within the class of mental “I”-ascrip-
tions. For this reason, I have argued, neither psychological expressivism nor the default
view is equipped to offer a satisfactory solution to it.
5. A Neo-Expressivist View of Avowals

Let us take stock. We have rejected the materialist introspectionist view on the grounds
that, in its eagerness to put mental and non-mental ascriptions on an ontological par, it
failed to respect those features of avowals that set them apart from certain nonmental
present-tense self-ascriptions. Simple avowal expressivism addresses that problem by
claiming that avowals, unlike all other ascriptions, are simply expressions of avowed
mental states rather than descriptive reports about them. But the price it pays is that of
compromising the semantic continuities between avowals and other ascriptions. Our
study of psychological expressivism and the default view revealed that it is misguided to
seek an account of the contrast between avowals and other ascriptions that is fashioned
after anti-realist accounts in other areas.
Returning to our earlier separation of the positive expressivist claim from the negative
ontological claim, as well as from the negative semantic claim, we can see in outline what
shape an adequate solution to our puzzle may take. The positive expressivist claim
seems to hold much promise for explaining the epistemic asymmetries between
avowals and other ascriptions. But the temptation to put it to the service of an anti-
realist agenda, by committing to the negative ontological claim, should be resisted, on
pain of incoherence. And, to avoid compromising the semantic continuities, the
47
Wright 1998, p. 41. For further discussion of the default view, see Boghossian 1989, McDowell 1998,
Fricker 1998, Bilgrami 1998, and Moran 2001. The discussion here follows closely my discussion in Bar-On
2004, pp. 347–50, 410–12.
180 DORIT BAR-ON
expressivist claim must be decoupled from the negative semantic claim.48 My main task
in this section is to flesh out this outline and present what I regard as a viable neo-
expressivist account of the special security of avowals.
As a preliminary, let us return to the case of self-ascriptions issued through proprio-
ception or kinesthethesis. We should note that such self-ascriptions, which surely
represent genuine truth-evaluable reports, nevertheless share a certain epistemic feature
with avowals. Proprioceptive and kinesthetic self-reports are “identification-free”:
epistemically speaking, they do not rest on a recognitional identification of the subject
of the utterance or thought. That is to say, in normal circumstances, if I say (or think):
“My legs are crossed” or “I’m spinning around,” my utterance/thought does not rest on
my recognizing some individual as myself. I do not identify someone as being me and take
that person’s legs to be crossed. I have no more reason for thinking that someone’s legs are
crossed than whatever reason I have for thinking that my legs are crossed. This is what
renders bodily self-reports of this kind “immune to error through misidentification” (to
use terminology originally employed by Gareth Evans and Sidney Shoemaker).49
The following features of ascriptions that are immune to error through misidentifi-
cation (IETM) are crucial for our present purposes. (i) The class of ascriptions that are
IETM is heterogenic; it includes mental self-ascriptions (“I have a toothache”),
perceptual self-reports (“I see a canary”), as well as a variety of bodily self-reports
(“I am sitting down,” “I am bleeding,” “My legs are crossed”). (It also includes some
demonstrative ascriptions, such as “That is moving fast.”) (ii) When a self-ascription of
the form “I am F” is IETM, although I may fail to be F, I cannot have reason for
thinking that someone is F that isn’t at the same time reason for thinking that I am F. And
I cannot as much as doubt that it is me who has the relevant properties without doubting
that someone has them. Thus, although my self-ascription may be in error, the error is not
one of recognitionally mis-taking one individual for another; it is not an error through
misidentification. (iii) Whether or not a self-ascription is IETM depends not on its semantic
content, but rather on its “epistemic pedigree.” One and the same (semantically
individuated) self-ascription can be IETM or not, depending on the basis on which it
is made. For example, if I recognize that my legs are crossed by seeing someone in a
mirror and taking her to be me (perhaps mistakenly), my self-ascription will not be
IETM. (iv) In the case of avowals, though, I seem to enjoy additional, ascriptive security:
not only are avowals immune to error of misidentification, but unlike all nonmental
self-reports (as well as mental third-person ascriptions or past-tense self-ascriptions or
evidential mental self-ascriptions), I submit that they are also immune to error through
misascription. When avowing “I’d like some water,” for example, not only do I have no
48
Wright, too, separates off the positive expressivist claim as one that a realist need not deny, see Wright
1988, p. 30.
49
For discussion of the phenomenon of immunity to error through misidentification, see Wittgenstein
1958, 66–7, Evans 1982 esp. Chapter 7, sec. 2, Shoemaker 1968, and Wright 1998, pp. 18–20. I discuss the
phenomenon and its relevance to explaining the security of avowals in Bar-On 2004, Chapters 4 and 6.
reason for thinking that someone wants some water, other than whatever reason I have for
thinking it’s I who wants water. I also have no reason for thinking that I want some
substance that isn’t also a reason for thinking that I want water. And I have no reason for
thinking that I have some attitude or other toward water that isn’t also a reason for
thinking that I want water. As I self-ascribe a mental state in the normal, ‘avowing’ way
(e.g. declaring “I have a terrible headache!,” walking into a room and saying “I am so
annoyed at this mess!,” looking at the sky and saying “I think it’s going to rain,” etc.), I do
not latch on to a present state of mine and take it to be M (a headache, or annoyance, or
thought that p). I have no reason for thinking that I am in some state or other, over and
above or separately from whatever reason I have for thinking I am in M.50 If my avowal is
false, this will not be due to a recognitional error of mis-taking one of my mental states for
another. I can make such recognitional errors, however, if I were to self-ascribe M on the
basis of some evidence, observation, interpretation, therapy, etc.
To say that avowals are immune to error through misascription is, then, not to say that
they are absolutely infallible or incorrigible. It’s just to say that they are protected from a
certain array of epistemic errors (and thus corrections)—a much wider array than all other
ascriptions, including, specifically, proprioceptive and kinesthetic self-reports, as well as
evidential mental self-ascriptions. If I say or think: “My legs are crossed” (in the normal
way), then even if I cannot be misidentifying who it is whose legs are crossed, if my legs
are not crossed, my self-report will be mistaken, precisely because I have misidentified
the state of my legs: I will have mistaken one state of my limbs for another. In the case of
avowals, though, I suggest, I am protected from epistemic error in the ascriptive part as
well. To reiterate, my avowal is not guaranteed to be correct and forever beyond
correction. As we saw, in certain circumstances we may even be led to question an
avowal of pain. But I would argue that these are not circumstances in which the avower
is simply falsely ascribing to herself a state on the basis of a mistaken belief that she is in
some other state. (After all, in the circumstances, she would be equally disposed to emit a
spontaneous yelp. And, on pain of regress, it is implausible to suggest that the yelp would
be the result of a false belief based on how her present state appears to her.)
The characterization of avowals’ distinctive security as a matter of their unique
immunity to error through misascription provides a suitably tempered interpretation of
the claim that avowals are incorrigible and indubitable, an interpretation which does
not require invoking either Cartesian privileged access or, indeed, any distinctively
secure epistemic basis on which avowals supposedly rest. Though on this characteriza-
tion avowals are not portrayed as absolutely incorrigible or indubitable, we can better
understand the specific ways in which they are not open to epistemic challenge, doubt,
and criticism, unlike other empirical ascriptions. The characterization also gives a more
precise sense to the intuition that avowals are epistemically more immediate or direct
50
If, as Wright says, the avowal is groundless, then I have no reason for thinking that I am in
M. Alternatively, it may be suggested that I do have a reason—though it isn’t a justifying reason: it is simply
my being in M that gives me reason. (I consider this suggestion in Chapter 9 of Bar-On 2004.)
182 DORIT BAR-ON
than other ascriptions. However, by itself, the characterization does not give us a full
explanation of avowals’ special security. For one thing, we need to understand the
source of the additional immunity to error that avowals enjoy.51 Why is it that avowals
are not only immune to error through misidentification? Moreover, we have seen that
immunity to error, in general, is no guarantee of truth. Yet avowals contrast with all
other ascriptions in being very strongly presumed to be true.
It is at this point that I think we should co-opt the key insight from simple avowal
expressivism: the idea that avowals’ security is to be explained by appeal to their
expressive character, rather than by appeal to the security of the epistemic basis on
which they are made (be it introspection, or any other kind of epistemic basis). Like
simple avowal expressivism, avowal neo-expressivism maintains that the epistemic sim-
plicity and immediacy of avowals is best explained by seeing them as direct expressions of
subjects’ self-ascribed mental states. But in contrast to simple avowal expressivism, neo-
expressivism maintains that avowals—unlike natural expressions, and like various
mental reports—represent genuine, truth-evaluable ascriptions. In what follows,
I briefly introduce certain key distinctions that would allow us to give more precise
sense to the neo-expressivist claim that avowals are similar to natural expressions in
being directly expressive of subjects’ mental states, even though, semantically speaking,
they are truth-evaluable self-ascriptions. I will then return to the question of the
explanation of avowals’ security.
The first distinction is one borrowed from Wilfred Sellars 1969, between two
different senses (or kinds) of expression.
a-expression: in the action sense, a person expresses a state of hers by intentionally doing something.
For example, when I intentionally give you a hug, or say: “It’s so great to see you,”
I express in the action sense my joy at seeing you. (One may also express one’s feeling
of sadness in the action sense by letting tears roll down her cheeks, instead of wiping
them off and collecting oneself.)
s-expression: in the semantic sense, e.g., a sentence expresses an abstract proposition, thought or
judgment by being a (conventional) representation of it.
For example, the sentence “It’s raining outside” expresses in the semantic sense the
proposition that it is raining at time [t] outside place [p].52
51
In the case of proprioceptive reports, the source of their immunity to error through misidentification
has to do with our possessing special mechanisms for obtaining information concerning our own bodies. See
Evans 1982, Chapter 7.
52
Sellars also speaks of expression in the causal sense, where an utterance or piece of behavior expresses an
underlying state by being the culmination of a causal process beginning with that state. (For example, one’s
unintentional grimace or shaking hands may express in the causal sense one’s pain or nervousness, respec-
tively.) For present purposes, we can set aside ‘c-expression,’ though it plays a role in the full neo-expressivist
account. See Bar-On 2004, pp. 216f., 248f., and passim.
As we consider the respects in which avowals may resemble natural expressions, we

should be focusing on natural expressions that fall under a-expression—where the
relevant mental state is the rational cause of, or reason for, one’s expressive behavior.
Furthermore, as we consider these cases, we must distinguish between the act of
expressing and its product. Like other nominals in English (and other languages), such
as “statement,” “assertion,” “ascription,” “report,” as well as, indeed, “expression,”
etc., “avowal” admits of two different readings. It can be read as referring to someone’s
act of avowing, which is an event in the world with a certain causal history and certain
action-properties. But it can also be read as referring to the result or product of such an
act—a linguistic (or language-like) token, an item with certain semantic properties.
The important similarities between avowals and other intentionally produced expres-
sions (intentionally produced natural expressions included) require thinking of avowals
as acts, whereas the salient dissimilarities come to the fore when we think of avowals as
products.
The product of an act of avowing, unlike a smile or a wince, or even a verbal cry
such as “Ouch!,” is a semantically articulate self-ascription, an item with semantic
structure and truth-conditions. It is a product whose properties allow it to serve, and be
caught up, in other kinds of distinctively linguistic (and mental) acts. Importantly,
avowals, understood as products, s-express (in Sellars’ semantic sense) self-ascriptive
propositions, to the effect that the avower is in some state. And the mentalistic terms
that constitute parts of avowals semantically represent the relevant states. Natural
expressions such as a facial expression, a gesture, a bodily movement, a demeanor, an
inarticulate sound, when understood as products, and however produced—do not on
their face s-express anything. There are no semantic conventions in virtue of which
laughter represents amusement, no linguistic rules in virtue of which a hug signifies, or
refers to, joy at seeing the person hugged, or a cowering demeanor stands for feeling
threatened.53 Thus there are notable differences between avowals and acts of natural
expression in terms of their products. But I think that the expressivist insight regarding
avowals should be understood, in the first instance, as a claim about the relevant acts,
not about their products. The claim is that there are notable similarities between the act
of avowing a state and the act of giving it a (so-called) natural expression; e.g. between
the act of saying: “I really like this!” (or, alternatively, “This is wonderful!” or
“Yippee!”), and the act of giving an ear-to-ear smile, or a firm handshake. And this
claim could be true, even if there were systematic differences between the products of
acts of avowing and the products of naturally expressive acts.
53
Although, of course, there are gestures and other non-verbal expressions that are governed by socio-
cultural conventions. The conventions do not, however, assign semantic content to the gesture, but rather set
up a (pragmatic) connection between the making of the gesture (e.g. tipping one’s hat) and being in the
relevant state, or having the relevant attitude, sentiment, etc. (e.g. feeling respect).
Also, a case can be made that, e.g., an animal’s alarm call has semantic content—representing, e.g., threat
from above or whatever. Still, unlike an avowal, it cannot be plausibly taken to express in the semantic sense a
proposition about the relevant mental state of the animal issuing the call.
184 DORIT BAR-ON
Thus consider a paradigm case of expressive behavior: a small child, Jenny, eagerly
reaching for a teddy bear. Jenny simply wants the toy, and her reaching for it gives
direct expression to her desire, quite independently of any judgment on her part to the
effect that she wants the teddy. Jenny wants the teddy and her (non-verbal) behavior
shows it; she does nothing to hide her desire for teddy. And in normal circumstances,
her audience will be directly responsive to the state they perceive her to be in; barring
contravening reasons, they will simply hand her the toy. But now consider another
episode in which Jenny emits a certain sound (“Uh!”), or calls out: “Teddy!” as she
reaches for the toy. And finally, consider an episode in which she avows: “I want
Teddy,” perhaps with no reaching at all. Intuitively, the verbal emissions, just like the
reaching behavior, which we would consider a natural expression of Jenny’s desire,
‘come directly from’ the child’s desire. They seem equally ‘pressed out’ from her, and
they appear no more driven by a prior deliberation, consideration, or determination
regarding how things are presently with her. The verbal utterances seem to be equally
expressive of her desire for the teddy. Granted: when saying “I want Teddy!” Jenny
makes intentional use of conventional verbal means that are acquired. If her saying
“I want the Teddy” itself constitutes expressing her desire, the expression will not be a
natural one (at least in one sense of the term). But notice that the same would hold of
her saying “Daddy!” to express her happiness at seeing Dad, or her saying “Ouch!” to
express her pain. Very early on in our lives, we begin to use acquired, conventional
means—both words and gestures—to give expression to our states.
To vary the example, consider a linguistic utterance such as “It’s so good to see you!”
It typically serves to express the speaker’s joy. But, it seems, one can equally express her
joy by avowing “I am so glad to see you!” Armed with the act/product distinction, as
well as the distinction between a- and s- expression we can capture the similarities and
differences between the two utterances. They can both be seen as instances of the same
type of act, and they both a-express the same type of mental state. But the products of
these two acts, though both linguistic, are (semantically) different. The first sentence
token s-expresses the proposition that it is very good to see one’s hearer, in virtue of the
rules of English. The second sentence token is self-ascriptive: it s-expresses the proposi-
tion or thought that the speaker is very glad to see her hearer. By contrast, the product of
an act of giving your friend a cheerful hug is not governed by any syntactic or semantic
rules of English, and it does not s-express anything. Still, it can equally be seen as an act
of the same type, which a-expresses your joy at seeing your friend.
What this illustrates is that our ordinary everyday notion of expression applies readily
to non-verbal and verbal expressions alike, and to verbal expressions of a wide variety.
The commonsense notion is that of an act, whether linguistic or nonlinguistic, whose
point is none other than either to make public, put forth, share, or air, or else just to
give vent or voice, rather than to offer a descriptive report, make an assertion, or inform
someone about the speaker’s present thoughts, feelings, emotions, or attitudes. The
notion seems to apply even when the expressive vehicle used is not ‘natural’ but is
rather introduced in some conventional way and is acquired. It is also applicable to acts
performed in the absence of audience, as when someone spits at a photograph or

caresses an object while alone, as well as to mental acts such as saying to oneself: “Oh
my god!” or “That bastard” or “What a mess!”
On avowal neo-expressivism, then, avowals emerge as a certain class of expressive
acts; acts in which a subject gives linguistic vent to present mental states. These are acts
in which subjects speak their minds, instead of giving non-linguistic expression to it, and
the linguistic vehicles they use are self-ascriptive. Acts of avowing are epistemically
unmediated; but for all that, they have products (sentence tokens, in the case of speech,
articulate thought tokens, in speechless cases) that can be true or false. And we have
seen that this is not without precedent. When I exclaim “This dog is scary!” I may be
giving voice to my present fear, using a linguistic vehicle—a sentence token—that is
true or false, depending on whether the designated dog is scary or not. Similarly, when
I avow “I am scared of this dog,” I may be giving voice to my fear, using a sentence
token that is semantically about me and a state of mine, and is true or false depending
on whether I am scared of the designated dog or not.54
We are now in a position to return to one of our earlier questions. We have noted
that, even though what avowals s-express are not a priori, or necessary propositions,
and although when avowing subjects make epistemically unmediated, ‘groundless’
claims, avowals are quite rarely challenged or corrected, and not because they cannot
be negated (like moans or verbal expletives), it is not simply that avowals go unchal-
lenged; they are also strongly presumed to tell us true things about their subjects. On the
neo-expressivist account, the reason is this. When I avow, I a-express the very same state
whose presence makes true the proposition s-expressed by the avowal (understood as
product). To take me to be avowing, then, is to take it that I am in the designated state,
which is to take the avowal to be true.55 (We have already touched on the question
how the expressivist idea helps explain the various epistemic asymmetries.)
Our puzzle regarding avowals was this:
Why is it that avowals, understood as true or false ascriptions of contingent states to an individual,
are so rarely questioned or corrected, are generally so resistant to ordinary epistemic assessments,
and are so strongly presumed to be true?
This puzzle invites a search for a distinctive feature(s) of avowals that would explain
why they enjoy the level and kind of security that we ordinarily assign to them. If we
are not to turn our back on the puzzle, as the default view arguably does, yet at the
54
The observation that is crucial for the neo-expressivist account is that, in point of its potential to express
the subject’s (self-ascribed) present state of mind, a self-ascription such as: “I am so annoyed at you”
(produced in speech or in thought) is no different from other verbally expressive acts, e.g.: “This is very
annoying,” or “Enough!” or even “Ugh!”.” In Bar-On 2004, Chapter 7, I argue that it is further plausible that
these latter expressions can be expressive of a subject’s mental state in the same sense as intentional acts of
natural expression, such as letting out a sound or making a face or a gesture that are not conventional. But the
further claim is not strictly part of the core neo-expressivist thesis.
55
The complementary question that then arises for the neo-expressivist is how to understand the
possibility of false avowals. I discuss this issue at length in sections of Chapters 7 and 8 of Bar-On 2004.
186 DORIT BAR-ON
same time improve on the materialist introspectionist view of first-person authority, we

must show why it should be reasonable for us to treat avowals as protected from certain
kinds of epistemic mistakes and criticisms, even if not absolutely infallible and incorri-
gible, and to assign avowing subjects inalienable and non-transferable first-person
authority. On the neo-expressivist story, it is reasonable for us to assign inalienable
and non-transferable first-person authority to subjects, because this authority is assigned
on the strength of recognizing their avowals as acts in which they directly express the
very states they self-ascribe through the products of those acts at the same time as they self-
ascribe them.
6. Expressivism, Truth, and Realism

As mentioned earlier, Wright does not think avowal expressivism is ruled out of court
by the Geach point, but he has other worries about the expressivist proposal. I believe
avowal neo-expressivism is equipped to address the remaining worries Wright raises for
avowal expressivism.56 But my concern here is rather with Wright’s reasons for taking
expressivism in general to represent a ‘faux pas.’ Deploying the distinctions that
(I believe) allow avowal expressivism to rise from the dead, might we not be able to
articulate a neo-expressivist paradigm that is properly expressivist but neither requires
denying truth-evaluability nor is driven by an antecedent anti-realist agenda?
First, let us note that avowal neo-expressivism, like simple avowal expressivism
properly construed,57 is neither motivated by nor committed to psychological anti-
realism. The view is strictly designed to explain the epistemic asymmetries between
avowals and other ascriptions. Like simple avowal expressivism, the neo-expressivist
view does not explain avowals’ distinctive features by appeal to their secure epistemic
basis, but instead appeals to their expressive character. However, unlike the simple
view, it does not take the security to be a matter of the semantics or ‘grammar’ of
avowals understood as products; for avowals qua products are said to share truth-
conditions with other ascriptions (e.g. “I am tired” avowed by me is true iff DB is
tired). The neo-expressivist positive claim is confined to what avowals a-express and
does not concern what avowals (or rather their products) s-express. It does not get
coupled with a negative semantic claim (a denial of truth-conditions and thus of truth-
evaluability).
But suppose we now ask: What are the ‘truth-makers’ of avowals—what renders an
avowal true or false? The answer, as far as neo-expressivism is concerned, is straight-
forward: facts about the mental lives of subjects. There is no presupposition that such
facts are problematic, nor even that there is any systematic ontological contrast between
psychological and other facts. Moreover, facts about the mental life of any subject can
in principle be reported by anyone, including the subject herself. It’s just that when
56 57
See footnote 28. See above, Section 3.
avowing, subjects do not set out to issue reports about their states of mind. Suppose
I avow: “I am tired” (instead of just stretching and giving a big yawn), or “I’m feeling
upset about my brother’s not showing up to my party” (where presumably I couldn’t
engage in any non-verbal behavior expressive of the relevant state, though I could
express my state by saying: “It’s very upsetting that my brother didn’t show up to my
party”). I can turn around and make a past-tense descriptive report made true by the
same state of affairs: “I was tired then,” or “I felt upset about . . . ”. Indeed, at the time
I issued an avowal, I might have issued a present-tense evidential self-report instead.
Had circumstances been different, I might have said “I am tired” on the basis of seeing
my reflection in the mirror, or I might have concluded that I am feeling upset about my
brother’s absence on the basis of a therapy session, discussion with a friend, etc. The
neo-expressivist explanation of avowals’ distinctive security focuses directly on episte-
mological and action-theoretic differences between avowals and their corresponding
self-reports, rather than on any alleged ontological differences between the mental and
other realms.
Unlike the answer given by wholesale psychological expressivism to the (essentially
epistemological) puzzle about avowals’ security, the neo-expressivist answer avoids
‘teetering into incoherence,’ since it does not presuppose that there are no states for any
psychological sentences to describe. Moreover, unlike the default view, it is not
committed to the strong dependence of psychological facts on the judgments of the
subjects who harbor psychological states. On the default view, being in a mental state is
partially constituted by thinking that one is in the mental state, so that facts in the
mental realm are determined in part by what self-ascribers take them to be. But there
are good reasons to avoid tying the presence of all mental states to the presence of
thoughts about them. For one thing, ‘basic’ mental states, such as hunger, thirst, pain,
etc. are ones that we seem able to share with non-human animals and pre-cognitive
children, who are presumably not equipped to entertain thoughts about their own
mental states. And when it comes to sophisticated psychological states, such as an
occurrent desire to have one’s political rival be embarrassed on a given occasion, it
seems quite possible for someone to have that desire without realizing it. Moreover, it
seems perfectly possible for one to form the judgment that one is in some psychological
state even though one is not. Passing judgment on a contingent matter of fact is a risky
business, in the case of mental states as elsewhere. However reliable one is, one can get
things wrong. There are all sorts of ways of coming to make a judgment that one feels
depressed, or hopes for something, or wants something, or even is in pain, and they can
lead us astray. There may well be particular mental phenomena that do fit Wright’s
characterization of the judgment-dependence of mental ascriptions, where one’s
judgment that one is in the mental state actually makes it the case that she is, or at
least partially constitutes her being in the state.58 But I do not believe this to be the
58
See Taylor 1985, pp. 100–1, Moran 1988, and Tanney 1996.
188 DORIT BAR-ON
general case. In general, I think we should allow that judging that one is in some mental
state does not make it so.
All this tells against conceptually tying the very existence or presence of mental states
to subjects’ judgments. And nothing in the neo-expressivist view of avowals requires
such a conceptual link. Not only does the view allow for the possibility of being in a
mental state without forming any judgment about it, and of forming false judgments
about one’s present mental states on this or that basis, but the view also allows for the
possibility of false avowals. Unlike views that explain avowals’ security by appeal to
the ‘logical grammar’ of avowals in particular or of mental discourse more generally,
neo-expressivism does not associate with avowals any conceptual guarantee of truth.
It also does not require accepting that mental states in general are of subjects’ own
making.59
In the particular case of avowals, there may be, then, good reasons for the neo-
expressivist to avoid strongly anti-realist views of mentality, without yet committing to
any particular view of the nature of mental states. Taking the lead from the neo-
expressivist view of avowals, we may generalize, and consider the following combination:
(i) Endorsing a positive expressivist claim concerning a given discourse D, i.e.
maintaining that utterances/thoughts that involve the ‘signature’ terms or
concepts of D characteristically (and perhaps essentially) serve to a-express
certain mental states of the speakers/thinkers.
(ii) Rejecting the negative semantic claim, i.e. maintaining that utterances/
thoughts in D are truth-evaluable in that they s-express propositions that can
be true or false.
(iii) Rejecting the negative ontological claim, i.e. refusing to commit to there being
no D-facts of the matter, and no D- objects, properties, or states to constitute a
proper subject-matter for discourse D.
In order for this combination to present a viable option, the expressivist claim has to
be put forth for some particular explanatory purpose other than the traditional expressi-
vist purpose of explaining how utterances of a given discourse can have a point even
though there are no facts for them to be about. It would be particularly instructive,
I think, to sketch such a neo-expressivist option for the case we originally contrasted
with the case of avowals: that of ethical discourse.60
According to a view we may dub ethical neo-expressivism, traditional ethical expressi-
vism was correct to identify an important, perhaps essential, expressive function of
ethical discourse and reflection: ethical claims (whether made in speech or in thought)
serve to express certain of our conative attitudes. Armed with the distinctions intro-
duced when presenting avowal neo-expressivism, we can retain this key idea without
59
For further discussion, see Bar-On 2004, Chapters 8 and10.
60
The application of the neo-expressivist framework to ethical discourse is the topic of Bar-On and
Chrisman 2009.
compromising the palpable semantic continuities between ethical and non-ethical

claims. As in the case of avowals, we can distinguish between the act of making an
ethical claim and the product of this act. An ethical claim considered as a product—i.e.
a sentence- (or thought-) token that essentially employs ethical terms (or concepts)—
can be said to s-express a true or false proposition. However, as with avowals, what is s-
expressed by a claim does not settle what mental state is characteristically a-expressed
by acts of making the claim. With Hume, we may think that purely cognitive states
(such as beliefs or judgments) cannot by themselves motivate or explain action, and
furthermore, we may think there is a non-accidental connection between sincerely
making an ethical claim and being motivated to act in its accord. The expressivist claim
tells us that ethical utterances (in my terminology) a-express the very same states whose
presence is required for understanding the perceived motivational force of such
utterances.61 As Ayer already saw, this expressivist insight is best captured without
supposing that (again, in my terminology) the products of ethical claims s-express
propositions that self-ascribe those states.62 Thus, as long as we’re talking about
expressive acts, we can agree with Ayer that ethical claims betray conative attitudes,
not because they report them, but because they (a-) express them directly. What the
neo-expressivist goes on to add is that we need not go on to accept Ayer’s claim that
ethical claims (understood as products) are not truth-evaluable because they do not (s-)
express any true or false propositions.
We can present ethical neo-expressivism as a combination of the following three claims:
(i) Ethical claims understood as acts are different from ordinary descriptive claims
in that agents making them (in speech or in thought) essentially a-express
conative attitudes.
(ii) Ethical claims understood as products are semantically continuous with ordinary
descriptive claims in being truth-evaluable, embeddable in conditionals, nega-
tion, logical inferences, etc. This is because ethical claims s-express true or false
propositions.
(iii) It is a matter for further metaphysical investigation to determine what (if
anything) makes ethical propositions true or false: whether there are ethical
properties or states of affairs, and if there are, what their nature is and to what
extent they are mind- (or judgment-) dependent remains to be told.63
61
They do so, whether or not they also express a belief whose content is given by the s-expressed proposition.
For some discussion, see Bar-On and Chrisman 2009.
62
Ayer 1946, pp. 104–8.
63
As noted above, in the ethical case, unlike in the avowal case, we are not courting incoherence by
leaving open the possibility that there are no ‘truth-makers’ for the relevant propositions. One may wonder,
however, what we are to make of the proposition allegedly s-expressed by an ethical claim, if we are to leave
open the possibility that there are no ethical properties or states of affairs. For our purposes here, we can
specify this proposition disquotationally. So, for instance, we’ll say that the claim “Tolerance is a virtue”
s-expresses the proposition that tolerance is a virtue. (To say this is not to commit to a disquotational theory of
the truth of ethical (or other) claims. It is just to avoid committing to the possibility of any paraphrastic
semantic analysis thereof.)
190 DORIT BAR-ON
Ethical expressivists, it seems, should welcome the possibility of decoupling the key
expressivist insight from unwelcome semantic commitments. But why should anyone
be drawn to expressivism about ethical discourse once it is divorced from the denial of
ethical facts? The answer already suggested is that the expressivist claim may be
explanatory of certain puzzling features of ethical discourse. Quite independently of
any ontological qualms about ethical facts or properties, we may find it illuminating to
recognize that ethical claims a-express conative states, since that may explain the
apparently non-accidental connection between ethical claims and motivation (at least
within a Humean framework). On the ethical neo-expressivist proposal, to take
someone as making a genuine ethical claim is ipso facto to regard her as giving voice to
a state whose presence explains why she is suitably motivated when making the claim.
Finally, I think Wright should welcome the introduction of a new expressivist
paradigm, since it may help capture several insights he himself has offered in his
extensive discussions of expressivism, realism, and truth.64 In particular, the new
paradigm seems very much in the spirit of Wright’s minimalist proposal, which allows
us to preserve truth-evaluability and truth talk wherever we find the right assertoric
trappings, while leaving it open whether we should come down on the side of realism
or anti-realism. What the new paradigm adds to the mix is the positive expressivist
claim; a claim that, I hope to have shown, we often have reason to invoke.65
Bibliography
Austin, J. L. 1961 “Performative Utterances,” in J. O. Urmson and G. J. Warnock (eds.)
Philosophical Papers, Oxford, Clarendon Press, pp. 233–52.
Ayer, A. J. 1946 Language, Truth, and Logic, 2nd edn., London, Gollancz.
Bar-On, D. 2004 Speaking My Mind: Expression and Self-Knowledge, Oxford, Clarendon Press.
Bar-On, D. 2012 “Externalism and Skepticism: Recognition, Expression, and Self-Knowledge,”
in A. Coliva (ed.) Self-Knowledge and the Self, Oxford, Oxford University Press, pp. 189–211.
Bar-On, D. and Chrisman, M. 2009 “Ethical Neo-Expressivism,” in R. Shafer-Landau (ed.)
Oxford Studies in Metaethics, Volume IV, pp. 133–66.
Bilgrami, A. 1998 “Resentment and Self-Knowledge,” in C. Wright, B. C. Smith, and
C. Macdonald (eds.), Knowing Our Own Minds, Oxford, Oxford University Press, pp. 207–41.
Blackburn, S. 1984 Spreading the Word, Oxford, Oxford University Press.
Blackburn, S. 1993 Essays in Quasi-Realism, Oxford, Oxford University Press.
Blackburn, S. 1998 “Wittgenstein, Wright, Rorty and Minimalism,” Mind 107, pp. 157–81.
Boghossian, P. 1989 “Content and Self-Knowledge,” Philosophical Topics 17, pp. 5–26.
Davidson, D. 1973 “Radical Interpretation,” Dialectica 27, pp. 313–28.
Davidson, D. 1984 “First Person Authority.” Dialectica 38, pp. 101–11.
Dennett, D. 1987 The Intentional Stance. Cambridge, Mass., MIT Press.
64
See Wright 1992 and the essays collected in Wright 2001 and Wright 2003.
65
I wish to thank Matthew Chrisman, Peter Hanks, and Ram Neta for helpful comments on previous
drafts. Many thanks to Jussi Suikkanen for both his comments and his bibliographical help.
Evans, G. 1982 The Varieties of Reference, Oxford, Clarendon Press.

Fricker, E. 1998 “Self-Knowledge: Special Access vs. Artefact of Grammar: A Dichtomy
Rejected,” in C. Wright, B. C. Smith, and C. Macdonald (eds.) Knowing Our Own Minds,
Geach, P. 1960 “Ascriptivism,” The Philosophical Review 69, pp. 221–5.
Gibbard, A. 1990 Wise Choices, Apt Feelings. Oxford, Clarendon Press.
Gibbard, A. 2003 Thinking How to Live Cambridge, Mass., Harvard University Press.
Hacker, P. 1993 Wittgenstein: Meaning and Mind, Oxford, Blackwell.
McDowell, J. 1998 “Response to Crispin Wright,” in C. Wright, B. C. Smith, and
Mackie, J. 1977 Ethics: Inventing Right and Wrong, London, Penguin.
Moran, R. A. 1988 “Making Up Your Mind: Self-Interpretation and Self-Constitution,” Ratio
NS 1, pp. 135–51.
Moran, R. A. 2001 Authority and Estrangement: An Essay on Self-Knowledge, Princeton, NJ,
Princeton University Press.
Ridge, M. 2006 “Ecumenical Expressivism: Finessing Frege,” Ethics 116, pp. 302–36.
Rorty, R. 1970 “Incorrigibility as the Mark of the Mental,” Journal of Philosophy 67, pp. 399–424.
Sellars, W. 1969 “Language as Thought and as Communication,” Philosophy and Phenomenological
Research 11, pp. 506–27.
Shoemaker, S. 1968 “Self-Reference and Self-Awareness,” Journal of Philosophy 65, pp. 555–67.
Stevenson, C. 1944 Ethics and Language. New Haven, Conn., Yale University Press.
Tanney, J. 1996 “A Constructivist Picture of Self-Knowledge,” Philosophy 71, pp. 405–22.
Taylor, C. 1985 “The Concept of a Person,” in Human Agency and Language: Philosophical Papers,
Wittgenstein, L. 1958 The Blue and Brown Book, New York, Harper & Row.
Wittgenstein, L. 1974 Philosophical Grammar, Oxford: Blackwell.
Wright, C. 1987 “On Making Up One’s Mind: Wittgenstein on Intention,” in P. Weingartner
and G. Schurz (eds.) Logic, Philosophy of Science and Epistemology, Proceedings of the XIth
International Wittgenstein Symposium, Vienna, Holder-Pickler-Tempsky, pp. 391–404.
Wright, C. 1988 “Realism, Antirealism, Irrealism, Quasi-Realism,” in H. Wettstein, T. Uehling
and P. French (eds.) Realism and Anti-Realism, Midwest Studies in Philosophy, vol. XII, Min-
neapolis, University of Minnesota Press, pp. 25–50.
Wright, C. 1989 “Wittgenstein’s Rule-Following Considerations and the Central Project of
Theoretical Linguistics,” in A. George (ed.) Reflections on Chomsky, Oxford, Blackwell,
pp. 233–64.
Wright, C. 1991 “Wittgenstein’s Later Philosophy of Mind: Sensation, Privacy, and Intention,”
in K. Puhl (ed.) Meaning Scepticism, Berlin, de Gruyter, pp. 126–47.
Wright, C. 1992 Truth and Objectivity. Cambridge, Mass., Harvard University Press.
Wright, C. 1995a “Truth in Ethics,” Ratio 8, pp. 209–26.
Wright, C. 1995b “Can There Be a Rationally Compelling Argument for Anti-Realism about
Ordinary (‘Folk’) Psychology?.” Philosophical Issues 6, pp. 197–221.
Wright, C. 1996 “Précis of Truth and Objectivity,” Philosophy and Phenomenological Research 56,
pp. 863–8.
192 DORIT BAR-ON
Wright, C. 1998 “Self-Knowledge: The Wittgensteinian Legacy,” in C. Wright, B. C. Smith,

and C. Macdonald (eds.), Knowing Our Own Minds, Oxford, Oxford University Press,
pp. 13–46.
Wright, C. 2001 Rails to Infinity: Essays on Themes from Wittgenstein’s Philosophical Investigations.
Wright, C. 2002 “What Could Anti-Realism about Ordinary Psychology Possibly Be?,” The
Wright, C. 2003 Saving the Differences: Essays on Themes from Truth and Objectivity, Cambridge,
SECTION III
Truth, Objectivity,
and Relativism
7
Some Remarks about Minimalism
Simon Blackburn
I
In this essay I want to develop some thoughts about the difficult issue of deflationism in
the theory of truth.1 I offer the essay to a collection dedicated to Crispin Wright with
some trepidation. For his own work on truth, particularly in the wonderful Truth and
Objectivity, has powerfully developed a subtle kind of position, which he indeed calls
minimalism, but which may well escape the doubts and difficulties I am going to
explore. I shall say something about that at the end of this essay.
For all the work that has been done on this subject, particularly in the last decade or
so, I find it one on which it is possible to be in two minds. Indeed I should confess at
once it is a subject on which I fear I am in two minds. I shall introduce this by talking
about a position of my own, but then generalize the problem. On the one hand, a
position I hold dear, namely expressivism as a theory of normativity, can profit greatly
from deflationism. For deflationism enables the expressivist to fend off a familiar kind of
complaint, which is that he rides roughshod over our everyday confidence that there
are moral truths. For the deflationism enables us to say that given that we have a
syntactically appropriate sentence with which to voice our attitudes, we can add ‘is
true’ or ‘it is true that’ with no further cost. There is a story to be written, in this view,
about the ethical proposition, and how it holds its place as a focus for discussion and
thought. But there is no last chapter to be written about ‘what makes such a proposi-
tion true’. There is nothing occult or Platonic or mysterious waiting to puzzle us,
but also no need to struggle with implausible reductions in order to find ‘naturalistic’
truth-conditions. If a David Armstrong or a David Lewis comes along demanding a
‘truth-maker’ we can profit from deflationism, and simply say that what makes it true
that honesty is good is that honesty is good. Nothing else needs to be said, wearing
allegedly metaphysical hats, or allegedly scientific hats. Of course, as first-order mor-
alists rather than theorists of the nature of morality—painters rather than anatomists in
1
To ward off misunderstanding, I shall use ‘deflationism’ and ‘minimalism’ more or less interchangeably,
except when talking explicitly about Wright’s own version of minimalism, which is rather different.
196 S IM O N B L AC K BU R N
Hume’s terms—we can reflect on what makes honesty good, and sketch whatever
connections impress us between honesty and other things such as respect, integrity, or
happiness. But that is a different discussion.
Again, expressivism can make use of deflationism in warding off that other hoary old
war-horse, the charge of relativism. For relativism really needs some kind of transition
from ‘they (the others) believe that p’, where p is some moral claim that we ourselves
do not hold, to something along the lines of ‘it is true for them that p’ or ‘p is just as true
as –p’ or ‘–p is no more true than p is’. This transition may, just, seem plausible if we are
gripped by a ‘realist’ picture, and the absence of some solid, metaphysical object-like
fact is troubling us. For then it might seem that –p is no better ‘grounded’, no more
solid or in touch with reality, than p, and this is what the relativist is trying to express.
But once this picture disappears, as it should do given deflationism, everything changes.
‘–p is no more true than p’ for instance becomes at best something like an expression of
agnosticism or indifference, and in turn this is a moral attitude that there is no reason to
adopt, just because some hold one attitude and others hold the contrary; it would
amount to distributing equal sympathy to each side, and there is not always any reason
to do that.
So far, then, deflationism seems a Good Thing.
On the other hand, lurking in the background there is a big bad bug. This is the
familiar thought that expressivism needs some kind of contrast between commitments
which are suitable and those which are not suitable for its distinctive treatment.
Normativity is supposedly suitable, and a variety of familiar reasons suggest why. But
there are many other candidates. Suppose we say, for instance, that conditionals do not
have truth-makers, but express commitments to ratios of probabilities. Or, suppose we
say that probability judgments again do not have truth-makers, but express idealized
betting dispositions. Or, along with Wittgenstein, that ‘my attitude towards him is an
attitude towards a soul. I am not of the opinion that he has a soul’. In each case not only
the plausibility, but the very content of the position seems to require a contrast. It is a
contrast that can be put a number of ways: between belief and attitude, or representa-
tion and the adoption of a stance, or between beliefs with truth-makers from beliefs
that do not need them. Whichever way we put it, it seems to require some kind of
distinction between representative states of mind, and states of mind that, although
couched in propositional form, and accepting truth-functional and other embeddings
just as much as they accept the truth predicate, are nevertheless not to be thought of as
representational.
But is this distinction sound, given deflationism? Any story about it will depend on
some conception of ‘representation’, the function with which the other things are
contrasted. And the worry arises that this conception of representation must be (a)
robust enough at least to generate a space for functions with which it contrasts and (b) if
so, a dagger in the heart of deflationism, since it must be closely allied to, if not identical
with, a similarly robust conception of truth. This is not the weak charge that is
sometimes made that it is, for instance, empiricist prejudice, or scientism, or at any
SOME REMARKS ABOUT MINIMALISM 197
rate something Bad that selects some commitments as bona fide representations, only to
contrast others as having some other function. We need not worry very much about
this vague charge. For the sources of the theory can look after themselves, in our best
theories of our own perceptual abilities, sensitivity to features of things, or our best
science of our surroundings and our responses to them. But the more specific version of
the charge will be that any appropriate theory of representation must itself betray
deflationism. It is itself a theory of truth, or at least a substantial chunk of a real, robust,
story about truth. And it is this specific charge that raises unease, an unease which
Crispin Wright was quick to express, and that I now need to explore.
As is well known Wright’s ‘disciplined syntacticism’ in effect gives a minimal
account of (‘minimalizes’) any notion, such as representation, which could be used
to support the contrast and it thereby closes off any space for expressivism. I have in the
past sheltered behind the authority of Ramsey and Wittgenstein in opposing this
conclusion.2 Ramsey is clearly a deflationist about truth who nevertheless indulges
highly developed functionalist stories about such states of mind as belief in a probability
judgment or belief in a generalization. And in order to defend Wittgenstein from
Wright’s charge of inconsistency I suggested that we split the theory of the proposition
or commitment, from the theory of truth. The former can be as ‘maximalist’ as we
wish, so that propositions or contents are many-sorted: some arise as foci for thought
about betting rates, or ratios of confidences, or as reflections of attitude, or perhaps as
instruments for getting us from data to predictions, as one kind of pragmatism about
scientific theory would suggest. Whereas others are more conventionally ‘representa-
tional’. Their job is to reflect the world, or tell it like it is. But all this, or so I have
argued, can coexist with the idea that the last thing to say about truth is entirely
deflationist.
In fact, this kind of story is not so far from Crispin Wright’s own preferred
approach. Taking the theory of identity as a model, Wright has argued for a
many-sorted but univocal truth predicate. There are different things to say about
truth, as there are about identity, in different applications. But the underlying notion
and the kinds of inference it supports are always the same. Now the difference
between this and a many-sorted approach to propositions is pretty fine. Where
Wright would say that truth in some applications goes along with ‘wide cosmologi-
cal role’ or ‘cognitive command’ whereas in others it does not, the latter would say
that some propositions have wide cosmological role than others, or that failure to
agree on them leads more directly to a charge of cognitive incompetence than it does
in the case of others. An out-and-out deflationist about truth can say that kind of
thing, just as people making the original distinction between primary and secondary
properties did not bring in any focus on truth as such, but only on the kinds of
proposition said to be true.
2
See, for instance, the interchange with Crispin Wright in Mind, 1998.
198 S IM O N B L AC K BU R N
It is indeed not so very clear what is gained or lost if we redirect attention from the
content of what is said, the proposition, to the truth predicate as it applies to the
different propositions. It may sound as if any philosophy that emerges in one story
could be translated without loss into the other. Or it may be that there is an actual
disadvantage in switching attention from the proposition to the truth predicate attach-
ing to it. The move invites a charge of a kind of double counting. Consider, say, the
difference between ‘the sky is blue’ and ‘the block of granite is a cube’. If Locke is right
there are asymmetries of ‘cosmological role’ here, with the shape of the block
something ready to cause all sorts of consequences, whereas the blueness of the sky
causes fewer. Or, we might find an asymmetry of cognitive command, being sympa-
thetic to the idea that a colour-blind being or on with very different colour sensitivities
can still function perfectly well in our world, while one blind to the shape of things
could not.3 But having satisfied ourselves about that asymmetry, why go on to find
another similar asymmetry when we talk of the truth that the sky is blue or the block is
cubic? Why shift the asymmetry back to different realizations of implementations of
the truth predicate, when it already has a perfectly good home in the content of the
propositions themselves?
Wright, as I apprehend him, believes that there is an advantage in applying the
many-sortedness to truth rather than simply to the propositions said to be true, because
he believes that truth is not to be entirely deflated in any case. He takes seriously the
normative function of truth, and the consequent thought that if truth is normative then
it must be something that is normative. Normativity cannot just float in a vacuum, as it
were. It must arise from some kinds of property that truth has. Otherwise, how could it
be important whether we have it, and how could it be a duty to confine ourselves to it,
or the internal goal of belief that it hits upon it?
II
These different bugs may, indeed, be preludes to a general worry about deflationism.
The worry is that what deflationism solves in one chapter, it presupposes in another.
For instance, when a deflationist like Horwich presupposes some notion of ‘proposi-
tion’ or ‘judgment’, then the charge would be that this notion is itself dependent
upon some conception of truth, or of aiming at the truth, or in a nutshell, depends
upon some conception of representation. If, as in Field’s version, deflationism avoids
talking of propositions but talks instead of ‘utterances-as-understood’ then the same
doubt arises. An utterance is understood only by a subject capable of judgment,
and the powers involved in doing that will very likely turn out to be powers of
representation.
3
See especially Jonathan Bennett’s classic article, Bennett 1965.
Peter Strawson mocked a philosophy of truth that presupposes a concept of judg-

ment, in a remark that Keith Simmons and I quote in our introduction to our Oxford
Readings volume on Truth:
And it is indeed very strange that people have so often proceeded by saying “Well, we’re pretty
clear what a statement is, aren’t we? Now let us settle the further question, namely what it is for a
statement to be true.”4
It is important to realize that this remark embraces not only substantive or ‘robust’
theories of truth, but any deflationism that simply turns its back on the further question,
if, unhappily, the material necessary and sufficient for answering it already went into
the story about what a statement or judgment or proposition is. Strawson’s gibe will
not be blunted just by avoiding nouns such as ‘statement’ or ‘proposition’, perhaps
through a misdirected fear of ontological profligacy, or through a misdirected Quinean
fear that their associated identity conditions are too slack to allow us to use them. For
the same worry can be put in terms of verbs. One who says something, speaks truly if
things are as he says they are, and Strawson’s gibe is that it is not much use saying
nothing about ‘truly’ if we have already handed ourselves a full understanding of saying
how things are. And it is here, I think, that contemporary deflationism can seem to be
badly compromised.
To avoid Strawson’s problem, it is clearly tempting to stick to predicating truth of
sentences. But each of Horwich and Field, who are the two deflationists about whom
I shall have most to say, had a good reason for moving from good old-fashioned
sentences, relying on a Tarskian version of the T-schema as the schema whose
instances together give a recursive definition, or an infinitary definition of truth,
albeit for one language at a time. The defect of deflationism cast that way is that the
individual instances of the T-schema seem to be contingent, and furthermore they
give us contingencies of which additional knowledge does not augment anyone’s
grasp of the truth predicate. 5 It is not just, as is sometimes carelessly said, that Tarski
‘relativizes truth to a language’, for that can sound fairly harmless. It sounds harmless
when it is taken to mean no more than that the truth of the overall biconditional
depends upon the language of which the sentence disquoted is supposed to be a
member. But it is not at all harmless when it implies that speakers of different
languages cannot share a conception of truth. This is far from harmless; in fact it is
fairly obviously false. It is avoided, as I have said, by Horwich who treats propositions
and not sentences, and in a different way by Field, who in effect prevents us from
raising the topic by restricting us to a conception of truth that applies only to utterances-
as-understood.
4
Blackburn and Simmons 1999, p. 21.
5
Of course, one can treat T-sentences as part of what Carnap called ‘pure’ semantics, in which case they
are analytic, a priori, or part of the definition of the abstract language, say ‘English’. This just relocates the
crucial contingencies, which come back when we ask whether any empirically given group speaks Enlish as
defined.
200 S I M O N B L AC K BU R N
But the question we are posing still arises. Isn’t the notion of a proposition, or
equally the notion of a ‘sentence as understood’ itself going to need unpacking in terms
of some notion of representation or truth? Let us consider Field’s response first. He
himself is forthright about taking on the challenge of saying something about meaning
to complement deflationism about truth. In that connection he talks interestingly of
the creeping threat that robust notions of ‘representation’ pose to his kind of deflation-
ism. His own preferred theory of meaning centres upon ‘conceptual role’, but expands
therefrom:
conceptual role isn’t enough: it is both ‘internalist’ and ‘individualist’: and a plausible deflation-
ism is going to have to give to content both ‘externalist’ and ‘social’ aspects.6
Let us concentrate upon the first. Field recognizes that there are ‘external’ correla-
tions between circumstances and dispositions to utter: I am a very accurate indicator of
whether there is rain falling on my head, so I tend to believe that there is rain falling on
my head if, and only if, there is rain falling on my head. And if I tend only to utter the
appropriate sentence when I believe in its truth, then my disposition to utter will itself
correlate with rain falling on my head.
This is simply a correlation, there to be observed; and a deflationist is as free to take note of it as is
anyone else, and as free as anyone else to deem it an ingredient of what he calls content.7
The threat is now apparent. For why aren’t these relations, which Field calls
‘indication relations’, themselves sufficient to give birth—no doubt a birth followed
by a life of refinements and subtleties—to the very notion of a truth-condition that
deflationism exists to deny?
This is in fact how I presented the matter a long time ago in Spreading the Word,
highlighting the way in which we see ourselves as responding to the facts. It is an integral
part of our theory of the world that the cat being in the garden causes many things: ‘the
shadow on the parsley, the absence of the birds, or my belief that it is there’.8 But then
our conception of ourselves as responding to things and facts about things seems a
promising starting point for a theory of representation, and eventually, perhaps, an
inflationary account of truth.
Obviously Field has to resist this reasoning. His principal objection is subtle. He
acknowledges, as we have just seen, that the deflationist can recognize facts about
indication, and can “even use the disquotation schema to formulate the distinction
between one of his belief states indicating its own truth conditions and its indicating
something else . . . ”.9 This distinction is illustrated in turn with two examples. One is a
case of systematic exaggeration. Consider distance. It may be that my saying something
is n metres away correlates with it being f(n) metres away, where fx starts dropping off
rapidly from x after a few metres. Thus, my saying that things are 100 metres away will
6 7
Field 1999, pp. 356–7. Ibid., p. 357.
8 9
Blackburn 1984, p. 246. Field 1999, p. 358.
actually correlate with them being (say) 75 metres away, and so on. The other is an
example of indirect causation, so my beliefs about events in Bosnia may not correlate at
all with events in Bosnia, but only with newspaper reports of events in Bosnia.
Field’s admission that a deflationist can make these distinction is welcome. It would
be a fatal blow to the position if it couldn’t find the resources to distinguish between
the circumstance in which someone’s remark that an object is 100 metres away is true,
and the circumstance in which his saying it is a reliable indicator that the object is 60
metres away. Or, between distinguishing when it is right to say that there has been a
massacre in Bosnia, namely when there has, and the circumstance in which it may be
explicable or justifiable for people to be saying that there has, when there has not,
because they have received misleading reports. ‘But’, says Field:
What [the deflationist] can’t do, it seems to me, is say that this distinction is of much explanatory
importance: for that would give truth conditions (rather than indication relations) a central role
in the theory of mind; and the claim that truth conditions have a central role in the theory of
mind is the defining characteristic of inflationism.10
Field admits that this looks quite implausible, but claims that it can grow upon one
“after one has spent some time trying to say just what its explanatory importance is”.11
But I think this idea should not grow upon us.
The claim, remember, is that the distinction between a belief state reliably indicating
its own truth-condition, and it reliably indicating something else is of little explanatory
importance ‘in the philosophy of mind’. I find this claim quite hard to interpret, and
very hard to accept in any feasible interpretation. Consider, for instance, the systematic
exaggerator. Suppose he is the caddy to a golfer, and constantly overestimates the
distance to the green, telling his player that it is 200 metres when it is in fact 150, 300
when it is 200, and so forth. His mistakes have consequences; when he is trusted his
employer loses the game and he loses his job. Perhaps his mistakes have interesting
mind/brain causes: he is astigmatic or strangely habituated or whatever. Does his trait
play a ‘central role in the theory of mind’? It is very hard to know what this demands.
He could, certainly, go on living in a fool’s paradise in which his inadequacy seldom or
never comes to light, and so behave much as someone who gets distances right
behaves. We can live with a great many false opinions lurking in our heads, but
undetected if circumstances for voicing them or acting upon them are sufficiently
rare. This would be mere chance, and equally likely there are not only social events (his
losing his job) that the quirk rather obviously explains, but more overtly mental events,
such as his belief that he has been unjustly sacked, or his depression, or his envy or
hatred of others more successful. Even in the event that his quirk remains idle, for he is
seldom or never called upon to judge distances, it is still potentially explanatory of all
manner of things.
10 11
Ibid. Ibid.
Turning to the Bosnia example, we get the same structure. It is no doubt true of me
(or was in the 1990s, to which Field’s example belongs) that my beliefs about Bosnia
correlate as well or better with newspaper reports than with actual events on the
ground, and we know why that is so. But suppose I am now put in the position of
reporting on events in Bosnia, am taken there, and made to file eyewitness reports.
Suppose I only file reports that correlate with what (other) newspapers are saying, and
not with events on the ground. This trait may be explicable, for instance if I only hang
out in the bar with the other reporters or embed myself with the same lying military
spokesmen as they do. Is it of little or no explanatory importance? In favorable
circumstances it explains my losing my job, once more. In unfavorable circumstances
it might start a war.In these cases the consequences come about through social
mechanisms, but of course, they do not have to. The propensity to get distances
wrong could easily affect you without your having misled other people: you misjudge
the distance and fail to hit the deer or jump the river or avoid the rain. We need to
avoid any model where belief is, as it were, purely ‘inside the head’, so that it may make
no difference whether my beliefs are generally true. ‘Inside the head’ a belief is a belief,
with the same mental relations whether or not it is true or false. But beliefs are not
inside the head. They are states of active engaged agents, and it very much makes a
difference to these agents whether they are true, or false. A belief, as Scottish philoso-
pher Alexander Bain well said in the nineteenth century, ‘is a preparation for action’,
and while the relation between truth of belief and success in action is not easy to state
accurately, few doubt that it exists, and indeed that it is only through its existence that
we have any chance of seeing our cognitive capacities as adaptations.12 In short we
must not overlook the myriad small successes that constantly attend our actions only
because perceptually and in more advanced ways, we get things right.
Not only are these things so, but we are always aware that they might be so. The
distinction between a remark being true and it being only an indication of something
else is one that is often in our minds. ‘I might have been misled by the newspapers’ is
not a thought that belongs solely to philosophers, and nor is ‘My eyesight might be
letting me down’. In other words, the distinction is explanatory of thoughts about our
own fallibility or security. It is required to explain a range of attitudes we may hold to
ourselves as good judges of the truth. It explains things in the mind.
Of course, and to repeat, often I am unaffected by truth or falsity. I believe that
O. J. Simpson was guilty, but things are unlikely to go better for me if this belief is true,
or worse if it is false. But it is easy to envisage holistic resources to cope with this. It is
also easy to imagine social resources. That is, things go better if we believe that people
are guilty if and only if they are, rather than in some skew subset of cases.
It seems to me therefore that Field has nodded. Explanatory importance cannot
be the issue, for apart from anything else almost any fact can assume almost any
12
I give one way of developing a connection in Blackburn 2005.
explanatory importance, given the right circumstances, and almost any fact can be
imagined to explain very little, again given the right circumstances. But then, if the
difference between resources open to the deflationist and those needed for inflationism
are not categorized in these terms we are left wondering how they are to be categor-
ized, and Field, as I understand him, offers us no further help. The crucial concession, it
seems to me, is that we must adopt a theory with the resources to distinguish between
truth-conditions and mere indication relations. Once that is conceded, the inflationist
cannot be charged with making this distinction ‘more explanatory’ than it is. It seems
that all sides need to join in suggesting what the distinction comes to, but recognizing,
as I have suggested, that reliable indication is likely to be a core ingredient in any
identification of truth-conditions.
III
If we turn to Horwich we find a rather different way of resisting the inflationist
potential in the notion of a proposition or judgment. Again, however, the key issue
concerns explanation.
So consider, as an essential part of any theory of the proposition, the component of
simple reference: ‘rain’ in English refers to rain. Horwich claims that it is permissible
here to talk of a relation, between in this case an English word and rain. However:
The question of whether or not this relation of reference is constituted by some underlying causal
relation—or by some other non-semantic relation—is an entirely separate issue. And part of the
deflationary position, as I see it, is that the reference relation is very unlikely to have any such
underlying nature. For it is plausible that the explanatory basis for all facts regarding reference is a
theory whose axioms are instances of the disquotational or equivalence schemata. This is because,
on the one hand, such axioms appear to suffice to explain all other facts about reference; and on
the other hand, it is not likely that these facts will themselves be explained in terms of something
more fundamental.13
Once more it is not easy to interpret this. In particular it is not easy to see what is
meant by claiming that the explanatory basis for all facts regarding reference is the
described theory. For that theory is in essence just a conjunction of statements of the
very kind in question: ‘rain’ refers in English to rain & ‘snow’ refers in English to snow
& so on. It is, to say the least, stretching things to say that such a conjunction ‘explains’
the truth of its individual conjuncts.
Horwich’s modesty here (in Dummett’s sense) sits uneasily with his pursuit of a ‘use’
theory of meaning. If the last thing to say is that ‘rain’ refers in English to rain, then
there is no need to pursue a story about aspects of the use of the term that explain or
even ‘constitute’ this contingent fact. But in fact Horwich’s pursuit of a use theory is
surely well-motivated, for the modesty suggested by this passage is a false modesty. The
13
Horwich 1998, p. 124.
classic way of showing this is by imagining a slightly variant language, let us call it
Englich, which is like English except that the word ‘rain’ (the syntactically identified
string of letters) does not refer to rain, but let us say, to hail, while the word ‘hail’ refers
to rain. The ‘theory’ of Englich is just as simple as the theory of English. So the
question arises: what makes it true that we speak English, and not Englich? It is no good
here citing ‘axioms’, for one set gives one result and the other gives the other result. It is
no good saying that it is ‘analytic’ or somehow guaranteed by logical principles
governing disquotation that we speak English and not Englich, for that is just false.
We clearly could have spoken Englich (this shows that nothing is gained by defining a
language in terms of syntax and semantics, making it analytic that ‘bears’ refers in
English to bears, for instance. The contingency just shifts its place, requiring us to ask in
virtue of what contingencies it is true to say that we speak English—the so-called
‘actual language’ relation). The obvious gap requires some kind of story to fill it:
something to do with causation, proper function, dispositions, indication relations, and
the like. And then, as with Field, the suspicion arises that a sufficient investigation of
these relations gives us a substantive or robust, non-deflationary account of reference.
Horwich’s fuller position is not in fact as modest as the above quotation suggests. His
general line is not that there is nothing more to be said than that ‘rain’ refers to rain, but
rather that there are different things to be said about different instances of reference. On
this account the vice is abstraction, not immodesty. He acknowledges that for each word
there is a “certain explanatorily basic regularity governing its overall deployment” but
says that basic use regularities ”need have no common form”, in particular, “they need
not relate the words they govern to the members of their extensions”.14 On this
account, immodesty is fine, but one should be immodest in different way on different
occasions. We have a kind of pluralistic immodesty.
Pluralistic immodesty collapses into total modesty only if we recoil from the vice of
abstraction to the opposite extreme of extreme particularism. That is, we not only give
up any attempt to say what reference is in general, but give up any attempt to find a
commonality underlying the different modest things we say. On this line, there is
something to say about ‘rain’—that it refers in English to rain—and ‘hail’—that it refers
in English to hail. But we give up any attempt to see these as instancing a common
pattern. It is as if we see only a subject term (the words) and two entirely distinct
monadic predicates, rather than one relation applicable to two different ordered pairs.
When he compares his overall use theory of meaning to deflationism about truth, it
sounds as though this may be what Horwich intends. For a defining characteristic of
deflationism about truth is that there is indeed no common pattern discernible in the
different instances of the T-schema. As I put it in Spreading the Word, ‘penguins waddle’
is true iff penguins waddle, and ‘snow is white’ is true iff snow is white, but a defining
characteristic of deflationism is that there is ‘nothing much in common’ between these
14
Horwich 1998, p. 113.
two, just as there is nothing much in common between deciding whether penguins
waddle and snow is white.15
However it stands with truth, it is hard to believe in this extreme particularism
when it comes to reference. Surely there is the possibility of some uniform pattern at
least distinguishing families of co-explicable instances of the reference schema? The
awful vice of abstraction, as Berkeley would have it, might lead someone to think
that the very same kind of thing explains the fact that ‘2’ refers to the number 2 and
that ‘rain’ refers to rain. And this may be a mistake. But there may be no mistake in
thinking that the very same kind of thing that explains ‘rain’ referring to rain also
explains ‘hail’ referring to hail and ‘snow’ referring to snow. Are patterns really
impossible? On the face of it, and in spite of Berkeley, they look to be as innocent as
any other kind of generality. We might say something like this: once the learner can
reliably associate the term ‘t’ with the weather condition it is used to indicate, as part
of a norm-governed pattern of communication, she is deemed a fully fledged
member of the linguistic community, meaning t by ‘t’—and that would be a schema
equally applicable to rain, snow, and hail. As in the case of Field and the idea of
building truth-conditions partly upon indication relations, further refinements could
easily follow on.
In spite of this side of his thought, I do not read Horwich as really intending to prove
or even argue that pattern is impossible. The very idea of a use theory suggests
otherwise. Whether Berkeley liked it or not, theories only work by generalization.
IV
Horwich runs together two different concerns when he treats these issues. One is the
Berkeleian hostility to abstraction, or to a ‘general account’ of the structure of ‘x means
F’. The other rather different strand is hostility to the idea that the basic regularities of
use governing words should relate the words they govern to members of their
extensions. Consider, for example:
For the basic use regularities of different words—like different laws of nature—need have no
common form; and they need not relate the words they govern to members of their extension.16
Now it is obviously possible to separate these claims. As I have said, we may admit
the lack of common form without going wholly particularist. We might work in terms
of families of words, admitting different kinds of explanation. Almost any theorist
would admit this over, say, the difference between the form of explanation of use one
might give of logical operators, as opposed to the form one might give of proper names
of persons, for example. We might similarly sort one particular category, say that of
predicates, into different families: many theorists believe in a different explanation
15 16
Blackburn 1984, p. 230. Horwich 1998, p. 113.
of the way natural kind terms work from that in which other terms work, for instance.
But what of relationship to members of the extension?
Here it seems to me Horwich may be right, and certainly for the purpose of this essay
I have no quarrel with his claim. But notice how weak the claim is. It is only that the
fundamental acceptance properties governing the use of a term need not relate the
understanding user to the extension. But it clearly can do so, for certain terms, or certain
families of terms. For example, Horwich’s own example of the basic acceptance
property for a simple colour term does so:
The explanatorily fundamental acceptance property underlying our use of ‘red’ is (roughly) the
disposition to apply ‘red’ to an observed surface when and only when it is clearly red.17
If this simple stimulus (redness) response (disposition to acceptance) account is

fundamental for some terms, a question arises whether some version may apply to
bigger families than Horwich, in his deflationist guise, seems to admit. Indeed, so much
seems implicit in the general treatment of proper names, where a good candidate for
the fundamental acceptance property of a sentence predicating F of a would be the
disposition to accept it when and only when a is clearly F.
It is, I think, a difficult question how far this betrays the advertised deflationism
about reference. Clearly, if we raise the question of how, non-miraculously, someone
might have this disposition, we enter the territory of causal relations, the ability to
recognize features that subjects possess, and mechanisms that enable transmission and
indication relations. Equally, if we consider putative counterexamples, such as know-
ing what ‘Sherlock Holmes’ means, we would need a theory of understanding in the
absence of extensionally certified dispositions of application. All these issues will take us
back into the same inflationist territory that threatened Field.
For my immediate defensive purpose, however, we can put this on one side. For one
thing that becomes visible, given a pluralistic story that includes potentially non-
relational accounts of various terms, is that it opens up space for precisely the kind of
theory exampled by expressivism. It could be, that is, that there is a family of terms, say
moral or ethical terms, whose use requires a different kind of description from that of
others, and that these descriptions do not consist in relating understanding users to
members of their extensions. The expressivist says that this is indeed so: the fundamen-
tal stories about the use of these terms requires concentrating upon their practical
function in directing choice and action. Furthermore, at least for the most general or
‘thin’ ethical terms, this story does not require that the user is in any relationship at all to
features of action or properties of character or whatever it is that actually deserve
describing in those terms. The user can be wrong, almost without limit, about which
kinds of things to value or to permit or to require from others, but moralize away
merrily using the same terms with the same function as the rest of us.
17
Horwich 1998, p. 45.
A distinctive claim of expressivism, however, will be that this kind of account should
not be given of all predicates. There are those, such as ‘red’ in the account above, for
which the fundamental story is quite different. It has no relationship to anything like
choice or motivation, and it does have a clause relating the user to the actual extension
of the term. If ‘representation’ has its home in these cases, then we have the requisite
contrast.
A philosophy that applauds the emerging pluralism, sees it as sufficient to vindicate
everything that the expressivist wants to say, but which also downplays or denies any
residual or contrastive conception of representation is offered in a number of highly
illuminating papers, by Huw Price.18 Price believes that expressivism survives, and
indeed survives best, in a climate of ‘functionalist pluralism’, which charts the multitude
of functions that different kinds of assertion serve, but without at any point locating
those functions by contrast with a quotidian, everyday notion of representation or
truth. Rather, those notions themselves come to be ‘constructed’ or at least understood
in the light of the various conversational and other demands that discourse puts upon
us. I do not have the space here fully to discuss the merits of this approach (nor its
cousin, the pragmatism of Bob Brandom). It is notable in not demanding any mini-
malist ideology (instead it has a pragmatist underpinning) and therefore escapes the
difficulties that, if I am right, threaten Horwich and Field more clearly. The salient
question for Price will be whether the pragmatist can sketch the functions in question
without at some point locating subjects of the discourse in a world whose fundamental
denizens and features they respond to, converse about, and represent. Is it possible that
the creatures of the left, which the pragmatist likes—expression, discourse, norms, and
practices—depend upon the ground on the right, the land of reference, representation,
fact, truth, truth-makers, and ontology? Or do they suffice somehow to construct and
legitimize these otherwise suspicious notions, at least in some cases? Or is there, as
Richard Rorty supposed in his more messianic moments, an inalienable gap between
them? As I see it, as with Horwich and Field, the dilemma will be that either the
suspicious notions are constructible out of the materials the pragmatist does employ, or
they escape his grasp but to that extent leave him unable to give a fully satisfactory
account of ourselves as speakers and hearers in a common world. If the former, then the
old contrast that expressivism wants is constructible within his picture; if the latter then
the old contrastive expressivism may look better after all.
V
All these issues serve to point up the question of when anyone deserves calling a
deflationist. Consider the atomic sentence ‘a is F’. For the deflationist the last word
about truth is, in Horwich’s version, our disposition to accept this instance of the
18
See Price 1988, 1997, 2003.
schema: the proposition that a is F is true if and only if a is F. The opponent insists on a
theory of judgment, and we are now imagining such a thing being provided. Its clauses
are going to include something like:
The basic acceptance property governing ‘a . . . ’ is that the user is disposed to assent to ‘a . . . ’
when and only when a is clearly . . .
The basic acceptance property governing ‘ . . . is F’ is that the user is disposed to assent to ‘ . . . is F’
when and only when . . . clearly is F.
Do these compromise deflationism? The thought that they do must be this: you cannot
be a deflationist about truth and an inflationist about the other semantic relations, since
we know that these are interdefinable. An atomic sentence ‘a is F’ is true if the
reference of ‘a’ falls within the extension of ‘F’. If the clauses of the use theory inflate
these sub-sentential semantic relations, then they also inflate truth.
I think there is a schism here in the deflationist camp. It is at least noteworthy that
early deflationists contrast deflationism about truth with whatever thick or robust stories
might be given about reference. Typically, it was the elusive nature of facts that
bothered them. Things are fine, and so in principle are the relations whereby words
pick out things. Why shouldn’t they be—for after all you can design physical instru-
ments and processes that select or pick out things from other things? So, for instance,
when the later Wittgenstein recants on the Tractatus conception of sentences as pictures
of facts, he explicitly contrasts elusive facts with everyday things and complexes of
things (for example, you can move a complex of things, but you cannot move a fact).
But it is not at all clear that the clauses suggested above do compromise deflationism.
For they don’t give us explicit definitions of semantic relations between sub-sentential
parts and things.
Still, there remain grounds for unease. Imagine a toy language, perhaps used by a
machine. The machine can recognize certain individuals, a, b, c . . . in the sense that
faced with one of them it comes up with the correct label. We know the physical
processes underlying this capacity—suppose it has a gallery of images and a way of
checking optical input against them. Furthermore it has ‘predicates’, and rules of
application: it can sort items in the domain into ones that are F and ones that are
not. Perhaps, for instance, it operates a colour checking procedure and sorts things
by colour. Putting these capacities together, it can issue reports, such as ‘a is F’, or ‘not
(b is F)’. There is nothing mysterious in the capacity, since we understand the
component recognitional part, and the component selection mechanism.
For this simple device we have a ‘theory’ of reference and a ‘theory’ of predication,
in the sense that we understand why it recognizes what it does, and why it classifies as it
does. The operation of the device is wholly transparent. Suppose now it outputs the
sentence ‘a is F’, and someone asks me what that means. I reply that ‘a is F’ is true if and
only if a is F. Do the things we have already said about the ‘fundamental acceptance
property’ or basic ‘use’ of this sentence compromise a deflationist understanding of this
T sentence?
It seems to me that they do. Or, rather, they remove any point from highlighting
deflationism about truth. The deflationist may gain the position that there is no last
chapter waiting to be written about truth. But that is because rather substantial first
chapters needed to be written about the semantics of the components of the sentence.
It seems to me, therefore, that each of Field and Horwich, our most distinguished
contemporary deflationists, are occupying unstable positions. I side with Davidson in
believing that the tasks that they admit suffice to accomplish the tasks they want to deny;
the theory of the proposition cannot be given without a theory of representation, and
this in turn will provide the material for subsentential semantic relations. An across-
the-board deflationism is not possible. Or at least, if it is possible, that is only by adopting
a more substantial ‘modesty’ than anyone in the contemporary debate embraces.
Where do these points leave the debate between myself and Wright, about the
possibility of expressivism? They show limits to the compatibility between expressivism
and deflationism, as we might have expected. They have me conceding that a ‘maximal-
ist’ theory of the proposition will compromise a deflationism about truth, in effect
conceding that Wright was correct in seeing a fissure in the simple position that each
of Ramsey and Wittgenstein seemed to occupy. More generally, they show me advanc-
ing doubts about the tenability of deflationism in general, at least as it is developed by two
of its leading exponents. Wright’s own version of minimalism remains outside the target
area, being not as deflationist as all that. I do not find it easy to judge the remaining issue
of whether to point many-sortedness at propositions or at the kinds of truth with which
different families of propositions consort, but in the absence of further thought about the
implications of a decision, I find I can live with the issue unresolved.
My main point is now concluded. But as a consequence I should say that the
particular threat I started with, that deflationism is both needed by an expressivist
theory of normativity, but also denied by it, can now be deflected. The denial part
shows the expressivist embracing a particular theory of the functioning of ethical
predicates, and one that describes their use in terms that contrast them with other
predicates.19 The acceptance part shows that enough can be said about their use to see
them fitting without problem into the standard T-schema, and understanding the fact
that they do is all that is required for an understanding of normative truth.
Bibliography
Bennett, J. 1965 “Substance, Reality, and Primary Qualities”, American Philosophical Quarterly 2,
pp. 1–17.
Blackburn, S. 1984 Spreading the Word: Groundings in the Philosophy of Language, Oxford, Oxford
University Press.
19
I should say that throughout this essay I have talked as if there is a particular family of ethical predicates.
This is a simplification, although not one that is here significant. In reality the function of evaluation can be
communicated in many different ways, and without using a special vocabulary. It can be carried by ordinary
predicates in particular uses, or used with particular intonations.
210 SIMON BLACKBURN
Blackburn, S. 2005 “Success Semantics”, in H. Lillehammer and H. Mellor (eds.) Ramsey’s

Legacy, Cambridge, Cambridge University Press, pp. 22–36.
Blackburn, S. and Simmons, K. (eds.) 1999 Truth, Oxford, Oxford University Press.
Field, H. 1999 “Deflationist Views of Meaning and Content”, reprinted in S. Blackburn and
K. Simmons (eds), Truth, Oxford, Oxford University Press, pp. 351–91.
Price, H. 1988 Facts and the Function of Truth, Oxdord, Blackwell.
Price, H. 1997 “Naturalism and the Fate of the M-worlds”, Proceedings of the Aristotelian Society
Supp. Vol. LXXI, pp. 247–67.
Price, H. 2003 “Truth as Convenient Friction”, Journal of Philosophy 100, pp. 167–90.
8
Objectivity, Explanation, and
Cognitive Shortfall
Stewart Shapiro
People are inclined to think there is a world of facts as opposed to a world of words
which describe these facts. I am not too happy about that.
What rebels in us . . . is the feeling that the fact is there objectively no matter in
which way we render it. I perceive something that exists and put it into words.
From this, it seems to follow that something exists independent of, and prior to
language; language merely serves the end of communication. What we are liable to
overlook here is the way we see a fact—i.e., what we emphasize and what we
disregard—is our work.
Waismann (1968, pp. 137, 140)
1. Objectivity
Since antiquity, philosophers and scientists have been trying to say something about
how the world is, in itself, independent of the perceptions, minds, conventions,
representation schemes, etc., of the human inquirer. Through science, we seem to
have made progress on this. For example, we know a lot—or seem to know a lot—
about how the perception of color, shape, and solidity works, and, in particular, about
which features of physical reality give rise to various perceptions. But is it possible to
obtain a completely objective picture of the world? Clearly, our best theories and
explanations are due to both the nature of the non-human world and the nature of
human knowers. As John Burgess and Gideon Rosen put it,1 “our theories of life and
matter and number are to a significant degree shaped by our character, and in particular
by our history and our society and our culture.” This is not to say that the world itself is
somehow shaped by “our character.” It is just our theories of the world that are so
shaped. To what extent does this fairly obvious observation undermine objectivity?
1
Burgess and Rosen 1997, p. 240.
212 S T E WA RT S H A P I RO
There are a number of different and mutually competing traditions that have it that,
in one way or another, there just is no way to sharply separate the “human” and the
“world” contributions to our theorizing. Many versions of idealism give the world
itself a human character, holding that the world itself is shaped by our judgments,
observations, etc. Putting that aside, we have Kant’s doctrine that the ding an sich is
inaccessible to human inquiry. We approach the world through our own categories,
concepts, and intuitions. In a deep sense, we cannot get beyond those, to the world as it
is, independently of said categories, concepts, and intuitions. Among contemporary
philosophers, a widely held view, championed by W. V. O. Quine, Hilary Putnam,
Donald Davidson, and Burgess and Rosen, is that there is, to use a crude phrase, no
God’s eye view to be had, no perspective from which we can compare our theories of
the world to the world itself, to figure out which are the “human” parts of our best
theories and which are the world parts. From this perspective, there simply is no sense
of completely describing the world independently of the perspectives, theories, beliefs,
form of life, or whatever, of the theorist.
This conclusion is of-a-piece with the celebrated attack on the analytic–synthetic
distinction. Quineans agree with the patent truism that the truth or falsehood of any
given sentence is due to both what the words in the sentence mean and the way the
world is. They argue, however, that there is no clean way to separate those factors from
each other. This at least strongly suggests that it is impossible to completely separate out
the part of our true representations that is due to the way the world is and the part that
we, the human describers, contribute, in forming and deploying our concepts. It is a
patent truism that if we try to say something about the world, we have to use language.
And this language invariably contains mind-dependent features. If nothing else, the
meaning of the words is, at least in part, a human contribution. To paraphrase Friedrich
Waismann, language was developed to serve human interests. The way we describe the
world, and the way we see the world, is “our work.”
It is still a jump to go from these observations (if that is what they are) to the lack of
objectivity, and even to the lack of complete objectivity, in our theorizing. Moreover,
to say the least, the attack on the analytic–synthetic distinction has not won universal
approval in the world of philosophy. But I think it can be agreed that there is, at least,
an interesting and important problem of determining how much objectivity is possible.
Waismann asks, “What, then, is the objective reality supposed to be described by
language?” He points out that language “contributes to the formation and participates in the
constitution of a fact; which, of course, does not mean that it produces the fact.”2
In discussing the issue of metaphysical vagueness, Mark Sainsbury3 sums up the
situation nicely. According to Sainsbury, the very question of whether our language
is vague because the world is vague, or whether vagueness is a purely linguistic
2
Waismann 1968, pp. 140–1. 3
Sainsbury 1994, }7.
O B J E C T I V I T Y, E X P L A N AT I O N , A N D C O G N I T I V E S H O RT F A L L 213
phenomenon, due solely to the kinds of languages we have evolved and perhaps the
kinds of beings we are:
suggests that we have a choice of picture. According to one, our world, before we find it, is an
undifferentiated sludge. In finding it, we divide it up. If we divide with vague tools, we see the world
as containing vague objects . . . but the explanation will lie with us and our tools and not with the
world. According to the alternate picture, the world is a certain way before we find it. Our job is to
fashion concepts to mirror it. Because it contains vague objects, we find vague objects, and fashion
vague concept[s] to match . . . I doubt whether either picture is intelligible . . . We cannot think of
our world except through our concepts.
One who goes down the Kantian/Waismannian/Quinean road may end up as an

idealist, holding that there just is no mind-independent world. For what it is worth,
I would resist that. I see no reason to challenge the naive common sense view that the
world “is a certain way before we find it,” and that our finding it a certain way does not
make it that way. Before conscious, sentient humans came to the universe, it contained
electrons, stars, black holes, and all kinds of plants and animals. And before I came to
the universe, it contained cities, parks, and my ancestors.
Ordinary, everyday discourse employs a notion of objectivity which at least seems to
be coherent and useful. Intuitively, there is a difference in kind between utterances like
“Cows have a split hoof and chew their cud” and “Water contains hydrogen” and
utterances like “Small, rectangular eyeglasses are cool” and “Broccoli is disgusting,” or
“Broccoli! Yuch!” The distinction plays a central role in both everyday and intellectual
discourse, and I presume that it is not an illusion. Well, what is this notion of
objectivity, and how is it properly deployed? How can it be deployed, if something
in the neighborhood of the Kantian/Waismannian/Quinean perspective is correct?
Along similar lines, there is at least a prima facie difference between sports like soccer
and racing, which at least seem to have objective criteria of success and failure, and sports
like gymnastics and figure-skating which rely more directly on the opinions of judges. To
be sure, even with electronic detection, races still require human judges, to decide close
cases, or to interpret the electronic data. Moreover, even in gymnastics and figure skating,
the judges have to be trained, and presumably success in those sports is not completely
subjective—whatever that means. Still, there is a nagging intuition that racing judges and
soccer referees are usually trying to adjudicate something objective—who crossed the line
first, whether the entire ball crossed the entire goal line, whether a pass is offside, etc.4 The
task at hand here is to try to say something about what that means.
Crispin Wright’s Truth and Objectivity is the most comprehensive account of objec-
tivity that I know of. Through this and subsequent work, Wright provides a wealth of
detailed insight into the underlying concepts and, in particular, into the notion of
4
Of course, soccer referees also have to decide things like whether a player handles the ball deliberately
and whether a foul is intentional. Consideration of these cases would not take us too far afield, but the present
focus lies elsewhere.
independence from human judgment, convention, and the like. According to Wright,
objectivity is not a univocal concept. There are different notions or axes of objectivity,
and a given chunk of discourse can exhibit some of these and not others. The purpose
of this essay is to explore aspects of two of these axes. In both cases, we quickly run up
against the Kant/Waismann/Quine perspective, and we have to explore the extent to
which we can sort out the aspects of our best theories, explanations, and, indeed,
statements, that are tied to human interests, perspectives, and judgments, and the parts
that are tied to the way the world is, independent of said interests, perspectives, and
judgments. We also encounter delicate issues in epistemology and the philosophy of
science along the way. The terrain is most interesting.
According to Wright, the first hurdle for a non-objective account of any discourse is a
principle of epistemic constraint (EC). Relatively early in the book, it is schematized thus:
If P is true, then evidence is available that it is so.5

A bit later, EC is reformulated as a biconditional:
P $ P-may-be-known.6
I presume that the notion of “evidence” in the first version means something like
“conclusive evidence, sufficient for knowledge.” Otherwise, EC is less interesting for
our purposes. Possible differences between the two formulations involve intricate issues
in epistemology, some of the very same issues that we will encounter below. We will
go with the second version.
The principle of epistemic constraint underlies Michael Dummett’s semantic anti-
realism,7 a thesis that all truths are knowable. Indeed, for Dummett, truth itself has an
epistemic component. Wright is officially neutral on this global thesis of epistemic
constraint. He leaves open the possibility that some chunks of discourse are epistemi-
cally constrained and some are not. Prima facie, humor and moral discourse are both
epistemically constrained. Is it even conceivable that there could be a fact about what is
funny, or what is morally unjust, that no one can know? On the other hand, basic
science is perhaps epistemically unconstrained, if anything is. There very well might be
some facts about the physical universe that cannot be known, statements for which
there is no conclusive evidence. Perhaps facts outside our light cone qualify.
It seems to follow from the very meaning of the word “objective”—independence
from human opinion and judgment—that if epistemic constraint fails for a given area of
discourse—if there are true propositions in that area whose truth cannot become
known—then that discourse can only have a realist, objective interpretation. Wright
puts it well:
To conceive that our understanding of statements in a certain discourse is fixed . . . by assigning
them conditions of potentially evidence-transcendent truth is to grant that, if the world
5 6
Wright 1992, p. 41. Ibid., p. 75.
7
See, for example, Dummett 1973 and Tennant 1997.
co-operates, the truth or falsity of any such statement may be settled beyond our ken. So . . . we
are forced to recognise a distinction between the kind of state of affairs which makes such a
statement acceptable, in light of whatever standards inform our practice of the discourse to which
it belongs, and what makes it actually true. The truth of such a statement is bestowed on it
independently of any standard we do or can apply . . . Realism in Dummett’s sense is thus one
way of laying the essential groundwork for the idea that our thought aspires to reflect a reality
whose character is entirely independent of us and our cognitive operations.8
Wright argues, however, that the converse can fail: epistemic constraint is not sufficient
for the failure of objectivity. Even if a given discourse is epistemically constrained, it is
consistent to
retain the idea that [the] discourse is representational, and answers to states of affairs which, on at
least some proper understandings of the term, are independent of us. For example, in shifting to a
broadly intuitionistic conception of, say, number theory, we do not immediately foreclose on
the idea that the series of natural numbers constitutes a real object of mathematical investigation,
which it is harmless and correct to think of the theoretician as exploring.9
When epistemic constraint does hold, then Wright’s other criteria—width of cosmo-
logical role, cognitive command, and the Euthyphro contrast—come into play. Here
lies the multi-dimensional notion of objectivity. This essay has nothing directly to say
about the Euthyphro contrast, and the concomitant notion of judgment dependence.
Present concern is with cosmological role and cognitive command.
2. Explanation
Wright10 defines a chunk of discourse to have “wide cosmological role . . . just in case
mention of the states of affairs of which it consists can feature in at least some kinds of
explanation of contingencies which are not of that sort—explanations whose possibili-
ty is not guaranteed merely by the minimal truth aptitude of the associated dis-
course.”11 The idea is that a discourse is apt for a realist construal, on this axis, if
statements in that discourse occur essentially in explanations within a range of other
discourses.
To illustrate the criterion, Wright points out that the wetness of some rocks can
explain “my perceiving, and hence believing, that the rocks are wet,” “a small . . .
child’s interest in his hands after he has touched the rocks,” “my slipping and falling,”
and “the abundance of lichen growing on them.”12 Statements about rock wetness
thus occur in explanations concerning perception, belief, the interests of children, the
human abilities to negotiate terrain, and lichen growth. Discourse about rock wetness
thus has wide cosmological role and, of course, it is most natural to regard that
8 9 10
Wright 1992, p. 4. Ibid., p. 5. Ibid., p. 196.
11 12
Ibid., p. 198. Ibid., p. 197.
particular discourse as objective. Wet rocks are not wet because we perceive them or
judge them to be wet. It is the other way around.
Wright quickly adds that the explanations in question do not have to be causal.
Otherwise, only science-like discourses that traffic in causality would have a chance of
passing the test. Mathematics figures in explanations of all sorts of phenomena
throughout the sciences and everyday discourse:
it is notable . . . that the citation of mathematical facts does contribute, seemingly, to other kinds
of explanation . . . (It is because a prime number of tiles have been delivered, for instance, that a
contractor has trouble using them to cover, without remainder, a rectangular bathroom floor,
even if he has never heard of prime numbers and never thought about how the area of a rectangle
is determined.)13
According to Wright, mathematics enjoys wide cosmological role. What of moral

discourse? One might think that the moral corruption of a given person helps explain
why she has no friends, or why no one will give her a loan. Wright argues that in such
cases, the proper explanation does not invoke moral facts, as such. The proper
explanation for why this person lacks friendship and credit is not that she is wicked,
but that people believe her to be wicked. Clearly, the person would be shunned just as
much if she weren’t actually wicked, but was nevertheless believed to be. And she
would not be shunned if she were wicked but was not believed to be so.
Examples like this one lead to Wright’s nuanced articulation of the criterion. The
width of cosmological role is “measured by the extent to which citing the kinds of states
of affairs with which [the discourse] deals is potentially contributive to the explanation
of things other than, or other than via, our being in attitudinal states which take such
states of affairs as object.”14 He argues that moral discourse fails the criterion.15
Wright’s favorite example of a discourse that fails to be objective is humor. And, of
course, common sense, or intuition, supports this. Surely, there is no sense in which a
story can be funny objectively, independent of the attitudes, judgments, forms of life of
humans, or whatever. What would it mean to say that? Discourse about humor is thus
used as a test case to help articulate the various axes of objectivity.
It seems to me that humor does enjoy some cosmological role. Some time ago,
I went to hear a stand-up comedian, and laughed so hard for so long that I got a
stomach ache. I was sore for several hours afterward, so much so that I regretted going
to the performance. On another occasion, during a solemn event, I was ruminating
about a joke I heard, and I could not help laughing out loud. So how does one explain
my soreness and my rude and embarrassing behavior—and the subsequent blush?
13
Wright 1992, pp. 198–9.
14
Ibid., p. 196.
15
This case, and Wright’s conclusion here, is controversial. There is an ongoing debate over whether
moral discourse figures in legitimate explanations of non-moral facts, beyond attitudes and the like. These
discussions are often taken to bear on the objectivity of moral discourse, as Wright’s criterion would predict.
Prima facie, one explains both events by citing facts about humor. I was sore because the
comedian was funny, and I acted as I did, and blushed, because the joke was funny.
This example is different from the one above that invokes morality, since the
explanation here does not turn on beliefs, or at least not on beliefs about what is and
what is not funny. Suppose that, for some reason, I formulated a belief that the
comedian (or the joke) was funny but he (or it) wasn’t—so that my belief would be
false. Surely, I would not have laughed so hard, or perhaps not at all, just because
I believed that the comedian or joke was funny. Or suppose that I formed no beliefs
about them at all. I suspect that I would have laughed just as hard and blushed the same.
Admittedly, the explanation in these cases does seem to involve some sort of
“attitudinal state.” Arguably, humor just is an attitudinal state, or closely relates to
one or several such states. I suppose that the explanation is that I found the comedian
and the joke funny (and perhaps not that they were funny). But is this last an attitudinal
state “which take such states of affairs as object”? Is it a propositional attitude? If so,
then what are the relevant propositions? Again, as I recall, I was not even thinking
about whether the comedian or joke was funny. I was just laughing. Frankly, I am not
sure about what it takes for an explanation to turn on “attitudinal states which take
such states of affairs as object,” but perhaps that is just my own ignorance.
To be sure, even if this example is conceded, it does not follow that humor has a
particularly wide cosmological role. Its explanatory role may be very limited.
Wright argues that cosmological role underlies, or should underlie, accounts of
objectivity in terms of abduction, or inference to the best explanation. It does not
matter much, he argues, whether the proposed explanations are “best,” only that they
are successful explanations at all. Clearly, the plausibility of cosmological role as a
criterion of objectivity depends on the nature of explanation. We thus come to some
deep and troubled waters. What is it to explain something? What is it for a proposed
explanation to be successful—to be an explanation? This, of course, is one of the most
fundamental issues in the philosophy of science and in epistemology generally, and
there is nothing close to a consensus in the extensive literature.
Intuitively, to explain something is to give a reason for it, to answer a “why
question.” Although dictionaries are often poor sources of philosophical insight, we
might note that according to Webster’s Twentieth Century Unabridged Dictionary, to
explain is to “make plain, clear, or intelligible; to clear of obscurity.” Entry 3b of the
Oxford English Dictionary is almost identical: to explain is “to make plain or intelligible,
to clear of obscurity or difficulty.” Of course, it will not do to leave things at this basic
or pre-theoretic level. We need an account—an explanation—of the relationship
between explanation and objectivity, if there is one.
On this everyday, intuitive conception, explanation is context sensitive. What
makes for an explanation—what makes for clarity, intelligibility, lack of obscurity,
removal of difficulty—depends on the interests, goals, and background assumptions of
the person asking for the explanation, the person who wants to know why. What will
make the matter intelligible for her? What, exactly, is the “difficulty” that needs to be
cleared up in the case at hand? When asked for an explanation of a devastating fire in a
building, a chemist might be satisfied to learn that it was caused by the combustion of
exposed gasoline. This particular fact does not clear the matter up sufficiently for the
police. Or consider a more scientific example. Suppose that someone notices an
interesting phenomenon in nature, say that an interesting color spectrum is reflected
under certain circumstances. The physicist attributes the matter to quantum effects of a
certain substance. She produces a differential equation, and shows how the spectrum
fits its solutions. She is rightly satisfied with this explanation. But this would not satisfy a
layperson, who cannot follow the details of the mathematical argument, or does not
understand the scientific theory underlying it. For our layperson, the differential
equation does not make the non-mathematical phenomenon “plain, clear, or intelli-
gible,” nor does it make the phenomenon “clear of obscurity,” as Webster and Oxford
both put it. The reason is that the proposed “explanation” is itself too obscure—too
unplain, unclear, and unintelligible—for him to understand it. I presume that one
cannot claim that a sentence makes something intelligible if she cannot understand the
sentence, or if the sentence itself is pretty obscure.
We can extend the example. Suppose that there were a potential branch of
mathematical science that would account for a huge variety of phenomena, in a
wide variety of areas. But suppose that the theory is enormously complicated. The
shortest statement of any of its basic principles, in any language a human could
understand, would require more symbols than there are molecules in the solar system.
For all we know, the universe itself just might be that complex, and there just is no
humanly understandable theory that accounts for it. Intuitively, the enormously
complex theory would be objective, since it gives an accurate description of a
presumably mind-independent world. Arguably, however, the theory would fail the
criterion of cosmological role, as the latter has been formulated so far. The grand theory
would decidedly fail to make anything “plain or intelligible” nor would it clear
anything of “obscurity or difficulty.” I presume that a sentence cannot make anything
intelligible if no one can understand the sentence. The phenomenon to be explained
would be just as obscure as before. Any “difficulties” we had with it would remain.
Cases like this question the role of cosmological role as a criterion of objectivity, so long
as we stay with the everyday, dictionary conception of explanation. What do our interests,
and what makes for human intelligibility, have to do with how the world is, in itself,
independent of human activities, judgments, and the like? Mathematics often finds
wonderfully concise ways to describe phenomena, some of which are too complex to
be described any other way. A differential equation can code or express a wealth of
physical detail. It is hard to see how the fact—if it is a fact—that human beings sometimes
find mathematical explanations clear and intelligible should count in favor of the objec-
tivity of mathematics. What does it matter what we find plain, clear, or intelligible?
I thought that objectivity is related to what is independent of human interests and goals.
It is only a contingent fact about us that we have the abilities and the limitations we
do. Any techniques we use to understand things within our limitations need not shed
light on how the world is independently of those abilities and limitations. I presume
that Wright never intended to invoke this ordinary, interest relative notion of expla-
nation, right out of the dictionaries. One obvious response to this observation is to
reformulate the criterion of cosmological role. Perhaps we should replace explanations
for flesh and blood human beings with explanations-in-principle. In effect, we invoke
idealizations similar to those invoked in mathematical theories of computability,
deducibility, and the like. We envision beings who are like us, but do not have the
limitations of life-span, attention span, memory, and available material.
So the question underlying cosmological role is what our idealized counterparts
would find “plain, clear, or intelligible” and “clear of obscurity.” I am not sure how
helpful, or even coherent, this maneuver is. I presume that our interests are tied to the
limited kinds of beings were are. What are the interests of our highly idealized counter-
parts? Would they care about the same things we do? What would they find “diffi-
cult,” or “obscure,” or otherwise in need of explanation? Perhaps the move should be
to envision beings who have the same interests that actual, flesh and blood, humans
have, but do not have their limitations—if this make sense.
On the philosophical front, none of these maneuvers help much. To repeat, if
explanation is in fact tied to interests, whether they be ours or those of our idealized
counterparts, then we can and should wonder what it has to do with objectivity. One
option, I suppose, is to argue, or merely assert, that a proposition or an area of discourse
is objective—independent of human (or idealized human) judgment and the like—just
because it figures in what we human beings (or our idealized counterparts) find clear
and intelligible, and free from obscurity. This would be a pre-established harmony,
suggesting that our minds are tuned into the ultimate fabric of the universe. If we (or
our idealized counterparts) get it—if we (or they) find certain things intelligible—then
the universe must be so, independent of us and our judgments and interests. I do not
know how to argue for or against this, but it strikes me as hubris.
For a different tactic, which perhaps reduces the hubris, one can leave the notion of
explanation at an intuitive, unidealized, human level, and concede its psychological
underpinning: to explain is to make clear, plain, or intelligible, just as Webster and
Oxford contend. Nevertheless, the tactic goes, an explanation is not satisfying, or should
not be, unless there is some objective relationship between the explanation and what it
explains. A good explanation needs to engage the non-human world at some level. On
such a view, the original formulation of the width of cosmological role is too crude. It
is not the case that everything that occurs in an explanation is relevant to objectivity,
perhaps for reasons given above.
To illustrate the distinction, one thesis of Steven Yablo’s fictionalism16 is that
mathematics plays only a representational role in science, and, thus, in scientific
explanations. The idea is that mathematics allows us to state or express features of the
16
Yablo 2001, 2006.
non-mathematical world—features that are difficult or perhaps impossible to express

otherwise, given the sorts of being we are. If Yablo is correct, then perhaps mathemat-
ics fails the test for objectivity. Arguably, what is merely representational in this sense
need not be objective, even if it figures in explanations. But perhaps other aspects of
explanation are explanatorily relevant, and perhaps bear on objectivity.
But what are those other aspects? A philosopher who is inclined this way has the
burden of articulating which parts of an explanation are explanatorily relevant, in the
proper sense, and which are explanatorily superfluous, playing only a representational
role. For this account to be helpful here, one must do this articulation without
presupposing that we already have a clear account of what objectivity is.
Perhaps a better route toward giving cosmological role a decent chance of being an
axis of objectivity would be to move away from the above intuitive accounts of
explanation. We need a better, or at least more relevant, account of what it is to
explain. For what it is worth, some of the other entries in the Oxford English Dictionary
seem to point in this direction. According to entry 3a, to explain is “to unfold
(a matter),” “to give details of, enter into details respecting,” and entry 5 is “to make
clear the cause, origin, or reason of; to account for.” This, of course, is fine for a
lexicographer, but the philosopher needs more than these hints.
There are several longstanding philosophical traditions that allow for an objective
concept of explanation, distinct from (but perhaps related to) the everyday interest-
relative, psychological notion above. The Aristotelian notion of final cause seems
related to, perhaps identical with, what may be called objective explanation. The
final cause of a substance is its underlying reason or purpose. Perhaps the same goes
for events or what it is that we ask for explanations of. If so, we can paraphrase Wright
thus: a discourse has a wide cosmological role to the extent that mention of the states of
affairs of which it consists feature in final causes of a wide range of contingencies which
are not of that sort.
Of course, this resolution of our problem requires one to accept an Aristotelian
metaphysics. We have to think of substances, events, or whatever, as having purposes
objectively—independent of human intentions and desires, and, indeed, independent of
human designs and purposes. A western theist might invoke the designs and purposes
of the Deity, and thus solve our problem quickly. Presumably, God’s designs and
purposes are indeed independent of, or at least prior to, human interests, forms of life,
and the like. But theology aside, there is some work to do. This is not the place, and
I am not the philosopher, to further articulate the metaphysics, let alone adapt it for
present purposes.
Bernard Bolzano’s ground–consequence relation gives us another avenue to explore
(if, in fact, the route is different).17 In light of common folk scientific theories, one can
infer the temperature in a room from the height of mercury in a thermometer placed
17
Bolzano 1837.
there, and one can infer the height of the mercury in the thermometer from
the temperature in the room. The inference relation is thus symmetric, modulo the
background theory. In contrast, the ground–consequence relation is asymmetric.
According to Bolzano, the temperature is the ground and the mercury-height the
consequence, and not vice versa. The temperature is the reason for the height of the
mercury; the height of the mercury is not the reason for the temperature. All this
sounds right, at least for this example. If the ground–consequence notion is coherent,
and serviceable, then perhaps we can say that a discourse has a wide cosmological role
to the extent that mention of the states of affairs of which it consists feature in the
Bolzano-grounds of contingencies which are not of that sort.
Bolzano’s notion of “ground” is similar, perhaps identical, to Gottlob Frege’s notion
of the “ultimate ground,” or “justification” of a true and knowable proposition.18 It
has its roots in rationalism. Frege writes:
The aim of proof is, in fact, not merely to place the truth of a proposition beyond all doubt, but
also to afford us insight into the dependence of truths upon one another. After we have
convinced ourselves that a boulder is immovable, by trying unsuccessfully to move it, there
remains the further question what is it that supports it so securely?19
The notion of ultimate ground figures heavily in Frege’s accounts of analyticity and
a priority:20
Now these distinctions between a priori and a posteriori, synthetic and analytic, concern, as I see
it, not the content of the judgment but the justification for making the judgment. Where there is
no such justification, the possibility of drawing the distinction vanishes . . . When a proposition is
called a posteriori or analytic in my sense, this is not a judgment about the conditions,
psychological, physiological and physical, which have made it possible to form the content of
the proposition in our consciousness; nor is it a judgment about the way in which some other
man has come, perhaps erroneously, to believe it true; rather, it is a judgment about the ultimate
ground upon which rests the justification for holding it to be true.
This means that the question is removed from the sphere of psychology, and assigned, if the
truth concerned is a mathematical one, to the sphere of mathematics. The problem becomes, in
fact, that of finding the proof of the proposition. If, in carrying out this process, we come only on
general logical laws and on definitions, then the truth is an analytic one.21
Frege’s is quite explicit here that the notion of “ground” is an objective relation,
having little to do with human psychology. It is clearly not a matter of what we find
compelling, but of how the truths themselves are structured. Frege’s unrelenting assault
against psychologism plays a key role here.
18
Frege 1884. 19
Ibid., }2.
20
Frege adds a note that he “does not mean to assign a new sense” to terms like “analytic,” but “only to
state accurately what earlier writers, Kant in particular, have meant by them.” This exegetical question is
pursued extensively in Burge 2000.
21
Frege 1884, } 3.
Despite the use of epistemological terms like “proof ” and “justification,” Frege’s
relation of grounding or dependence is as much metaphysical as it is epistemic.22 As
Frege himself notes, the relation has nothing to do with how people in fact come to
believe propositions. It is also not a question of whether we know, for example, that
7 + 5 = 12. Skepticism and fictionalism aside, there is no serious question that we do
know that. Nor is it a question of how we happen to know that 7 + 5 = 12. That
proposition was known long before the foundational work began. Moreover, for most
of us, this knowledge need not, and in fact did not go via the proposed founding
definitions. Frege’s dependency relationship thus seems to require a distinction be-
tween the state of knowing, or the state of being justified, and the ultimate or objective
ground or justification of a proposition.
Frege writes: “we are concerned here not with the way in which [the laws of
number] are discovered but with the kind of ground on which their proof rests; or in
Leibniz’s words, ‘the question here is not one of the history of our discoveries, which is
different in different men, but of the connection and natural order of truths, which is
always the same’.”23 As indicated by the reference to Leibniz, Frege here invokes a
longstanding rationalist principle that truths are objectively ordered. The proposal
under discussion is to invoke the ordering to articulate (or correct) Wright’s criterion.
The thesis would thus read that a discourse has a wide cosmological role to the extent
that mention of the states of affairs of which it consists features in the “proof ” or
ultimate ground of contingencies which are not of that sort.
It is easy to see how wide cosmological role, in either the Aristotelian, Bolzanian, or
Fregean/rationalist sense is relevant to objectivity. Indeed, the notions are designed to
be relevant to objectivity. It is stipulated that the ordering itself is objective. But it is not
clear that the notion is of any help in articulating the notion of objectivity, via the axis
of cosmological role. Right now anyway, I do not see how one can further articulate
the notion of objective or ultimate ground without a prior notion of objectivity on
board. Let me spell out the problem a bit.
First, one might think that a proposition has a Fregean ground only if it is objective.
On this view, only objective matters have objective explanations. One cannot even ask
for the ground, or ultimate justification, of non-objective matters. There simply is no
ground for a statement like “Monty Python’s Life of Brian is hilarious.”
Perhaps this is too extreme. It is plausible that the ultimate grounds of humor-facts,
and the like, lie in human sensitivities, or in social evolutionary pressures. That is,
humor-facts, and the like, may have grounds, in the Aristotle–Bolzano–Frege sense.
Still, one might think that only objective matters can serve as grounds for other
propositions. That is, a proposition P can be the objective ground for a proposition
Q only if P is objective.
22
Thanks to Penelope Maddy and Michael Detlefsen here.
23
Frege 1884, }17, Leibniz, Nouveaux Essais, IV, }9; see Burge 1998, 2000 and Jeshion 2001, 2004.
If this is correct, then cosmological role is useless as a criterion, crux, or axis of

objectivity. Only objective matters serve as explanations, in the relevant sense. In
effect, the account of objectivity, via cosmological role, presupposes objectivity.
Moreover, on this view, why does the width of cosmological role matter? On the
theses in question, a proposition figures in any explanation of the relevant sort only if it
is objective. Why does it matter whether it figures in lots of such explanations, in a
wide variety of discourses?
To give the width of cosmological role, as interpreted presently, a chance of playing
the role that Wright assigns to it as an axis of objectivity, we have to allow that non-
objective matters can serve as grounds of other propositions. That is, some (presumably
non-objective) matters have non-objective justifications or grounds. Then we argue,
or claim, that if the pronouncements of a given chunk of discourse figure in a wide
variety of such explanations, then that discourse is objective, on this axis. But what is
the argument or motivation for the width of cosmological role, as presently inter-
preted, as an axis of objectivity? Once we allow non-objective matters into the
objective network of objective grounds, final causes, or whatever, why should one
think that items which ground a lot of different types of propositions are themselves
objective? Since, on the option in question, non-objective facts provide the objective
grounds for propositions, why can’t non-objective facts provide a lot of such grounds,
in a wide variety of discourses?
I do not claim to have presented a full-fledged knock-down argument. It is perhaps
still open for a theorist to articulate an Aristotelian, Bolzanian, or Fregean notion of
cause or ground which does not presuppose the general, intuitive notion of objectivity.
Then, perhaps, such a notion could be used to shed light on what it is to be objective
(along one axis). Meanwhile, let us take a look at other accounts of explanation.
The sub-discipline of philosophy which has given the most attention to explanation
is, of course, the philosophy of science. Perhaps we can get some help from there. The
literature is extensive, and I cannot do much more than scratch the surface—if that.
The idea is to try to extend insights concerning scientific explanation to explanation
generally.
On some irrealist views about science, such as constructive empiricism, explanations
do not have to be true.24 I presume that such accounts of explanation are of no help
here. If one is not going to agree that a proposition or theory that successfully explains a
wide range of contingencies is likely to be true, then why should one think that the
putative explanation is objective, true independent of human interests and judgments?
Let us begin with the early deductive-nomological model of scientific explanation.25
The idea, roughly, is that to explain an event or phenomenon is to deduce a description
of it from Laws of Nature. For example, we explain the tides and the orbits of planets
when we can deduce them from initial conditions and Newton’s Laws of motion. So
24 25
e.g. Van Fraassen 1980. See, for example, Hempel and Oppenheim 1948.
perhaps we can say that a discourse has wide cosmological role if a wide range of
phenomena can be deduced from the Laws within the discourse (together with initial
conditions, definitions, and the like). How does this proposal fare as an articulation of
objectivity, along this one axis?
Clearly, the crucial notion here is that of a Law of Nature. There is, of course, no
shortage of literature on that topic either. The usual contrast is between a Law and an
accidental generalization. According to classical mechanics, it is a Law that all objects
attract each other. It is, I hope, an accidental generalization that every President of the
United States has been male (although it is not exactly a coincidence, of course).
Note that the present proposal would restrict cosmological role to those discourses
that have Laws. If a discourse does not have Laws, then it cannot have a cosmological
role at all, let alone a wide one. Are there Laws of humor? Or of taste? Or of etiquette?
It may be that having Laws is a sufficient, and perhaps necessary, condition for a
discourse to be objective. If so, we may be in a position much like that with objective
explanation above. We may have to know what objectivity is before we can know
what the Laws are, and thus with what explains what. To find the deductive-nomo-
logical account of explanation helpful in articulating cosmological role, one has to
provide an account of Law that does not presuppose a prior account of objectivity.
Another problem with the Aristotle–Bolzano–Frege account may emerge here as
well. If it is necessary to be objective in order to have Laws at all, then why would the
width of cosmological role matter? A discourse is objective if it has some Laws and thus
can figure in explanations at all.
In any case, the deductive-nomological account of (scientific) explanation has been
roundly criticized, at least as being a necessary condition on scientific explanation. The
role of deduction in the account disqualifies common statistical explanations. One cannot
say, for example, that smoking explains cancer since one cannot deduce the presence of
cancer from any Laws about smoking and cancer. Some life-long smokers never get
cancer. Wesley Salmon26 proposed that some explanations can go via statistical relevance
instead of Laws. The idea is that the proposed explanation makes the explanandum
more probable than it would otherwise be. We can say that smoking explains cancer if
it is more probable that someone will get cancer if she smokes than if she does not.
The notion of statistical relevance seems to let in explanations from prima facie non-
objective discourses. The presence of a funny commedian makes it more probable that
someone will laugh, and perhaps it explains the laughing. As I learned, the presence of
an exceptional commedian makes it more likely that I will get a certain kind of stomach
soreness than it would otherwise be. Similarly, the presence of some disgusting broccoli
on my plate makes it more likely, and perhaps explains, that I will be in a bad mood the
rest of the evening, and that I will say things I will regret later, and that my wife will get
angry at me.
26
Salmon 1971.
This move is perhaps promising, for our purposes. The notion of probability
invoked here is statistical, and is thus seemingly objective. Yet, we do not need to
have an articulated notion of objectivity in order to characterize the notion of statistical
relevance.
Clearly, statistical relevance cannot be a sufficient condition on explanation, since it
is usually symmetric. The presence of lung cancer makes it more likely that a given
patient is or was a smoker. But one would not think that the cancer explains the
smoking. Similarly, the fact that I laughed inappropriately at a solemn event makes it
more likely that I recently heard a funny joke, but the laugh does not explain the
funniness of the joke. So this account of explanation must be supplemented somehow.
If this can be done in an acceptable and illuminating manner, then one might attempt
to articulate Wright’s criterion of cosmological role along theses lines: a discourse is up
for an objective interpretation, on this axis, if sentences in the discourse make sentences
in other discourses more likely, given certain other conditions.
Again, for this plan to work in the present context, the “other conditions” in
question should not presuppose the notion of objectivity. To be crude, it would not
help to say that a discourse is explanatory only if it is both statistically relevant and
objective. Ideally, we’d like the conditions to be met for at least some discourses that
are, intuitively, not objective. If the conditions in question can indeed be specified in a
way that does not beg the question, then the issue would come down to whether the
advocate of this option can argue that if the conditions are met in a large number of
statistically relevant cases, involving a wide range of situations—a wide range of
explananda—then the discourse in question satisfies at least some intuitive conception
of objectivity.
More recently, Salmon27 proposed another model of scientific explanation. The
idea is that an explanation of a given event traces the causal processes that brought
about the event, and those processes and events that make up the event itself. The
explanation shows how the event fits “into a causal nexus,” as Salmon put it.28 This
plan disqualifies any discourse that does not deal in causality, but perhaps it is at least a
partial account of explanation. For present purposes, the thesis would be that a
discourse is objective, on this axis, so interpreted, if events described by the discourse
fit into the causal nexus of a wide range of events from a wide range of discourses.
The focus now turns to the notion of causality—yet another notion with a long and
troubled history. And the reaction should be familiar. One might hold that causes must
themselves be objective. In that case, the width of cosmological role seems to be useless
as an articulation of objectivity. We must have a prior account of objectivity before we
can understand what it is for an event to cause another. Moreover, the width of
cosmological role is irrelevant. A discourse is objective if the events it describes figure
in any causal nexus.
27 28
Salmon 1984. Ibid., p. 9.
But perhaps notion of causality is not directly bound to objectivity. Arguably, some
prima facie non-objective discourses can figure in a causal nexus. The funniness of a
commedian caused me to have stomach cramps, and the disgustingness of the broccoli
caused my bad mood and the subsequent marital strain. But if this is admitted, then
why think that if a discourse fits into the causal nexus of a wide range of events then it is
objective?
Let’s take up one final attempt to explain explanation. A popular account is that good
scientific explanations serve to unify various phenomena.29 Accordingly, an event or
discourse explains another if the former is used to derive the latter and relate it to a wide
range of other phenomena. A paradigm case is the role of classical theory of gravity to
explain how and why objects seem to fall, the tides, the orbits of the planets, etc.
Philip Kitcher motivates the criterion: “Science advances our understanding of
nature by showing us how to derive descriptions of many phenomena, using the
same pattern of derivation again and again, and in demonstrating this, it teaches us
how to reduce the number of facts we have to accept as ultimate.”30
Of course “ultimate” is not the same as “objective,” but perhaps we are in the right
ballpark. It depends on whether “ultimate” is to be understood in psychological,
human-interest-relative terms. The issue now is whether unification is a criterion of
objectivity. Why think that just because a chunk of discourse, or a given theory, serves to
unify a disparate range of phenomena, that it is objectively true, independent of our
theorizing, form of life, etc.?
Recall the dilemma that got this project started. Burgess and Rosen31 articulate the
truism that “our theories of life and matter and number are to a significant degree
shaped by our character, and in particular by our history and our society and our
culture.” Of course, if scientists do their job well, our theories are also shaped by the
world, as it is, independent of us and our history and culture. The question at hand
concerns the extent to which we can separate out these factors in our best scientific
theories: “to what extent does the way we are, rather than the way the world . . . is,
shape our mathematical and physical and biological theories of the world?”
Unification is of a piece with simplicity, or perhaps it is just a species of simplicity.
A more unified theory, or explanation, is simpler than one that is more gerrymandered.
It is generally agreed that simplicity and unification are criteria of scientific success.
Other things equal, simpler and unified theories are better. But it seems to me that both
of these criteria sit more comfortably on (or closer to) the “human” than the “world”
side of the mix. It is easy to see why we human scientists prefer simple and unified
theories and explanations. As Kitcher put it, unification “teaches us how to reduce the
number of facts we have to accept as ultimate” (emphasis added). It is essentially the
same reason that carpenters prefer dry wood. It is easier to work with. But do we have
reason to believe, in addition to this, that a unified theory or explanation is more likely
29 30 31
See Kitcher 1989. Ibid., p. 423. Burgess and Rosen 1997, p. 240.
to be objectively true? Do we have reason think that the world itself is simple and
unified, that there are indeed only a few ultimate facts and primitives? Is the world itself
designed, or does it just happen to have, the sort of structure that is conducive to
human understanding, namely a world with only a few primitives. Why think we live
in a world that can be understood with a simple, unified theory?
The above journey does little more than broach or gesture toward a deep and
complex potpourri of issues, but it is about all I can do here. I’ll rest content if the
reader is convinced that the issues surrounding the width of cosmological role need
further development, and that this development cannot proceed without taking sides
on a range of contentious issues in metaphysics and the philosophy of science. To
summarize, it all comes down to what it is to explain. A common pattern emerged
from many of the options broached above. It may be, first, that only objective matters
can serve as explanations at all. Then there are two problems. First, one must already
have an account of what it is to be objective before one can determine what counts as
an explanation and thus with what counts as having a wide cosmological role. This
would make cosmological role useless as a criterion or axis of objectivity. This may be
overcome, either by formulating the notion of explanation without presupposing a
prior notion of objectivity, or perhaps by thinking of objectivity and explanation as
interlocked notions used to illuminate each other. In these cases, however, the width of
cosmological role is irrelevant. A discourse is objective if its propositions figure in any
explanations. If, on the other hand, we allow non-objective matters to serve as
explanations, then why does the width of cosmological role count as an axis of
objectivity? Why think that a discourse that figures in a lot of explanations is more
likely to be objective, in the intuitive sense of being independent of interests, judg-
ments, and the like?
3. Cognitive Shortfall
With cognitive command, another of Wright’s axes of objectivity, we also encounter
deep, difficult, and contentious issues, this time in epistemology. I will do my utmost to
produce more light than heat in briefly covering this most interesting terrain.
Suppose that the purpose of a given area of discourse is to describe mind-
independent features of some mind-independent reality—understood intuitively.
It follows that if two people disagree about something in that area, then at least one
of them has misrepresented that reality. In typical cases, one of them has a cognitive
shortcoming. Suppose, for example, that two eyewitnesses disagree on the question of
whether the bandits left a crime scene in one or two cars. Then, vagueness aside, at least
one of the witnesses did not look carefully enough, had faulty eyesight or memory, had
an obstructed view, or something else along those lines. On the other hand, if a
discourse does not serve to describe a mind-independent realm, then disagreements
in that discourse need not involve cognitive shortcoming on the part of either party.
To pursue a standard example, two people can disagree about whether a given cake is
tasty without either of them having any cognitive shortcoming. One of them may have
a warped sense of taste, or no sense of taste, but there need be nothing wrong with his
cognitive faculties. He can perceive, think, and reason as well as anybody.32
Wright writes that
A discourse exhibits Cognitive Command if and only if it is a priori that differences of opinion
arising within it can be satisfactorily explained only in terms of “divergent input”, that is, the
disputants working on the basis of different information (and hence guilty of ignorance or
error . . . ), or “unsuitable conditions” (resulting in inattention or distraction and so in inferential
error, or oversight of data, and so on), or “malfunction” (for example, prejudicial assessment of
data . . . or dogma, or failings in other categories).33
In other words, if cognitive command fails, then (cognitively) blameless disagree-

ment is possible, or at least it cannot be ruled out a priori.
In contemporary philosophy, non-cognitivism about a given discourse is the view
that assertions in the discourse do not serve to state objective facts. Varieties on that
theme include expressivism and projectivism. Typically, the assertions in the discourse
have the grammatical form of declarative sentences, such as “murder is wrong” and
“Harpo Marx was funny.” So one can speak of “disagreement,” at least prima facie.
The non-cognitivist presumably holds that it is possible for such “disagreements” to be
cognitively blameless.
The purpose of this section is to explore the viability of cognitive command as an
axis of objectivity. We focus on defeasible discourses. Suppose, for the sake of
argument (at least), that it is possible for two scientists to each reach reflective
equilibrium, but come to conflicting conclusions, perhaps because they made different
tradeoffs in the process. The equilibrium envisioned (or postulated) here is more
robust, and more subtle, than that of empirical equivalence and Quinean under-
determination of theory by data. In light of the Quine–Duhem thesis that data is
theory laden, our scientists need not agree on what the data is, or on how it should be
described.34 So they need not see themselves as presenting rival theories to account for
the same data. That is, each of our scientists need not see the other’s theory as
empirically equivalent to his or her own, and there may be no neutral perspective
from which to make a judgment of empirical equivalence.
Let us further assume that no further available data will knock either of our scientists
out of reflective equilibrium, and thus break the deadlock. In other words, we assume
that, for each of them, the overall epistemic situation will not improve, and perhaps
cannot improve. Their theories are rough equals on whatever criteria are used for
32
This is not to say that every disagreement about taste is cognitively blameless. Given some delicate
psychological feedback mechanisms, or some indelicate snobbery, someone may find some poor whiskey
tasty if he misreads the label on the bottle, or is otherwise mistaken about its origin. This once happened to
me.
33
Wright 1992, p. 92.
34
Thanks to Marcus Rossberg here.
assessing scientific theories. Finally, suppose that one of our scientists, Barbie, says P,
and the other, Ken, says ¬P. Let as also assume, for now, that each attaches the same
meaning to P. That is, we begin with the assumption that the disagreement is genuine,
although we will challenge this a few times below.
On the combination of assumptions in play here, it seems that there is be nothing to
fault either scientist. Each has followed the proper methodology flawlessly, and so each
displays no cognitive shortcoming. Yet they come to different conclusions using the
same, fallible methods. So, if the foregoing, admittedly simplified assumptions are
possible, then cognitive command seems to fail for science. Blameless disagreement is
possible. Or so it seems. I submit that there are a number of competing philosophical
conclusions one might draw in this case, with perhaps no clear winner.
Perhaps we were wrong to agree that Ken and Barbie attach the same meaning to
the sentence P. According to some holistic assumptions about semantics, the meaning
of a term is tied to the theory in which it is embedded.35 From a perspective like that,
perhaps, Ken and Barbie do not really disagree; they just talk past each other. If so, then
perhaps cognitive command is saved. There is no blameless disagreement in this case,
since there is no genuine disagreement. I submit that there is a high price to this
conclusion, at least in this case. Once we wax that holistic concerning meaning, is there
ever any disagreement about anything? We have encountered another instance of a
painfully familiar sub-theme of this essay. The underlying issues are too delicate and
complex to be carried much further here, due to lack of space and my own compe-
tence. For now, let us stick to our assumption that Ken and Barbie mean the same thing
by P, and disagree on its truth-value.
In that case, one might conclude that science is not (completely) objective, at least
concerning the matter that separates Barbie and Ken. On this view, there simply is no
fact of the matter, independent of the intellectual lives of scientists, whether P or ¬P is
true. This is of a piece with the motivating theme of this essay, the seemingly inability
to completely separate the “human” and the “world” contributions to our theorizing.
Perhaps Barbie and Ken come to different judgments concerning the simplicity or
unification of a given sub-theory, or that one of them opts for a simpler or unified
theory at the expense of some other criterion. How does that decision bear on the
objective truth of the respective theories? Do we know, or believe, that the universe is
simple or unified in a certain way? Once again, simplicity and the like seem to sit closer
to the “human” than the “world” side of the mix—even if we accept the thesis that
there is no sharp way to separate out the items in the mix.
35
We touch upon yet another deep and interesting philosophical topic here. When is it that two people
disagree? Intuitively, and as a first pass, Ken disagrees with Barbie if he endorses a proposition that is logically
incompatible with a corresponding proposition that Barbie endorses. This, of course, raises issues concerning
the nature of propositions, and how token utterances manage to express propositions. These questions are
crucial in deciding between contextualist and relativist accounts of various topics. See MacFarlane 2007.
An extreme version of this first option would be to deny that there is a fact of the
matter—objective or otherwise—whether P or ¬P. A less extreme version would
allow that there is a fact of the matter, but that it is not objective—not independent of
the human theoretical situation. Or one might go relativist, holding that P is true for
Barbie, and ¬P is true for Ken.
There is a version of relativism that may allow us to reintroduce objectivity into the
situation, at the expense of one of our pet assumptions. It may be that Barbie and Ken
are onto different aspects of an objective physical world, much as in the story in which
blind people around an elephant are onto different aspects of the (presumably objec-
tive) “reality.” One of them describes the trunk, saying it is rough and thick; another
describes an ear, saying it is smooth and thin; a third describes a leg, saying it is thick and
solid, etc. Once again, if we look at it this way, the disagreement between Barbie and
Ken is not genuine.36 Since they are talking about different aspects of reality, then they
do not contradict each other. At some level, they do not attach the same meaning to
the terms, contra our assumption.
Note that some of the options broached so far deny the objectivity of science, at least
in part. This flies in the face of a strong intuition that there is a physical world, not of
our making. According to this intuition, our two scientists, Ken and Barbie, are trying
to describe the world’s mind-independent properties, and at least one of them is
mistaken (perhaps faultlessly). Of course, this intuition is not sacrosanct, and many
philosophers have turned to views like instrumentalism and constructive empiricism,
not to mention idealism, in response to various considerations, including the under-
determination of theory and the role of criteria like simplicity and unification in
scientific methodology. Most versions of the first option do not go nearly that far,
proposing only that at times, science is not completely objective.
Nevertheless, let us explore some options that allow one to maintain an intuitive
realism for science. Science may or may not be holistic, but, presumably, philosophy is.
Maybe someone can maintain the naive realism and adjust elsewhere.
The most straightforward move for a scientific realist is to hold that if the present sort
of conflicting equilibria are possible, in the extreme idealized sense contemplated here,
then epistemic constraint fails for science.37 Given the realism—and bivalence—either
P is true or else ¬P is, but we just do not know, and indeed cannot know, which of our
two scientists is right. If, for example, Barbie is the one who got things right, then she
just got lucky. If P is in fact unknowable, then, of course, Barbie does not know that P.
36
Thanks to Carrie Jenkins here. To add one more option to the mix, someone might suggest a dialetheic
resolution of the situation: P and ¬P are both true, of the same reality, in the same sense. In this case as well,
there is no disagreement, since, according to the dialetheist, Ken does not say anything that is logically
incompatible with what Barbie says. Mark Silcox and Jon Cogburn (2006, pp. 374–5) discuss, and reject, a
suggestion (due to Brandon Cooke) that cognitively blameless disagreement concerning literary interpreta-
tion motivates dialetheism. Silcox and Cogburn argue that in such cases, the parties are talking about different
objects, and so are not disagreeing.
37
Thanks to Crispin Wright here.
No one does. On this combination of views, we can rely on Wright’s thesis that the
criteria of cognitive command (as well as the width of cosmological role and the
Euthyphro contrast) is an axis of objectivity only if the discourse is epistemically
constrained. On this option, the failure of cognitive command is simply irrelevant to
the matter of objectivity: science is objective just because it is epistemically uncon-
strained.
There is another option, which also enjoys some intuitive plausibility. Here, again,
delicate issues in epistemology come to the fore. One might think that epistemic
constraint has not been ruled out by our scenario of blameless disagreement, even
assuming the objectivity of science. Suppose, again, that Barbie is the one who is in fact
correct. One might argue that, lucky or not, she does know that P. By hypothesis, P is
true. Moreover, Barbie believes P, and this belief was arrived at by correctly—
flawlessly—following the best scientific procedure available. What more does it take
for (albeit fallible) knowledge?
As often happens in philosophy, it is easier to ask questions than to answer them. It
can be agreed, I think, that under the prevailing assumptions, Barbie has justified, true
belief that P is true. But it is also generally held nowadays that justified, true belief is not
sufficient for knowledge, especially in defeasible areas like science, thanks to Gettier.
Barbie’s warrant for P consists of its place in her overall theory, found to be in
reflective equilibrium, meeting the relevant scientific criteria as well as possible. On the
assumptions in question, there seems to be a defeater to her warrant for believing P,
namely Ken’s having come to the opposite conclusion, making different tradeoffs
along the way. If Barbie were to become aware of Ken’s overall account, then perhaps
she should retract her claim to know that P. If she is honest in assessing the situation,
she may even agree with the foregoing hypothesis that her epistemic state is no better
than Ken’s. Presumably, Ken would similarly retract his belief in ¬P.
Notice that this contradicts the foregoing hypothesis that the epistemic situations for
Barbie and Ken will not improve. Admittedly, nothing breaks the tie, but something
has come along to move Barbie and Ken to be agnostic about P, and, arguably, this
improves the situation, in good Socratic fashion. Barbie perhaps realizes that she does
not know, and so becomes agnostic.
Actually, I am not sure that this is correct, or even that it is a natural conclusion to
draw. Perhaps we are not being sufficiently holistic concerning the epistemic situations
here. I submit that Barbie does not need to concede that Ken’s situation is as good as
her own, even after she reads and absorbs Ken’s account of the matter. 38 How one
assesses a given piece of evidence depends, in part, on what other beliefs one has, and
on what one considers the “data” to be. Just as there need not be any theory-neutral
description of the data, there may be no theory-neutral account of epistemic assess-
ment—no theory-neutral way to assess the situation. My above hypothesis that the
38
Here we scratch the surface of another important issue in epistemology, the proper weight to be placed
on informed disagreement. See, for example, Christensen 2007 and Elga 2007.
theories score as rough equals is made from an external perspective, one that neither
Ken nor Barbie completely share. Perhaps it is the perspective of a third scientist who is
reviewing their respective work.
To keep the discussion going, and explore the options at hand, let us agree that
Barbie maintains her belief in P, and Ken maintains his belief in ¬P, in whatever
epistemic situation we envision for them. To be more specific to the case at hand, our
working hypothesis is that for Barbie to maintain P and for Ken to maintain ¬P is not
for either of them to exhibit a cognitive shortcoming.
Our third option, then, is that even if, in the scenario in question, Barbie should not
claim to know P once she becomes aware of Ken’s account of things, she knows P
anyway. Maybe the potential defeater—Ken’s theory—does not undermine Barbie’s
justification, on whatever the correct account of justification is. By hypothesis, the
counter-evidence is misleading, since Barbie is correct in her belief. So we maintain
that Barbie does indeed know P, this despite Ken’s admittedly flawless perspective to
the contrary.
We can get a bit sharper here at the cost of getting more controversial. Suppose that
the correct epistemology (for science) is some sort of reliabilism, an externalist view.
From such a perspective, one might think, Barbie has knowledge of P simply because
her belief in P was generated by a reliable (albeit fallible) “mechanism,” scientific
method itself. For Barbie, at least, there are no Gettier-style defeaters. Her methods are
generally reliable, and they worked properly this time to produce a true belief. For a
reliabilist, what else does it take for Barbie to know?
To summarize, an advocate of this third option maintains the intuition that science is
objective and that epistemic constraint holds for it; and hopes (or argues) that the
correct epistemology cooperates with these conclusions. In this case, blameless dis-
agreement is nevertheless possible. So if the third option prevails, then cognitive
command is not a reliable indicator of objectivity, contra one of the themes of Wright
1992. Cognitive command fails for one of the paradigms of objectivity, science itself.
Chapter 4 of Wright 1992 deals with the aforementioned Quine–Duhem phenom-
enon that observation is theory-laden. Wright comes to tentative conclusions similar to
those reached here. If observation is indeed theory-laden in the indicated way, then it is
hard to maintain both that epistemic constraint holds in science and that cognitive
command is a criterion of objectivity.
A referee suggested that the negative conclusion here is reminiscent of the failure of
Logical Positivism. In particular, the attempt to invoke cognitive command to distin-
guish objective from non-objective discourses is of a piece with the attempt to
characterize the notion of cognitive significance. In both cases, the holistic nature of
the enterprise compromises the attempt at a clean distinction. The problem we face
here is that it seems impossible to distinguish failures of cognitive command that are
due to the non-objective nature of the subject-matter—disputes concerning taste and
humor or, to use the positivist example, the excesses of metaphysics—from failures of
cognitive command that are due to the highly theoretical nature of the subject matter,
and its holistic epistemology. We have again encountered what I call above the
Kantian/Waismannian/Quinean perspective.39
As the song goes: come right back. There may be a fourth option, and this will be
the last considered here. I promise. Maybe we can maintain that science is objective
and epistemically constrained, and that cognitive command holds after all. The claim is
that if Barbie does in fact know P, then perhaps the other scientist in the scenario, Ken,
exhibits a cognitive shortcoming. He believes an objective falsehood, ¬P, and, by
hypothesis, P is knowable, since Barbie knows it. What else does it take to exhibit a
cognitive shortcoming?
One intuition behind this thought is that if we are going to wax externalist
concerning knowledge, and grant that Barbie knows P, then perhaps we should go
externalist concerning “cognitive shortcoming” as well. Ken is at “fault” simply
because he denies something that can become known (on the presumed externalist
account of knowledge). Ken does not have to be aware, internally, that he is at fault.
We can, of course, define our terms as it pleases us, but it does not seem appropriate or
useful to extend the notion of “cognitive shortcoming” that far. By hypothesis, both
scientists acted in the epistemically most responsible manner possible. So in that sense,
neither can be faulted. As noted above, Ken was simply unlucky. Should this bad luck
count as a cognitive shortcoming, as a blameworthy disagreement? If it is not Ken’s fault,
then why is he to be blamed, just because he happens to believe something whose
negation is knowable? Perhaps “cognitive shortcoming” has to be understood internally.
This fourth orientation is explicitly ruled out by Wright’s later work. William
Taschek and I once argued that, in general, epistemic constraint entails that cognitive
command holds.40 In other words, if all truths in an area of discourse are knowable,
then there can be no blameless disagreement in that area. It follows that cognitive
command is not really a different axis of objectivity, over and above epistemic
constraint, or so we suggested.
Our argument does not assume bivalence, and uses only intuitionistic logic. Here it
is, in schematic form:41
1. For any statement S, if S then S is knowable. (Epistemic Constraint)
2. Subject A sincerely believes P, subject B sincerely believes ¬P, and neither A
nor B have any cognitive shortcomings, nor have they less then fully exercised
their capacities. (premise)
3. Suppose P (assumption)
4. It is knowable that P. (from 1&3)
39
The referee also suggests that even if Wright gives up on trying to distinguish subject-matters by levels
of objectivity—one of the key goals of Wright 1992—the notion of cognitive command has a lot of
explanatory value. Presumably, the notion would still help to distinguish disputes within various dis-
courses—be they about humor, taste, metaphysics, grammatical deep structure, or subatomic particles.
40
Shapiro and Taschek 1996.
41
Ibid., p. 85
5. Subject B has come to believe something that is not only false but is the
negation of something knowable. (from 2&4)
6. B has a cognitive shortcoming. (from 5)
7. Contradiction (from 2&6)
8. ¬P (from 4–7, discharging 3)
9. It is knowable that ¬P. (from 1&8)
10. Subject A has come to believe something that is not only false but whose
negation is knowable. (from 2&9)
11. A has a cognitive shortcoming. (from 10)
12. Contradiction (from 2&11)
13. Epistemic constraint is inconsistent with the failure of cognitive command.
(from 1–12, discharging 1,2)
In his detailed reply to our argument, Wright 2001 proposed that the quantifiers in
the formulation of cognitive command should be understood constructively, via
intuitionistic logic. For cognitive command to hold, Wright argues, it is not enough
that blameless disagreement be ruled out. By the above argument, that much happens
whenever epistemic constraint holds. The proper criterion for cognitive command is
that whenever there is disagreement, then it is possible to assign blame to one or the
other participant. The view
that cognitive command holds whenever conflict of opinion is possible . . . demands, in the
presence of [epistemic constraint] that there be an identifiable shortcoming in [one of the]
conflicting opinions . . . So although indeed in position to rule out the suggestion that any
disagreement is cognitively blameless . . . we remain . . . unentitled to the claim that there will
be a cognitive shortcoming in any difference of opinion . . . We remain so unentitled precisely
because that would be a commitment to a locatability claim . . . The immediate lesson is that it is an
error (albeit a natural one) to characterize failures of cognitive command . . . in terms of the
possibility of blameless differences of opinion . . . We do know [that blameless disagreement is not
possible]. But that is not sufficient for cognitive command.42
On this articulation of the present axis of objectivity, the proper slogan for cognitive
command is not that blameless disagreement is impossible, but rather that wherever
there is disagreement, there is locatable blame.
The constructive reading of cognitive command is present in the 1992 book, when
Wright takes up a different matter:
it cannot be . . . that [a cognitive] shortcoming may be indefinitely unidentifiable. For if that were
so—if a situation were possible where we could definitely say that one of a pair of disputing
theorists was guilty of cognitive shortcoming, but there was no way of saying who—then there
would be no identifying the winner (if any) in the dispute either; and that would be as much as to
allow that a true theory might be unrecognizable after all. When truth is regarded as essentially
42
Wright 2001, p. 86.
epistemically constrained, Cognitive Command requires the identifiability of cognitive shortcom-

ing whenever it occurs.43
OK. Now let us return to Barbie and Ken. One can still maintain that, even with the
constructive reading, cognitive command still holds. The constructive reading requires
that the Ken’s cognitive shortcoming can be identified. Well, it can be. Barbie identifies
it. By hypothesis, she knows P, and thus she knows that Ken is in error. All she needs to
do is to point to the place in Ken’s reasoning where he differs from her. As it happens,
that difference led to Ken to an error, and so it is a cognitive shortcoming. It can be
located, since Barbie can locate it, and, in a sense, she can know that she has done so.
Since she knows P, she knows that Ken’s move was in error. What else does it take to
identify a shortcoming in this admittedly defeasible area of inquiry? To use Wright’s
terminology, she correctly identifies the winner—herself.
I presume that this last move is not in the spirit of Wright’s reply to Shapiro and
Taschek 1996, or the above comments from Wright 1992 concerning the identifica-
tion of the winner. Perhaps the underlying assumption is that in a sense, cognitive
shortcoming has to be understood internally—no matter what we say about knowl-
edge. Any cognitive shortcoming should be accessible, internally, to a reasoner who is
sufficiently attentive and careful. The said reasoner should be able to recognize that he
is in error. So if Ken has a cognitive shortcoming, he should be able to locate it himself.
Or perhaps the constructive requirement on cognitive command is that the shortcom-
ing be locatable by someone neutral to the dispute. For cognitive command, as
interpreted constructively, to hold, it must be that a disinterested and dispassionate,
but fully rational, third party should be able to determine which of Barbie and Ken has
a cognitive shortcoming.
Wright proposes that failures “of cognitive command . . . must be viewed as situa-
tions where we have no warrant for a certain claim.” The question here is, who is the
“we.” Given the assumptions under our fourth interpretive option, it can’t be Barbie,
or any of her colleagues. I propose that the “we” must be a neutral referee, perhaps a
third scientist writing a review of Ken’s and Barbie’s work. The prevailing assumption
of the scenario—which we may call robust reflective equilibrium—is that this neutral
party cannot adjudicate the dispute, and thus locate any cognitive shortcoming in
Ken’s or Barbie’s work. By hypothesis, both of them followed the best available
procedure flawlessly. Our hypothetical referee cannot identify a shortcoming unless
she could somehow identify which of them is right, and by hypothesis, this is just what
our referee cannot do. Such is defeasible discourse, under the assumption of conflicting
reflecting equilibria.
Let me briefly summarize the options on the table. First, one might claim that the
discourse in question is not objective, simply because cognitive command fails. One
can either deny that there is a fact of the matter at all, or just that there is no objective
43
Wright 1992, p. 163.
fact of the matter. Second, one might claim that the discourse in question is objective,
but that epistemic constraint fails for it. The true proposition in question is unknow-
able. Third, one can maintain that the discourse in question is objective and that
cognitive command fails anyway—in which case cognitive command does not track
objectivity. We have a counterexample to Wright’s thesis. I take the fourth option—
where the incorrect party automatically has a cognitive shortcoming—as ruled out, if
we properly understand the constructive nature of Wright’s notion of cognitive
command.
One definitive meta-conclusion of this section is that the issues with Wright’s
detailed account of cognitive command invoke delicate matters in epistemology, as
well as issues concerning what counts as “cognitive.” The suitability of cognitive
command as an axis of objectivity depends on how these controversial matters are
articulated and adjudicated.
Note
This essay develops some issues that are briefly broached in my “The Objectivity of Mathemat-
ics,” (2007). I am indebted to the philosophy department at the University of Notre Dame, the
University of Paris 7, and the Arché Research Centre at the University of St Andrews for
devoting sessions to the project. Feedback was most helpful. Thanks also to Marco Panza, Bob
Hale, Fraser MacBride, Carrie Jenkins, Crispin Wright, and two anonymous referees. I especially
appreciate the spirit of collegiality.
Bibliography
Benacerraf, P. and Putnam, H. 1983 Philosophy of Mathematics, 2nd edn., Cambridge, Cambridge
University Press.
Bolzano, B. 1837 Theory of Science, trans. R. George, 1972, Berkeley, University of California
Press.
Burge, T. 1998 “Frege on Knowing the Foundation,” Mind 107, pp. 305–47, reprinted in Burge
2005, Chapter 9.
Burge. T. 2000 “Frege on Apriority,” in P. Boghossian and C. Peacocke (eds.) New Essays on the
A Priori, Oxford, Oxford University Press, pp. 11–42; reprinted with new appendix in Burge
2005, Chapter10.
Burge, T. 2005 Truth, Thought, Reason: Essays on Frege, Oxford, Oxford University Press.
Burgess, J. and Rosen, G. 1997 A Subject with No Object: Strategies for Nominalistic Interpretation of
Mathematics, Oxford, Oxford University Press.
Christensen, D. 2007 “Epistemology of Disagreement: The Good News,” The Philosophical
Review 116/2, pp. 187–217.
Dummett, M. 1973 “The Philosophical Basis of Intuitionistic Logic,” in Truth and Other Enigmas,
Cambridge Mass., Harvard University Press, 1978, pp. 215–47; reprinted in Benacerraf and
Putnam 1983, pp. 97–129.
Elga, A. 2007 “Reflection and Disagreement,” Noûs 41/3, pp. 478–502.
Frege, G. 1884 Die Grundlagen der Arithmetik, Breslau, Koebner, trans. J. L. Austin, 1960 The
Foundations of Arithmetic, 2nd edn., New York, Harper.
Hempel, C. and Oppenheim, P. 1948 “Studies in the Logic of Explanation,” Philosophy of Science
15, pp. 135–75.
Jeshion, R. 2001 “Frege’s Notions of Self-Evidence,” Mind 110, pp. 937–76.
Jeshion, R. 2004 “Frege: Evidence for Self-Evidence,” Mind 113, pp. 131–8.
Kitcher, P. 1989 “Explanatory Unification and the Causal Structure of the World,” in P. Kitcher
and W. Salmon (eds.) Scientific Explanation, Minneapolis, University of Minnesota Press,
pp. 410–505.
MacFarlane, J. 2007 “Relativism and Disagreement,” Philosophical Studies 132, pp. 17–31.
Sainsbury, R. M. 1994 “Why the World Cannot Be Vague,” The Southern Journal of Philosophy 33
supplement, pp. 63–81.
Salmon, W. 1971 “Statistical Explanation,” in W. Salmon (ed.) Statistical Explanation and
Explanatory Relevance, Pittsburgh, University of Pittsburgh Press, pp. 29–87.
Salmon, W. 1984 Scientific Explanation and the Causal Structure of the World, Princeton, Princeton
University Press.
Shapiro, S. 2007 “The Objectivity of Mathematics”, Synthese 156, pp. 337–381.
Shapiro, S. and Taschek, W. 1996 “Intuitionism, Pluralism, and Cognitive Command,” Journal
of Philosophy 93, pp. 74–88.
Silcox, M. and Cogburn, J. 2006 “Computability Theory and Literary Competence,” British
Journal of Aesthetics 46, pp. 369–86.
Tennant, N. 1997 The Taming of the True, Oxford, Oxford University Press.
Van Fraassen, B. 1980 The Scientific Image, Oxford, Oxford University Press.
Waismann, F. 1968 “Verifiability,” in A. Flew (ed.) Logic and Language, Oxford, Basil Blackwell,
pp. 117–44.
Wright, C. 2001 “On Being in a Quandary: Relativism, Vagueness, Logical Revisionism,” Mind
110, pp. 45–98.
Yablo, S. 2001 “Go Figure: A Path through Fictionalism,” in P. A. French and H. K. Wettstein
(eds.) Figurative Language: Midwest Studies in Philosophy 25, Oxford, Blackwell, pp. 72–102.
Yablo, S. 2006 “The Myth of the Seven,” in M. E. Kaledron (ed.) Fictionalism in Metaphysics,
9
How to Formulate Relativism
Carol Rovane
1. Introduction
There is no very clear, generally accepted formulation of the doctrine of relativism that
fulfills all of the following desiderata: (1) the doctrine emerges as coherent; (2) it emerges
as enjoining a distinctive metaphysical commitment; (3) it emerges as controversial in
ways that make some of the traditional forms of opposition to it understandable—not, of
course, the forms of opposition that would have us dismiss the doctrine out of hand as
incoherent, but rather forms of opposition that view relativism as ruled out by some
other highly valued metaphysical commitment.
My aim in this essay is to begin the task of formulating the doctrine in a way that has
a realistic hope of satisfying these desiderata. I say begin the task, because the formula-
tion on offer will be left underdeveloped in many crucial respects. Still, it is a proposal
that seems to me worth exploring. Until some such proposal is fully implemented,
there is little interest or point in attempting to evaluate the doctrine.
When we lack a clear formulation of a philosophical doctrine, we can often make
headway towards elucidating it by starting with a widely accepted and paradigmatic
example. But there are several reasons for caution against placing undue weight on
examples in philosophical discussions of relativism. First, even among those who are
sympathetic to relativism there isn’t perfect agreement about what counts as a paradig-
matic example. I say this even though many recent discussions have focused so often
(indeed almost exclusively) on so-called “disputes of inclination,” in which two parties
are alleged to affirm and deny a single proposition concerning some matter where
gustatory values are concerned, such as, say, whether rhubarb or fish sticks are delicious.
These recent discussions are just the latest installment in a longer history of debate on the
subject that in the twentieth century took on a much wider range of examples drawn
from cultural anthropology, philosophy of science, ethics, and politics. There is a real
possibility that those examples do not all illustrate a single phenomenon. So if we wish to
argue that any one of the phenomena that were illustrated in them is the one that most
deserves to be viewed as a paradigmatic instance of relativism, we shall have to explain
why and, in that case, the real point of departure for our effort to formulate the doctrine
H OW T O F O R M U L AT E R E L AT I V I S M 239
will not be the illustrative example itself but rather our explanation of why we view it as
an illustrative example. What is more—and this is my second caution—once we confine
our attention to a specific example (or sort of example), we shall find that it can be
interpreted in different ways that reflect different intuitions about what relativism
involves, and we shall have to explain why we advocate one interpretation over others.
So yet again the example itself will not be the real basis from which our effort to
formulate the doctrine is proceeding, but rather our explanation of why we interpret the
example in the way we do. A third caution is this: not only can we interpret any given
example in different ways so as to illustrate different intuitions about relativism, we can
also interpret it so that it fails to illustrate any recognizably relativist position at all.
Opponents sometimes take this fact—that we are never forced to interpret a given
example as illustrating relativism—as a reason to dismiss the doctrine out of hand.
However, this response fails to take into account the possibility that we may not be
forced to give a non-relativist interpretation either. This seems to me to be a real
possibility, since those who think we are forced to give a non-relativist interpretation
are very likely bringing to bear assumptions and commitments that relativists would be
within their rights to regard as question-begging. This raises an important question
about who bears the burden of proof in debates about relativism, and a related question
about what form such proof should take. Obviously, we cannot properly address these
questions without first arriving at a satisfactory formulation of the doctrine itself, so that
we can be clear about what, exactly, is at issue in such debates. So I conclude that we
cannot reasonably expect to get very far on this prior task of formulation just by staring at
examples. A better approach would be to start with some of the things that philosophers
have said about the doctrine, in order to gain a better sense of what it is (more or less
exactly) that we would be trying to find examples of.
Among the many philosophers I have read on the subject, I have admired and been
instructed measurably more by Crispin Wright than any other. He has never under-
estimated the difficulty of formulating the doctrine of relativism in a satisfactory way.
Yet he has always remained open-minded about the prospects for doing so, and he has
offered some extremely interesting and creative proposals about how we might go
about doing so.
As I’ve said, recent discussions tend to conceive relativism as arising with so-called
disputes of inclination, disputes in which, it is alleged, that it is very intuitive that both
of the disputants can be right even though they disagree. From here on, I shall refer to
this intuitive conception of relativism as the Disagreement Intuition. Wright takes this
intuition seriously on several related grounds. He takes it for granted that disputes of
inclination are familiar occurrences in our actual, lived interpersonal dealings, and that
we generally experience them as irresoluble. He also takes it for granted that we do not
generally see their irresolubility as signaling an absence of knowledge on anyone’s
part—as if what we needed was more inquiry or a better theoretical framework in the
light of which one party or the other (or perhaps both) might emerge as mistaken. If
I were to say that rhubarb is delicious and Wright were to deny it, and if I initially
240 C A RO L ROVA N E
responded by trying to insist that there is a truth at issue that he is missing, he would be
within his rights to insist on his own view of the matter. Furthermore, he would not
merely be claiming a right to make his own mistakes in the spirit of freedom of
opinion; he would be expressing his conviction that even if there is a sense in which
I may be correct to say that rhubarb is delicious, this doesn’t rule out his being correct
to deny it. So according to Wright—and he is very much in step with a prevailing
philosophical consensus—what the relativist needs to make sense of is the possibility of
such irresoluble disagreements.
I want to underscore that it is crucial that the irresolubility in question should
emerge as having metaphysical and not merely epistemic significance. For suppose
that the reason why disputes of inclination and other similar disagreements cannot be
resolved is the fact that we are subject to epistemic limitations that prevent us from
figuring out which party (if any) is mistaken and which (if any) is right. This wouldn’t
by itself occasion any distinctive or controversial distinctive metaphysical commitment
of the relativist—the idea that we may be subject to epistemic limitations, even
insuperable ones, is surely available to opponents of relativism. Thus the only way in
which the irresolubility of certain disagreements could bear a distinctive metaphysical
significance is if both parties could actually be right, at least in principle.
Insofar as Wright has not been entirely dismissive of relativism, it is because he views
the framework of anti-realism as providing some prospects for making sense of this
metaphysical possibility. The natural thought here is that the truth in such domains is
whatever ideally informed subjects would be warranted in asserting, and so if it ever
were to happen that such ideally informed subjects were warranted in asserting
contrary propositions, they would both be right. (In the space I have and with much
to discuss ahead, obviously I have had to expound this natural thought much more
crudely and briefly than is found in Wright’s elaborate and interesting writings about
the connection between anti-realism and relativism.)
It seems that the reasons I have just reviewed are all good reasons for taking the
Disagreement Intuition seriously, especially in the domain of value, and it isn’t
surprising that they are widely accepted among contemporary philosophers who
write on the topic of relativism. Many of them find it plausible that the truth
concerning certain evaluative matters is a matter of what ideally informed evaluators
would judge to be valuable, and that if such ideally informed evaluators were to
disagree, they would both be right. Furthermore, whenever we find that a philosophical
position in value theory fails to guarantee against the possibility of such metaphysically
irresoluble disagreements, it is routinely classified as “relativist” on that score.
But the Disagreement Intuition seems to face, at first sight, an obvious logical
difficulty: parties who disagree affirm and deny the same proposition, and to allow
that they can both be right seems, on the face of it, to allow an exception to the law of
non-contradiction. The most common strategy of response to this logical difficulty is
to try to avoid outright contradiction through a ‘semantical’ maneuver, in which the
truth of the parties’ respective claims is portrayed as relative to different contexts. This
“semantical relativist” strategy invokes a second intuitive conception of the doctrine of

relativism, which I will call the Relative Truth Intuition. The Relative Truth Intuition
can claim some independent motivation, since it is strongly evoked by the very name
of the doctrine of relativism, and also figures in much popular discourse about it. Yet it
is not widely regarded by contemporary philosophers as capturing an independently
distinctive metaphysical commitment of the relativist. In their view, only the Disagree-
ment Intuition can capture that commitment, and their main interest in the Relative
Truth Intuition lies in whether it can help us to make adequate philosophical sense of
the Disagreement Intuition.
Wright has argued on more than one occasion that the various versions of such a
‘semantical’ relativism currently on offer do not succeed on this score. His arguments are
so convincing that they should really just be cited rather than reviewed and discussed in
detail. Accordingly, I won’t rehearse them here. I’ll simply lay out what I see as the most
basic reasons why we shouldn’t expect that the Relative Truth Intuition can help us to
make adequate philosophical sense of the Disagreement Intuition.
My own firm conviction is that the Disagreement Intuition clearly saddles the
relativist with having to allow exceptions to the law of non-contradiction, and it
isn’t reasonable to expect that we can elaborate it in a way that avoids this intolerable
consequence. Nevertheless, much of Wright’s work on the subject of relativism has
been devoted to exploring the extent to which it can be avoided, albeit without
invoking the Relative Truth Intuition. I think he would concede that his earliest efforts
in this regard vindicate my conviction. I’m not sure whether he would concede that his
more recent efforts, which invoke the idea of a Quandary, also vindicate it—though
that is what I shall argue.
If I’m right that we cannot reasonably expect to rescue the Disagreement Intuition
from its logical difficulty, it doesn’t follow that there is no coherent way to formulate the
doctrine of relativism. There is yet another intuition about its content that we might
explore, which I’ll call the Alternatives Intuition. What I mean by “alternatives” is this:
alternatives are truths that cannot be embraced together. It seems to me that the Alternatives
Intuition actually subsumes the Disagreement Intuition, in the sense that the latter offers
a particular way of understanding how alternatives could arise: it says that some truths
cannot be embraced together because they stand in logical conflict. However, it is
striking that in the main twentieth-century debates about relativism, alternatives were
not generally conceived in that way. Consider, for example any of the following:
Carnap’s linguistic frameworks, the different conceptual schemes that were described
by cultural anthropologists, the theoretical paradigms that figured in Kuhn’s accounts of
scientific rationality, Goodman’s “ways of worldmaking.” None of these candidates for
the status of ‘alternatives’ were portrayed as standing in direct logical conflict.
The hard question confronting the Alternatives Intuition is whether there is any
other way of making sense of why truths cannot be embraced together, besides their
being in logical conflict. Prima facie, it is hard to see what else could. This fact helps to
explain why the Disagreement Intuition has such a grip on us; but it is also a fact that
fuels one major line of objection against relativism. The objection puts forward a
dilemma for alternativeness: any pair of truth-value-bearers is either inconsistent or
consistent; if they are inconsistent then they cannot be embraced together but they
cannot both be true; if they are consistent then they can both be true which means that
they can also be embraced together; in neither case do we have alternatives in the sense
required by the Alternatives Intuition.1 Anyone who thinks that the two horns of this
dilemma exhaust the logical possibilities, and who regards the reasoning of the first
horn as correct, must conclude that there isn’t any scope for alternativeness at all, and
indeed that we cannot make sense of the very idea of alternativeness.
It appears to be a daunting task to make adequate philosophical sense of the idea of
an alternative in the face of this dilemma. Nevertheless, there are two points that speak
in favor of trying to do so: first, the Alternatives Intuition really does seem to be a
central intuition about relativism, and second, insofar as it makes no direct mention of
logical conflict, it does not automatically fall prey to the logical difficulty faced by the
Disagreement Intuition, viz., its susceptibility to falling foul of the principle of non-
contradiction.
The first point is confirmed by the fact that the dilemma for alternativeness is so
frequently raised by opponents of relativism in informal philosophical discussions. It
isn’t always articulated in exactly the terms I have used. But more than once I have
heard it voiced in the following way: “What is it that is supposed to make so-called
alternative conceptual schemes competitors? Is it that they logically conflict? If so, then all
we have is an ordinary disagreement, which is surely too innocuous to amount to a
relativism; on the other hand, if they don’t logically conflict, what makes them
competitors? Why can’t they both be true? And if they are both true, where is the
relativism? I simply do not see what the relativist’s position is supposed to be!” This
couldn’t be a constant and relevant complaint against relativism unless the Alternatives
Intuition about what the doctrine of relativism is, were indeed central to relativism; and
so it does seem worth exploring.
The second point, that the Alternatives Intuition can be articulated without directly
mentioning logical conflict, is not significant unless we can identify some other reason,
aside from logical conflict, why it might be the case that some truths cannot be
embraced together. This is not possible unless there is a third possibility that the
dilemma for alternativeness overlooks, which is the possibility that, in some cases,
truth-value-bearers are neither inconsistent nor consistent.
It shouldn’t be hard to see that if it ever were the case that truth-value-bearers were
neither inconsistent nor consistent, then they wouldn’t stand in any logical relations at
all. This means that they wouldn’t together fall under any of the normative constraints
that follow upon the presence of logical relations; they would be, as I shall put it,
1
What I mean by “truth-value-bearer” is anything that is capable of being true or false, such as sentences,
utterances, propositions, and beliefs. It doesn’t matter for my purposes in this essay which things actually bear
truth-values, and so I shall refer to them all indifferently as truth-value-bearers.
normatively insulated from one another. Here is how such normative insularity would
provide for alternativeness: since normatively insulated truth-value-bearers would not
stand in the logical relation of inconsistency, they would not together fall under the law
of non-contradiction and, as a result, they could both be true without violating that
law; since they would not stand in the logical relation of consistency either, they would
not together fall under the law of conjunction and, as a result, that law wouldn’t license
us to conjoin them even if they were both true; it follows such normatively insulated
truth-value-bearers could be true but not conjoinable—that is, they could qualify as
alternatives in the sense I have defined, of truths that cannot be embraced together.
When we construe the doctrine of relativism as involving a commitment to the
existence (or perhaps just the coherent possibility) of alternatives, and when we
construe alternatives in terms of normative insularity, the doctrine emerges as having
a distinctive metaphysical significance which is also, at the same time, a distinctive
practical significance.
The metaphysical significance of relativism, so formulated, is best appreciated by
reflecting on what the doctrine asks us to give up. If there are no alternatives, then all of
the truths can in principle be embraced together, and this invites us to think that there
is a single, consistent and complete body of truths. Metaphysically speaking, this is a
conception of the world as one in roughly the sense that Wittgenstein articulated in the
opening of the Tractatus, when he said that the world is all that is the case.2 On my
proposed formulation, then, a coherent relativism would be asking us to give up that
unimundial view in favor of a multimundial view instead. It would be claiming that the
world is not one but many, because there isn’t one single and complete body of truths,
but rather many incomplete bodies of truths that cannot be embraced together.
Most philosophers presume that any plausible formulation of the doctrine of
relativism would have to reveal it as a doctrine that stands opposed to realism. But if
I’m right, then relativists are not directly calling into question whether reality is mind-
independent in the way that realists claim. What they are most directly calling into
question is whether reality is one in the sense I just explained above. As it turns out, it is
possible to deny the oneness of reality without also denying its mind-independence—in
other words, an argument for relativism needn’t be an argument against realism. But this
is not a point that I have room to develop in this essay.3
The converse point, that an argument against realism needn’t be an argument
for relativism, is much more familiar and not at all controversial. Many varieties of
2
I have been arguing that we should see this issue—the oneness of the world—as the real dividing issue
between relativists and their opponents since my first publication about relativism in Rovane 2002. In that
paper and all I have published since on the topic I have noted the affinity between unimundialism and
Tractarian realism. For this reason I initially didn’t see the need to introduce the term “unimundialism” at all.
I subsequently came to see that the term “realism” is so firmly associated with the idea that reality is mind-
independent that we need another term entirely to capture what I have found referred to what I take to be
the “Tractarian condition” of oneness.
3
See my Rovane, 2010b.
anti-realism obviously do not challenge the oneness of the world in any way—for
example, Berkeleyian idealism does not. All the same, traditionally, most arguments for
relativism have proceeded from an anti-realist premise. I will conclude my arguments
in this essay by explaining why it seems to me that the very same picture of anti-realism
that seems to afford the possibility of metaphysically irresoluble disagreements is better
conceived as affording the possibility of normative insularity instead, and therewith,
relativism conceived as multimundialism.
2. Why the Relative Truth Intuition Fails

to Save the Disagreement Intuition
If I say that rhubarb is delicious and Wright denies it, we seem to disagree. It also seems
that each of us is within our rights to stick to our own view of the matter, even after we
have learned of the other’s apparently opposed view. This isn’t because each regards
the other as epistemically inferior in some way; it is rather because we think that there is
no one set of gustatory standards to which all parties must answer in order to have true
views about what is delicious. Wright need only answer to his personal gustatory
standards, and I need only answer to mine. But then the logical difficulty confronting
the Disagreement Intuition looms: if we have properly answered to our respective
gustatory standards, we have both made true claims, and this appears to violate the law
of non-contradiction.
Semantic relativists have proposed that we can avoid an outright violation of that law
by portraying the truth of our respective claims as being relative to our respective
gustatory standards. There is no outright contradiction if we say that the proposition
‘Rhubarb is delicious’ is true relative to Rovane’s gustatory standards while the
proposition ‘Rhubarb is not delicious’ is true relative to Wright’s gustatory standards.4
It has been often objected that this proposal is inadequate because it fails to preserve a
sense of genuine disagreement between Wright and me. And it is sometimes objected as
well that the proposal invites us to think that Wright and I actually agree—the thought
here is that if he and I came to see what is right about the proposal, then he and I could
agree that our respective claims are true relative to our respective gustatory standards.
Some semantic relativists think it is a sufficient response to these objections to point out
that Wright and I have been portrayed as affirming and denying the same proposition. In
their view, even if we can be said to agree, we would really be agreeing to disagree.
As we evaluate this response, we should reflect on an important contrast between
two kinds of case. In some cases, parties who disagree may have good reason to believe
that neither of them will ever be able to convince the other, and this may lead them to
conclude that there is no point in continuing to argue about the matter over which
they disagree, even though they both agree that they cannot both be right. This case
4
The central work here is Kölbel 2002.
fully deserves to be classified as one in which the parties agree to disagree. But disputes
of inclination are supposed to be quite different. In such disputes, what the parties agree
on isn’t only that they can’t convince each other, but also that it wouldn’t really be
appropriate to try to do so, for they already believe that they are both right (or at least
can be). So although it could fairly be said that they too, like the parties in the case I just
considered, would conclude that there is no point in continuing to argue about the
matter in dispute, it would be for a quite different reason—they would agree that there
isn’t anything for them to continue arguing about. For some of us, this very feature of
disputes of inclination—and all other apparent disagreements that are alleged to give
rise to relativism—suffices to disqualify them as disagreements at all. This point will
figure centrally in my own arguments to come, for the multimundial view.
The point is further supported by the fact that the metaphysical scenarios that are
envisaged in connection with disputes of inclination can be fully captured without
portraying the parties as affirming and denying the same proposition. In the example at
hand, instead of portraying Wright and me as making contrary claims which are true
relative to our respective gustatory standards, we might just as well suppose that we
don’t really mean the same thing by the word “delicious”—that what I am claiming is
that rhubarb is delicious-by-the-lights-of-Rovane’s-gustatory standards, while what Wright is
claiming is that rhubarb is not delicious-by-the-lights-of-Wright’s-gustatory-standards. This
portrayal makes completely clear that there is no real disagreement here. And yet it
does not call into question any of the metaphysical assumptions that drive the semantic
relativist’s portrayal. On both of these portrayals, Wright’s and my claims must answer
to our respective gustatory standards in order to be true, and there are no further facts
to which those standards must themselves answer.
This way of removing the appearance of a disagreement is sometimes called “con-
textualism.” It has also been compared to “indexical relativism”—a form of relativism
on which it isn’t only the truth-values of utterances that are relative to context but also
their meanings, as we find in the case of utterances involving indexical pronouns such
as “I,” “here,” and “now” whose reference varies with context of use. Most contem-
porary philosophers who subscribe to the prevailing consensus view of relativism reject
such contextualist and indexical relativist portrayals of what is going on in disputes of
inclination. There are three main grounds on which they do this: linguistic, practical,
and metaphysical. None of these grounds is entirely compelling.
Linguistic data have been cited to show that ordinary speakers do not regard words
like the English word “delicious” as having meanings that vary with context. Yet even if
we take these linguistic data at face value, they are hardly decisive. As Wright has pointed
out, the interpretive approach to meaning that has been advocated by Davidson and
others supports the idea that meaning is always highly sensitive to and variable with
context.5 Furthermore, linguistic intuitions often reflect commonsense assumptions that
5
For a complete account of the reasons in favor of this approach to meaning, as well as its implications, see
Bilgrami 1992.
cannot withstand philosophical scrutiny, and when those assumptions are corrected that
will mandate changes in our standing linguistic practices; and if those changes are
implemented, then presumably, the intuitions of ordinary speakers will change. In the
example under discussion, when Wright and I initially confront our dispute over rhubarb
we will most likely presume that we are using the word “delicious” with the same
meaning. But if we then go on to explore our differences, and come to agree that both of
our claims can be true because they answer to different gustatory standards, then it will be
appropriate for us to conclude that we never really had been talking about the same thing
even though it initially seemed to us that we were. Wright had been talking about
whether rhubarb is delicious-by-the-lights-of-Wright’s-gustatory-standards, while I had
been talking about what is delicious-by-the-lights-of-Rovane’s-gustatory-standards. To
view our ‘dispute’ in this way is to allow that we had been wrong to view ourselves to as
having a genuine disagreement.
Wright still sees a disagreement here, partly on the ground that there is an evident
practical conflict between us.6 If I find rhubarb delicious, then I have reason to eat
rhubarb, and also to buy it and cook it, and if I am a gardener, to cultivate it. If Wright
doesn’t find it delicious, then he doesn’t have the same reason to do these things; other
things being equal, he will not eat, buy, cook, or cultivate rhubarb. Furthermore, if he
really hates rhubarb, and if we happen to share a kitchen and a garden, he may actually
oppose my cooking and cultivating it. So it is clear that our gustatory attitudes may put
us into direct practical conflict. It is also clear that if we are in practical conflict, then we
are in disagreement about something, namely, whether to do (or allow to be done) the
very actions over which we are in conflict. But it is not clear that the practical conflict
between Wright and me over rhubarb presupposes that he and I are affirming and
denying the same proposition when I say it is delicious and he says that it isn’t. Two
agents can have reason to carry out (or promote) opposed courses of action simply
because they are pursuing different values that don’t themselves logically conflict.
Suppose, for example, that I decide to grow rhubarb because I want to eat food that
I grow myself and I’ve learned that it is easy to grow rhubarb, while Wright decides not
to grow rhubarb because he wishes to grow a cash crop and rhubarb doesn’t pay. Here
Wright and I have decided to carry out opposed courses of action, and yet there is no
direct logical conflict between our background evaluative attitudes. That is, no single
proposition is being affirmed and denied when I claim that it is good to grow rhubarb
because it is easy to grow, and Wright claims that it is not good to grow rhubarb
because it is not a cash crop. And it seems to me that the case of our gustatory attitudes
is well understood along these lines. If I wish to pursue what is delicious-by-the-lights-
of-Rovane’s-gustatory-standards, this may lead me to carry out courses of action that
are in some sense “opposed” to Wright’s, when he pursues what is delicious-by-the-
lights-of-Wright’s-gustatory-standards. But it doesn’t follow that we ever affirm and
6
Wright 2006.
deny a single proposition concerning whether rhubarb is delicious in a univocal sense.

Those who are impressed by linguistic data may want to cry foul here. How can I say
that the case is well understood along these lines given that ordinary speakers generally
take words like “delicious” to have a univocal and constant meaning across contexts?
My answer is that if we do not disambiguate the term, and if we also allow that Wright
and I can both be right, then we shall have a violation of the law of non-contradiction.
In addition, we shall obscure the metaphysical facts as they are being assumed here for
the sake of argument in the example at hand—the fact that our respective claims must
answer to different gustatory standards in order to be true.
This brings us to the main metaphysical ground on which it is most commonly
insisted that we should try to preserve the appearance of disagreement in disputes of
inclination, and in other similar cases that we intuitively associate with the doctrine of
relativism: we should do so simply in order to capture the doctrine’s intuitive content
when it is conceived in accord with the Disagreement Intuition. We’ve seen that the
semantic relativist’s strategy, as standardly applied, fails on this score. An important
refinement of that strategy has been offered that claims to do better—namely, “assess-
ment relativism.”7
The refined strategy proceeds from a particular diagnosis of why the more standard
strategy fails to preserve a genuine sense of disagreement in disputes of inclination: the
standard way of portraying the truth of the claims involved as relative leaves the
disputants no ground on which to assess one another’s claims as false, and hence
mistaken. This, of course, is exactly what we have seen in the example under
discussion: once I come to view the truth of Wright’s claim that rhubarb is not
delicious as relative to his gustatory standards, then I must assess it as true and I have
no ground on which to assess it as false or mistaken. But according to the refined
strategy, I do still possess a basis on which I could in principle assess it as false, for it is
false by the lights of my own gustatory standards. So what seems to be needed is a way
of portraying truth as relative that does not deprive me of this basis.
Accordingly, the refined strategy suggests that we portray truth as relative not to the
context that a party occupies when she makes a claim—her context of “use”—but as
relative to the context that a party occupies when she assesses the truth of a claim—her
context of assessment. The suggestion is that if claims about deliciousness are true
relative to contexts of assessment rather than to contexts of use, this will direct Wright
and me to assess each other’s claims by the lights of our own gustatory standards, and
hence as false, thereby preserving our sense of being in disagreement. At the same time,
we can also view our disagreement as relativism-inducing, because we can allow that
there is a sense in which our respective claims are true, insofar as they are true relative to
our own contexts of assessment.
7
See MacFarlane 2005, 2007, 2008. For further discussion of both his and Kolbel’s versions of the
semantic relativist strategy see Rovane 2010a.
It seems to me that assessment relativism may provide a formally adequate way of

capturing the Disagreement Intuition. But it cannot be deemed wholly adequate
from a philosophical point of view unless it provides for a way in which we could live
relativism in accord with that intuition. In order to do this, it must enable parties who
face disputes of inclination to attain a coherent view of their situation, including a
coherent stance toward one another. Assessment relativism does not provide this. It
instructs Wright and me each to allow that there is a sense in which the other’s claim
is true, because it is true by the lights of the gustatory standards that the other
employs—both as guides to gustatory evaluation and as guides to assessing the
truth of claims about gustatory value. The question is, why should either of us regard
it as appropriate to assess the other’s claim as false as well? I can see why it might seem
appropriate to answer this question in the way suggested by assessment relativism:
surely, each of us ought to assess the truth of gustatory claims on the basis of gustatory
standards that we actually embrace, and if Wright and I were to do that, then each of us
would of course assess the other’s claim as false. However, such an assessment would
fail to take into account the fact (assumed in the example) that each of us believes that
the other is not actually answerable to the gustatory standards that we ourselves
embrace. I want to insist that leaving this out of account would be a breach of
rationality.
The requirement of rationality I have in mind here is rather like the requirement to
arrive at and act upon all-things-considered judgments in practical deliberation. Of
course, this is a requirement that we rarely if ever live up to—indeed perhaps we never
can live up to it. All the same, it seems to me that the requirement still functions as a
regulative ideal, and that whenever we deliberate in its light, we recognize that we
ought to take into account all of the saliently relevant information that we possess
before deciding upon a course of action. I think there is a parallel requirement or ideal
that governs our assessments of truth, which instructs us to assess the truth of any given
claim in the light of all of the saliently relevant information that we possess. But when
I assess the truth of Wright’s claim about rhubarb in the way that assessment relativism
recommends, I am not doing this, and deliberately so. Assessment relativism explicitly
instructs me to assess his claim on the basis of my own gustatory standards, and to leave
out of account my belief that Wright answers to different gustatory standards from
mine. But if I aim to do the rational thing here, I must take that belief into account,
along with all of my other relevant beliefs. And then I will have a positive basis of my
own to assess his claim as true, and not just to allow that he has such a basis; and
furthermore, I will no longer have any basis on which to assess his claim as false, or to
view myself as having a genuine disagreement with him.
So once again, it is hard to retain a sense of disagreement in disputes of inclination
when we grant that both parties are right. But what I really want to emphasize here, is
that when assessment relativism tries to offer a way of retaining a sense of disagree-
ment, it asks us to transgress a very basic principle of rationality (viz., the ‘all things
considered’ requirement upon assessment of the truth of claims), and so it doesn’t
ultimately afford a coherent stance for the relativist to take toward allegedly relativism-
inducing disagreements.
Before moving on, let me relate this complaint about assessment relativism to the
dilemma for alternativeness. Assessment relativism aims to get round the difficulty
presented by the dilemma’s first horn—that inconsistent claims cannot both be true—
by stipulating that there is no context of assessment relative to which inconsistent
claims are both true. If we assume that I am bound, by my commitment to my own
gustatory standards, to assessing Wright’s claim that rhubarb is not delicious as false,
then the only way in which this aim can be realized is by portraying my context of
assessment in such a way that I cannot also, at the same time, acknowledge the basis on
which Wright’s claim is appropriately assessed as true. My complaint takes for granted
that I can acknowledge this. I believe I am right to take this for granted, since otherwise
we would be conceiving my situation in a way that would deprive me of any insight
into why my (alleged) disagreement with Wright threatens to be relativism-inducing—
it would deprive me of being able to live relativism. However, my complaint doesn’t
recommend that I wallow in contradiction—that I conclude that Wright’s claim is
appropriately assessed both as false and as true. The recommendation is to think
through the implications of all of my relevant beliefs concerning matters of gustatory
value in order to find the best way out of the threatened contradiction. When we think
them through, we find little reason to assess Wright’s claim on the basis of my gustatory
standards.
It is of course a peculiar thing to have gustatory standards, and to assess the truth of
one’s own claims concerning matters of gustatory value in accord with them, and yet
not appeal to those very standards when one assesses the truth of claims concerning
matters of gustatory value when they are lodged by others. Or to put it another way, it
is peculiar to assess the truth of others’ claims on the basis of standards that one does not
oneself embrace. But it seems to me that this just is the peculiarity of the relativist’s
stance. And it isn’t an incoherent stance unless it is incoherent to suppose that such
standards are not necessarily universal.
At this stage of the dialectic, it is open to opponents of relativism to respond by
urging that, for all I have said so far, my proposed way of conceiving the example at
hand is compatible with their view of it. They want to say that Wright and I either
disagree or agree. And they want to bring to bear the very same reasoning that we find
in the dilemma for alternativeness: on the one hand, if we disagree, then our claims
about rhubarb logically conflict and can’t both be right, and furthermore, if our claims
are entailed by our respective gustatory standards then those standards also logically
conflict and can’t both be right; on the other hand, if our claims are both right, then
they can’t logically conflict, and so they must be interpretable in such a way as to
remove the appearance of logical conflict—as I proposed when I proposed disambig-
uating the term “delicious.” The point of contention between me and opponents of
relativism concerns what follows from the very last step in this chain of reasoning—that
Wright and I aren’t really lodging contradictory claims. Opponents of relativism take
this to indicate that we really agree, while I deny that this is so. However, I haven’t yet
offered any reason for denying it (though I did hint at a reason when I raised the
possibility that I might assess the truth of Wright’s claim about rhubarb on the basis of
gustatory standards that I do not myself embrace). All I claim to have accomplished in
this section is to drive home how difficult it is to elaborate the Disagreement Intuition
in such a way as to afford an adequate philosophical account of what is going on in the
sorts of cases that are supposed to give rise to relativism-inducing disagreements. When
we invoke the Relative Truth Intuition in order to avoid the threat of contradiction
that follows upon the Disagreement Intuition, we generally find ourselves unable to
preserve a sense of genuine disagreement.
3. Why Being in a Quandary is Not Being a Relativist

As I’ve said, Wright also rejects the semantic relativists’ strategies for saving the
Disagreement Intuition, and his discussions of these matters are the most detailed and
thorough and convincing that I have read.8 Yet he continues to regard the Disagree-
ment Intuition as the intuition about relativism that is worth trying to elaborate. In his
view, if there is no coherent way to make sense of relativism-inducing disagreements,
then there is no coherent way to formulate the doctrine of relativism.
Wright’s interest in arriving at such a formulation goes together with an indepen-
dent and more longstanding interest in providing a satisfactory account of anti-realism.
This isn’t surprising, given the natural thought I articulated at the outset about how
anti-realism might give rise to relativism-inducing disagreements. Let me repeat it: in
domains where anti-realism holds, the truth is whatever ideally informed subjects
would be warranted in asserting, and it follows that if two such ideally informed
subjects were to disagree, then they would both be right.
Because this natural thought so obviously brings contradiction in train, Wright
doesn’t endorse it as it stands. Instead of directly asserting that both parties to a
disagreement can be right, he has suggested instead that a disagreement can be
“faultless,” by which he means that there aren’t any facts in the light of which either
party can be said to be guilty of any cognitive shortcoming. Intuitively, I would have
thought that in domains where anti-realism holds, to be free of all cognitive short-
comings would be tantamount to being right. But that leads straight back to the
contradiction that is so evident in the crude picture. Rather than rest there, Wright
has persisted in investigating ways in which we might avoid this contradiction, taking
intuitionist mathematics as his model. This approach seems highly sensible since
intuitionists and anti-realists alike place epistemic constraints on truth. Furthermore,
intuitionists have revised classical logic with the express aim of avoiding certain contra-
dictions that can be derived within classical mathematics via the method of reductio ad
8
See especially Wright 2007, 2008.
absurdum, and this raises the question whether we might be able to avoid certain
contradictions that can be derived within the framework of anti-realism by confining
ourselves to intuitionistically valid inferences.9
It lies beyond my concerns in this essay to address the many formal issues and
technical details that arise in connection with Wright’s appeal to intuitionism. I want
to register just two points. First, he himself has shown that his early account of anti-
realism, which affords the possibility of “faultless” disagreements, does entail contra-
dictions, and not all of them can be avoided by confining ourselves to intuitionistically
valid forms of inference.10 Second, his subsequent efforts to avoid such contradictions
by invoking the quasi-technical idea of a Quandary are subject to the usual disjunctive
objection: either they fail to do justice to the Disagreement Intuition, or they fail to
avoid all contradiction. Let me explain.
Wright defines Quandaries as follows:
P presents a Quandary for a thinker T just when the following conditions are met:
(i) T does not know whether or not P

(ii) T does not know of any way of knowing whether or not P
(iii) T does not know that there is any way of knowing whether or not P
(iv) T does not know that it is (metaphysically) possible to know whether or not P
Note that the satisfaction of each of those four conditions would be entailed if a fifth condition
were satisfied:
(v) T knows that it is (metaphysically) impossible to know whether or not P.
But that is not to be feature of a Quandary as I intend the notion to be understood; rather, in a
Quandary we have uncertainty through and through; so we do not even have the certainty of
undecidability. Note also that it is important that the notion of possibility appealed to is of the very
weakest—metaphysical—genre.11
It would appear that a Quandary is as close as logic will permit us to get to affirming a
contradiction without actually doing so. For if we cannot affirm both halves of a
contradiction as true, the next closest thing would be to keep an open mind about
both—that is, to leave it open that each might be true even though they cannot be true
together. And when we face a Quandary we face a prospect of un-ending open-
mindedness, even though we don’t know for sure that the matter about which we are
in a Quandary can never be resolved.
There is also one crucial respect in which facing a Quandary is very much like facing
the sort of disagreement that is alleged to give rise to relativism: when we believe that
9
Wright spells out his account of anti-realism in Wright 1992. He has explored the implications of his
account for relativism in a number of papers that I have cited or will cite in the course of this essay.
10
Wright 2006.
11
Wright 2001.
we face a Quandary concerning whether P or not-P, we have no idea about what

course of inquiry we could possibly follow in order to resolve the question whether
P or not-P; and arguably, the same is true when two parties believe that they face a
relativism-inducing disagreement—they too have no idea about what course of inqui-
ry they could possibly follow in order to resolve their disagreement. Thus, when
I claim that rhubarb is delicious and Wright claims that it isn’t, it is not as though we
can begin to argue about standards of evidence, or experiments that we might conduct,
in order to discover some sort of fact in the light of which one or the other (or both) of
us might be mistaken. Indeed, Wright himself reports real perplexity when he tries to
envisage what sort of property or fact there could possibly be that would, if it existed,
provide a ground for resolving our dispute.12 Insofar as we share Wright’s metaphysical
perplexity, we may be inclined to infer that disputes of inclination really do satisfy all
the conditions of a Quandary, including even condition (iv) which says that when we
face a Quandary concerning P or not-P, we do not know whether it is metaphysically
possible to know whether P or not-P. For if we cannot so much as envisage what sort
of property or fact would, if it existed, settle whether P or not-P, it would seem to
follow that we do not know whether it is metaphysically possible for us to know
whether P or not-P.
Yet it is precisely with condition (iv) that disanalogies between Quandaries and
disputes of inclination loom. It is very much part of the common understanding of why
disputes of inclination might constitute relativism-inducing disagreements that the
parties involved have some standing on which to claim knowledge. Perhaps philoso-
phers with extremely high standards for knowledge would want to deny this. But even
they should concede that the epistemic situation is quite close to knowledge. The
parties in disputes of inclination are supposed to be within their rights to retain
confidence in their own beliefs even after they have encountered their dispute. And
this isn’t supposed to be because each regards the other as being in an epistemically
inferior position. Each is supposed to regard the other as “cognitively blameless.” So
the curious thing about disputes of inclination is the entitlement to persist in one’s own
belief in the face of the well-grounded beliefs of others who apparently disagree with
one. We couldn’t possibly have such an entitlement without backing from an anti-
realist metaphysics according to which there are no further facts, beyond those that are
already at the disposal of the disputing parties, to which they are answerable. And
surely, anti-realism supports the idea that it most definitely is metaphysically possible to
attain knowledge in the matter under dispute—which clearly speaks against the idea
that we satisfy condition (iv), or indeed have any of the kinds of uncertainty that
characterize a Quandary.
It might be suggested that the parties in disputes of inclination ought to withdraw
their claims to knowledge simply on the ground that their claims contradict each other
12
Wright 2006.
and therefore cannot both be true. If we were to accept this suggestion, perhaps we
might also find a compelling ground on which to view the parties in disputes of
inclination as faced with Quandaries. But I want to insist that when we view disputes of
inclination in this way, we are not regarding them as disagreements of the sort that
intuitively give rise to relativism. We are regarding them as ordinary disagreements in
which it is assumed that both parties cannot be right. This means that there is something
for them to resolve, however ill equipped they may be to do so. In fact, it seems to me
that Quandaries are Quandaries precisely because they present us with something to be
resolved. The idea that we may never be able to resolve them and, moreover, that we
don’t even know whether it is metaphysically possible for us to resolve them, doesn’t
speak at all against that. Yet that is precisely what is supposed to be missing in disputes of
inclination and other cases of alleged disagreement that are supposed to give rise to
relativism. In such cases, there isn’t supposed to be anything to be resolved.
Wright is very explicit on the point that he intends the uncertainty of a Quandary to
be thoroughgoing. So perhaps he intends that our uncertainty should extend to this last
matter, concerning whether or not disputes of inclination present us with anything to
be resolved. I just observed that this attitude would not be in keeping with the way in
which such disputes are typically viewed by would-be relativists, who adopt an anti-
realist view in the light of which they think it is clear that such disputes don’t present us
with anything to be resolved. But in any case, the troubling issue before us is this: the
very idea that such disputes do not present us with anything to be resolved is bound up
with the idea that there are no further facts to which the parties are answerable, from
which it follows that both parties can be right—an apparent exception to the law of
non-contradiction. So it would appear that to be in a Quandary about whether or not
disputes of inclination present us with anything to be resolved is to be in a Quandary
about whether there can be exceptions to the law of non-contradiction—and the
threat of contradiction that attends the Disagreement Intuition hasn’t been avoided.
Also note: the logical offense here cannot be avoided by saying that only the possibility
of a contradiction has been allowed, and that no actual contradiction has been affirmed.
On the usual modal reading of the law of non-contradiction as a necessary truth, it rules
out the possibility of contradiction.
The upshot of this discussion is that the conviction which I asserted at the outset in
my introductory remarks looks to be vindicated. Whenever we try to preserve the
Disagreement Intuition without falling into contradiction, we get one of the following
results: either we fail to preserve a sense of disagreement in the cases that are alleged to
be relativism-inducing, as we found in connection with the semantic relativists’
strategies; or we end up portraying those cases as ordinary disagreements in which
both parties cannot be right, as we found above when we portrayed disputes of
inclination as presenting direct Quandaries concerning the matter under dispute.
Correlatively, whenever we manage to do full justice to the Disagreement Intuition,
we end up saddling the relativist with having to allow that there can be exceptions to
the law of non-contradiction.
Quite apart from the threat of contradiction, I myself think that the gravest threat to
the Disagreement Intuition is the one I first raised in response to the standard strategy of
the semantic relativist. We can always easily remove the appearance of disagreement in
disputes of inclination and other allegedly relativism-inducing disagreements where the
parties agree that they can both be right, since all we need to do is to disambiguate
terms in such a way as to clarify that their respective claims are not really answerable to
the same facts. (I didn’t there put the point in terms of “answering to facts” but rather in
terms of “meeting standards.” But that is a harmless difference since, in the example
under discussion, the only ‘facts’ at issue concern what standards of gustatory value
Wright’s and my claims about rhubarb would have to meet in order to be true.)
Although the availability of such a portrayal does pose a threat to the Disagreement
Intuition, it doesn’t necessarily preclude a relativist understanding of the example.
Speaking for myself at any rate, I don’t lose my impression that what the relativist
intuitively has in mind is still at work in the example, even after the appearance of a
genuine disagreement has been removed. And I don’t think we should give up this way
of seeing the example. After all, the two portrayals—the one that preserves the
appearance of disagreement and the one that doesn’t—are metaphysically equivalent.
They make the very same anti-realist metaphysical assumptions about what sorts of
facts do and don’t exist, and what is required in order that Wright’s and my respective
claims be true.
Many philosophers—both would-be relativists and opponents of relativism—
assume that when we remove the appearance of disagreement in this example, we
must lose our impression of relativism because we will have portrayed the parties in the
apparent dispute as actually agreeing. But it seems distinctly odd to suppose that the case
must be one of agreement since, as I just pointed out, we are still assuming the same
metaphysical facts in the case that seem to many give rise to a relativism-inducing
disagreement. It shouldn’t be surprising, then, if on closer inspection we find that it is
not really a case of agreement at all.
When parties agree, they hold exactly the same beliefs. Whereas, in the context of
my dispute with Wright, when I concede that he may be right to say that rhubarb is
delicious by-the-the-lights-Wright’s-gustatory-standards, I do not thereby come to
hold his evaluative belief. I can see why it may seem otherwise, since I myself can truly
say that rhubarb is delicious by-the-the-lights-of-Wright’s-gustatory-standards. But if
I really did share his evaluative belief, then I would take it as a basis for my own
deliberations about whether to do such things as buy, cook, grow, and eat rhubarb.
And as we are now imagining the example, I do not do this. That is why Wright could
correctly point out that our respective attitudes lead us to carry out opposed courses of
action. I argued above that this doesn’t suffice to show that he and I disagree about
whether rhubarb is delicious in a univocal sense, since we might end up pursuing
opposed courses of action even when we are acting on the basis of attitudes that don’t
logically conflict, such as the belief that rhubarb is easy to grow and the belief that it is
not a cash crop. (As an aside, I might add that we could end up in practical conflict
even if we were in perfect agreement about the gustatory value of rhubarb. If we both
agreed, on the basis of the very same evaluative beliefs, that it is delicious in the very
same sense, our agreement might well lead us to want to eat the very same rhubarb. Of
course, these are not opposed courses of action in the sense Wright intended, but
another kind of practical conflict. My point above was that opposed courses of course
of action need not indicate the presence of an underlying evaluative conflict in a
logical sense, as distinct from mere difference. I’m raising this additional point about
practical conflict in order to drive home that we ought not to conflate issues of logical
conflict vs. accord on the one hand, and issues of practical conflict vs. accord on the
other.) It may seem that I am stretching a point here, especially since linguistic
intuition is so firmly on the side of the idea that Wright and I really do disagree.
But I’ve said all that I have to say against wisdom of portraying us that way if we want
also to allow that we can both be right. I take the intuition that we can both be right as
the dominant one when we are concerned to characterize the peculiar stance of the
relativist. And my present point is that to say that we are both right is not quite the
same thing as saying we agree. If we agreed, we would take the very same attitudes as
bases for our respective deliberations. But Wright and I do not each come to adopt the
other’s beliefs about the gustatory value of rhubarb when each allows that the other is
right. Where rhubarb is concerned, Wright does not deliberate from the belief that
rhubarb is delicious-by-the-lights-of-Rovane’s-gustatory-standards. The only sense in
which he “believes” that rhubarb is delicious-by-the-lights-of-Rovane’s-gustatory-
standards is that he takes this to explain Rovane’s behavior; otherwise, the basis of his
own evaluative responses to rhubarb is confined to his belief that it is not delicious-by-
the-lights-of-Wright’s-gustatory-standards. (Parallel remarks of course apply to my
beliefs about rhubarb.)
It might be asked: since neither of us is prepared to evaluate rhubarb on the basis of
the other’s belief, why isn’t it correct to say that each of us rejects the other’s belief ?
That is certainly one way to read the example. But if what we mean by rejecting the
other’s belief is dismissing it as false, we would be giving up on the relativist intuition
that both parties can be right, and we would be moving to a non-relativist picture of
the situation being envisaged, on which Wright and I have an ordinary disagreement
in which it is presumed that we cannot both be right. My concern here is to
make sense of how we can both be right, without thereby coming to construe our
situation as one of ordinary agreement, any more than we construe it as one of ordinary
disagreement.
The next section will pave the way for such a relativist reading of the example, on
which Wright and I neither disagree nor agree—a reading on which our respective
beliefs, when true, qualify as alternatives. We’ll see that encountering alternatives would
be nothing like facing a Quandary, though I think this should already be apparent from
what I have said in this section.
4. Alternatives
At the outset I defined alternatives as truths that cannot be embraced together. I went
on to raise a dilemma for the very idea of alternativeness, which assumes that every pair
of truth-value-bearers must be either inconsistent or consistent, and then goes on to
reason as follows: the inconsistent ones can’t ever both be true while the consistent
ones can always be embraced together, and so in neither case do they qualify as
alternatives as I’ve defined them. I also noted that if there is any room at all for
alternatives, then there must be a third possibility that the dilemma overlooks and
that there is only one such possibility to considered, which is that in some cases truth-
value-bearers are neither inconsistent nor consistent. I went on to make a few
preliminary remarks about what relativism amounts to when it is formulated as carrying
a commitment to the existence, or perhaps more weakly, the possibility, of alternatives.
Let me now elaborate.
The first thing to appreciate is that when truth-value-bearers are neither inconsistent
nor consistent, they fail to stand in any logical relations at all. It follows that the primary
dividing issue between relativists and their opponents concerns whether logical rela-
tions run everywhere, among all truth-value-bearers. Opponents of relativism insist
that they do, and this brings in train a metaphysical commitment that relativists reject. If
logical relations run everywhere, among all truth-value-bearers, then all of the true
ones must be consistent and conjoinable, which means that there is a single, consistent,
and complete body of truths. This amounts to a metaphysical commitment to the
oneness of the world, or unimundialism. In contrast, relativists deny that there is a single,
consistent, and complete body of truths, and they affirm instead that there are many,
incomplete bodies of truths that cannot be conjoined; in other words, they hold that
there are many worlds rather than one, or multimundialism.
I hope that no one is misled into thinking that the notions of completeness and
incompleteness that figure in this characterization of unimundialism vs. multimundi-
alism have anything to do with completeness and incompleteness in the logicians’
sense. The dividing issue between relativists and their opponents, as I’m construing it, is
not whether all of the theorems of a given formal language are also tautologies. The
issue concerns, as I’ve said, the oneness of reality. As l’ve noted, the philosophical
conception that is closest to this unimundial conception is the one that Wittgenstein
announced at the start of the Tractatus, when he declared that the world is all that is the
case—for I take his “all” to refer to the single, consistent, and complete body of truths.
I suspect that formally inclined philosophers will worry that this conception of a totality
of all of the truths is bound to be formally problematic. But we’ll see that the interest of
the idea doesn’t really hinge on its formal tractability. It is best thought of as a regulative
ideal that imposes a distinctive stance toward inquiry and interpersonal relations. The
practical significance of relativism lies in the fact that it rejects this unimundial stance in
favor of a multimundial stance which is not constrained by the unimundial ideal of a
single, consistent, and complete body of truths.
Let me first describe the unimundial stance. When we assume that logical relations
run everywhere, among all truth-value-bearers, we must acknowledge that these
logical relations impose normative constraints at every turn in our inquiries. Whenever
we come across a new truth-value-bearer that we haven’t entertained before, the
question automatically arises whether it is consistent or inconsistent with what we
already hold true. If we find that it is inconsistent with what we already hold true, then
we must choose among the following options: either we must reject it as false or we
must revise our standing beliefs so as to accommodate its truth, or we must suspend
judgment on it along with those of our standing beliefs with which it conflicts; if we
find that the truth-value-bearer is consistent with what we already hold to be true, then
it is a candidate for belief—though we would generally need additional warrant for
believing it, beyond its mere consistency with what we already believe.
Just as no one should be misled into thinking that my initial characterization
of unimundialism vs. multimundialism has anything to do with completeness and
incompleteness in the logicians’ sense, no one should be misled into thinking that
my characterization of the unimundial stance is bound up with the goal of arriving at a
complete description of the world. That problematic epistemic goal is not foisted on us
when we embrace the unimundial ideal of a single, consistent, and complete body of
truths. Qua regulative ideal, the unimundial ideal isn’t properly thought of as a goal at
all; its practical significance is more properly thought of in negative terms, as registering
normative constraints—the constraints I just described above.
These same normative constraints that govern a unimundialist’s approach to inquiry
will naturally carry over to interpersonal relations. If we are committed to unimundi-
alism, then whenever we encounter another person, we must always assume that her
beliefs are either inconsistent or consistent with our own. If they are inconsistent, then
either we must reject hers as false or we must revise ours in order to accommodate hers
as true, or we must suspend belief on the matters about which we are in conflict; and if
they are consistent, then they are candidates for belief by us. However, as in the case of
our own inquiries, we might require additional warrant for taking on another person’s
beliefs, over and above mere consistency with our own prior beliefs—whether this is so
will depend upon what sort of warrant we think is generally provided by the testimony
of others.
To sum up this brief account of the unimundialist’s distinctive stance toward
the world, inquiry, and interpersonal relations: as unimundialists, we view the world
as one; and precisely because we view the world as one, we view every truth-
value-bearer as something to be assessed together with the rest, and we view every
person as someone whom we might teach or from whom we might learn.
A word now about the distinctive character of the multimundialist’s stance. When
we are multimundialists, we hold that logical relations do not run everywhere,
among all truth-value-bearers. It follows that we might encounter a truth-value-
bearer that doesn’t stand in any logical relations to our current standing beliefs. This
means that it will be normatively insulated from our beliefs in the following additional
ways, over and above failing to stand in any logical relations to them: it wouldn’t
provide any basis for revising our standing beliefs in any way, and likewise, our
standing beliefs wouldn’t provide any basis either for rejecting it as false or for
embracing it as true. Thus, if we were to encounter an alternative in the context of
inquiry, we would be encountering a truth-value-bearer that could not be assessed
together with our own beliefs due to its normative insularity, with the result that it
could not be embraced together with them even if it was true. This means that to
encounter an alternative is to encounter a truth-value-bearer that lies entirely outside
the scope of our own inquiries.
Just as the normative constraints that unimundialism imposes on inquiry would carry
over to our interpersonal relations, so also the sort of normative insularity that follows
upon multimundialism has important implications for interpersonal relations. If there
are normatively insulated alternatives—truths that cannot be embraced together—it
follows that some truths may fail to be universal in the social sense that they would not
be truths for everyone. As multimundialists, we would have to allow that others may
have their truths which are not for us and vice versa. It also follows that there may be
persons whom we cannot teach and from we cannot learn—not on all matters, but on
those matters concerning which our beliefs are normatively insulated from theirs.
So to sum up what it would mean to adopt the relativist’s multimundial stance: we
would not view the world as one, in the sense of viewing all truths as assessable and
embraceable together from the same point of view; because this is so we would not
view all persons as inquirers into the same truths; we would allow rather that we may
have truths that are not truths for others, and that others may have truths that are not
truths for us.
By now it should be fully clear why encountering alternatives would be nothing like
facing a Quandary. Relativism would not induce any kind of uncertainty, let alone the
metaphysically deep uncertainty of a Quandary. It would induce something quite
else—a form of extreme epistemic indifference, which would be socially manifested as
an epistemic disengagement from others. (Note that this form of epistemic indifference
and disengagement is compatible with a high degree of curiosity and interest and even
concern for those others whose attitudes and inquiries are normatively insulated from
our own.)
Perhaps I have underestimated the advantages of Wright’s proposal. But let me
register, in any case, that my own proposal offers what I see as a highly sensible way to
re-understand why and how the situations that are generally—but in my view mistak-
enly—described as “disputes of inclination” may be relativism-inducing: they are occa-
sions on which it makes sense to adopt the multimundial stance. Going back to Wright
and me on rhubarb. We can suppose that he and I inhabit different gustatory worlds,
which are literally constituted by different matters of gustatory value—the one world by
what is delicious-by-the-lights-of-Wright’s-gustatory-standards and the other world
by what is delicious-by-the-lights-of-Rovane’s-gustatory-standards. If I should come
to learn about what is delicious-by-the-lights-of-Wright’s-gustatory-standards, he and
I can both see that this has absolutely no bearing on what is or can be true concerning
what is delicious-by-the-lights-of-Rovane’s-gustatory-standards. In other words, when
I want to learn about matters of gustatory value, no normative constraints on me are
forthcoming from Wright when he reports about what is delicious-by-the-lights-of-
Wright’s-gustatory-standards. No gustatory enlightenment is forthcoming from him
either, for whatever truths he knows lie entirely outside the proper scope of my own
gustatory inquiries, in the sense that they can never be a basis for curing my gustatory
ignorance or correcting my gustatory mistakes. And if he in turn is in any way ignorant
or mistaken about matters of gustatory value, nothing I know can provide a basis on
which to correct that. In short, we have nothing to teach or learn from each other,
because the truths with which we are concerned are not universal.
One may reasonably doubt whether there is really any scope for anyone to be wrong
concerning matters of gustatory value. If there isn’t, then such matters could not
occasion relativism as I have formulated it, for I take relativism to be concerned in
some way with the nature of truth, and where there is no possibility of error there is no
truth. The more interesting and serious candidates for relativism lie in the domains of
morals, politics, and artistic aesthetics. If relativism can arise in these domains, it is
because there are social contexts in which local issues arise, about which there can be
local agreement and disagreement and local error, where these social contexts are
circumscribed or closed off from one another so that what is true in one context places
no normative constraints on what can be true in the others. This needn’t mean that the
occupants of one context cannot learn about what is morally, politically, or aesthetically
true in other contexts. It would mean that such learning would have only detatched
explanatory value—it would belong to sociology, or anthropology, or psychology.
It wouldn’t provide moral, political, or aesthetic illumination, because it wouldn’t help
one to navigate the moral, political, or aesthetic choices that arise locally, within one’s
own social context. That is why one could learn about the truths in question without
thereby coming to embrace them together with one’s own—that is the sense in which
one could regard them as alternatives.
This may all seem very implausible. But if so, I think that just signals that the doctrine
of relativism itself may seem implausible. It was never part of my aim to formulate it in
such a way that it would emerge as plausible. On the contrary, I aimed to capture a
distinctive and controversial metaphysical commitment that we intuitively associate with
the doctrine of relativism.
It may be the Disagreement Intuition has become so entrenched that it has effec-
tively dislodged the Alternatives Intuition. In that case, it may be impossible for me to
convince my contemporaries that I have succeeded in my aim. But I can still claim
history to be on my side. My proposed formulation of the doctrine of relativism, as
carrying a commitment to the existence, or perhaps just the possibility, of alternatives,
constitutes a return to the main terms of debate about relativism that predominated in
the twentieth century. When Davidson famously argued against relativism in his
Presidential address to the APA his stated target was the “very idea” of a conceptual
scheme.13 But it was alternative conceptual schemes on which he mainly focused. He

specifically defined alternativeness as arising when different languages or theories are
true but not translatable, which is a reasonable approximation of what I mean by the term
“alternative.”14 As Davidson rightly saw, this was a primary commitment of relativists
who took their inspiration from Kuhn’s philosophy of science and from cultural
anthropology, who typically expressed it by saying that theories or languages are
incommensurable.15 Part of what made for such incommensurability was a failure to
share meanings, as when Kuhn insisted that there is no meaning-invariance across
scientific paradigms. But of course, if paradigms do not share meanings—if they are not
translatable—then they cannot logically conflict. So if they are somehow alternatives to
one another—if they cannot be embraced together—it can only be because they are
normatively insulated from one another. That, on my reading, is what Kuhn claimed
(though since he didn’t go so far as to suggest that distinct paradigms can be equally
true, he wasn’t a true relativist by my lights). A commitment to alternativeness is
also evident in two influential examples of relativism coming from the tradition of
logical empiricism, namely, Carnap’s linguistic frameworks and Goodman’s ways of
worldmaking.16 Finally, there is significant overlap between the idea of alternativeness
13
See Davidson 1984.
14
It might be wondered whether Davidson’s notion of alternativeness is really equivalent to mine.
Although he was explicit that alternatives are equally true, he was not explicit that they cannot be embraced
together. Yet this was implicit in his discussion. We can see why by considering what would follow if we
supposed that alternative conceptual schemes could be embraced together. In that case, there could in
principle be a single, more comprehensive language in which we could express them both. But Davidson was
clearly not arguing against this possibility. In fact, it would have been absurd for him to argue against the
possibility that different parts of a single language may be true but not translatable. Insofar as they differ in
content and are not logically false, it is bound to be the case that different parts of the same language will be
true but not translatable. The point I’m working up to here isn’t that different parts of the same language
actually satisfy Davidson’s criterion for alternativeness. He was very definite that he was referring to whole
languages that are true but not translatable. My point concerns what distinguishes these two cases of non-
translatable truths—the ones that are expressible only in different languages vs. the ones that are expressible in
different parts of a single language. I can see no other answer but that the former cannot be embraced together
while the latter can. If I am right, then the idea that alternatives cannot be embraced together is an unstated
presupposition of Davidson’s entire discussion.
15
It is worth registering that the early Feyerabend did not use the term “incommensurability” to mean
alternativeness in my sense. In Feyerabend 1965 he explicitly claimed that theories are incommensurable
when they employ conflicting principles. He went on to recommend that we proliferate incommensurable
theories in order to enhance our chances of exposing error, through decisive experiments that would give us
grounds for choosing between them. His primary aim in recommending such methodological pluralism was to
avoid dogmatism—uncritical acceptance of a theory merely because the theory is able to satisfy its own
standards. This recommendation simply wouldn’t make any sense unless he assumed that incommensurable
theories cannot be equally true, unlike alternatives in my sense. (N.B. We might doubt whether there can be
incommensurability in Feyerabend’s sense, insofar as he agrees with Kuhn that different scientific paradigms
don’t share meanings. For when there aren’t shared meanings it is unclear how there can be conflict, as
opposed to mere difference. I take it that the idea of a decisive experiment is supposed to clarify this in the
following way: holders of incommensurable theories would employ different terms to describe the same
experiment, but they could still agree that the results of the experiment would simultaneously confirm one of
their theories while infirming the other.)
16
The inspiration for relativism from Carnap comes from his insistence that certain questions can have
objective answers only when they are raised internally, within a given linguistic framework. See Carnap
1950. For Goodman on relativism, see Goodman 1978.
in my sense and Williams’s idea of a merely notional confrontation.17 Though I do think all
of these positions differ in one way or another from the particular characterization of
‘alternatives’ I have offered, and therefore from the conception of relativism that I have
said owes to the notion of alternatives, they all make implicit or explicit appeal to the
rough idea of alternatives that is to the Alternatives Intuition, as I have called it and not
the Disagreement Intuition or the Relative Truth Intuition.
5. Anti-Realism and Normative Insularity

I want to close by returning to the very rudimentary picture of anti-realism that seemed
to entail that there can be relativism-inducing disagreements in which both parties are
right. My responses to this picture have been normatively driven: we ought not to
countenance exceptions to the law of non-contradiction and, so, we ought to find a
better way to understand what is going on in the situations being envisaged. My
recommendation was to remove the appearance of contradiction and then do what we
can to preserve and make sense of our impression that the situations are of the sort that
the relativist intuitively has in mind—and I have since shown how we can do this by
invoking the idea of alternatives as arising with normative insularity, rather than logical
conflict
But suppose we were overwhelmingly impressed by the reasons for taking the
appearance of contradiction seriously in these cases, and concluded that we ought
not to accept my recommendation to remove it, but should instead try to accommodate it
in as coherent a way as we can manage. When I try to think through what would be
involved in attempting such an accommodation, I find that I am eventually brought
back to my proposal to posit normative insularity.
Currently, the only proposal I know of to accept exceptions to the law of non-
contradiction concentrates on the case of paradox.18 It may seem that relativism-
inducing disagreements are paradoxical, since they seem to embroil us in allowing
that contradictory claims can both be true. But in the case of paradox, a single subject,
with a single point of view, finds reason to accept both P and not-P; whereas in the case
of relativism-inducing disagreements there are two subjects neither of whom is willing
17
See Williams 1981. There is one crucial difference: he initially specified that a merely notional
confrontation would arise only when one system conflicts with another. Yet at no point in the rest of his
account did he appeal to the idea of conflict. He emphasized instead that it is impossible to “go over” from
one system of beliefs to another without “losing one’s grip on reality.” He also emphasized that it is either
impossible or pointless for us to try to rationally appraise the systems of belief to which we stand in merely
notional confrontation. These characterizations make clear that systems in notional confrontation would
exhibit the main feature of alternativeness, which is that they cannot be embraced together. It is less clear that
they would, or even can, exhibit the other feature, of being equally true. In fact, Williams studiously avoided
pronouncing on this matter at all. See Rovane 2008 for further discussion of his important contribution to
this topic.
18
I have in mind Graham Priest.
to accept both P and not-P—that is one of the necessary conditions for the presence
of a disagreement, over and above the presence of a contradiction. So I’m going to
set aside the case of paradox as too dissimilar, and concentrate instead on the specific
way in which the elementary picture of anti-realism seems to entail that there could
be exceptions to the law of non-contradiction, in order to try to imagine what
would be involved in trying find a coherent way to accommodate this apparent
implication.
According to that picture, the truth in domains where anti-realism holds is whatever
ideally informed subjects would be warranted in asserting, and we cannot rule it out as
metaphysically impossible that such subjects might be warranted in affirming and
denying the same proposition. But the supposition has never been that if such subjects
came to see that this was their situation, then each ought to embrace the other’s claim
and wallow in the resulting contradiction. The supposition is rather that they disagree.
This means that they must recognize the law of non-contradiction as operative
within their respective points of view—accepting its injunction against embracing true
claims together. Yet they would also have to arrive at a somewhat non-standard
understanding of why they should accept this. On our usual understanding, one reason
why we ought not to violate the law of non-contradiction is our interest in truth. The
law of non-contradiction tells us that at most one of a pair of inconsistent claims can be
true and, so, if we want to believe only truths then we mustn’t believe inconsistent
claims. In contrast, if we were to allow that inconsistent claims can be equally true, we
would need a different reason to refrain from embracing inconsistent claims together.
We might simply insist that there is always something absurd about embracing a
contradiction. I find it difficult to evaluate this claim. On my understanding, what is
absurd about embracing a contradiction is precisely the fact that we would be embrac-
ing two claims as true when they can’t both be true. Yet there may be room for the
thought that contradiction itself is absurd, regardless of its implications (or lack thereof )
about truth. My own first impulse is to reject this thought on the ground that it doesn’t
provide any explanation of why contradictions are absurd. But, on reflection, I think
this impulse should probably be resisted since I cannot conceive a neutral standpoint
from which to think about why we ought to avoid affirming contradictions that
doesn’t already incorporate, and indeed exploit, a commitment to avoid affirming
them, and this may suffice to show that we would be within our rights to insist that we
can and should retain the law of non-contradiction, even as we give up our usual
understanding of how contradiction relates to truth. If that is so, then perhaps we could
have a reason, albeit a non-standard one, to refrain from embracing inconsistent claims
together, even when we allow that both are true—as is supposed to be the case in
relativism-inducing disagreements.
However, this is not the end of the departures that we would have to make from our
usual understanding of the relevant logical matters, in order to accommodate this
supposition. Just as we would have to revise our understanding of how contradiction
relates to truth, likewise, we would have to revise our usual understanding of how
conjunction relates to truth. On our usual understanding, if two claims are true then
their conjunction is true and, because this is so, it is always permissible to conjoin true
claims. But if we were to allow exceptions to the law of non-contradiction, then
we would have to deny this. That is, if we were to allow that inconsistent claims can be
true, then we would have to deny that their conjunction is either true or permissible.
I suspect that there is more than one way in which we might formally accommodate
these required departures from our usual understanding of how the laws of non-
contradiction and conjunction relate to truth. Since I am not a logician, I cannot
identify them all, let alone assess them. But it seems safe to assume that the best strategy
would be to try to retain as much of our usual understanding as we can. As far as I can
see, the most natural way to do this would be through a principle of logical confinement.
Such a principle would confine the laws of logic so that they hold only within a given
epistemic context, in which a particular set of internally consistent normative standards
prevails. This would allow us to understand and apply ordinary laws of logic, such as
the laws of non-contradiction and conjunction, in the usual way within an epistemic
context, without allowing any exceptions to them there. The only change it would
require is that inferential practice would be confined within contexts, in the sense that
we could never draw inferences involving truth-value-bearers that belong to distinct
contexts.19
Aside from helping to solve some of the logical difficulties that would follow upon
allowing apparent exceptions to the law of non-contradiction, there is another point in
favor of such a principle of logical confinement. The principle is independently invited
by the very metaphysical commitments of the crude picture of anti-realism that invites
us to allow such exceptions. Consider again Wright and me on rhubarb. The reason
why we arrive at contrary judgments concerning whether or not rhubarb is delicious is
supposed to be that each of us employs and answers to a different set of gustatory
standards. These different standards are naturally thought of as constituting distinct
epistemic contexts that each of us occupies. Within our respective contexts, we can
apply ordinary laws of logic in the usual way, with the usual understanding of how they
relate to truth. For example, if I come across an inconsistency in my own gustatory
beliefs, I must infer that one (or more) of them is false. But when I encounter a claim of
Wright’s that is inconsistent with my beliefs, I shouldn’t draw the same inference;
I must leave it open that his claim may be true, because the law of non-contradiction is
confined in such a way that it doesn’t apply across contexts but only within them.
Once we have introduced the principle of logical confinement, we have at our
disposal a somewhat different account of why we might hold that we shouldn’t
embrace certain truths together in domains where anti-realism holds. It is not incon-
sistency per se that stands in the way. It is rather this: embracing those truths together
would amount to conjoining them in accord with the law of conjunction; but in
19
I have no doubt that logicians have already devised non-standard logics for this sort of purpose.
domains where strong anti-realism holds, the application of that law, and indeed all
laws of logic, should be confined within a given epistemic context, and should never be
applied across distinct contexts. Thus, given the principle of logical confinement, we
ought not to embrace truths drawn from distinct contexts together even if they are not
inconsistent. That is why it is not inconsistency per se that, on this account, stands in the
way of embracing certain truths together.
However, there is a deeper point yet to be made in connection with the principle of
logical confinement. I’ve formulated it as a principle that confines inferential practice to
epistemic contexts (in domains where anti-realism holds). I’ve granted, for the sake of
argument, that inferential practice might be so confined while logical relations like
consistency and inconsistency are not similarly confined. But we should doubt whether
this makes sense. For what more is there to the reality of logical relations than their normative
force, as that force is reflected in the laws of logic? More specifically, what more is there to the
reality of a logical relation like consistency than that it licenses and prohibits inferences
in accord with the law of conjunction? And what more is there to the reality of a logical
relation like inconsistency than that it licenses and prohibits inferences in accord with
the law of non-contradiction? I take the burden of these questions to be that there is no
metaphysical room for logical relations that carry no logical force. (N.B. This meta-
physical point leaves plenty of normative room for revisionist understandings of the
laws of logic, such as those proposed by intuitionists. All it requires is that we see logical
laws as registering the normative force of underlying logical relations, with the
implication that any revised understanding of those laws must bring in train a revised
understanding of those relations).
If I’m right about this deeper point, then we have the conclusion that I promised. If
we think through what would be involved in accepting the elementary picture of anti-
realism, and trying to accommodate its apparent implication that there can be relativ-
ism-inducing disagreements in which both parties are right, we find reasons to deny that
there is any real inconsistency present in the situations that are supposed to induce
relativism,, but rather normative insularity instead. In other words, we find reasons to
set aside the appearance of disagreement, and to charaterize the situation as involving
alternativeness in the sense that I have elucidated.
I repeat that my aim in this essay has not been to find grounds on which to accept or
reject relativism. I have merely tried to formulate the doctrine itself, so that we are clear
about what such grounds are supposed to speak for or against. My proposal is that we
should elucidate the doctrine as involving a commitment to alternativeness, because this
is a common commitment running through disparate accounts and conceptions of the
doctrine, and because it is one of the main targets of arguments against the doctrine,
which I have summed up in the form of a dilemma for alternativeness. It shouldn’t be
surprising—and the dilemma brings this out—that the primary challenge that we face
when we are trying to elucidate the doctrine of relativism is a logical one, the challenge
of making sense of what might stand in the way of embracing certain truths together.
I have suggested that normative insularity is what might stand in the way. Although the
suggestion successfully avoids the dilemma for alternativeness, it leaves us with formal
issues yet to be addressed, and it also leaves us with many open questions concerning
what sorts of metaphysical conditions might generate normative insularity. Neverthe-
less, it does succeed in identifying the fundamental dividing issue between relativists
and their opponents. It is not whether reality is mind-independent, but whether it is
one in the sense that is captured by the unimundial ideal of a single, consistent, and
complete body of truths. To construe relativism as rejecting this ideal, by affirming that
some truths are normatively insulated from one another, redeems the idea of alter-
natives in a way that I hope has enough freshness and depth to generate further
philosophical inquiry into its implications, so that we might, once those implications
are thought through, come to the point where the doctrine of relativism can eventually
be assessed.
Bibliography
Carnap, R. 1950 “Empiricism, Semantics and Ontology,” Revue Internationale de Philosophie 4,
pp. 20–40, reprinted in R. Carnap, 1956 Meaning and Necessity, Chicago, University of
Chicago Press, pp. 000–000.
Davidson, D. 1984 “On the Very Idea of a Conceptual Scheme,” in Inquiries into Truth and
Interpretation, Oxford, Oxford University Press, pp. 183–98.
Feyerabend, P. 1965 “Problems of Empiricism,” in R. Colodny (ed.) Beyond the Edge of Certainty,
Englewood Cliffs, N.J., Prentice-Hall, pp. 145–260.
Garcia-Carpintero, M. and Kölbel, M. (eds.) 2008 Relative Truth, Oxford, Oxford University
Press.
Goodman, N. 1978 Ways of Worldmaking, Cambridge, Mass., Hackett.
Kölbel, M. 2002 Truth Without Objectivity, London, Routledge.
MacFarlane, J. 2005 “Making Sense of Relative Truth,” Proceedings of the Aristotelian Society CV,
pp. 321–39.
MacFarlane, J. 2007 “Relativism and Disagreement,” Philosophical Studies 132, pp. 17–31.
MacFarlane, J. 2008 “Truth in the Garden of Forking Paths,” in M. Garcia-Carpintero and
M. Kölbel (eds.) Relative Truth, Oxford, Oxford University Press, pp. 81–102.
Rovane, C. 2002 “Earning the Right to Realism and Relativism in Ethics,” Philsophical Issues 12,
pp. 264–85.
Rovane, C. 2008 “Did Williams Find the Truth in Relativism?” in D. Callcut (ed.) Reading
Bernard Williams, London, Routledge, pp. 43–69.
Rovane, C. 2010a “Relativism Requires Alternatives, Not Disagreement or Relative Truth,” in
S. Hales (ed.) A Companion to Relativism, Oxford, Basil Blackwell, pp. 31–52.
Rovane, C. 2010b “Why Scientific Realism May Invite Relativism,” in M. De Caro and
D. MacArthur (eds.) Naturalism and Normativity, New York, Columbia University Press,
pp. 100–20.
Williams, B. 1981 “The Truth in Relativism,” in Moral Luck, Oxford, Oxford University Press,
pp. 132–43.
Wright, C. 2001 “On Being in a Quandary: Relativism, Vagueness and Indeterminacy,” Mind
110, pp. 45–98.
Wright, C. 2006 “Intuitionism, Realism, Relativism and Rhubarb,” in P. Greenough and
Wright, C. 2007 “New Age Relativism,” Philosophical Issues 17/1, pp. 262–83.
Wright, C. 2008 “Relativism about Truth Itself: Haphazard Thoughts about the Very Idea,” in
M. Garcia-Carpintero and M. Kölbel (eds.) Relative Truth, Oxford, Oxford University Press,
pp. 157–86.
SECTION IV
Warrant, Transmission Failure,
and Scepticism
10
When Warrant Transmits*
James Pryor
I
We can ask about doxastic warrant—which of your beliefs are reasonable, or episte-
mically appropriate?—and we can ask about prospective or propositional warrant—
what things do you have warrant to believe?
You have all things considered warrant to believe Q when Q is the right, or
epistemically appropriate, thing for you to believe overall. There should be a story
about why Q is the right thing to believe: what factors warrant or count in favor of
believing Q. As theorists we want to be able to discuss such factors, while allowing that
they may be opposed by other factors that warrant believing alternative hypotheses. Or
they may be undermined by information that makes them count less strongly in favor
of believing Q. In the vocabulary I prefer, we say that the factors at least constitute
prima facie warrant to believe Q—and leave it open to further discussion how
complex bodies of prima facie warrant should add up.
Sometimes your warrant to believe Q includes warrant to believe some premises
that entail or ampliatively support Q. For example, what warrants you in believing
An animal is in the pen may be some visual warrant to believe That’s a zebra in the pen.
Let W be your warrant to believe Q, the factors that count in favor of believing Q.
I’ll represent the “counting in favor of ” relation with a hollow arrow, thusly:
Q. An animal is in the pen.
* Thanks to audiences at UNAM, CUNY, Rochester, Indiana, Reed, Buffalo, Hong Kong, Sydney,
Canberra, Seoul, Boulder, St Andrews, and Brown; and especially to Martin Davies, Earl Conee, Sinan
Dogramaci, Yuval Avnur, Nico Silins, Matt Kotzen, Patrick Hawley, Paul Boghossian, and Crispin
Wright for their comments. (And one more generous interlocutor whose comments I can no longer
attach a name to.)
270 J A M E S P RYO R
In this case, W consists in visual warrant to believe a premise P that entails Q. We can spell
that out like this:
W. [Visual warrant to believe P] where P is: That’s a zebra in the pen

It will be useful to be able to switch easily between talking about the premise P, and
talking about the warrant W you have for Q by virtue of having certain warrant to
believe P. Sometimes this is done by overloading the interpretation of a label: using it
sometimes for warrants, other times for premises. For instance, we might write:
Z. That’s a zebra in the pen.

and alternately use “Z” to designate the premise P, or the warrant W. This shortcut can
ease discussion and is common in practice, so it will be useful to continue it here.
Context should always settle when we’re talking about a warrant, and when about a
proposition. Despite this terminological convention, though, we should not forget the
substantive difference between warrants and propositions. The warrant Z is not itself a
proposition or premise. Neither is the premise Z ever a warrant to believe Q: not even
when it entails Q.1 Rather, it’s the visual (or whatever) warrant you have for premise
Z (which we may also use “Z” to designate) that constitutes a warrant to believe Q.
On many views, warrants don’t always consist in warrants to believe supporting
premises. For example, what warrants you in believing It looks as though there’s a zebra in
the pen may just be your visual experiences, or perhaps the fact that you have those
experiences. Not any premises about those experiences, that you need warrant for
believing. Let’s label the warrant you have here E, and the hypothesis it supports
LOOKS. I’ll represent such warrant with brackets, like this:
E. [Visual experiences as of a zebra]
LOOKS. It looks as though there’s a zebra in the pen.

In such cases I’ll only use “E” to designate the warrant your experiences give you to
believe LOOKS, not to designate any proposition.
1
There is a view that denies this; see Silins 2005, }4.2 and his references to Klein in fn. 19.
W H E N WA R R A N T T R A N S M I T S 271
Sometimes we’ll be considering factors that warrant belief in a conclusion only

against the background of other warrants you possess. For example, on many views, your
visual experiences won’t be enough by themselves to warrant believing That’s a zebra in
the pen. But they will warrant that belief against the background of warrant R to believe
your vision is reliable. I’ll represent that like this:
Here E on its own isn’t warrant to believe Z: only R+E together count in favor of
Z. But R has more of a background role than E. So long as you do possess warrant R,
you might reasonably believe things on the basis of your experiences without ever
considering or having any beliefs about how reliable your senses are. That is, so long as
you have warrant to believe your senses are reliable, your perceptual beliefs might not
need to be based on that warrant. E might on its own be an adequate basis for believing Z.
As you’ll see, this is not my own view of perceptual warrant; but we want to have the
apparatus on hand to talk about it. Another epistemic relationship discussions some-
times attend to is when some facts help enable the presence of some warrant, without
themselves being the possession of a warrant. (see my discussion of “warrant-making
factors,” later.) I have not given this relationship a place in these diagrams.
II
Dretske says that reasoning like this:
Zebra-1. [Visual experiences as of a zebra]
Zebra-2. That’s a zebra in the pen.
This arrow represents not a “counting in favor of,” but a move in reasoning
Zebra-3. That’s not a mule cleverly disguised as a zebra.

fails; and he says that its failure illustrates a violation of Closure. That is, it may be that
premise Zebra-2 entails Zebra-3, and you have warrant (namely, Zebra-1) to believe
Zebra-2, without you having warrant to believe Zebra-3.
Wright is sympathetic to the claim that Zebra reasoning would be defective: you
couldn’t reasonably base a belief in Zebra-3 on this kind of inference. But he resists
the claim that this needs to be a violation of Closure. If you have warrant to believe
Zebra-2, then perhaps you must also have warrant to believe Zebra-3. But it doesn’t
follow that your warrant to believe Zebra-2 constitutes your warrant to believe Zebra-3. It
may rather be that you’d never get to the point of having warrant for Zebra-2 without
some other warrant for Zebra-3 antecedently in place. So what we have here isn’t a
failure of Closure, but rather just a failure of your warrant to transmit through the
reasoning from Zebra-2 to Zebra-3. Because Zebra-2 isn’t itself part of your warrant to
believe Zebra-3, you can’t reasonably base a belief in Zebra-3 on Zebra-2.2
For this to be a useful diagnostic tool, we need a systematic story about when we
should expect warrant to transmit, and when we shouldn’t. What features of this case
are behind our judgment that you can’t reasonably base a belief in Zebra-3 on Zebra-
2? And if we’re to avoid violating Closure, why is it that Zebra-3 so conveniently
happens to be inependently warranted every time we get to the point of having
warrant for Zebra-2?
Here’s one natural story. As we said at the end of the previous section, perhaps your
visual experiences don’t justify belief in Zebra-2 on their own, but only against the
background of other warrants you possess. So our diagram ought really to look like this:
Zebra–1
Zebra–2
We might express what bothers us about the inference from Zebra-2 to Zebra-3 by
saying “You were already taking Zebra-3 for granted, when you moved to Zebra-2!”
2
I’ll be drawing on Wright 2000a, 2002, 2003, 2004, and 2007. Klein 1981 had a preliminary version of this
notion (“evidence for” closure, as opposed to “being justified” closure), and Wright began articulating it in his
1985. See also Davies 1998, 2000, 2003a, and McKinsey 2003. These authors contrast failures of Closure to
transmission-failure: see, for example, Wright 2000a, pp. 140–1, 143, 157; 2002, pp. 331–2, 335; 2003, pp. 57–
8, 61, 67–8. But whether these really do diverge is sensitive to what we mean by “Closure.” You do have
warrant to believe each of Zebra-2 and Zebra-3; so it’s true we don’t have any failure of this principle:
Closure for Propositional Warrant. If you have warrant to believe some premises, which you recognize
entail Q, then you also have warrant to believe Q.
However, some have argued that Closure is better understood as a principle about doxastic warrant, like:
Closure for Doxastic Warrant. If you have warrant to believe some premises, and correctly deduce Q from
them, then the belief in Q you so form will be a warranted belief.
And in the Zebra case, a subject could correctly deduce Zebra-3 from Zebra-2, but wouldn’t be able to
reasonably base a belief in Zebra-3 on that deduction. Arguably, this Closure principle will fail whenever
warrant-transmission does. (See Silins 2005, }5 for relevant discussion.)
The phenomenon we’re discussing is also sensitive to what kind of warrant you have for Zebra-2. If your
warrant for Zebra-2 wasn’t coming from Zebra-1, but rather from the fact that the animal just brayed in the
distinctive way that only zebras can, then plausibly you would be in a position to eliminate the mule
hypothesis, and so would be able to infer Zebra-3 from Zebra-2.
So perhaps Zebra-3, or something close to it, is part of the background warrant that’s
needed to provide your warrant to believe Zebra-2:
Zebra-1
Zebra-3
Zebra-2
Arguably, any case with this structure will bother us in the same way. When Zebra-3 is
among the background you need antecedent warrant for, to have the warrant you do
to believe Zebra-2, then that warrant for Zebra-2 won’t itself further warrant or count
in favor of Zebra-3. Consequently, you won’t be in a position to reasonably believe
Zebra-3 on the basis of Zebra-2. We can call this the Background Warrant (BW)
Model of transmission-failure:
Zebra-1
Zebra-3
Zebra-2
This arrow means the reasoning would not transmit the displayed
warrant for Zebra–2 to Zebra–3.
Zebra-3
Some tricky details aside, this is the most plausible and recognizable form of the
phenomenon Wright has in mind. Our central two questions in this essay will be: Q1.
Does Moorean reasoning, like:
Moore-1. [Visual experiences as of hands]

Moore-2. I have two hands.
Moore-3. I’m not a disembodied handless spirit, who is merely hallucinating hands.
exemplify the same BW Model? I say it does not. We’ll focus even more on: Q2. Can
we identify some other general model of transmission-failure that Moorean reasoning
does exemplify? I will float and criticize some candidates. In the end, I don’t think
anyone has yet identified a pattern of transmission-failure that we have any consensus
Moorean reasoning exemplifies.
We’ll begin by sorting out various details.
III
Let’s say you have an earned warrant when the warrant is constituted at least in part by
some experience or awareness or insight or understanding that you have. Typical cases
of earned warrant will be when you perceive that something is so, or when past
observations warrant believing further conclusions. But introspective warrant can be
earned, too: the experience of having a headache (or on some views awareness of your
headache) gives you warrant to believe you have a headache. A priori warrant can be
earned, too: your understanding of the relation 6¼, the singleton function, and the
empty set, together give you warrant to believe 6¼{}.
We can separate warranted beliefs into those which are based on premises, or inference from
other information you possess; and those which are not. We can separate warrants to believe
into those which are constituted in part by warrant you have to believe other things, and
those which are not. These contrasts will not coincide. Recall our earlier example:
R
In this scenario, your belief in Z may not be inferred from any premises. Nonetheless,
part of the warrant you have for Z is the warrant you have to believe R, that your
senses are reliable. You need warrant to believe R, though you don’t need to actually
believe or even consider R, for your belief in Z to be warranted.
I’ll say that a belief is inferentially warranted if it’s warranted by virtue of being
inferred from other warranted beliefs. Derivatively, we can say that a warrant is
inferential when it’s constituted by the availability of some inference to the subject
from other, antecedent warrants. On the other hand, a warrant is mediate just when
it’s constituted (at least in part) by warrant you have to believe or doxastically accept
other things. What our example illustrates is that a belief can be non-inferentially
warranted, and so its warrant be non-inferential, even though that warrant is still
mediate. You might infer Z from R and some claim about what experiences you’re
having; but in the case we’re considering, the warrant you have for Z isn’t constituted
by the availability of such an inference. It’s constituted by your merely having the
experiences you do, and also having warrant to believe R.
A warrant is immediate when it’s not constituted by warrants to believe or dox-
astically accept other things. Recall the example:
LOOKS. It looks as though there’s a zebra in the pen.
Here, what warrants you in believing LOOKS is just your having certain experiences. Or
on some views, it’d be your introspective awareness of having them. In neither case
is warrant to believe any other claims involved. Warrant of this sort would be immediate.
In addition to earned warrants, some philosophers countenance the possibility of

unearned warrants. These will be warrants that aren’t constituted by any experience
or awareness or insight or understanding the subject has. For example, some philoso-
phers talk of us having default justification to believe that our senses are reliable, and that
we’re not brains in vats: these hypotheses, they say, are just intrinsically epistemically
probable, without the subject needing to do or experience anything to make them so.3
Some philosophers say that cogito beliefs are warranted just by virtue of being hyper-
reliable, that is, impossible to form falsely.4 And so on.
It’s controversial whether there are any unearned warrants.5 But many philosophers
do countenance some. Wright is among them. He calls unearned warrants “entitle-
ments”; he reserves the term “justification” for earned warrants; and he uses “warrant”
for the broad class that includes both earned and unearned varieties. I prefer the term
“justification”; I think it fits history and broader usage better to apply that term to any
kind of warrant to believe or doxastically accept things. If there are entitlements or
unearned warrants, I’d call them “justification,” too. However, in the present discus-
sion I will stick close to Wright’s own usage.6
Recall again the example:
I said that your warrant here to believe Z is mediate. This will be so regardless of
whether your warrant to believe R is earned or unearned. Even if what warrants you in
believing R is some default presumption in its favor, rather than any experience or
insight you’ve had, it’s still true that part of what warrants you to believe Z is warrant to
believe something else. So your warrant to believe Z is mediate.
On the other hand, some theorists including me think our warrant for some
perceptual beliefs is immediate. If Z is such a belief, then your warrant would instead
be structured like this:
3
See for example White 2004, Davies 2000, Field 2000, Field 2005, Cohen 1988, esp. }V, Cohen 1999,
Cohen 2000, Zalabardo 2005, and White 2006.
4
See Pryor 2006a and 2006b for discussion.
5
Conee and Feldman’s “evidentialism” might be understood as the doctrine that all warrant is “earned,”
in the way we’re understanding that here (see their 2004).
6
In his 2007, Wright says he uses “warrant” only for all things considered epistemic statuses; but I’m not
following him in that regard. I will speak both of all-things-considered and prima facie warrants. Burge 1993
and 1996, Dretske 2000, and Peacocke 2004 use the terms “warrant” and/or “entitlement” in ways that
roughly parallel Wright’s (but there are also differences). Other philosophers use these terms for a variety of
different epistemic statuses: for example, Meiland 1980; Pollock 1983; and Plantinga 1993a, 1993b. Wright
himself uses the notion of “warrant” differently in his 1991.
I’m not sure whether claims about zebras are immediately warranted in this way, but
I do think that some perceptual beliefs are. At a minimum, our simplest perceptual
beliefs, like There is light ahead, are. I call this view dogmatism about perception.
If dogmatism about perception is right, then some instances of Moorean reasoning
won’t have this structure:
Moore-2. I have two hands
but will instead have this:
and so neither Moore-3, nor anything else, will be among the background warrant you
need, to have the warrant you do for Moore-2. There is no such background warrant. If
dogmatism is right, then, these instances of Moorean reasoning won’t exhibit the BW
Model of transmission-failure. It remains to be seen whether there’s some other model
of transmission-failure that they do exhibit.
A Moorean dogmatist thinks that there isn’t—that, whatever its faults, the Moorean
reasoning is capable of transmitting warrant to Moore-3. That reasoning can be a way to
acquire some warrant to believe Moore-3; and it can be reasonable to believe Moore-3
on the basis of it.
I’ve defended Moorean dogmatism elsewhere.7 My preference is for a version of
dogmatism where your perceptual warrant comes from the phenomenology of your
7
See Pryor 2000 and 2004.
experience, which I think can be shared between veridical perceivers and hallucinators.
However, our discussion here will have mostly to do with dogmatism in general: with
any view that holds that your perceptual warrant is immediate. This includes reliabilist
and disjunctivist versions of the view, too; and views which would explain your
perceptual warrant in terms other than what the phenomenology of experience is like.8
It’s not straightforward which perceptual beliefs are good candidates to be immedi-
ately warranted. In the Zebra reasoning, I think your warrant does plausibly depend on
some background assumptions. This is because Zebra-2 goes beyond what’s really
reported by your experiences. If it is a cleverly-disguised mule, or a fur-covered robot,
we wouldn’t say that you’ve misperceived it. The error wasn’t in what you saw, but in
what you went on to believe. On the other hand, if you’re having illusory experiences
as of, say, a light ahead, you are victim to a perceptual error. Things aren’t the way
vision is reporting them. So I’d say that There is light ahead is specifically reported by
your senses, and is immediately warranted. It’s not obvious which example I have two
hands is more like. But for expository convenience, I’ll treat this as a candidate to be
immediately warranted, like There is light ahead.
Care is needed not to confuse dogmatism with alternative epistemologies.
Dogmatism might naturally be summarized with the slogan “You’re allowed to take
it for granted that your senses are reliable, that you’re not a brain in a vat, and so on.” If
you acquire evidence of your unreliability, or that you’re envatted, that will undermine
your perceptual warrant; but no antecedent warrant to believe those undermining
hypotheses fail to obtain is needed. You’re “taking it for granted” that those hypotheses
are false in the sense of being vulnerable to them as defeaters, without having
antecedent warrant to rule them out.9
This slogan “You’re allowed to take it for granted . . .” can also be read differently. It
can be read to mean that you have an entitlement or warrant to believe that your senses are
reliable, that you’re not a brain in a vat, and so on—perhaps a warrant you didn’t need
to do anything to earn. This view is not dogmatism. It’s just one version of the
competing view that perceptual warrants are mediate. This is a popular alternative to
dogmatism; it’s the view Wright espouses, and that Davies held until around 2003.10
On this view, your perceptual warrants aren’t inferential. Your experiences as of hands
may warrant you in believing you have hands without you needing to perform any
inferences or even needing to think about whether you’re a brain in a vat. Neither do
you need to have any special experience or insight. The antecedent warrants in
question are unearned “entitlements.” But to be warranted in your belief that you
have hands, on the basis of your hand-like experiences, you do need to have antecedent
warrant to believe you’re not a brain in a vat. It can’t come from your perception-based
beliefs in the external world; it has to be in place independently.
8
For a list of dogmatist-like views, see <http://www.jimpryor.net/research/dogmatist.html>.
9
See here Davies 2004, and Davies 2000, p. 404; see also Burge 2003.
10
See fn. 3, above, for some other proponents.
On this sort of view, the Moorean reasoning exhibits the same BW Model of
transmission-failure that we’re attributing to Dretske’s Zebra reasoning.
I mentioned earlier that there are tricky details with applying the BW Model of
transmission-failure to Dretske’s example. Let’s confront some of them. Consider this
alleged part of the Zebra reasoning:
Zebra-1
Zebra-3
Zebra-2. That’s a zebra in the pen.
Do you really need antecedent warrant to believe you’re not looking at a disguised
mule, to be perceptually warranted in believing Zebra-2? What if that alternative is
beyond your ken—e.g. because you’ve never heard of mules, or because you never
would have guessed that animals can be disguised to look like other animals?
One response here would be to say that you can have warrant to believe things that
you’re not yet able to believe or even entertain, and to maintain that you do after all need
warrant to believe you’re not looking at a disguised mule, even when that alternative is
not one you’re yet in a position to consider. A different response would be to say that
the mule hypothesis is one you’d need antecedent warrant to rule out when it arises for
you as a question. Prior to that, perhaps you don’t. In cases where the subject is
proposing to reason from Zebra-2 to Zebra-3, however, the question whether it’s a
disguised mule has arisen for the subject, so in those cases her warrants will need to have
the structure diagrammed above.
Perhaps these responses can be satisfyingly developed. But there is some real doubt
whether it’s really Zebra-3 that our warrant to believe Zebra-2 depends on. Might it
not instead be something more general?11 Perhaps it’s rather this:
Zebra-4. Zoo pens don’t often contain animals disguised to look other than they are.
If the structure of the Zebra reasoning is really this:
Zebra-1
Zebra-4
Zebra-2
Zebra-3
then the BW Model we’ve articulated no longer applies.
11
See here also Silins 2005, pp. 78–9.
Perhaps Zebra-4 and Zebra-3 are closely enough related that some minor variant of
the BW Model could apply. The idea may be: Zebra-3 is really supported by Zebra-4,
or by the interaction of Zebra-4 with Zebra-1, and would already need to be so
supported for you to have the warrant you do for Zebra-2. So reasoning from Zebra-
2 to Zebra-3 wouldn’t track the real structure of your warrants.12 They’d really look
like this:
Zebra-1
Zebra-4
Zebra-2
Zebra-3
I think it will be extremely difficult to spell out in detail why such structures should
obstruct warrant-transmission from Zebra-2 to Zebra-3. In explicit reasoning we very
often “overshoot” the real warrants that we’re relying on. For instance, suppose I visit a
remote plain of Antarctica, and see what looks to be a woman ahead. I reason:
Plain-1. [Visual experiences as of a woman.]

Plain-2. There’s a woman walking up ahead.
Plain-3. So I’m not the first person to walk on this plain.
My reasoning here doesn’t accurately track the real structure of my warrants. We
can undermine my warrant to believe Plain-2—give me evidence that I’m
incapable of distinguishing men from woman at this distance—while intuitively
leaving in place the same warrant I have for Plain-3. So Plain-2 is really stronger
than anything my warrant for Plain-3 depends on. I might more accurately have
reasoned:
Plain-1. [Visual experiences as of a woman.]

Plain-4. There’s a person walking up ahead.
Plain-3. So I’m not the first person to walk on this plain.
Yet I’d be very reluctant to say my actual reasoning, from Plain-2 to Plain-3, is
unreasonable. I think I can acquire warrant to believe Plain-3 by reasoning in that
way, and can reasonably base my belief in Plain-3 on such reasoning.
What is the difference between the bad Zebra reasoning and the permissible Plain
reasoning? Arguably, both have the same structure:
12
Compare Wright 2002, pp. 334–5.
and we want to say your warrant for 3 strictly comes from the interaction of 1 with
some of the stuff in the background, like 4, rather than from 2. Yet in the one case
reasoning from 2 to 3 can still seem reasonable; in the other case not. What makes the
difference? If the BW Model of transmission-failure is to be extended beyond cases
where 3 itself is in the background, these details need to be ironed out.
I don’t think it will be easy to do that; I’m not sure it will even be possible. But for
our purposes we can set these difficulties aside. If dogmatism is true, then no model of
transmission-failure which involves background warrants for the 2-premise will apply
to Moorean reasoning. For dogmatism says our perceptual warrant for Moore-2 doesn’t
depend on any background warrant.
IV
When people articulate their qualms about dogmatism, they often seem to be applying
a general principle like this:
No Fallible Immediacy (NFI) If some warrant W could be possessed by you in situations other
than A—for example, if it could be possessed in each of situation A and B—then W can’t
immediately warrant you in believing that A obtains, rather than the alternatives.13
But when I articulate this principle, I find that few are willing to embrace it. They’re
right to hesitate. Many epistemologies on offer nowadays countenance some form of
immediate warrant, and often that warrant is thought to be fallible. Can you have
immediate but fallible warrant to believe you have a headache? If so, then let A be the
situation in which you genuinely do have a headache, and B the situation where you
don’t, but still are warranted in thinking you do, and we have a counterexample to
NFI. Can you have immediate but fallible warrant to believe the answer to a simple logic
puzzle is A? Perhaps you’re wrong, the answer is really B. The warrant you have is
13
Prima facie, this principle is what seems to be behind arguments like: “Dogmatism has no resources to
prevent the conclusion that I can acquire elementary knowledge by vision plus elementary inference of a
proposition whose being false would predict my enjoyment of exactly the same visual experience. That is
absurd” (Wright 2007, }IV.4).
possessible in situations (such as the real situation) where A is false; yet it’s an immediate
warrant for A. So we have another counterexample to NFI.
Indeed, Wright himself seems committed to denying NFI. For he thinks we have
unearned warrants to believe we’re not brains in vats, that we instead have genuine
perceptual contact with our surroundings. This warrant is fallible: we might be wrong.
But neither does Wright think it’s constituted by our warrant to believe other premises.
So these unearned warrants can be fallible yet immediate.
If there’s no general difficulty with fallible yet immediate warrants, we need to know
why there should be some special difficulty in the case of perception.
Perhaps the problem is that perceptual warrants are earned. Perhaps what’s true is
not NFI but a more restricted principle:
No Earned Fallible Immediacy (NEFI) If some earned warrant, W could be possessed by you in
situations other than A, then W can’t immediately warrant you in believing that A obtains, rather
than the alternatives.
Wright can countenance fallible yet immediate unearned warrants, while refusing to
allow fallible yet immediate perceptual warrants, because they would be earned.14
What could motivate this restriction? What keeps it from being ad hoc?
I’m going to suggest a line of reasoning which may be tempting people here. This
will be speculative. But I think it’s worthwhile articulating the reasoning, and saying
what’s wrong with it.
Consider a situation in which you already know ahead of time what warrant you’re
going to acquire for A. For example, suppose you already know that when you open
your eyes you’ll have visual experiences as of an amber wall. In such cases, it’s very
plausible that actually having the experiences shouldn’t make any epistemic difference as
to whether the wall really is amber. It shouldn’t make you any more confident that the
wall is amber, rather than beige but lit by tricky orange lighting.
But now suppose you don’t yet know what warrant you’re going to acquire when
you open your eyes. Actually having experiences as of an amber wall would then tell
you you’re among those situations where you will have these experiences. It’s tempting
to think that’s all it can tell you. For, after you factor in the knowledge that you’re
among the situations where you will have the experiences, what epistemic role is left
for the experiences to play? Didn’t we just decide, a paragraph ago, that actually having
the experiences shouldn’t make any further difference, beyond what you’ve now
factored in?
And now, haven’t we arrived at NEFI? All that your experiences can by themselves
warrant you in believing is that you have them. They can’t give you an immediate
warrant to believe the wall is amber rather than beige but lit by tricky lighting.
There’s a subtle mistake in this reasoning. The mistake is to equate “learning that
you will have the experiences” with “learning only that you will have the experiences.”
14
Thanks to Carrie Jenkins for identifying this strategy.
On a view where your experiences immediately warrant you in believing the wall is
amber, the information that you will have the experiences should support not just the
claim that you will have them, but also, prospectively, whatever claims those experi-
ences will immediately support. That’s why actually having the experiences makes no
difference: if you knew in advance that you’d have them, then you should already have
incorporated the warrant they give you to believe the wall is amber. It doesn’t follow
that your experiences only warrant believing you have those experiences; and hence,
only that you’re either seeing an amber wall or seeing a beige wall lit by tricky lighting. If
it’s true that your experiences only warrant that, it needs to be shown another way.
A different source of resistance, to Moorean dogmatism in particular, is just that it
licenses Moorean reasoning, and that reasoning sounds bad to many philosophers. We
need some account of why it should sound bad. The simplest hypothesis would be that
it really is bad.
Some philosophers are tempted to diagnose the flaw in Moorean reasoning in
probabilistic terms, and argue that this conflicts with what the Moorean dogmatist
says. Moore-3 can be formulated in such a way that its negation entails you’d have the
experiences you do. Then even if your experiences did confirm Moore-2, which
entails Moore-3, by orthodox Bayesian thinking those experiences should disconfirm
Moore-3 rather than supply warrant to believe it. Whenever not-M entails E but
M doesn’t, the probability of M given E should be lower than the prior probability
of M.15
Some philosophers want to explicate transmission-failure in probabilistic terms like
these.16 But I propose to set the probabilistic considerations aside for this discussion.
There are many complications at play: What’s the relation between probabilistically
confirming and supplying warrant? Should the epistemic effect of having experiences be
the same as conditionalizing on the premise that you have them? There’s so much to
discuss that we need to take it up in other forums.17 Here I’ll suppose that transmission-
failure is meant to be a novel diagnostic tool, not just another vocabulary for expressing
familiar probabilistic theorems.
The complications aside, I doubt that probabilistic considerations really could be the
source of our deepest-rooted intuitive resistance to Moorean reasoning. Intuitively, the
following reasoning seems just as problematic:

Moore-3*. I have two hands and I’m not a disembodied handless spirit, who is
merely hallucinating hands.
15
For a list of philosophers pressing or responding to this argument, see <http://www.jimpryor.net/
research/bayesian.html>.
16
For a list, see <http://www.jimpryor.net/research/transmission.html>.
17
See Pryor (ms-a).
Yet, whereas Moore-2 merely entailed the earlier Moore-3, it’s logically equivalent to
Moore-3*. So if your experiences supply any probabilistic confirmation to Moore-2,
they have to confirm Moore-3* to the same degree. There’s no basis for any probabilistic
critique of the step from Moore-2 to Moore-3*. Yet our intuitive resistance to this
reasoning and to the original Moorean reasoning seem to come from a common
source.
Where now are we? The options on the table are: Either (i) we accept Moorean
dogmatism, so Moorean reasoning can be reasonable. In that case, we owe an account
of why that reasoning seems intuitively to be problematic. Or (ii) we reject dogmatism
altogether, as a false epistemology of perception. Instead, we go for something like
Wright’s epistemology, and say that Moorean reasoning fails because it exemplifies the
BW Model of transmission-failure. We might combine with that, or pursue separately,
the idea (iii) that there’s something probabilistically objectionable about Moorean
reasoning. This needs much discussion, and we’re going to set it aside here. Or, (iv)
we say that even if dogmatism is true, and so Moorean reasoning doesn’t exemplify the
BW Model of transmission-failure, nonetheless there are other (non-probabilistic)
models of transmission-failure which it does exemplify. So even if dogmatism is true,
Moorean reasoning should still fail.
I’m trying to present dogmatism in a favorable light. But my goal here is not so
ambitious as to establish it is true. Rather, my goal is to eliminate option (iv). I think
future debate is best focused on the other options. (A fifth, higher-order option will be
discussed in }VII, below.)
I’m not sure whether Wright disagrees about this. Often he seems to be backing
option (ii); and in 2007 it’s hard to read him as entertaining anything else. But as we’ll
see, there’s enough interpretive unclarity that he may sometimes be read as backing
option (iv). My goal is to persuade you that that’s a bad strategy. If you want to charge
Moorean reasoning with transmission-failure, you’re best off arguing that dogmatism is
false and that Moorean reasoning exhibits the same BW Model of transmission-failure
that we’re attributing to the Zebra reasoning. (Or getting into the details of what
probability dynamics are at work.)
I’ve proposed a dogmatist defense of Moorean reasoning in my 2004. It will be
useful to quickly review what I said there; that will be Section V. In Section VI, we’ll
take up the interpretive task of sorting out exactly how Wright is conceiving of
transmission failure. In Section VII, we’ll consider and criticize possible models for
transmission failure other than the BW Model.
V
If the Moorean dogmatist is to be believed, Moorean reasoning can be reasonable.
Why then does it strike us as so intuitively problematic?
To answer this, we need to discuss three or four key questions. The first is:
What is the epistemic effect of false higher-order warrant? For example, consider
a mathematician who reasons correctly to P, but then acquires warrant to believe,

falsely, that the proof strategy she used is flawed. She now has higher-order warrant
to believe, falsely, that her original reasoning fails to warrant belief in P. How
confident should she be that P? Plausible things can be said on either side of this
question; but in the end, I and many others think the mathematician’s higher-order
warrant should render P somewhat less warranted for her. Warranted beliefs about
your warrants, even when those beliefs are false, can undermine the prima facie
support the lower-order warrants provide.
The second question is: What is the epistemic effect of reasoning that’s in fact flawed
but which seems compelling? Suppose our mathematician carefully reasons through
another proof, of Q, which does have some subtle flaw that she’s unaware of. Will she
have warrant to believe Q? Perhaps she has inductive warrant: she knows that proofs
she finds compelling, and proofs in this journal, have good track-records. But set that
aside. Does she have any non-inductive warrant to believe Q? Does the apparent
compellingness of the proof, or her experience of seeming to deduce Q in that way,
give any warrant for Q?
I don’t know what’s the right thing to say about this. I can be persuaded either way.
What we say about Moore’s argument needs to track what we say.
Suppose we say the mathematician does have some warrant for Q, despite the
flaw in her proof. Then consider a philosopher who’s reading skeptical arguments
to the effect that we have no perceptual warrant; or anti-dogmatist arguments to
the effect that NEFI is true, and so experiences don’t by themselves, immediately,
warrant any perceptual beliefs. If dogmatism is true, such arguments are flawed.
But the philosopher may not yet see the flaws, or may not see them as flaws. The
arguments may seem compelling to her. Then she’ll be in a situation like the
mathematician’s: she may be warranted in believing the skeptic’s conclusions,
despite the flaws in the arguments. And then, drawing on our answer to the first
question, the philosopher’s warrant for these false claims about her perceptual warrant
may well undermine that perceptual warrant. Even though her experiences really do
provide it.
That’s not to say that the skeptic is correct. The skeptic says that none of us are
perceptually warranted. We’re only here describing a mechanism by which people
who read skeptical arguments and find them compelling may end up in the kind of situation
the skeptic falsely claims we’re all in. Doing epistemology may be harmful to your
epistemic health. Other people, who are ignorant of the skeptic’s arguments, or are
(permissibly) unmoved by them, can retain their immediate perceptual warrant to
believe Moore-2, and can reasonably go on to infer Moore-3 from it.
On the other hand, though, perhaps flawed arguments can’t warrant belief in their
conclusion. In that case, the dogmatist is committed to saying that philosophers lack
warrant to believe the skeptic’s and the anti-dogmatist’s conclusions, even when the
arguments for those conclusions seem compelling. To sort out what’s going on in that
case, we need to attend to a third question.
Suppose you merely believe P, without warrant for doing so. You also recognize
that P entails Q, and nothing about that entailment serves to undermine your confi-
dence in P. What then is your epistemic relation to Q?
It doesn’t seem right to say that you have warrant for Q. All that Q has going for it is
that it follows from P, an unwarranted belief of yours. But neither does it seem enough
to say that Q is just unwarranted, in the way that any other arbitrary unsupported
proposition is. Your unwarranted belief in P seems to put some kind of pressure on you
to believe Q too. The overall most reasonable course may be for you to believe
neither. But if, failing that, you believe P yet resist believing Q, you’re manifesting
some further epistemic flaw. If you’re going to be unreasonable and believe P, then
you “ought” in some sense to believe Q too.
This phenomenon is familiar in the ethics and practical reasoning literature. In recent
years Broome has discussed how it operates in epistemology too.18 I’ve yet to find a
fully satisfying vocabulary for describing the phenomenon. Broome began by talking
about “normative requirements,” as opposed to “reasons”: your unwarranted belief in
P normatively requires you to believe Q, but since it’s unwarranted, it isn’t a reason for
Q. In my 2004, I talked about “rational commitment.” Both expressions are non-
optimal. For one thing, the general phenomenon here can also be present in cases
where the relation between P and Q is merely ampliative; so talk of “requirement” and
“commitment” is too strong. For another thing, talk of “commitment” may mislead.
We don’t want to say that the subject is conclusively under any rational obligation to
believe Q. Rather, she’d be best off believing neither P nor Q.19
Perhaps it’d be helpful to borrow an idiom from ethics, and call the relation between
your belief in P and a prospective belief in Q one of conditional or hypothetical
support. If you really had warrant for P, on the other hand, that warrant would
categorically support believing Q. Hypothetical support is partly a matter of what
such categorical support you would have, if you had that warrant for P. But it’s also
more than that: when your belief in P hypothetically supports believing Q, and you
nonetheless resist believing Q, you will thereby be exhibiting some kind of (categori-
cal) epistemic defect.
Some philosophers describe these phenomena in terms of relevance relations that
hold between P (or sometimes your belief in P) and the proposition Q, rather than a
prospective belief in Q. That may be adequate for some cases, but it won’t generalize
properly. Intuitively, the same phenomenon is at work when a subject believes,
without warrant, that on balance her evidence tells equally in favor of Q and not-
Q. In such a case, I’d like to say that the subject’s unwarranted belief about her
evidence hypothetically supports suspending belief whether Q. If we were restricted to
18
See the series of papers beginning with Broome 1999.
19
Also, one referee thought it more plausible to say it’s not your belief in P itself, but rather your doxastic
commitment to P, which is constituted or manifested by your believing, that “commits” you to any further
beliefs.
relations whose right-hand arguments were propositions, we wouldn’t be able to say

that. (In such cases, the subject’s unwarranted belief about her evidence may well
hypothetically support believing propositions about her relation to Q; but it should also
hypothetically support the prospective attitude of suspending belief whether Q itself is
true.) Similarly, on the left-hand side, if a subject is unwarrantedly agnostic about P,
that attitude may hypothetically support being agnostic about further matters. So the
phenomenon we’re talking about is fundamentally a relation between actual and
prospective attitudes, not just a relation between propositions.
Here’s another generalization of the phenomenon, which will be crucial to our
discussion. Suppose you have some warrant W that (categorically but prima facie)
supports believing Q. This warrant is vulnerable to being undermined by warrant to
believe U. Now suppose you merely do believe U, without warrant. Earlier, we said that
mere belief in P can hypothetically support believing Q. Similarly, I think mere belief
in U can hypothetically undermine the warrant W gives you to believe Q. In my
2004, I described this by saying your belief in U “rationally obstructs you” from
believing Q on the basis of W. You’re already manifesting a rational flaw by believing
U without warrant; but you’d be manifesting a further flaw if you went ahead and
believed Q on the basis of W, while maintaining your belief in U.
It’s clear to me that this is a flaw, but unclear how it relates to other, more familiar
epistemic statuses. Perhaps having doxastic warrant—warranted beliefs, not just warrant
to believe—requires that your beliefs not be hypothetically undermined in this way.
I’m not sure; but this is a proposal I do tentatively defend.20
In any event, let’s return to our philosopher who finds flawed skeptical or anti-
dogmatist arguments compelling. This philosopher unwarrantedly believes that she’s in
no position to have perceptual beliefs. She believes that her experiences don’t by
themselves, immediately, warrant any perceptual beliefs. Or she unwarrantedly believes
things that hypothetically support those conclusions. In response to our first question, we
said that warranted belief in such conclusions should (categorically) undermine the warrant
her experiences really do give her. Given what we’re saying now, unwarranted belief in
those conclusions (or in premises that hypothetically support them) should hypothetically
undermine in a parallel way. For a philosopher with such beliefs, it’d be epistemically
defective to believe things just on the basis of her experiences—even if those experiences
are in fact giving her categorical warrant to so believe.
I’ve described mechanisms by which philosophers who have certain epistemic
views—or beliefs that hypothetically support those views—would have their perceptual
warrant for Moore-2 categorically or hypothetically undermined.21 Those mechanisms
20
See Pryor 2004 and Pryor (ms-b). Bergmann 2004 and 2005 shares many important elements of my
view; though there are also some differences.
21
Note I have not proposed that the mere possession of critical faculties, or more conceptual sophistication,
makes you worse off. Those would be different sorts of claims. It’s only when your faculties are deployed in the
ways I described that you may make yourself epistemically worse off than necessary.
are genuine; but they can foster the illusion that your warrant for Moore-2 is structured
like we imagined your warrant for Zebra-2:
Moore-2
That is, they foster the illusion that your warrant for Moore-2 isn’t immediate, but
rather depends on antecedent background warrant to believe things aren’t as the
skeptic proposes. I want to emphasize we’ve now seen how these mechanisms may
be operating even if that’s false, and your warrant for Moore-2 is immediate. That may
not be enough to dispel the illusion, but I hope it will begin to counteract it.
Let’s confront an important fourth question. What if a subject doesn’t believe the
skeptic’s conclusion, but is merely agnostic as to whether it’s true? What if she doesn’t
believe any undermining hypothesis, but merely entertains doubts as to whether they’re
true?22 Should Moorean reasoning be rationally effective against her agnosticism or
doubts? Most likely it won’t be effective, and it can seem like that’s rationally appropri-
ate. Someone who considers Moorean reasoning from such a position shouldn’t be
persuaded by it to overcome her agnosticism or doubts. Can the dogmatist give any
account of this?
This is difficult, in part because it’s difficult to identify what agnosticism and doubts
really are. To begin, we should distinguish agnosticism from the state of having no doxastic
attitude at all towards an issue. A subject who’s never considered a question isn’t yet
agnostic about it; she has no doxastic attitude at all. Agnosticism is instead a particular kind
of attitude, one of suspending belief. As an initial approximation, we may construe
agnosticism as a kind of mixed partial belief: you partly believe things are one way, and
partly believe they’re the other way. This is only an approximation: as Scott Sturgeon has
argued, agnosticism seems to involve a kind of “committed neutrality” that a mere state
of mixed belief may lack. Still, it will be enough to start with.
As an initial approximation, “entertaining doubt” may also be construed as having a
kind of mixed belief. (Or perhaps it’s not mixed real belief, but instead mixed pretend or
suppositional belief.) This too is only an approximation, but again I think it will be
enough to start with.
Now undermining, like support, is a matter of degree. If a subject who all-out
believes an undermining hypothesis thereby hypothetically undermines his perceptual
warrant, a subject who partially believes (or partially pretend-believes) an undermining
hypothesis will thereby hypothetically undermine his perceptual belief to some degree.
22
See Wright 1985, p. 437: “One the hypothesis is seriously entertained that it is likely as not, for all
I know, that there is no material world as ordinarily conceived . . . ” See also his 2007, }IV.2, and White
2006’s three-card scenario.
So the same mechanisms we’ve been discussing will be operative; and I think they’ll
still contribute to the illusion that we need antecedent warrant to believe things aren’t
as the skeptic proposes.
At one point Wright says:
I cannot rationally form the belief that it is currently blowing a gale and snowing outside on the
basis of my present visual and auditory experience while simultaneously agnostic, let alone
skeptical, about the credentials of that experience.23
I agree with him; there’s nothing in this passage that a Moorean dogmatist needs to resist.
When you form any belief, it’s irrational to combine the belief with a continuing agnosti-
cism about the belief ’s warrant. Believing on the basis of some warrant would be
hypothetically undermined by such agnosticism. It doesn’t follow that you need antecedent
warrant to believe that your perceptual beliefs are warranted after all (that your experi-
ences have good “credentials”). For all Wright says here, subjects who have no doxastic
attitudes whatsoever about the credentials of their experience, or subjects who just uncritically
accept that their experiences warrant their perceptual beliefs, may be reasonable in holding
their perceptual beliefs. Not because that acceptance is covertly warranted—but rather
because they merely lack the doxastic attitudes that can hypothetically undermine the
warrant their experiences are immediately providing. Being perceptually warranted
doesn’t require us to have antecedently warranted positive higher-order beliefs, to the effect
that our experiences do give us warrant. But as we’re seeing, it may require us to lack
negative higher-order attitudes—warranted or not—that hypothetically undermine, and
so rationally interfere with, our first-order, immediate perceptual warrants.24
I don’t expect these explanations to satisfy everyone. In particular, if you’re sure it
must be less reasonable for the subject to believe Moore-3 on the basis of Moore-2 than
it is for her to believe Moore-2, these explanations probably won’t satisfy you. So be it.
They at any rate constitute some accounting for why Moorean reasoning should seem
sketchy to us, compatible with its retaining its capacity to transmit warrant. I don’t
expect what I’ve said here to be decisive.
VI
How does all this bear on Wright’s own understanding of transmission-failure?
Wright—and Davies also—alternate between different glosses on what transmis-
sion-failure consists in. One criterion they give is: Can reasoning through an argument,
23
Wright 2004, p. 193; compare 2003, p. 61; 2007, }} II, IV.1, and IV.2; 2000a, p. 142; and 2002, pp. 334–5.
24
This bears on the “Simple Elevation Hypothesis” Wright formulates in his 2007, }II. Roughly, that
principle says that if C are the conditions that confer some warrant W, then being in a position to warrantedly
believe C obtains suffices for being in a position to warrantedly believe you have W. I agree that warrant to
believe C obtains may warrant some subjects in believing they have W. But it can’t suffice to put them in a position to
warrantedly believe they have W. They’d need also to lack interfering higher-order attitudes, including
agnosticism.
relying on the indicated warrants for its premises, add up to (or constitute, or give you)
any new warrant to believe the argument’s conclusion?25 Another family of criteria asks
things like: Can reasoning in the indicated way rationally resolve doubt or agnosticism about
the argument’s conclusion?26
In Section V, above, I reviewed reasons for thinking that Moorean reasoning might
count as transmission-failing by the second criterion even if it didn’t by the first—that
is, even if it did give subjects new warrant for Moore-3, and the right kinds of subjects
(subjects without interfering higher-order attitudes) would be able to reasonably believe
Moore-3 on the basis of such reasoning. So if the notion of transmission-failure is to
track an argument’s inability to do that, then we should take the first criterion as
definitive, and set the second aside.27
But when does that first criterion apply? What sorts of arguments can and what sorts
can’t transmit warrant?
Wright has addressed that question in several places. In most of his discussion he
seems to be operating with some form of our BW Model of transmission-failure. He
does distinguish between two transmission-failure “templates.” He calls them the
information-dependent and the disjunctive template. (I’ll indicate Wright’s termi-
nology with this typeface.) These present some interpretive difficulties, but as we’ll
discuss, they might properly be understood as two subspecies of the BW Model. So it’s
not clear that Wright countenances any forms of transmission-failure other than the
BW Model. If he doesn’t, then he should agree that if dogmatism is true (and we set
probabilistic worries aside) Moorean reasoning should be allowed to transmit warrant.
However, there are also elements in Wright’s discussion that do suggest non-BW
models of transmission-failure.
Throughout his discussion, Wright deploys a contrast between information-
dependent and non-inferential warrants. Wright says that what makes a warrant
non-inferential is that the warranted belief not be based on any inference, and that,
furthermore, it not draw any of its support from “evidence” or “collateral informa-
tion.”28 In his earliest discussions, he uses “evidence” to mean “empirical evidence” (see
25
See Wright 2000a, pp. 142–3, 154; 2002, pp. 335, 342; 2007, }III; Davies 2000, pp. 388, 393, 394, 399,
400. Sometimes this is put differently: Can reasoning in the indicated way add up to a first warrant to believe
the argument’s conclusion? (Wright 2000a, pp. 141, 155; 2002, pp. 331–2; 2003, p. 57; Davies 2000, pp. 386,
388). It’s unclear whether every argument that passes the main criterion for transmitting warrant needs also to
pass that variant (see Davies 2000, p. 388 and fn. 17).
26
Wright 2000a, p. 140; 2002, p. 331; Davies 1998, pp. 348–9; Davies 2000, pp. 386, 387, 388,
394, 396–7, 398–9, 400, 401. A significantly different variant is: Can doubt about the conclusion be
rationally combined with preparedness to reason in the indicated way? See Davies 2000, pp. 397, 398–9,
401, 402–4, 410.
27
Though see Davies’s development of the second criterion in Davies 2009.
28
For his definitions of “information-dependent,” see Wright 2000a, p. 143; 2002, pp. 335–6; and
2003, p. 59. On “non-inferential,” see Wright 2002, p. 341 and 2003, pp. 60, 62.
2000a, p. 147). In later discussions, he counts earned a priori warrants as evidence too
(2004, pp. 174–5). But nowhere does he count unearned warrants as evidence. So now
consider beliefs whose warrant is non-inferential, in the sense I articulated in section III,
but still mediate, because it depends on antecedent unearned warrant to believe other
things. Wright will regard these beliefs as non-inferential and “groundless” (2000a,
pp. 152–3, 156–7; and 2003, pp. 66–67). In his view, that includes our most basic
perceptual beliefs. In this table:
What I call immediate What I call mediate but What I call inferential
warrants non-inferrential warrants warrants
Non-inferential warrants that

depend on antecedent unearned
warrants to believe other things
Non-inferential warrants that

depend on “information”
(that is, antecedent earned
warrants to believe other things)
the darkest gray cells represent Wright’s information-dependent warrants; the other
cells his non-inferential warrants.
In Wright’s vocabulary, our perceptual warrants have a humean or “i-ii-iii” struc-
ture when they’re “inferentially” supported by “defeasible evidence” (see 2002,
pp. 336 and 341; 2003, p. 60; 2004, p. 172). They have only a cartesian structure if
they’re “acquired not on the basis of evidence, but directly” (2003, pp. 60, 62; 2004,
p. 168). This bears on what sorts of skeptical argument, and what sorts of transmission-
failure, our perceptual beliefs are vulnerable to.
As I said, Wright outlines two templates for transmission-failure. The first is the
information-dependent template. This is instantiated when some premise in a piece
of reasoning is warranted in part by information, that is, antecedent earned warrants,
that includes the reasoning’s conclusion. That’s clearly an example of our BW Model
of transmission-failure:
information, or earned warrants

to believe B premise A
conclusion B
But the BW Model also would apply to cases where your reasoning depends on
antecedent unearned warrants that include the reasoning’s conclusion:
unearned warrants to believe B

premise A
conclusion B
An important question is whether this is all that Wright’s second template, his
disjunctive template, manages to cover.29 If so, then Wright’s two transmission-
failure templates are just subspecies of our BW Model.
However, Wright seems to want his disjunctive template to cover cases where the
relevant warrants are alleged to be genuinely immediate. So perhaps the disjunctive
template only overlaps with our BW Model, and includes more besides.
It’s not easy to determine which of these two possibilities obtains. My considered
view is that the former is correct: that the only models of transmission-failure Wright
successfully presents us with are BW Models. When he’s trying to apply his disjunctive
template to cases where your warrant for the relevant premises is alleged to be genuinely
immediate, he’s most charitably interpreted as arguing that the warrant in question can’t
really be immediate, but can at best be of the sort represented by the light gray cell in
our table. (I’m not sure whether this captures Wright’s own understanding of what he’s
doing.)
I’ll briefly present some evidence in support of this interpretation. But as I said, the
question is tricky. (I’ll reserve some of the harder evidence until later.)
To begin, Wright at some points (in 2003, at note 9 and pp. 67–8) says that
transmission-failure presupposes Closure. That is, in order for us to have a case where
you have warrant for A that fails to transmit to B, it must be that Closure is not violated:
you must also have some warrant to believe B. Why? Wright seems to be thinking that
29
For Wright’s exposition of the disjunctive template, see Wright 2000a, pp. 153–6; 2002, pp. 342–4;
and 2003, pp. 60–3.
if your warrant for A didn’t already require that warrant for B, you wouldn’t be in the
kind of position that blocked warrant-transmission. In other words, transmission-
failure must be due to B’s being among the background warrants that warrant you in
believing A. This is just our BW Model.
Let’s next consider Wright’s expositions of his disjunctive template. Wright is
supposing that he has some warrant W his possession of which is compatible with each
of A and C being the case. He believes A on the basis of that warrant, and reasons from
A to B. B is such that its falsity implies C is true; to simplify we can just set B equal to
not-C. Wright focuses on the question: what should make it reasonable for him to
believe A on the basis of W, rather than the more tentative belief A ∨ C? Here is a
sample of what he says:
[T]he more tentative claim would indeed be appropriate unless I am somehow additionally
entitled to discount alternative C . . . [I]n order for me to be entitled to discount C, and so
move past the disjunction to A, I have to be entitled to discount the negation of B, and therefore
entitled to accept B; for by hypothesis, if not-B were true, so would C be. So it would seem that
I must have an appreciable entitlement to affirm B already, independent of the recognition of its
entailment by A, if I am to claim to be warranted in accepting A in the first place . . . So my
warrant [for A] does not transmit [to B].30 31
Notice what he says: for the subject to reasonably believe A on the basis of W, he needs
to be antecedently “entitled to discount” not-B, and “entitled to accept” and “affirm”
B. Elsewhere Wright talks of having “antecedent reason to think” that B holds (2000a,
p. 156; see also 2003, p. 66 and 2002, p. 346). In other words, the subject’s warrant
for A must require him to have an antecedent warrant to believe or doxastically accept
B. Presumably, then, the warrant for A is not immediate. But it can still be non-
inferential in his sense: it can be a warrant from the light gray cell in our table.
That’s the most straightforward construal of Wright’s disjunctive template, and if it’s
correct, then this template will be just a subspecies of our BW Model. It won’t apply in
cases where your warrant for A genuinely is immediate:
30
Wright 2003, pp. 62–3, Wright’s italics; compare his 2000a, p. 155 and 2002, p. 343).
31
An aside: Wright says in a note that this argument relies on a closure step: if you need antecedent
warrant to believe not-C (because your warrant for A draws support from that antecedent warrant), and you
appreciate that not-B implies C, then you will also need some warrant to believe B. However, I think that’s not
enough for Wright’s purposes. The fact that you must in this situation have some warrant to believe B does
not entail that this warrant has to be antecedent to A, nor that your warrant for A will constitutively depend
on it too. So we wouldn’t yet have reason to deny that warrant could be transmitting from A to B. Wright is
on better footing if he assumes that your antecedent warrant for not-C not merely implies you have some
warrant for B; but assumes, more strongly, that your warrant for not-C will transmit to B. Then it will follow,
in the situation he describes, that you have a warrant for B which is antecedent to A. I think even more will be
needed to clinch his case: having a not-C-derived, antecedent-to-A warrant for B is compatible with also
having an A-derived warrant for B. So Wright ultimately needs to show that the warrant you have for
A includes the not-C-derived, antecedent to-A warrant for B. That would make it clear why your warrant for
A should be blocked from transmitting to B. This will be easiest when B implies, or otherwise supports, not-
C. When B is merely implied by not-C, more work will be needed.
[experiences]
However, there are reasons to hesitate: notice in the second-to-last sentence of the
quoted passage, Wright doesn’t say he needs an antecedent warrant to believe B if he is to
be warranted in accepting A. He says he needs it in order to claim to be warranted in
accepting A. Is this difference significant? Should we take Wright to be talking not about
what it takes to be warranted in believing A, but rather about what it takes to have some
higher-order warrant, to the effect that one is warranted in believing A?
If the claiming warrant talk was supposed to mean that, Wright didn’t do much
to highlight it. He mostly writes as though being in a position to claim a warrant
were just a matter of having a warrant—or perhaps, having an earned warrant. For
example, he says:
Whenever I get in a position to claim justification for a proposition, I do so courtesy of specific
presuppositions . . . for which I will have no specific, earned evidence . . . Again: whenever
cognitive achievement takes place, it does so in a context of specific presuppositions which are
not themselves an expression of any cognitive achievement to date.32
He speaks here as though getting in a position to claim justification were the same
thing as “cognitive achievement tak[ing] place.” Yet he uses “cognitive achievement”
to cover getting evidence for first-order claims like P, or coming to know P (see e.g.,
2004, p. 175). It’s not said to require higher-order warrants.
Wright elaborates on how he understands the notion of claiming warrant in his
2007. There he confirms that it should not be understood as just equivalent to believing, or
acquiring warrant to believe, that one has a warrant. Instead, the main role of the notion
is to express a kind of internalist warrant. It’s unclear to me whether he wants being in a
position to claim warrant to at least always include having a higher-order warrant.
There are other higher-order threads in Wright’s discussion. Consider these
corresponding sentences from his different expositions of the disjunctive template:
[W]hat, in the circumstances, can justify me in accepting A? Why not just reserve judgment
and stay with the more tentative disjunction, “A or C”?33
[W]hat, in the circumstances, can justify me in accepting A? Should I not just reserve
judgment and stay with the more tentative disjunction, either (I have warrant for) A or C?34
32 33 34
Wright 2004, p. 189. Wright 2000a, p. 155. Wright 2003, 62; my underlining.
[W]hat, in the circumstances, can justify me in taking it that I have a warrant for A? Why
not just reserve judgment and stay with the more tentative disjunction, Either (I have a
warrant for) A, or C?35
Wright doesn’t say anything to highlight these changes; and neither does he say
anything to motivate focusing on higher-order questions in this context. Moreover, his
dialectical settings demand that we shouldn’t be focusing on higher-order questions.
Wright’s expositions begin with questions about whether you’ve acquired warrant to
believe that certain animals are not disguised mules, or that the wall you see is not white but lit
by tricky red lighting. He says that you may warrantedly be taking it for granted that the
italicized claims are true, but it would be “absurd to pretend that you had gained a
reason for thinking so.” He then says his disjunctive template is suggested by, or is a
generalization of, what’s going on in such examples. Up to that point, everything was
formulated in terms of what first-order warrants we possess.
So I don’t think there’s enough in the texts to support reading much into Wright’s
higher-order-ish talk. But we’ll revisit this issue in the next section.
We’ve been discussing how to interpret Wright’s notion of non-inferential
warrant, and his disjunctive transmission-failure template. There are also some
other, more minor, interpretive issues to sort out.
One of these is that, in his 2004, Wright hesitates to speak of being entitled to believe
anything. What’s driving him is a concern whether the notion of “belief ” can be used
in a way that permits one to rationally believe things on mere “entitlements,” that is,
unearned warrants (2004, pp. 175–7). When the only warrants one has are unearned,
perhaps we should instead speak of trusting that things are as those warrants testify. If
so, then Wright’s account of perception should be revised to say: our perceptual
warrants require us to have antecedent entitlements to trust that our senses are
reliable, that we’re not brains in vats, and so on.
This talk of entitlements to trust might evoke the notion of “taking for granted”
that we discussed in }III, above. But that’s not what Wright means. His notion of
trusting that P is still a species of “accepting,” which is a general kind of action-
guiding attitude. trusting is like belief in that it’s a form of acceptance that excludes
doubt and withholding judgment (2004, pp. 192–4). At one place Wright characterizes
entitlement as a “norm of doxastic acceptance” (2004, note 1). So I think we can group
belief and trust together as two species of doxastic acceptance. The sense in which a
dogmatist says you “take it for granted” that you’re not a brain in a vat, and so on, is
instead just being vulnerable to that as a defeater. It’s no kind of doxastic attitude
towards the hypothesis that you’re not a brain in a vat.
If Wright’s notion of trusting is still a form of doxastic acceptance, then I don’t
think the contrast between it and belief bears much on the issues we’re discussing. If
there are differences between different kinds of doxastic acceptance, they’re not
35
Wright 2002, 343; my italics.
differences I care about. I’m interested in the general question of what we have warrant
to doxastically accept in any form.
Another interpretive issue concerns the fact that Wright seems to understand
entitlements as warrants that don’t “bear on the likely truth” of what they support
(2004, pp. 177, 182). This is puzzling. Most epistemologists would agree that there are
norms of doxastic acceptance that don’t bear on the likely truth of one’s belief, but
they’d deny that these are properly regarded as epistemic warrants. (That doesn’t have to
mean they have no relevance to epistemology.) For instance, Pascalian considerations
might mandate believing or at least doxastically accepting that God exists, but they
wouldn’t do anything to make that hypothesis epistemically more probable, and so
they shouldn’t be counted as epistemic warrants that God exists. I’ve been assuming that
Wright does intend his entitlements to be a kind of epistemic warrant. But I’m not
entirely sure that’s correct. I’d find the indispensability considerations he adduces in his
2004 more convincing if they weren’t supposed to constitute epistemic warrants (see
also 2000a, p. 157; 2003, pp. 66–7, 68). Perhaps this is a sign that Wright and I are
operating with fundamentally different understandings of what an epistemic warrant
needs to be. Well, I’ll hope there’s enough overlap in our understanding that I haven’t
seriously misrepresented him.
Summarizing: based on the evidence we’ve reviewed so far, Wright’s two transmis-
sion-failure templates seem to be subspecies of our BW Model. As we’ve noted many
times, that model can’t apply when your warrant for the relevant premises is genuinely
immediate.
VII
Could there be other models of transmission-failure, which apply even when your
warrant for the premises is immediate?
Our Background Warrant Model says:
... B
if B plays this role,
A
than this reasoning doesn’t transmit warrant

B
If B is among the background you need antecedent warrant for, to be warranted by

W to believe A, then your W-based warrant for A can’t transmit to B.
Some of Wright’s higher-order-ish talk suggests a different, Higher-Order Back-
ground Warrant (HOBW) Model:
if B comes here . . . W
. . . or here
W warrants believing A A
then this reasoning doesn’t transmit

warrant
B
If B is among the background you need antecedent warrant for, to be warranted in
believing you are warranted by W to believe A, then your W-based warrant for A can’t
transmit to B.
The suggestions are strongest in Wright’s 2002 exposition of his disjunctive template.
Later in that same paper he also writes:
[A] warrant, W, for a belief, A, cannot transmit to any of its consequences, B, if —in context—
one would need an entitlement (earned or standing) to B in order to defend the claim that
conditions for the acquisition of W were satisfied.36
In other words, transmission from A to B fails when an antecedent warrant for B is

needed to support the higher-order claim that you have acquired (or are at least in a
position to acquire) warrant W to believe A.
What could motivate such a restriction on warrant-transmission? After the work of
Alston and Audi,37 there’s now broad agreement among epistemologists that higher-
order epistemic statuses may be harder to achieve, and may need to draw on more
resources, than their first-order counterparts. So why should facts about what’s needed
to achieve the higher-order warrant that you are warranted by W to believe A bear on what
the first-order warrant for A is itself able to transmit to?
Even if we were to grant the HOBW Model, it’s not clear to me that Moorean
reasoning as understood by dogmatists would instantiate it. Some dogmatists, like
myself, are internalists about warrant. They should say: look, for W to warrant you
in believing you have hands, it’s not a requirement that your senses really be reliable, or
that you really fail to be a brain in a vat. So why, for you to warrantedly believe that W so
warrants you, should it be a requirement that you have antecedent warrants to believe
those things? It ought to be enough that you merely lack warrants to believe their
negations. More needs to be done to persuade us internalist dogmatists that antecedent
warrant for Moore-3 is even needed for the higher-order claim that your experiences warrant
you in believing Moore-2.
36
Wright 2002, p. 345.
37
See the essays in Alston 1989, especially Essay 6; and Audi 1993, especially Chapters 4 and 11.
What may make the HOBW Model attractive is the hovering thought that
dogmatism can at best be an answer to first-order skepticism; that it’s going to let
second–order skeptics have run of the house.38 This thought goes: even if a lack of
perception-independent warrant to believe we’re not brains in vats doesn’t preclude us
from having perceptual warrants, it will preclude us from warrantedly believing we have
them. Dogmatists can at best give us “animal warrants” to believe we have hands; not
warrants that survive reflective ascent.
But I see no reason to concede that. Why can’t our immediate perceptual warrants
be available to us for settling higher-order questions, too?
What is true is that it’s easier to give an accounting of our first-order perceptual
warrants than it is to give an accounting of our higher-order warrants. Our first-order
warrants are constituted by everyday experiences. Our higher-order warrants are
constituted by lengthy philosophical discussions, of which the present discussion is
just one piece. But if dogmatism is correct, shouldn’t it be possible for us (eventually) to
be warranted in believing it is? (These philosophical discussions ought to lead some-
where!) Why can’t our immediate perceptual warrants be part of our (more dearly
won) warrant to believe dogmatism?
I suppressed the HOBW Model of transmission-failure until now for expository
convenience. It was useful for us to get clear on the prospects of the BW Model first.
But as I’ve just argued, I don’t think the HOBW Model necessarily changes our
dialectical position.
What I think will be more promising to consider are some other models, which try
to motivate obstacles to Moorean transmission without making claims about what
higher-order warrants we’re in a position to have. These other models may, on
some views, end up entailing the HOBW Model. But since they don’t rest on any
contested claims about what higher-order epistemic positions we’re in, they may be
dialectically more useful.
In the framework we’re working with, warrants are constituted by factors that count
in favor of believing something. Let’s call these warrant-making factors (on analogy
with the notion of “truth-makers”). Warrant-making factors may include things like:
having antecedent warrant to believe such-and-such premises. They may include
things like: having such-and-such experiences. On reliabilist theories, they will include
things like: having reliable vision. And so on.
This enables us to formulate one new model of transmission-failure.
Warrant-Making Factor (WMF) Model: If B is among the factors that make W be a warrant for A,
then your W-based warrant for A can’t transmit through reasoning to the conclusion that B obtains.39
38
See Wright 1985, p. 444; 1991, p. 89; and 2007, }II.
39
Compare Wright’s repeated insistence that warrant cannot transmit to the “external preconditions for
the effectiveness of your [belief-forming] method” (2000a, p. 154; 2002, p. 342; 2003, pp. 61, 70; see also his
remarks on “operational necessity” in the same papers). See also Davies 1998, p. 354 and 2000, p. 404; and
Principle QB in McLaughlin 2000.
This is different than the BW and HOBW Models, because B need not include your
having warrant to believe anything. B may instead be a fact like: your vision is reliable,
or: you have such-and-such experiences. For example, the WMF Model would count
this reasoning as transmission-failing:
[Visual experiences as of hands]
I see two hands.
I have visual experiences.
In the Gettier literature, one sees the term “defeat” used in such a way that defeating
facts merely have to obtain, in order to neutralize an epistemic status you’d otherwise
possess. I use the term differently. I only use “defeat” to describe a kind of warrant or
evidence subjects get. When warrant-makers give you prima facie warrant to believe
A, that warrant is vulnerable to being outweighed or undermined by other defeating
evidence. Let’s focus on the undermining kind of defeat. This is exemplified in the
following sort of case: you seem to see Tom at the library. Then you learn that Tom is
one of a set of identical triplets. Intuitively, this new information doesn’t tell you that
Tom wasn’t at the library. At least, not directly. It might on balance be reasonable for
you to reduce your confidence that Tom was there. But what the new information
more directly testifies to is that you weren’t in a position to visually tell that Tom was at the
library. In doing that, it undermines your initial visual warrant to believe Tom was there.
When a hypothesis U is such that evidence for it will undermine a warrant W for
A, I call U an undermining hypothesis for W. Such hypotheses lower the likelihood
that in having W you’d thereby be perceiving, or more generally, in a position to tell, that
A. On the other hand, I’ll call a hypothesis an anti-underminer with respect to W and
A when (i) it entails that some underminers of W’s support for A are false; and (ii) it’s
not itself such an underminer.
This enables us to formulate another new model of transmission-failure:
Anti-Underminer (AU) Model: If B is an anti-underminer with respect to W and A, then your
W-based warrant for A can’t transmit through reasoning to the conclusion that B.40
Things can be said to motivate each of these new models. As we saw in }III, above, it
can be natural even for a dogmatist to describe you as “taking for granted” that your
senses are reliable, and so on. And it can be natural to worry whether arguments can
40
See Wright’s characterization of his “triads” in 2007.
be effective whose conclusions are already “taken for granted,” or “presupposed,” in

that way.41
Though these proposals can be motivated, I do not think they can be correct. I have
counter-examples.
Consider:
Turing-1. Being able to grasp entailments is sufficient for being able to rationally
follow an argument.
Turing-2. Turing machines are able to grasp entailments.
Turing-3. So Turing machines are able to rationally follow arguments.
Never mind how we might establish Turing-2. Just assess whether this argument seems
to be objectionably question-begging. Most of my consultants judge it is not.
And yet, if you learned that you hadn’t properly followed the argument, that would
undermine how confident you ought to be in its conclusion (at least, on the basis of so
reasoning). If you learned that you weren’t capable of following the argument, all the
more so. Now it’s certainly epistemically possible that you are merely a Turing
machine, so if you acquired reason to believe (let us suppose, falsely) that mere Turing
machines are never able to rationally follow arguments, it should to some degree have
that undermining effect. What that means is that the conclusion of this argument is one
kind of anti-underminer for the warrant the argument gives you. (It’s an anti-under-
miner for the warrant you get by reasoning through or following the argument, not your
warrant for any of the premises.) So the AU Model would count this argument as
transmission failing. Yet intuitively, it seems like such reasoning should be okay.
If in fact you are merely a Turing machine, then the conclusion of the argument may
be not just an anti-underminer, but a warrant-making factor as well. In that case, the
WMF Model would also count the argument as transmission-failing. But again, it
seems like it ought not to be so counted.
Consider a different case. You’re looking in a zoo pen, and your mad scientist friend
proudly announces that any animals in the pen have been turned invisible. You inform
him he’s wrong: you see rats in the corner, so there are some visible animals in the pen
after all. Your reasoning might be represented like this:
Rat-1. [Visible experiences as of rats]

Rat-2. I see rats in the corner.
Rat-3. So there are some visible animals in the pen.
41
Wright usually uses the vocabulary of “presuppositions” to refer to antecedently warranted background
premises: see for example 2000a, p. 148; and his repeated references to presupposing warrants and entitlements:
2000a, p. 159; 2002, p. 345; 2003, pp. 64, 66, 75. See also Davies 1998, pp. 350–2, 354. However, as Wright
points out in 2007, }I, “presupposition” can also be understood in the more lightweight way being envisaged
here. See also the references in fn. 9, above; and Alston 1986, who says that when I form perceptual beliefs,
I’m “assuming” or “proceeding as if ” sense perception is reliable (pp. 323–4, 327–9), without needing to
have any antecedent warrant to believe that it is (pp. 331–2).
Assess whether this argument seems objectionably question-begging. Most of my

consultants judge it is not.
And yet, to the extent that you have reason to believe any animals in the pen are
invisible, your visual warrant to believe Rat-2 is undermined. So Rat-3 is an anti-
underminer for the warrant Rat-1 gives you to believe Rat-2. So the AU Model
would count this argument as transmission-failing, too. Yet intuitively, it seems like
such reasoning should be okay.
The natural move at this point will be to try to refine the transmission-failure models
to avoid these counter-examples. For example, we might try to replace the AU Model
with:
If B is an anti-underminer for W whose non-obtaining would give you the same experiences you
actually have then your W-based warrant for A can’t transmit through reasoning to the
conclusion that B.
This would no longer count the Rat argument as transmission-failing, since if the pen
did only contain invisible animals, we wouldn’t expect to have experiences of rats. It
would only impugn arguments like this:
Rat-1. [Visible experiences as of rats]

Rat-2. I see rats in the corner.
Rat-3*. So it’s not the case that: there are no visible animals in the pen and I am
merely hallucinating rats.
and that seems close enough to Moorean reasoning that we have no independent
intuition it should be okay.
Perhaps some such refinement can work. But I’m pessimistic, for three reasons.
In the first place, the refinement I articulated wouldn’t help with the Turing
example.
In the second place, what may make such refinements attractive are the probabilistic
considerations that I deferred for discussion elsewhere. Those probabilistic considera-
tions allegedly threaten the possibility of something’s warranting you in believing Q,
when not-Q’s truth would entail that you have the same warrant. I acknowledge the
prima facie force of those considerations; and I have responses to them; but they require
extended discussion. Our question here is whether the notion of transmission-failure
gives us any further diagnostic tool, or whether it’s just piggy-backing on the probabi-
listic considerations. I suspect that efforts to refine the AU Model in the way I described
will lead to our just piggy-backing.
As I said in section IV, I doubt that the probabilistic considerations really do reflect
our deepest-rooted intuitive resistance to Moorean reasoning, anyway, for we feel that
same resistance to arguments where the probabilistic considerations are not applicable.
My third doubt about the strategy of refining the AU Model is that it raises serious
motivational difficulties.
It was possible to motivate the AU and WMF Models in the first place because they
reflected notions of “taking for granted” or “presupposing” that we already had some
intuitive grasp of: it can seem natural to doubt whether arguments can establish their
own “presuppositions.” That can seem like a covert form of question-begging.
This motivation wouldn’t have been available if we didn’t have some folk under-
standing of anti-underminers or warrant-making factors as “presuppositions,” and had
just used some technical notion instead. It doesn’t seem natural to doubt whether
arguments can establish their own XYZZYs. The refined AU Model strikes me as
going down that path. I don’t think we have any intuitive notion of “presupposing”
such that the relation specified in the original AU Model and the relation specified in
the refined model count as different kinds of presupposing. No, the sense in which one can
naturally be counted as “presupposing” B in the refined model is the same as in the
original model; it’s just that in the refined model, B meets extra conditions besides. Yet
our counter-examples show that it’s a mistake to count “presupposing” your conclusion
in that sense as any form of begging a question. The proponent of the refined model
needs to scratch up some further motivating story. I’m not sure there is any (non-
probabilistic) such story.
Perhaps all that will be said for a refined story is that it extensionally captures the
kinds of arguments, like Moore’s, that one doesn’t like.
I understand and respect the view that, when B is the kind of conclusion we’re
considering, the subject’s warrant W for A can’t be immediate, but instead requires her
to have antecedent warrant to believe B. As a dogmatist, I don’t agree with that view.
But I recognize it as a viable theoretical option. In that case, Moorean reasoning would
fail because it exemplifies the simple BW Model.
If on the other hand we evaluate Moorean reasoning from the dogmatist’s perspec-
tive, it seems to me more promising to diagnose our intuitive resistance to the
reasoning as the product of illusion. I don’t see much money in the project of coming
up with (non-probabilistic) models of transmission-failure that impugn Moorean
reasoning even when our perceptual warrant is immediate.
Bibliography
Alston, W. 1989 Epistemic Justification, Ithaca, Cornell University Press
Audi, R. 1993 The Structure of Justification, Cambridge, Cambridge University Press.
Bergmann, M. 2004 “Epistemic Circularity: Malignant and Benign”, Philosophy and Phenomeno-
logical Research 69, pp. 709–27.
Bergmann, M. 2005 “Defeaters and Higher-Level Requirements”, The Philosophical Quarterly 55,
pp. 419–36.
Broome, J. 1999 “Normative Requirements” Ratio 12, pp. 398–419.
Burge, T. 1993 “Content Preservation”, Philosophical Review 102, pp. 457–88.
Burge, T. 1996 “Our Entitlement to Self-Knowledge”, Proceedings of the Aristotelain Society
XCVI, pp. 91–116.
Burge, T. 2003 “Perceptual Entitlement”, Philosophy and Phenomenological Research 67,

pp. 503–48.
Cohen, S. 1988 “How To Be a Fallabilist”, Philosophical Perspectives 2, pp. 91–123.
Cohen, S. 1999 “Contextualism, Skepticism, and the Structure of Reasons”, Philosophical
Perspectives 13, pp. 57–89.
Cohen, S. 2000 “Contextualism and Skepticism”, Philosophical Issues 10, pp. 94–107.
Conee, E. and Feldman, R. 2004 Evidentialism, New York, Clarendon Press.
Davies, M. 1998 “Externalism, Architecturalism, and Epistemic Warrant”, in C. Wright,
B. C. Smith, and C. Macdonald (eds.) Knowing Our Own Minds, Oxford, Oxford University
Press, pp. 123–54.
Davies, M. 2000 “Externalism and Armchair Knowledge”, in P. Boghossian and C. Peacocke
(eds.) New Essays on the A Priori, Oxford, Oxford University Press, pp. 384–414.
Davies, M. 2003a “The Problem of Armchair Knowledge”, in S. Nuccetelli (ed.) New Essays on
Semantic Externalism and Self-Knowledge, Cambridge, Mass., MIT Press, pp. 23–55.
Davies, M. 2003b “Externalism, Self-Knowledge, and Transmission of Warrant”, in M. Frápolli
and E. Romero (eds.) Meaning, Basic Self-Knowledge, and Mind: Essays on Tyler Burge. Stanford,
CSLI Publications, pp. 105–30.
Davies, M. 2004 “Epistemic Entitlement, Warrant Transmission, and Easy Knowledge”,
Proceedings of the Aristotelian Society Supp. Vol. LXXVIII, pp. 213–45.
Davies, M. 2009 “Two Purposes of Arguing and Two Epistemic Projects”, in Ian Ravenscroft
(ed.) Mind, Ethics, and Conditionals: Themes from the Philosophy of Frank Jackson, Oxford, Oxford
Dretske, F. 1970 “Epistemic Operators”, Journal of Philosophy 67, pp. 1007–23.
Dretske, F. 2000 “Entitlement: Epistemic Rights without Epistemic Duties?” Philosophy and
Phenomenological Research 60, pp. 591–606.
Field, H. 2000 “A Priority as an Evaluative Notion”, in P. Boghossian and C. Peacocke (eds.)
New Essays on the A Priori, Oxford, Oxford University Press, pp. 117–49.
Field, H. 2005 “Recent Debates about the A Priori”, Oxford Studies in Epistemology 1, pp. 69–88.
Ginet, C. 1980 “Knowing Less By Knowing More”, Midwest Studies in Philosophy 5, pp. 151–61.
Hawthorne, J. 2004 Knowledge and Lotteries, Oxford, Oxford University Press.
Klein, P. 1981 Certainty: Refutation of Skepticism, Minneapolis, University of Minnesota Press.
McKinsey, M. 2003 “Transmission of Warrant and Closure of Apriority”, in S. Nucetelli
(ed.) New Essays in Semantic Externalism and Self-Knowledge, Cambridge, Mass., MIT
Press, pp. 97–116.
McLaughlin, B. 2000 “Self-Knowledge, Externalism, and Skepticism”, Aristotelian Society
Supplemental Volume 74, pp. 93–117.
Meiland, J. 1980 “What Ought We to Believe? The Ethics of Belief Reconsidered”, American
Philosophical Quarterly 17, pp. 15–24.
Peacocke, C. 2004 The Realm of Reason, New York, Oxford University Press.
Plantinga, A. 1993a Warrant: The Current Debate, New York, Oxford University Press.
Plantinga, A. 1993b Warrant and Proper Function, New York, Oxford University Press, New York.
Pollock, J. 1983 “Epistemology and Probability”, Synthese 55, pp. 231–52.
Pryor, J. 2000 “The Skeptic and the Dogmatist”, Noûs 34, pp. 517–49.
Pryor, J. 2004 “What’s Wrong With Moore’s Argument?”, Philosophical Issues 14, pp. 349–78.
Pryor, J. 2006a “Hyper-Reliability and A Priority”, Proceedings of the Aristotelian Society CVI,
pp. 327–44.
Pryor, J. 2006b “More on Hyper-Reliability and A Priority”, available online at <http://www.
jimpryor.net/research/papers/More-Hyper.pdf>.
Pryor, J. (ms-a) “Problems for Credulism”, available online at <http://www.jimpryor.net/
research/papers/Credulism.pdf>.
Pryor, J. (ms-b) “Hypothethical Oughts”, available online at <http://www.jimpryor.net/
research/papers/Hypothetical.pdf>.
Silins, N. 2005 “Transmission Failure Failure”, Philosophical Studies 126, pp. 71–102.
Sturgeon, S. 2012 “Confidence and Coarse-Grained Attitudes”, Oxford Studies in Epistemology 4.
White, R. 2006 “Problems for Dogmatism”, Philosophical Studies 131, pp. 525–57.
Williamson, T. 2000 Knowledge and Its Limits, Oxford, Oxford University Press.
Wright, C. 1985 “Facts and Certainty”, Proceedings of the British Academy 71, pp. 429–72.
Wright, C. 1991 “Skepticism and Dreaming: Imploding the Demon”, Mind 100, pp. 87–115.
Wright, C. 2000a “Cogency and Question-Begging: Some Reflections on McKinsey’s Paradox
and Putnam’s Proof ”, Philosophical Issues 10, pp. 140–63.
Wright, C. 2000b “Replies”, Philosophical Issues 10, pp. 201–19.
Wright, C. 2002 “(Anti-)Sceptics Simple and Subtle: G. E. Moore and John McDowell”,
Wright, C. 2003 “Some Reflections on the Acquisition of Warrant by Inference”, in
S. Nuccetelli (ed.) New Essays on Semantic Externalism and Self-Knowledge, Cambridge, Mass.,
MIT Press, pp. 57–77.
Wright, C. 2004 “Warrant for Nothing (and Foundations for Free)?”, Proceedings of the Aristotelian
Society Supp. Vol. LXXVIII, pp. 167–212.
Wright, C. 2007 “The Perils of Dogmatism”, in S. Nuccetelli and H. Seay (eds.) Themes from
G. E. Moore: New Essays in Epistemology and Ethics, Oxford, Oxford University Press.
Zalabardo, J. 2005 “Externalism, Skepticism, and the Problem of Easy Knowledge”, Philosophical
Review 114, pp. 31–61.
11
Wright on Moore*
José L. Zalabardo
1. Transmission
Jim’s teacher has just given him his marked maths exam. Jim knows (because he is
looking at it) that his mark is 7.25 out of 22. He also knows (because the teacher just
said it) that the pass mark is 35%. Does Jim know he has failed? No, he doesn’t. Not yet.
As you would expect from his mark, Jim is not very good with numbers. He’ll need a
few minutes with pencil and paper to work out that 7.25 is less than 35% of 22. Only
then will he know that he has failed.
This case exemplifies a common and important phenomenon: someone recognizes
the validity of an inference from a set of premises that he knows, and in so doing he
acquires knowledge of the conclusion. Jim knows that his mark is 7.25 out of 22 and
that the pass mark is 35%. Then, by virtue of his calculation, he comes to recognize the
validity of the inference from these premises to the conclusion that he has failed,
thereby coming to know the sad truth.
It is undeniable that there are many cases, like Jim’s, in which recognizing the
validity of an inference from known premises brings about knowledge of the conclu-
sion, as this is the most natural characterization of what goes on when we acquire
knowledge by deductive inference. But can knowledge always be acquired in this way?
Does recognition of the validity of an inference from known premises always bring
about knowledge of the conclusion?
There are cases in which this method of knowledge acquisition is trivially ruled
out—namely cases in which the subject already knows the conclusion. If Jim’s exam
had been marked “7.25 out of 22: Fail” by his perfectly reliable teacher, and Jim had
formed the belief that he had failed upon reading this, he would thereby have come to
know that he had failed. So even if he had later performed his calculation, this wouldn’t
have resulted in knowledge acquisition. Notice that circular arguments fall under this
category automatically. If the conclusion is one of the premises, you can’t know the
premises without knowing the conclusion. Hence recognition of the validity of the
inference won’t furnish you with knowledge of the conclusion.
* I have benefited from discussion of these ideas with Mike Martin. I’m also grateful to Ross Ford.
WRIGHT ON MOORE 305
Is this the only way in which recognition of the validity of an inference from known
premises can fail to produce knowledge of the conclusion? This question is answered in
the affirmative by a principle which, following Crispin Wright, I shall call Transmission.1
Wright formulates the principle in terms of the notion of warrant. His conception of
warrant is loose and open-ended. In some contexts, he seems to treat it as interchange-
able with evidence, reasons, or grounds,2 but the possibility of warrant in the absence of
these is one of the most prominent features of his epistemology.3 It is unquestionable that
he sees the issue of the transmission of warrant as relevant to that of the acquisition of
knowledge by inference.4 This suggests that he regards warrant as playing an important
role in the contrast between knowledge and mere true belief, but he doesn’t provide an
explicit account of how he takes warrant and knowledge to be related. In this essay I am
going to simplify matters by providing an explicit definition of the former in terms of the
latter. I shall assume the tripartite conception of knowledge as a species of true belief, and
use the term warrant to denote the relation between a subject S and a proposition p whose
instantiation or otherwise when S truly believes that p will determine whether S knows
that p. I am also going to assume that warrant facts are not primitive—that whenever
S has warrant for p there is a different fact by virtue of which the warrant fact obtains.
I shall refer to the fact that plays this role in each case as the warrant-constituting fact. I want
to emphasize that this is unlikely to be the conception of warrant that Wright is
employing. Hence my conclusions will bear on Wright’s views only indirectly—to the
extent that his conception of warrant approximates mine.
We can now provide the formulation of Transmission that I am going to use in this
essay. For the sake of simplicity, I shall provide a formulation that covers only one-
premise inferences:
T1 If p is a proposition that S believes and for which S has warrant (and q is a proposition for
which S doesn’t have warrant), then by recognizing the validity of the inference from p to q,
S will acquire warrant for q.
Transmission plays an important but complex role in the assessment of some prominent
sceptical arguments. In a recent exchange, Fred Dretske and John Hawthorne have
used the label heavyweight for propositions that Hawthorne characterizes in the follow-
ing terms: p is a heavyweight proposition just in case “we all have some strong
inclination to say that p is not the sort of thing that one can know by the exercise of
reason alone and also that p is not the sort of thing that one can know by use of one’s
perceptual faculties (even aided by reason)”.5
1
See Wright, 2000, 2002, 2003.
2
See, e.g., Wright 2002, pp. 333–4.
3
Wright first put forward the idea in Wright 1985. The thought is developed in new directions in Wright
2004 where he adopts the term ‘entitlement’ to refer to non-evidential warrant.
4
See, e.g., Wright 2002, p. 332: “Intuitively, a transmissible warrant should make for the possible
advancement of knowledge, or warranted belief [ . . . ]”.
5
Cf. Hawthorne 2005, p. 33. See also Dretske 2005.
306 J O S É L . Z A L A B A R D O
The inclinations that define heavyweight propositions are exploited by a familiar

line of sceptical reasoning seeking to conclude that we can’t know either other
propositions that we are intuitively inclined to regard as easily knowable (call them
lightweight). The reasoning challenges the knowability of a lightweight proposition by
finding a heavyweight proposition entailed by it, and arguing that since the latter can’t
be known, the former can’t be known either. Thus, the thought goes, since you can’t
know that there is a material world, you can’t know that you have a broken fingernail,
or since you can’t know that the world didn’t come into existence five minutes ago,
you can’t know that you had eggs for breakfast this morning. I shall refer to sceptical
arguments based on these ideas as Cartesian scepticism.
The relevance of Transmission for the assessment of Cartesian scepticism lies in the
fact that, if Transmission holds, then we seem to have at our disposal a very simple
account of how we can obtain warrant for a heavyweight proposition H that the
sceptic singles out to challenge our knowledge of a lightweight proposition L: We can
acquire warrant for H by recognizing the validity of the inference from L to H. Thus,
by recognizing the validity of the inference from the premise that you have a broken
fingernail (which you might know by sense experience) to the conclusion that there is
a material world, you would be able to acquire warrant for this conclusion, and by
recognizing the validity of the inference from the premise that you had eggs for
breakfast this morning (which you might know by memory) to the conclusion that
the world didn’t come into existence five minutes ago, you could obtain warrant for
this conclusion. Let me refer to inferences from a lightweight proposition to a
heavyweight proposition entailed by it as Moorean inferences, since G. E. Moore’s
famous proof of an external world seeks to put such an inference to anti-sceptical use.6
However, the effectiveness of Moorean inferences against Cartesian scepticism is not
universally acknowledged, since many philosophers have a strong inclination to reject
the idea that we can come to know heavyweight propositions in this way. Let me refer
to those who object to the possibility of acquiring warrant for heavyweight proposi-
tions by recognizing the validity of Moorean inferences as anti-Mooreans. In order to
vindicate their position, anti-Mooreans need to provide a satisfactory explanation of
why this method of warrant acquisition fails, given that, if we had warrant for the
premise of a Moorean inference, and Transmission held, recognition of the validity of
the inference would seem to result ineluctably in the acquisition of warrant for its
conclusion.
There are two explanations of this failure which, I take it, the anti-Moorean should try
to avoid. One is the sceptic’s. For the sceptic, the reason why recognizing the validity of a
Moorean inference never provides us with warrant for its heavyweight conclusion is that
we don’t have warrant for its lightweight premise, as would be required for Transmission
to apply. The other is the wholesale rejection of Transmission, which seems to jeopardize
6
See Moore 1939.
WRIGHT ON MOORE 307
the possibility of acquiring knowledge by deductive inference. The only obvious way of
avoiding these two unpalatable extremes would be to argue that Transmission holds in
some cases—those in which knowledge acquisition by deductive inference seems un-
problematic—but fails in others—including Moorean inferences. Notice, however, that
the anti-Moorean can’t simply stipulate that Transmission applies in all cases except in
Moorean inferences, as this would provide no explanation of why knowledge acquisition
doesn’t ensue in these cases. The anti-Moorean who pursues this line needs to provide a
principled way of drawing the distinction between legitimate and illegitimate applica-
tions of Transmission. She needs, in effect, to provide an instance of the schema:
If p is a proposition that S believes and for which S has warrant (and q is a proposition for which
S doesn’t have warrant), then by recognizing the validity of the inference from p to q, S will
acquire warrant for q just in case X.
I shall refer to the just in case X clause in this schema as the limitation clause.7 The
resulting principle will have to satisfy two conditions. First, it has to be extensionally
adequate, i.e. condition X has to be violated by Moorean inferences, but satisfied by
intuitively legitimate cases of inferential warrant acquisition. Second, it has to render
it intelligible that warrant acquisition is possible in the latter cases but not in the
former.
This anti-Moorean agenda has been largely set by Crispin Wright in an influential
series of papers.8 My goal here is to provide a critical assessment of some aspects of this
project. I shall consider first some difficulties faced by some limitation clauses that
Wright can be read as advocating. Then I shall argue that the anti-Moorean might not
need limitation clauses—that in the presence of certain plausible assumptions, the anti-
Moorean should have no trouble combining the possibility of knowing lightweight
propositions with the unrestricted applicability of Transmission. I shall end by consid-
ering a parallel issue that arises when we shift our attention from the transmission of
warrant to the transmission of evidential support.
2. Wright’s Limitation Clauses

It is not easy to find a formulation of a limitation clause that can be unquestionably
attributed to Wright. This is due to a large extent to the fact that he has presented his
ideas in connection with several different issues, including the tension between
semantic externalism and privileged access, Hilary Putnam’s proof that we are not
envatted brains, G. E. Moore’s proof of an external world, and the disjunctivist
response to scepticism.9 Rather than trying to defend the attribution to Wright of a
7
Martin Davies has used the term ‘limitation principle’ to refer to principles specifying conditions under
which warrant transmission would not be effected. See, e.g., Davies 1998.
8
See Wright 1985 the papers cited in fn. 1.
9
The first two are the main focus of Wright 2000. The last two are discussed in Wright 2002.
specific limitation clause, I am going to consider the general phenomenon with which
he has sought to motivate his proposals, and raise some issues for the limitation clauses
that emerge from this approach.
Wright’s basic idea is that the adequacy for warrant transmission of an inference from
p to q will be undermined when the truth of q, or the subject’s warrant for q, is
somehow presupposed by the subject’s warrant for p. Perhaps the version of this
thought most strongly suggested by Wright’s work yields a limitation clause dictating
that recognition of the validity of an inference from a warranted premise won’t give the
subject warrant for the conclusion when the subject’s warrant for the premise requires
that the subject has independent warrant for the conclusion.
We can provide a formulation of this proposal based on our assumption that warrant
facts are not primitive. According to the resulting limitation clause, recognition of the
validity of an inference from a warranted premise won’t give the subject warrant for the
conclusion when the constituting fact for the subject’s warrant for the premise requires
that the subject has independent warrant for the conclusion. This clause generates the
following amended version of Transmission:
S will acquire warrant for q just in case the constituting fact of S’s warrant for p doesn’t require
that S has independent warrant for q.10
It has been argued that this principle would rule out the possibility of acquiring warrant
for a heavyweight proposition by recognizing the validity of a Moorean inference only
on a questionable conception of warrant.11 Consider the Moorean inference from the
premise that I have a broken fingernail to the conclusion that there is an external world.
In order for this inference to fall under the limitation clause of T2, it would have to be
the case that the fact that constitutes my warrant for the proposition that I have a
broken fingernail requires that I have independent warrant for the proposition that
there is an external world. But on most contemporary accounts of warrant, my warrant
10
There is a weaker and a stronger reading of the notion that a fact F requires that a subject S has warrant
for a proposition p. On the weaker reading, what this amounts to is that S’s having warrant for p is a necessary
condition for F, or, equivalently, that F is a sufficient condition for S’s having warrant for p. It is unlikely that
this is the right reading of the notion as it figures in the limitation clause that I am tentatively attributing to
Wright, since it doesn’t grasp the idea that what the warrant-constituting fact requires is warrant for the
conclusion that is independent. On the weak reading, the warrant-constituting fact for the premise will count
as requiring that the subject has warrant for the conclusion when it is also a warrant-constituting fact for the
conclusion. Thus suppose that the fact that constitutes S’s warrant for the proposition that it is raining also
constitutes her warrant for the proposition that there is precipitation. Then the fact that constitutes the
subject’s warrant for the former requires, in this sense, that the subject has warrant for the latter. On the
stronger reading, F requires that S has warrant for q only when S having warrant for q is an integral part of
F. That is, F is of the form F' & S has warrant for q, where F' doesn’t constitute S’s warrant for q. Then S’s
warrant for q—its constituting fact—will have to be independent of F. The limitation clause that I am
attributing to Wright calls for a reading along these lines.
11
See Brown 2003; Pryor 2004.
WRIGHT ON MOORE 309
for the former is constituted by facts that neither include my having warrant for the
latter nor give me warrant for it.
The point can be illustrated by considering a reliabilist account of warrant, although
it applies to a wide range of positions. On the reliabilist account, if I formed the belief
that I have a broken fingernail as a result of the operation of my sensory devices, and
my sensory devices are reliable, then this fact constitutes my warrant for the proposition
that I have a broken fingernail. But this warrant-constituting fact doesn’t seem to
require (either in the weak or the strong sense of the notion, see fn. 10) that I have
warrant for the proposition that there is an external world. It follows that this Moorean
inference would not fall under the limitation clause of T2, and hence that T2 would
sanction the acquisition of warrant for the proposition that there is an external world
through recognition of the validity of this Moorean inference. If this point is accepted,
we will have to conclude that T2 is too strong (i.e. its limitation clause too weak) to
serve the anti-Moorean’s purposes, as it fails to rule out the acquisition of knowledge of
heavyweight propositions by recognizing the validity of Moorean inferences.
I am very sympathetic to this point. However, here I want to pursue a different line
of reasoning. I want to argue that when we read T2 in terms of the notion of warrant
with which I am operating, we have at our disposal a more direct route to the
conclusion that T2 is not sufficiently weak. Let L be a lightweight proposition for
which subject S has warrant, let F be the fact that constitutes S’s warrant for L, and let
H be a heavyweight proposition entailed by L. Now, suppose that F requires that S has
warrant for H. It follows that F is a sufficient condition for S having warrant for
H. Then we have that the fact that gives S warrant for L entails that S also has warrant
for H, and hence that S’s acquiring warrant for H by recognizing the validity of the
inference from L to H is trivially excluded: S cannot acquire warrant for H in this way
because she already has it.
What this reasoning suggests is, in effect, that on the conception of warrant with
which I am operating, T2 is not weaker than T1. Every putative case of warrant
transmission that is excluded by T2 would also be excluded by T1. For if the fact
that constitutes S’s warrant for L entails that S has warrant for H, then it can’t be the
case that S has warrant for L but not for H, as required for the applicability of T1.
Therefore, T2 is equivalent to T1. This result would only fail to obtain on a concep-
tion of warrant according to which it is possible for a subject who has a warrant for a
proposition to acquire another warrant for that same proposition, or for a subject who
has some warrant for a proposition to acquire more warrant for that same proposition.
There are no doubt properties for which the term ‘warrant’ could be used which
exhibit these features. But the conception of warrant with which I’m operating doesn’t
make room for these possibilities. On this conception, warrant is a relation between
subjects and propositions that a subject either bears or doesn’t bear to a proposition. If
S bears this relation to p, then there is no sense in which S could come to bear this
relation to p ‘twice over’ or ‘to a greater extent’. It could happen though that S has
warrant for p by virtue of warrant-constituting fact F1, and another fact F2 comes to
obtain with the power to constitute the fact that S has warrant for p, so that even if F1
didn’t obtain S would still have warrant for p by virtue of F2. We shall consider later on
whether Moorean inferences can supply us with alternative sources of warrant for their
conclusions in this way.
As I mentioned above, the claim that having warrant for a lightweight proposition
requires having warrant for the heavyweight propositions that it entails is rejected by
many contemporary epistemologists. However, there is another relation between the
conclusion of a Moorean argument and warrant for its premise that is unquestionably
exemplified. Typically, if H is a heavyweight proposition that figures in a Cartesian
sceptical challenge to S’s knowledge of a lightweight proposition L, we have that if
H were false S wouldn’t have warrant for L. In other words, if F is the fact that
constitutes S’s warrant for L, then F entails H. If my warrant for the proposition that
I have a broken fingernail is supposed to result from the operation of my sensory
devices, then if there were no external world, not only would I have no broken (or
unbroken) fingernails. In addition, I would have no sensory devices whose operation
could supply me with warrant for the proposition that I have a broken fingernail. The
fact that constitutes my warrant for the proposition that I have a broken fingernail
entails that there is an external world. We can use this idea to formulate a revised
transmission principle:
S will acquire warrant for q just in case the constituting fact of S’s warrant for p doesn’t entail q.
Is this the transmission principle that the anti-Moorean is looking for? Defending an
affirmative answer to this question would require showing that the putative cases of
warrant acquisition that it excludes are illegitimate. One way one could try to discharge
this task would be to argue that when the warrant-constituting fact for the premise
entails that the conclusion is true, it requires that the subject has independent warrant
for the conclusion.12 But this argument for T3 would face problems similar to those
that afflicted T2, as the accounts of warrant that are incompatible with T2 are also
incompatible with this defence of T3.
T3 has been challenged on the grounds that some of the instances of inferential
warrant acquisition that it excludes seem intuitively legitimate.13 These challenges accuse
the principle of being too restrictive. Here I want to pursue a different line of reasoning.
I am going to argue that T3 is not restrictive enough—that it sanctions instances of
inferential warrant acquisition that the anti-Moorean would want to proscribe.
We can make the point with the help of Dretske’s zebra example.14 Suppose that
Fred goes to the zoo and upon seeing the animals in the enclosure marked ‘zebras’ he
12
Sometimes Wright seems to advance this line of reasoning. See, e.g., Wright 2002, pp. 342–3.
13
See Davies 1998, p. 352; Pryor 2004, pp. 358–9.
14
See Dretske 1970.
WRIGHT ON MOORE 311
forms the belief that they are zebras. If we use once more the reliabilist account of
warrant as an illustration, if the perceptual devices with which Fred has formed this
belief are reliable, then he will have warrant for the proposition that the animals are
zebras, and if this proposition is true, then Fred’s belief will have the status of
knowledge. Suppose now that Fred comes to recognize the validity of the inference
from the proposition that the animals are zebras to the proposition that they are not
cleverly disguised mules. I take it that the anti-Moorean won’t want to treat this as a
legitimate case of inferential warrant acquisition.15
The question that we need to ask is whether this case would fall under T3’s
limitation clause. It should be clear that it doesn’t. The fact that Fred formed his belief
with reliable perceptual devices does not entail that the animals he is looking at are not
cleverly disguised mules. It does entail that such deception is sufficiently rare, but not
that it never occurs, or that it is not occurring on this occasion. It follows that the fact
that constitutes his warrant for the premise of Fred’s inference could obtain even if the
conclusion were false. Therefore, T3 would sanction Fred’s piece of reasoning as a
legitimate case of warrant acquisition. By recognizing the validity of the inference, Fred
would acquire warrant for the conclusion.16
Other cases that the anti-Moorean would want to proscribe do fall under T3’s
exclusion clause. Thus, for example, one could argue that the inference from the
proposition that the animals in the enclosure are zebras to the conclusion that not all
the zebra-looking animals in the world are cleverly disguised mules would be ruled out
by the principle. For if Fred lived in an environment in which this conclusion was false,
his zebra-detecting sensory devices would be highly unreliable. This might be seen as a
partial vindication of the principle, since it excludes some of the cases that the anti-
Moorean wants to rule out. Against this, however, one could argue that a satisfactory
diagnosis of the pathology afflicting the inference from the premise that the animals are
zebras to the conclusion that not all the zebra-looking animals in the world are cleverly
disguised mules should mention a deficiency that is also present in the inference from
that premise to the conclusion that the animals in the enclosure are not cleverly
disguised mules. The fact that T3 treats the latter inference as acceptable should be
seen as undermining the claim that the principle accurately identifies the source of the
problem afflicting the former.17
15
Wright clearly wants to exclude this kind of case. See Wright 2002, pp. 342–44.
16
The falsity of the conclusion would entail that Fred doesn’t know the premise, since it would entail that
the premise is false, but this observation doesn’t give rise to a plausible limitation clause, since proscribing
inferential knowledge acquisition whenever the negation of the conclusion entails the negation of the
premise would rule out all instances in which the inference is valid. See Pryor 2004, p. 358.
17
Notice that, on other conceptions of warrant, T3 would sanction even Moorean inferences with the
negations of traditional sceptical hypotheses as conclusions. Take, e.g., Jim Pryor’s dogmatism, conceived as
an account of warrant. If my warrant for the proposition that I have a broken fingernail is constituted by the
fact that I have the phenomenology of seeming to ascertain with my perceptual episodes that this proposition
is true, presumably this warrant-constituting fact does not entail that there is an external world, since an evil
demon victim would have the same phenomenology. See Pryor 2000, 2004, pp. 356–7.
3. Transmission and Closure

In the preceding section I have argued that the features of the relationship between
premise and conclusion of a Moorean inference on which Wright has focused won’t
easily yield a limitation clause that serves the anti-Moorean agenda. I don’t claim to
have delivered a fatal blow to the project. Perhaps the difficulties that I have raised can
be overcome with a more sophisticated account. And even if this line of thought didn’t
work, other authors have approached the task of finding limitation clauses from points
of view that fall outside the scope of my discussion so far.18
I am not going to pursue further the assessment of specific limitation clauses. Rather,
in this section I am going to make a suggestion that might enable the anti-Moorean to
call off the search for a limitation clause. I am going to argue that, in the presence of a
principle that Wright and many others accept, together with some plausible assump-
tions, endorsing transmission in its unrestricted form (T1) doesn’t force us to condone
the instances of warrant acquisition that the anti-Moorean wants to rule out.
The saving principle is Closure. Thanks to Wright we have a clear understanding of
the distinction between Transmission and Closure.19 Transmission, as we have seen, is
a principle concerning a specific method of warrant acquisition—namely, the acquisi-
tion of warrant for the conclusion of a valid inference with known premises as a result
of recognizing the validity of the inference. Closure does not concern this or any other
specific method of warrant acquisition. What Closure dictates is simply that certain
warrant claims about a subject are incompatible. It says that if p entails q, then we can’t
ascribe to a subject warrant for p and for the proposition that p entails q while not
ascribing to her warrant for q. We can formulate the principle as follows:
If p entails q, and S has warrant for p and for the proposition that p entails q, then S has warrant
for q.
Wright has expressed unambiguously his support for Closure, and in spite of some
distinguished detractors, the principle enjoys widespread support.20 Here I am not
going to defend Closure. My goal will be to argue that, if the principle is true, then the
anti-Moorean might have no need to limit the applicability of Transmission.
The basic idea is very simple. Let L be a lightweight proposition, and H a heavy-
weight proposition entailed by L. If Closure holds, then someone who has warrant for
the proposition that L entails H won’t be able to acquire warrant for H by recognizing
the validity of the inference from L to H. For either she doesn’t have warrant for L, or
she does, in which case, by Closure, she also has warrant for H. Either way, she doesn’t
satisfy the antecedent of T1, and recognition of the validity of the inference from L to
H will fail to provide her with warrant for H.
18
See Davies 1998, 2003.
19
See, e.g., Wright 2002, pp. 331–2.
20
The main dissenting voices are, of course, Fred Dretske and Robert Nozick. See Dretske 1970 and
Nozick 1981, Chapter 3.
WRIGHT ON MOORE 313
Hence, in order to show that recognition of the validity of the inference from
L to H can’t provide a subject with warrant for H, it will suffice to show that,
prior to recognizing the validity of the inference, the subject already has warrant for
the proposition that L entails H. If the anti-Moorean can show that this is the
situation in each of the cases that she wants to exclude, she will no longer need a
limitation clause, and she will be able to embrace transmission in its unrestricted
form (T1).
The possibility that a subject has warrant for an entailment even though she hasn’t
recognized the validity of the corresponding inference shouldn’t be problematic.
I think it is natural to say that I have warrant, say, for the proposition that the Axiom
of Choice entails Zorn’s Lemma, since the result is stated in every set theory manual
I’ve checked, and is accepted by every competent logician I know. However, it would
be odd to say that I recognize the validity of the inference from Choice to Zorn’s
Lemma, since I have never worked through a proof and I have no intuitive under-
standing of why the former necessitates the latter.
Furthermore, validity recognition is a conscious episode that involves consciously
entertaining the proposition that the premise entails the conclusion. However, it is
widely accepted that we can believe, and have warrant for, propositions that we
haven’t consciously entertained, and I can see no reason to reject this possibility
when the proposition in question has the form of an entailment.
However, for the anti-Moorean it is not enough to argue that warrant for an
entailment is in principle possible when the subject hasn’t recognized its validity.
What she needs to argue is that this is the situation for every Moorean inference to
which she takes exception. In the remainder of this section I want to sketch a line of
reasoning that might lead to this result.
If p entails q, let’s say that the entailment from p to q is epistemically transparent (or that
p transparently entails q) just in case anyone who possesses the concepts that figure in
p and in q has warrant for the entailment. Clearly, if the entailment from p to q is
epistemically transparent, then anyone who is in a position to recognize the validity of
the inference from p to q will already have warrant for it, since such a subject will have
to possess the concepts that figure in p and in q. This means that the anti-Moorean
could secure the result she needs if she could argue that in every objectionable
Moorean inference the premise transparently entails the conclusion.
I am not going to provide a full defence of this claim, but I want to suggest that the
thought is perfectly plausible. Take, for example, the proposition that I have a broken
fingernail and the proposition that there is an external world. I think it is natural to
suppose that anyone who has the concepts that figure in these propositions will believe
that the former entails the latter—whether or not she has consciously entertained the
entailment, let alone recognized its validity. For it seems to be a condition on having
these concepts that one has the belief that connects them in this way. And I think it is
just as natural to say that, in normal circumstances, someone who has these concepts
knows the entailment—that warrant for the entailment is a direct consequence of the
fact that the subject possesses the concepts that figure in it, even if the subject hasn’t
recognized its validity, or even consciously entertained it.21
I think that a similar claim can be made with respect to other Moorean inferences
that have played a prominent role in sceptical debates. If you have the concepts that
figure in the proposition that the seaweed was washed up by the tide some hours ago
and in the proposition that the world didn’t come into existence five minutes ago, then
you have warrant for the proposition that the former entails the latter. And if you have
the concepts that figure in the proposition that all As are Bs and in the proposition that
some properties are invariably co-instantiated with others, then you have warrant for
the corresponding entailment. And the verdict is equally plausible in Dretske’s cases. If
you have the concepts that figure in the propositions that the animals in the enclosure
are zebras and that they are not cleverly disguised mules, then you have warrant for the
proposition that the former entails the latter. And if you have the concepts that figure
in the propositions that the wall is red and that the wall is not white but lit by a red
light, then you have warrant for the corresponding entailment.
These remarks hardly add up to a satisfactory defence of the claim that in every
objectionable Moorean inference the entailment from premise to conclusion is episte-
mically transparent. My main contention is that if this claim were correct, then the
anti-Moorean who subscribes to Closure would have no need for limitation clauses.
Even under T1, the Moorean inferences to which she objects would fail to sustain the
acquisition of warrant for their conclusions. Let me summarize how this result is
reached. Let L be a lightweight proposition, and let H be a heavyweight proposition
transparently entailed by L. If S has the concepts that figure in L and H, then she has
warrant for the proposition that L entails H, and then, by Closure, if she has warrant for
L she also has warrant for H. Hence, either she doesn’t have warrant for L or she has
warrant for H. Either way she doesn’t satisfy the antecedent of T1, and hence she
wouldn’t come to have warrant for H by recognizing the validity of the inference from
L to H.
The anti-Moorean strategy that I have outlined might seem to play into the hands of
the sceptic. For once we have accepted Closure and the epistemic transparency of
problematic Moorean inferences, we are forced to accept that you cannot have warrant
for the lightweight premise of an objectionable Moorean inference unless you have
warrant for its heavyweight conclusion. And this is all the sceptic needs to run her
familiar line of reasoning from the unknowability of heavyweight propositions to the
unknowability of the lightweight propositions that entail them.
21
Paul Boghossian has suggested that whenever belief in a proposition is required for having the concepts
that figure in it, the proposition might have warrant for the subject who has the concepts, and hence the
belief. He writes: “There may be beliefs that are such that having those beliefs is a condition for having one of
the concepts ingredient in them. Any such belief, it seems to me could plausibly be claimed to be default
reasonable. For if it really were part of the possession condition for a given concept that to possess it one had
to believe a certain proposition containing it, then that would explain why belief in that proposition is at least
presumptively (though defeasibly) justified”—Boghossian 2000, p. 240.
WRIGHT ON MOORE 315
In one respect, this complaint is incontestable. As soon as the anti-Moorean who

follows this strategy accepts that the conclusion of an objectionable Moorean inference
can’t be known, she will have to accept that the premise can’t be known either. But
this leaves the anti-Moorean one route open to try and avoid widespread scepticism.
She can avoid this outcome if she can resist the claim that the heavyweight propositions
on which the sceptic focuses are unknowable.
The prospects for this line of thought turn on the extent to which warrant requires
evidence, since the sceptic’s defence of the unknowability of the propositions on which
she focuses rests on the contention that we cannot obtain adequate evidence for
them.22 Hence, if the anti-Moorean could show that one can have warrant for these
propositions in the absence of adequate evidence, she would be able to stop the
sceptic’s reasoning on its tracks.23 The idea that warrant for the negations of sceptical
hypotheses might not require evidence has long been a central ingredient of Wright’s
epistemology. Thus, in “Facts and Certainty” he suggests that we might be able to
overcome Cartesian scepticism “if it could be reasonable to accept a group III proposi-
tion [roughly, the negation of a sceptical hypothesis] without reason; that is, without
evidence”.24 His proposal at the time as to how this situation might come about
involved the idea of treating these propositions as “partly constitutive” of the concepts
that figure in them.25 More recently he has explored other possible accounts of the
phenomenon.26 In any case, the possibility of warrant in the absence of evidence is a
straightforward consequence of mainstream externalist epistemologies, since they
typically postulate sufficient conditions for knowledge that can be exemplified when
the subject has no evidence. Once this possibility is accepted in principle, one could
argue that some heavyweight propositions prominent in sceptical debates are particu-
larly suited for this treatment.27
I’d like to end this section by returning to a point that I mentioned briefly earlier on.
A critic might concede that I have successfully shown that recognition of the validity of
a Moorean inference can never enable a subject to acquire warrant for its conclusion, so
to speak, de novo, since the conditions that make this method of warrant acquisition
possible obtain only when the subject already has warrant for the conclusion. Never-
theless, the thought goes, nothing that I have said undermines the possibility that by
recognizing the validity of a Moorean inference, a subject obtains an additional source
22
In Zalabardo 2006, I have attacked what many have seen as the most effective line of support for the
idea that knowledge requires evidence.
23
For this approach to Cartesian scepticism, see Vogel 2000, p. 606.
24
Wright 1985, p. 450.
25
Ibid., p. 453.
26
Cf. Wright 2004.
27
Cf. my Zalabardo 2005, }4, where I offer an explanation of why this might be so. I am not suggesting
that this move can provide a satisfactory solution to sceptical problems, but I do think that the only hope for
keeping them alive is to construe them, as Wright suggests, “not as directly challenging our having any
warrant for large classes of our beliefs but as crises of intellectual conscience for one who wants to claim that
we do”—Wright 2004, p. 167.
of warrant for the conclusion, over and above the warrant she already had, and this
possibility is certainly against the spirit of the anti-Moorean position.
This objection can be easily rebutted. If a subject S who has a source of warrant A for
a proposition p is to count as having acquired an additional source for warrant B for p,
we would expect the following counterfactual to hold: if A hadn’t obtained, but B still
obtained, then S would still have warrant for p. Hence, if S’s recognition of the validity
of the inference from L to H is to provide S with an additional source of warrant for H,
it would have to be the case that, even if S had not had warrant for H prior to the
episode of validity recognition, this episode would still give S warrant for H. But
according to the anti-Moorean strategy that I have presented this counterfactual is
clearly false. For if S didn’t have pre-existing warrant for H, then, by Closure, she
wouldn’t have warrant for L either, in which case recognizing the validity of the
inference from L to H would not furnish her with warrant for H.
4. Evidence Transmission
So far I have been concerned exclusively with the question, whether recognition of
the validity of a Moorean inference with a known premise can give us knowledge of
its conclusion. This is not the only issue under discussion in the recent literature on
Moorean inferences. Another prominent issue in these debates is whether a piece of
evidence can support the premise of a Moorean inference while not supporting its
conclusion, for a subject who knows that the former entails the latter. This is the
issue on which James Pryor has focused in a recent paper. Pryor formulates the issue
in terms of the notion of justification, but he construes justification as “the quality
that hypotheses possess for you when they are epistemically likely for you to be
true”,28 and it is natural to say that a piece of evidence supports a hypothesis for you
when it increases the extent to which the hypothesis has this quality for you. Pryor
writes:
Pre-reflectively, it seems like Moore’s perception of his hands should give him more justification
to believe he has hands than he’d have without it. And we grant that hands are external objects
(and that Moore knows them to be so). Yet many are reluctant to accept that Moore’s perception
of hands gives him more justification to believe there’s an external world. Why do we hesitate? If
something gives you justification to believe P, and you know P to entail Q, then shouldn’t it give
you justification to believe Q, too?29
Let me refer to any case in which a piece of evidence E provides a subject S with
support for p but not for q, even though S knows that p entails q, as a case of evidence
transmission failure. The main goal of Pryor’s paper is to argue that, when p and q are the
28
Pryor 2004, p. 352.
29
Ibid. p. 350. Notice that Wright has explicitly denied that this is the issue with which his discussion of
Moorean inferences is concerned. Cf. Wright 2002, pp. 334–5.
WRIGHT ON MOORE 317
premise and conclusion of a Moorean inference that S knows to be valid, and E is a

typical piece of evidence for p, evidence transmission doesn’t fail—i.e. E also supports q
for S:
I think that Moore does have perceptual justification to believe he has hands, and I think his
justification to believe that does transmit to the hypothesis that there’s an external world. So
Moore can acquire justification to believe there’s an external world by having experiences of
hands and reasoning in the way he does.30
And again:
I think you genuinely do get justification to believe the external world exists from your
perceptual justification to believe hands exist.31
Pryor takes issue with a line of reasoning that he finds in Wright and Davies for the
view that transmission fails in these cases—roughly the thought that in these cases your
evidence for the premise will only do its job if you have antecedent evidence for the
conclusion. I find Pryor’s attack’s on this line of reasoning highly convincing. What
I don’t see is that the shortcomings of the Wright/Davies defence of evidence
transmission failure force us to accept Pryor’s conclusion. My goal in this section is
to provide an account of why evidence transmission might fail in these cases which
doesn’t rest on any of the assumptions to which Pryor takes exception.
Suppose that a piece of evidence E makes a proposition p more likely for you to be
true, and that you know that p entails another proposition q. Does it follow that
E makes q more likely for you to be true? There is a fallacious line of reasoning that
makes an affirmative answer to this question seem irresistible. Let’s represent how likely
a piece of evidence X makes a proposition ç for a subject S as a rational number
between 0 and 1, PS(ç/X). We can reason as follows. If S knows that p entails q, we
have that Ps(q/p) = 1. Then for every piece of evidence X, PS(q/X) PS(p/X). Hence,
in particluar, PS(q/E) PS(p/E). Therefore, E makes q at least as likely to be true for
S as it makes p to be true for S, and if E makes p more likely to be true for S, it also
makes q more likely to be true for S.
But this line of reasoning is incorrect. The problem is that it rests on an over-
simplified account of when a piece of evidence makes a hypothesis more likely to be
true for a subject.
Consider the following case. You are told that a box contains ten objects, that half of
them are red and that three of the red objects are balls. Now, one of the objects is taken
from the box. If you were now told that the object is red (R), would this make it more
likely for you that the object is a ball (B)? As I have described the case, the answer is
surely ‘yes’. Before you are told that the object is red, the information at your disposal
doesn’t make the object more likely to be a ball than not. But the evidence that the
30 31
Pryor 2004, p. 351. Ibid., p. 352.
object is red does make it more likely to be a ball than not, since PS(B/R) = 0.6. Hence
R makes B more likely for you to be true.
However, consider now a slightly different case. This time, in addition to all the
information that you were given in the previous case, you are told that seven of the
objects in the box are balls. Suppose, as before, that an object is taken from the box. We
can ask again the same question: if you were told that the object is red, would this
make it more likely for you that the object is a ball? I want to argue that in this case
the question would have to be answered in the negative. Before you are told that the
object is red, there is for you a seven in ten chance that the object is a ball. If you were
told that the object is red, the probability for you of the object being a ball would
decrease—to six in ten. Hence the information that the object is red would not make it
more likely for you that it is a ball. In fact, it would make more likely for you that the
object is not a ball, since upon learning that the object is red, the probability for you of
its not being a ball would increase from three in ten to four in ten.
What this suggests is that when a subject S already assigns a probability to a
hypothesis H, prior to acquiring a piece of evidence E, whether E would make
H more likely to be true for S is not determined exclusively by how likely E makes
H for S – PS(H/E). It is determined, rather, by how this likelihood relates to the
likelihood of H for S prior to acquiring E – PS(H). E makes H more likely for S to be
true just in case PS(H/E) > PS(H).32
Once we adopt this construal of when E makes H more likely for S, the argument
that I provided against evidence transmission failure can no longer be run. Now the
assumption that E makes p more likely to be true for S is cashed out as PS(p/E) > PS(p).
Since S knows that p entails q, we have, as before, that PS(q/E) PS(p/E). This entails,
of course, that PS(q/E) > PS(p), but it doesn’t follow from this that E makes q more
likely to be true for S. For that we would need to show that PS(q/E) > PS(q), but we
have no way of establishing this. Our assumptions are perfectly compatible with the
possibility that PS(q/E) PS(q). Whenever this situation obtains, we’ll have a case of
evidence transmission failure.
Let’s review now how these considerations apply to some of the cases that interest
us. Take, first, Dretske’s zebra case. Fred’s zebra-like sense impressions will make the
proposition that the animals are zebras more likely to be true for him just in case the
likelihood for him that this proposition is true, given these sense impressions, is higher
than the likelihood for him that the proposition is true prior to having these sense
impressions. And it is hard to deny that his sense impressions satisfy this condition.
Suppose that Fred goes to the zoo and sees an enclosure covered with a screen. It is
32
This construal corresponds to the notion that Earman labels ‘incremental confirmation’ (see Earman,
1992, pp. 66–7). As Earman points out (p. 67), it is equivalent to what he calls the ‘likelihood criterion’, i.e.
PS(E/H) > PS(E) (provided that none of the probabilities involved are zero). My construal derives from a
familiar measure of evidential support first put forward, as far as I know, by Carnap, in the preface to the
second edition of Carnap 1962 , pp. xv–xvii—i.e. PS(H/E) – PS(H). E makes H more likely for S to be true,
on my construal, when this function has a positive value.
WRIGHT ON MOORE 319
undeniable that, whatever probability he ascribes at that point to the animals behind
the screen being zebras, it would be increased if the screen was removed and he had
zebra-like sense impressions. However, consider now the situation with regard to the
proposition that the animals in the enclosure are not cleverly disguised mules. We can
expect, given his background information, that the probability that he assigns to this
proposition before the screen is removed is very high. However, and this is the crucial
point, when the screen is removed and he receives zebra-like sense impressions, the
likelihood for him of the proposition being true would not increase. In fact, we can
expect this probability to decrease—even if only slightly. After all, the animals
wouldn’t be cleverly disguised mules unless they produced zebra-like sense impres-
sions. Hence such sense impressions rule out only possibilities that would make it true
that the animals are not cleverly disguised mules (e.g. elephant-looking elephants,
tiger-looking tigers . . . ). Therefore, under normal circumstances, we can expect Fred’s
zebra-like sense impressions to make the proposition that the animals in the enclosure
are zebras more likely to be true for him, but not to have this consequence for the
proposition that they are not cleverly disguised mules, even though he knows that
the former entails the latter. Dretske’s case exhibits evidence transmission failure.33
A similar situation obtains with regard to the pairs of propositions that figure in
Cartesian sceptical arguments. The likelihood for me that I have a broken fingernail,
given sense impressions as of a broken fingernail, can be expected to be higher than the
likelihood for me of this proposition before having any relevant sense impressions.
Nevertheless, the likelihood for me of the proposition that I am not a disembodied
victim of an evil demon would not be affected by my broken fingernail-like sense
impressions. The likelihood for me of this proposition before I look at my fingernails,
whatever it is, will remain unchanged after I look and receive these sense impressions.
The evil-demon hypothesis, like other Cartesian sceptical hypotheses, is precisely
designed to have this feature. Hence these hypotheses can be expected to generate
cases of transmission failure.
I’d like to close this section by considering how these ideas apply to Wright’s
soccer example.34 Notice first that the evidence of someone kicking a ball between
two posts can be expected to make the proposition that a soccer goal is being scored
more likely for you. To see this, we just need to compare the probability that a
subject can be expected to ascribe in normal circumstances to the hypothesis that
a soccer goal is being scored prior to ascertaining that a ball is being kicked between
two posts with the probability that she can be expected to ascribe to this proposition
after making this discovery. It is clear that the former, whatever it is, will be lower
than the latter.
33
Pryor is undoubtedly aware of the phenomenon that I am using to explain evidence transmission failure,
since it is exemplified by his case of Clio’s pet (cf. Pryor 2004, pp. 350–1). However, he dismisses this explanation
out of hand.
34
See Wright 1985, p. 436, 2000, pp. 141–3.
Now, with respect to the proposition that a soccer game is being played, the same
situation obtains. In normal circumstances, the evidence that a ball is being kicked
between two posts will make this proposition more likely for a subject. Just think of
how likely it is for you that a soccer game is being played at an arbitrary moment and
how likely this is to be happening when a ball is being kicked between two posts.
I can’t see how the latter could fail to be higher than the former—even taking account
of the fact that some of the times when a ball is being kicked between two posts no
soccer game is taking place.
Interestingly, this will remain so even if the circumstances are assumed to be
abnormal in the way that Wright envisages. Suppose that a soccer film is being
rehearsed in the park, and an instance of a ball being kicked between two posts is as
likely to be a film rehearsal as a goal being scored in a real soccer game. It remains the
case that the likelihood that a soccer goal is being scored and a soccer game being
played in the park at an arbitrary moment is lower than the likelihood of this happening
when a ball is being kicked between two posts. Hence the evidence that a ball is
being kicked between two posts will still make the propositions that a soccer goal
is being scored and that a soccer game is being played more likely for you. It will also
make more likely for you the proposition that a rehearsal of the football film is taking
place, since the likelihood that this is happening at an arbitrary moment is also lower
than the likelihood that it is happening when a ball is being kicked between two posts.
However, there is no reason why a piece of evidence should not increase the likelihood
for you of two incompatible hypotheses.
Nevertheless, there are other circumstances in which evidence transmission between
these propositions would fail. Suppose that the soccer film has an obsessive director,
and the goal-scoring scene is constantly being rehearsed in the park, except when the
pitch is being used for a real game of soccer. For a subject who knows this to be
the situation, the discovery that a ball is being kicked between two posts will still
make the proposition that a soccer goal is being scored more likely, since even in these
circumstances it is less likely that a soccer goal is being scored at an arbitrary moment
than when a ball is being kicked between two posts.35 However, for such a subject the
discovery that a ball is being kicked between two posts would decrease the likelihood
that a game of soccer is being played. For her, the likelihood that a soccer game is being
played in the park at an arbitrary moment is higher than the likelihood of this when a
ball is being kicked between two posts, since she believes that a game of soccer is on
whenever it’s not the case that a ball is being kicked between two posts. Hence, in
these circumstances, the discovery that a ball is being kicked between two posts would
provide the subject with evidence for the hypothesis that a soccer goal is being scored,
but not for the hypothesis that a soccer game is being played, even if she knows that the
former entails the latter, and evidence transmission fails once again.
35
Unless a huge number of goals is scored with other parts of the body, but we will disregard this
possibility.
WRIGHT ON MOORE 321
5. Conclusion
I have considered two issues regarding the relationship between a lightweight propo-
sition L and a heavyweight proposition H entailed by L. The first is whether someone
who knows L can come to know H by recognizing the validity of the inference from
L to H. The second is whether a piece of evidence that makes L more likely for you
must also make H more likely for you if you know that L entails H. It is often assumed
that one can only answer these questions in the negative by imposing implausible
internalist constraints on when L can be known or supported by a piece of evidence.
This assumption can take the form of a dilemma: either we accept these constraints or
we answer our questions in the affirmative, maintaining that transmission (of warrant
or of evidence) never fails. I find both horns of this dilemma unappealing. The
internalist constraints are indefensible, but the idea that Moorean inferences can
sustain the acquisition of knowledge of, or evidence for, their conclusions is deeply
counterintuitive.
My main goal in this essay has been to sketch strategies for avoiding this dilemma—
for vindicating the possibility of transmission failure without invoking the internalist
constraints on knowledge or evidence often associated with it. These strategies, if
successful, will enable us to reject the internalist constraints, as I think we should, while
embracing what pre-reflective intuition surely singles out as the default position—that
Moore obtains neither knowledge nor evidence from his proof.
Bibliography
Boghossian, P. 2000 “Knowledge of Logic”, in P. Boghossian and C. Peacocke. (eds.) New
Essays on the A Priori, Oxford, Oxford University Press, pp. 229–54.
Brown, J. 2003 “The Reductio Argument and Transmission of Warrant”, in S. Nuccetelli (ed.) New
Essays on Semantic Externalism and Self-Knowledge, Cambridge, Mass., MIT Press, pp. 117–30.
Carnap, R. 1962 Logical Foundations of Probability, 2nd edn., Chicago, University of Chicago
Press.
Davies, M. 1998 “Externalism, Architecturalism, and Epistemic Warrant”, in C. Wright,
B. C. Smith, and C. Macdonald (eds.) Knowing Our Own Minds, Oxford, Oxford University
Press, pp. 321–62.
Davies, M. 2000 “Externalism and Armchair Knowledge”, in P. Boghossian and C. Peacocke
(eds.) New Essays on the A Priori, Oxford, Oxford University Press, pp. 384–414.
Davies, M. 2003 “The Problem of Armchair Knowledge”, in S. Nuccetelli (ed.) New Essays on
Semantic Externalism and Self-Knowledge, Cambridge, Mass., MIT Press, pp. 23–55.
Dretske, F. 1970 “Epistemic Operators”, Journal of Philosophy 67, pp. 1007–23.
Dretske, F. 2005 “The Case against Closure”, in M. Steup and E. Sosa (eds.) Contemporary
Debates in Epistemology, Malden, Mass., Blackwell, pp. 13–26.
Earman, J. 1992 Bayes or Bust?: A Critical Examination of Bayesian Confirmation Theory, Cambridge,
Mass., MIT Press.
Hawthorne, J. 2005 “The Case for Closure”, in M. Steup and E. Sosa (eds.) Contemporary Debates
in Epistemology, Malden, Mass., Blackwell, pp. 26–43.
Moore, G. E. 1939 “Proof of an External World”, Proceedings of the British Academy 25,
pp. 273–300.
Nozick, R. 1981 Philosophical Explanations, Cambridge, Mass., Harvard University Press.
Pryor, J. 2004 “What’s Wrong With Moore’s Argument?”, Philosophical Issues 14, pp. 349–78.
Vogel, J. 2000 “Reliabilism Leveled”, Journal of Philosophy 97, pp. 602–23.
Wright, C 1985 “Facts and Certainty”, Proceedings of the British Academy 71, pp. 429–72.
Wright, C. 2000 “Cogency and Question-Begging: Some Reflections on McKinsey’s Paradox
and Putnam’s Proof”, Philosophical Issues 10, pp. 140–63.
Wright, C. 2003 “Some Reflections on the Acquisition of Warrant by Inference” in
Wright, C. 2004 “Warrant for Nothing (and Foundations for Free)?”, Proceedings of the Aristotelian
Zalabardo, J. 2005 “Externalism, Skepticism and the Problem of Easy Knowledge”, Philosophical
Review 114, pp. 33–61.
Zalabardo, J. 2006 “BonJour, Externalism and the Regress Problem”, Synthese 148, pp. 135–68.
12
Moore’s Proof, Liberals, and
Conservatives – Is There a
(Wittgensteinian) Third Way?*
Annalisa Coliva
In the last few years there has been a resurgence of interest in Moore’s Proof of the
existence of an external world, which is now often rendered as follows:1
(I). Here’s a hand
(II). If there is a hand here, there is an external world
Therefore
(III). There is an external world
The contemporary debate has been mostly triggered by Crispin Wright’s influential—
conservative—“Facts and Certainty” and further fostered by Jim Pryor’s recent—
liberal—“What’s Wrong with Moore’s Argument?”2 This debate is worth surveying
* I would like to thank in particular Akeel Bilgrami, Manuel Garcı́a-Carpintero, Sebastiano Moruzzi, Jim
Pryor, Carol Rovane, Barry Smith, Elia Zardini as well as people in attendance at the seminars I gave at the
Institute of Philosophy in London and at the Department of Philosophy of the University of Barcelona for
comments on previous versions of this essay. I would also like to express my gratitude to an anonymous
referee for valuable criticisms. Most of all, however, I would like to thank Crispin Wright for his many
comments and suggestions and for his unfailing support and encouragement during the years in which we
have been discussing these issues. This essay has been written during my tenure of an Alexander von
Humboldt Fellowship, at the Department of Philosophy in Heidelberg. My thanks to the Alexander von
Humboldt Stiftung for their financial support and to the Department of Philosophy in Heidelberg for
providing me with exceptional working conditions.
1
It should be kept in mind that this is not Moore’s original Proof and also that the Proof was not intended as
an anti-sceptical argument, but, rather as an anti-idealist one (see Moore 1942). I present, discuss, and criticize
the original version of the Proof in Coliva 2004, 2010c. In the latter writings (2004, pp. 402–4 and 2010c,
Chapter 1), I also discuss why, despite Moore’s own intentions, his Proof—as was presented in his 1939
paper—can be read as an anti-sceptical argument.
2
Wright 1985 and Pryor 2004, but see also Wright 2002, 2004a and Pryor 2000. The labels “conserva-
tive” and “liberal” are now standard for their respective positions as well as “sceptic” and “dogmatist”.
Wright, however, shares with a sceptic only the assumption that warrant for (III) is needed in order to have
warrant for (I), but he doesn’t endorse the view that this should lead to scepticism, viz. to the view that no
324 A N N A L I S A C O L I VA
with care because—so I shall contend—it will help us see that, in fact, it allows for
an important view that, so far, hasn’t been explicitly considered. The critical survey
will be the task of the next two sections, while, in the remaining two, I shall
outline the third view, enlarge on its consequences with respect to scepticism about
the external world, and discuss its bearing on Moore’s Proof. Let me stress from the
start that this is a programmatic exploration of such an alternative position, which
will have to be further developed in subsequent work.3 Be that as it may, as will
become apparent, the third view will be Wittgensteinian in spirit and in the
Appendix I will briefly compare it with Wittgenstein’s (more or less) official line
in On Certainty.4
1. The contemporary debate

As is familiar, Wright’s view is that the Proof fails because it is epistemically circular.5
According to him, perceptual experience can provide a (defeasible) warrant for the first
premise—which Wright doesn’t call into question—only in a conducive informational
setting including, in particular, the thesis that there is an external material world,
broadly manifest in ordinary sense experience. So Moore’s warrant for premise (I)
depends, in Wright’s conservative view, on his already having a warrant for the
conclusion, since it is only in the context of such anterior information that he justifiably
takes his sense experience as a warrant for “Here’s a hand”. The Proof accordingly,
though valid and proceeding from premises that there is no (non-sceptical) reason to
deny are known, fails to be rationally persuasive—it can’t produce a first warrant for
believing its conclusion.
Jim Pryor has contested this. He agrees that reason to doubt (III) would defeat the
warrant supplied by perceptual experience for (I). But he thinks that one could have
a—to be sure, defeasible—perceptual warrant for (I) just by taking one’s current sense
experience at face value, without the need for any prior and independent warrant for
(III), provided one doesn’t already have any such doubt. Hence, the Proof, in his view,
is not epistemically circular. What is true, he suggests, is that it is dialectically ineffective:
specifically, that it fails in the dialectical setting in which it is presented, since a sceptic
will already doubt its conclusion—that is, on his view, believe it is (likely) false—and,
warrant for (III) can be provided at all. Hence, the label “sceptic”, when applied to his view, may be
misleading; “conservative” might be better. There will be more about Wright’s overall position in the
following (see }3.1).
3
Some attempts in this direction are contained in my 2010a, 2011, forthcoming a, forthcoming b.
4
As is familiar, there is no consensus on what exactly Wittgenstein’s position in On Certainty was. For an
overview of various possible readings of it, see Moyal-Sharrock and Brenner 2005.
5
A number of recent publications testify to the interest this debate is enjoying. Just to mention a few
prominent examples, besides a number of papers written by Wright, Davies, and Pryor, which gave rise to,
and developed the debate, see Beebee 2001, Peacocke 2004, pp. 112–15, Schiffer 2004, Brown 2005, Silins
2005 and 2007, White 2006, as well as Wright 2007 and Pryor, Chapter 10 in this volume.
M O O R E ’ S P RO O F, L I B E R A L S , A N D C O N S E RVAT I V E S 325
hence, will refuse to regard one’s current experience as a warrant for (I).6 But
then, since, by the sceptic’s lights, there is no warrant for (I) in the first place,7
there is no warrant to be transmitted from the premises to the conclusion of the
argument either, which, therefore, will fail to establish that conclusion—at least,
to a sceptic’s satisfaction.
Let us now turn to consider who, between Wright—the conservative—and Pryor—
the liberal—has the better of their epistemological dispute: is antecedent warrant for
(III) a necessary part of the stage-setting for the evidential value of sense experience for
any proposition like (I) or is it sufficient merely to lack reason to doubt (III)?
2. The Wright–Pryor Dispute: Is Anyone Right?

Most of those who have tended to side with Pryor—philosophers such as Martin
Davies, Christopher Peacocke, and Tyler Burge8—are impressed by the fact that it
wouldn’t be truthful to our ordinary epistemic practice to think that only those who
can produce a warrant for a belief in the existence of the external world could have
(perceptual) warrant for an empirical belief such as (I). After all—they seem to think—
many would agree (though not Moore and Pryor) that practically nobody, apart from
perhaps some trained and resourceful epistemologist, could produce a warrant for the
belief in the existence of the external world. Still, we readily acknowledge that both we
and our fellow human beings have empirical beliefs which are perceptually justified.
Moreover we gladly admit that children, who may well lack the conceptual repertoire
necessary even to entertain the belief that there is an external world, regularly arrive at
perceptually justified empirical beliefs.
The trouble with this sort of consideration is that it mistakes the project in which
Pryor and Wright are—at least primarily—engaged. These philosophers aren’t princi-
pally interested in reconstructing our actual epistemic practices.9 Rather, their project is
normative:10 what they aim to understand are the warrants, if any, that are needed in
order for a belief to be, not justified by the lights of our actual epistemic practice,
6
It is contentious that a sceptic would deem (III) false—a sceptic is no idealist. Rather, he would regard it
as unwarrantable. But, for present purposes, we can omit this qualification. I explore this in more depth in
Coliva 2010a.
7
Or, while continuing to exist “in the abstract space of warrants”, it is rationally unavailable to him,
given his collateral beliefs. I discuss this point in more detail in Coliva 2010a. For present purposes this won’t
make any difference.
8
Davies 2004, pp. 226–30, 234–5, Peacocke 2004, p. 178. Burge 1993, pp. 458–9; 2003b, p. 264 has
offered considerations which support the impression that he too would favor Pryor’s position over Wright’s,
although he himself isn’t—obviously—taking issue with either.
9
See, for instance, Wright 2004a, pp. 204–5 and Pryor 2005, pp. 181–2. In conversation, however, Jim
Pryor has pointed out to me that, although his project remains normative, he is also interested in giving a
correct description of our actual epistemic practice.
10
Of course also Davies, Peacocke, and Burge are engaged in normative projects. So, really, it shouldn’t
be a consideration that Pryor’s proposal would be more truthful to our epistemic practice. At most, if it turned
out to be correct from a normative point of view, its descriptive adequacy would be a plus.
whatever that may be, but genuinely epistemically appropriate in the “abstract space of
justification or warrants”,11 as it were. Hence, it doesn’t matter whether we don’t
actually require subjects to be able to produce a warrant for a belief such as (III), in
order to credit them with a warrant for a belief of the kind of (I). The question, rather,
is whether, in general, a warrant for (III) is necessary for there being a warrant for (I) or
the myriad other propositions which we do indeed habitually take to be perceptually
justified.
Before attending to the task of assessing the Wright–Pryor dispute, however, it is
important to be explicit about what can be taken to be their common ground. First
they both conceive of warrant along internalist lines. That is to say, not as something
whose obtaining is merely due to favorable external conditions which may be totally
unknown to us, but, rather, as something whose obtaining is internally certifiable in
such a way that a (suitably conceptually endowed) subject can appeal to it to redeem or
claim one’s warrant for the belief in question. Secondly, it is important to keep in mind
that Wright and Pryor both allow that it is metaphysically possible, though maybe not
nomologically so, that a subject’s sense experience be indistinguishable from a first-
personal point of view whatever its causal origin might be.12
Let me add one last point in the way of clarification of the positions in play here.13
There is a possible ambiguity in saying—as Pryor maintains—that one can take one’s
experience at face value as a warrant for “Here is a hand” as long as one has no reason to
doubt that there is an external world. For this may be taken to mean either as long as,
lacking reasons to doubt, the assumption that there in an external world (although it
might be implicit) is (still) in place; or else, as long as no assumption is made and, in
particular, as long as there is no reason to think that it might be false that there is an
external world. Thus, on the one hand, there is the view according to which one’s
experience can be a warrant for “Here is a hand” only if, while lacking reasons to
doubt, the assumption that there is an external world is in place. On the other hand,
there is the view according to which one’s experience can be such a warrant even
when no assumption about the existence of an external world is in place, and when, in
particular, there are no reasons to think it might be false.14 Now, not only do I think
11
See fn.7. This twist of phrase is due to Davies 2009. It is in fact equivalent to talk in terms of propositional
as opposed to doxastic and rationally available warrants.
12
It has to be stressed that this isn’t necessarily a fall-back into seduction by the “highest common factor”,
denounced by McDowell 1982 (although in conversation Jim Pryor has manifested his sympathies with such
a view). Even if disjunctivism about perceptual experience were true and seeing and hallucinating were two,
mutually exclusive mental states, it would remain that it is metaphysically possible that a subject could not be
able to tell which one he is in. This point is convincingly argued for in Wright 2002, pp. 344–5.
13
I am grateful to Jim Pryor for pressing me on the need of making the following explicit.
14
Talk of assumptions should not immediately lead one to think of a subject’s mental attitude. There may
be assumptions which are needed to have certain warrants—think of the role of axioms in formal theories, or
of certain background assumptions in scientific ones—even when we are merely dealing with “the abstract
space of warrants”. There will be more about this in the following. As to whether talk of “reasons to doubt”
involves a psychological attitude in its turn, it depends on whether we are dealing with “the abstract space of
reasons” or with doxastic reasons. Here, I’m not specifically talking about either, but it should be kept in
that this second reading is what Pryor actually maintains, but, in effect, for reasons
which will become apparent in the following,15 I think that this is what he has to
maintain, in order for his account of Moore’s Proof ’s failure to be at least prima facie
plausible.
So, granting this assumption, let’s consider—if indeed it is intelligible—the case of a
genuinely open-minded subject, who neither believes nor disbelieves that there is an
external world, not even implicitly or tacitly. Now, there is a major difficulty in seeing
how, as Pryor believes, his occurrent experience as of a hand in front of him can give
him an immediate warrant to believe (I) just so long as he has no reason to doubt (III).
For, if Pryor is right, this warrant, even if defeasible, must be such that, once he has it, it
gives him justification for “Here is a hand” and allows him to discard beliefs which are
incompatible with it, such as (I*) “My current experience as of a hand is envatted
experience”. But Pryor agrees that it is metaphysically possible that one’s experience be
subjectively indistinguishable whether produced by a meddlesome scientist in a sce-
nario of handless envatment or by normal perception of external things. So, how can
such an experience immediately—that is, independently of any other assumption—
give one a warrant for (I) and also disconfirm (I*)? Furthermore, assuming closure, how
could having an experience which would be subjectively indistinguishable from one
occurring in a scenario of brain envatment disconfirm that very hypothesis?16
As mentioned, another assumption in common between Wright and Pryor is that
warrant is conceived along internalist lines, as something we can use to claim (if
appropriately conceptually endowed and defeasibly as it may be) that the belief it
warrants is true. So, now, consider again an open-minded subject. Since, according to
Pryor, no view on (III) is needed in order to have a perceptual warrant for (I) when
one’s experience takes a certain course, it would seem, at least prima facie, that his model
would allow for a situation in which such a subject, who attentively thinks about the
matter (and considers a hand a physical object), could profess open-mindedness with
respect to (III) (as well as with respect to other propositions such as (III*) “My sense
organs are now working properly”), while fully understanding its content, and yet, on
the basis of his current experience, also claim to have warrant for (I). But I think we
mind that Pryor deals with both and since he thinks that there are no non-doxastic reasons to doubt about the
existence of an external world, he mostly concentrates on doxastic ones. In that case, one’s propositional
warrant for (I) would not be defeated, but simply unavailable to the subject, given his collateral beliefs. See
Pryor 2004, pp. 362–8. I discuss this issue in more depth in my Coliva 2010a, 2011 and forthcoming a, b.
15
See }5 and fn. 48.
16
Here is an analogous thought. Suppose that there can be cleverly disguised mules that look just like
zebras. Now, how could one’s experience as such give one an immediate warrant for “Here is a zebra” rather
than for “Here is a cleverly disguised mule”? If we are inclined to think that it can, it is simply because we are
already implicitly assuming that the experience we are having is caused by zebras rather than by cleverly
disguised mules. Similarly, assuming closure, how could that experience just by itself disconfirm the very
hypothesis that we are surrounded by cleverly disguised mules? Wright doesn’t have a similar problem
because, on his view, in the case of Moore’s Proof, we in fact have an entitlement for (III). Hence, we are
entitled to discard the possibility of uncongenial scenarios and are therefore entitled warrantedly to believe (I)
on the basis of our current sense experience.
would regard it epistemically inappropriate for a subject to claim “I have no view about
whether there is an external world (or about whether my sense organs are now working
properly), but nonetheless I am (or take myself to be) warranted, on the basis of the
experience I am currently having, to believe that here is a hand.” To stress, the point isn’t
that such a subject would continue to remain open-minded about the existence of an
external world. Indeed, by running the Proof, he would, according to Pryor, acquire a
warrant to believe that there is an external world (or that his sense organs are working
properly). Rather, the point is that before running the Proof, on Pryor’s view, he could
find himself in the situation just described: one of overt open-mindedness with respect
(III) (or III*) and yet of internally certified warrantedness of (I).17
There is a lot more to say.18 But I think these considerations suggest that Pryor may
well be wrong in maintaining that it suffices, in order to have a perceptual warrant for
(I), merely that one’s experience assumes a certain course in conditions where there are
no reasons to doubt (in the sense we have specified) that there is an external world.
Against that, I want to set the idea that it is only in a context in which it is positively
assumed, no doubt most of the time implicitly, that a subject’s experience is produced in
what one takes to be the ordinary way, by causal interaction with a world populated by
physical objects, that one can rationally claim, or grant, the having of perceptual
warrant for a proposition like (I). Pryor’s conception of the conditions which are
required to accompany the experience is too liberal, for, to repeat, on that conception it
would seem enough in order to acquire a warrant to believe (I) simply to have a certain
kind of perceptual experience, while having no view on (III). The next question to ask
is: if Pryor is wrong, does it ipso facto follow that Wright’s conservative view isn’t?
3. A Third Way?
To suggest that the evidential value of experience depends on (tacit) acceptance that
there is an external world does not appear to force us to admit, with Wright, that a
warrant for the belief in the existence of the external world is needed in order to have a
(perceptual) warrant for (I). It seems that there may be an intermediate position: if
merely assuming a thesis, implicitly as it may be, is not equivalent to having (or to
assuming one has) a warrant for it, then the erroneousness of Pryor’s position doesn’t
entail the rightness of Wright’s. The problem, however, is that of properly character-
izing such an intermediate view, since a number of possibilities are open.
Before attending this task, I wish to clarify what I take to be a constraint on any
appropriate characterization of the third way; namely, that it should be suited to meet
the sceptical challenge. Let me briefly say what, in this context, this challenge amounts
to. I think that we can distinguish two kinds of scepticism.19 On the one hand, there is
17
See also Wright 2007, pp. 34, 39.
18
For further reasons to complain with Pryor’s position, see Wright 2007.
19
Wright 2004a, pp. 167–75 and Coliva 2008.
Cartesian scepticism, which, by raising the hypothesis that we may be dreaming right
now (or, equivalently, that we might be brains in a vat), asks us to provide warrants for
any specific empirical belief we may have. Since, on that hypothesis, our experiences
would be subjectively indistinguishable from what they actually are, notwithstanding
their different causal origin, a Cartesian sceptic would then conclude that, in fact, we
have no such warrant.20 Since this would be true for each and every empirical belief we
may have, a Cartesian sceptic can then generalize and claim that we have no warrant for
our general belief in the existence of the external world.21
On the other hand, however, there is another kind of scepticism, elicited from the
work of Hume and typified, ironically, by Moore’s Proof, as Wright reads it. Such a
kind of scepticism doesn’t traffic in dreams, demons, and brains in a vat. Rather, it
asks us to provide a warrant for the general belief in the existence of the external
world, taken as such. It is important to emphasize here that the kind of warrant a
sceptic asks us to provide is epistemic. By this, I mean that the possession of such a
warrant, for a given belief, should give one a justification to believe in the truth of
what is warranted thereby. To meet this challenge, however, it seems that all we can
do is to derive a warrant for (III)—the belief that the class of physical objects isn’t
empty—from our perceptual warrant for a belief about one of its instances, like (I)—
the belief that here’s a hand. But, according to a Humean sceptic (as well as Wright),
such a perceptual warrant is only possible in a context where (III) is assumed. (III),
however, can rationally be assumed—both according to the sceptic and Wright—only
if we have a warrant for it. But we can have no such an independent warrant for (III).
Hence, the Humean sceptic concludes that no warrant for (III) can be provided at
all.22
The constraint I want to impose on what would count as a suitable characterization
of the third way is that it be somehow capable of confronting the Humean sceptical
challenge. But, as we know at least since Strawson’s work on scepticism,23 there can be
direct and indirect responses to scepticism. A direct response would be one in which we
try to meet the sceptical challenge head-on and try to provide an epistemic warrant for
(III)—that is, a warrant that, no matter whether perceptual or otherwise, would give
one a justification to believe that (III) is true. An indirect response, in contrast, would be
diagnostic, but it could take at least two forms: it could either try to show that the
sceptic’s challenge is, after all, meaningless, incoherent, or self-refuting;24 or else, it
20
Or, alternatively, if warrant is conceived in an externalist fashion, that we may have such a warrant if
conditions are favorable, but that we have no way of claiming—that is, of rationally reassuring ourselves—that
we have it, for nothing in our subjective experience allows us to discard the hypothesis that it may be
produced by a dream. For an in-depth treatment of this issue see Wright 2004a and Coliva 2008.
21
Or again (cf. previous footnote), that we can’t claim such a warrant.
22
Or, alternatively (cf. the two previous footnotes), that we can’t claim such a warrant.
23
Strawson 1985, p. 3. Ironically, Strawson seems to think that naturalism can be a response to scepticism,
when in fact it is simply the result of having given in to it.
24
Putnam, Davidson, and perhaps Wittgenstein in On Certainty could be taken as examples of one version
or other of this kind of indirect response to scepticism.
could try and provide a strategy of “damage limitation”,25 that is, a strategy which,
while granting the good-standing of the sceptical challenge, and, furthermore, that no
epistemic warrants for (III) can be provided, would show that the allegedly disastrous
consequences of such a concession wouldn’t follow.
What are these consequences? I think we can safely call them “postmodernist”:26 not
so much that our shared practices of formation of empirical belief are somehow up in
the air, in danger of annihilation by the sceptical challenge—or put a little less
figuratively, that in fact we have no epistemic warrants, of an ordinary perceptual kind,
for our everyday empirical beliefs. Rather, that whatever these practices may be and
whatever purposes they may accomplish, our acceptance of their foundations isn’t
rationally grounded, for we don’t have, nor can we have, an epistemic warrant for
them. Thus, our taking for granted that there is an external world wouldn’t be
grounded in reasons and evidence, but in “something animal”—something brute and
not rational—like an ingrained psychological mechanism, or, perhaps, a “form of
life”.27
Alternatively, the consequences of the concession that our acceptance of presuppo-
sitions such as (III) isn’t epistemically warranted may be considered to be this: we take it
for granted that there is an external world because this is the “hinge” of our practice of
forming, assessing, and withdrawing from empirical beliefs on the basis of empirical
evidence. Such a practice, in its turn, is something we find it useful or convenient to have,
given our various purposes in life. After all, it does serve us reasonably well. So, taking
for granted one of its presuppositions, such as that there is an external world, is
something we do only practically rationally—that is to say, in virtue of the expected
utility of having the practice of which (III) is a presupposition. Much contemporary
thought, mostly influenced by Nietzsche and Weber, would indeed embrace such a
view.
Just to anticipate a little, in the following, I will sketch a strategy for making out the
third way, which would be one of “damage limitation”. In particular, I will try to
maintain that our acceptance of (III) is epistemically unwarrantable yet rational and,
moreover, rational by the very lights of epistemic rationality itself.28
With this in mind, let us briefly consider how an externalist characterization of the
third way would fare. According to such a position, we should say that, as long as the
25
Wright 2004a, p. 206.
26
I hereby take the liberty to appropriate a label used many times by Crispin Wright at least in
conversation.
27
Or, to use an expression dear to Simon Blackburn, albeit allegedly first used by the Queen after Princess
Diana’s death, such takings-for-granted would be the result of “dark forces at work, of which we know
nothing” (Blackburn 2005, pp. xv, 66). Hume can be taken as maintaining the naturalist view that our belief
in the existence of the external world is something we have in virtue of our psychological constitution.
Wittgenstein, according to Strawson’s reading of On Certainty, in contrast, can be taken as the supporter of a
different kind of naturalism, according to which such a belief is something we have in virtue of belonging to a
certain form of life. There will be more on Wittgenstein’s alleged naturalism in the Appendix.
28
If this sounds oxymoronic, wait until }3.2.
assumption is merely true—that there is indeed an external world, interacting with

one’s sense organs in much the way one normally supposes—one is warranted in taking
one’s experience at face value to support beliefs such as (I). So the warranting potential
of experience depends on what one takes for granted and on whether what one takes
for granted is true—what kind of world one actually lives in.
The externalist interpretation, however, is problematical when taken in connection
with scepticism, for at least the following reason. If the externalist proposal is somehow
put at the service of Moore’s Proof, in order to provide a direct response to scepticism, it
wouldn’t turn that Proof into a cogent argument. For reflect: the perceptual warrant we
would have and claim for (I) and which would transmit to (III) would be conditional
upon relying on (III)’s truth, with no particular warrant or justification for doing so.
Yet, one can’t produce and claim a first warrant for a conclusion when warrant for it
would in fact depend on already assuming that very piece of information.29
It appears, then, that the third view will best be developed within the broad family of
internalist positions: as long as one merely takes it for granted, trusts, or accepts that there is
an external world, one will be rationally entitled to take one’s perceptual experience at
face value as a warrant for one’s empirical beliefs. Characteristically, the notion of
taking for granted (or of trusting or accepting) is construed as an attitude of acceptance
of a proposition which does not depend on evidence, nor implies its truth:30 if we take
P for granted (or, equivalently, we trust in it or assume it), then we act as if P were true,
although we have no evidence for it—nor, in the relevant case, can we acquire any—
and even if P might in fact be false—but, notice, that in the relevant case, we can
acquire no evidence that P might be false either. This attitude, however, will be
regarded by the third way as grounding the possibility of acquiring evidence for (or
against) other (empirical) propositions. The suggestion, then, will be that some pro-
positions have to be taken for granted in order for other propositions to be empirically
warranted (or disconfirmed). In particular, a proposition such as (III) has to be taken for
granted in order for a proposition of the kind of (I) to be perceptually warranted (or
disconfirmed by empirical evidence).31
29
There will be more on this in }5. Notice, moreover, that if things were otherwise, a whole host of long-
lasting philosophical problems—such as the existence of other minds, the existence of the past, and the thesis
of the uniformity of nature—could easily be solved: it would be enough merely to design logically valid
arguments the warrantedness of whose premises depended on the truth of their conclusions. For instance, (I)
Here is a person in pain; (II) If there is a person in pain here, then other minds exist; therefore, (III) Other
minds exist. It is worth stressing that this is not what Moore is doing, at least not overtly. For Moore, contrary
to the kind of externalist I’m thinking of here, does not hold that (III) is true and thus safely assumed, and that
that, in turn, allows one to have and claim warrant for (I) (nor does Pryor, for that matter).
30
See Wright 2004a, pp. 175–8, 183 and fn. 36.
31
As opposed to what Davies 2004, p. 230 maintains and which elicits his scepticism with respect to such a
third way. He writes, in the course of developing his idea of a negative entitlement, which I will not discuss
here: “Switching to the negative notion of entitlement, we could add that the thinker is entitled not to
bother about, nor even to consider, that his perceptual apparatus might not be operating properly. But we
must not slide from this to the idea that, since the thinker does not doubt that his perceptual apparatus is
working properly, he assumes this. For the thinker need not be capable of adopting any attitudes towards that
It is worth stressing that construing acceptance as an attitude towards a proposition

does not mean to say that subjects should explicitly entertain and assent to the
proposition in question, or that they should at least have the conceptual resources
necessary simply to entertain it. Rather, they could be seen as implicitly (yet positively)
assuming a given proposition in virtue of being immersed in a practice, which has that
acceptance as a precondition of its rational intelligibility. To illustrate: my dialling a
certain international code to call my best friend when I am abroad can be seen as
rational only on the assumption that I am taking for granted that she is in that other
country, even if I am not explicitly considering that presupposition, let alone assenting
to it. Indeed, even a child who, for the same purpose, dialled that code, while (perhaps)
even lacking the conceptual repertoire necessary to entertain that presupposition,
could be granted with its acceptance, because he could be seen as taking part in a
communal practice whose rational intelligibility rests on such an assumption.32
So, the notion of acceptance I am working with is admittedly psychologically very
thin, but I think this is the best I can offer in order to bring what I take to be the right
theoretical model to bear on—or, at any rate, in line with—real-life situations. Were
one to find this proposal unsatisfactory, one could simply hold on to the idea that on
the third way, the architecture of the abstract space of warrants is such that in order for
experience to be a warrant for specific empirical beliefs, certain assumptions have to be
in place—that is, they must be accepted, as it were, by an ideal epistemic subject.
Granting this, however, the salient problem with the proposal is to make it suitable
to confront the sceptical challenge. After all, as we saw, the sceptic need not contest
that there are certain things we take for granted and that the existence of an external
world is just one of those. Indeed, he may even claim, with Hume, that this is precisely
what scepticism—correctly understood—leads us to see: that our ordinary empirical
inquiries aren’t rationally grounded, for we don’t have, nor can we have, a warrant for
them. If the third way is to do better, it needs to be made out that “animal trusting” is
not merely all that is available when reasons and justifications fail, but that attitudes of
trusting somehow underwrite the very possibility of empirically justified belief. Thus,
the only way in which trusting, or taking for granted, could be of avail in this
connection, would depend on the possibility of redeeming its rationality, contrary
appearances notwithstanding.
3.1 Wright’s entitlements?

Wright’s most recent work seems to go in this direction too: according to Wright, we
can face the sceptical challenge and win if we broaden the class of warrants to
countenance both evidential—that is, perceptual—and non-evidential warrants. In the
proposition.” Notice, however, that Davies supports his notion of negative entitlement on the basis of
considerations of psychological adequacy. This kind of consideration, however, isn’t immediately relevant to
the normative enterprise at issue in this kind of debate.
32
In On Certainty Wittgenstein often points in this direction too. See for instance OC }}148, 174, 360.
relevant class of cases, in Wright’s view, while no evidential warrant would (or could)
be available, a non-evidential one would (perhaps33) be attainable. Hence, when it
comes to a belief such as (III) we may have a non-evidential warrant—that is, an
entitlement, in Wright’s terminology—to trust such a presupposition, and in the context
of such rational trust, thereby rationally take our specific experiences to support
particular empirical beliefs.
Now, I don’t wish to discuss Wright’s proposal in detail, but some things are worth
noticing. If Wright’s non-evidential warrants were supposed to be genuinely epistemic
warrants—that is, warrants that, though non-perceptual, would certify (albeit defea-
sibly) that what they warrant is in fact the case—I don’t think they would have been
vindicated by Wright’s various strategies for redeeming them—entitlements of cogni-
tive project, of substance, etc. For, as connoisseurs of Wright’s most recent work will
know, his entitlements make only for the rational permissibility, given certain practices
and conceptual schemes we need or want to hold on to,34 of accepting their presupposi-
tions. The thought is that if we want or need to hold on to them, then we have no
other option but to accept 35 their presuppositions. But the fact that a story can be told as
to why accepting (III) is rational, if we need or want to hold on to a conceptual scheme
where physical objects are countenanced, doesn’t eo ipso give us anything that would
certify that (III) is indeed the case. So, I think that Wright has given us no epistemic
warrant, properly so regarded, for accepting (III). Hence, if Wright’s recent work is
intended as a way of filling out the details of a conservative and direct response to
scepticism—one according to which non-evidential, still entirely epistemic warrants
for (III) are provided—I think it is not successful.
But I think Wright is aware of this.36 So I deem he is actually trying to do something
different, namely to stick to the structural project of the conservative view—which is
that of providing a warrant for propositions like (III)—while in fact offering a warrant
other than epistemic—that is, a rational entitlement—which is different from a
genuine epistemic warrant, both because it is not evidential and because it does not
bear on the truth of what is warranted thereby. If so, Wright’s strategy would be
structurally conservative but, in effect, an indirect response to scepticism, as it would merely try
33
Wright 2004a, pp. 200–3, 203 in particular himself acknowledges that his way of redeeming the
rationality of “There is an external world”, by means of what he calls “entitlement of substance”, is not
watertight. I discuss the details of this kind of entitlement in Coliva 2007, 2012 and forthcoming a, b.
34
Wright 2004a, p. 192: “If a cognitive project is indispensable, or anyway sufficiently valuable to
us . . . we are entitled to . . . [its] presuppositions without specific evidence in their favor.”
35
See Wright 2004a, p. 189: “one cannot but take certain things for granted”.
36
See Wright 2004a, p. 206 where he actually says: “In general, it has to be recognised that the unified
strategy [viz. his strategy of response to scepticism both of Cartesian and Humean kind by means of the appeal
to rational entitlements] can at most deliver a sceptical solution . . . Sceptical solutions concede the thrust of the
sceptical arguments they respond to . . . The unified strategy likewise concedes the basic point of the sceptical
arguments to which it reacts, namely that we do indeed have no claim to know, in any sense involving
possession of evidence for their likely truth, that certain cornerstones of what we take to be procedures
yielding knowledge and justified belief hold good.”
to limit the damage of the sceptical considerations.37 To repeat, this would mean
that while Wright thinks that warrant for (III) can be provided, it would not be a
warrant which would give one reason to think that (III) is true. Rather, it would be a
warrant, which would make it rationally permissible to assume (III) and, in light of
that assumption, rationally permissible to go on acquiring the usual perceptual
warrants for empirical beliefs such as (I), that we take our perceptual experiences to
provide us with.
This, in fact, helps us see the structural difference between his project and mine. On
one possible interpretation of Wright’s position, he thinks that acceptances can be
rational just in case we can provide a rational entitlement for them—that, as we saw, he
thinks of as some kind of warrant, albeit different from a genuinely epistemic warrant,
evidential or otherwise and hence as something that, once you have it, justifies your
acceptance of, in our case, (III). This may seem rather intuitive. After all, it is conceived
on the model of what makes our ordinary empirical beliefs rational; namely, the fact
that we have warrants for them, only, in this latter case, genuine epistemic warrants of a
perceptual kind. But, first, it is notable and relevant to our present purposes, that
Wright’s apparent concession that we need a warrant for (III)—albeit neither an
evidential nor an epistemic one—thereby, in turn, attaining a position in which we
can acquire perceptual warrant for ordinary empirical beliefs such as (I), seems to go
against the spirit of the third way I am trying to canvass. For the latter’s distinctive
contention, if it can be sustained, is precisely that no warrant for such takings-for-
granted is required before they can be rational.
Second, I think there are other ways, different from being variously warranted, in
which an acceptance can be rational: if, for instance, that acceptance is itself constitutive
of rationality. Think, for example, of accepting the pattern of inference which consti-
tutes modus ponens. What one might easily end up saying is that accepting Q, given the
previous acceptance of P and “If P, then Q”—when one is considering that general
scheme of inference and not one of its specific instantiations—is part of what constitutes
being rational. There is no need to think that accepting it is rational only if we can
discover some kind of warrant for it. Now, suppose you meet someone who tells you,
“Look! We accept Q, given our acceptance of P and of ‘If P then Q’. This, however, is
a brute acceptance, which cannot be warranted, hence it is not rational.” Suppose we
were to respond by saying “True, that’s what we do. But to do that is part of what
being rational actually consists in. So, that acceptance is rational after all, even if we have
37
Let me register at this point that there might be worries about the nature of the rational entitlements
Wright is producing. Some theorists, in particular Pritchard 2005 and Jenkins 2007, have claimed that
pointing out that accepting presuppositions such as (III) is simply something we can’t but do if we need or
want to engage in certain epistemic practices, seems to suggest that accepting them is rational in virtue of the
utility those practices have in our life, or the purposes they serve. Hence, the kind of rational entitlement
provided for them would be merely practical. If this were the case, then Wright’s strategy wouldn’t supply a
counter to the “postmodernist” implications of the Humean sceptical challenge, for, as we saw, one form
those implications could take is precisely that we are merely practically warranted in accepting presupposi-
tions such as (III).
no warrant for it.” Now, if we were right about that, I think we would have responded
to our opponent. This, however, doesn’t mean that we would have produced or
discovered any kind of warrant that makes that acceptance rational. For to point out
some kind of conceptual or grammatical connection between our acting in a certain way,
like accepting Q (given our acceptance of P and “If P then Q”), and being rational is
not to provide anything like a warrant for that acceptance, capable of turning it into
something rational. Rather, to point that out, is simply to register that warrants and
justifications have come to an end, yet our actions, like accepting Q (given our
acceptance of P and of “If P then Q”), are what determines, in a given context, what
being rational consists in. So, they are not rational in virtue of something else, whatever
that something else might be—an evidential warrant or Wright’s rational entitlements.
Rather, they themselves constitute or determine what being rational consists in.
Let me try to clarify the proposal by contrasting it with Wright’s. One could (I think
correctly) put this alternative view by saying that its central contention is that there are
acceptances that are rational in virtue of being constitutive of (our notion of) rationality.
One could then go on to say that, if this is the case, what the alternative proposal makes
out is a warrant for them—although not an evidential one—and that it is the presence
of such a warrant that turns what would otherwise be a-rational acceptances into
rational ones. If this were true, then the present proposal would differ from Wright’s
only in detail—in the way in which, in effect, it makes out an entitlement in the
relevant case—not in structure.
But now consider the following “in virtue of which” kind of explanation: we
could say that the liquid in the bottle sitting on the table is water in virtue of its being
H2O. Now, obviously, when we say this we don’t mean to say that there is something
else—H2O—which attaches to the liquid in the bottle and turns it into water. Rather,
what we mean to say is merely that being H2O is what being water consists in.
Similarly, by simply saying that an acceptance is rational in virtue of its being
constitutive of rationality one is not thereby committed to saying that there is
something—its being constitutive of rationality—over and above the acceptance
itself, which turns it into a rational one. All one would in fact be saying is what
being a rational acceptance consists in.
Furthermore, it is important to note that by explaining that the liquid in the bottle is
water in virtue of its being H2O, one isn’t providing any warrant for the fact that that
liquid is water, but only, at most, for one’s claim that it is. Similarly, I think there is
actually no reason to hold that by giving an account of why a certain acceptance is
rational one would have provided a warrant for the acceptance itself. Rather, all one would
have provided is, at most, a warrant for claiming that the acceptance is a rational one.
Actually I think that this last point doesn’t apply just to the explanation of the
rationality of certain acceptances I have proposed and will further develop in the
following, but to Wright’s as well. In particular, since by simply accounting for why
a given acceptance is rational—either following Wright’s suggestions or mine—we
wouldn’t thereby produce a warrant for it, but we would simply provide ourselves
with a warrant to claim that it is, I actually suspect that his notion of entitlement—
conceived of as a kind of warrant that has always been present at first-order, alongside
with acceptances, thereby making them rational, and which he has had the merit of
discovering (or of making out) for all of us, so that we can now actually claim it, thus
redeeming it at second-order—is entirely spurious. If the analogy with the explanation
of why something is water is correct, we can now see that while Wright may have
given us a warrant for our claim that certain acceptances are rational, he has certainly not
discovered warrants that have been there all along, accompanying our acceptances and
making them rational. So, in effect, entitlements would be the product of a projection
error: the error of projecting what constitutes a warrant to claim that a given acceptance
is rational (i.e. the explanation we may offer of why that acceptance is rational), back
onto the level of first-order warrants for that very acceptance, which, if it existed at all,
would (allegedly) be capable of turning that otherwise a-rational acceptance into a
rational one.
So, to summarize this perhaps intricate discussion: I have claimed that (i) there aren’t
and there need not be any warrants, of any kind whatever, to make certain acceptances
rational, insofar as the latter are themselves constitutive of rationality; and (ii) that when
we point that out, or indeed provide ourselves with a warrant to believe that those
acceptances are rational, perhaps along the lines explored by Wright in his discussion of
rational entitlements, we don’t provide anything like a warrant for those otherwise a-
rational acceptances, which would make them rational. Rather, we simply provide
ourselves with a warrant to claim that those acceptances, which have always been
rational independently of any justification we may have to maintain they are, are
indeed so.
There is obviously a lot more to be said, which would deserve a separate treat-
ment.38 Nonetheless, I hope that, whether or not I have succeeded in casting some
doubt on the legitimacy of Wright’s notion of entitlement, what we have just seen is
enough to clarify the structural differences between his project and mine: Wright
thinks that in order for our acceptances to be rational, we must provide warrant for
them. That warrant can’t be evidential, hence it must be something else—a rational
entitlement—which, moreover, does not give one a justification to believe that what it
warrants is true, but simply makes it rationally permissible to take a certain assumption
for granted. In contrast, I think acceptances may be rational—and, moreover, as I will
argue, rational from the point of view of epistemic rationality itself—even if no
warrant—evidential or otherwise—can be provided for them, so long as it can be
claimed that such acceptances are constitutive of (epistemic) rationality.
Going back to the third way, I think something broadly analogous to what we have
observed about accepting the conclusion of modus ponens can in fact be said about our
acceptance of (III). If so, that acceptance will be rational, even if we have no warrant for
38
For some other objections to Wright’s notion of entitlement, especially to “entitlement of substance” as
he calls it, see Coliva 2007, for more general criticisms, see Coliva 2012 and forthcoming a, b.
it. Furthermore, the story we will tell in order to bring out its rationality could be used
to counter the potentially “postmodernist” implications of the sceptical challenge,
without thereby producing or discovering any first-order warrant—that is, any extra
“something in virtue of which” our acceptance of (III) would be rational.
3.2 The epistemically rational legitimacy of assuming (III)

So, now, let me turn to the attempt to redeem the rationality of accepting (III), even if
no epistemic warrant for it, evidential or otherwise, can be provided. I think that in
order to meet this challenge we should ponder further on the role a presupposition
such as (III) plays for us. What accepting (III) allows us to do, first and foremost, is to
have the basic epistemic practice of forming, assessing, and withdrawing from empirical
beliefs on the basis of perceptual evidence pro or against those beliefs. To say that this
epistemic practice is basic means to say that it is presupposed by all other practices of
formation of empirical beliefs. So, for instance, we can arrive at beliefs about the
material world around us also through the aid of instruments. But, clearly, we would
have to rely on the fact that our senses, empowered, for example, by the use of a
telescope, actually put us in contact with physical objects whose distance makes them
invisible, or only partially visible, to the naked eye.
In this connection, the case of a religious believer who accepts what he takes to be
the word of God as revealed in the Bible, and who, like Cardinal Bellarmine, may
(allegedly) refuse to look into the telescope and to trust his senses to form beliefs about
the planets, is often discussed as a case of acceptance of a different epistemic practice of
formation of empirical beliefs. While it is discussible whether it is indeed such a different
epistemic practice,39 it remains that it wouldn’t be basic. For in order to accept that the
planets aren’t made of the same substance that constitutes mountains and lakes, on the
basis of the Holy Scriptures, one needs to read the Bible (or whatever other source of
information is supposed to ground that belief), as well as understand what it means.
Hence, one needs to trust that one’s senses put one directly in contact with physical
objects such as token bibles, their written pages, etc., that then provide the evidence on
which one’s peculiar views on the nature of the skies are formed.
Similar considerations can be put forward also for those practices of formation of
empirical belief which most of us would deem totally irrational, like consulting
horoscopes or making divinations. For they would presuppose trust in the fact that
our senses put us in touch with physical objects: after all, we would obtain the relevant
information by looking at stars, planets, at the interiors of animals, and so on. Hence,
even in these cases we would need to form empirical beliefs on the basis of the
39
See for instance Boghossian 2006, p. 104. According to Boghossian, Cardinal Bellarmine would in fact
use his senses to form ordinary beliefs about the sun, the stars, and the planets, but, trusting the Bible, he
would then refuse to use his senses (empowered by the telescope) to form beliefs about their physical nature.
This, according to Boghossian wouldn’t mean that Bellarmine holds on to a different epistemic practice, but
only that he thinks, contrary to most of us, that what the Bible says about the nature of the skies is true.
deliverances of our senses, from which we would then infer, according to some
(bizarre) theory, that it is going to rain soon, or that we are going to have a productive
year, or a miserable one.
To say that our practice of forming empirical beliefs on the basis of our perceptual
experience is basic then entails that it is presupposed by all other procedures of
formation of empirical beliefs we have and, in this sense, that it is central to our form
of life. Whether it is also universal, or even necessary, is another matter. I think it can be
maintained that it is universal in the sense that it is shared by all human beings in virtue
of the kind of creatures we are. But perhaps we might imagine beings different from us
that have no senses whatever but are somehow able to have intellectual knowledge of
objects around them.40 So maybe this practice is universal, given the kind of creatures
we are, but contingent on precisely these features of our nature. Be that as it may, in
order to defend the rationality of accepting its presuppositions we don’t need to assume
(or to prove) its necessity.
What is crucial to realize, however, is that the basic epistemic practice of forming,
assessing, and withdrawing from empirical beliefs on the basis of perceptual evidence,
of which (III) is constitutive, is not just one practice we have—no matter how central it
is to our overall system of formation of empirical beliefs. Rather, it is the very practice
that contributes to the individuation of what we take epistemic rationality to be. To be
epistemically rational simply is to be able to form, assess, and withdraw from empirical
beliefs on the basis of one’s perceptual evidence. Since there wouldn’t be any epistemic
rationality if we did not trust in the existence of an external world, it is constitutive of
epistemic rationality, as we ordinarily understand it, that we accept what in effect
enables it, such as the existence of an external world.
A short detour is necessary at this point. If idealists or phenomenalists feel outraged at
the idea of epistemic rationality depending on assuming (III), they should realize that
this is merely what our notion of epistemic rationality is like:41 we do indeed think that
perceptual experience can certify the truth of the fact that there is a hand—taken as an
object existing independently of our minds—where we seem to see it. If we are serious
about the fact that it is a metaphysical possibility that experiences be indistinguishable
with respect to their causal origin, then that may be so only if, as we have seen, the
assumption that there is an external world is in place. Hence, if such theorists were to
insist on the possibility of epistemic rationality independently of assuming (III), they
would have to concede that they are indeed using or talking about a different notion of
40
I am not going to inquire any further whether this possibility is really conceivable and coherent. At least
prima facie, it seems so.
41
Of course this is not an empirical claim. The view I am putting forward is precisely that acceptance of
(III) is constitutive of epistemic rationality. I can’t argue for this here, but I have done it elsewhere. See Coliva
2009, 2010 b, c where I actually maintain the even stronger claim that phenomenalists and idealists would not
be able to deliver a coherent notion of epistemic rationality even by their own standards, i.e. even if
experiences were not taken to bear on the existence of mind-independent objects, but on the existence of
“objects” understood along idealists’ and phenomenalists’ favored lines.
epistemic rationality. One, in particular, in which perceptual evidence doesn’t really

speak to the truth of empirical beliefs, but, rather, to the coherence between our sense
impressions over time. So, they would be revisionists both with respect to our concep-
tual scheme, since no mind-independent objects would be countenanced, and with
respect to our notion of epistemic rationality. Luckily, however, we are dealing with
Humean scepticism, which isn’t revisionary of our concepts (and practices), as it leaves
everything as it is, while simply pointing out that our most basic assumptions, such as
(III) aren’t warranted, and that, therefore, aren’t rationally held.
So what I would like to suggest, by analogy with the case of accepting (the
conclusion of) modus ponens we reviewed in the previous section, is that to say that
the acceptance of (III) is constitutive of epistemic rationality makes it rational in its turn
and, moreover, rational—that is rationally mandatory—by the very lights of epistemic
rationality itself. Thus, the thought is that although we have no epistemic warrant—
evidential or otherwise—for a proposition like (III), accepting it can be seen as,
nevertheless, fully epistemically rationally legitimate, and not merely practically so, just
because such an acceptance is indeed constitutive of our very notion of epistemic
rationality.
It must be stressed that, on my view, the notion of epistemic rationality is not given
absolutely, but always within a system or practice—that is, the system or practice in
which we produce evidential warrants in favor or against empirical propositions. In
order for that system or practice to be possible, it must be assumed that there is an
external world, given the preconditions on which we have agreed, at least for the sake
of argument, namely that perceptual experiences could be subjectively indistinguish-
able no matter what their causal origin might be, that we are working with an
internalist notion of warrant, that the only “good” warrants in town are evidential
ones and that such warrants can’t be provided for propositions like (III) along the lines
indicated by Moore’s Proof. Now, clearly, that assumption isn’t irrational for it isn’t
made against any evidential warrant we may have—as we have repeatedly pointed out,
any evidential warrant would depend for its existence on making such an assumption,
in turn. Nor does it necessarily have to be seen as a-rational, like something brute and
instinctive we simply do, as a consequence of our psychological constitution or of our
upbringing within a community of beings who share a certain form of life. Nor does it
have to be considered rational only by the lights of practical rationality. Rather, it can
be seen as part of epistemic rationality, even if it is not warrantable, as it is constitutive
of it: not making that assumption would deprive us of the system or practice within
which that notion has a home, and, consequently, of that very notion too, in such a
way that we could no longer impugn the rational legitimacy of that assumption. For, in
order to have our notion of epistemic rational legitimacy at all, that assumption has to
be in place. Therefore, that assumption is indeed part of our system or practice of going
about forming, assessing, and withdrawing from empirical beliefs on the basis of
warrants, for it is its condition of possibility. Yet, clearly, it is not just like any other
element within that system or practice, for warrants for or against it cannot in fact be
given. Still, it does not fall outside it either; rather, it is part of it, as it were, at its limit,
because it makes that system or practice possible. This is why it can be seen as
epistemically rationally legitimate—indeed, mandatory—in its turn, even if it is un-
warrantable.
Hence, to sum up, I am maintaining that our notion of epistemic rationality is in fact
wider than may have been realized. In particular, it extends not only to perceptually
warranted empirical beliefs, but also to the acceptance of those presuppositions that
make the acquisition of those warrants possible in the first place. Let me emphasize
again that to say that an acceptance is epistemically rationally legitmate if it is constitu-
tive of epistemic rationality is not intended to provide a (non-evidential) warrant for it,
that is to say, something else in virtue of which it turns out to be rational, but only to
point out an aspect of what its being epistemically rationally legitimate consists in.
4. The Third Way and Scepticism

If what I have just said is along the right lines, where does it leave us with respect to
scepticism? I think that we can agree with Wright that the (Humean) sceptic’s mistake
consists in making an erroneous inference from a true observation when he moves
from the absence of epistemic warrants for propositions like (III) to the conclusion that
taking them for granted is not rational, or only practically so.42 Still, although I agree
with Wright’s general diagnosis, I think the reason why the (Humean) sceptic goes
astray is not that he has too narrow a conception of warrant and that, after all, we have
some kind of warrant for (III)—albeit different from genuinely epistemic, perceptual
warrant—which makes its acceptance rational. On the contrary, I am quite convinced
that warrants, in this area, are just what a sceptic takes them to be—that is, the normal,
evidential warrants that bear on the truth of what is warranted by their means.43
Hence, a Humean sceptic’s mistake doesn’t depend on too narrow a conception of
warrant, but, rather, on too narrow a conception of our very notion of epistemic
rationality. For, if the third way is right, epistemic rationality extends also to the
acceptance of those presuppositions, such as (III), which are constitutive of it. To
notice this is to give ourselves the means—in effect a rational warrant—to affirm,
contrary to scepticism, that accepting these presuppositions is rational after all.
However, if it is right that epistemic rationality extends also to the acceptance of
presuppositions such as (III), then it is indeed ironic that the philosophical figure, who
is usually considered the champion of epistemic rationality, can actually survive only on
42
Actually Wright only points out the erroneous inference from the absence of empirical warrants for
(III) to the conclusion that accepting it is not rational. But we saw before that the “postmodernist”
implications of the Humean sceptical challenge could take the form of exposing the merely practical
rationality of accepting (III).
43
I can’t develop the point here, but I suspect that also other forms of entitlement, namely those recently
developed by Burge 2003a, Peacocke 2004, and Davies 2004, may turn out to be problematical when taken
in the context of the sceptical challenge, as I have described it.
the basis of a misconception of that notion! But this is an irony which can easily be
explained: scepticism challenges “epistemological realism”—the view, as I understand
it,44 according to which even our most basic acceptances, such as (III), can be grounded
in evidence. Rightly finding that view untenable, a sceptic seems to give way to a form
of “epistemological irrealism”—the view according to which those acceptances aren’t
rationally grounded at all, or are simply practically so. What I think is true, in contrast,
and brought out by the third way, is that neither epistemological realism nor episte-
mological irrealism accurately describe our epistemic situation. For that situation is in
fact a form of “epistemological internal realism”, or even of “epistemological anti-
realism”: from within our actual epistemic practice, to accept that there is an external
world, even if no warrants for doing so are attainable, is what enables the deployment
of what we take epistemic rationality to be and, for that very reason, is itself rationally
legitimate and, furthermore, epistemically so.
5. The Bearing of the Third Way on Moore’s Proof

In order to assess the bearing of the third way on Moore’s Proof it is necessary to
consider what a proof, in general, is and should accomplish.45 On reflection, it is quite
clear that a proof should provide one with a warrant to believe its conclusion, so that if
one didn’t have that before, one would acquire it; or else, if one had it already, one
could reinforce it.46 It should also be clear that proofs are arguments we design and put
forward in order to be able to claim the rational legitimacy of our belief in their
conclusions. So proofs are procedures which produce warrants for their conclusions
thereby giving one the possibility to appeal to them to claim that one’s belief in them is
warranted.
If so, it is fundamental, in order for a proof to be cogent, that warrant for its
conclusion be not prerequisite for having warrant for its premises. Flouting this
requirement would indeed give rise to the kind of transmission failure made familiar
by Wright’s writings on the topic. Furthermore, it would clearly make the proof
44
I am using this label differently from Michael Williams 1996, according to whom “epistemological
realism” is the view that even independently of contextual factors, there is a fact of the matter as to what kind
of justification a belief requires; and, in particular, that our beliefs about the external world must always be
justified by sensory experience if they are to amount to knowledge. This is not the place for a detailed discussion
of Williams’ views, but notice that, on my notion of epistemological realism, its denial isn’t ipso facto an
endorsement of epistemological contextualism—the view that Williams favors. I am willing to grant that only
perceptual evidential grounds can turn empirical beliefs into knowledge.
45
This very issue has been prominent in Davies’ recent writings. See, in particular, Davies 2004 and 2009.
For a discussion and criticism, see Coliva 2010a.
46
Hence, in general, a proof doesn’t necessarily depend on being open-minded with respect to its
conclusion, no matter how one construes the notion of open-mindedness. Davies 2004, p. 240 construes
it as an attitude of either open disbelief in it, or as the attitude of considering that the conclusion “may very
well be false”. In Coliva 2010a I cast doubt on the legitimacy of this understanding of the notion of open-
mindedness. It seems to me that open-mindedness should more naturally be taken to involve having no
attitude with respect to the truth and warrantedness of a given belief.
unavailable as a means to reassure ourselves of the rational legitimacy of our belief in its
conclusion. For we could then claim to have warrant for its premises, which should
transmit to the conclusion, only by already being in a position to claim that we have a
warrant to believe the latter.
Now I agree with Wright’s view that there is something epistemically wrong with
Moore’s Proof. The point is to see whether our diagnoses will converge or not, given
that we conceive differently of the structure of empirical warrants. To repeat, while
Wright thinks that antecedent warrant for (III) is necessary to have warrant for (I),
I think that merely assuming (III) is enough to that effect. That is to say, it is only in a
context where (III) is assumed (however implicitly that might be) that one’s sense
experience will constitute a warrant for a belief such as (I).
Let me approach this final issue in a slightly indirect way. As we saw in section 3, an
externalist construal of the third way would try and provide warrant for the belief in
the existence of an external world by saying that given that we are in the favorable
circumstance of being suitably causally interactive with a world populated by physical
objects, we do have a perceptual warrant for “Here is a hand”, when we seem to see a
hand in front of us, which would then transmit to (III), across the entailment. But,
obviously, the proof of (III) allegedly provided by Moore’s argument would then be
conditional upon correctly relying on (III)’s truth, which is exactly what the Proof
should have provided warrant for. That is to say, the proof of “There is an external
world” would remain conditional upon assuming what was to be proved to no less a
degree than if that conclusion itself had figured among the premises.
Now two things are worth noticing. First that, like in any formal proof, the
conclusion cannot figure among the premises used to prove it. For, otherwise, the
proof would obviously be circular and hence would not provide any warrant to believe
its conclusion.47 To see clearly that this would be the case, consider that, on this
picture, relying on (III)’s truth is what rationally allows us, given a certain course of
experience, to enter the other premises (in fact (I), which, by rational reflection, allows
us to enter (II)) in the proof and therefore get to the conclusion (III). So, in effect, the
Proof would proceed as if it had the following structure:
P
Q
Q!P
--------
P
Secondly, consider a context in which one may want to propound the Proof so
conceived to claim that one’s belief in (III) is warranted. What one would end up
47
In logic one could prove a proposition from itself but, obviously, this would not give one any (new)
warrant to believe it. Since we are here dealing with the power of proofs to generate warrants to believe their
conclusion and claim them, we can discard this case as irrelevant to our present purposes.
saying is that given that (III) is true, then one has warrant for (I) and that that warrant
transmits to (III). But, clearly, one’s argument would remain conditional upon cor-
rectly considering (III) true and the obvious question to raise would be with what right
one can do so. This, however, is a question that cannot be answered by Moore’s Proof
itself, which can only get started, on this picture, by relying on (III)’s truth. Lacking
such an answer, the externalist construal of Moore’s Proof could do nothing to give one
the means to claim a first warrant to believe its conclusion.
Now, I think something broadly analogous to what we have just seen would happen
on the kind of reconstruction of the structure of Moore’s Proof the third way would
command. For, on the third way, the warrant for its first premise would depend on
assuming (III). To repeat, it is only in a context in which (III) is taken for granted (as
implicitly as it may be) that one’s current sense experience can be taken to bear on the
truth of (I) rather than on that of, for instance, “My current experience as of a hand is
envatted experience” (on the assumption that such an experience could be subjectively
indistinguishable, its different causal origin notwithstanding). If so, the proof of (III)
would be conditional upon assuming exactly what was to be proved. Again, this would
make the Proof immediately circular, for the reasons just reviewed. For assuming (III) is
what would rationally allow us to enter the other premises and get to the conclusion in
something like the previous way.
What we should ask, then, is what would happen if, when conceived in the way
recommended by the third view, we run the Proof to claim our warrant for (III). What
we would end up saying is that assuming that there is an external world, one would be
entitled to take one’s experience as of a hand as a (defeasible) warrant for (I), which
would then transmit to (III) across the entailment. But the obvious question to ask is,
once more, with what right one could be assuming (III). This, however, is a question
which cannot be answered by Moore’s Proof itself, since the latter can be run only if
that assumption is already made. That is why, I think, Moore’s Proof would prove, once
more, unable to provide a first warrant to claim that accepting its conclusion is the
rational thing to do. For, again, we could claim warrant for (I) and hence for (III), on
this diagnosis, only by being already warranted in claiming that assuming (III) is
rationally legitimate.
I conclude, therefore, that Moore’s Proof cannot give one any first warrant to believe
(III) or to claim that accepting it is rational. As we may put it, Moore’s Proof is
epistemically inert, when first warrants are concerned: if you don’t already assume its
conclusion, it cannot give you reasons to believe it, nor can it give you reasons to claim
that assuming that conclusion is rationally legitimate, if you don’t already have an
independent warrant to think that it is.
Now, it should be clear that there are obvious analogies between Wright’s diagnosis
of Moore’s Proof ’s failure and mine. But couldn’t it perhaps be the case that we have in
fact hit upon two different species of transmission failure? In general, as we have seen,
transmission failure occurs when a proof or, more generally, an argument cannot
generate any first warrant one may have to believe its conclusion in such a way that
one could appeal to it to reassure oneself of the rational legitimacy of holding that
conclusion. This, on Wright’s view, occurs if and only if warrant for the conclusion is
presupposed in order to have warrant for the premises and, correspondingly, when
claiming the latter would depend on already being in a position to claim the former. By
contrast, on the third way, transmission failure would occur when the structure of
warrants is different—when warrant for the premises merely presupposes the assumption
of the conclusion. Of course, when the vindication of one’s warrant for the conclusion
is concerned, the two views would return a similar verdict on why Moore’s Proof fails,
for they would both say that claiming that one has warrant to believe its premises
presupposes being able to claim that one’s acceptance of its conclusion is indeed
rational. But, again, the ways they go about motivating the rationality of that accep-
tance are substantially different: in Wright’s case, on his understanding of the matter, it
is necessary to exhibit the special kind of warrant that has always accompanied that
acceptance, whether we were able to recognize it or not, thus redeeming it at, as it
were, second order. In my case, in contrast, one has to provide an independent a priori
argument and, therefore, produce an a priori warrant to claim that those otherwise
warrant-free assumptions are indeed rationally made.
Obviously a lot more should be said to defend the view that these are really two
different species of the same kind of phenomenon which may generically be called
“transmission failure”. For instance, it should be further investigated whether other
non-contentious examples of either, but in particular of the second kind of it I am
tentatively putting forward, could be provided. So let me end with a somewhat
provisional note by saying that on the third way the Proof wouldn’t fail because, as
Wright has it, it can’t produce a—as it were—first warrant to believe its conclusion,
since one should already have a warrant to accept (III) in order to be perceptually
warranted in believing its premise. Nor would it fail only when taken in a certain
dialectical—sceptical—context, as Pryor maintains.48 Rather, it would fail because, as a
matter of fact, it would exhibit an even deeper and more basic kind of circularity than
the one exposed by Wright—namely, that which is brought about by flouting the
requirement that the conclusion of a proof shouldn’t be relied on, as implicitly and
unwarrantedly as that might be, in order to have warrant for its premises.
We can thus visualize the various diagnoses of Moore’s Proof ’s failure as follows:
48
Notice that the two different verdicts elicited by Pryor’s model and mine on the reasons as to why
Moore’s Proof fails clearly testify to the difference of our respective overall positions. In particular, when in
section 2 I said that on his view having no reasons to doubt that there is an external world must mean, in
effect, having no view about it, my reason for saying so was that otherwise he could not maintain that the
Proof is epistemically in good standing and merely dialectically faulty. For, as we can now see, if having no
reasons to doubt were taken to mean (implicitly) positively assuming, Moore’s Proof would turn out to be
epistemically circular, although in a sense different from Wright’s. The point can also be put in the form of a
dilemma: either Pryor holds on to his diagnosis of Moore’s Proof ’s failure, in which case having no reasons to
doubt must mean having no view; or else, if having no reasons to doubt is, on his view, equivalent to
(implicitly) positively assuming, he would have to give up his account of the Proof ’s failure.
Conservatives Liberals The third way

W(III) No doubt(III) Assuming (III)
W(I) Here’s a hand W(I) Here’s a hand W(I) Here’s a hand
# (II) If here’s a hand, # (II) If here’s a hand, # (II) If here’s a hand,
# there is an external world # there is an external world # there is an external
world
W(III) There is an external W(III) There is an external (III) There is an external
world world world
Circularity: W(III) is No circularity. But a Circularity: (III) must be
already sceptic assumed in order to have
needed to have W(I) would doubt (III), hence W(I).
would
regard (I) as unwarranted.
Transmission failure I Dialectical ineffectiveness Transmission failure II
As emphasized there is a lot more to say about this second notion of transmission
failure, or indeed about transmission failure tout court. For our present purposes,
however, there is no need to expound on it now.49 My only aim in this essay has
been to sketch a view that, I hope, will repay closer scrutiny and investigation. For, if it
can be sustained, it offers the prospect of giving an alternative characterization of a
number of issues which are prominent in epistemology and which are raised by
Wright’s work and by the work of those who, like Pryor, have engaged with it.
Namely, the topics of the structure of empirical warrants, of the epistemic cogency of
proofs in general, and of Moore’s Proof in particular, as well as the issue of offering a
diagnosis of why scepticism about the existence of an external world is untenable.
Appendix: The Third Way and Wittgenstein’s

On Certainty
Here I would like briefly to compare the third way I have canvassed with some of the remarks in
Wittgenstein’s On Certainty. As is familiar, there is an interpretation that goes back to Strawson,50
according to which Wittgenstein was in fact maintaining a form of naturalism. Passages like }}204,
358, and 35951 are usually taken as suggesting that, according to Wittgenstein, trusting in the
49
I do so in Coliva 2011, where I also take up the issue of whether this second kind of transmission failure
is compatible with retention of the Principle of Closure for warrant under known entailment—whence the
‘?’ in the last line of Moore’s Proof reconstructed along my favorite lines.
50
Strawson 1985.
51
“Giving grounds, however, justifying the evidence, comes to an end;—but the end is not certain
propositions’ striking us immediately as true, i.e. it is not a kind of seeing on our part; it is our acting, which lies
at the bottom of the language-game” (OC }204).
“Now I would like to regard this certainty, not as something akin to hastiness or superficiality, but as a form
presuppositions of our language games is something we simply do, as part of our “form of life”,
and should thus be seen—equivalently, it would seem—“as something animal”.52
I don’t think we can be content with such a position if it is meant to be some kind of response
to scepticism: as we saw, a sceptic could simply acknowledge and in fact insist that it is the very
point of sceptical arguments to show that at the foundations of our epistemic practices there are
presuppositions—like (III)—that we simply assume—let it be because of some kind of ingrained,
animal “instinct”, perhaps due to our overall psychological constitution, as Hume would have it,
or because of our having been brought up within a certain community, which shares that
psychological make-up, as well as a lot of practices and, more generally, “language games”. But it
seems to me quite clear that the whole point of On Certainty is to resist scepticism (and to criticize
Moore as well).53 So, if what we were given was in fact a form of naturalism, then little would
have been done to combat scepticism.
Furthermore, it is true that Wittgenstein writes about trusting in the presuppositions of our
language games as something “animal” and as belonging to “a form of life”. But it shouldn’t be
overlooked that at }358 he also adds “this is very badly expressed and probably badly thought as
well”. I don’t know whether Wittgenstein was dissatisfied with his own way of putting things
because he was somehow aware of the sceptical (or “sceptical friendly”) implications of the
“naturalistic passages” in On Certainty; but, whether or not that was his motivation, the point is
that it should have been.54
The third way I have maintained would in fact remedy the seeming sceptical outcome of
Wittgenstein’s allegedly official line. For it is true, as Wittgenstein has it, that we neither have nor
can have epistemic warrants for the “hinges” of our “language games”. Still, in contrast with the
radical, “postmodernist” outcome of the sceptical considerations, the third way makes for the
possibility that these acceptances are rational after all. So, as we saw, the sceptic would continue
to win as far as lack of epistemic warrants for them is concerned, but he wouldn’t be justified in
claiming that our accepting them lies outside the scope of epistemic rationality. For, if the third
way is on the right track, such an acceptance would lie only, as it were, at the limit of epistemic
rationality, not outside it. The sceptic’s mistake, therefore, would consist in not seeing the width
of our notion of rationality, thereby declaring irrational what is merely epistemically unwarrant-
able (as well as irrefutable) and indeed necessary in order to have the notions of epistemic
rationality and warrant he wants to make use of.
Interestingly, at various points in On Certainty, Wittgenstein prefigures what I take to be the
gist of the third way. He writes:
It belongs to the logic (italics mine) of our scientific investigations that certain things are in deed not
doubted. (OC }342)
of life. (That is very badly expressed and probably badly thought as well)” (OC }358).
“But that means I want to conceive of it as something that lies beyond being justified or unjustified; as it
were as something animal” (OC }359).
But see also OC }}144, 148, 196, 232, 402.
52
I don’t actually think that these two twists of phrase are equivalent: a form of life includes cultural as
well as natural elements; or else, in McDowell’s (1994) terminology: elements which belong to our “second
nature” as well as to our first, merely animal one.
53
See Coliva 2010c.
54
For an anti-naturalist reading of Wittgenstein, see Marconi 1987, pp. 127–8 and Wright 2004b.
Everything descriptive of a language game is part of logic. (OC }56, italics mine)
[S]omewhere I must begin with not-doubting; and that is not, so to speak, hasty but excusable:
it is part of judging. (OC }150, italics mine)
The “hinges”, among which “There is an external world” would figure,55 can thus be seen as
part of “logic” and, in this sense, their acceptance would be rational. Despite the fact that, as is
well known, for Wittgenstein they aren’t elements within the language games in which reasons
for and against empirical propositions are produced, and therefore reasons for or against them
can’t be produced either, they are still part of “what is descriptive of a language game” and hence
of “logic”. So they do not lie outside it. Thus, it seems that if they still belong to “logic” but aren’t
within the language game, unlike ordinary empirical propositions, they can only lie at, as it were,
its limit.
Of course, “logic” is here safely replaceable by “grammar”—in the typically Wittgensteinian
sense of the term—and we know that Wittgenstein wanted to regard these “hinge” propositions
as somewhat like empirical propositions (OC }}401–2) which, however, have lost—or have
never played, so far—that role (OC }}96, 97), for they are not subject to verification and control,
unlike genuine empirical propositions. Rather, on his view, they play a different role—viz. that
of founding rules of all our language games (OC }95). Rules, we might say, of evidential
significance—that is, of what our sense experience is going to bear on—rather than rules which
simply allow or exclude some conceptual combinations, like “patience is played by one person
alone” or “an object can’t be red and black all over at one and the same time”.56
This is not the place to dwell further on this aspect of Wittgenstein’s thought. What I would
like to stress, however, is that there is some textual evidence (though no doubt contradicted at
other places) that for Wittgenstein too their acceptance—the acceptance, that is, of what he
considered as propositions about objects (OC }402), which played the role of rules and not of
genuine empirical propositions—was indeed rational, and not merely something brute and
instinctive; or else, merely the product of our upbringing within a certain form of life. The
third way I have canvassed capitalizes on these hints.
Furthermore, it is noteworthy that the third way defends the rationality of accepting (III) by
appealing to the conceptual or grammatical link between that acceptance and our notion of
epistemic rationality—an appeal that would be in keeping with the later Wittgenstein’s way of
dissolving so-called “philosophical problems”, and, in particular, with his rule-following con-
siderations, at least according to some interpretations of those passages.57
Another point of comparison between Wittgenstein’s remarks in On Certainty and the third
way concerns the allegedly relativistic implications of his thought many scholars have thought they
discerned.58 This is a very delicate issue, whose proper treatment will have to be deferred to
55
This is a contentious claim, which I defend in Coliva 2010c, Chapter 3.
56
I am grateful to Crispin Wright for suggesting this difference to me. Moyal-Sharrock 2005 too draws a
similar distinction.
57
Most notably, Backer and Hacker 1984.
58
Prominent examples are Rorty 1979, p. 317 and, more recently, Boghossian 2006, pp. 69–80. For
present purposes I am passing over the important fact that Wittgenstein might have been quite inclined to
relativism regarding the presuppositions of what I shall call, collectively, “non-basic” language games (see his
example of physics vs. consulting oracles, OC }}609–12, but also OC }}132, 336, 671) but much more
cautious about those of our most basic language games, such as our practice of holding, assessing, and
another occasion.59 Here, however, I would just like to table some considerations. According to
relativist interpretations of his thought, Wittgenstein would maintain that if we somehow needed
or wanted to change our “language games”, we would be at liberty to do so, thus abandoning their
presuppositions. But we could only do so—the train of thought continues—if nothing forced us to
hold on to them, come what may; and if nothing so forces us, it is because these presuppositions are
neither true, nor false;60 neither warranted, nor unwarranted;61 neither rational, nor irrational.62
They stay put only as long as we need or want to engage in the practices they are constitutive of.
Perhaps the circumstances in which we would be willing to give them up are far-fetched, yet it is
only if we viewed the presuppositions of our language games as lying outside the domain of
warrant, rationality, and truth-assessibility that we could have such room for manoeuvre.
There is something right and something wrong here: what is right, I think, is that we don’t,
nor can we have perceptual warrants for such presuppositions and, thus, nothing which would
bear on their truth. This, in turn, also helps us see why, for Wittgenstein, properly speaking, they
are neither true nor false: since we can’t have warrants either for or against them, we can’t claim
they are either true or false. Yet, to think that they may be true or false, independently of our
possibility of acquiring evidence for or against them, would simply mean to hold on to a realist
conception of truth that, arguably, the later Wittgenstein was distancing himself from. What is
wrong is to think that, as a consequence of their being unwarranted (and unwarrantable), these
presuppositions aren’t rationally held, if the third way is somehow right; and also that our
accepting them is merely conventional or motivated simply by practical considerations.
Still, to say that presuppositions such as (III) are rationally held doesn’t commit one to saying that
the hinges can’t be changed at all. If “something really unheard of ”63 were to happen—that is, if,
for some reason, we could no longer be the epistemic creatures we in fact are—then of course the
presuppositions of our present (basic) epistemic practice, which is itself constitutive of epistemic
rationality, may no longer be in place. It is a matter of some controversy whether Wittgenstein
thought that these radical changes, which would result in totally different systems of justification
and, consequently, in totally different notions of rationality, were really conceivable, or were just
remote metaphysical possibilities, actually beyond our real understanding.64 Be that as it may, the
fact that epistemic practices might be totally different from what they are now, and hence that there
might be a totally different notion of rationality, shows merely that neither the former nor the latter
are metaphysically grounded. Holding this form of anti-foundationalism, however, seems still a
long way short from thinking that there actually are radically different systems of justification,
which are all equally legitimate. If so, there is little room for maintaining that Wittgenstein was a
relativist.65 Rather, it appears that, at most, he was merely an anti-foundationalist.
withdrawing from empirical beliefs on the basis of perceptual evidence, of which (III) is an example.
However, what I will be saying primarily concerns these basic presuppositions.
59
I actually develop this theme in Coliva 2010b, c.
60
OC }205 (but contrast with OC }206).
61
OC }359 (but see also OC }}307, 599).
62
OC }559, but notice that the English translation obscures this point because it renders the German
“vernünftig” “unvernünftig” with “reasonable” and “unreasonable”, respectively. This is not a mistake, but a choice
of words that makes it less clear that Wittgenstein is here talking about rationality and not mere reasonableness.
63
OC }513 (italics mine), and that may well be impossible to conceive given our present form of life.
64
Stroud 1965 denies it, Marconi 1987 doesn’t. In Coliva 2010b, c I side with Stroud.
65
Marconi 1987, pp. 122–33 argues that Wittgenstein was a “virtual ” relativist. I take issue with that in
Coliva 2010b, c.
Bibliography
Backer, G. and Hacker, P. 1984 Scepticism, Rules and Language, Oxford, Blackwell.
Beebee, H. 2001 “Transfer of Warrant, Begging the Question and Semantic Externalism”,
Philosophical Quarterly, 51, pp. 356–74.
Blackburn, S. 2005 Truth, London, Penguin.
Boghossian, P. 2006 Fear of Knowledge, New York, Oxford University Press.
Brown, J. 2005 “Doubt, Circularity and the Moorean Response to Scepticism”, Philosophical
Perspectives 19, pp. 1–15.
Burge, T. 1993 “Content Preservation”, The Philosophical Review 102, pp. 457–88.
Burge, T. 2003a “Perceptual Entitlement”, Philosophy and Phenomenological Research 67, pp. 503–48.
Burge, T. 2003b “Reply to Martin Davies”, in M. J. Frapolli, and E. Romero (eds.)
Meaning, Basic Self-Knowledge and Mind: Essays on Tyler Burge, Stanford, CSLI Publications,
pp. 250–7.
Coliva, A. 2004 “Proof of an External World: Transmission Failure, Begging the Question or
Dialectical Ineffectiveness? Moore, Wright and Pryor”, in A. Coliva and E. Picardi (eds.)
Wittgenstein Today, Padova, Il Poligrafo, pp. 397–415.
Coliva, A. 2007 “Lo scetticismo sul mondo esterno”, in A. Coliva (ed.) Filosofia analitica. Temi
e problemi, Roma, Carocci, Chapter 8.
Coliva, A. 2008 “The Paradox of Moore’s Proof of An External World”, Philosophical Quarterly
58, pp. 234–43.
Coliva, A. 2009 I modi del relativismo, Roma-Bari, Laterza.
Coliva, A. 2010a “Moore’s Proof and Martin Davies’s Epistemic Projects”, Australasian Journal of
Coliva, A. 2010b “Was Wittgenstein an Epistemic Relativist?”, Philosophical Investigations 33,
pp. 1–23.
Coliva, A. 2010c Moore and Wittgenstein. Scepticism, Certainty and Common Sense, Basingstoke,
Palgrave Macmillan.
Coliva, A. 2011 “The Varieties of Failure (of Warrant Transmission—What Else?)”, Synthese,
On-line first, DOI: 10.1007/s11229-011-0006-6.
Coliva, A. 2012 Scetticismo. Dubbio, paradosso e conoscenza, Roma-Bari, Laterza.
Coliva, A. forthcoming a “Moderatism, Transmission Failures, Closure and Humean Scepticism”,
in D. Dodd and E. Zardini Contemporary Perspectives on Scepticism and Perceptual Justifica-
tion, Oxford, Oxford University Press.
Coliva, A. forthcoming b Within a System. A Plea for Moderatism, ms.
Davies, M. 2004 “Epistemic Entitlement, Warrant Transmission and Easy Knowledge”, The
Aristotelian Society Supp. Vol. LXXVIII, pp. 213–45.
Davies, M. 2009 “Two Purposes of Arguing and Two Epistemic Projects”, in I. Ravenscroft
(ed.) Minds, Ethics, and Conditionals: Themes from the Philosophy of Frank Jackson, Oxford,
Jackson, F. 1987 Conditionals, Oxford, Blackwell.
Jenkins, C. 2007 “Entitlement and Rationality”, Synthese 157, pp. 25–45.
Kelly, T. 2003 “Epistemic Rationality as Instrumental Rationality: A Critique”, Philosophy and
Phenomenological Research 66, pp. 612–40.
McDowell, J. 1982 “Criteria, Defeasibility and Knowledge”, Proceedings of the British Academy 68,
pp. 455–79.
McDowell, J. 1994 Mind and World, Cambridge, Mass., Harvard University Press.
Marconi, D. 1987 L’eredità di Wittgenstein, Roma-Bari, Laterza.
Moore, G. E. 1939 “Proof of an External World”, reprinted in Philosophical Papers, London,
Allen & Unwin, pp. 126–48.
Moore, G. E. 1942, “A Reply to my Critics”, reprinted in P. A. Schillp (ed.) The Philosophy of
G. E. Moore, 1999, Evanston, Ill., Northwestern University Press, pp. 533–677.
Moyal-Sharrock, D. 2005 Understanding Wittgenstein’s On Certainty, Basingstoke, Palgrave
Macmillan.
Moyal-Sharrock, D. and Brenner, W. H. (eds.) 2005 Readings of Wittgenstein’s On Certainty,
Basingstoke, Palgrave Macmillan.
Peacocke, C. 2004 The Realm of Reason, Oxford, Oxford University Press.
Percival, P. 2002 “Epistemic Consequentialism I”, Proceedinsg of the Aristotelian Society Supp.
Vol. LXXVI, pp. 121–51.
Pritchard, D. 2005 “Wittgenstein’s On Certainty and Contemporary Anti-Scepticism”, in
D. Moyal-Sharrock and W. H. Brenner (eds.) Readings of Wittgenstein’s On Certainty,
Basingstoke, Palgrave Macmillan, pp. 189–224.
Pryor, J. 2004 “What’s Wrong with Moore’s Argument?”, Philosophical Issues 14, pp. 349–78.
Pryor, J. 2005 “There is Immediate Justification”, in M. Steup and E. Sosa (eds.) Contemporary
Debates in Epistemology, Oxford, Blackwell, pp. 181–202.
Rorty, R. 1979 Philosophy and the Mirror of Nature, Princeton, NJ, Princeton University Press.
Schiffer, S. 2004 “Skepticism and the Vagaries of Justified Belief ”, Philosophical Studies 119/1–2,
pp. 161–84.
Schillp, P. A. (ed.) 1999 The Philosophy of G. E. Moore, Evanston, Ill., Northwestern University
Press.
Silins, N. 2005 “Transmission Failure Failure”, Philosophical Studies 126/1, pp. 71–102.
Silins, N. 2007 “Basic Justification and the Moorean Response to the Skeptic”, in
T. Szabo–Gendler and J. Hawthorne (eds.) Oxford Studies in Epistemology, vol. 2, Oxford,
Stalnaker, R. 2002 “Epistemic Consequentialism II”, Proceedings of the Aristotelian Society Supp.
Vol. LXXVI, pp. 153–68.
Strawson, P. 1985 Skepticism and Naturalism: Some Varieties, London, Methuen.
Stroud, B. 1965 “Wittgenstein and Logical Necessity”, Philosophical Review 68, pp. 504–18.
White, R. 2006 “Problems for Dogmatism”, Philosophical Studies 131/3, pp. 525–57.
Williams, M. 1996 Unnatural Doubts: Epistemological Realism and the Basis of Scepticism, Princeton,
NJ, Princeton University Press.
Wittgenstein, L. 1922 Tractatus Logico-Philosophicus, trans. D. F. Pears and B. F. McGuinness,
1979, London, Routledge & Kegan Paul.
Wittgenstein, L. 1969 On Certainty, Oxford, Basil Blackwell.
Wright, C. 1985 “Facts and Certainty”, Proceedings of the British Academy 71, pp. 429–72.
Wright, C. 2004a “Warrant for Nothing (and Foundations for Free)?”, Proceedings of the Aristotelian
Wright, C. 2004b “Wittgensteinian Certainties”, in D. McManus (ed.) Wittgenstein and Scepticism,
London, Routledge, pp. 22–55.
Wright, C. 2007 “The Perils of Dogmatism”, in S. Nuccetelli and G. Seay (eds.) Themes
from G. E. Moore: New Essays in Epistemology and Ethics, Oxford, Oxford University Press,
pp. 25–48.
13
Wright Against the Sceptics
Michael Williams
1. Introduction
Crispin Wright has thought long and hard about scepticism. He offers both a powerful
analysis of how the most significant sceptical arguments work and a general strategy for
blunting their destructive force.1
Wright’s sceptic follows what I shall call the Cornerstone Strategy. In Wright’s termi-
nology, a cornerstone is a deep presupposition of the justificatory procedures appropriate to
some area of discourse. The sceptic argues that, since cornerstones lack justification
(Wright likes to say “warrant”), cornerstone-dependent beliefs lack warrant too. So for
example, the belief that experience is a more or less reliable guide to at least the gross
features of our immediate environment is a cornerstone for our acquiring justified beliefs
about the external world. Since we cannot validate the reliability of experience, none of
our beliefs about the world around us is justified. Wright replies that this argumentative
strategy involves a lacuna. Cornerstones are rationally warranted, but in a way that the
sceptic overlooks. They are rational but unearned entitlements.
Wright’s way with scepticism is influenced by his reading of Wittgenstein, particu-
larly On Certainty.2 Since I too find inspiration in On Certainty, I am sympathetic to a
number of things that Wright has to say. Nevertheless, I find his response to scepticism
unsatisfying. Wright’s anti-sceptical ambitions are strangely modest: his response to
scepticism, he tells us, is only an exercise in “damage limitation”.3 But if an argument
contains a lacuna, surely it is fallacious, and how can a fallacious argument do damage?
The answer must be that, while the sceptic over-reaches, his argument contains a
sound core. Accordingly, whether Wright’s response to scepticism is satisfactory will
depend on how much damage he lets the sceptic get away with. Too much, I shall
argue. Wright’s response is far more concessive that it should be. I shall conclude by
suggesting how a much less concessive response may be attempted.
1
See in particular Wright 1985, 2004a, 2004b.
2 3
Wittgenstein 1969. Wright 2004a, p. 206.
W R I G H T AG A I NS T T H E S C E P T I C S 353
Coming to grips with Wright’s anti-sceptical strategy proves to be a tricky business.

Not only is Wright’s response to scepticism more concessive than a brief description
suggests, it is also much more complex. Indeed, I think that Wright’s presentation of his
position is misleading in at least two, intimately related ways.
1. Wright’s strategy seems simple and direct. The sceptic claims that certain beliefs are
unwarranted: Wright replies that this is not so, that the sceptic has an impoverished
conception of how warrant comes to be possessed. But Wright’s approach cannot be
that simple. If he were simply supplying what the sceptic claims we don’t have, Wright
wouldn’t be engaging in damage limitation. Consider the distinction between earned
and unearned income. Money is money: it buys the same goods and services however
I come by it. So if my salary is replaced by dividends and capital gains, I am not worse
off. Maybe I’m better off: I have the money but I don’t have to work for it. A private
income is not a form of financial damage limitation. Warrant, as Wright understands it,
cannot be like that. The crucial feature of entitlements, I shall claim, is not that they
replace the earned warrants the sceptic denies us with unearned warrants of the same
kind but that they offer earned warrants of a different kind.
2. Wright’s argument is informed by a crucial but under-emphasized distinction
between philosophical scepticism and scepticism about philosophy. By “philosophical scepti-
cism” I mean the view that we do not have any knowledge of (say) the external world:
more radically, that our beliefs about the world are not so much as justified. By
“scepticism about philosophy” I mean the meta-sceptical view that we cannot explain,
in a “philosophical” way, how it is that our beliefs about the world are justified, or
amount to knowledge. As we shall see, Wright’s theory of entitlements, though
naturally first taken as a response to philosophical scepticism, is really a response to
scepticism about philosophy. As a result, Wright’s response to philosophical scepticism
is cursory and largely independent of his theory of unearned warrant. But even as a
response to scepticism about philosophy, that theory is too concessive and, arguably,
unsuccessful in its own terms.
That is the prospectus. Let me try to make good on it.
2. Scepticism: Sources and Limits

Why take scepticism seriously in the first place? I suggest two conditions that a seriously
problematic version of scepticism must meet.
The first condition is that the sceptic’s conclusions must appear to be unacceptable
because seriously damaging. If we found a particular sceptical claim benign, or at least
innocuous, we would not worry about how to avoid or accommodate it. Nor would
there be much to a sceptical problem if the sceptic’s conclusion could be headed off by a
trivial modification of our pre-theoretical thinking about knowledge or justification.
Sceptical claims must threaten the intellectual integrity of our practices of epistemic
evaluation in important ways. Or they must threaten our understanding of those
354 MICHAEL WILLIAMS
practices in a way that is itself disquieting, for example by seeming to deprive us of a

kind of understanding that we have good reason both to want and to expect to be
available. Call this the Unacceptability Condition.
The second condition is that the sceptic’s arguments must grip us, at least when we first
encounter them. However damaging the sceptic’s conclusions may be, they will
amount to a threat only if we suspect that we are committed to them. Thus the sceptic’s
arguments must at least appear to be compelling because intuitive. By calling them
“intuitive”, I mean that the sceptic must derive his conclusions from what anyone one
can see, on reflection, to be features of ordinary epistemic concepts. As Barry Stroud puts
it, the sceptic’s damaging conclusions are troubling because they appear to follow from
“platitudes we would all accept”.4 Call this the Intuitiveness Condition.
The two conditions are related in an obvious way. Assuming that we have reason to
value (say) the distinction between justified and unjustified belief, as ordinarily drawn,
arguments for justification-scepticism must threaten that distinction and not some
caricature of it. If sceptical arguments are not intuitive, sceptical conclusions will not
be damaging. (At least, it will take a lot of argument to show that they are.)
Let us look at Wright’s reconstruction of the sceptic’s argument, in the light of these
two conditions. Wright says that a proposition is a cornerstone for a given region of
thought “just in case it would follow from a lack of warrant for it that one could not
rationally claim warrant for any belief in the region”.5 The most compelling arguments
for scepticism now go by way of two lemmas. The first is that a certain proposition that
we accept is indeed a cornerstone for a given region of thought. The second is that we
have no warrant for accepting it. The sceptic concludes that no beliefs in the region are
warranted.
We can see from this that Wright’s understanding of scepticism is broadly similar to
mine. In arguing for the first lemma, Wright’s sceptic is trying to show that his
argument does not distort ordinary epistemic procedures, thus that it meets the
Intuitiveness Condition. And since his conclusion threatens to erase the fundamental
distinction between warranted and unwarranted belief, it surely seems damaging.6
So much for cornerstones in general. Moving closer in, Wright sees two ways of
implementing the Cornerstone Strategy, each yielding a distinct kind of sceptical
argument. Wright calls the two kinds of argument “Cartesian” and “I-II-III”. For
reasons that will become apparent, I shall call the members of the I-II-III family Locality
arguments.
In giving the Cornerstone strategy a Cartesian twist, the sceptic first argues that it is a
cornerstone for many of my beliefs that I am not cognitively disabled or detached from
reality in some pervasive way, as I would be in certain sceptical scenarios: for example,
if I were trapped in a persistent, coherent dream, if I were the victim of a malin génie, or
if I were a brain a vat. He then argues that we can have no warrant for ruling out such
4 5 6
Stroud 1984, p. 82. Wright 2004a, pp. 167–8. Ibid., p. 206.
W R I G H T AG A I NS T T H E S C E P T I C S 355
systematic disablement, precisely because it is systematic. For example, suppose that

I think up a test for whether I am dreaming right now: pinching myself, say. The
evidence garnered by executing such a test can be no stronger than my independent
grounds for thinking that the test was properly carried out: a fortiori, no stronger than
my warrant for thinking that it was carried out at all and not merely dreamt. I can
acquire a warrant for the proposition that I am not dreaming only if I already have one.
The sceptic concludes that no warrant for this cornerstone proposition can possibly be
obtained.
The Locality version of Cornerstone Strategy is typified by Hume’s argument for
inductive scepticism. In the simplest case of inductive inference, we take a natural
pattern observed in some finite number of cases (without observed counter-examples)
to provide a defeasible warrant for an unqualifiedly general claim: we go from “All
observed F’s are Gs” to “All Fs are Gs”. Such an inference is reasonable only if the
world contains many exceptionless regularities: that is, inductive inference presupposes
what is sometimes called the “Uniformity of Nature” thesis. The Uniformity thesis
does not make inductive inferences deductively valid: nothing can do that. Rather,
its role is to “provide an informational setting in which the observed pattern of
co-occurrence between Fs and Gs defeasibly warrants generalization”.7 If the thesis
holds, our observational evidence may have presented instances of a genuine regularity.
Of course, our sample of observed instances can be misleading, reflecting no genuine
regularity: that is why inductive warrants are defeasible. But if the Uniformity thesis is
false, there will be no regularities to be right or wrong about. In this case, an observed
correlation will provide no warrant for a general conclusion. However, while the
Uniformity thesis is a cornerstone for beliefs concerning empirical generalizations, it
is itself an empirical generalization, hence justifiable—if at all—only by inductive
inference. Since all such inference presupposes the thesis, it seems that the thesis itself
cannot be warranted in a non-circular way. It is a cornerstone for which we can have
no warrant.
Hume’s inductive scepticism offers a template for a whole family of sceptical
arguments. Let P be any proposition concerning some perceptible fact about the
world around me: e.g. that I have two hands. Then consider:
I. My current experience is in all respects as if P.
II. P.
III. There is a material world.
Of course, experience sometimes misleads, so the warrant that I provides for P is
defeasible. Nevertheless, I remains the best possible evidence anyone could have for
P. But as with inductive evidence, the warrant that I provides is “information-
dependent” rather than unconditional. The inference from I to P is a good one only
7
Ibid., p. 170.
if experience is a reliable, albeit imperfect, guide to the world around us: a fortiori, only
if there is a world to be right or wrong about. III is thus a cornerstone for beliefs about
the material world. But how can I learn that there is a material world, except by way of
acquiring beliefs about particular worldly facts? There is no other way: that there
is a material world is an empirical fact, verifiable by experience, if verifiable at all.
We cannot try to warrant belief in the material world without presupposing that
very belief. Here, too, we find a cornerstone for which we can have no warrant.
Parallel arguments can be mounted with respect to our knowledge of other minds,
or of the past. My knowledge of another person’s inner life is necessarily indirect.
However, observing another person’s behavior warrants my crediting him with
subjective states only if such states have some characteristic outer expression: a fortiori,
only if there are such states to be expressed. Or again, my apparent memories warrant
beliefs about past events only if they are reliable indicators of what went before: i.e.
only if there is a past and the world did not, as Russell once suggested, come into
existence five minutes ago complete with delusive memories and other purported
but misleading traces of considerable antiquity. Fill in the details. Our commitments
to the existence of other minds and the reality of the past emerge as further
warrantless cornerstones.
Arguments following the I-II-III template presuppose that beliefs in the target-area
are justified, if at all, by inference from some restricted range of information. Does this
compromise their intuitiveness? If, according to our ordinary way of thinking, what we
typically know through perception is how things are in the world, the sceptic’s
argument may not trouble us. Wright replies that, while the “justificational architec-
ture exploited by I-II-III scepticism” might be challenged in certain applications, there
is no prospect or repudiating it generally. Even if we have some direct knowledge of
worldly facts, we do not have much. In this sense, informational impoverishment is the
fate of any finite inquirer. Wright calls this feature of our epistemic situation cognitive
locality, hence my name for the arguments that exploit it.
Although Wright distinguishes Cartesian from Locality arguments, his fundamental
objection is to the shared strategy they exemplify. That is why his theory of unearned
warrants constitutes a unified response. However implemented, the Cornerstone Strat-
egy is fallacious in virtue of ignoring the possibility of entitlements. But how are
entitlements themselves possible? Wright offers four suggestions.
a. Strategic Entitlement. I am stranded on a desert island. I have no idea whether the
fruits I find there are safe to eat. But since not eating them will ensure starvation, eating
them is strategically rational: it dominates not eating them. Reichenbach suggests
treating inductive inference along similar lines. If nature is uniform, induction will
work. If nature is not uniform, no method of ampliative inference will work. Accord-
ingly, adopting inductive methods is a dominant strategy. But whereas my entitlement
to eat the problematic fruits is restricted to a special context, inductive methods are
needed across the board. My entitlement to them is absolute.
W R IG H T AG A I N S T T H E S C E P T I C S 357
b. Entitlement of Cognitive Project. Wittgenstein notes that,

One cannot make experiments if there are not some things that one does not doubt . . . If I make
an experiment I do not doubt the existence of the apparatus before my eyes. I have plenty of
doubts, but not that.8
Not only would doubting the existence of the apparatus disable me with respect to
conducting experiments, I would not be able to muster the commitment needed to
carry through a cognitive project if I had doubts rationally committing me to ques-
tioning the project’s significance or competence. In the context of the experiment,
then, such scruples can reasonably be set aside, unless there are positive grounds for
entertaining them. By thus reasonably excluding doubt, entitlements of project war-
rant genuine acceptance and not mere acting on certain assumptions. Wittgenstein
finds such entitlements pervasive: whenever we conduct an inquiry (or offer a justifi-
cation), we will rely on presuppositions that are exempted from scrutiny. This does not
mean that they cannot be questioned or justified, only that to make them the object of
investigation would be to create a new context of inquiry with presuppositions of its
own. However, according Wright, entitlement of cognitive project is of limited
significance, for while it makes some headway against Cartesian scepticism, it does
little to blunt the force of Locality arguments. For the purpose of getting on with an
experiment, I can justifiably ignore worries about whether I am dreaming right now.
But all I am presupposing here is that my perceptual faculties are functioning properly
on the present occasion; and the question raised by a Locality argument for external world
scepticism concerns my warrant for supposing that they ever function properly.
c. Entitlement of Rational Deliberation. We are not just inquirers but actors, faced with
decisions about what it is rational to do. Since we cannot deliberate without anticipat-
ing the results of various proposed courses of action, we are entitled to inductive and
abductive methods (hence to the presupposition of a nomically regular world). Further,
since those methods depend on evidence previously gathered, entitlement of rational
deliberation gains a foothold with respect to scepticism about the past.
d. Entitlements of Substance. We have made some progress against Locality-based
scepticism. However, in contrast to the broadly methodological commitments involved
in inductive and abductive inference, accepting the existence of the world involves a
conception of the nature of reality, hence ontological commitment. In an unexpected
twist, Wright suggests that we might locate the source of entitlement to this substantial
commitment in cognitive locality itself. We cannot in general repudiate cognitive
locality without giving up the very idea of a world extending spatially and temporally
(perhaps infinitely) beyond one’s “cognitive horizon”. Thus in the very acknowledge-
ment of our informational limitations, we conceive ourselves as always occupying a
particular place in the world. Cognitive locality thus implies an unearned warrant for
8
Wittgenstein 1969, } 337.
accepting the existence of the world. The very idea that generates the sceptical problem
offers the key to its solution.
3. Varieties of Justification
If Wright’s response to scepticism is an exercise in damage limitation, what exactly is the
damage that needs to be limited? Or putting the question the other way round, what
do entitlements give us, and how is it less than we hoped for (or thought we had)? To
clear these matters up, we must make some distinctions.
Following James Pryor, let us distinguish between our having reason to believe
something and our being justified in believing it.9 One can have reason to believe
p without being justified in believing p. For one thing, one does not always believe
what one has reason to believe. For another, one can be unjustified in believing what
one has reason to believe: for example, if one has other beliefs (unjustified but firmly
held) that either undermine one’s reasons for believing p or imply that p is false. Notice
that to say that someone has no reason to believe p can be to make a factual claim. But to
say that he is not justified in believing p is to make an evaluative or normative claim. It is
to imply that (in some way or other) he is not entitled to believe p.
We can distinguish two kinds of reasons. Although the distinction is not altogether
easy to make in a completely precise way, almost all epistemologists distinguish epistemic
from pragmatic reasons. Roughly, to have an epistemic reason to believe p is to have
evidence that makes p (more) likely to be true, whereas to have a pragmatic reason to
believe p is for believing p to have some kind of desirable payoff, whether or not p is
true. For example, their alleged consoling effects are a pragmatic reason for believing
various religious dogmas but provide no epistemic warrant. We might say that, whereas
epistemic reasons are rooted in our evidence, or in the reliability of our faculties,
pragmatic reasons are rooted in our interests.
The distinction between epistemic and pragmatic reasons suggests a correlative
distinction between ways of being justified in believing p: epistemically if one’s reasons
are epistemic, pragmatically if they are pragmatic. However, not all cases of being
justified in believing something are cases of having reason to believe it. If one has
reason to believe p, and one’s so doing is not blocked by other beliefs, then one is
positively authorized to believe p. However, there is a weaker notion of being justified,
connected with the commonsense idea of entitlement. One can be entitled to do
something if it is not prohibited. On this weaker conception, justified beliefs are beliefs
that are rationally permissible: there are no reasons, epistemic or pragmatic, for not
holding them, though perhaps one has no positive reason to hold them. This is default
entitlement. Even more weakly, a belief may be justified if it is non-culpable. Perhaps, in
9
Pryor 2004, see especially p. 352. Pryor is concerned exclusively with epistemic justification. He has
some very interesting things to say about Wright’s idea sceptical arguments exploit “transmission failure”.
WRIGHT AGAINST THE SCEPTICS 359
some way, believing p is not a good thing to do. But I can’t be blamed because I have a
good excuse: for example, I couldn’t help it.
Wright holds that entitlements give us “warrant for nothing” or “foundations for
free.” They are not earned by doing any “specific evidential work”.10 By itself,
however, this doesn’t tell us much. For one thing, entitlements are not the only
examples of unearned warrant that Wright recognizes. According to the justificational
architecture exploited by Locality arguments, earned warrants are acquired by infer-
ence from beliefs for which the warrant is (at least in some cases) unearned. Thus, while
warrant for beliefs concerning particular worldly facts is earned by inference from
experiential evidence, whatever experience tells us it tells us “directly,” i.e. non-
inferentially. No evidential work is required. So aren’t all foundations “for free”?
Experiential knowledge involves “direct” awareness of how things appear. On this
view, experiential beliefs are, as I like to say, intrinsically credible, so that the unearned
warrant attaching to experiential beliefs goes beyond both non-culpability and default
entitlement. Though non-inferential, experiential beliefs enjoy positive epistemic
authorization. What about Wright’s entitlements? They are not intrinsically credible,
since for all we know they could be false. But neither is our commitment to them
merely default or non-culpable. In treating entitlements as commitments possessing
unearned warrant, Wright does not mean that they are propositions that we (permissi-
bly) accept without reason, or that we just can’t help accepting. His point is that our
reasons are not evidence. Our reasons for accepting cornerstones are pragmatic rather
than epistemic.
We are now in a position to unpack the idea of damage limitation. The sceptic is
correct in arguing that we have no epistemic reasons for accepting cornerstones. He is
therefore correct to conclude that our accepting them is epistemically unjustified.
Indeed, from a purely epistemic standpoint, we have no right to accept them. As Wright
candidly notes, his strategy “concedes the basic point of the sceptical arguments to
which it reacts, namely that we do indeed have no claim to know, in any sense
involving evidence for their likely truth, that certain cornerstones of what we take to
be procedures yielding knowledge and justified belief hold good”.11 Following
through on this concession to the sceptic, Wright argues that we should not think of
cornerstones as beliefs, but rather as the objects of rational trust. So conceived, entitle-
ments should be understood in terms of the more generic notion of acceptance
indicated by such ordinary phrases as “acting on the assumption that” or “taking it
for granted that”.12 This is revealing. Such a conception of acceptance is appropriate to
commitments that are only pragmatically warranted.
We now see how, in Wright’s eyes, the sceptic over-reaches: he infers an unquali-
fied normative conclusion from a restricted epistemological claim. Having shown that
cornerstones are not epistemically justified, he concludes that they are unjustified sans
10 11 12
Wright 2004a, p. 174. Ibid., p. 206. Ibid., p. 203.
phrase. This is not so. Cornerstones have rational warrant, albeit pragmatic rather than
epistemic.
I should stress that Wright himself does not characterize the warrant for cornerstones
as pragmatic. There is some justification for this, in that entitlements involve pragmatic
reasons of a special kind. In a crucial respect, these reasons differ from such merely
pragmatic considerations as the consolations of religion. Cornerstones are presupposi-
tions of rational inquiry. They give us a chance of discovering truths in various regions of
discourse, if there are truths to be discovered. The interests that cornerstones serve are
thus, at least in part, epistemic. However, the fact that the reasons for accepting
cornerstones reflect epistemic interests does not make them epistemic reasons.
Before turning to the question of whether there is anything damaging in Wright’s
concession to the sceptic, I shall make two critical comments on the way Wright
presents his anti-sceptical strategy.
First, Wright mis-characterizes his response. According to Wright, the sceptic goes
wrong by overlooking the possibility of unearned warrants. This is wrong. Wright’s
entitlements are not propositions accepted for no reason. They are not default credible.
Accordingly, they are not really unearned. Rather, they are not earned by doing
“specific evidential work”. They are earned—or are at least earnable—by showing
how they further essential epistemic interests. But this raises a question, namely,
whether to possess warrant for a cornerstone one must oneself appreciate the reasons for
that warrant. To answer that one must is characteristic of an internalist approach to
reasons. Wright is understandably reluctant to set the bar for warrant so high. He finds
an analogy with basic logic suggestive: when someone follows a valid pattern of
inference, “we will naturally credit [her] with warrant to proceed as she does, even if
she has given no explicit thought to that way of proceeding and would not have the
slightest idea how to answer if a request for justification was made”.13 Cornerstone
commitments are warranted if there are good pragmatic reasons for them. But not
everyone need appreciate those reasons: that is the business of philosophers. Thus
Wright adopts an externalist attitude towards pragmatic reasons. Obviously, Wright’s
externalism points to a way in which, for most people, warrant for cornerstones is
unearned. So I can sum up my criticism by saying that Wright’s idea of foundations for
free runs together two distinctions: that between two kinds of reasons (epistemic and
pragmatic) and that between two ways of understanding the relation of reasons to
doxastic justification (internalist and externalist).
My second point is that, even if some sceptics over-reach, not all do. Wright’s
blanket charge of fallacy is thus itself a step too far. Some sceptics readily concede that
our beliefs are justified in various non-epistemic ways: that we have some kind of
pragmatic justification—or at least psychological excuse—for holding them. As Wright
recognizes, Hume is a case in point. Not only is Wright’s understanding of how
13
Wright 2004a, p. 204.
scepticism arises modelled on Hume’s, in the broadest strategic terms his response is
Humean also.
The Hume–Wright claim that our reasons for our deepest presuppositions have no
bearing on the truth of such commitments is surprising enough to count as seriously
sceptical. Certainly, it runs counter to our pre-theoretical views. Wright is correct to
see himself as engaged in damage limitation.
4. Two Forms of Scepticism

The sceptic claims that cornerstones are not supported by epistemic reasons. He
concludes that we are not epistemically justified in accepting them. On my reading,
Wright concedes all this, parting company with the sceptic only when the latter
suggests that we have no rational entitlement whatsoever to our cornerstone commit-
ments. But although the points Wright concedes to the sceptic may seem surprising, are
they really damaging?
There is a case for answering “No”. With respect to everyday inquiries, we might
argue, the everyday distinction between justified and unjustified beliefs holds up. That
distinction is made within the framework of our cornerstones. Scepticism reveals the
limits of epistemic justification. In making peace with the sceptic, we must recognize that
everyday epistemic assessment takes place within those limits.
This relaxed response to the sceptic’s view of the epistemic status of cornerstones
implies that epistemic justification is not damaged by recognizing its limits. Naturally,
the sceptic will demur. According to the sceptic, if cornerstones are epistemically
unjustified, so are beliefs that depend on them.14 As Wright says, the upshot of a
Cartesian argument is meant to be “a challenge to our possession of warrant for any of
the large class of dependent beliefs in question”.15 If I have no warrant to suppose that
I am not dreaming, I have no warrant for any belief (based on my current experience)
about the external world. Similarly, the result of a Locality argument is supposedly that
“There is no warrant for any type-II proposition” (non-basic proposition in the
relevant region).16 Oddly, however, if we take Wright’s account of sceptical argumen-
tation au pied de la lettre, these conclusions are not available. Wright defines a corner-
stone for a given region of discourse as a proposition we accept such that, if we lack
warrant for it, we cannot rationally claim warrant for any belief in the region, not that we
do not possess warrant. In Wright’s reconstruction, the sceptic’s argument for the claim
that everyday beliefs are unjustified is fallacious even when restricted to epistemic justification
and independently of the availability of any kind of warrant for cornerstones. Taking Wright’s
definition of cornerstones at face value, the most the sceptic can conclude from our
lacking epistemic justification for cornerstones is that we cannot rationally claim
warrant for cornerstone-dependent beliefs, not that we do not possess it. I would not
14 15 16
Ibid., p. 172. Ibid., p. 168. Ibid., p. 172.
mention this oddity if I thought it was just a slip. But it isn’t. It is the clue to what
Wright takes the sceptic’s most serious threat to be.
Now I do not deny that sceptics sometimes argue in the way Wright suggests. To
show that we do not have warrants for everyday beliefs, they argue that, when pressed,
we cannot produce them. In particular, we cannot give evidence for the reliability of
fundamental information-gathering procedures. But sceptics who argue this way do
not see themselves as arguing fallaciously. This is because they take it for granted that
we possess only such warrants as we can rationally claim. The arguments Wright
attributes to the sceptic presuppose, though they do not defend, an internalist concep-
tion of epistemic justification.
Why does Wright not call attention to the sceptic’s internalist proclivities? The
answer is that he does—eventually, and not quite in such stark terms. For as it turns out,
Wright is not disposed simply to hand the sceptic the move from our lacking epistemic
warrant for cornerstones to our lacking epistemic warrant for dependent beliefs. This
reluctance emerges in his discussion of what he calls the “leaching” problem. Let P be
some ordinary non-basic belief that we are entitled to regard as knowledgeable and C a
cornerstone for P; let warrant include both evidence and entitlement; and let warrant
be closed under known logical consequence. Then it seems that the following claims
are in order:
1. If we run a risk in accepting C, then we run a risk in accepting P.
2. We do run a risk in accepting C.
3. P is known.
According to Wright, this is “near enough” an inconsistent triad. Much of the point
of the concept of knowledge is to single out a state in which belief is safe. So we seem
to be saying that accepting P is both safe and risky.17
Wright’s problem is not happily stated. The issue is not knowledge (understood as
risk-free belief ) specifically but epistemic justification generally. Thus the starting point
for the problem is not that we take some risk in accepting C, but whether C is to any
degree epistemically likely. Insofar as the problem is one of knowledge, it arises because
knowledge has traditionally been taken to involve epistemically justified, true belief;
and entitlement implies none of these. Wright agrees with the sceptic that cornerstones
might be false and we can’t do anything to show that they aren’t. He further agrees that our
lack of evidence for cornerstones degrades their epistemic status: rational trust is not
rational belief. The leaching problem arises because it seems natural to suppose that this
epistemic degradation must seep from cornerstones into beliefs that depend on them.
How can the epistemic status of a dependent belief be higher than that of the corner-
stone on which it depends?
17
Wright 2004a, pp. 208–9.
This may be why Wright states the problem in terms of knowledge. For if knowl-
edge depends on justification, it surely depends on epistemic justification; and corner-
stones are only pragmatically justified. How, then, can cornerstone-dependent beliefs
be epistemically justified? There is only one way. Our possession of positive epistemic
authorization for dependent beliefs must require only that the relevant cornerstones be
true, not that they be warranted. This is the line that, at least tentatively, Wright is
inclined to take. Reacting to the prima facie inconsistent triad, he says that “What
inherits the risk we run in trusting C without evidence is not our belief that P—for we
may in fact have reliable evidence for P—but our belief that we have reliable evidence
for it.”18 This introduces an element of externalism into Wright’s account of our
possession of epistemic warrant for ordinary non-basic beliefs.19
This is all very puzzling. If Wright is willing to adopt externalism, why doesn’t he
accuse the sceptic of a quite different fallacy from that of ignoring the possibility of
entitlements? If Wright thinks that our ordinary conception of justification is extern-
alist, why doesn’t he say that the sceptic simply cannot get from our lacking epistemic
warrant for cornerstones to our lacking warrant for everyday belief, whether or not
cornerstones are warranted in some other way? Indeed, if he thinks that externalism is
the correct account of our understanding of justification for everyday beliefs, why
doesn’t he accuse the sceptic of failing to meet the Intuitiveness Condition, by virtue of
substituting an unmotivated and implausible epistemic internalism for our less demand-
ing, everyday conception of epistemic warrant? Come to that, why does Wright
develop the theory of entitlements at all, except perhaps as a kind of pragmatic add-
on to a fundamentally externalist theory of warrant? Of course, taking refuge in
externalism would raise questions about the damage done by sceptical arguments. If
the sceptic’s demonstration that cornerstones lack epistemic warrant does not threaten
the epistemic status of everyday beliefs—for which epistemic warrant can permissibly
be understood in an externalist way—what does it threaten?
We find a clue in Wright’s reluctance to acknowledge the extent to which his
response to philosophical scepticism depends more on his externalist conception of
epistemic justification than on his theory of pragmatic entitlements. Perhaps “reluc-
tance” is too weak, for in his explicit discussion of externalist approaches to epistemol-
ogy, Wright characterizes his own reflections as “very much internalist” in spirit. In
Wright’s view, “entitlements . . . in contrast with any broadly externalist conception of
18
Ibid., p. 209.
19
I say “an element of externalism” because I do not wish to imply that Wright embraces a purely
externalist (or perhaps purely reliabilist) account of warrant for everyday, dependent beliefs. In particular, I do
not wish to suggest that our possession of non-epistemic warrant for cornerstones is irrelevant to our possession
of epistemic warrant for beliefs depending on them. Even if ordinary beliefs are epistemically authorized by
the truth of cornerstones, that authorization might be rationally blocked by an absence of rational trust.
Maybe there is some epistemic defect in relying on procedures you have no reason to trust, even if they are in
fact trustworthy. I am not taking a stand on this. All I want to claim here is that Wright’s response to
philosophical scepticism (the claim that we have no warrant for various everyday beliefs) essentially involves
an element of externalism, and that this externalist move is independent of his theory of entitlements.
warrant, are essentially recognizable by means of traditionally internalist resources—a

priori reflection and self-knowledge—and are generally independent of the character of
our actual cognitive situation in the wider world”.20
Wright acknowledges that philosophers sympathetic to externalism will be likely to
question the point of such reflections. As he himself puts the objection:
If knowledge, and justification, are essentially environmental—are constituted by (perhaps
reflexively inscrutable) contingencies of our cognitive powers and the way they enable us to
interact with the external world—then no mere sceptical paradox, developed in the armchair,
can show that we have no knowledgeable or justified beliefs. So why bother trying to make out
entitlements?21
His response is that

what is put in doubt by sceptical argument is—of course—not our possession of any knowledge or
justified belief—not if knowledgeability, or justification, are (sic) conceived as constituted in
aspects of the external situation in which we come to have a belief . . . What is put in doubt is
rather our right to claim knowledge and justified belief. It is this which the project of making out
entitlements tries to address and which . . . externalism is impotent to address.22
This passage picks up what previously seemed to be Wright’s misformulation of the

sceptic’s argument; and in so doing, it reveals the damage that the sceptic threatens us
with: damage that externalism cannot limit, much less repair. The sceptic threatens to
foreclose the possibility of our ever attaining an adequate philosophical understanding of
the epistemic warrant that (we like to think) attaches to responsibly held everyday
beliefs. If we are willing to see our ordinary idea of epistemic justification as partly
externalist, we can secure the possibility of epistemic and not merely pragmatic
justification for everyday beliefs. But while we can make this move to head off
philosophical scepticism, we have done nothing to head off scepticism about philosophy.23
That scepticism about philosophy is Wright’s real concern explains his apparent
wavering between internalism and externalism. Wright’s view is that, even though
externalism may have something to tell us about everyday epistemic procedures,
internalism is the only possible approach to philosophical reflection on them. As
philosophers, we would like more than the possibility or even the fact of epistemic
warrant. We would like to show, in some general way, that we have such warrant.
Failing that, we would like at least to put ourselves in a position to rationally claim
warrant. However, if the sceptic is right, we will never be in such a position. So even if
the sceptic fails to deprive us definitively of epistemic justification, he threatens to
preclude our achieving any satisfactory epistemological self-understanding.
Here we find further indications of Wright’s deep affinity with Hume. Hume, too,
thinks that, while scepticism has no seriously destructive consequences for everyday
20
Wright 2004a, pp. 209–10; qualification, n. 27.
21 22 23
Ibid., p. 210. Ibid. Cf. Stroud, 2000.
life, it is fatal to a certain kind of philosophical project: that of legitimating or

rationalizing ordinary beliefs and inferential procedures, as opposed to explaining
their psychological origins. Where Wright differs from Hume is in thinking that,
although scepticism does some damage to our philosophical ambitions, the damage is
not catastrophic.
Wright suggests, plausibly, that there are two kinds of epistemic virtue. One kind has
to do with our intellectual integrity in the matter of forming beliefs or claiming knowl-
edge. The other has to do with our beliefs’ situational provenance. Let us call these aspects
of epistemically virtuous belief-formation responsibility and reliability. Descartes’s project
is to bring these virtues into harmony: to show that if we behave responsibly in forming
and retaining our beliefs, we will hold beliefs that are both certain and true. For
Wright, scepticism stands as a permanent barrier to the achievement of this ambitious
goal. In that way, scepticism does irreparable damage to the quest for philosophical
understanding. However, the damage can be limited. To be sure, Wright cannot prove
that we have knowledge (or epistemically justified belief) of the externalist kind that he
seems willing to countenance: that would require him to show that cornerstones are
true, which he agrees cannot be done. Indeed, he inclines to the view that cornerstones
are “beyond evidence”,24 which suggests that they can be neither confirmed nor
disconfirmed. Thus, it is in philosophy that we feel the effects of acknowledging the
limits of epistemic justification: effects that we need not feel in everyday life. But
Wright does not conclude that, as philosophers, we can only hope for a “fortunate
cognitive situation”.25 We can do more than hope that our cornerstones are true: we
can lay rational claim to them. We therefore have a rational claim to knowledge of the
world, even if having a rational claim does not amount to a proof. Scepticism’s
destructive force with respect to philosophy is significantly reduced, though not wholly
dissipated.
5. How Successful?
Even though, according to Wright, the real damage done by scepticism is to philoso-
phy, rather than to everyday knowledge and justification, his position is still too
concessive. It is also problematic methodologically.
Wright tries to drive a wedge between two questions: whether we possess justified
beliefs in everyday matters and whether we are rationally entitled to claim that we do.
As I put it, he works with an implicit distinction between philosophical scepticism and
scepticism about philosophy. However, I do not think that these two forms of
scepticism are so easily kept apart. This is why it is important not to concede too
much to the sceptic, even with respect to our prospects for a philosophical understanding of
everyday knowledge.
24 25
Wright 2004a, p. 204. Ibid., p. 211.
Wright takes cornerstones to have pragmatic but not epistemic warrant. Not
surprisingly, this general idea has antecedents in the Pragmatist tradition: for example,
in William James. W. K. Clifford ends his essay “The Ethics of Belief ” with the claim
that “It is wrong in all cases to believe on insufficient evidence.”26 In “The Will to
Believe”, James replies that Clifford’s rule is irrationally restrictive. Clifford tells us that
we ought never to hold a belief on insufficient evidence. But consider the following
situation:
— You are confronted with a live hypothesis: one that you take seriously (i.e. regard
as having some chance of being true).
— The choice between accepting and rejecting it is forced: there is no third way.
— The choice matters: serious consequences attend it.
— The question “cannot by its nature be decided on intellectual grounds”.
In such a situation, James argues, keeping the question open is as risky as deciding Yes
or No. It is “attended with the same risk of losing the truth”. Accordingly, in such a
situation we “not only lawfully may, but must, decide”.27 Wright gives us a sophisti-
cated updating of James’s response to Clifford, with this difference: that where James
defends the will to believe, Wright holds that we should not really think of cornerstones
as beliefs. Presumably, Wright shares Clifford’s view that rational belief is closely tied to
epistemic justification.
Many philosophers find James’s response to Clifford unsatisfying. The worry is that
pragmatic justification is too easy to come by. For example, where Clifford was
something of a religious sceptic, on the grounds that there is no evidence that any
theological claims are true, James wanted to argue that religious commitment could be
rational, even if religious dogmas were beyond evidence. But is that really the end of
the matter? If it is, pragmatic justification looks like an exercise in wishful thinking.
Now one might suppose that Wright is not vulnerable to this objection. In defence
of religion, James emphasizes its consoling or morally bracing effects. By contrast,
entitlements are presuppositions of rational inquiry, not just assumptions that make us
happier or more socially useful (though of course they may do that too). But this
doesn’t settle the question. Forget the consoling aspects of religious belief, if such there
are: there is no obvious reason why theological commitments should not be entitlements
of cognitive project. After all, they are presuppositions of theological inquiry, in just the
way that other cornerstones are presuppositions of investigations in physics or history.
If there are theological truths to be discovered, then taking appropriate theological
commitments on board will make possible theological discoveries. It is essential not to
lose sight of the radically non-epistemic character of Wright’s “unearned” warrants. As
I argued earlier, their being epistemically motivated does not make them epistemic reasons,
as most philosophers understand “epistemic reasons”. So even if, in the context of
26 27
Clifford 1999, p. 96. James 1948, p. 89–90.
philosophical inquiry, Wright’s foundations are not exactly for free, we can acquire
them without working all that hard.
Wright’s inspiration, Wittgenstein, is often suspected of encouraging epistemic
relativism. This is no accident. If cornerstones really are beyond evidence, as Wittgen-
stein is often taken to suggest, the prospect of relativism looks all too real. In taking his
cue from Wittgenstein, Wright falls in with this conception of cornerstones. In
Wright’s view, cornerstones are epistemically isolated. For this reason, the problem of
alternative cornerstones has no obvious solution.28
Perhaps Wright could argue that some cornerstones are more fundamental than
others. Some (the uniformity of nature, the existence of the external world) are implicit
in rational inquiry as such, whereas others are only cornerstones for certain special (and
thus possibly dispensable) investigations. This is too large an issue to take up here. I shall
make some brief remarks that bear on it at the end of this essay. For now, let me just say
that I don’t think that the idea that cornerstones (any cornerstones) are open to rational
criticism is easily to combined with the idea that they are epistemically isolated. Wright
is concerned with profane inquiries, which he no doubt regards as fundamental. Others
may think that sacred matters are equally (or more) important. Criticized on profane
grounds, they can plead epistemic isolation on behalf of their cornerstone attachments.
As rooted in our interests, pragmatic warrant is interest-relative, even if the interests in
question are our interests as inquirers.
I suggest, then, that Wright’s meta-sceptical concession has first-order implications.
The problem of alternative cornerstones, with their capacity to warrant all sorts of
controversial first-order commitments, would have to be resolved at the philosophical
level, if it could be resolved at all. To resolve it, we would have to know where we
have, or can rationally claim, epistemic warrant. It is not enough to say that, depending
on the facts of the case, some of our beliefs are warranted and some aren’t. So for
example, if God exists, and some people really do hear His voice in their hearts, then
they acquire religious knowledge by attending to what it says: just as, if there is a
material world more or less adequately represented by experience, we learn all sorts of
facts first hand. But we want to know when warrant is genuine and when it isn’t. To do
this, we would need to reconcile responsibility with reliability. If the sceptic is to be
believed—and Wright thinks that he is—we cannot do this. Externalism—mere
philosophical hope—is no help. But internalistically rationalizing our entitlements is
no help either.
Let me now take up the methodological problem. My question is whether Wright
can avoid externalism, thus mere hope, with respect to his own (supposedly internalist)
philosophical reflections.
Wright’s strategy is to offer a theoretical defence of a kind of pragmatic entitlement.
Ideally, this defence would call on only such classic internalist resources as reflection
28
I do not myself think that Wittgenstein is a relativist. See Williams forthcoming.
and self-knowledge. However, Wright recognizes that an exception might have to be

made for entitlements of substance, for which the defence turns on our conceiving
ourselves as always already in the world. In my view, this is not a trivial matter. Nor is
the worry about staying within the methodological bounds of epistemic internalism
confined to entitlements of substance.
Let us return to Wright’s affinity with Hume. Hume argues that making inductive
inferences and believing in external things belong to human nature. Since we cannot
be blamed for what we cannot help, these commitments are at least non-culpable. But
Hume also argues that, to the extent that we could approximate abandoning ordinary
beliefs and inferential procedures, doing so would be pragmatically unreasonable
because suicidal. At the same time, Hume insists that these psychological and pragmatic
considerations do not touch the sceptic’s theoretical (Hume says “philosophical”) point
that we have no epistemic reasons to believe that there is an external world or that it is
inductively tractable. In that sense, our beliefs and inferences rest on presuppositions
that are epistemically unjustified. Wright differs from Hume principally in taking issue
with the psychologistic character of Hume’s response to scepticism. Where Hume places
most emphasis on the involuntariness of fundamental commitments, thus on their non-
culpability, Wright argues that cornerstones objects of rational trust because positively
(if pragmatically) authorized. Because of this difference of emphasis, Wright is more
hopeful than Hume with respect to the prospects of legitimating ordinary beliefs and
procedures in some very general way. This is the point of his theory of entitlements.
All forms of entitlement are connected with our need to engage in rational deliber-
ation. Whence the need? Wright’s answer is that rational deliberation is bound up with
rational agency and “rational agency is nothing we can opt out of ”.29 Why? One
possibility is that rational agency is an inbuilt feature of human nature: we just aren’t (or
aren’t entirely) creatures of habit and instinct. However, if this is Wright’s view, his
strategy is Humean at one remove, still appealing in the last analysis to a basic
psychological fact. This won’t do. Perhaps, then, Wright’s thought is that we cannot
rationally opt out of rational agency. However, if this is Wright’s idea, he is already
thinking of us as thrown into a world we never made, which we imperfectly control,
and which we must cope with on the basis of limited information. In other words, the
world-picture contained in the cornerstone propositions we are trying to justify is
presupposed all along by their pragmatic warrant. Why isn’t this circularity as bad as
that displayed by attempts to give cornerstones a direct epistemic justification?
This is not such a problem for Hume. In giving a pragmatic defence of inductive
inference and belief in an external world, Hume relies on commitments that he has
found, independently, to be inevitable. (If they were a matter of choice, they would be
weakened by exposure to sceptical argumentation. But they aren’t weakened. So . . . ).
Wright, however, wants to distance himself from Hume’s psychologistic strategy.
29
Wright 2004a, p. 198.
This may not be so easy. Indeed, working within Wright’s self-imposed constraints,
I do not think that it is possible at all.
Wright’s response to the sceptic is far too concessive.
6. What Went Wrong?

To understand why Wright is so concessive to the sceptic, we must go back to the
question we began with: Why is scepticism a problem in the first place? I suggested two
conditions: that sceptical conclusions be unacceptable because damaging, and that
sceptical arguments appear to be compelling because apparently intuitive. I also sug-
gested that Wright’s understanding of scepticism is broadly along these lines. However,
there is a crucial difference between Wright’s views and my own.
Like many philosophers, Wright thinks of scepticism as a kind of paradox. Thus in an
early treatment of scepticism he writes,
The best philosophical paradoxes . . . signal genuine collisions between features of our thinking
that go deep. Their solution has therefore to consist in fundamental change, in taking up
conceptual options which may have been overlooked . . . [T]he traditional sceptical arguments,
in their strongest forms, are such paradoxes.30
Here, Wright concedes that the best sceptical arguments present genuine (and not
merely prima facie) paradoxes. So whereas I think that even the best sceptical arguments
only seem to be intuitive, Wright thinks that they are intuitive, hence that their
conclusions are truly damaging. In consequence, Wright’s response to scepticism is
inevitably revisionary, inviting us to think about justification in a new way. And since
the revisions must be to deep features of our ordinary ways of thinking—Stroud’s
platitudes we would all accept—his response is inevitably concessive.
Wright’s thinking has evolved since he wrote the words just quoted. In particular, as
we have seen, he has reservations about the sceptic’s internalist proclivities, though he
does not exploit these reservations to question the intuitiveness of the sceptic’s
argument. More than that, he still treats the sceptic’s argument as intuitive in regard
to the features that, in my view, really matter.
Let us revisit Wright’s two ways of implementing the Cornerstone Strategy.
Cartesian arguments make play with sceptical scenarios: that I am trapped in a
persistent, coherent dream, or that I am a brain in a vat. According to the sceptic,
though it is a cornerstone for the warrant of ordinary empirical beliefs that such
sceptical possibilities do not obtain, I cannot rule them out. But why do I need to?
They are, after all, bare logical possibilities. It is certainly not part of our ordinary
conception of epistemic justification that all logically possible alternatives to a given
empirical belief must be ruled out, on pain of that belief’s being unwarranted. So at a
30
Wright 1985, pp. 429–30.
minimum, the sceptic needs to explain why we should take his scenarios seriously.
Wright offers no suggestions as to what such an explanation would look like.
We can go further: sceptical scenarios can be ruled out: by background knowledge.
While it may be a bare logical possibility that I am a brain in a vat, it is not a serious
possibility, for the simple reason that no such technology yet exists. As for dreaming,
I know quite a bit about dreams in general and my own dreams in particular. For
example, my dream-imagery is predominantly visual, though occasionally it involves
auditory and kinaesthetic elements. But one thing it never involves is serious pain. So
for me, the “pinch” test works: if it really hurts, I am definitely not dreaming. Wright
says that there can be no test for whether I am dreaming, since the warrant for
accepting the results of such a test can be no stronger than my antecedent warrant
for believing that the test was competently executed; and the strength of this warrant
will be zero unless I am already warranted in believing that I did not dream that
I executed the test. However, in my case, the pinch only yields a positive result when
competently executed. If I don’t really pinch myself, it doesn’t hurt. If it hurts, I know
I carried out the test and so I know I am not dreaming.
I am not suggesting for a moment that these simple-minded remarks dispose of
Cartesian scepticism. The problem with them is obvious: my ways of “ruling out”
sceptical possibilities depend on my already knowing a good deal about the world
around me. Indeed, in the case of the pinch test, I am not only presupposing that there
is an external world, I am taking for granted quite a bit about the kind of experiences
that do and do not reveal it to me. In other words, from the sceptic’s point of view, my
warrants for discounting his troubling scenarios presuppose the kind of knowledge (or
justified belief ) that he means to call in question. Fair enough. But what this shows is
that the function of sceptical scenarios is to question whether we have any reason to
believe that experience, considered in isolation, testifies to anything concerning the
external world, even that there is one. The same goes for putative traces to the
existence of the past, and so on for other sceptical problems. Of course, experience
will tell us all kinds of things about the world, assuming that we are not trapped in a
sceptical situation. The problem, however, is to show that we are not so trapped,
without relying on antecedent knowledge of the world: i.e. on the supposition that
experience is a generally reliable guide to things around us. We can conclude that
Wright’s “Cartesian” scepticism is either easily dismissed or is Locality scepticism in
another rhetorical guise. Accordingly, Locality scepticism is what we need to focus on.
The defining feature of Locality scepticism is, obviously, its reliance on cognitive
locality. We might think, then, that this is the idea we need to probe. However, far
from questioning cognitive locality, Wright defends it. This defence is essential to his
position. Without the informational restrictions imposed by cognitive locality, we
wouldn’t need cornerstones, as Wright understands them. At the same time, the same
restrictions ensure that cornerstones are beyond epistemic justification, and perhaps
beyond evidence altogether. But although commitment to cognitive locality is the
driver of Wright’s concessions to scepticism, is cognitive locality really an evident

commitment of our ordinary conception of epistemic justification?
Wright thinks that cognitive locality is “the circumstance that only a proper subset of
the kinds of states of affairs which we are capable of conceptualizing is directly available
at any given stage in our lives, to our awareness”.31 Since what we see, hear, etc. for
ourselves is (very) limited, much of what we believe depends on inference; and
wherever there is a distinction between non-inferential and inferential belief, we can
locate the justificational architecture that Locality arguments exploit. This argument
moves far too quickly. The problem is clearest with respect to the Locality argument
for external world scepticism. However, although I won’t develop the argument here,
the problem I am going to point to applies to Locality arguments generally.
Wright thinks that an acknowledgement of cognitive locality is implicit in our
conception of ourselves as always being at a particular place at a particular time: what
we can see for ourselves is positionally constrained. This is true: I can’t be in two places at
once. But this is only one out of many constraints. When I go to a football match, what
I can see of it depends on where I’m looking at any given moment, how good a view
I have of the pitch, how sharp my eyes are, and how well I understand the game. Of
course, limitations like these can be remedied. If I can’t see well, I can get a better seat
or new glasses. If I don’t know what to look for, I can learn. Furthermore, such
contingent limitations on “seeing for oneself” do not set any principled limits to the
kind of facts that one can personally determine. Trained observers can just see things
that other people have to work out by inference or have explained to them. The
informational restriction needed to motivate a Locality problem with respect to
knowledge of the external world off the ground is different in both respects. My
alleged dependence on “experience” for evidence about how things are in the world is
permanent and irremediable: I can’t move to a better seat. Also, it is a restriction on the
kind of fact capable of being directly known. Roughly, I can know “directly”—i.e.
non-inferentially—how things appear to me. To get any further, I will need to rely on
cornerstones, even though I can have no epistemic reason to accept them. This is not
cognitive locality: it is cognitive subjectivity. Indeed, it is cognitive imprisonment within
our subjectivity.
Wright’s conception of the “justificational architecture” exploited by Locality
scepticism embodies foundationalism of a familiar kind. The critical feature of this
architecture is that it is substantive rather than merely formal. It is not the view that, at
any given time, there are lots things that are, in various ways, reasonably treated as
default credible, while others derive their credibility from the evidence that our default
entitlements provide. This formal “foundationalism” is broadly correct: there are
indeed, in a sense, unearned entitlements, though this status is by no means the
exclusive property of Wright’s cornerstones. Wright’s view, however, is that we can
31
Wright 2004a, p. 173.
specify, in terms of the kind of fact they concern, the kinds of beliefs that need support
(by evidence or by reasons of some other variety) and the kinds of beliefs that are fit to
provide it. So whereas I think that the inferential/non-inferential distinction is funda-
mentally methodological, Wright thinks that it is ontological. That is what I mean by
saying that his foundationalism is substantive, not merely formal.
Substantive foundationalism is indefensible. There is no special kind of belief that is
intrinsically non-inferential. What needs evidence and what doesn’t is always and every-
where a matter of context. The key point was made long ago by Austin:
[T]he idea that statements about ‘material things’ as such need to be verified is just as wrong as,
and wrong in just the same way as, the idea that statements about ‘material things’ as such must be
based on evidence. And both ideas go astray, at bottom, through the pervasive error of neglecting
the circumstances in which things are said—of supposing that the words alone can be discussed, in a
quite general way.32
In questioning Wright’s conception of cognitive locality, then, I am not questioning it

in a local application. I am not saying that the foundations of knowledge need to be set
higher up: say, at the level of beliefs about familiar objects in our immediate surround-
ings. My reservations concern the very idea of a substantive justificational architecture.
I identified two features in virtue of which a sceptical argument presents a problem:
it must appear to be an intuitive argument for an unacceptable conclusion. Wright attacks
scepticism at the wrong point, conceding the intuitiveness of the sceptic’s argument
while trying to limit the damage. But the sceptic’s argument isn’t intuitive in the first
place. Rather, it is an artefact of a contentious theorization of our ordinary practices of
epistemic assessment. What scepticism threatens, therefore, is not the coherence or
intellectual integrity of ordinary epistemic claims: it threatens the sceptic’s—and a
certain traditional epistemologist’s—implicit theory of knowledge.
It may seem that, in drawing this conclusion, I am agreeing with Wright. As we have
seen, Wright’s own final position is not straight philosophical scepticism but a miti-
gated scepticism about philosophy. However, while my views and Wright’s overlap to
a limited extent, in the end they are very different.
If I am correct in thinking that cognitive locality, as Wright understands it, is by no
means an obvious feature of our epistemic situation—if indeed we think of ourselves as
having an epistemic situation—why does Wright find the idea so compelling? The
answer to this question lies in Wright’s conception of philosophy. Wright thinks that
the resources for philosophical understanding are internalist—reflection and self-
knowledge. That is to say, a subjective starting point is the starting point, not for
everyday justification, but for doing philosophy.
There are two reasons why a philosopher might take this view. The first has to do
with the peculiar generality of the understanding that epistemologists of a certain stamp
are tempted to seek. They want to understand some special kind of knowledge—say,
32
Austin 1962, p. 118.
knowledge of the external world—as such. As I sometimes say, they impose a totality
condition on a properly philosophical understanding of our knowledge of the world.
Wright’s commitment to this conception of epistemology shows up in his connecting
cornerstones with very broad “regions of discourse”. Which of course turn out to be
the familiar ones: the external world, the past, and so on. But epistemologists of this
stamp also want a special kind of explanation. Broadly speaking, they want a legitimating
explanation: an explanation that each of us could give to himself and which would
bring to light the epistemic credentials of the beliefs in question. Wright accepts this
too, with the proviso that he hopes for no more than a rational claim to knowledge of
the world, rather than a proof that we have it. Putting these points together, the
philosophers I am describing want us to be able to explain to ourselves how it is that we
know, or justifiably believe, anything whatsoever about the external world, and to do so
in a way that vindicates our self-conception as knowers. But it readily appears that such
an explanation cannot take any knowledge of the world for granted without lapsing
into circularity: in effect, this is what Wright takes sceptical arguments to show. That
means that a satisfactory explanation would have to consist in a legitimation of
objective knowledge on the basis of subjective resources. Wright thinks, correctly,
that no such explanation is to be had, as long as we insist on equating rational
entitlement with epistemic warrant. The sceptic is right to insist that we will never
prove or even make probable the claim that responsibility guarantees reliability. The
best we can do is to show that we can, in good conscience, take them to go together.
I conclude that Wright’s subjective turn has nothing to do with commonsense
conceptions of locality. There is no fact of cognitive locality, as Wright needs to
understand locality. Rather, cognitive locality is built into Wright’s conception of
the task of philosophy. Cognitive locality, as subjective confinement, is at best an
entitlement of cognitive project. Wright thinks that the project is a good one, so the
entitlement holds. I think the project is misguided. This is the fundamental issue on
which we part company.
Wright thinks that scepticism about philosophy is damaging because it precludes a
kind of understanding that it is at least coherent to hope for. But it is sensible to be
disappointed at a failure of understanding only if there is something to understand, and
I don’t think that there is. The fatal mis-step is made in the very acceptance of such
traditional “regions of discourse” as “the external world” or “the past”. Why think of
such classifications as picking out genuine kinds, as opposed to loose aggregates? No
one tries to understand anything whatsoever that happens in Baltimore, or anything
whatsoever that happens on a Thursday. Deploying a universal quantifier is not
sufficient to mark out an area of inquiry. We find kinds where we can locate some
kind of theoretical integrity. The question to ask, then, is what binds together beliefs
about the external world? This classification brings together an extraordinary congeries
of subjects—history, physics, biology—not to mention a myriad of unconnected,
mundane factual commitments. The classification possesses no evident topical unity.
Quite so: the unity is epistemological. From the outset, “beliefs about the external
world” is meant to contrast with “the evidence of experience”: the distinction is the
inferential/non-inferential distinction, interpreted ontologically rather than methodo-
logically. If we drop the idea of cognitive locality, the conclusion to draw is not that we
will never be able to understand knowledge of the external world, in an appropriately
“philosophical” way. The correct conclusion is that there is nothing to understand. The
category is bogus.33
These thoughts suggest a final word on cornerstones. Wright thinks of cornerstones
as beyond evidence. I take it that he means to put them beyond confirmation and
disconfirmation. Like the poor, they are with us always. However, deep epistemic
presuppositions are candidates for such extreme epistemic isolation only if they are very
thin and abstract. Serious cornerstones—the deep presuppositions of real-life investiga-
tions—are much richer. Because of this, they are often revised on the basis of results
that they themselves throw up. Moreover, cornerstones of one kind of inquiry can be
modified in the light of results obtained by another: as carbon dating or spectral analysis
can lead us to revise our ideas about historical artefacts and documents. The presuppo-
sitions of real-world inquiries are eminently revisable. Maybe we will always think that
there is a world to investigate. But whether that world must always be thought of as
“material” is doubtful.34 (I don’t think that the world is even “external” in the relevant
sense.) But once we agree that cornerstones can be criticized, we must be open to the
possibility that entitlements of cognitive project can lapse, either under the weight of
the internal problems they generate or under pressure from criticism from without.
The epistemological project, as Wright conceives it, is rickety on both counts.
Where a project responds to a felt need, criticism is more effective when offered in
conjunction with an alternative vision. So if we follow the line of thought I have been
adumbrating, what picture of knowledge and justification are we left with? Wright
himself provides the answer to this question. He writes:
[W]e should view each and every cognitive project as irreducibly involving elements of
adventure—I have, as it were, to take a risk on the reliability of my senses, the conduciveness
of the circumstances, etc.35
However, cognitive adventuring is not a matter of taking a flyer, in the bare hope that
things are as they should be. As Wright explains, “warrant is acquired whenever
investigation is conducted in a fully responsible manner”. Full responsibility does not
require the investigation of every possible way things might have gone wrong: that is
33
I develop this argument at length in Williams 1996. See especially Chapter 3. I develop the argument in
a somewhat different way, starting from “regress” rather than I want to make it clear, then, that in questioning
Wright’s conception of cognitive locality, I am not questioning it in a local application. I am not saying, as
some say, that the foundations of knowledge need to be set higher up: say, at the level of beliefs about familiar
objects in our immediate surroundings. My reservations concern the very idea of substantive foundationalism
“Cartesian” problems, in Williams 2002.
34
Wittgenstein thinks that bedrock presuppositions get modified in the course of inquiry. See the
metaphor of the river bed }} 91–9 of Wittgenstein 1969.
35
Wright 2004a, p. 190.
impossible, as scepticism shows. Rather, full epistemic responsibility requires only that
we have looked at “any of the presuppositions about which there is some specific
antecedent reason to entertain a misgiving”.36 In other words: we deal with problems
as they arise; otherwise, if it ain’t broke, don’t fix it. Of course, in investigating the
warrant for a belief, when that warrant has become problematic, we inevitably create a
context with unexamined presuppositions of its own: we are always adventuring. But
this does not matter: if the results of this subsequent investigation come into question,
we can investigate their presuppositions. This is not a vicious regress of justification, just
the open-endedness of inquiry.
In his brief remarks on this alternative picture, Wright sketches what is now called a
“default and challenge” conception of epistemic justification.37 On this conception,
we can legitimately presuppose reliability by virtue of (a) conducting ourselves in a
responsible manner and (b) being committed to vindicating this presupposition, should
there arise some reason to question it. Properly worked out, the default and challenge
conception shows how responsibility and reliability can be harmonized, which for
Wright is epistemology’s highest ambition.
Wright does not agree, but only because he mishandles the idea of “limits to
justification”. On the default and challenge conception of warrant, what we can
look into is always in part a function of what we are leaving alone. Investigating the
presuppositions of a given inquiry typically launches a new inquiry, with presupposi-
tions of its own. But when Wright says that scepticism shows us that inquiry must
always take place “within such limits”, he has something else in mind. He is thinking of
absolute, context-invariant limits. That is, he is thinking of cornerstones as embedded in
a permanent, context-invariant and ontologically substantive justificational architec-
ture. But the whole point of the default and challenge conception is to provide a radical
alternative to that picture of our epistemic situation. Unwilling to recognize this,
Wright concedes far more to the sceptic than anyone should. There is a lot to the
idea of unearned warrant, but only when it is detached from the traditional assumptions
about knowledge and justification that Wright continues to make.
Bibliography
Austin, J. L. 1962 Sense and Sensibilia, Oxford, Oxford University Press.
Brandom, R. 1994 Making It Explicit: Reasoning, Representing, & Discursive Commitment, Cam-
bridge, Mass., Harvard University Press.
Clifford, W. K. 1999 “The Ethics of Belief ”, in The Ethics of Belief and Other Essays, Amherst,
NY, Prometheus Books, pp. 70–96.
36
Ibid., p. 191.
37
The name is due to Robert Brandom. See Brandom 1994, Chapter 4. However, I think that the idea is
present in Austin.
James, W. 1948 “The Will to Believe”, in Essays in Pragmatism (edited by A. Castell), New York,
Hafner, pp. 88–109.
Pryor, J. 2004 “What’s Wrong with Moore’s Argument?”, Philosophical Issues 14, pp. 349–78.
Stroud, B. 1984 The Significance of Philosophical Scepticism, Oxford, Oxford University Press.
Stroud, B. 2000 “Understanding Human Knowledge in General”, in Understanding Human
Knowledge, Oxford, Oxford University Press, pp. 99–121.
Williams, M. Unnatural Doubts 1992 Oxford, Blackwell; paperback edition, Princeton, NJ,
Princeton University Press 1996.
Williams, M. 2002 Problems of Knowledge, Oxford, Oxford University Press.
Williams, M. forthcoming “Knowledge without Epistemology: Why (Wittgensteinian) Con-
textualism is Not Relativism”.
Wittgenstein, L. 1969 On Certainty, Oxford, Basil Blackwell.
Wright, C 1985 “Facts and Certainty”, Proceedings of the British Academy 71, pp. 429–72.
Wright, C. 2004a “Warrant for Nothing (and Foundations for Free)?”, Proceedings of the Aristote-
lian Society Supp. Vol. LXXVIII, pp. 167–212.
Wright, C. 2004b “Wittgensteinian Certainties”, in D. McManus (ed.) Wittgenstein and Scepti-
cism, London, Routledge, pp. 22–55.
REPLIES
Crispin Wright
Foreword
More than a quarter of a century ago in a collection I edited to celebrate the centenary of
the first publication of Frege’s Grundlagen der Arithmetik, I wrote that “Philosophical
achievements can be measured in different ways. One important measure is the quality
of the philosophy which a philosopher inspires.”1 I must ask the reader to trust that I was
not in penning that sentence foresightedly preparing for the immense compliment that
this anthology of new essays centred on my work pays me by that criterion. I am
immensely grateful to all the authors for contributing so interesting and challenging a
set of papers, and to Annalisa Coliva for patiently bringing the volume together. Special
thanks go to Paul Boghossian for critical feedback on all four chapters of my Replies.
It is both a relief and a regret that it would have been quite impractical, within the
time and space at my disposal, to attempt to respond to the essays in the detail that the
quality of their discussions and criticisms of my ideas deserve. I do not know, indeed,
how well I could have managed if my time and space were limitless, for the questions
on which the contributors have tended to focus are some of the most obdurate in
contemporary philosophy, and many of the criticisms go right to the heart of my
proposals on these issues. What follows, accordingly, is more of a somewhat eclectic set
of reflections on landscapes and ‘matters arising’ than the kind of attempt, more usual in
festschriften, to respond seriatim and even-handedly to one’s commentators. The effect
may be, perhaps, to make some of my reflections more self-contained, and perhaps
more easily read on that account. But I am very aware of the many interesting and
pertinent lines of thought to which I have attempted no response. May there be other
opportunities!
1
C. Wright 1984 (ed.) Frege: Tradition and Influence, Oxford, Blackwell.
Replies Part I: The Rule-Following
Considerations and the
Boghossian, Peacocke, Horwich, and Bilgrami
Following a Rule: Two Problems

I first read Wittgenstein’s Remarks on the Foundations of Mathematics (RFM) in the summer
of 1965. It took about a week of afternoons in the Fellows’ Garden of Trinity College,
Cambridge, during a spell of particularly beautiful weather. The effect was to turn my
thinking about mathematics upside down. The initial focus of my interest had been on
the question—crucial to an understanding of Mathematical Intuitionism, on which
I was contemplating a thesis—of how one might coherently conceive of each instance
of a general arithmetical proposition, say Goldbach’s Conjecture, as determinate in
truth-value yet question the validity of the Principle of Bivalence in relation to the
general proposition itself. It had never occurred to me to ask what were the presupposi-
tions, or hostages, of the idea that each such instance is determinate in truth-value,
notwithstanding the human impossibility of an actual verification in all but a paltry
initial segment of the countably infinite range of cases involved. But it became clear that
the question had, in effect, very vividly occurred to Wittgenstein. I had not at that time
looked at the Philosophical Investigations at all, and Wittgenstein’s editors had (most
unhelpfully) omitted from the version of RFM then available the sustained discussion
of rule-following that, in the third edition, became section VI. But the general tendency
of his thinking was nevertheless unmistakable. To regard each instance of Goldbach’s
Conjecture as determinate in truth-value, irrespective of any ability of ours actually to
determine its truth-value—or even, for all but a small initial segment of cases, to come
within range of understanding a formulation of the statement concerned—seemed to
call for a kind of objective reach of meaning that would inevitably relegate our cognitive
relationship to it, in understanding, to something not unjustly assimilated to Plato’s
conception of the strange yet convenient (putative) sensitivity of our minds to the
requirements of the Forms. But if, to the contrary, we were to try to regard meaning as
wholly grounded in socio-linguistic practices, rather than as somehow driving them
from above, how might we make sense of this reach of meaning, to which the
380 CRISPIN WRIGHT
conception of an arbitrary in principle decidable arithmetical statement as already

determinate in truth-value was seemingly committed?
The issue is hardly as plain as day, but it struck me then, as now, as raising some of the
most profound questions to which analytical philosophers are challenged to respond,
fully deserving of the preoccupation—indeed, obsession—that it became for many
philosophers in the last three decades of the twentieth century. It is rare, perhaps
unheard of, for an important philosophical concern to spring, originally minted, from
the mind of a single thinker, but I share Saul Kripke’s opinion that in his musings on
the issues in this vicinity, Wittgenstein in effect invented a new philosophical problem,
or cluster of problems, for which—like all the deepest philosophical problems—it is
extremely difficult both to capture exactly the right formulation and to fix on a settled
response.
In my own early work, including Wittgenstein on the Foundations of Mathematics, the
problem shaped itself as concerning the idea of what I termed investigation-independence:
the idea that the meaning of a statement can, in collaboration with the facts, take care
of its truth-value in a way that is constitutively independent of any propensity of
judgment of ours. If one combines that, at first sight platitudinous and anodyne
thought with the second apparent truism that we give our statements meaning not
in the sense of annexing them to Platonic contents that already sustain investigation-
independence but by constituting their meanings in our communicative practices, then
we immediately confront the difficulty, prominent also in Kripke’s discussion,1 of
explaining how it is that anything we do, or can do, amounts to the constitution of unique
meanings, autonomously so conceived. This concern is certainly an important element in
Wittgenstein’s own discussion, but it should be noted that it is first and foremost a concern
about linguistic content—though it spreads to psychological content and the relation
between, for instance, the formation of an intention and what is subsequently adjudged to
fulfil it. It is happily captioned as a problem about rule-following only insofar as the contents
of rules, too, like all contents, come within its shadow—or because one is thinking of the
meaning of an expression as akin to a rule for its correct use.
There is, however, a second aspect of the ‘rule-following considerations’ that only
later, and gradually, began to distinguish itself for me. And this does indeed have to do
with rule-following per se. The two aspects are contrasted in my paper on Chomsky2 by
the questions:
How can I tell which rule I (used to) follow? (Or: how is the rule that I used to follow to be
identified?)
versus:
How can I tell what the rule I grasp requires of me here?
1 2
Kripke 1982. Wright 1989, p. 243.
REPLIES 381
In that paper I suggested that Wittgenstein’s thought about a regress of interpretations

(Philosophical Investigations }}198–201) belongs with a difficulty attending the second
question, rather than—as on Kripke’s account—with the first. This difficulty, put in
the most general terms is the following: if the transition from a rule ‘in mind’—let it be,
for the sake of argument, a fully determinate, if you like Platonic rule—to its (correct)
application in a given particular context is a mediated transition: one bridged by
supplementary cognitive activity, then that activity better not itself involve further
rule-following, on pain of making all rule-following involve further rule-following,
and hence regressive and impossible.
The distinctness of this issue is somewhat obscured in Wittgenstein’s own exposition
in those particular paragraphs, where the mediating activity is represented as one of
interpretation, so that it is the question of the determinacy of the content of the rule that
still seems to be in play. Nevertheless the second aspect is different; and crucially so. It
will be my main focus in the following remarks.
Throughout the mid-to-late 1980s Kripke’s book generated an avalanche of com-
mentary, mostly expressly concerned only with a character called ‘KW’, and mostly
ignoring most of the other commentary. Everyone seemed to want to define a personal
escape route from the ‘Skeptical Paradox’. After my paper on Chomsky, I felt I’d had
my fill of the debate and took a sabbatical from the topic for a double septennium. It
was a series of conversations and exchanges with Paul Boghossian3 on Lewis Carroll
and the epistemology of basic inference that stimulated me to think afresh about the
issues and I returned to them in a paper I presented at the NYU Mind and Language
seminar in 2002.4 The principal new development in my thinking in that paper was a
much clearer focus on the second question. The last part of the paper canvassed what is
there termed the Modus Ponens model of rule-following, according to which a rule is
conceived as a general conditional prescribing a mandate—that is, a requirement,
permission or prohibition—for an action A:
If P, then !A
on the condition P, and the following of the rule as essentially an inferential transition
from a state of cognition of the satisfaction, in the particular circumstances, of its
antecedent P to a detached recognition of (or action upon) the mandate concerned.
Evidently, the operation of such a rule is going to require an anterior capacity of
judgment whether or not P. But if such is the basic form that anything properly
regarded as rule-following has to take—the suggestion of the paper was that it is hard to
see any alternative—and if we propose to think of all competent linguistic practice
as consisting in rule-following, the question looms how to conceive of the rules
3
See e.g. Boghossian 2000, 2001. Relevant papers of mine are Wright 2001, and Wright 2004.
4
“What is Wittgenstein’s Point in the Rule-following Discussion: Five Themes”, referred to by both
Boghossian and Paul Horwich in their contributions to this volume. A descendant of this paper was
subsequently published as Wright 2007.
382 CRISPIN WRIGHT
governing the most basic and primitive expressions of the language, or indeed any
expressions an accurate statement of whose satisfaction-conditions would have to take
a homophonic form. Suppose, for example we conceive of our respective uses of ‘pain’
as governed by, if by any rules, then, inter alia, by the rule:
If in pain, then assent to “I am in pain”.
Then the cost of thinking of competent operation with such expressions as a form of
rule-following is going to be that grasp of the concepts they express must be viewed as
cognitively anterior to mastery of the relevant expression—exactly the conception of
linguistic competence, resting on a prior capacity for fully articulated thought, which is
expressed in the passage from St Augustine with which the Philosophical Investigations
begins and from which, as I put it, that text is a “journey of recoil”.
The conclusion is a dilemma: either linguistic competence is not, au fond, a matter of
rule-following, or some quite other model of rule-following is demanded. Christopher
Peacocke’s and Paul Horwich’s chapters in this volume may be seen as attempting to
outline (very different) responses that embrace the second horn. I’ll come to them
shortly. But the problem is not confined to understanding the sense, if any, in which
basic linguistic competence can be rule-informed. Indeed, as Paul Boghossian’s essay
forcefully brings out, it seems to embrace not merely basic linguistic competence but
rational thought itself.5
The Regresses
To bring the issues into focus, let us step back and try to say at an intuitive level what it
is to follow any rule. Suppose you are playing chess and wondering whether now is a
good time to castle. Is it permissible to do so? The rule for castling says, roughly:
If neither the king nor one of its rooks has so far moved, and if the squares between them are
unoccupied, and if neither the king nor any of those squares is in check to an opposing piece, and
if the king and the rook in question are on the same rank,6 then one may castle, that is, move the
king two spaces towards the rook and . . .
Your following this rule, presumably, will consist in your recognizing what it has to say
about your present situation in the game and then acting in the light of that informa-
tion. So, for example, you reflect:
(Minor premise) In this game so far, neither my king nor queen’s rook have yet been moved,
the squares between them are unoccupied, . . . etc.
And you draw the conclusion:

I may castle now.
5
See also Boghossian 2008.
6
Exercise for reader: do you see why it is necessary to include this condition?
REPLIES 383
Several points are salient. First, as in the Modus Ponens model, the rule is given as a
general conditional content, whose antecedent portrays the kind of circumstances (the
triggering circumstances) in which the permission articulated in its consequent is acti-
vated. Second, the ability to follow the rule involves knowing what this conditional
content is, and attempting to uphold it—to avoid making moves that “break” it, as we
are wont to say—in the course of the game. Of course you may succeed in this attempt
without in any way overtly distinguishing yourself from someone who is ignorant of
the rule for castling but can otherwise play chess, namely by never availing yourself of
the permission that the rule grants. So—a familiar but vital point—we may not equate
following a rule with acting in the manner that it requires: conforming to a rule is one
thing, following it something different.7 Third—or so it is natural to think, though this
will prove to be a crucial and controversial point—following the rule essentially
involves (very elementary, perhaps unreflective) reasoning. The rule supplies the
major premise for what is, in effect, a step of modus ponens, the minor premise
being supplied by your observation of the board configuration and memory of the
game so far. Fourth, the output of this minor feat of information-processing is a
judgment—“I may castle now”—on which you may or may not act. That you are
following the castling rule explains your making this judgment by giving your reasons
for it; and if you act on the judgment, that too will be rationally explained by your
following the rule in question. Finally, the rule is normative in the sense that both the
output judgment, and any action upon it, are assessable as correct or not in the light of
the rule: you may be mistaken in thinking that the trigger conditions are met, or you
may correctly judge that you may now castle but bodge the actual move (maybe you
inadvertently move the king only one space), but in either case, the propriety of
describing your move as incorrect depends on bringing it within the purview of the
rule—that is, on viewing it as an attempt to follow the rule.
Here are the key points again. Rules are general conditionals, connecting a trigger
condition with a mandate. Following a rule requires somehow accepting its content and
subjecting one’s practice in relevant respects to its requirements. What one then does is
assessable as correct or not according as it squares with the output of the rule for the
circumstances in which it is done. This output is a matter of entailment, and following
the rule accordingly involves somehow tracking this entailment—so presumably infer-
ence. Rule-following is rational activity: that one is following a particular rule con-
tributes to the reasons one has for doing what one does, and in part rationally explains
one’s performance.
I said “so presumably inference”. But that is going to be a key issue. Provided we
stall on any firm commitment to the role of inference, this model—the italics in the
previous paragraph lend themselves to the acronym, “the ACRE model”—is effec-
tively that proposed by Boghossian in his present chapter. And it surely hits off one very
7
Note that I do not say, “something more”. You can be following a rule yet fail to conform to it, as a
result of a mistake, inattention, and so on.
384 CRISPIN WRIGHT
intuitive notion of rule-following. But how extensively can it be coherently applied?

All believing, we tend to think, is assessable as rational or not. More, it’s intuitive8 to
suppose that all believing is potentially rational or not at the point of ingress, so to
speak: that a belief ’s being rationally managed is one question and its being rationally
entered into is another, and that each of these dimensions of constraint engages any
particular belief of any particular thinker. What Boghossian calls the “Rulish” picture
of belief essentially has it that both dimensions of rationality are matters of successful
rule-following. Entry-rationality, in particular, will consist in the following of correct
epistemic rules in the formation of the belief. Can this idea be sustained under the aegis
of the ACRE model of what rule-following consists in?
Well, we have noted that there will be a bind on any account whereby the transition
from any rule to its (correct) application in a given particular context is a mediated
transition: one bridged by supplementary cognitive activity involving further rule-
following. Wittgenstein’s “regress of interpretations” is one illustration of this pattern
and now we have set ourselves up for another, as Boghossian notes.9 For the ACRE
model, harmless and truistic though it seems at first blush, does seem to require that rule-
following essentially involves inference. How are we to understand the model except like
this: that a rule connects with the practices it rationalizes and explains by providing a
general instruction—hence a premise—for conclusions about the mandates it issues in
particular sets of circumstances, so that following it involves drawing and recognizing such
conclusions. But drawing conclusions is inference, and inference, surely, is itself a kind of
rule-following—specifically, the following of rules of inference! In following the rule for
castling at a particular situation in the game of chess, I in effect reason from the rule as
premise, together with a description in relevant respects of the circumstances of the game,
to a conclusion about whether I may castle at that particular juncture. But in drawing this
conclusion, I also follow the rule of modus ponens—and according to the ACRE model
interpreted as incorporating inference, this secondary act of rule-following will itself
involve reasoning from a set of premises, one of them a general statement of the rule of
modus ponens, to a conclusion—presumably, that the drawing of the original conclusion
is admissible—distinct from the original conclusion about the admissibility of castling.
And now this inference in turn, once assimilated to an act of rule-following and subjected
to the ACRE model so understood, calls for a further inference—still with a statement of
the rule for modus ponens as its major premise, but with a differing minor premise and a
further novel conclusion, viz. that the drawing of the conclusion of the second inference is
admissible. And so on. The regress is vicious, since the inferences concerned all involve the
drawing of different conclusions. In brief: if, as may seem wholly natural, even unavoid-
able, it is built into the ACRE model that following a rule always involves inference, and if
inference is always a form of following a rule, then it follows that any act of rule-following
requires another distinct from it, and not relevantly easier than it. Moreover every act of
8
Though of course denied by some distinguished philosophers, including Quine and Davidson.
9
This volume, pp. 40–1; see also Boghossian 2008, p. 492 and following.
REPLIES 385
rule-following in the series thus generated is distinct from all those that precede it in the
series. That makes following a rule an example of a “supertask”. We cannot perform such
tasks.
Call that the Inference problem. It’s bad, but it is not the full extent of the woes. As we
in effect already noted in connection with the Modus Ponens model and St Augustine,
there is also a Minor Premise problem. The ACRE model, inferentially interpreted,
involves that an epistemic rule connects with a judgment it rationalizes and explains
via a minor premise, that its trigger condition obtains. However the model won’t
rationalize the output belief unless this premise too is accepted in whatever sense the
rule itself is accepted. If my judgment that it is legitimate to castle now is to be
rationalized and explained by my following the rule for castling in terms of the
ACRE model, then I need to be credited with a judgment about the trigger condi-
tions—the board configuration and the history of the game—as well as an acceptance
of the rule. But any old judgment? The Rulish picture has it that all rational belief is
belief formed by following basic epistemic rules. Yet however epistemically virtuous
the rules concerned, the quality of the beliefs that they deliver must obviously be
hostage to the quality of one’s judgments of the truth of the minor premises to which
they are applied. It is only if the minor premises are themselves rationally accepted that
the application of the rules will guarantee the rationality of the beliefs thereby formed.
But if acceptance of the minor premise is to be rational, and if all rational judgment is
conceived as entered into in the mode depicted by the ACRE model, specialized to
epistemic norms, then another regress is launched: to arrive at any rational judgment,
I must first make another, independent rational judgment.
So this nexus of problems is the second aspect of the ‘rule-following considerations’
that I referred to—a second fundamental form of paradox, additional to and indepen-
dent of the problem of scepticism concerning the determinacy of rules emphasized in
Kripke’s exegesis and my own first book. The core problem is that three very natural
assumptions about what rule-following consists in, and about inference, viz.
that the essence of rule-following is captured by the ACRE model;
that the ACRE model requires that the application of rules everywhere involves
inference; and
that inference is itself a form of rule-following;
are jointly inconsistent with the very possibility of finite beings’ following a rule—any
rule. In consequence, if rational belief-formation is a matter of the following of
(correct, or rationally acceptable) epistemic rules, then rational belief-formation is
also impossible. And if basic linguistic competence is a matter of the following of
rules that constitute the proper use of the language concerned, then there is no such
thing as basic linguistic competence.
A less hysterical conclusion is this: if each of those three theses is faithful to our actual
concept of rule-following, then it will indeed be impossible to follow a rule, of any kind,
so long as doing so is thus conceived. The challenge, then, would be to explain what to
386 CRISPIN WRIGHT
give up, how to refashion our concept in such a way as to conserve at least something
of the central role we ascribe to it in linguistic competence, intentional action, rational
belief-formation and management, and the subjection of multifarious of our practices
to standards of correctness.
At the conclusion of his present chapter, Boghossian tentatively suggests that, natural
though it undoubtedly is, the idea to expunge may be that applying a rule has to be a
matter of inference from the rule.10 I agree that we must leave no stone unturned in the
presence of paradox, but that suggestion seems to me too difficult. How can someone
be properly said to be following a general rule through a series of cases if she is not in
effect instantiating its content to them in what amount to successive inferential steps?
Boghossian himself was surely right to stress that a rule-follower must in some sense
accept any rule that she is following, and surely in following it she must somehow process
its instructions in such a way as to bring it to bear on new specific cases? Is there any
intelligible model of such processing that does not involve inference? Of course, how
exactly we should understand the notion of inference is itself a vexed issue. But the
point is not whether we call this “processing of instructions” inference or not, but that,
however described, it cannot be itself any form of rule-following or we will in effect
reinstate the first of the two regresses. Nor can it in general involve processing a
judgment that the trigger condition for the rule is satisfied or—at least on the Rulish
picture of rational belief —we will reinstate the second of the two regresses.
A more plausible direction is that the mistake—or the thing we need to revise—is
the supposition that inference is everywhere and essentially a kind of rule-following. In
that case, we can allow the inferential interpretation of the ACRE model and avoid the
first regress. But still the point will remain that in order to lead to rational belief, any
inference—whether it involves rule-following or not—will still require rationally
accepted premises. So more would still need to be said to staunch the second regress.
And if inference is not a matter of rule-following, then presumably the Rulish picture
of rational belief will be wrong for some kinds of inferentially formed belief. So in what
sense will such beliefs be rational? And in any case, if inference is not (always) a case of
rule-following, what is it when it is not?
Neither Paul Horwich nor Christopher Peacocke directly addresses this cluster of
problems in their chapters, but the views they respectively defend represent nonethe-
less two of the most salient directions of possible reaction. Horwich would be content,
I surmise, with the ACRE model as an account of what he terms explicit rule-following,
but he contrasts that with the wide range of cases—par excellence, rational belief-
formation and language mastery—where normative constraints seem to be in play and
we are attracted to the idea of a kind of implicit rule-following. And of this idea he
10
That rule-following might, in the most basic case, consist in actions for which reason is somehow
directly provided by a state of acceptance whose content is that of a (general) rule, without intermediary
inferential processing—that such a connection may be a primitive ingredient in rational action—was
apparently mooted by Tyler Burge in conversation and is the hesitant parting shot in Boghossian 2008.
REPLIES 387
moves very quickly to a deflated account, whereby the ACRE model’s core idea of
acceptance of the rule—of the performances in question as sourced in an information
state that generates mandates over performance—is simply dropped in favor of a
conception of performance as governed by ideal laws only in the manner in which
patterns in nature are governed by covering laws in natural science. Stones do not carry
the information conveyed by a statement of the laws of gravitation to which their
motion conforms. If acceptance is essential to rule-following, then Horwich offers
regularity, not rule-following. The idea that the performances concerned are neverthe-
less subject to normative constraint is grounded, in Horwich’s view, not in the existence
of mandates issuing from internalized rules but in our customary practice of holding
them open to appraisal, qualification, withdrawal, and apology. This critical aspect of
our practice is likewise subject no doubt to meaning-constituting covering laws; but is
not informed by the content of those laws in the mode of explicit rule-following.
Horwich’s primary concern in his present contribution is with meaning, rather than
rational belief. But one can foresee the outline of the corresponding view of rational
belief. If entry-rational beliefs are properly required to be beliefs sustained by the
processing of reasons for them, under the aegis of correct epistemic norms whose
contents are represented in the mind, then the view will be that we should give up on
the notion and accept that most of our beliefs are, simply not entry-rational. Their
formation may still present an example of implicit rule-following in Horwich’s deflated
sense: the patterns of their formation may be subject to natural law, and their accep-
tance and rejection held subject to criticism and correction. But the rationality of a
belief will not in general equate with its being formed in the light of rational constraints.
On such a view, the ground is cut from under the regressive paradoxes. I shall later
touch on the matter of whether it involves surrender of anything that we might
understandably wish to keep.
Tacit Knowledge
Peacocke’s proposal is quite another thing. While his discussion is primarily directed at
the other half, so to speak, of the rule-following problem: that connected with
meaning- and content-scepticism, and at the development of a framework in which
investigation-independence can be saved without any epistemically problematic Pla-
tonization of rules, meaning, and concepts, the apparatus he outlines is nevertheless a
prototype of a second salient form of response to the collapse of the generalized ACRE
model. This form of response is to propose that the generalized model fails not by dint
of misconstruing entry-rational belief as an effect of our processing the requirements of
norms of rational belief-formation, somehow internalized, but in a misunderstand-
ing—encapsulated in the exposition of the model in personal-level terms—of the
nature of the information-states involved and the thinker’s relation to them. For
Peacocke, the crucial step to a coherent view of the matter is their reconstrual as states
of tacit knowledge.
388 CRISPIN WRIGHT
One crux here lies in the interpretation of ‘tacit knowledge’. In an early exchange
with myself about rule-following and semantic theory,11 Gareth Evans too took
recourse to the idea of competent speakers’ states of tacit knowledge of the axioms
of a semantic theory for their language. But for Evans, talk of such states stood to be
interpreted dispositionally, specifically in terms of dispositions to understand whole
sentences in certain ways. A speaker’s tacit knowledge, for example, that “London”
refers to London was identified with a disposition to construe the truth-conditions of
sentences containing “London” in referential position in certain ways. Semantic
theory, on Evans’ view in that paper, was to be construed as a purely descriptive
enterprise, specifically as the project of giving systematic, compositional syntax-based
descriptions of speakers’ pattern of interpretation of (novel) sentences in their language.
This conception was precisely intended to contrast with the project of systematic
formulation of a body of information whose realization in the cognitive architecture
of a speaker would somehow explain his mastery of the language concerned—the
conception of semantic theory to which, in my half of the exchange, I had worried was
destined for a losing contest with Wittgenstein’s discussions of rules.
It is otherwise with Peacocke. On his account—moving now to rational judgment
rather than language mastery—the invocation of tacit knowledge is precisely aimed at
explanation where Evans was content with description.12 The invocation of tacit
knowledge is to assist with the explanation both of what entry-rationality consists in
and of how thinkers are enabled to arrive at entry-rational judgments.
According to Peacocke, concepts are identified by their “fundamental reference rules”
and grasp of a concept consists in tacit knowledge of such a rule for it. For example, the
fundamental reference rule for the observational concept round is, he offers, that some-
thing falls under it just if it has the same shape as things are represented as having when
they are experienced as round.13 In moving to an entry-rational non-inferential
11
Evans 1981.
12
I am simplifying some of the subtlety in Evans’ discussion. Evans was clear that we should understand
the disposition associated with a semantic axiom as a single state lying at the causal source of its manifestations
in a speaker’s linguistic practice; and clear too that in order to legitimate thinking of the relevant dispositions
that way, it was crucial that there should be additional evidence, going beyond the patterns in their construal
of whole sentences, for speakers’ possession of such dispositional states. He anticipated that such additional
evidence would be forthcoming from patterns of acquisition (on learning) and decay (say, in brain-damaged
subjects) of whole-sentence competences. The important point is, however, that even with dispositions
robustly so conceived, the theory of meaning, as conceived in his paper, is still a descriptive project, with no
answer essayed to the question, “How are speakers able to parse novel utterances?” It is a further step—
indeed, a jump—to the thought that the states concerned should be conceived specifically as information-
bearing states, rather than so-far unspecified states with the appropriate causal powers. (Some of the issues here
connect with the contemporary debate concerning knowledge-how and information that has been stimu-
lated by Stanley and Williamson 2001.)
13
There are, naturally, major methodological questions about how such rules are to be identified.
Peacocke offers several examples: the fundamental reference-rule for the perceptual demonstrative that F,
for instance, is alleged to be that where “the apparently perceived object is given in way W in perception”,
a thinker’s tokening of it will refer to the unique object perceived in way W by the thinker at the relevant
time. These proposals seem, by comparison with the familiar kind of homophone risk-free formulations of
1970s semantic theory, to involve substantial reflective analysis, executed in the armchair, and one might
REPLIES 389
judgment—say, that [that object is round]—a thinker draws on his tacit knowledge of the
fundamental reference rules for the concepts involved, possibly in combination with
further information, thereby arriving at an appreciation that a certain state or states—
perhaps an experience of the object as round—gives reasons for thinking that the relevant
judgment is true.14 This ‘drawing’ and ‘arriving at an appreciation’ is psychologically real,
but it is not, Peacocke is quite explicit, a matter of conscious inference or thought. “There
is a subpersonal transition from the content of the tacit knowledge to the appreciation that
a certain kind of perceptual experience makes a perceptual application of the concept
round rational in the given circumstances.”15
In sum: the judgment that P is entry-rational when it is made on the basis of a state
S that is recognized to provide reason for P by a suitable derivation from the funda-
mental reference rules for the concepts in P, together perhaps with other information.
The relevant S may consist in other beliefs, and the judgment that P may then be
inferential.16 But it may also consist in an episode of sense experience, or of apparent
memory, or perhaps a ‘finding obvious’.
One immediate question is how this kind of model is supposed help with issues of
content-scepticism and investigation-independence. Obviously it cannot do so unless
the place occupied by a fundamental reference rule in the kind of information
processing that Peacocke envisages is filled by something whose content is determined
in ways that are proof against the familiar sceptical arguments. Yet about this crucial
issue Peacocke’s discussion impresses as more than a little casual. He is right, of course,
that the content of a subpersonal state is not something that we need to constitute by
linguistic practice, nor fix by our thoughts. But the question still arises, what factors do
contribute to the fixation of the content of such a state, and how determinate a content
are they empowered to fix? Peacocke raises the issue only in the closing paragraphs of
his essay, and briskly canvasses the idea that an account based on the causal explanatory
role of such states has “reasonable prospects”. One might well be pessimistic about that.
A cardinal part of the causal explanatory role of such states, if Peacocke is right, is in
explaining the formation of content-bearing attitudes and the utterance of content-
expressing sentences. It is quite unclear how any account of the causal explanatory roles
of the relevant tacit knowledge states could succeed in fixing determinate contents for
them if this aspect of their roles was to be left out of account. But, if it is to be taken into
account, how is any path back from these explananda to the content of the underlying
wonder if that methodology can be reconciled with their role as the putative contents of tacitly known states.
Such concerns are taken up in Peacocke 2008 but I cannot pursue the issues here.
14
“The content of the tacitly known reference rule itself also contributes to an explanation, non-causal, of
why the experience of an object as round is a good reason for the judgment that the perceptually given object
is round, when the thinker is entitled to take experience at face value.” This volume, pp. 55–6.
15
This volume, p. 56.
16
It won’t be, though, if first-order beliefs can rationalize their second-order counterparts—a kind of
model of intentional psychological self-knowledge that Peacocke himself has long advocated.
390 CRISPIN WRIGHT
tacit knowledge states supposed to work unless fully determinate conceptual contents are
already ascribed to the former?
I won’t pursue this matter further here, however. The issue I wish to consider in more
detail is how Peacocke’s model might help with the regresses. Obviously enough the
Minor Premise problem is addressed immediately: the model is explicit that the triggers for
rational movement to new beliefs need not themselves be beliefs, nor any states for which
the question of antecedent reason arises in the way required to launch the Minor Premise
problem. But the situation with the Inference problem is less clear. While the ACRE
model represents the cognitive processes involved in rule-following in personal terms—the
states of mind and movements of thought it posits are those of a fully intentional subject—
there is no immediate prospect of assistance with the Inference problem merely in the
suggestion that we should recast the model in terms of subpersonal processes and states. As
Boghossian observes, if subpersonal rule-following involves subpersonal inference—as on
Peacocke’s account it seems it does—and subpersonal inference is itself a form of sub-
personal rule-following, then the threat of the regress remains. Subpersons—or anyway,
finite populations of subpersons—are no better at supertasks than persons are. The
Inference problem will not be helped by retaining the structure of the processes of rule-
following that the ACRE model postulates and merely insisting that the model must
sometimes be realized subpersonally rather then personally. If going subpersonal can help,
it will be because it somehow allows us to refashion that structure.
So, does it? The question turns on the extent to which the subpersonal operations
involved in the acquisition of rational belief on the kind of account proposed by
Peacocke are themselves to qualify as rational, and what it can legitimately mean so to
describe them. Let’s prescind for a moment from the fundamental reference rules
central to Peacocke’s account, and consider the matter in the most abstract terms. Let
S1 be a state of registration (subpersonal acceptance, or programming) of the rule:
If P, then !A
where A is a type of action and ‘!A’ expresses some form of mandate for a performance
of A. Let S2 be, correspondingly, a state of registration of an instance of the trigger
circumstance, P.
Suppose Agent starts in these two subpersonal states (and is in no other state which is
material to the rationality of his doing A) and then does do A. Does he act rationally? In
one—deflated—sense, yes, of course: he does something for which he has reason,
something that is rationally sanctioned by his subpersonal states of information. But that
is not the sense of ‘rationally’ that interests Peacocke. Peacocke requires that in order to
qualify as rational, specifically: as the rational formation of a belief, an action must be
sourced in the information states that provide its mandate, even if the sourcing is to
operate entirely at the subpersonal level. The question is whether this requirement can
be fleshed out without generating an analogue of the Inference problem.
Let’s go carefully. For Agent’s token of A to be so sourced, it needs to be an action on
an information state S3 of his cognitive system which registers that there is the mandate:
REPLIES 391
!A. (No doubt that notion—what constitutes action on a mandating information

state—may prove troublesome under philosophical pressure, but I am going to take
it for granted here.) The question is: how did Agent get into S3? Well, if he did so
rationally, it must have been that he in some sense ‘transitioned’ from S1 and S2 to S3.
So let A* be his token of that transitional act type, and ask: was A* performed
rationally? Again, in the deflated sense, yes. The movement from information states
S1 and S2 to S3 is licensed by the entailment of the content of S3, viz. [!A], by the
contents of S1 and S2, viz. [If P, then !A] and [P]. But again, the requirement is not
merely that the transition, A*, be one which is licensed by an entailment, and is in that
sense of a rational type in the context, but that Agent’s token of it be rational: that it
have been performed because of that entailment. And that in turn requires that there
have been an information state S4 of Agent’s cognitive system which registers that there
is a mandate—!A*—for the transition A*, and that the token transition have been an
action on that information state. Suppose so. Then the question is: how did Agent get
into S4? Well, if he did so rationally, it must have been that he in some sense
transitioned from ulterior states whose contents entail the content of S4. Let A** be
Agent’s token of that transitional act type and ask: was A** performed rationally? . . .
So there is still a regress. Going subpersonal doesn’t help, at least on the assumptions
made. Those assumptions were: (i) that any token action of the system, including in
particular any particular transition from one set of information states to another, is
rationally performed only if mandated by a suitably specific information state; (ii) that
the achievement of such a specific information state will in general be possible only by a
transition from states encoding relevant (general) information to the specific informa-
tion state in question. Under these assumptions, one token rational transition demands
the prior performance of another . . .
It does not seem to me likely that (ii) is negotiable. If that is right, then going
subpersonal can assist with the Inference problem only if we relax the requirement that
a movement of the system counts as rational only if it is made on a suitably mandating
specific information state. We have to allow that some of the informational transitions in a
rational system are merely licensed, in the sense above—that they merely conform to rational
pattern—rather than being mandated, in sense of: made because the system has entered
into an information state that encodes the relevant license, on which it then acts. In a
nutshell, a rational system must, at some level exhibit mere regularities of rationality—
brute rationality—rather than any subpersonal analogue of personal rule-following.
This brings us to the brink of what should be a much fuller discussion, which
I cannot undertake here, but let me pursue the matter just a little. It seems to me that
the immediate effect of the foregoing reflections is to blow away the motivation for the
thoroughgoing rationalist position that Peacocke attempts to take. We have to
acknowledge that psychological representations, whether personal or subpersonal, of
principles which we would like to regard as rules of rational belief-formation—rules of
logical inference, and the kind of ampliative rules of transition which philosophers
have been tempted to postulate for perceptual justification, induction, and perhaps
392 CRISPIN WRIGHT
self-knowledge—cannot coherently be regarded as always playing an indispensable

part in the cognitive economy of a rational thinker who moves to rational beliefs in the
ways they prescribe. Peacocke’s model of rational belief-formation involves play with
principles additional to these, viz. his fundamental reference rules for the concepts
involved in a judgment which a rational thinker allegedly processes in arriving at an
‘appreciation’, as he puts it, that a certain state constitutes sufficient reason for accep-
tance of the judgment. This appreciation should be thought of as, in our terms, an
information state mandating acceptance of the judgment. But since we know that some
of the rational transitions involved in the accomplishment of such a state must be brutely
rational, without the source of mandates achieved by processing via representations of
the relevant rules, why believe, or require, that it is any different with fundamental
reference rules? Why should they feature in the cognitive architecture of a rational
thinker, as items of his tacit knowledge, rather than merely encode patterns of belief-
formation which, no doubt as a result of training, he is disposed to display? Why does
there have to be any cognitive state mandating the making of the judgment at all?
The doubt tends in the direction of the sort of view I take to be Horwich’s, where
the conception of a rational thinker as one whose judgments are characteristically
reached by processing of information states encoding the requirements of rationality is
supplanted by that of a thinker who is merely characteristically disposed to judge what
it is rational for him to judge, for who-knows-what cognitive psychological or any
other cause. And clearly, purely as far as the thinker’s performance is concerned, there is
not going to be anything to choose between a Peacockian form of explanation of his
judgments, and another—austere dispositional—model which will continue to find
work for the idea of registration of trigger conditions, but, rather than view the
consequent transition to belief as driven by an apparatus of cognitive psychological
states encoding the content of logical, epistemic, and fundamental reference rules,
views it simply as the exercise of a bare disposition to conformity with the rules
concerned. Once again: dispositions in general do not call for explanation by means
of states of a system conceived as encoding the content of rules prescribing the kind of
performance that the disposition in question is a disposition to execute. Why should it
be any different with the causation of what we regard as rational belief-formations?
What Peacocke’s programme appears to need is a well-motivated distinction in the
respective roles to be played by epistemic norms and fundamental reference rules. It has
to be acknowledged, on pain of the regress, that in order to contribute to the rational
formation of beliefs, transitions in the cognitive processing of a rational thinker do not
need to be mandated by specific information states sourced in tacit knowledge of
appropriate epistemic norms. I take that to be proved by the reflections above. But it
can still consistently be insisted that specific judgments made at the end of the line, as it
were, do require a mandating information state, and that it must be achieved by
suitable processing of fundamental reference rules if it is to subserve genuinely rational
belief. But how might this distinction be motivated?
REPLIES 393
There is, indeed, a further concern: on Peacocke’s view, the rationality of a thinker’s
judgments now becomes a question of the structure of the actual scientific explanation of
his making them, no matter how impeccably rational they impress, in ordinary
intellectual commerce, as being. Once we recognize that we are stuck with the idea
that certain kinds of informationally unconstrained tendencies are an essential part of
rational judgment, it seems to be an open scientific question how far that phenome-
non—absence of informational mandate—extends. The austere dispositional view says:
why not all the way? To insist that that answer be resisted, is to endorse a conception of
rational judgment as not merely an ability but an ability that needs to be grounded in a
certain way in order to count—as if adroit shot selection at tennis had to be grounded
in a certain kind of cognitive processing before it counted as a genuine skill. This, it
may be contended, is to confuse the (possible) explanation with the datum to be
explained. Peacocke’s conception leaves our rationality—contrast: our behaving as if
rational—hostage to empirical scientific (cognitive psychological) fortune.
But there is also, it needs to be said, a very significant cost to the drift towards
austerity. Indeed, I think the dialectical situation here brings out real tensions in our
basic thinking about human rationality. In other spheres of good performance—for
example, acting well ethically—the contrast between doing the right thing and doing it
for the right reasons is paramount. There is all the moral difference in the world
between action for the right reasons and action of the same type, in the same context,
on quite other or no good reasons. It is an analogue in the epistemic sphere—in the
ethics of belief —of this idealized conception of fully responsible moral agency that is
the casualty of the regressive argument. The austere dispositional view embraces the
limitation. But what is lost thereby is akin, in the doxastic sphere, to loss of the kind of
value associated with courage in contrast to recklessness. That is a major loss, and it
would be a prize worth having if there were some principled way to limit the damage.
The Normativity of Meaning

The normativity of meaning is a lynchpin of Kripke’s Wittgenstein’s ‘Skeptical Argu-
ment’.17 It is, for example, because meaning is supposedly normative, and because that
is taken to involve that understanding involves sensitivity to norms, that a dispositional
account of meaning—the natural avenue to explore after the apparent failure of
antecedent use and mental facts to determine meaning—is allegedly foreclosed.18
17
I shall use the American spelling, and the corresponding ‘Skeptical Solution’, whenever referring to the
dialectic in Kripke’s exegesis.
18
It is true that the Kripkean argument also makes play with the finitude of our dispositions, in contrast
with the presumed infinitary determinacy of meaning, manifest in, for example, the determinacy in truth-
value of the infinite range of quantifier free arithmetical equalities. But this is actually just the same point
again. For meanings to be determinate in infinitely many cases is for them to contribute towards the
determination of the truth-values of infinitely many sentences, irrespective of our dispositions of assessment
of those sentences. The truth-values so determined will then supply a standard of correctness for those
dispositions.
394 CRISPIN WRIGHT
Dispositions can at best determine what we are inclined to say; meanings, by contrast,
determine what we ought to say.
The claim has been widely accepted since Kripke’s discussion. But it is not incon-
trovertible. Certainly, the concept of linguistic meaning is inextricably linked with that
of correctness in linguistic practice. It is only utterances that have meaning for which
there is any such thing as correct or incorrect use; and whether an utterance is correct
or not depends on what it means and how matters in the world stand in relevant
respects. But this is not enough to ensure that meaning is normative in more than a
deflated sense. A conception of normativity serving merely to respect these correctness
platitudes needn’t amount to anything that would pre-empt dispositional or other
forms of naturalistic construal. Indeed, the Skeptical Solution itself, whereby the letter
of the platitudes is (as intended, anyway) saved, illustrates the point. The idea of the
normativity of meaning presupposed in the Skeptical Argument is a full-blown
conception, according to which grasping a meaning is a state of information involving
internalization of a standard of correctness, a principle that can be used in conjunction
with an appreciation of the relevant facts to determine whether or not (there is reason to
regard) a token utterance is true. So regarded, meaning stands to correct linguistic
performance precisely in the role of a rule to a practice it governs. The moral of the
preceding section of these remarks, however, was that it is incoherent to regard the
operation of standards of correctness as everywhere involving their internalization in
determinate information states of practitioners. That conclusion was forced for the case
of rational belief-formation. But in parallel, there has to be space for a conception of
what it is to understand an expression which centralizes the ability to make correct use
of it without viewing it as informed by an internalized rule determining what is and
isn’t correct use.19 The normativity of meaning, on the full-blown interpretation, is no
platitude. It is contestable, and is indeed contested in his present chapter and elsewhere
by Paul Horwich.
Meaning-Intentions as Degenerate
Let me, though, before turning to Horwich’s proposals, respond to the very different
challenge to the normativity of meaning developed by Akeel Bilgrami. In commentary
on Kripke’s book written shortly after its appearance,20 I proposed that the letter of the
Skeptical Argument might be addressed simply by invoking our intuitively immediate
19
It is a corollary of this that Kripke’s Skeptical Argument, if successful, undercuts, in a way, its own
objection to dispositional views. It is the full-blown interpretation of the normativity of meaning that is
beyond dispositional construal. Once that interpretation is abandoned, there is space for broadly dispositional
accounts that place normativity not at the source but in the character of the practices—retraction, criticism,
endorsement, and so on—that the meaning-constituting dispositions are dispositions to engage in. If that is
correct, then the distinctively non-factualist tenor of the Skeptical Solution is unmotivated. Although he
finds no use for the notion of a disposition in his proposals, I take Horwich’s proposal to make essentially this
move.
20
Wright 1984.
REPLIES 395
self-knowledge, in the ordinary case, of our own intentions—and hence that the real
issue raised, when one is challenged to adduce facts in virtue of which one means (or
meant) one thing by an expression, rather than another, is to understand better the nature
of immediate (or recollected) intentional self-knowledge and how it is feasible. Bilgrami
accepts this broad thought, identifying or assimilating what it is to mean an expression in
a particular way with a kind of intention—what Boghossian in his contribution to this
volume calls the Intention View21—at least for the purposes of his present essay.
Normativity comes into question, in Bilgrami’s view, not because the assimilation of
meaning to an intention, or any other intentional state, is in question, but because the
relevant intentions, the ones that look best suited to determine meaning, are, Bilgrami
contends, a “degenerate” case of intention, and in effect non-normative.
A natural candidate for a meaning-constituting intention for, say, a predicate would
be the intention to respect a certain satisfaction-condition in one’s use of it—for
instance, to apply “square”, or to acknowledge it as correctly applied, just to square
things. That is an intention one could fail to live up to. But as a proposal for what it is to
mean square by “square”, there is, as Bilgrami observes, a difficulty with it.22 We do not
want to make it a condition of understanding “square” that a speaker intend to be
honest; or—if this is something he could anyway strictly intend—to make no mistakes.
Even a persistent failure to apply, or to acknowledge as correct applications of “square”
to square things need not enjoin that a speaker has failed to use “square” with the
meaning: square. The explanations of the aberrant uses may lie elsewhere. If meaning
so-and-so is an intention-like state, it is, at any rate, not the intention to use the
expression correctly, or to comply with certain standards, or uphold certain regularities
in one’s use of it but rather, as an approximation, to accept that one’s uses are properly
assessable by those standards, or open to assessment as correct or not according to
whether they maintain those patterns. Meaning square by “square” is thus better
assimilated to the intention to allow that one’s uses of “square” are answerable to the
21
Boghossian argues that the ultimate objection to any Intention View is that it must succumb to the
Inference problem. This needs a more extended discussion, but let me observe that it cannot in general be
correct to think of action upon intention as involving any kind of inference. Whatever the conclusion of such
an inference, one would still have to act on that; there simply has to be some notion of inferentially
unmediated action on an intentional state if ordinary psychological explanations are ever to serve to
rationalize an agent’s behavior. And once that is granted, why cannot ordinary intentions be such directly
action-connected states?
Boghossian’s point, however, is not so easily finessed. For in the case of meaning-constituting intentions,
the contents concerned will presumably be conditional and general: the intentions in question will be intentions
to abide by a rule, or maintain a certain pattern in one’s linguistic practice, and it is unclear what sort of model
might be given of action upon these intentions without involving processing of triggers and simple inference.
I am inclined to stand by the proposal that the problem raised by Kripke’s skeptic must, insofar as linguistic
practice is rational, intentional activity, ultimately concern our knowledge of the intentional states that
underlie linguistic performance, including meaning if that is one such state. But I agree with Boghossian that
simply falling back on a comparison between meanings and, presumably, general intentions holds no evident
promise of help with the regresses.
22
The relevant intention is, of course, to be construed as conditional: roughly, that if the question arises,
then to apply, or acknowledge the application . . . etc. One would need to be very busy otherwise.
396 CRISPIN WRIGHT
relevant satisfaction-condition than to have them conform to it. It is a standard-

determining intention, not an intention to comply with standards.
This point seems very well taken. But now Bilgrami suggests that the adjustment
is crucial in the following respect. Its effect is to take us from intentions which an
agent may, notwithstanding effort to the contrary, fail to comply with to intentions
which there is no such thing as genuinely having yet failing to comply with. And surely in
that case, he argues, we are no longer dealing with anything properly regarded as
normative. Anything that constitutes a norm has to be something that one can fail to
live up to.
It might be wondered whether an intention that is degenerate in this sense—an
intention the mere having of which on a particular occasion ensures that one lives up to
it on that occasion23—could be any real intention at all; if not, then the real tendency
of the point, if sustained, is that the original assimilation of meaning and intention is
basically misconceived. But Bilgrami goes in the other direction. Meaning, in his view,
is an intentional state; but it is not a normative state.24
The suggestion that meaning-intentions are peculiarly non-normative is intriguing
and worth further scrutiny. Bilgrami’s argument for it recapitulates core themes from
his Belief and Meaning.25 How might I intend my use of an expression to be
answerable to a certain satisfaction-condition and yet it fail to be so answerable?
More generally, how might I intend a certain performance to be answerable to
certain standards of evaluation and yet fail to make it so? Suppose I set myself to
cook a chicken chasseur successfully by Cordon Bleu standards (minimal passing
level). I might produce so bizarre and inappropriate a performance that it falls
outside the range of evaluability by the standards in question (I deep-fry an old
inner tube in castor oil, say). Or I may not know what Cordon Bleu standards
involve, so that I am in no position to set them for myself. Bilgrami envisages two
corresponding ways in which my linguistic performance might fail to be answerable
to a satisfaction-condition that I intend it to answer to. I might misspeak myself,
producing a malapropism, or nonsense word, and so produce a performance that is
not evaluable by the satisfaction-condition I intended for my uses of an expression
I thereby fail to use. Or—the case on which he expends most attention—I may not
23
The reference to the occasion is essential, of course. There is no difficulty, in the cases concerned, with
the idea of a change of mind.
24
The reader might wonder whether, if Bilgrami is right, there are not resources here to deflect those of
Kripke’s arguments against dispositional construals of meaning which presuppose its normativity. For if
meaning is a non-normative mode of intending, there is no force in the charge that the much-vaunted
normativity of meaning goes missing on any dispositional construal. It is missing anyway.
But there is some doubt about the dialectical relevance of this thought. If someone is content to accept
some form of Intention View of meaning, he is presumably not going to be drawn to a dispositional account
in any case.
25
Bilgrami 1992.
REPLIES 397
know, for semantic externalist reasons, what that satisfaction-condition actually is,
so that it is not by that standard that I intend my uses of the expression to be
evaluable as correct or not. Nevertheless, according to the usual externalist con-
tention, it will be in fact to the externally determined standard that my uses will be
accountable.26
Bilgrami’s response to the second case is to repudiate the externalism that suppo-
sedly opens the gap. The burden of the greater part of the argument of his essay is
that any form of semantic externalism strong enough to render one’s own meanings
opaque to one, so that the intention to use an expression with a particular satisfaction-
condition may be out of accord with the actual satisfaction-condition that, in one’s
mouth, it has—any such semantic externalism will be impotent to draw the
proper and necessary distinction between irrational belief on the one hand and
beliefs that are wayward through ignorance or error about a posteriori matters on
the other. In his view, only a conception of meaning that, in the fashion of Frege’s
notion of sense, renders it transparent to the thinker can properly subserve that
distinction.
To engage this claim in any depth would take me far beyond the pardonable scope
of these comments. Let me merely anticipate the externalist reply that the distinction
that Bilgrami rightly demands be safeguarded marches in step with the distinction
between glitches in one’s belief system that are apprehensible a priori and those whose
disclosure requires a posteriori investigation. If irrationality is essentially an a priori
detectable vice, then when content is fixed non-transparently to a thinker, the
presence of certain contradictory beliefs in his system need reflect no irrationality.
But no valid transition is apparent from that concession to the conclusion that none of
the vices of irrationality, including the presence of contradictions, will any longer be a
priori detectable in a system of belief involving externally determined contents. So
unless we are dealing with a form of externalism whose effect is to undercut the a priori
altogether, the expectable upshot will be at most to redraw the boundaries of irrational
belief management, rather than to obliterate them.
26
For what it is worth, I have some difficulty seeing that either type of case really does frustrate the kind of
intention that is relevant. If I suffer a slip of the tongue, or other misspeaking, I fail to use words that
I intended to use. That isn’t the same thing as failing to implement the standing conditional intention to have
my use of those words answer to certain satisfaction-conditions. To frustrate that conditional intention, it has
to be that I do utter the intended words but that they somehow fail to be answerable to the relevant
condition. As for the externalist case, the question is: what is the satisfaction-condition that I intend my uses
to answer to? If it is an externally—socially or causally—determined standard, then presumably my intention
succeeds. (Ignorance of a standard need be no barrier to an intention to be held answerable to it.) If it is an
internally determined, ‘narrow’ or idiosyncratic condition, why should others’ holding my use accountable
to the external standard count as frustrating that intention? The impact of externalism in this case seems to be,
not to provide a way that the meaning-determining intention can be frustrated after all, but rather to bar its
qualifying as meaning-determining.
398 CRISPIN WRIGHT
Let it be, though, that Bilgrami is right in his principal contention: the thought that to
intend that one’s uses of “square” be answerable to the satisfaction-condition: x is square,
is not normative since not an intention that one can fail to live up to. I think we should
pause before drawing the conclusion that this undercuts the normativity of meaning, even
when meaning-fixing intentions are construed as standard-setting intentions. There is an
underlying issue here about the sense, if any, in which intention, even regular non-
degenerate intention, is properly described as normative in any case. That one goes into a
context with a certain intention does not per se set a standard of correctness for one’s
behavior in that context. A failure to act on the intention may reflect a change of mind, or
distraction, or some misapprehension about what it takes to implement the intention in
the circumstances, or some puzzling form of akrasia. But there is no clear sense in which,
in any of these cases, one acts incorrectly. If assimilating meaning to intention is, inter alia,
meant to provide for the normativity of meaning in a sense relating to the setting of
standards for correct linguistic practice, it is the contents of the relevant intentions, rather
than merely their intentionality, so to speak, that is going to do the work. What is required
to provide for correctness and incorrectness is not—or not immediately—that the inten-
tions be ones one can have yet fail to live up to, but rather that the standards they are
intentions to answer to be ones that one can fail to live up to. The intentions concerned
can be degenerate in Bilgrami’s sense and yet deliver all the normativity that we could
want. That will be the case if, even if the relevant intentions cannot themselves fail to be
lived up to, the standards which they are intentions to answer to can.
There is more to say. I am sure I am not the only philosopher to have been worried
by the looseness of the notion of ‘the normativity of meaning’ as it has featured in these
debates over the last thirty years. When some parameter is described as normative over
a certain practice, the generic idea is that it contributes towards determining some
notion of good practice; but ‘good’ may be filled out in ways that involve any or all of
correctness, value, or support by reasons. Thus each of rules (correctness), morality
(value), and intentional states like belief and desire (reasons) are said to be normative.
This is a landscape littered with tissues of confusion. The mis-assimilation, noted by
Bilgrami, of meaning square by “square” to the intention—crudely—to apply the word
just to square things betrays a conflation of the first two notions (the setting of a
standard is confused with the desirability, presumably derived in this instance from the
value of truth, of incurring positive evaluations by that standard). It is a further
question, to be broached shortly, whether meaning—that is, the state of understanding
an expression in a particular way—should be considered as contributory to a subject’s
reasons for her linguistic actions. Belief is normative in that way; but it has nothing
especially to do with correctness, or value.
Meanings as Fixed by Natural Law

On Horwich’s view, the Skeptical Argument goes wrong in the assumption that
because meaning is a normative concept, meanings—if they exist—have to be available
REPLIES 399
to speakers as guides and justifiers of their uses of language. For Horwich, as noted, the
meaning-constituting facts reside in the ideal laws governing speakers’ uses of expres-
sions. These are natural laws, broadly comparable in status to those of any natural
science. It is true, and important, that in linguistic practice we in many respects act as if
we were following known rules constitutive of correct ‘play’: we criticize and correct
ourselves and others, we qualify and reformulate what we say, we stick to it or retract it.
But this is, I take Horwich to say, a basic propensity. These corrective aspects of our
practice are “fundamental to our language-game”27 and while, with all other aspects,
they are to be explained in terms of relevant meaning-constituting fundamental
regularities, the template of this explanation is not that of intentional rule-following;
and the details of what we find acceptable practice are not in general grounded in
psychological states personally or subpersonally encoding the relevant laws.
Horwich’s views are complex and are developed more extensively elsewhere.28 He
recognizes, of course, that his stance involves revision of many of our preconceptions
about meaning and would doubtless respond that one effect of the Kripkean paradox is
to teach us that such preconceptions have to be given up. The relevant questions are
therefore whether the broad approach can be fit for purpose—that is, whether it
restores a notion of determinacy of meaning that is proof against further sceptical
challenge—and whether the preconceptions whose sacrifice it involves are, more than
preconceptions, essential ingredients in the notion. I will briefly canvass one doubt on
each score. They share a common root.
The common root is the point that language is intentional, rational activity. The
sayings and writings that constitute public linguistic practice are actions, up for rational
explanation by the citation of intentional states of the language user. That is enough to
call into question whether we should expect any simple regularities in speakers’ public
linguistic practices of the kind Horwich’s account rests upon. Holistic patterns of
explanation will operate: what speakers are prepared to say will depend not just on
meaning but variable collateral beliefs and desires, and their meanings will consequently
surface in their linguistic behavior in the essentially variable and indirect way that, for
example, their beliefs and desires do. The answer to the question, how will someone use
“square” who means by it: square—like the answer to the question: how will he behave
if he thinks the ice is thin—can only be: ça depend (on what he thinks about all kinds of
things, what he wants, and how he understands other expressions).29
27
Compare Wittgenstein RFM (VI, 28).
28
See especially Horwich 2005. The normativity meaning is focused on in Ch. 5.
29
The reader should not get the impression that Horwich himself passes this holism by. See for instance
Horwich 2005, p. 41. But he believes its impact is qualified if we focus on activity at the level of thought, and
in a language of thought. At this level, considerations like what a speaker wants to communicate and how he
wants to go about it will not impinge, and the patterns assumed by his use of expressions will consequently be
uncomplicated by such factors. But that does not affect the basic point. For even someone’s patterns of
(involuntary) acceptance of an ‘internal sentence’ will still be complicated by lots of other factors, including
background information, weighting of evidence, risk aversion, and collateral mistakes, which may vary even
though his understanding of the sentence does not.
400 CRISPIN WRIGHT
Now obviously, if there are no expectable straightforward regularities in the use of

an expression which competent speakers—or thinkers—generally can be expected to
exhibit in common and which might be encapsulated in some foreseeable kind of ideal
governing law,30 then it is forlorn to hope that determinacy of meaning might be
grounded in the specifics of such underlying governing laws. Absent any reason to
think that the use of an expression allows of any purely natural scientific systematic
description, the only model we have of a systematic description of linguistic compe-
tence is that of the familiar kind of theory that assigns semantic significance to words and
modes of sentence composition from the start. So Horwich, the concern is, doesn’t do
enough to support the basic assumption of his view that there actually are any ideal laws
governing our use of words, fit to constitute determinate meanings.31
The idea that a thinker’s understanding of an expression combines with other aspects
of her intentional psychology to determine her reasons for employing it one way or
another is deeply engrained in our intuitive thought about these matters. It belongs
with it that rational linguistic practice is informed by an awareness of meaning. Hor-
wich’s putative ideal laws of use, by contrast, are nothing of which we are aware. He
anticipates this point, and finesses it by the counter that, in his view, we will be self-
aware of meanings only where their application can be assimilated to cases of explicit
rule-following in his sense.32 It is a datum, on the other hand, that where we are at best
merely implicitly following rules, we are not aware of what the rules are. But this view
of the matter may seem better adapted to a speaker’s relation to, for example, the
grammar of her language than to her understanding of individual expressions. The
grammar is identified theoretically on the basis of competent speakers’ individual
impressions of grammaticality and ungrammaticality. A thinker’s grammar will indeed
shape the detail of her linguistic behavior, notwithstanding the fact that what it shapes is
intentional action, subject to the holistic variations just noted. But—the crucial
point—the patterns that the grammar provides a theoretical systematization of are
invariant under those variations. What sentence I am prepared to produce, or assent to,
in given circumstances may vary with my beliefs and desires; but the patterns consistent
with grammaticality in my language will not so vary but will, ideally, be manifest,
whatever sentence is concerned. Grammar does not interact holistically with aspects of
my personal psychology in shaping my linguistic behavior in the way they interact with
each other. By contrast, my understanding of “square” does, as normally conceived, so
interact. It is not given to me as a hypothesis best explaining my impressions of
acceptable and unacceptable uses of “square”. Rather, as least as ordinarily conceived,
I choose my words, and do so in the light of knowledge about what I am trying
30
Horwich does not say how he understands the notion of a governing law, but we can take it that such a
law should at least subsume a range of regularities in the phenomena it ‘governs’.
31
Horwich’s grounds for this assumption are of course developed in other writings; see especially 2005,
Chapter 2.
32
See }8 of his essay, at p. 83 of this volume.
REPLIES 401
to achieve, beliefs about how things stand circumstantially in relevant respects, and
knowledge as to what my chosen words will say.
In sum. So long as we stand by the idea that the way a speaker understands an
expression contributes to the explanation of her production and reception of speech
acts as a reason-giving factor, alongside belief, desire, and intention, there is little cause
to expect the kind of regularities of linguistic behavior that could underwrite Hor-
wich’s ideal governing laws, or support their candidacy to constitute determinate
meanings. Conversely, the putative meanings constituted by such laws cannot be
items for ordinary psychological self-knowledge, any more than grammar can, and it
is consequently deeply revisionary of our ordinary ideas about rational linguistic agency
to propose that it is such laws that constitute the determinacy of our meanings.
I’ll return to this latter objection after further review of some of the issues
concerning self-knowledge of meaning in Part II of these Replies.
Bibliography
Boghossian, P. A. 2000 “Knowledge of Logic”, in P. A. Boghossian and C. Peacocke (eds.) New
Essays on the A Priori, Oxford, Clarendon Press, pp. 229–54.
Boghossian, P. A. 2001 “How are Objective Epistemic Reasons Possible?”, Philosophical Studies
106/1–2, pp. 1–40.
Boghossian, P. A. 2008 “Epistemic Rules”, Journal of Philosophy 105/9, pp. 472–500.
Evans, G. 1981 “Semantic Thery and Tacit Knowledge”, in S. Holtzman and C. Leich (eds.).
Wittgenstein: To Follow A Rule, London, Routledge & Kegan Paul, pp. 118–37.
Horwich, P. 2005 Reflections on Meaning, Oxford, Oxford University Press.
Press.
Peacocke, C. 2008 Truly Understood, Oxford, Oxford University Press.
Stanley, J. and Williamson, T. 2001 “Knowing How”, Journal of Philosophy 98/8, pp. 411–44.
Wittgenstein, L. 1956 Remarks on the Foundations of Mathematics, 3rd edn (Oxford: Basil Blackwell
1981).
Wright, C. 1980 Wittgenstein on the Foundations of Mathematics, Cambridge, Mass., Harvard
University Press.
Wright, C. 1984 “Kripke’s Account of the Argument Against Private Language”, Journal of
Philosophy, 81, pp. 759–78.
Wright, C. 1989 “Wittgenstein’s Rule-Following Considerations and the Central Problem of
Theoretical Linguistics”, in A. George (ed.) Reflections on Chomsky, Oxford, Basil Blackwell,
pp. 233–64.
Wright, C. 2001 “On Basic Logical Knowledge”, Philosophical Studies, 106, pp. 41–85, reprinted
in J. Bermudez and A. Millar (eds.) Reason and Nature, Oxford, Clarendon Press, pp. 49–84.
Wright, C. 2004 “Intuition, Entitlement and the Epistemology of Logical Laws”, Dialectica 58,
pp. 155–75.
Wright, C. 2007 “Rule-Following Without Reasons: Wittgenstein’s Quietism and the Consti-
tutive Question”, Ratio 20/4, pp. 481–502.
Replies Part II: Knowledge of Our
Own Minds and Meanings
Bar-On and Smith
The Problem of Self-knowledge

Although some philosophers may have preferred to think otherwise,1 it is not philosophy
but part of the ordinary folk notion of the mental that each of us stands in a special
cognitive relationship, denied to others, to our own thoughts, hopes, imaginings, and
sensations—that our mental lives are directly available to us and only indirectly available to
others, that “You cannot really know what another is thinking”, whereas of one’s own
thoughts one cannot but be aware. This notion is a premise for the other minds problem. It
is captioned in much early analytical philosophy of mind by the phrases, “privileged
access”, and “first-person authority”. The conception of the problem of self-knowledge
addressed in the work of mine2 that Dorit Bar-On’s and Barry Smith’s chapters respond
to is essentially that of achieving a satisfactory perspective on what is right about the
idea of “privileged access”, a perspective to head off the beckoning slide into other
minds scepticism, and the problems brought to the fore by Wittgenstein’s discussion of
private language, while at the same time finding a place for (perhaps disinfected versions
of) the seemingly undeniable asymmetries that motivate the folk notion.
It was, I think, Ryle who introduced the use of the term, ‘avowal’, in this context.
Since The Concept of Mind, the major tendency of philosophical discussion of psychological
self-knowledge3 has been to focus on its linguistic expression and on the asymmetries as
reflected in characteristics of the competent use of avowals, in contrast to the competent
ascription of mental states to others. If I say that you are in pain, or that you are expecting a
postal order in the mail, or anxious about the economic condition of English Premier
League football, I make a claim for which it is appropriate to ask for my reasons. But if
I avow any of those claims of myself, the request for reasons, it is commonly said, “makes
1
I am thinking of John McDowell, but perhaps unfairly.
2
There isn’t a lot: essentially just the Whitehead Lectures, and the substantially overlapping “Self-
knowledge: The Wittgensteinian Legacy”. But I have thought, and continue to think about the issues a
lot more than this rather meagre return indicates.
3
In general, I’ll just say “self-knowledge”.
REPLIES 403
no sense”. Again, if I make any of those claims about someone else, it is straightforward to
specify a context in which you might, without doubting my sincerity, reasonably wonder
whether what I was saying was true. But when the corresponding claims are made of
myself, it is at least not straightforward, and may in context be impossible, to envisage
circumstances in which, without doubting my sincerity, you might reasonably wonder
about the truth of what I have said. If, finally, I have a headache, or if I am concerned
about the financial predicament of English Premier League football, you would normally
expect me to be in a position to say so, and, ceteris paribus, willing to say so. But there would
be no normal expectation that you would be in position, or willing, to say so on my behalf.
These are, at first exposure, points about our linguistic practice. No doubt they
could stand refinement and nuance to accommodate details of context and variations
the kinds of mental state concerned, but there seems no prospect that they should
simply prove illusory on closer examination. At this point though, we confront a fork.
What comes first here: the linguistic practice, or the thoughts of the thinker manifested
in that practice? The problem of self-knowledge will look different depending on how
one chooses. On the first option, we will tend to think of matters—very much as Ryle
and Wittgenstein did, and as I think Bar-On does too—as centred upon explaining
avowals. The problem will be viewed as that of accounting for the distinctive aspects of
the “grammar” of avowals, in contrast to the competent other-ascription of mental
states; and a range of candidate explanations will become salient, for instance those
falling under the broad rubric of expressivism, which will seem point-missing on the
second option. On the latter option, by contrast, the relevant features of avowals will,
from the outset, be seen as reflections of certain special aspects of the epistemic
character of the self-directed thoughts they express. And it will seem overwhelmingly
natural to suppose just what the folk notion does suppose: that selves characteristically
know of the states that give rise to avowal in a way that involves no inference or
independent reasons, and which is characteristically very secure; and that the states of
the relevant kind are typically salient to their subjects. And the problem will then be to
account for these apparent epistemic advantages in a way that is proof against a slide
into Cartesian privacy and its associated nemeses.
The triptych of immediacy, authority, and transparency that I have used to configure the
explananda of the problem allows of interpretations to accord with either alternative:
language first, or thought first. Let us pick up the old term ‘privileged access’ as a name for
the triptych on its epistemic interpretation. It’s worth pausing a moment over the notion
of immediacy as an ingredient of privileged access. The distinction between inferential
and non-inferential judgment is itself in need of clarification but immediacy should here
not be equated with non-inferentiality in any case.4 If, for example, we take perceptual
4
McDowell, in “Response to Crispin Wright” (1998), takes it that this distinction rapidly undercuts any
diagnosis of the appeal, for folk philosophy, of a Cartesian, observational model as an apparent explanation of
privileged access. I reply that the distinction is a difficult one, and it is only too plausible that ordinary thought
would miss it.
404 CRISPIN WRIGHT
judgment as a paradigm of the non-inferential, there is still scope for views which hold that
perceptual judgments nevertheless have a basis in reasons: reasons provided not by other
beliefs but by states of perceptual awareness (or perceptual seeming). Such a state—its
non-doxastically seeming to one that P—is commonly regarded as rationalizing the belief
that P even in cases where it does not consist in a perceptual awareness that P. But for
psychological states, in contrast, there is in general no plausible candidate for such a
mediating, rationalizing state. In the case of sensation, for example, there is no evident
distinction to be drawn between a non-doxastic seeming that, for example, one has an itch
between one’s toes and one’s actually having the itch. And in the case of, say, the belief
that English Premier League football is financially unsound, there seems no sense at all to
be attached to the idea of a non-doxastic seeming that one has that belief —though one may
of course believe that one believes it.
Constitutive Views
My exploration of so-called Constitutive views was fashioned as a strategy for evading
the Language First or Thought First dilemma. In effect, the idea was to point out a
third, deflationary option: to side with Thought First, but without the invocation of
privileged access. On such a view we will essay to regard those of a subject’s self-
directed thoughts that potentially issue in avowal as indeed having properties that
underwrite the characteristic features of avowals, but not the properties associated with
privileged access, indeed not properties of epistemic provenance at all. The choice of
the term, “constitutive”, was perhaps not entirely felicitous. At any rate, it has
encouraged unnecessary criticisms. The idea was not at all that a toothache, say,
might be in part constituted by the judgment that one has a toothache, in a metaphysi-
cal sense like that in which the identity of the singletons of Cicero and Tully might be
conceived as in part constituted by the identity of Cicero and Tully—as if the fact of
the toothache itself had a judgment as a component. The suggestion was, rather, that
the distribution of truth-values among propositions concerning a subject’s mental states
is, a priori and necessarily, constrained by what she herself takes to be the truth about
her mental states, and that this is a conceptually basic or primitive point, rather than a
consequence of some independently accountable aspect of epistemic advantage which
selves enjoy in relation to their own states of mind. The master thought of Cartesian-
ism, epitomized by the metaphor of the inner theatre and the very etymology of the
term, “introspection”, is that there is indeed such an advantage. The project of any
Constitutive view is to do without that, or anything like it, while yet avoiding falling
back on the Language First alternative.
I think it is an open question what is the most robust formulation of a view of this
type. In my own earlier work I experimented with two.5 One was to invoke the model
5
A third form of Constitutivism is of course at the heart of Akeel Bilgrami’s 2006. I have no space to
compare and contrast the details here.
REPLIES 405
of response- or judgment-dependence explored in Truth and Objectivity and elsewhere.

According to that model, certain subject-matters—perhaps colour and secondary
qualities generally, perhaps certain kinds of value—have the feature that the extensions
of their signature concepts are determined, at least in part, by suitably constrained
sensory, affective, or doxastic responses of ours; in the case of judgment-dependence,
by the very (suitably constrained) judgments of ours about what the concepts in
question apply to. “Suitably constrained” means that the responses in question only
count as appropriately extension-determining if they are elicited under certain inde-
pendently specifiable optimal, or normal conditions. Thus, as a putatively prototypical
example, the surface of an object observed under optimal lighting conditions by an
attentive, normally visually functioning individual is red if and only if he is thereby
affected with the appropriate distinctive visual experience. To regard the extension of
red as response-dependent is to regard (a suitable specification of ) this rubric as an a
priori conceptual necessity, and one whose necessity is owing, moreover, not to the
fact that redness infallibly causes experiences of the relevant kind under the conditions
in question, but to the fact that there is nothing more to an object’s being red, under the
conditions in question, than its so appearing to a normally sighted subject—the well-
known form of contrast first outlined in Plato’s Euthyphro.
This broad idea can naturally be—and has been—implemented by a number of
specific templates differing in detail. To bring it to bear to deflate the pressure towards
invoking privileged access, a template involving some form of biconditional depen-
dence will clearly be required—so that there is a chance of mimicking both transpar-
ency and authority—and the relevant response will have to be that of judgment, since
the connection that we are trying to safeguard is between the psychological facts and
the subject’s impression of them. Because the required connection has to be highly
reliable, the relevant optimal conditions will have to be ones that normally obtain, or
can very easily be brought about. But we do not want complete reliability, since
failures in the direction of transparency may occur in, for example, conceptually
limited subjects, and failures in the direction of authority may occur as a result of
phenomena of priming, self-deception, and so on.
That is one constitutivist model. But there is a problem with any such proposal, first
pointed out by Paul Boghossian.6 Simply: in this case the relevant responses have, as
noted, to consist in certain kinds of judgment, and judgment is of course itself an
intentional psychological phenomenon. It is important to be clear why exactly this is a
problem, given that on some models of response-dependence,7 subjects’ responses
under optimal conditions are in any case thought of as determining only part of the
extension of a relevant concept. Such partiality is not the concern. On such models,
there is still scope for this partial determination to be from without, as it were, by facts of
6
See Boghossian 1989, pp. 544ff. Also the present volume at p. 34.
7
Those that work with what in Truth and Objectivity I termed Provisional Equations (or ‘provisoed
biconditionals’).
406 CRISPIN WRIGHT
a different kind. But on any judgment-dependence proposal about ordinary psycholo-

gy of the broad stripe presently considered, the responses—the determining facts—are,
qua judgments, further examples of facts of the very kind supposedly being determined.
The proposal is committed to the thought that the distribution of truth-values among
ascriptions of (first-order) judgment to a thinker is determined, in part, by his second-
order judgments. What determines the distribution of truth-values among ascriptions
to her of these second-order judgments? If we reapply the model, it looks as though we
now need to fall back on her third-order judgments . . . and the beckoning regress, and
associated spectre of ungroundedness, threatens to undermine the very dependence
postulated by the Constitutive view. But if the model is not to be reapplied, then the
theorist is building in an unexplained exception to the judgment-dependence proposal,
and now faces embarrassing questions about what fixes the distribution of truth-values
among ascriptions of the relevant second-order judgments, and about their first-
personal epistemology.
It was these considerations that moved me towards a different implementation of the
basic thought of the Constitutive view, one that promises to finesse Boghossian’s
observation. Donald Davidson is famous for the interpretationist proposal that the facts
about a subject’s mental states do not outrun whatever would be opined by an ideal
radical interpreter, operating under conditions of maximum information. This kind of
proposal, of course, has problems of its own but—assuming that they are not in the end
lethal—it looks to allow for a ready extension, not apparently envisaged by Davidson
himself, to address the present concerns. Simply: it can be taken to be a constitutive
principle of best psychological interpretation that the interpreter must maximally respect
the express self-conception of the interpretee (authority) and must minimize the extent
to which unacknowledged mental states are ascribed to the interpretee (transparency),
whilst otherwise making the best possible sense of what she says and does. In effect,
rather than being charged simply with making the most satisfactory overall sense of the
subject’s linguistic and non-linguistic behavior, the interpreter is required, in doing so,
to keep to a minimum the extent to which the mental states ascribed conflict with, or
exceed, those the subject herself is willing to acknowledge. Perhaps better: compliance
with this constraint is taken as a necessary condition of “making the best possible sense”
of what the subject says and does. The constraint is to operate holistically, with data of
apparent acknowledgement themselves regarded as defeasible and open to interpreta-
tion as part of the same exercise. That is enough to parry the immediate concern about
ungroundedness that confronted the judgment-dependence model: the determination
is achieved at the level of whole end-product, rather than step-by-step by individual
judgments.8 To be sure, there are still concerns about underdetermination, and about
8
To parry, but not perhaps to finally assuage: for have we not merely replaced the dependence of the
subject’s states on her second-order attitudes with a dependence of them on the attitudes of a third party—the
hypothetical best interpreter?
REPLIES 407
the inexplicit character of the methodology of best interpretation, but they were there
from the start.
What is the relation of any of this to Wittgenstein’s thought? There is of course no
basis in his texts for ascribing to him any explicit form of response-dependence or
interpretationist proposal. But a contention that I think he did hold—or which anyway
falls out very quickly from an application of the metaphilosophy of the Philosophical
Investigations to the present issues—is, though similarly deflationary in spirit, in tension
with the Thought First aspect of Constitutivism in any case. It is the contention that
our difficulties in this area result from the misguided quest for philosophical explanation of
what are in fact basic features of the “language game” of ascription of psychological
states to oneself and others. In Wright 1998 I called the position that develops out of
this idea the Default View. The Default View is a Language First view. It holds that the
authority, for example, that attaches to a subject’s self-conception of her own mental
states is an epiphenomenon of the authority invested, by the rules of the language
game, in what she has, sincerely and comprehendingly, to say about her own mental
states. This is not expressivism, though it is consistent with an expressivist embellish-
ment of certain types of avowals—which, to the gratification of commentators such as
Bar-On, Wittgenstein himself here and there famously seems to go in for.9 On the
Default View, the aspects of linguistic practice that (misguidedly) motivate the notion
of privileged access are not to be explained in terms of other aspects of their semantic
role, for example by their being “expressions”. They are not to be explained at all. It is
rather that by being primitively endowed with authority, they naturally acquire the
informational significance possessed by the canonical behavioral expressions of a sub-
ject’s mental states.
So let’s take an overview of the landscape of the ways of avoiding privileged access
just distinguished. Neither version of the Constitutive view—judgment-dependence
or interpretationist—is directly concerned with the phenomenon of avowal. On either
proposal, a metaphysical thesis is being advanced about the nature of the psychological:
that what is true about a subject’s psychological states is constitutively conditioned by
what she herself, in normal circumstances, takes to be the truth about them. These
proposals are no more concerned with our characteristic ways of speaking than are the
contentions that colour is response-dependent, or that linguistic meaning is interpret-
ation-dependent. However, either contention will entail, and in that sense explain, the
special status accorded to a subject’s avowals. Expressivism also gives an explanation of
that—but one which derives not from the nature of the subject-matter, but from the
9
The most famous passage is of course Philosophical Investigations }244: “ . . . how does a human being learn
the meaning of the names of sensations?—of the word ‘pain’ for example? Here is one possibility: words are
connected with the primitive, the natural, expressions of the sensation and used in their place. A child has hurt
himself and he cries; and then adults talk to him and teach him exclamations and, later, sentences. They teach
the child new pain-behavior.” But this is as close as Wittgenstein gets to an explicit expressivism, and there is
nothing in the passage or its surroundings to exclude the “new pain-behavior” consisting in a practice of
assertion.
408 CRISPIN WRIGHT
alleged distinctive semantic role of avowals. But on the Default View, as a special case of
Wittgenstein’s excoriation of the urge to philosophical explanation, the distinctive
characteristics of avowals really are grammatically primitive, as it were; there is no
explaining them and if any plausibility attaches to response-dependence, or interpreta-
tionist, or expressivist views, it is because they offer prismatic, somewhat distorted
reflections of this primitiveness.
Now, it is not completely clear whether, when Bar-On is critical of what she
understands by the Default View, and when Smith is critical of what he terms my
“deflationary” approach, either has in mind the distinctions just noted. Smith’s princi-
pal concern is with self-knowledge of meanings, but his complaint against my discus-
sions seems to be a generic concern, applicable to all the views above: it is that such
views do not really address the question of self-knowledge. If this complaint is addressed
against Constitutive views as characterized above, then it is misguided. There are
conceptions of knowledge to hand—those which, for example, centralize reliability,
or safety—which will readily allow those of a subject’s opinions which meet appropri-
ate conditions, determined by a conception of their subject-matter as judgment-
dependent, or as interpretation-dependent in the kind of fashion proposed, to rank
as knowledgeable. It is of course true that knowledge, so achieved, will not be a matter
of keeping track. But of course the whole point of such views is to repudiate the notion
that self-knowledge is, in the basic avowable case, a matter of such keeping track. The
objection is yet more tendentious if directed against the Default View, a major part of
whose point is, in effect, precisely to repudiate the idea of self-knowledge in the first
place; on this view, the special place occupied by avowals in the language game of
ascription of psychological states is exactly not a reflection of the knowledgeability of
their authors.10
For her part, Bar-On’s objections to the Default View are various. Sometimes it
seems that it is some kind of constitutive view that she has in mind, as when she
complains that it is not clear what exactly is being offered by way of analytical
reconstruction of the truth-conditions of ascriptions of mental states11 The objection
that seems to weigh most forcefully with her, however, is different and, I think,
important. Not all self-ascriptions of mental states exhibit the characteristic marks
of avowals. There are ascriptions of mental states to oneself made at times other
than that of the occurrence of the state concerned; and there are what Bar-On
calls “reportive” such ascriptions based on reasons, interpretation, and evidence.
No account, she charges, which simply privileges those of a subject’s
opinions about herself which are formed under good but normal conditions,
10
Thus Philosophical Investigations }246: “It can’t be said of me at all (except perhaps as a joke) that I know
that I am in pain. What is it supposed to mean—except perhaps that I am in pain?”
11
This volume, p. 177.
REPLIES 409
or privileges her self-conception in the interpretation of her by others, or simply

privileges what she has to say about herself, can provide the resources to explain this
distinction.
Is this a compelling objection? On the face of it, it would seem that it cannot be: that
if we had an adequate characterization of the data that set the problem—that is, of the
range of cases, whether in speech or in thought, that apparently manifest immediacy,
authority, and transparency—then this characterization, whatever it is, could simply
be built into the appropriate formulation of the relevant template for judgment-
dependence, or the appropriate constraint on interpretation, or the appropriate
specification of the rules of the language game of ascription of psychological states.
Only characterize the problem properly, it might be thought, and any view of
the kinds I have been distinguishing will have been given the resources to respond to
Bar-On’s concern.
This looks like the right direction for a reply to Bar-On’s objection to take. But a
caveat is called for. What exactly is an avowal? Is there a complete and correct
characterization of this class of utterances? The interest of the judgment-dependence
and interpretationist proposals is in a certain class of opinions, rather than their linguistic
vehicles. But it looks as though it would be difficult to say which exactly the relevant
opinions are—which are the opinions which, when formed under suitable conditions,
are to be conceived as partially determining the facts of a subject’s ordinary psychology,
or as fixing those aspects of her self-conception with which an optimal interpretation is
constrained to accord—other than by saying that they are those opinions whose
expressions constitute a proper avowal? But then we have to reckon with the following
possibility: that a serious attempt to give a comprehensive account of which are the
first-personal self-ascriptions that exhibit the characteristic features must, if it is to do
better than a mere illustrative list, advert to properties of these utterances which favor
an account that is at odds with the broadly deflationary kinds of proposals that I have
been reviewing. It might be charged, for example, by someone with vestigial Cartesian
sympathies that there is ultimately no characterizing the class of first-personal self-
ascriptions that may be properly avowed other than in epistemological terms: an
utterance of this kind can be properly avowed just when the subject stands in the
appropriate distinctively first-personal epistemic relation to the state of affairs it depicts.
At the other extreme, and more congenially to Bar-On, it might turn out that there is
no comprehensively characterizing the relevant class save by reference to the notion of
an utterance that expresses, according to a preferred account of that notion, the relevant
mental state; that the characteristic marks of avowals are grounded in, and their extent
determined by, the illocutionary act of expression. Accordingly, it does seem fair to say
that a proponent of either form of deflationary view does have a pressing obligation to
show that a characterization is possible of the target range of utterances, or opinions,
which avoids letting in the opposition by the back door, as it were. I would accept the
criticism that, at the time of writing, this obligation has not really been discharged.
410 CRISPIN WRIGHT
Expressivism
There has been no more determined or systematic a supporter of expressivism as a
response to the problems of self-knowledge than Bar-On. Her contribution to this
volume outlines key features of her major elaboration and defence of the view offered
in her 2004 and elsewhere. The key move is one that I myself argued for in Wright
1998 and I agree with her that it is a vital part of any viable expressivism in this region.
This is to jettison altogether the idea, originally the cardinal point of, for example,
ethical expressivism, of any tension between the allegedly expressive role of avowals
and the acknowledgement of a domain of ordinary psychological truth and falsity. The
thesis has to be not that there is no such thing as genuinely true or false psychological
assertion, including assertion about one’s own case, but that those aspects of our
discourse which give the impression of being an especially authoritative genre of
such assertion are actually not assertoric at all, serving rather to, in a relevant technical
sense, give expression to the beliefs, or feelings, that, prima facia, they seem to state. In
those parts of her chapter, accordingly, where Bar-On is arguing that a competitive
expressivism about avowals must distance itself from the non-factualism of traditional
expressivist views of, for example, moral value, she is, at least as far as this reader is
concerned, preaching to the converted.
That agreed, I am however doubtful about Bar-On’s suggestion that expressivism
about avowals, so liberated from any non-factualist implication, can then comfortably
coexist with a minimalist conception of truth and truth-aptitude for ordinary psycho-
logical discourse in general. In Wright 2002 I argued in some detail that minimalism
about ordinary intentional psychology is actually a dubiously stable position dialecti-
cally, since it appears to entail a corresponding minimalism about the very distinction
between merely minimally truth-apt discourses and those that discharge a more
robustly representational role, and that it is the hallmark of the former that they fail
to exert cognitive command, that is, that there is no a priori guarantee the disputes
within them need involve anything worth regarding as a cognitive shortcoming. The
argument I gave is quite complex, and is no doubt open to challenge. But unless it is
addressed, the bottom line threatens that to regard ordinary psychological discourse as,
though truth-apt, merely minimally so is a commitment to regarding that classification
of it as rationally unforced. That is not a contradiction, but it does raise the question, of
one who takes this view, why exactly they are doing so—something close to, though
not quite, a kind of Moorean paradox.
Much better, then, if expressivism in the present context is yoked to a realist view of
ordinary psychological claims—as indeed, in Bar-On’s own exposition, it is. I would
not still wish, without attempting to respond in detail to the development and defence
in Bar-On 2004, to stick to my sometime description of this type of view as a “dead
duck”. But I continue to have a number of reservations that go to the basic structure of
REPLIES 411
the view and are independent of its detail, and which strike me as serious enough to
make it unlikely that any version of it can be fit for purpose. Bar-On herself has
fashioned replies to some of these objections, which she refers to but does not enlarge
on in her present chapter.12 So I’ll provide merely the briefest outline of them before
moving to a new point.
There is much to be said about what notion of expression best serves the expressivist’s
theoretical purposes, and Bar-On’s discussions contribute considerably to the clarifica-
tion of this issue. But for the purposes of any expressivist realism equipped to dissolve the
problem of self-knowledge in the kind of way anticipated, the essential point has to be
the following. Selves must turn out to stand in no form of epistemically superior
relationship to their own mental states that would ground the distinctive features—
those suggestive of privileged access—of avowals. Those features have to be explained by
showing how they flow from the expressive function of avowals. So expression has to be
understood in a way that makes that possible. The idea of any distinctively privileged
first-person epistemology of the psychological has to be undercut; that is the whole point of
the view. This, of course, doesn’t debar one who regards avowals has having a straight-
forwardly assertoric role from acknowledging that they also have an expressive func-
tion—just as a realist about moral value is not debarred from allowing, if it is true, that
moral assertions characteristically also serve to express feelings of moral approval and
disapproval. But that acknowledgement does not amount to expressivism in the sense
here of interest. The crucial move must be to utilize the expressive character of avowals
to obviate any impression that their distinctive features stem from some kind of
recognized first-personal epistemic superiority—in effect, to show the problem of
self-knowledge, conceived as I have outlined it, is an illusion.
I am afraid I remain utterly sceptical that this can be done. In earlier work, besides
emphasizing that the distinctive attributes of authority, immediacy, and transparency
seem to belong as much to the unvoiced impressions a subject has of her own mental
states as to her expressions of them, I noted how the idea that all a thinker might have
to go on, as far as her strictly epistemic situation is concerned, in her assessment of her
own mental states are the public expressions that are available to anyone, leads to
absurdity in certain kinds of scenario, where the outward behavior manifesting a
certain psychological state—perhaps acute social embarrassment—is actually more
salient to onlookers than to the subject who is trying to conceal it.13 One, as it seems
to me, rather desperate response to the first of these considerations is to attempt to
regard token thoughts concerning one’s own mental state as themselves a kind of
expression—a kind of interior avowal, if you will.14 But such a move, it should be
evident, even if allowed, makes no progress at all with the second problem. For one
thing, it seems that my awareness of a toothache is hardly likely to be compromised if
12
This volume, p. 176, fn. 40.
13
Thus the torture victim case in Wright 1998, p. 37.
14
See this volume at p. 185 and Bar-On 2004, p. 23.
412 CRISPIN WRIGHT
I fail to token any relevant thought—say, “This tooth hurts like Hell!”—and thus
simply don’t express it at all, even inwardly. And even if I do happen to ‘inwardly
express’ my pain by such a thought, and so have that additional datum, that will hardly
restore my epistemic situation with respect to the pain to one of superiority over
onlookers unless I am aware of the thought—and aware of it in what appears to be a
distinctively first-personal way, since the case is, by hypothesis, one where I don’t
articulate it publicly.
Let me, though, move to the advertised new point. Behavior that is prototypically
expressive of a mental state, in the kind of way Wittgenstein was thinking of —
grimacing and hopping about while clutching a stubbed toe, scratching at a mosquito
bite, or a child’s yowling after it has dropped its ice-cream on the path—is involuntary.
More sophisticated kinds of expressive behavior—for instance, swearing, or hurling
down the newspaper when reading the football results, or grabbing the bars of one’s
cell and making as if to shake them loose—are often, by contrast, intentional: subject to
voluntary control and in some sense selected. However such behavior, although
intentional, does not, insofar as it is expressive of the reactions or emotions portrayed,
amount to action supported by a structure of reasons—beliefs, desires, and intentions—that
coordinate into a practical syllogism. In that sense, such actions are not properly
classified as rational actions, though they are rationally intelligible acts. For convenience,
let’s term the three categories involuntary, intentional, and purposive respectively. Now in
principle, a vocalization with a truth-evaluable content can belong to any of the
categories. Wittgenstein’s idea is at least coherent: there is no reason a priori why an
episode of linguistic behavior, of a kind acquired as a replacement for the natural
expressions of a pain, would have to be intentional, let alone purposive. The point that
needs to be noted, however, is that the mode of evidential significance possessed by such a
performance must vary as a function of which category it is taken to belong to. It is of
course possible to simulate pain behavior, or voluntarily “let it out”, or do so as part of
an attempt at communication with a foreign doctor. But part of the irresistible
evidential force of pain behavior, at its most convincing, is precisely the consideration
that it is conceived as involuntary, as a result of the pain’s surging to the surface, as it
were, so the question of insincerity does not arise. Avowals, however, cannot in
general acquire their evidential force on that model, even when the mental state in
question does have natural untutored forms of behavioral expression. For we can take it
as a datum that avowals will only exceptionally be involuntary. Usually they will be at
least intentional, even if not purposive. But most often they will be regular speech acts,
underwritten by a practical syllogistic structure of reasons.
What follows? Granting that much is still consistent, no doubt, with there being
some theoretically pointful notion of expression such that a deliberate, purposive
avowal, underwritten by a structure of practical reasons, may still count as an expression
of the relevant mental state—that state, namely, that satisfies the truth-conditional
content which, Bar-On’s kind of expressivism grants, is associated with the avowal.
What seems clear, though, is that this notion of expression, if such there be, won’t be
REPLIES 413
able to deliver the intended upshot that the impression of epistemic superiority
conveyed by misassimilating an avowal to an assertoric report should turn out as
illusory. To appreciate this we have merely to notice that among the subject’s set of
reasons for a sincere purposive avowal, confidently made, will have to be an awareness
of the relevant psychological state. Your General Practitioner asks you, “Point out
where it hurts”. Clutching your lower right abdomen, you say, “It hurts here,
especially when I poke it”. That, by any usual standards, is an avowal. We can suppose
it is sincerely made. What are the reasons that support it? Well, they’d better include
the desire that your GP has an accurate impression of your symptoms, your consequent
intention not to mislead him by what you say, and your awareness of when and where it
hurts. Maybe, I say again, there is some theoretically worthwhile sense in which your
response to the doctor’s question expresses your abdominal pain; maybe there isn’t. But
if there is, then whatever it is, it’s not available to defuse the idea of privileged access.
Rather, in order to understand your avowal as rationally performed, we need to
presume that you have a kind of reliable awareness of the character of your pain that
is denied to the doctor in advance of your expression of it in answer to his question.
The point is good for any performance, linguistic or otherwise, that is naturally
conceived as giving deliberate expression to a state of mind. We can, without loss of
generality, take any such expression to be the first occasion of giving vent to the state of
mind in question. But then the relevant practical syllogism that rationalizes the subject’s
performance will have to include a belief about her state of mind as a precursor to its
receiving outward expression. In short: to make practical rational sense of deliberate
avowals, or any other form of deliberate expression of our mental states, involves
adverting to beliefs of the subject about our own mental states which need to be
regarded as in good standing if she is to be regarded as acting well, rationally speaking,
but which cannot coherently be regarded as grounded in ways that are appreciable in
principle by any observer of her performance.
Knowing One’s Own Meanings

Barry Smith’s interesting and thoughtful chapter is given to the special case of self-
knowledge of meaning. His chapter is concerned with what he calls the Reconciliation
Problem, that is, the problem of reconciling self-knowledge of meanings, conceived as
exhibiting the phenomena of privileged access, with the essential publicity of mean-
ing—the idea not merely that anything one can mean by an expression is available to
be meant by anyone else, but likewise that the fact of one’s meaning it is publicly
available too. In short, the problem is that of reconciling the publicity of what one
means with the availability of one’s meanings to oneself in the manner distinctive of
intentional self-knowledge.
A crucial question, of course, is where the notion of the publicity of meaning springs
from. It’s an axiomatic common theme in Wittgenstein, Quine, Dummett, and David-
son. For Quine, and Davidson, the motivation is, au fond, metaphysical-naturalist:
414 CRISPIN WRIGHT
there is, in their view, nothing for one’s meaning something in particular by a particular
expression to consist in if the expression’s having that meaning does not make some
distinctive, identifying impact on one’s behavior. It is different for Wittgenstein, and
following him Dummett: for these philosophers, publicity is a consequence of the idea
that there is nothing to meaning save what is understood, and that understanding is a
kind of knowledge-how, rather than knowledge-that—a complex of operational skills
associated with the expression in question. Operational skills, in their very nature, have to
be manifestable.
But these philosophers take quite different views of the other half of the reconcilia-
tion problem. Quine and, I believe, Davidson would simply reject any notion of
privileged self-knowledge of meaning; all there is is the homophonic knowledge
generated by disquotation into one’s own idiolect. Wittgenstein was exercised by
the phenomenology of the apparent transparency of one’s own meanings—for in-
stance, in the phenomenon of “grasping in a flash”.15 But one senses that he regarded
this as a puzzling, easily misunderstood phenomenon which gets in the way of a correct
philosophical take on the notions of meaning and understanding, rather than a datum
which any satisfactory account has somehow to integrate. Dummett, by contrast,
regards what he calls the transparency of meaning to the thinker—more specifically,
the transparency of Fregean senses—as an indispensable component of any satisfactory
account of what it is to understand a language. On a view like Quine’s, and Davidson’s,
I take it, there is no real reconciliation problem in the first place. The self-knowledge
component of the problem is a chimera. It is on a Dummetian view that the problem is
indeed acute, since on the one hand it is being asserted that meanings are by their very
nature public, that understanding is essentially practical knowledge, and on the other
that they are available to privileged access. There appears, on the face of it, no way that
these claims can be made to cohere.
Smith’s ambitious essay sets itself to provide a solution to this problem. His key
move, however, is to resist the “exteriorization” of meaning that is the common theme
running through Quine, Dummett, Wittgenstein, and Davidson. This leaves the status
of the publicity of meaning somewhat in shadow. If meaning is indeed an inner
psychological phenomenon, then the idea that it is also essentially public cannot be
motivated by the kinds of considerations that moved those philosophers. Smith does
not explain why he nevertheless accepts it, so that a reconciliation problem still arises.
But he believes he can safeguard the publicity of meaning, even starting from a
conception of meanings as interior.
Smith’s leading idea is to apply to the purpose a notion of epistemic entitlement along
the lines that I have myself tried to defend in various recent papers.16 A key point
for his project is one often stressed by John McDowell, that the phenomenology
of understanding the speech of another involves immediacy—we does not, save
15
See e.g. Philosophical Investigations }}138–99, 191 and 197.
16
I pick up the discussion of this notion in Part IV of my Replies in this volume..
REPLIES 415
exceptionally, actively interpret the speech of another, but simply get her meaning by
listening. Exactly what philosophical force is carried by this observation has always
seemed somewhat moot to me. But Smith’s proposal is that it is this: that in parsing the
speech of another, one simply superimposes one’s own idiolect onto her words,
spontaneously understanding them as if produced by oneself. The crucial suggestion
is then that if, as he takes the case to be, we are each of us defeasibly entitled to the
general claim that anyone who uses an expression that is currency in one’s idiolect
means by it what one means oneself, then I can fetch up with knowledge of what you
mean by a particular expression—to wit, exactly the same as I mean, which, our
starting point was, I already know.
Smith’s invocation of the notion of epistemic entitlement is somewhat briskly done,
but I will not here delve into the question whether the case can be developed in such as
way as to bear a clear structural analogy to other cases where, as I have argued,
entitlements can be used to underwrite claims to knowledge. It is worth stressing,
though, that in my own view, it is only claims to knowledge that can be underwritten
in this way. The upshot of Smith’s argument, if successful, ought to be, not that others’
meanings are indeed available to us, but only that we are in position rationally to claim
that they are. Whether that is a sufficiently robust sense of “publicity” to address any
remaining reconciliation problem will depend on what Smith doesn’t supply, namely a
motivation for continuing to regard meanings as publicly available, once the “exter-
iorizing” move is rejected.
There may, however, seem to be a residual problem about the ability of Smith’s kind
of account to secure publicity in any worthwhile sense. If all goes well, each of us winds
up entitled to claim to know that others share our respective idiolectic meanings. So
your meanings, so I am entitled to claim, are available to me. But publicity had better
involve that availability is reciprocal—that I should also be entitled to claim to know
that my meanings are available to you. How does that work?
Here is a line that Smith might try. Suppose I am indeed entitled to claim to know
that my meanings are the ones that you express when you use the same vocabulary.
You too have the same entitlement. So you are entitled to claim to know that your
meanings are shared by me. But these meanings, I can reflect, are mine—or so I am
entitled to claim. So the meanings that you are entitled to claim to know are shared
with others, including myself, will be—or so I should think—the very ones that
I mean. So hey presto, I am entitled to claim to know that my meanings are available
to you.
The main assumption of Smith’s project is that we do indeed have privileged access
to meaning. This has, of course, to relate to knowledge of idiolectic meaning—no one
thinks they have privileged access to the meanings of Latin, or Dutch sentences. Smith,
however, nowhere says what form he conceives self-knowledge of idiolectic meaning
to take. Presumably it has to be propositional knowledge: the characteristic marks of
immediacy, transparency, and authority engage opinions that so-and-so is the case. But
then, what are the propositions in question like? It is at this point that one may begin to
416 CRISPIN WRIGHT
have some doubt about the reality of the very phenomenon—of first-personal
privileged access to idiolectic meanings—which Smith sets himself to reconcile with
publicity. The knowledge in question has, presumably, to be metalinguistic—the
propositions concerned have to be propositions about one’s idiolect, characterizing
the meanings of its expressions, appropriately mentioned. But then the following point
gives concern: there is no essential role for such knowledge in rationalizing one’s uses
of the idiolect. Rationalizing the use of token expressions in the idiolect will require
citation of various germane beliefs, desires, and intentions, and the citation of states of
understanding. But understanding is not metalinguistic knowledge: to understand a
language is not to have knowledge about it but to be able to use it to say what, modulo
one’s other reasons, it is rational for one to want to say. The process of deliberation that
leads up to the decision to say something need not, on pain of incompleteness, involve
any semantic beliefs—any more than it needs syntactic ones. Of course the point
remains that a normal adult speaker will typically have many beliefs of a semantic
character. But they will be beliefs about the proper use of English, or French. To make
rational sense of my use of a particular French word on a particular occasion, it may
well be necessary to ascribe to me a metalinguistic belief about its meaning; and the
reasons leading to my choice of the word may well involve such a belief. But the
metalanguage in which these beliefs are formulated, if it is to be the language in which
I articulate my reasons, had better be my idiolect. What is quite obscure is what role
there is for metalinguistic beliefs in my idiolect about my idiolect. But such, it seems,
would have to be the character of the beliefs that Smith regards as setting the problem
of reconciliation.
The first-personal knowledge that Smith’s chapter is premised upon thus begins to
seem fugitive on close inspection. Knowledge of one’s intentional states in general, of
course, does involve knowledge of their content. But this is knowledge of mental
content, not linguistic content. My realization that I was just now mistakenly thinking
that tomorrow is Sunday does not come to me idiolectically garbed, as it were, and
awaiting identification of its content via an application of my knowledge of what
thought is expressed by “Tomorrow is Sunday” in my idiolect. Knowledge of the
contents of one’s thoughts does not proceed through knowledge about one’s idiolect.
But if the latter is needed to explain neither how one expresses oneself, idiolectically,
nor how one knows what one thinks, what is it for? And why suppose we have it?
If this line of thought is correct, then we must revisit the somewhat negative
assessment, reached at the end of Part I of these replies, of Paul Horwich’s idea of
meanings as constituted in regularities of use, formulable in terms of ideal laws of which
ordinary speakers may be quite unaware. It does seem plausible that the competent,
purposive use of language must draw on some form of knowledge of meanings that is
available to a subject at an intentional level—that there is a sense in which the subject
“chooses her words”, in the light of her beliefs and goals, and no corresponding sense
in which, for example, she “chooses a syntax” to frame the way the expresses herself.
The syntax shapes her linguistic competence, but is not drawn upon, as personal-level
REPLIES 417
information, in its routine exercise. I canvassed it as an objection to Horwich that

meaning seems to be different in just this respect: that if we are to explain a linguistic act
as rational, we will need to cite the subject’s understanding of it, alongside her other
beliefs and desires, in a suitable practical syllogism, and that Horwich’s account seems
therefore to err in placing meanings beyond the sphere of ordinary personal knowl-
edge. But the upshot of the foregoing consideration of Smith’s project is that we have
no plausible model of what knowledge of meaning, as personal self-knowledge, might
consist in. The model that thrusts itself forward is essentially metalinguistic, and it
cannot be knowledge of that character that the rational use of language essentially
draws upon.
So is Smith’s project misconceived, and is the objection to Horwich disarmed?
Maybe. The question, I think, is whether some other, personal-level model of
knowledge of meaning can be given to do justice to the idea that knowledge of
one’s own meanings somehow enters into the rationalization of intentional linguistic
activity, at a personal level but in, as it were, quotation-free form, without taking the
shape of propositional metalinguistic knowledge. If any such idea can be made
coherent and plausible, there will be a question whether such knowledge comes within
the scope of ordinary psychological self-knowledge of the kinds relevant here. If it
does, the objection to Horwich’s approach will stand, and Smith’s issue about recon-
ciliation with publicity will arise.
Bibliography
Bar-On, D. 2004 Speaking My Mind, Oxford, Clarendon Press.
Bilgrami, A. 2006 Self-Knowledge and Resentment, Cambridge, Mass., Harvard University Press.
Boghossian, P. A. 1989 “The Rule-Following Considerations”, Mind 98/392, pp. 507–49.
McDowell, J. 1998 “Response to Crispin Wright”, in C. Wright, B. C. Smith, and
Wright, C. 1998 “Self-Knowledge: The Wittgensteinian Legacy”, in C. Wright, B. C. Smith, and
C. Macdonald (eds.) Knowing Our Own Minds, Oxford, Oxford University Press, pp. 15–45.
Wright, C. 2002 “What could Anti-Realism about Ordinary Psychology Possibly Be?”, The
Replies Part III: Truth, Objectivity,
Realism, and Relativism
Blackburn, Shapiro, and Rovane
Background
The metaphysical debates about realism are some of the oldest in philosophy, and some
of the most intractable. And although the specific emphases and titles of realism’s
many antagonists over the years—the idealist, subjectivist, non-factualist, expressivist,
non-cognitivist, constructivist, projectivist, instrumentalist, nominalist, quasi-realist,
irrealist . . . —have varied with the period and the subject-matter, the analogies are
striking enough to have encouraged Michael Dummett to hope that there might be a
single underlying issue whose proper clarification could serve both to fix the right rules
of engagement wherever such a debate is joined and to determine winners and losers.
Dummett’s proposal, famously, was that this one single, underlying issue—the “Key to
All Mythologies”—concerned the proper model of meaning, that is, the form that
should be assumed by a theory of meaning for discourse concerning the subject-matter
concerned, with the prototype for such a debate supplied by the opposition between
classical realism in the philosophy of mathematics and the constructivism of the
Mathematical Intuitionists, which Dummett reformed as a contest between truth-
conditional and proof-conditional accounts of mathematical propositional content.
As mentioned elsewhere in these Replies, my own first research interest in philoso-
phy was the issue between Platonism and constructivism about pure mathematics. And
on first encounter, Dummett’s proposal had all the resonance of a major breakthrough.
Indeed, the experience of first reading Dummett’s early papers on these issues was, as
the late David Pears once remarked, akin to having someone remember to turn on a
light in a darkened cellar in which one had been groping around looking for the way
out. Dummett’s model thus became for a time my major philosophical preoccupation
and many of my early papers focused on the sympathetic criticism and development
of his marquee ideas about ‘anti-realism’, the connections with Wittgenstein’s later
ideas about meaning, assertibility-conditions versus truth-conditions, and the mean-
ing-theory based revision of classical logic.
REPLIES 419
But many problems emerged with this approach. Although, as I still think, the
Dummettian anti-realist had the better, albeit by a close decision on points, of the
critical general debate about the global adequacy of semantic theory informed by
classical (bivalent, or evidence-transcendent) truth-conditions, no general systematic
assertibility-conditional approach was ever developed to oppose the classical style of
semantic theory.1 And there seemed good reason for that. For one thing, the condi-
tions of epistemically proper assertion of empirical statements generally are contextual:
they depend on variable features of a thinker’s collateral beliefs and information. So the
meaning of a statement has to be thought of as just one parameter whose values feed
into the determination of its assertibility in a context, rather than as constituted in its
assertibility-conditions themselves. For another, the connection between statement
meaning and truth-conditions seems, at one level, to be merely the stuff of platitude: to
know the meaning of a statement S is to know what it says—to know, for some
content P, that S says that P. But if S says that P, then if and only if it is the case that
P will S be true. So a theory that determines what S says will, willy nilly, settle a truth-
condition for S. This simple manipulation of truisms seemed to imply that any general
model of statement meaning could not avoid truth-conditionality—there seemed to be
no other option if a systematic theory was to be possible at all.
Now, the last line of thought is something that contemporary deflationists about
truth, like Hartry Field and Paul Horwich, and assertibilists like Robert Brandom,
perceive themselves as needing to resist, and of course alternative—inferentialist,
conceptual role, and use-theoretic—conceptions of meaning are developed and sup-
ported in the writings of these philosophers (though the charge of asystematicity does,
I think, remain largely unaddressed). My own eventual reaction to it, though, was
different. Broadly, it was that Dummett had been right to discern an implication of
realism, intuitively understood, in the idea that the statements of a discourse might
allow of evidence-transcendent truth-values, but mistaken to suppose that the discard-
ing of this idea should enforce a departure from truth-conditional semantics. Rather,
the question should be: granted that meaning should everywhere be construed truth-
conditionally, what specific conception of truth should the statements in a given contested
region of discourse be regarded as subject to? The Intuitionists, in particular, should be
interpreted not as supporting a local case of a generalized assertibility-conditional
semantics, but a non-classical conception of mathematical truth.
There were other reasons for dissatisfaction with the Dummettian paradigm. Dum-
mett himself always wrote as if anti-realist views might be merely locally correct, and it
seemed essential that a good account of the debates should safeguard this possibility.
But the arguments he proposed for ‘dethroning’ classical truth-conditional semantics—
especially, the considerations about the acquisition and manifestation of a grasp of
evidence-transcendent truth-conditions—seemed globally effective or not at all. Of
1
As Timothy Williamson has never tired of complaining. See e.g. his notorious Williamson 2006.
420 CRISPIN WRIGHT
course, that still left open the possibility of other more local considerations that might
justify selectively opposing the idea that truth could intelligibly outrun all (in principle
available) evidence—special considerations available for judgments of humour, for
example, or colour. But, even acknowledging that, the preoccupation with evidence-
transcendent truth (or even more committally, as Dummett himself preferred, Bivalence)
seemed foreign to many of the targeted debates in any case: those between realists and
instrumentalists about science, for example, or realists and projectivists about ethics, or
modality. The unreconstructed idea that in a given region of discourse, we make judg-
ments that purport to represent and are answerable to facts that are ‘not of our making’
didn’t seem to depend on the conviction that those facts may transcend the evidence. And
the various anti-realists seemed, very often, to be targeting not that conviction but the idea
that the discourse in question was in the business of hitting off facts at all.
The programme of Truth and Objectivity grew out of these concerns. My overall aim
was to develop a framework to build on Dummett’s proposal but accommodate the
fact that a number of different cruces seemed to be discernible in the various discourse-
local contests about realism and to integrate the debates more tightly with the
metaphysics of truth. One key element in the new project was a conviction that the
traditional debates about truth—the contests between correspondence, coherence,
pragmatism, deflationism, and all that—had suffered from a conflation of concept
with property. The deflationists, it seemed to me, had been more or less right about
the thinness of the concept of truth—though wrong, perhaps, in the suggestion that it is
quite simple enough to allow of all aspects of its employment to be fully captured by the
Disquotational Scheme. But they were mistaken in moving from that to the claim that
truth, as a property of sentences, or propositional contents, could have no metaphysical
substance. Rather, my idea—the key idea in what has come to be known as alethic
pluralism—was that truth could have variable substance in different regions of discourse,
with any property a candidate to constitute truth in a given region which, when
constrained by additional principles specific to it, provided a model of the distinctive
behavior of the concept. And the differences among these local properties might then
serve to interpret the realism debates: in effect, what would be at issue would be the
character of the truth-property actually implicit in the practice of a certain discourse or—
when the debates take a potentially revisionary cast—the truth-property appropriate to
that discourse.
This proposal reflected a degree of attraction to deflationism—though an official
deflationist would naturally scorn the indulgence not just in a substantial property of
truth but a potential multiplicity of them! Another thinker who has been unwilling to
sign up to deflationism proper—to “minimalism” as, following Paul Horwich, he likes
to style it—but whose very influential writings in the area also reveal such a qualified
attraction is of course Simon Blackburn. But Blackburn has little sympathy for the idea
of pluralism about truth, preferring instead to look for ‘metaphysical contour’ in
distinctions between different kinds of truth-apt propositional contents. It may be of
some interest to dwell a little on the issues here and the options available.
REPLIES 421
Deflationism and Expressivism

Deflationism about truth endorses each of three claims:
(1) First, the concept of truth is exhaustively characterized by its role in the
Disquotational Scheme, ‘P’ is true if and only if P.2
(2) Second, there is no property of truth save in the ‘abundant’ sense in which
any significant grammatical predicate may be associated with a property whose
nature is fully transparent in the satisfaction-conditions of that predicate, so that no
interesting question can arise about the nature of the property, or what possession
of it consists in.
(3) Third, truth-aptitude travels in tandem with what Blackburn has nicely called
“propositional surface”: any properly intelligible discourse deals in truth-apt sen-
tences if it provides the resources for their embedding in the familiar ways under,
inter alia, negation, the conditional, modal operators, and propositional attitude
constructions.
The minimalism of Truth and Objectivity—contrast the ‘minimalism’ of Paul Hor-
wich, which is simply deflationism by another name—endorsed the third of these
claims, qualified the first and, en route to the alethic pluralism proposed in that book,
flatly rejected the second.3 But the early prototype of expressivism—about, say, ethics
or aesthetics—advanced in the writings of some of the logical positivists and their
immediate successors, but traceable perhaps to the Tractatus, rejected all three claims,
denying that the apparent ‘statements’ made in the discourses in question were so much
as truth-apt at all, and dismissing the appearance to the contrary as grammatical illusion.
Notwithstanding that such sentences indeed displayed “propositional surface”, their
actual use was not to state anything true or false, even deflationarily true or false, but
merely to give vent to attitude and feeling.
As is long familiar, this kind of proposal runs headlong into a serious problem: what
can it say about the role of ethical, or other allegedly purely ‘expressive’ sentences,
when embedded—and unendorsed—in the wider range of kinds of constructions that
go with propositional surface, most notoriously in the context of reasoning from and to
statements in which they occur under negation or as the antecedent of a conditional or
as mere suppositions? Blackburn’s landmark discussion in Spreading the Word4 was
2
Or in Horwich’s version, the Equivalence Schema for propositions,
It is true that P iff P.
3
The argument for doing so, in brief, was that deflationism can consistently countenance no norm over
the acceptance of sentences that qualify by its lights as truth-apt other than warranted assertibility; but that the
Disquotational Scheme itself, while requiring that we think of truth as normative, also enforces a potential
extensional contrast between ‘true’ and ‘warrantedly assertible’. It’s my impression that this line of argument,
developed in Wright 1992, Chapter 1, has generally been found persuasive by non-deflationists and neutrals,
but has had little impact on the thinking of the committed, like Horwich and Hartry Field.
4
Blackburn 1984.
422 CRISPIN WRIGHT
I think one of the very first attempts to take this ‘Frege–Geach’ problem head-on and
try to show how an expressivism worth the title could handle it.5 But the ‘quasi-
realism’ he offered there importantly qualified the original expressivist stance, since it
was part of its project that even the application of the truth-predicate to ethical sentences
might have a kind of earned intelligibility.
Note that the intelligibility had to be earned: it would need constructive philosophy
to justify the display of propositional surface by a discourse that was not in the business
of representing facts. The primary use of the truth-predicate still remained its annexure
to linguistic representation. So the predicability of ‘true’ was not, in Spreading the Word,
the relatively easy, promiscuous thing that deflationism, and Truth and Objectivity, took
it to be. Still, the concession was implicitly made that there could be legitimately
propositionally surfaced discourses in which the function of the truth-predicate—or a
predicate behaving just like it—was detached from the marking of correspondence
between thought and reality: a concession that marked at least a small step in the alethic
pluralist direction, and away from the original expressivism of the positivists.
So, the expressivism of Spreading the Word affirmed both that the truth-predicate
sometimes marks a substantial correspondence property (contrary to the second defla-
tionary thesis) and hence, since the Disquotational Scheme is powerless to explain the
distinction between such uses of ‘true’ and others, that there is more to truth than is
captured by that Scheme (contrary to the first deflationary thesis). Finally, aptitude for
characterization as ‘true’ or ‘false’ was allowed to spread across propositionally surfaced
discourses only to the extent that their propositional surface could somehow be
legitimated.6 All three characteristic theses of deflationism were still repudiated.7
What is the relation of Blackburn’s current expressivist views, a quarter of a century
later, to those of Spreading the Word? His very welcome present essay starts out with
what he confesses he feels as a dilemma. The dilemma is that, on the one hand,
expressivism is all about making a distinction between those regions of discourse
which are in the business of genuine representation and whose truth-predicate is
accordingly to be understood in correspondence terms, and those which are not—
5
The general reception of Blackburn’s book was that this effort was not successful. Blackburn himself in
effect acknowledged this when he revisited the issue in Blackburn 1988, which offers significantly different
proposals. But these too ran into technical snags. See Hale 1993.
6
One radical such form of legitimation would be the provision of a systematic reductive paraphrase
whereby the appearance of propositional surface would be eliminated, so that one might, e.g., attempt a
systematic reconstrual of moral discourse as a whole into a language of imperatives, say. But an informal
philosophical account of how talking ‘as if ’ there were moral facts could achieve certain characteristic
purposes of moral discourse—in effect, a kind of fictionalist account—would presumably also be a propos.
I think it is fair to say that the project of quasi-realism, in the hands of its author, has shifted over the years
from the former paradigm to the latter.
7
The position of Truth and Objectivity was thus, on these issues, intermediate between that of hard-line
deflationism and Blackburn’s original expressivism, endorsing the promiscuity of truth-aptitude but agreeing
with Blackburn that the predicate, “true”, functions differently in different regions of discourse, sometimes
denoting a property in the Church of Correspondence, sometimes belonging only to the trappings of
propositional surface.
REPLIES 423
and any such distinction is beyond the pale for deflationism. But on the other hand, he
professes himself drawn to the resource of deflationism which is—he suggests—only
too convenient for the expressivist when he is challenged to explain—or explain
away—the nature of the truth-values and truth-makers that seem to be called for
merely by the propositional surface of moral discourse or other expressivist targets. On
a deflationary understanding of truth, and truth-aptitude, the answer to this challenge
can be very brisk: nothing substantial—non-disquotational—need be said. By contrast,
the position of Spreading the Word involved a commitment to give a non-deflationary,
philosophically constructive response.
Now, there is, of course, a philosophical dilemma here only if one has a lingering shine
for the notion that the deflationary answer might after all be fit for the purpose. If so,
and if expressivism too incorporates insight, then there has to be a way of reconceiving
the expressivist proposal so as to finesse the clashes reviewed above between the two
views. Is there a possible revision in the conception of the obligations of expressivism,
properly consonant with its central ideas, which would allow an expressivist to avail
himself of the promiscuity of truth and truth-aptitude, shared by hard-line deflationism
and Truth and Objectivity? Can an expressivist be a deflationist after all?
The root problem for the combination is that expressivism wants to make a
distinction, between representational and merely expressive discourses, which the
promiscuity of truth and truth-aptitude generated by deflationism seems certain to
obliterate. Obviously, there can be only one possible general form of accommodation:
to make the relevant distinction in some way that does not commit one to thinking of
truth everywhere as representation. The alethic pluralism of Truth and Objectivity is
exactly one way of doing that, whereby the contrast between genuinely representa-
tional and expressive discourses is carried not by a difference between truth-aptitude
and the lack of it but by the character of the truth-properties respectively associated
with them. But that, of course, is not an accommodation that a real deflationist can
make. And Blackburn does not want to be an alethic pluralist. So the interesting
question is: is there another way, consistent with the operation of a deflationary truth-
predicate across the board?
The greater part of Blackburn’s present essay is taken up with criticism of the
ramification of deflationary conceptions of truth into deflationary conceptions of
semantic notions generally that one finds in the developed work of Hartry Field and
Paul Horwich on these issues.8 Blackburn’s basic point is that deflationism needs to say
something about the truth-bearers as well as about truth. And the challenge for defla-
tionism, when this need is taken seriously, is that the bearers of truth will presumably
be, one way or another, semantically individuated (at least on any plausible account).9 If
the principles of semantic individuation—what determines the distinction between
what is said by one sentence and what is said by another, for example—involve appeal
8
See for instance Field 1994 and Horwich 1998.
9
Propositions for Horwich, sentences individuated by meaning for Field.
424 CRISPIN WRIGHT
to real semantic word-world relationships, then much of the metaphysical motivation

for deflationism in the first place—the attempt to finesse such questions insofar as they
seem to be forced on us by the traditional issue concerning the nature of truth—will be
frustrated. The problems will, in effect, recur at, so to say, the sub-sentential level and
will promise to be no more tractable than before.
I cannot attempt to elaborate or engage with this major issue here, except to say that the
challenges Blackburn outlines to this more general deflationism strike me as sharp. It is, of
course, because they are sensitive to this general kind of worry that deflationists tend to
propose the kind of non-referential semantic theories that they do, in terms of conceptual
or inferential role (Field, Brandom) or regularities of use (Horwich). Blackburn’s suspicion
is that such proposals cannot prove satisfactory, or at least not comprehensively so. And the
reason, if I read him correctly, is broadly because there is no evident alternative to a
referential semantics if we are to have the resources to draw certain distinctions which, at
least in certain regions of discourse, seem fundamental—for example, the distinction
between the kind of situation which someone’s applications of a particular shape-predicate
may characteristically reliably indicate (because of the likelihood of illusions of perspective,
say) and the kind of situation which actually suffices for its application.
Whether or not ultimately compelling, what is striking is that the kind of objection
that Blackburn here lodges against a generalized deflationism highlights a possibility, or
so it might be thought, for selectively combining expressivist views of different regions of
discourse with an across-the-board deflationism about truth. The idea would be that
propositional surface, and with it deflationary truth-aptitude, may be common to very
different kinds of sentential contents, some grounded in the referential semantic properties
of the constituents of the sentences which express them, and others belonging with an
expressive, or otherwise functional, characterization of the role of, for example, the
signature predicates of the discourse concerned. If this were so, then when an expressivist
is challenged to explain what constitutes the difference between those regions of discourse
which serve our attempts to represent objective matters of fact and those where we give
vent, rather, to our attitudes and values, and to do so in a way consistent with both types of
sentence being alike apt just for deflationary truth, he can reply that the difference is
grounded in the differences between the two kinds of proposition concerned, which are
in turn reflected in the ways in which the meanings of the sub-sentential constituents of
statements in the two kinds of discourse are respectively determined.
Both this proposal and that of Truth and Objectivity would separate the idea of serious
representation from that of truth-aptitude. But whereas I proposed to leave in place the
connection with a certain kind of truth-aptitude, the outlined proposal breaks the ties
altogether: the contrast between the kinds of discourse for which an expressivist view is
apt and others is to be made by reference to the kinds of propositional contents in
which they deal, and the taxonomy of these kinds in turn is to be charted by reference
to distinctions in the ways in which the meanings of sentences in the kinds of discourse
concerned are respectively determined by the kinds of semantic properties possessed by
their sub-sentential features.
REPLIES 425
If this is not a complete misinterpretation of Blackburn’s intent, then it is notable—

perhaps ironic—that it represents what is in effect a reversion to a variant of Dummett’s
master thought, that the real issues here have to do with the proper model of meaning.
Of course, a semantic theory which delivers the kind of contrast that Blackburn—so
interpreted—wishes to make will have to be very different from the usual kind of
homophonic, truth-theoretic line of goods on which debate was centred in the 1970s.
That style of compositional semantics will mask the kinds of distinction that Blackburn
is anxious to see drawn, so the relevant sort of theory will have to be, in the
terminology of that time, immodest or “full-blooded”.
I have no wish to prejudge what might be accomplished by serious theoretical
endeavour in this direction. But I do remain sceptical about the consistency of drawing
the contrasts that expressivism needs in this kind of way while clinging to a purely
deflationary conception of truth. The evident difficulty of such a view will be that
when a discourse receives a properly referential semantics, it’s going to be hard to
interpret the truth of, say, a simple singular term-predicate sentence in purely defla-
tionary terms, in the presence of the consideration that the sentence will be true just
when the object denoted by its subject term possesses the property associated with its
predicate—which, where substantial referential relationships are involved, amounts to
correspondence to the worldly situation of the relevant real object’s having the relevant
real property. And that is precisely what won’t be said when, by contrast, the semantics
of a given predicate is explained expressively from the start, rather than in terms of
reference to a worldly property. In short, my expectation is that Blackburn’s own
canvassed style of objection to deflationary semantics, will come back—if good—to
bite him and enforce his adoption of a non-deflationary conception of truth in all cases
where it grips, that is, in all cases where, there is theoretical need for referential
semantics, non-deflationarily construed. Truth for those discourses cannot be defla-
tionary if semantics is not. Granted only, then, that this does not happen exception-
lessly—that such discourses contrast with others for which an expressivist semantics is
apt—the result is going to be to impose a form of alethic pluralism, at least to the extent
of the concurrence of correspondence and deflationary conceptions of truth.
I therefore think that the differences between Blackburn’s present conception of
these matters and that proposed in Truth and Objectivity have effectively become quite
small. If the distinctions between regions of discourse which, according to the vision of
Truth and Objectivity, involve different truth-properties are to be sustained by differ-
ences in the kinds of propositional contents involved—well, I have just argued that a
systematic account of those differences in kinds of propositional content will itself
impose corresponding differences in one’s conception of truth and truth-makers for
the contents in question, ensuring in particular that a deflationary conception of truth
will be at most only locally correct. Conversely, alethic pluralism of the stripe proposed
in Truth and Objectivity will have to address the question of what it is about a discourse
that fits its characteristic statements to qualify for one rather than another among the
426 CRISPIN WRIGHT
plural kinds of truth—and the answer to that, whatever the detail, will presumably
have to advert to systematic differences in the kinds of sentential content trafficked in.
Maybe, then, there is still a debate to have about what comes first: whether it is
differences in the kinds of propositional content concerned, independently explicable
in terms of semantic theory, that ground the applications of different truth-properties;
or whether it is, rather, the applicability of different truth-properties, grounded in the
discriminations made by various cruces of the kind that Truth and Objectivity tries to
describe, that imposes differences in the way we should think of the contents
concerned and hence—perhaps—the style of systematic semantic theory that should
account for them. I confess to a continuing inclination to the Truth and Objectivity form
of view, since I remain a little sceptical about the prospects for systematic but immodest
expressivist semantics. But the matter is for further attention.
Marks of Realism
Although it is my impression that it has sometimes been read as an anti-realist tract, the
real impetus of Truth and Objectivity was anti-quietist—or metaphysically activist. The
central idea, again, was that illumination of the issues between realism and its oppo-
nents, and a consequent vindication of the philosophical authenticity of the debates,
might be found by clarifying a range of relevant variations in the local profile of truth.
The main question thus became, what characteristics or marks of a region of discourse
might underwrite a realist or, broadly, anti-realist conception of truth for its proposi-
tions?
I proposed four. One was the Dummettian point: we are implicitly thinking of truth
in a particular discourse in a realistic spirit if we conceive that some at least of the
relevant propositions may, or must, have determinate truth-values irrespective of our
ability to determine what those truth-values are or even to accumulate evidence one
way or the other. Such a conception leaves us no alternative but to think of the source of
such truth-values as lying in a reality constituted independently of our cognitive
endeavours. The latter notion, however, does not seem to require the possibility that
truth be evidence-transcendent. Even in regions where, as we think, best enquiry must
lead us to the truth, the essence of realism is still intact, one would suppose, if such
enquiry can be conceived wholly as responsive to the subject-matter of the discourse,
whose nature is accordingly a matter of discovery. That contrast—the contrast between
enquiry conceived as wholly responsive to, and reflective of, matters constituted
independently of it, and procedures of opinion-forming where, in some sense, the
thinker ‘has his thumb on the scales’—seemed to me to be exactly the contrast that
Plato had Socrates debate with Euthyphro. It provided a second crux whose clarifica-
tion entered the agenda of my book.
Evidence-transcendence and the Euthyphro contrast both concern the nature of the
relationship between best possible enquiry and its outcome. Another, older perspective
on these debates, saw them as more directly concerned with the nature of truth itself.
REPLIES 427
Of course the older debate did not anticipate the possibility of alethic pluralism; for
Hegel, Frege, Bradley, Russell, and Ramsey, there was a decision to be taken about the
nature of truth across the board. The essence of realism was conceived as commitment
to, in effect, a correspondence conception of truth—a conception of truth consisting in
the successful representation by thought of a reality standing independent of it in the
manner in which the scene it depicts stands independent of a photograph. But the
trouble with this always was that to speak of truth as ‘correspondence to fact’ seems
more of a platitude than a piece of metaphysics—a thesis that adds only a modicum of
pomposity to the remark that a statement, or belief, is true if things are as it says they
are, if it “tells it like it is”. Not much to disagree with there. And the classical attempts
to give metaphysical substance to the platitude—to say in an illuminating, independent
way what exactly the relation of “correspondence” is, and what the nature of the other
term in the relation, the ‘facts’—petered out in vacuity, or implausibility, for all but the
cases typified by felines on hearthrugs.
Truth and Objectivity tried to do better by canvassing two further realism-relevant
cruces, each highlighting a feature that a discourse might lack but whose possession
would serve to give some additional substance to the correspondence platitude.
Cognitive Command proposed a connection between the idea of genuine representation
and the occurrence of some form of cognitive shortcoming in circumstances where our
representations fall into conflict. Width of Cosmological Role, focusing on the second
term of the correspondence relation, proposed that more than platitudinous substance
is given to the idea of ‘the facts’—the things which our thought may serve to
represent—if they can be seen to play a more active role in the life of the world than
merely serving as the, so to speak, internal accusatives of those representations.
To address the spectre of quietism, a successful articulation of any proposal of this
kind must pass through three stages. At the first, the task is to frame a relatively clear,
relatively interesting initial characterization and show how the crux outlined connects
with the characteristic intuitive realist imagery of mind-independence, answerability to
a reality ‘not of our making’, and so on. At the second stage, this initial characterization
should become the focus of further refinement and development; in the case of
evidence-transcendence, this might involve clarifying the modalities involved in the
claims that, in the region in question, truth can, or cannot, outrun all evidence, and
subjecting the relevant notion of evidence to further specification; in the case of
Euthyphro, it would involve doing work on cashing out the metaphor of contrasting
directions of fit between best opinions and the facts.
However, both these stages might be illuminatingly accomplished and yet leave us
no wiser when it comes to determining how to debate, still less settle a debate,
concerning how a given, contested discourse fares in respect of the crux in question.
One might for example be perfectly clear what it would be for the truth of a particular
empirical scientific theory, say, to be potentially evidence-transcendent, and yet have
no idea how to determine whether it would be justified to suppose that the theory in
question, understood correctly, did indeed have that potentiality. It is one thing, in
428 CRISPIN WRIGHT
other words, to achieve a relatively exact account of a characteristic whose possession

by a particular region of discourse would mandate thinking of it in a realist, or anti-
realist, way, but another thing to nail down how to resolve the question whether the
discourse concerned does indeed have that characteristic—where this issue in turn
breaks into the descriptive question whether our actual practice of the discourse
manifests the implicit ascription of the characteristic, and the normative question
whether it ought to. The third stage of clarification of a crux is thus to explain how
to prosecute an intelligent debate on the question whether the crux is indeed satisfied
by a particular region of discourse, and how in principle to determine the winner of
such a debate.
There is, of course, a pessimistic induction immediately to hand about the prospects.
In the case of evidence-transcendence, for example, Dummett’s own third-stage
suggestion was, as noted, that the debate should be taken into the theory of meaning:
the way to determine whether the notion of truth engaging a given region of discourse
was or was not potentially evidence-transcendent was to see whether the best theory
of meaning for that region would be based on potentially evidence-transcendent
truth-conditions or not. But as already noted, Dummett’s own, famous anti-realist
arguments—the considerations concerning acquisition and manifestation of under-
standing—seemed designed to show that evidence-transcendent truth-conditional
theories of meaning were never the best way to approach the project of giving a theory
of meaning, and hence that anti-realism should prevail across the board. For most of
those who, in its heyday, took an interest in the Dummettian conception of the issue,
this global anti-realist conclusion was too big a pill to swallow.
Dummett’s perspective gives rise, in fact, to a dilemma of which that is one horn. If,
as Dummett seemed to intend, the anti-realist argument is simply that there is no such
thing as understanding a statement—any statement—in such a way as to grasp the
possibility of its being true beyond all possible evidence, then, should the argument
succeed, that seems more of a reason for misgivings about the suggestion that realism is
satisfactorily explained as involving that commitment than cause to congratulate the
anti-realist. The project, after all, was to give sufficiently concrete sense to the realist
and the anti-realist standpoints to subserve intelligent local controversy. If, on the other
hand, the anti-realist grants that a grasp of evidence-transcendent truth-conditions is
possible, but contends that it is not realized in the case of a specific region of discourse,
then presumably the argument must then turn on the characteristics of competent
participation in discourse in that region. And that threatens to make the outcome of the
debate turn on the features of the actual linguistic practice concerned, which in turn
threatens to undercut, what was a central feature of Dummett’s account, the potential
of anti-realism to be revisionary of the linguistic practice (specifically, the inferential
practice), concerned. Yet the revisionary potential of the anti-realism of the Intui-
tionists is surely a non-negotiable requirement on any satisfactory account of it.
This example nicely illustrates the potentially very tricky character of the enterprise,
as one tries (i) to give sense to an original opposition largely left at the level of metaphor
REPLIES 429
and image, (ii) to provide additional clarification of what is in dispute, (iii) to under-
write significant, assessable debate about it, and (iv) to conserve any canonical implica-
tions, like the logical revisionism of the Intuitionists, of the unreconstructed initial
views. The quietist charge, that the debates between “realists” and “anti-realists” are in
the end without substance, or are anyway futile, is not going to succumb easily. Did
Truth and Objectivity manage any progress?
Significant doubts are expressed in Stewart Shapiro’s characteristically trenchant
essay.
Cognitive Command
The Cognitive Command constraint, recall, represents an attempt to put to the service
of realist/anti-realist debate the seemingly platitudinous reflection that where devices
of any kind whose function it is to represent states of affairs of a certain sort—cameras,
fax machines, wax tablets—conflict in their representations of a single scene or object,
it has to be true that one or another device, or process, has succumbed to some kind of
malfunction or shortcoming. So too, then, if thinkers, targeted upon the same ques-
tion, deliver discordant verdicts about it, and the question is one where our belief-
forming methods are thought of as apt for the production of genuine representations in
thought of self-standing matters: some kind of cognitive shortcoming on one, or both,
sides has to be involved. Conversely where, as we conceive, disagreement need not
betray such a shortcoming, it is inappropriate to think of the discourse in question as
representational; and the truth-predicate which engages with its statements had better
not be thought of as correspondence.
Shapiro focuses on a problem case for this idea that was already raised in Truth and
Objectivity.10 It is generated by (one understanding of ) the widely accepted thesis that
the data for empirical science are ‘theory-laden’—that what a thinker may correctly
report as having been observed in a particular context will be a function, in part, of the
theoretical commitments that she brings to the context. This thesis opens up—or so let
us grant11—the theoretical possibility of a certain kind of intractable disagreement
about issues in scientific theory. Two theorists may be destined to disagree, no matter
how thoroughly, extensively, and well they each investigate, because they cannot
agree about the data which such investigations throw up—because they begin by
bringing different background theories to the interpretation of those data. Yet there
might never be anything to choose between the resulting theories: each theorist might
be doing as well as it is possible to do, proceeding in a methodologically unimpeachable
way and achieving overall unimprovable results in internal theoretical equilibrium. If
10
He does acknowledge this (this volume, p. 232).
11
The implication is certainly not immediate, but it is not implausible. There is more detailed discussion
of it in Chapter 4 of Wright 1992, but I won’t attempt to take that further here.
430 CRISPIN WRIGHT
we grant that this is possible, should we conclude, under the aegis of the Cognitive
Command constraint, that theoretical science is a non-representational project?
The conclusion is certainly premature. The scenario described is one where there is
indeed no operational shortcoming on the part of the proponents of the conflicting
theories. Each is proceeding methodologically correctly by accepted standards. Each is
constructing a theory which, holistically assessed, has no superior. But there will still be
cognitive shortcoming if best method itself is here prone to cognitive shortfall. Consider
a scenario where the outcome of casting dice is considered by a tribe to be a good
predictor of the weather. Imagine that two members each roll a die in the appropriate
manner and wind up with conflicting predictions. Neither is guilty of any operational
shortcoming, but no conclusion should be drawn on that account about the repre-
sentationality of their meteorological discourse. The discourse is representational, and
the disagreement does involve cognitive shortcoming—it resides in reliance upon a
belief-forming method whose production of results converging with the facts is in the
lap of the gods and beyond the control of the investigator.12
However there are several matters arising. First, this way of squaring the troublesome
disagreement scenario with Cognitive Command does of course presuppose the adop-
tion of a realist view of scientific theory (or weather forecasting). It is because it
preconceives of scientific theoretical statements as apt for the representation of matters
of real fact that it is able to identify a potential cognitive shortcoming in reliance on the
output of best scientific method. In short, a presupposition of scientific realism is being
used to support the claim that scientific theory satisfies the Cognitive Command
constraint, and hence that methods that may irresolubly lead to conflict about the
scientific facts involve cognitive shortfall. That is all right. But it does mean that the
grounds for the preconceived realism have to be found elsewhere. The Cognitive
Command constraint is not, in this kind of case, going to deliver tools to assist the
resolution of the philosophical debate about realism. The constraint is, in this kind of
case, of no help in determining winners and losers.
I think I was under no illusion about this when writing Truth and Objectivity. There,
I saw the problem raised by the relevant kind of disagreement scenario rather differ-
ently to the way that Shapiro sees it. In my view then, the most pressing problem was
to reconcile the scenario with a scientific realism that, perhaps impressed by the
Dummettian anti-realist critique of evidentially unconstrained conceptions of truth,
held both that scientific theorizing was a robustly representational activity, and that all
the truths (and all the falsehoods) in which it was competent to deal had to be
evidentially identifiable. That kind of view is, it seems to me, in difficulty in the
12
Of course, we will want to say that, in addition, there is the shortcoming of a quite irrational reliance
upon a totally unsuitable method; and there isn’t that in the case of best scientific enquiry. But remember that
the Cognitive Command constraint requires that it be a priori that cognitive shortcoming be involved in the
formation of divergent beliefs. And it isn’t a priori that short-term weather forecasting by the rolling of dice is
not a reliable, inductively supportable method.
REPLIES 431
troublesome scenario if the core idea of Cognitive Command is to be respected. The

problem is that the constraint seems to jeopardize anything intermediate between
science’s being apt for the representation of potentially evidence-transcendent matters
of fact and its being merely minimally truth-apt. It was, in other words, the more
moderate, non-Dummettian forms of scientific realism that seemed to be in danger.
Shapiro suggests that the problem lies with the inability of the Cognitive Command
constraint as formulated to differentiate between cases where it fails as a result of the
non-objective nature of the subject-matter and cases where it fails due to the holistic
character of the evidence—the circumstance that the proper acceptability of pieces of
putative evidence is a function of the antecedent commitments of the enquirer.
I acknowledge the distinction, of course, but it would be hasty to conclude either
that failures of Cognitive Command in the latter kind of case are, so to say, realistically
neutral, or that the constraint gives us no grip on the former kind of case. It remains
reasonable to look askance at the claim to representationality when Cognitive Com-
mand fails in a context where cognitive access, if we had it, would be non-inferential
(comedy). And with respect to putatively realism-neutral failures due to holism of
evidence, it is exactly an alleged failure of that kind that lies at the very heart of the
famous scepticism of Quine, and Davidson, about the factuality of meaning and of
ordinary intentional psychology respectively. Those scepticisms are driven precisely by
the thoughts first that Cognitive Command, in effect, fails, for holistic reasons, for
ascriptions of meaning, and of intentional states; and second that there is no prospect of
an independent account of the subject-matter of those discourses to, as it were, shore
up their representationality in any case and allow us to say that the possible divergences
which cause the problem are properly attributed to limitations in the method. Of
course, in those two cases, there is some independent plausibility in the ideas, respec-
tively, that there can be no more to the meaning of expression than somehow surfaces
in its observable use, and no more to the intentional states of a subject than somehow
surfaces in her observable behavior. So in those cases, the anti-realist argument runs
from a failure of Cognitive Command plus scepticism that a realist conception of the
subject-matter can be independently motivated.
So how do matters now stand with Cognitive Command? It was a suggestion,
certainly never proved, of Truth and Objectivity that the constraint represents, as it were,
First Base—the first test to be passed if a case is to be made that a certain discourse is
more than merely minimally truth-apt. That suggestion can still remain on the table.
What we learn from reflection on the troublesome disagreement scenario is that
sometimes the question whether a discourse satisfies Cognitive Command may have
to be explored via considerations that argue directly, and independently, for or against a
more robust form of realism. Quine’s view was, in effect, that no such argument could
be given for the case of meaning and translation, whereas realism remained the natural
and proper view for the scientific theoretic enterprise. That is the whole point, in his
way of thinking, of the distinction between the arguments for the indeterminacy of
translation and the arguments for the underdetermination of scientific theory by data. But
432 CRISPIN WRIGHT
the question, of course, is what further cruces might be invoked to discipline the
drawing of that distinction: which are the cases where realism is well-motivated despite
the lack of any independent a priori guarantee that cognitive shortcoming must be
involved in conflicts of opinion? That, I think, is the most important question to
emerge from Shapiro’s discussion.
Wide Cosmological Role

Width of Cosmological Role was originally characterized like this:
Let the width of cosmological role of the subject matter of a discourse, be measured by the extent
to which citing the kinds of states of affairs with which it deals is potentially contributive to the
explanation of things other than, or other than via, our being in attitudinal states which take such
states of affairs as object.13
This proposal emerged from my attempt to refine Gilbert Harman’s objection to

various forms of realism about ethics that, suspiciously as he suggested, ethical “states
of affairs” have no part to play in the best explanation of ethical opinion—contrary to
what one would expect if such opinion was properly conceived as knowledgeably
responsive to such states of affairs. As critics observed, Harman’s idea is open to various,
more or less lethal difficulties.14 There are, in particular, various problems with his
invocation of the notion of best explanation. For example, the best explanation of any
belief based on defeasible evidence will be the believer’s possession and appreciation of
the evidence, and will thus be consistent with the falsity of the belief. So in cases—for
example, theoretical science—where the strongest possible evidence is invariably
defeasible, Harman’s constraint, conceived of as a necessary condition for the appro-
priateness of a realist attitude, threatens to exclude anything more robustly realist than
the constructive empiricism of van Fraassen. It was this kind of wrinkle that the cited
formulation of Width of Cosmological Role was designed to avoid. It seemed to me
that the kernel of insight in Harman’s suggestion was this: that the idea that our
opinions in some region are indeed responsive to an objective subject-matter attains
substance only if the subject-matter concerned is operative in the explanation (best, or
merely good enough) of more than the formation of those opinions.
Shapiro, though, has a worry about the invocation of the notion of explanation per
se, best or otherwise. It is that whether it is indeed a (good enough) explanation of a
given explanandum, E, to cite states of affairs, S1 . . . Sn, is a notoriously interest-relative
matter. Shapiro gives the example of the explanation of a devastating fire in a building:
a perfectly good physical explanation, such as might satisfy a chemist, in terms of a
combination of evaporated petrol and ageing electrical circuitry may very well not
satisfy the police. In general, whether something explains something else depends on
13
Wright 1992, p. 196.
14
Some details about the dialectic it gives rise to are given in Chapter 5 of Wright 1992.
REPLIES 433
the purposes, and indeed the intellectual compass, of an enquirer. Moreover purposes
and intellectual compass can cut across each other: What if what I want is the best
available scientific explanation of a given phenomenon, which unfortunately I am too
dim to understand?
So, what exactly is the problem? Shapiro is undoubtedly right that what we will treat as
a satisfactory explanation of a given phenomenon is surely context-sensitive in the
kinds of ways thus briefly gestured at. But it is not obvious that this reflection actually
does very much to draw the sting of the Wide Cosmological Role constraint, rather
than merely invite a modification of its formulation. The original challenge to the
moral realist, for example, was: Look, if you want to conceive of moral states of affairs
as situations independent of human sensibility to which, in the best case, our moral
reactions and opinions are responsive, then you owe an account of what (else) might be
attributable to their influence and workings, other than those reactions and the
formation of those opinions. Well, taking on board the point about interest-relativity,
the challenge merely becomes that of characterizing a possible set of interests (and
intellectual accomplishments) which would allow a thinker rationally to receive the
citation of moral facts as explanatory of non-moral facts other than those constituted in
the moral opinions and reactions of human subjects. That still seems to be a challenge
with teeth. After Shapiro’s observation, it is no clearer how to meet it than it was
before.
There is, though, a deeper point that may be elicited from Shapiro’s discussion.
Someone who advances anything in the spirit of Harman’s original proposal, or the
wide cosmological role constraint, as a condition on the appropriateness of a realist
view of a certain discourse almost certainly has in mind a notion of explanation that
reflects a putative objective explanatory order of things, the notion of a cosmos in which
events and states of affairs do not obtain for no reason, but feature upstream and
downstream of each other in determinate explanatory relationships. It is of course quite
consistent with such a metaphysical conception that what we treat as an explanation of a
particular situation may vary with the context of our interests and understanding. But if
we think of the objective explanatory order as a lattice-like structure, what will so vary
as a function of the context will be what node or nodes, upstream of but variously
laterally and vertically connected with the target node, we choose to cite.
On this picture, explanatory relationships are, quite consistently with the interest-
relativity of explanation, wholly objective. But the picture is not forced on us. Anti-
realism about explanatory relationships is one possible anti-realism. What is the status of
the Wide Cosmological Role constraint, or anything of its ilk, from the perspective
of such an anti-realism? One can still make a distinction, of course, between those kinds
of situations, or events, of which we find it intelligible that they should have the kind
of ramified explanatory connections which the wide cosmological role constraint calls
for, and those where that does not seem intelligible. But, absent the framework of an
underlying objective explanatory order of things, the distinction now seems to emerge
as a fact about our concept of the various subject-matters in question, rather than
434 CRISPIN WRIGHT
something fitted to underwrite well-conceived contributions to the traditional debates

about realism and their metaphysical aspirations.
The point generalizes. Whatever crux is proposed as necessary, or (partially) suffi-
cient, respectively, for the appropriateness of an anti-realist, or realist, conception of a
certain subject-matter, there will be an, as it were, second-order question about the
status, in the light of that and other cruces, of judgments about the satisfaction of that
very crux in particular cases. The tacit aspirations of the traditional debate unite the
anti-realist with the realist in a kind of underlying metaphysical realism: for each, the
desired end-product is to determine the real metaphysical status of the subject-matter
with which a particular discourse treats and our relationship to it. Shapiro’s concern
about the interest-relativity of explanation may be misguided in detail for the reason
cited. But the concern about interest-relativity is prima facie gripping precisely be-
cause, as Shapiro divines, the constraint has its intended purport only if its appeal to
explanatory relationships is understood to advert to objective connections that are ‘out
there’, and interest-relativity looks as though it may subvert that. The traditional
aspiration seems to demand a realist view of the satisfaction of the conditions articulated
in each and every correctly conceived realism-relevant crux. It has to be a fully
objective question whether Cognitive Command is satisfied, whether Wide Cosmo-
logical Role is satisfied, and so on.
Now, the objectivity of these questions is to be assessed in the light of the very same
cruces? If so, then it looks as though we launch a regress. To see this, let’s simplify and
suppose that Cognitive Command is our sole crux. Then if P is a representative moral
judgment, we have to ask not merely whether
(R) P exerts Cognitive Command

is true, but whether it is objectively true—at least if the answer is to have the intended
metaphysical significance. That will involve asking whether
(S) R exerts Cognitive Command

—and again, it seems the answer will have the intended significance, both for R and for
P, only if objective; that is only if
(T) S exerts Cognitive Command,

has an objective status . . .
Whether such a regress is harmless from the perspective of our philosophical
aspirations depends on whether the various questions in the ascent can be answered
somehow at one fell-swoop, or whether they can at best be addressed piecemeal. And
of course the issue becomes yet more complicated once a plurality of realism-relevant
cruces is admitted. Now, one would suppose, every ascription of satisfaction, or non-
satisfaction of a given crux should itself satisfy every crux necessary, and at least one
sufficient, for enough objectivity to safeguard the intuitive metaphysical realism that
drives the debates. Is that aspiration at all likely to be met?
REPLIES 435
So, pressing Shapiro’s concern, we uncover a complex- and awkward-looking

problem. How are any of these disputes to be settled if a verdict, however apparently
well motivated, must first be assessed in the light of the same constraints? How are we
ever to finish the discussion?
The most basic goal of the project of Truth and Objectivity was that of developing a
passably clear account of the content of the contrast between realist and anti-realist
views of a certain subject-matter. That goal is uncompromised by the worry just
developed. There is no need to despair that with care, we can indeed, via painstaking
articulation of the cruces proposed, and perhaps others, provide a worked-out devel-
opment of our conception of what is at stake in the disputes about realism. But that was
hardly the limit of our ambition. We wanted to understand better how different
regions of our thought may actually engage with the world in different ways. To do
that, we need not merely to make concrete and specific our inchoate conceptions of
differing such modes of engagement but to get into position to argue, convincingly and
objectively, where the differences fall—how it is, respectively, with physics, set-theory,
ethics, taste, colour, and logic. And here the traditional aspiration of the metaphysical
cast of mind requires that “How it is” means: how it objectively is—what we, objectively,
ought to recognize as the real and varying metaphysical predicaments of these various
regions of discourse. The more general underlying concern provoked by the develop-
ment of Shapiro’s point above is that even if we can get as far as an intuitively satisfying
articulation of a variety of relevant cruces, there may be no way of using it to
accomplish the fully objective insights that we seek. For if, in order for our delibera-
tions to count for anything, we have not merely to argue for particular verdicts about
particular discourses, but also to determine that those verdicts themselves fall on the
objective side of the relevant cruces, there is no end to what we have to determine
before a final verdict can be returned.
Alethic Relativism and Faultless Disagreement

In Truth and Objectivity three anti-realist ‘paradigms’—Expressivism, Error-theory, and
the Verificationist rejection of evidence-transcendence truth—were canvassed and
criticized in preparation for the proposal of the Minimalist-cum-Pluralist conception
of the realism debates there defended. With hindsight, the lack of any explicit
discussion of Relativism about truth—surely one of the oldest and most natural anti-
objectivist tendencies in all philosophy—may seem to have been a major omission.
The broadly Protagorean idea that, whether globally or locally, the idea of absolute
truth is an illusion—that there is only ‘your truth’ and ‘my truth’, or perhaps ‘our truth’
and ‘their truth’—continues to provide a standing temptation for the anti-realist
instincts of freshmen and the folk. Why didn’t it get a look-in, or at least a mention,
in Truth and Objectivity?
The answer is that, with the exception, perhaps, of the debates about ethics, where
relativism remained in focus largely as a result of the attention, sympathetic or critical,
436 CRISPIN WRIGHT
by writers such as Harman and Bernard Williams, the twentieth-century debates about
realism and objectivity had largely relegated the idea of relative truth to the scrapheap.
Non-factualist, non-cognitivist, and error-theoretic proposals had attracted develop-
ment precisely as more competitive—more resilient, and intuitively more felicitous—
ways of trying to do justice to the anti-realist impulse than anything that might be
provided by the dubious idea that the truth of a proposition would be best conceived as
involving an extra argument place, to be filled by a culture, or society, or, worst of all, a
single thinker. Had not serious doubts about the very coherence of this notion
emerged as early as Socrates’ dissection of it in the Theaetetus? And even in the case
of morals, where it continued to be debated, did not even the more defensible versions
of it have obviously unacceptable consequences?15
What a sea-change there has been. The second of the two decades that have elapsed
since the publication of Truth and Objectivity has seen a giddy upsurge of interest in
alethic relativism. However, as befits the contemporary philosophical milieu, the
driving force has been not so much a renaissance of confidence in the ability of
relativistic proposals to contribute to the traditional metaphysical debates—to provide
a stable, theoretically coherent home for anti-realist sympathies—as the thought that
relativism may have a part to play in descriptive philosophical linguistics; more specifically,
the suggestion, elaborated in a powerful and theoretically sophisticated way by John
MacFarlane and others, has been that that an empirically adequate semantics for certain
areas of discourse, including knowledge-ascriptions and talk of epistemic ‘mights’, will
need to incorporate the idea of the truth of an utterance as relative to a context of
assessment if it is to accommodate certain prima facie striking ‘data’ concerning the
linguistic practices in question. This data, principally involving retractions and reassess-
ments, but also embracing the kinds of things that can properly be said by eavesdrop-
pers,16 has seemed striking insofar as it apparently excludes any kind of indexical or
contextualist account of constructions for which a semantic invariantist story impresses
as inappropriate for other reasons.17
The issues raised by this ‘New Age’ relativistic tendency are in many ways orthogo-
nal to those concerning relativism as a broadly anti-objectivist stance. New Age
relativism is, as remarked, a descriptive thesis: a thesis that our actual discourse, in certain
regions of thought, displays patterns of which the best—empirically most adequate—
15
Perhaps most salient is the concern about the recoverability of any forceful notion of moral normativity.
That an argument is valid is a reason to accept its conclusion if you accept its premises. That it is valid by such
and such standards is no such reason unless you think the standards are correct. That the current (as I write
this) military intervention in Libyan air space is morally unacceptable is a reason to stop it, or to work to see it
stopped. That it is unacceptable by such and such standards is no such reason unless you think the standards
are correct. But moral relativism holds not merely that all moral truth is relative to standards but that there is
no further issue about the correctness of moral standards.
16
I suspect that the importance of eavesdropping data may have eluded many philosophers. It looms large
in Egan, Hawthorne, and Weatherson 2005.
17
It is my impression that this part of the argument—the basis for the disjunction: either Contextualism or
Relativism—has never quite received the clarity or attention needed to justify the revolution.
REPLIES 437
semantic theory will make central use of a notion of relative truth. It would be possible
to accept that view and at the same time maintain a revisionary stance towards those
aspects of our discourse, on the grounds that the actual subject-matter concerned was
fully objective, or in other ways unsuited for articulation in terms of relatively truth-apt
statements. Conversely, a traditional relativist about, say, ethics—like Harman—could
perfectly properly be quite undismayed by the consideration, if true, that our actual
moral discourse does not assume the patterns that would invite theoretical description
by the relativistic styles of semantic theory that MacFarlane and others have developed.
Traditional philosophical relativism is a normative thesis: a thesis about the proper way to
think about a certain subject-matter and its claims to objectivity. New Age relativism is
non-normative, and its advocates have mostly been unconcerned with issues
concerning realism and objectivity.18
Still, had New Age relativism come on the scene some fifteen years earlier, it’s hardly
likely that my book could have proceeded as it did. For, notwithstanding their
empirical semantic motivation, the ideas developed by the New Age relativists—
provided at least that they are coherent; provided that it makes sense to think of truth
as relative and to admit contents capable only of relative truth—are readily admissible
within the broad pluralistic framework of Truth and Objectivity and have, at least prima
facie, a natural site of application within that framework. That site is within the
space occupied by what my project regarded as merely minimally truth-apt discourses—
discourses which fail to qualify for Cognitive Command and (as the conjecture of Truth
and Objectivity had it) thereby fail to meet any other realism-relevant constraint. Just
how wide that space is, of course, is controversial but among its less controversial
members would be, for example, discourse about the comic, about the obscene and
revolting, and about the tasty. It is characteristic of such discourses that they may give
rise to what I have elsewhere called disputes of inclination:19 disagreements where one
thinker apparently takes the view that P and another that not P, and where there is little
plausibility in the idea that further information, or sophistication of a relevant sensibili-
ty, could justifiably lead to an assessment of one view or the other as superior. Some—
of course, not all—disagreements about the comic, or the tasty, seem to be like this. In
such cases, folk philosophy—this is not a linguistic datum, or an ‘intuition’, but a piece
of proto-philosophical theory—says that both opinions can be in good standing, and that it
can be perfectly rational for the protagonists to hang on to them, undismayed by the
apparently equally good standing of the dissenting opinion of the other. More specifi-
cally, the folk philosophical view is that such a case can manifest faultless disagreement:
there can be a genuine contradiction between the opinions concerned, neither need be
in error, and neither protagonist should feel that the credibility of his own opinion is
weakened by the situation. (I called these features Contradiction, Faultlessness, and
18
Other aspects of the relationship between traditional and ‘New Age’ relativism are helpfully discussed in
Boghossian 2006 and 2008.
19
Wright 2006.
438 CRISPIN WRIGHT
Sustainability respectively, and will retain these labels here.) The question on which the
idea of some form of relativization of truth might seem to hold some prospect of
assistance is whether, and if so how, this adventurous piece of folk philosophizing
might be developed and stabilized.
The problems about stabilizing the folk philosophical idea are obvious enough.
Suppose you and I, dining out, find ourselves in a fairly vivid dispute of inclination
about the merits of stewed tripe and onion in béchamel sauce. If my opinion, that
the dish is delicious, and yours, that it is disgusting (and hence not delicious J) are
genuinely contradictory, then logic decrees that they are not both true—and classical
logic decrees that one at least is false. Either way, the opinions involved cannot both be
in good standing unless ‘good standing’ is consistent with untruth. But it is no part of
ordinary thought in general to regard an opinion as in good standing if there is no better
reason to uphold it than to uphold a contrary. And by hypothesis neither you nor I can
adduce any consideration to break the tie; there can be vanishingly little room for
discussion in such a case, and the mere fact that an opinion is mine (or yours) is not, for
me (or you), or anyone else, a reason to regard it as true. The problem, in short, is to
make sense of the idea that logically conflicting opinions can remain in good standing
and tenable, once the grounds for them are exposed and found to be in relevant
respects symmetric and of matching strength. This is the problem that alethic relativism
might be supposed somehow to help with. It might be supposed to help because it
might be supposed that it allows us to regard each of the conflicting opinions as true—at
least, in the only sense of ‘true’ for which they are apt.
Before turning to whether relativism really does help, let’s ask: what else might help?
Well, not an invocation of dialetheism: that is, the idea that the disputed opinion might
be both true and false. That would certainly allow us to say that your opinion and mine
are on a par. But what about good standing? Even among those hospitable to the
possibility of ‘gluts’ of truth-values, there is little support for the idea that glutty
statements can be acceptable. The problem, rather, is to explain the sense—contrasting
with denial—in which such a statement should be rejected. There seems no prospect
that a dialetheic account can save Sustainability.
In other work20 I have argued that a broadly intuitionist framework can conserve
what might be regarded as the most important element in the folk philosophical view,
namely that in such a dispute there need be no presumption that either disputant in
particular need be at fault, either by the manner in which they arrive at their view or by
its misrepresentation of the facts. Such an intuitionistic account acknowledges that your
and my respective views about stewed tripe cannot both be true: that much is just a
consequence of the law of Non-Contradiction. But when Bivalence, and associatedly
the Law of Excluded Middle, fail for broadly intuitionistic reasons, that consideration
20
Wright 2006.
REPLIES 439
does not force us to say that one in particular of the disputants has to be incorrect. The
transition fails from Not-(A and B) to (Not-A or Not-B).
In effect, to say that much is just to draw out a consequence of what is involved in a
failure of Cognitive Command. The intuitionistic proposal is the natural treatment of
disputes of inclination for a supporter of the framework of Truth and Objectivity. When
Cognitive Command is missing for a certain range of statements, so is any a priori
guarantee that a dispute about one of them involves anything worth regarding as a
cognitive shortcoming. And once there is no presumption that one or the other
disputant in particular has to be guilty of such a shortcoming—no presumption that
distributively, as it were, either you are at fault or I am—then there is no rational
pressure on us individually to think, “ . . . and the guilty party could as well be me”, so
no pressure to qualify or abandon our respective views. So there is a case that the
intuitionistic proposal, if it can be otherwise well motivated, can capture (something
of ) each of three desiderata gestured at by the folk idea: the opinions in the dispute can
be allowed to be genuinely contradictory, yet there is no justified presumption that
either in particular has to involve any fault, and no reason for either protagonist not to
persist, even in the face of an opposing, no less well supported view.21
The intuitionistic proposal is thus, or so it seemed to me, a not-bad direction by
which to accommodate something close to the folk philosophical idea. But there is one
respect in which it may seem to come short of what is wanted if that idea is to be fully
accommodated. If my opinion is genuinely incompatible with yours, am I at least not
committed to regarding you as mistaken (and you, me)? Maybe there is no pressure,
flowing from a misguided acceptance a priori of Bivalence, to suppose that one of us in
particular has to be mistaken. But still, do we not both, just in taking a view, commit
ourselves to regarding the other as de facto mistaken? In short, it seems that the disputants
themselves cannot regard the dispute as faultless, even if there is no general philosophical
pressure, bearing upon a neutral witness, to suppose that there has to be fault in such a
dispute. So something important in the folk philosophical idea may seem to have been
lost. The scope for considered tolerance—part, plausibly, of what was meant to be
implicated by Faultlessness—conveyed in the acknowledgement that your opinion is just
as good as mine, has not yet been made available to those actually involved in the
disagreement.
Call this extra ingredient Parity. In effect, it is the requirement that Faultlessness be
appreciable, and endorseable, from the point of view not just of neutrals but of the
committed parties in a dispute of inclination. The folk philosophical thought is that
disputes of inclination can manifest each of Contradiction, Faultlessness, Sustainability,
and Parity. The intuitionistic proposal, it seems, cannot accommodate Parity. What,
I think, has not generally been appreciated with sufficient clarity is that relativism does
no better.
21
This is very brisk, I grant. A more careful development, and some responses to objections, are offered in
Wright 2006.
440 CRISPIN WRIGHT
Actually, we need to refer to two importantly different relativistic proposals that

have been distinguished in the recent discussions. MacFarlane’s own preference,
Assessment-relativism, is the proposal that, for statements in the discourse in question,
truth-value is a function of parameters fixed by the context of an assessor. Applied to
discourse of taste, this has the effect that your statement (or opinion) that tripe and
onion in béchamel sauce is disgusting may be correctly assessed as false by me, but true
by you, as a function of variation in the relevant parameter (standards of taste22) in our
respective contexts of assessment. According to Assessment-relativism, then, a given
historical token statement, or opinion, has no settled once-and-for-all truth-value—it
takes a truth-value whenever it is assessed, and what truth-value it takes depends on the
value of relevant parameters determined by the operative context of assessment, which
of course may vary.
The other relativistic proposal is what MacFarlane has chosen to call (most unhap-
pily, in my view) Non-indexical Contextualism.23 This is properly a version of relativism,
rather than contextualism as normally understood, since it is integral to it that the
content that I endorse when I affirm that tripe and onion in béchamel sauce is delicious
is indeed the very content that you implicitly deny when you affirm that tripe and
onion in béchamel sauce is disgusting. And the truth-value of the content so affirmed,
or denied respectively, is indeed, as before, a function of variation in the value of the
relevant parameter (standards of taste). What distinguishes the view from Assessment-
relativism is that the truth-value of my affirmation is to be determined relative to my
standards of taste, and the truth-value of your denial is to be assessed relative to yours.
So we can both be right! And now a given historical token statement, or opinion, does
have a settled once-and-for-all truth-value, as fixed by the values taken by the relevant
parameters in the context of its actual authorship, though there may be variation in
truth-value among other tokens of the same propositional content.
It is pretty immediate that Assessment-relativism is useless for the purpose of securing
Parity. By its rules, I am constrained to assess your opinion in the light of my standards,
rather than yours. So of course I will assess it as false. Since I assess my own as true, I can
then, surely, hardly regard your opinion as just as good as mine, and Parity is
surrendered from my point of view, the point of view of a participant in the dispute.24
22
I do not think it is at all clear in what sense basic taste is subject to standards, but I leave the issue aside for
present purposes.
23
The damage is probably done, but if it is not too late, let me put in a plea for “Author-relativism”.
24
Mark Richard, for one, is clear about this. He writes,
Suppose I think that Beaufort is a better cheese than Tome, and you think the reverse. Suppose (for reductio)
that each of our thoughts is valid—mine is true from my perspective, yours is from yours. Then not only can
I (validly) say that Beaufort is better than Tome, I can (validly) say that it’s true that Beaufort is better than
Tome. And of course if you think Tome is better than Beaufort and not vice versa I can also (validly) say that
you think that it’s not the case that Beaufort is better than Tome. So I can (validly) say that it’s true that
Beaufort is better than Tome though you think Beaufort isn’t better than Tome. From which it surely
follows that you’re mistaken—after all, if you have a false belief, you are mistaken about something. This line
of reasoning is sound no matter what the object of dispute. (Richard 2008, p. 132)
REPLIES 441
It is also lost from the point of view of any third party who happens to have his own
standards of taste. If they determine a view, he will be bound to disagree with at least
one of us. If his standards mandate neutrality on the matter of dispute, he will regard us
both as overstepping the mark. And if he has no relevant standards by which to form a
view, he will be in no position to judge ours as on a par.
This limitation of Assessment-relativism when it comes to sustaining Parity should
come as no surprise. It is a consequence of the very feature that MacFarlane designed
into his relativism in order to enable it to accommodate the, as it seemed to him,
compelling data about retractions, primarily in the case of epistemic modals. The
(alleged) phenomenon of the retraction of hitherto correctly asserted epistemic
modal claims purely on the ground of increased information, precisely amounts to a
disavowal of Parity in a prima facie dispute with one’s former self about the claim in
question.
Prima facie, however—as noted—Non-indexical contextualism does better. Now
your opinion and mine are both properly assessed relative to the standards of their
authors. So my opinion may be assessed as true—and your contrary opinion can also be
assessed as true, assuming that they are indeed sanctioned by the respectively different
standards involved. And indeed this upshot seems to chime with the promise of
relativism as intuitively intended: my opinion is ‘true for me’ and yours is ‘true for
you’. The trouble, though, is that this result is bought at the cost of surrender of aspects
of the interaction between contexts of propositional attitude and ascriptions of truth-
value which seem integral to a proper understanding of both, and whose compromise
wears a face of absurdity. In considering whether your opinion that tripe and onion in
béchamel sauce is disagreeable is true, I must answer affirmatively, since you are the
author of that opinion, so the relevant standards of assessment are yours, and you are
right (or so we are supposing) by those standards. But in considering whether to agree
with you—whether the proposition is true that tripe and onion in béchamel sauce is
disagreeable—I must answer negatively, since now the question is of my authorship
and the appropriate standards of assessment are mine. So I wind up affirming that you
believe something truly although what you believe isn’t true. That sounds like pretty
good nonsense, and it is certainly no intended aspect of the folk thought. But
something of the form will be affirmable by each participant in a dispute of inclination
if Non-indexical contextualism is correct.
Another who is clear-headed on the matter is Paul Boghossian who, citing the above passage from Richard
with approval, glosses the central thought as:
just because p is at best relatively true, and just because it is true from my perspective and false from yours, it is
not therefore right to say that our disagreement is faultless. For even if all of this is true, it will still be true that if
I validly (that is, truly, relative to my perspective) judge that p, then it will also be valid for me to judge that ‘It
is true that p’ and also ‘It is false that not-p.’ And if I can validly judge that ‘It is false that p’ then I must regard
anyone who believes that p to have made a mistake. (Boghossian 2011, p. 62)
Boghossian christens this line of thought the Argument from Immersion, and regards it as finally defeating any
claim of relativism to make sense of faultless disagreement.
442 CRISPIN WRIGHT
There is a joke that the celebrated Cambridge mathematician, G. H. Hardy, once

found something obvious after several minutes’ hard thinking about it. It may be that it
is beginning to seem obvious in that kind of way that that there is no squaring this
particular circle: that once Contradiction is accepted, Parity is simply a desideratum too
many. No doubt it is that thought that has prompted many philosophers to dispute the
reality of the disagreement in ‘disputes of inclination’ by proposing various kinds of
contextualist accounts of the content of the targeted claims that allow them to be
compatible with each other. If you and I are not really disagreeing, then of course we
are not committed, by our respective opinions, to regarding the other’s opinion as
inferior, and there need be no difficulty with any of Faultlessness, Sustainability, and
Parity. But this move gives up on Contradiction, and that we contradict each other in
such cases is as intuitive as any other datum of the problem. Is there any other
Contradiction-preserving option?
This is, in effect, the main issue addressed by Carol Rovane’s interesting and creative
essay. Relativism has little attraction for her, but she is interested in the question, what
is the most robust and potentially useful form of the view. Rovane argues that there is
an alethic relativistic option, multimundialism, which—though she does not express
herself in exactly these terms—promises to save Contradiction along with the other
three constraints.
It is, to be sure, a tinge disappointing to realize that this salvage will be bought at the
cost of surrender of the connection, so far taken for granted, between Contradiction
and disagreement. Rovane is explicit that, in her view, there is no saving faultless
disagreement. But even allowing that we are affirming contradictory opinions in our
dispute about tripe and onion in béchamel sauce, it does not follow—it is now
suggested—that we contradict each other. Let it be that our opinions are mutually
contradictory in the sense that neither of us could consistently add an endorsement
of the other to an endorsement of our own. Still there can be daylight between that
admission and the claim that we disagree: that is, the claim that we are committed to
regarding each other’s opinions as false, and hence rejecting Parity. Parity can be saved
if, although I could not add an endorsement of your opinion to an endorsement of my
own and remain consistent, still I am not, in endorsing my opinion, repudiating your
endorsement of yours. What is needed is that, where our standards vary, our taking the
views that we respectively do is simply to have no bearing on the propriety of the
other’s view. Our views are to be, in Rovane’s terminology, alternatives to each other
and thereby normatively insulated from each other.
How is this to work? According to multimundialism—as according to anything
worth calling relativism—the very content, tripe and onion in béchamel sauce is delicious,
that I affirm does indeed recur embedded in your denial. Our dispute does not have us
talking past each other in any sense that would be gratifying to contextualism. But
we are talking about different worlds. The multimundialist takes seriously—and must in
due course explain—the idea that relativistic parameters, for example, standards of
taste, literally contribute to the constitution of a world: a domain of facts, to which
REPLIES 443
statements informed by those standards are then answerable, but which is merely one
among a range of alternatives, constituted by differing standards, to which judgments
informed by those standards will answer in their turn. So we are talking past each other
in the sense that our claims, though semantically contrary, are answerable to different
worlds. And for relativism of this kind, in Rovane’s handling of it, there are no relations
of entailment or incompatibility between what is true in one such world and what is
true in another. That is why your opinion can be just as good as mine.
Rovane’s proposal highlights something interesting that has, in effect, been noted by
other recent commentators,25 viz. that, contrary to its standard presentations, alethic
relativism need actually involve no adjustment in traditional conceptions of the degree of
the truth-predicate—there need be no raising of the ‘adicity’ of truth. Someone who is
attracted to the idea of truth as consisting everywhere in correspondence, for instance,
can still be a relativist about certain kinds of truth if she is prepared to make the
metaphysical multimundialist move of abandoning the conception of a single Tractarian
all-encompassing totality of facts in favor of a many-worlds view of the relevant sub-
ject-matter.
There is a wave of obvious concerns about the interpretation, and legitimacy, of this
metaphysical move. How can there not be a single totality of all the facts that there
are—if the various denizens of the many worlds are indeed all facts? How are they, as it
were, to be segregated except notionally, by some form of subdivision of a more
comprehensive world? If that is how it goes, will not that more comprehensive world
then lurk in the background and offer absolute truth-makers after all? But if, more
exotically, we try to think of our respective worlds of taste, for example, as more than
notional segregations of the facts—as genuinely alternative complete determinations of
reality—what does that mean and how can we accomplish it? And even if we can, how
exactly would that ensure the normative insularity that Rovane canvasses? Don’t we
have to take it that our respective worlds somehow coexist if we are to regard each
other’s opinions as no worse than our own—as true of the world to which they relate?
To fix ideas, we can turn to the model of a prima facie similarly relativistic view of
weather reports. In Aberdeen, Scotland, on 23 January 2010 I report that it is snowing
hard with visibility down to less than 50 metres. In New York City on the same date
Rovane reports that it is sunny, crisp, and clear. Setting aside any semantics of these
remarks that interprets them as containing some form of inexplicit indexicality of place,
let us take it that the content of Rovane’s statement is the very same content that
I would affirm if I were to use her words in my location, and that it is a complete, truth-
evaluable content—something that can contribute, for instance, to a complete specifi-
cation of the content of a wish, or a belief. And let us assume the same, mutatis
mutandis, for my statement. So we are affirming contradictory contents. But, the
25
For example, by Beall 2006 and Fine 2005.
444 CRISPIN WRIGHT
relativist proposal is, these are contents that take a truth-value only relative to a
parameter of place.
Multimundialism is one option for the interpretation of this weather-report relativ-
ism. The idea of the many worlds required by the view may here be interpreted
perfectly straightforwardly. Rovane’s and my respective remarks are answerable to
different locations. Our remarks are mutually incompatible insofar as they cannot both
be true when directed at any single location. But that element of contradiction is quite
consistent with there being no good sense in which we disagree. Genuine disagreement
involves at least the potential, perhaps stubbornly unlikely to be realized, of change of
mind if one comes to see merit in the opposing view. But there is nothing for me to
learn about the status of my report by coming to know that Rovane is speaking truly.
The model is coherent enough—provided of course we grant the coherence of
its play with trans-locally invariant weather-report contents. (And, it hardly needs
emphasis, such semantic invariance is a feature of any interesting alethic relativism.) But
it had better not represent the multimundialist’s best attempt at doing justice to the folk
philosophical idea about disputes of inclination. Of course, it was announced in
advance that we were going to give up on the idea that genuine disagreement is
involved. But this model saves not even a vestige of the idea of disagreement. If, sitting
in the restaurant, menus in hand, we take ourselves to be discussing, and disputing, a
mutually understood topic, the merit of choosing stewed tripe and onion, we are as
confused—on this account of the matter—as Rovane and I would have been if our
remarks had been part of an argument on the telephone about the weather on 23
January 2010. Normative insularity, as captured by the weather-report model, is
nothing more surprising than what is involved between opinions about logically
independent subject-matters. In effect, the model loses contact with the idea that the
original prima facie dispute about taste does after all occur in what is conceived to be a
single, mutually understood conversation, directed at a single dish and its merit. It is
one thing to surrender the idea that real disagreement is involved in disputes of
inclination. But if an account is to do that, it had better do so in a way that avoids
convicting the antagonists in the dispute of purposes so egregiously crossed.
So I am inclined to raise the stakes. The folk philosophical thought is yet more
demanding than we thought: it demands each of Contradiction, Faultlessness, Sustain-
ability, Parity, and No Egregious Misunderstanding—a satisfying, stable account of it must
be consistent with the possibility that the participants in a dispute of inclination
mutually know what they are doing.
If multimundialism can do better in this last respect, it will be required that
something more be made of Rovane’s proposed idea of our respective worlds of
taste as alternatives. The weather-report model offers nothing by way of interpretation
of that. Different locations are not, I believe, in the sense that Rovane intends,
alternative. She insists at several places in her discussion that normative insularity, as
she intends it, is to involve suspension of relations both of incompatibility and compati-
bility. But her and my imagined remarks about the weather on 23 January 2010 are, on
REPLIES 445
the contrary, perfectly compatible. In the weather-report model, the locations are
partial: they coexist alongside each other, and there really is no conceptual difficulty
involved in, as it were, amalgamating the bodies of meteorological information
respectively associated with them. True, we cannot accomplish that amalgamated
body of information when the mode of expression is restricted to the (putatively)
relativistic propositions. But when I accept that it is clear, crisp, and sunny in New York,
I am agreeing with what Rovane says when she simply affirms it is clear, crisp, and
sunny. By contrast, I do not, in granting that tripe and onion in béchamel sauce is
disgusting by Rovane’s standards effectively agree with what she is saying when she
affirms the statement, unqualified, that it is disgusting.
The point on which the weather-report model breaks down most fundamentally is
that if gustatory standards may indeed be conceived as somehow determining a
‘locality’ of fact, they are not to be thought of as doing so in the way that places do.
If there are indeed permissible alternative such sets of standards, they are potentially
permissibly alternative ways of determining all the facts about what is tasty or not, and
in that sense compete over the determination of those facts. Nothing analogous to that
is involved in the restriction of claims of a certain kind to one location or another. New
York City and Eastern Scotland are different actual locations, and truths about the
weather conditions respectively obtaining at them hold of a single actual world. The
same point, granted, holds for truths about, respectively, what is sanctioned by your
standards of taste and what is sanctioned by mine, even if there is contradiction
between the propositions that our standards of taste respectively sanction. That my
standards validate the proposition that tripe and onion in béchamel sauce is delicious
and that yours validate its contradictory are both truths about the actual world. But the,
so to say, corresponding relativistic truths, that tripe and onion in béchamel sauce is
delicious, and that it is not, are, when conceived as by multimundialism, truths about
different complete worlds—albeit worlds that must in some sense be conceived of as
simultaneously real and existing. (For as noted, Parity, taken under the aegis of multi-
mundialism, requires that we each think of the other’s view as sustained by his/her
world of taste. And for that to be so, the worlds concerned must both exist.)
In summary. New York City and Eastern Scotland are simply different places,
disjoint regions of a single spatial reality. By contrast, the worlds of taste that Rovane’s
multimundialist is postulating had better be alternatives precisely in the sense that there
is no larger truth-making reality of which they are disjoint regions. This feature both
defeats the weather-report model and offers a promise that something might be said to
save the sense of difference of opinion—as opposed merely to opinion about different
subject-matters—that the folk philosophical idea about disputes of inclination involves.
Parity will require that the disputants each recognize that the other is, or anyway can
be, judging truly concerning his/her world of taste. The weather-report model has no
trouble with that. But the salvaging of the sense of disagreement requires that the
worlds be at the same time exclusive and, after a fashion, potentially complete. That
aspect totally eludes the weather-report model.
446 CRISPIN WRIGHT
Can it be captured? Well, maybe, if we can understand our respective standards of taste
as fashioning worlds of which each is taken, by its occupants, to exclude the other from any role in
the business of truth-making. Something like that, it seems to me, has to be the point of the
terminology of ‘alternatives’: there are many worlds of taste but, for each of the disputants,
only one that is empowered to determine the truth-values of judgments of taste.
A partial precedent for what the multimundialist seems to need is provided by the
way that a modal realist of the stripe of David Lewis thinks of possible worlds. For such
a modal realist, there is a sense—a transcendental sense—in which all possible worlds
are on a par: each is actual for its denizens, and the truths in each are as robust as the
truths in any other. For us, though, situated in the actual world, as we like to think, real
truth is truth at our world, the only truth-making world that there is, and truths at
other worlds are merely possibilities. On this conception, the opinion, of my Lewisian
counterpart in a world in which there are talking donkeys, that there are talking donkeys is
as robust a truth as my own opinion that there are none. Faultlessness, Sustainability,
and Parity hold from the transcendental perspective. But from the perspective of a
station at a particular world, only that world is actual and the truths that distinguish
other worlds from it are merely possible, actually false propositions. Contradiction, and
disagreement, belong with a committed, intra-world perspective.
Is this a chink of light? The obvious limitation for present purposes of the Lewisian
model of modal truth is that it is multimundialist precisely in the sense that it allows, for
compelling reasons, for no transworld travel, so no provision for dialogue, let alone
apparent dispute, between my counterpart and me. But maybe it could be argued that
there is no good reason to retain that feature when we are aiming for an account not of
modality but of disputes of inclination. Still, there remains no obvious prospect of
saving all the five facets of the folk philosophical idea at one pass, as it were. If we are to
adapt these ideas to the consolidation of that idea, we will have to say that the collective
appeal of the five facets results from a switching between committed and transcenden-
tal perspectives. From my—committed—perspective as we sit, looking at our menus,
there is only the world of taste that I inhabit: and, regarding our conversation from that
perspective, the fact is that we disagree, and you are wrong about stewed trip and onion
in béchamel sauce. But then I, as it were, sit back, and slipping into a transcendental
perspective, recognize a plurality of worlds of taste in which none is privileged for a
truth-making role, and each makes only for its local truths; and I recognize that, from
this perspective, neither of us is at fault, that both our views are, or can be, locally
correct, and just as good as each other.
That, roughly, is how I would myself propose to understand multimundialism. I am
not sure if Rovane would acknowledge any of it. In any case, I think a supporter of the
folk philosophical idea should find it wanting. For one thing, Parity was supposed to
hold from the perspective of the disputants, not a transcendental perspective whose
adoption involves disengaging from the dispute. I want to say, even while recognizing
that we are disagreeing, that your opinion is as good as mine. More, I want to say that it
is as good as mine on the matter in hand—and so that there is simply no analogue here of
REPLIES 447
weather-report locality, no proper place for the idea that we are speaking in, or of,
different worlds. To the extent that an account finds work for anything of that kind, it
is going to be in conflict with No Egregious Misunderstanding. After all, when I take
us to disagree, what do I think you are doing? What perspective, at that moment, do
I take you to occupy? Do I suppose that you intend to make a judgment about what
holds good in my world? Then I egregiously misunderstand. Do I suppose that you
intend to make a judgment about what holds good in your world? Then why am
I taking us to disagree?
Each of the three relativisms so far distinguished—assessment relativism, non-index-
ical contextualism (author-relativism), and multimundialism—variously stumbles here.
So, it is worth noting, does a fourth relativistic proposal which I mooted in the last
section of my 2006. That was the proposal to construe truth, in discourses apt to give
rise to disputes of inclination, as a form of superassertibility but then to allow (ordinary)
assertibility to fragment into a range of properties determined by the relevant non-
cognitive propensities of different participants in the discourse. Thus, very crudely: let
amusement be a non-cognitive response, and let claims of the form, ‘X is funny’, be,
absent reason otherwise, default assertible just when one finds oneself amused by
X. Since senses of humour may vary, and vary, by hypothesis, without cognitive
defect, ‘X is funny’ may be properly assertible by you but not by me. Assertibility
thus becomes a context-relative—indexical—property for reasons other than variation
in one’s information. Superassertibility idealizes away that latter kind of indexicality, by
requiring that in order to be superassertible, a statement must remain assertible under
arbitrary additions to, and improvements of, one’s information. But no matching
idealizing effect is thereby exerted on a non-cognitive assertibility base. So super-
assertibility potentially fragments too.
This fragmentation will allow us to give a quite literal construal of “true for me” and
“true for you”, and will make straightforward theoretical sense of the notion that your
verdict about stewed tripe and mine, though contradictory, are each true for their
respective authors. But again, the proposal limps when it comes to accounting for
Parity and No Egregious Misunderstanding. Suppose I bring an explicit clarity about
the superassertibilist-relativist proposal to the dinner table. When you affirm that
stewed tripe and onion in béchamel sauce is disgusting, you present your opinion as
true. How, when I take it that we are disagreeing, do I understand that claim? If I take it
that it is answerable to superassertibility on the basis of your relevant affective propen-
sities, well, that is certainly an opinion which I can regard as no worse than my own,
but why should I feel that that is anything with which I should want to disagree? So
understood, you are claiming that stewed tripe and onion in béchamel sauce is disgusting has a
property which, in denying that claim, I am not denying that it has. But if I take it that
your claim is answerable to superassertibility on the basis of my relevant affective
propensities, then I egregiously misunderstand.
So, no progress. Rovane herself, to stress, would probably say at this point that,
“I told you so.” Her express view, recall, is that alethic relativism, in its best form, must
448 CRISPIN WRIGHT
give up on the disagreement in ‘faultless disagreement’. How we react if we think she is

right will very much depend on what, in this area, we think that philosophy should try
to accomplish, and what should constitute satisfaction with its products. If our priority
is to articulate, in the sharpest possible theoretical form, what we conceive as the wisdom
incorporated in our ordinary take on the various issues that engage our philosophical
interest, then we will go on trying to understand and stabilize what I have called the
intuitive folk idea, and will merely dismiss the claim of relativism to be of any value in
that project. But we may be more doubtful of the claims to intuitive wisdom. If we
think it likely that folk philosophical ideas—slowly and haphazardly evolved under the
successive influence of the more or less primitive, often superstitious, variously theo-
logical and scientific images that have enjoyed temporary hegemony over the centu-
ries—are as likely to spawn paradox and incoherence as wisdom, then we may be
inclined to regard the various relativistic offerings as falling short merely of elements in
the folk thought that betray confusion, and the remaining question as being only which
of the relativisms is the best of an acceptably sub-optimal bunch.
Personally, I vacillate. But I still think it likely we can do better than relativism, in
any of the four forms proposed, for the purpose in hand. The matter demands a fuller
discussion, but let me close by outlining why.
The problem which seemed to require that we look past the intuitionistic proposal,
recall, was Parity: in opining that stewed tripe and onion in béchamel sauce is delicious,
I opine that it is false that it is disgusting, and hence that your opinion is false. So how
can I regard it as no worse than my own? But regarding your opinion as false
compromises its parity with my own only if ‘false’ carries its normal normative
punch. And a central contention of Truth and Objectivity, was that—at least over merely
minimally truth-apt discourses, where Cognitive Command fails—truth need carry no
payload of accurate substantial representation. When merely minimally truth-apt
claims are at stake, to regard a statement as false need not be to attribute any cognitive
fault to someone’s acceptance of it. So if there need be no other kind of fault, the way is
open for the idea that, in such a case, to describe an opponent’s view as ‘false’ is, in
effect, merely to record one’s disagreement with it, with no implication of any further
deficiency. There would be an imputation of fault, and hence a compromise of Parity,
only when ‘true’ demands a richer interpretation or when the disagreement itself has to
indicate fault. But that, where merely minimally truth-apt discourses are concerned, is
just what there need be no reason to suppose.26
Nor, on this account, is there any evident difficulty with finessing any issue of
Egregious Misunderstanding. The relativistic proposals all tried to accommodate Parity
26
The idea that a rescue of the idea of faultless disagreement might be accomplished by disarming the
truth-predicate of its usual ‘normative punch’ is also canvassed by Boghossian in section II of his 2011.
However, it is there developed under the aegis of explicitly relativistic norms of belief and assertion and, as he
in effect argues, thereby runs into the same difficulties in conserving the disagreement component that
multimundialism meets with.
REPLIES 449
by, in effect, one way or another, compatibilizing the disputants’ claims: by construing
the kind of truth they enjoy, or the kind of truth-makers that bear on them, as
capable of peaceful coexistence, even though the claims themselves are contradic-
tory. It then followed that in taking you to be making a claim that disputes mine,
I have to misunderstand the constraints to which, if you are clear-headed, you
intend it to answer. That, in essence, is the point that persuades Rovane that, for
any viable relativism here, disagreement has to go. But the minimalist proposal
avoids this bind. No relativized notions of truth now feature. When you affirm
the truth of your view, you are not to be interpreted as committed merely to its
satisfaction of a truth-concept that simply has no role in my assessment and whose
application I do not dispute. Rather just as, on the surface, it appears, you are
committed to an appraisal of stewed tripe that I do indeed reject. But, again, your
commitment to its truth need be no imputation of fault to me. You are indeed
committed to the falsity of my opinion. But, since this is merely minimal falsity,
and tagging my opinion as false is simply another way of expressing your
disagreement with it, with no implication of cognitive shortcoming, my view
can be none the worse for that.27
So, not only does relativism fail to improve on the intuitionistic view. It takes a
step back. But much more needs to be said. For one thing, not all the regions of
thought that may be taken to qualify as merely minimally truth-apt are liable give
rise to disputes in which we feel there is no good sense to be attached to the idea
of better and worse among conflicting opinions. Morals is an obvious exception,
for it is no part of folk philosophy to think of fundamental moral disputes as
disputes of inclination. There are variations in ‘metaphysical contour’ even among
merely minimally truth-apt discourses, and Truth and Objectivity did too little to
explore them. Still, I continue to think that the framework it outlined does provide
a good place to start, and in particular that it affords the best extant prospect of
making sense of aspects of our folk philosophical ideas about, as it were, merely
merely minimally truth-apt discourses. And I think, for the reasons that I have tried
to bring out, that alethic relativism has no part to play in that project. The
twentieth-century metaphysical debates were right to set it aside. Whether it is
better fitted for its more recently acquired linguistic-philosophical purposes is a
different question.
27
A full developement of this ‘minimalist’ way with faultless disagreement will need to address the point,
argued for in chapter 1 of Wright (1992), that any truth-predicate must function normatively in a way that
contrasts with the normative role of warranted assertibility. The question is whether truth’s possession of an
intrinsic normative role must vie with the thought that judging an opinion false need impute to criticism. I do
not think so, but must reserve the issue for treatment elsewhere.
450 CRISPIN WRIGHT
Bibliography
Beall, JC 2006 “Modelling the ‘Ordinary View’ ”, in P. Greenough and M. Lynch (eds.) Truth
and Realism, Oxford, Oxford University Press, pp. 61–74.
Blackburn, S. 1984 Spreading the Word, Oxford, Oxford University Press.
Blackburn, S. 1988 “Attitudes and Contents”, Ethics, 98/3, pp. 501–17.
Boghossian, P. A. 2006 “What is Relativism?”, in P. Greenough and M. Lynch (eds.) Truth and
Realism, Oxford, Oxford University Press, pp. 13–37.
Boghossian, P. A. 2008 “Epistemic Rules”, Journal of Philosophy 105/9, pp. 472–500.
Boghossian, P. A. 2011 “Three Kinds of Relativism”, in S. Hales (ed.) A Companion to Relativism,
Oxford, Blackwell, pp. 53–69.
Egan, A., Hawthorne, J., and Weatherson, B. 2005 “Epistemic Modals in Context”, in G. Preyer
and G. Peter (eds.) Contextualism in Philosophy: Knowledge, Meaning, and Truth, Oxford, Oxford
Field, H. 1994 “Deflationist Views of Meaning and Content”, Mind 103/411, pp. 249–85.
Fine, K. 2005 “Tense and Reality”, in Modality and Tense, New York, Oxford University Press,
pp. 261–320.
Hale, B. 1993 “Can There Be a Logic of Attitudes?”, in J. Haldane and C. Wright (eds.) Reality,
Representation, and Projection, New York, Oxford University Press, pp. 337–64.
Richard, M. 2008 When Truth Gives Out, Oxford, Oxford University Press.
Williamson, T. 2006 “Must Do Better”, in P. Greenough and M. Lynch (eds.) Truth and Realism,
Oxford, Clarendon Press, pp. 177–87.
Wright, C. 2006 “Intuitionism, Realism, Relativism and Rhubarb”, in P. Greenough and
Replies Part IV: Warrant
Transmission and Entitlement
Zalabardo, Pryor, Coliva, and Williams
When Does Warrant Transmit?

Not long ago I think it was widely assumed that recognizably valid reasoning from
known, or otherwise epistemically warranted premises, can always be enlisted to confer
knowledge of, or epistemic warrant upon a conclusion: that a recognized valid
deduction from warranted premises always transmits their epistemic credentials to the
conclusion. It was also widely assumed that this idea is tantamount to the principle of
Closure of knowledge, or epistemic warrant, over deduction: roughly, for the case of
knowledge, that the knowable consequences of knowable premises are likewise
knowable. Both points would nowadays be widely disbelieved. Closure is a weaker
principle than Transmission since it says nothing about warrant acquisition—about how
warrant may be acquired for the knowable consequences of knowable premises in any
particular case—while Transmission is generally recognized as open to counterexam-
ples, though individual cases—Moore’s Proof, McKinsey’s paradox, and Putnam’s
Proof that we are not Brains-in-a-Vat, for instance—remain controversial.
Much hangs on the correct classification of the controversial cases. The tenability of a
Dogmatist conception of perceptual justification is widely regarded as dependent on the
transmissiveness of Moore’s Proof; in the case of McKinsey, the issues impinge on the
epistemological architecture of intentional self-knowledge and of the a priori. But as Jim
Pryor’s chapter usefully brings out, the notion of warrant transmission, characterized at the
foregoing level of generality, may be thought to gesture at each of a number of (mostly)
inequivalent ideas. I’ll illustrate, going a little beyond his discussion, by citing six.
Here is the most immediately plausible proposal. Warrant for the (pool of undis-
charged) premises, [A1 . . . An], of an argument, may be regarded as transmitted to a
validly inferred conclusion, C, under these circumstances:
i. When the argument, together with the warrant for its premises,1 provides a
potentially first-time reason to believe C—in particular, when a rational subject
1
There is, of course, a relativity here: arguments are transmissive, or not, relative to a particular type of ground,
or combination of grounds, for their premises. There isn’t the same issue about the transmissiveness of Dretske’s
452 CRISPIN WRIGHT
who comes to know all of [A1 . . . An], could use the argument to learn
C. Transmissive arguments are those that are at the service of the extension of
knowledge by reasoning whenever their premises may be known.
That answer is potentially distinct from this:
ii. When the warrant for the premises, together with the reasoning, provides for a
new warrant to believe C—when a rational subject apprised of the warrant for
[A1 . . . An], could use the argument to add to the epistemic credentials of
C. Transmissive arguments are those that are at the service of the enhancement
of epistemic credentials by reasoning.
The potential for the distinctness of answer (ii) from answer (i) resides in the possibility
that an argument might somehow be at the service of providing additional warrant for
believing its conclusion even though not at the service of providing a first-time warrant for
that conclusion. There might be some arguments that, in a given evidential context, are
essentially apt for warrant enhancement, but not for first-time warrant creation.
The issue is tricky but here is one possible kind of example. Suppose I am trying to
count a rather mobile, smallish flock of sheep in a pen. And suppose you tell me that
there are fewer than thirty and, on counting as carefully as I can, I get the result twenty-
eight. I might reasonably take that result as corroborating what I have been told. But
I might also reasonably think that, had I lacked any independent information about the
number of the flock, I could not in the circumstances have trusted in the result of the
count sufficiently to conclude that their number was fewer than thirty. If that is not an
incoherent combination of epistemic attitudes, then the argument
(a) I have counted the sheep as best I could and got the result, 28;
So,
(b) There are 28 sheep in the pen;
Hence
(c) There are fewer than 30 sheep in the pen,
in combination with the empirical warrant for (a), is not at the service of the acquisition of
a first-time warrant for (c), though it may rationally be deployed to enhance independent
(testimonial) warrant for (c). The thought would be that the count and the testimonial
warrant support each other, and thereby strengthen my overall case for (c).
A Third Answer: C-Lowering

However, even if there are robust cases of this kind, so that answer (ii) is the more
fundamental, this should not encourage the general answer that warrant transmits in all
and only cases
Zebras argument, for example, if the warrant for its premise is not visual but testimonial, or based upon a
laboratory analysis of the creatures’ DNA. (In fact, the relativity goes deeper. More on this below; see note 4.)
REPLIES 453
(iii) When the argument, together with the warrant for its premises, is at the service
of improving one’s overall epistemic position with respect to its conclusion, C
at least, not if “improving one’s epistemic position” is identified with raising one’s prior
rational credence. Here I demur from a suggestion that José Zalabardo offers in the final
section of his essay. Zalabardo wants to argue that an epistemological externalist is in
position to agree that something is seriously amiss with Moore’s proof and similar
arguments—that a satisfactory diagnosis need not depend on our accepting what
Zalabardo views as objectionably internalist constraints on knowledge, or warrant,
whose accreditation might leave one vulnerable to undesirable side-effects (for in-
stance, scepticism!2). Zalabardo quite rightly observes that a valid argument to a certain
conclusion, C, together with the acquisition of certain grounds that properly enhance
one’s confidence in its premises, may nonetheless have the effect of enforcing a lowering
of one’s prior credence in C. Consider, for example, the argument (BIV) from
(a) I have two hands
to
(b) I am not a handless brain-in-a-vat whose visual cortex is being so stimulated as to
give me an experience as of seeing my two hands in front of my face.
Presumably the effect of an apparently normal visual experience as of my hands present
before my eyes will do nothing to lower, and may in the right context (perhaps after a
bloody car accident) raise, my prior credence in (a). But presumably it should lower my
prior credence in (b) which, before the occurrence of the experience, will have been
close to certainty, since I was then not so much as having the appropriate kind of
experience.
The phenomenon—that, in a certain informational context, evidence which raises
the joint probability of the premises of a valid argument to C may not raise, indeed may
lower, the prior probability of a validly derived conclusion C—is undeniable. Call it
C-lowering. Does C-lowering, though, provide an externalism-friendly diagnosis either
of what failure of transmission of evidential warrant consists in, or of the failings of
Moore’s proof?
2
Many philosophers, familiarly, have been tempted to regard scepticism as a kind of nemesis of
internalism. Acknowledging that the contrast between conceptions of knowledge, or epistemic warrant,
properly termed ‘internalist’ and ‘externalist’ respectively is not as clear as one might wish, I think there is
nevertheless an insight in this, which I’ll come to later in these comments. Let me say, though, that I am not
myself convinced that there is any good general debate to be had between internalist and externalist over the
proper analysis, or nature, of ‘knowledge’ or ‘justification’. The fact is there are various kinds of truth-relevant
epistemic good standing that a belief may enjoy, some to do, e.g., with its causal provenance and modal
profile, others to do with the awareness and good intellectual conscience of the believer. I think much recent
epistemology has suffered under the assumption that there is somewhere in the list of epistemic goods one
that underwrites all the rest—much in the way that ethics suffered in pursuit of a single, master account of the
nature of the moral good and of right action. But the truth in both cases is that we are dealing with a plurality
of values, or modes of good standing of action/belief, some of which may be best construed consequentially/
externally and others deontologically/internally.
454 CRISPIN WRIGHT
The questions are separable, but I think that both should get a “No”. Actually, I am
not sure I understand what is conceived as particularly externalism-friendly about the
proposal, since—as Zalabardo’s own concluding discussion of the Soccer case brings
out3—the question whether a particular piece of evidence that raises one’s prior
probability for some premises raises, lowers, or leaves unaffected one’s prior probability
for a validly inferred conclusion C is highly sensitive to background information—to
what else the subject takes herself to have reason to believe.4 It is not clear how this
point is to be accommodated by any of the usual externalist conceptions of the nature
of the warrant concerned.
Be that as it may, Zalabardo provides—as far as I can see—no explanation why, if we
replace (b) in (BIV) by the Moorean conclusion,
(m) There is an external material world,
the effect is that one’s prior probability for (m) will be lowered—or anyway unaffect-
ed—even in circumstances when one’s prior probability for
(a) I have two hands
is raised. The type of C-lowering illustrated by BIV is highly sensitive to the particular
form of the conclusion C in the argument concerned: in effect, there is a rational
lowering of the prior credence in C only when C’s negation predicts the evidence for
the argument’s premise(s). That is not a feature of Moorean arguments in general; in
particular, it is not a feature of Moore’s proof itself.
The point I want most to stress, though, is that C-lowering is not to be identified
with transmission failure in any case, since incoming evidence that lowers one’s prior
credence in C may yet provide a new—and perhaps still, in context, sufficient—reason
for accepting that conclusion. This will be the effect in general when the availability of
the relevant grounds for the premises of the argument itself tells against the obtaining of
a significant range of possible cases under which C would in any case have held good,
and thus adversely affects the latter’s prior probability. Here is a case. Suppose that
I deal you twenty-one cards face down from a standard Bridge pack, advising you—
this is your evidence, (a)—that exactly three aces—two blacks and a red—are among
3
This volume, pp. 319–20.
4
Here is a further example of such sensitivity: Presumably the perfectly zebra-like appearance of the
animals in the pen should raise your prior credence in the proposition that they are cleverly disguised zebra
lookalikes—after all, that was pretty unlikely but now the appearances at least are consistent with it,
something you might reasonably have expected would not be so. So Dretske’s argument, naturally under-
stood, exhibits the C-lowering phenomenon. But merely add as background information that, while frauds
of this kind have unfortunately occurred not infrequently in this particular zoo, the management would
never permit such a thing with any equine species, of which the wealthy owner is particularly fond, and the
perfect zebra-like appearance now becomes strong evidence that these are not cleverly disguised mules. In
those circumstances, the warrant transmits.
Corollary: warrant-transmission is relative not merely to the argument and evidence-type concerned, but
to collateral information. Below I shall suggest that there is a yet further relativity.
REPLIES 455
the dealt cards, and that exactly sixteen of the dealt cards are black. At my invitation
you then pick a card without turning it over. Presumably you can be reasonably fairly
confident—in the light of the chances of being wrong: 7 to 1—that you have not
selected an ace and very confident—with the chances of being wrong at 21 to 1—that
you have not selected a red ace. However, if I now add to your initial evidence the
further datum that the card you have selected is in fact red, chances of having selected
an ace, and hence—what is equivalent in the new informational context—chances of
having selected a red ace, shorten to 5 to 1; yet this is still enough, arguably, to ground
something approaching an expectation that the card, when turned over, will not prove
to be a red ace. Now, in another context, the grounds for this latter expectation—the
information that you have selected a red card, that just five of the cards overall are red,
and that they include just one ace—could have been acquired against a setting of
antecedent total ignorance about the suits and denominations of the twenty-one dealt
out cards. It would then have been intuitively perfectly proper to describe you as
having acquired a reason to think you had not selected a red ace where before you had
none. That reason is exactly the same as that which you finally have in the example as
developed above. It is just that acquiring it in the original context destroys a previous,
stronger reason for the same expectation. So warrant can transmit to the conclusion of
an argument even though its acquisition compromises a previously stronger epistemic
position with respect to that conclusion.5 C-lowering is thus a bad diagnosis of
transmission failure.
Transmission Failure and Externalism

My own suspicion, more generally, is that providing a convincing, recognizably
externalist diagnostic of what is going wrong in at least a decent cross-section of the
usual putative examples of non-transmissive arguments, is none too easy a thing to do.
What would it be, for example, for an argument to fail to transmit knowledge,
understood in any of the familiar externalist ways? Of course, we know to expect
failures of closure if knowledge is required to incorporate sensitivity of the kind
enshrined in the accounts of Dretske and Nozick—if, oversimplifying, one knows
that P only if, had P not been the case, one wouldn’t have believed it. But these
sensitivity-based analyses are now widely disbelieved; and in any case, it is, plausibly, a
constraint on the enterprise to provide an explanation of how knowledge-transmission
can fail even in an argument where closure is satisfied. But how will matters play on
other externalist conceptions of knowledge? Let a knowledge-transmissive argu-
ment—prescinding from any further subtleties—be one which is such that a thinker
who knows its premises and uses them to validly infer its conclusion, can thereby
5
There has been confusion about this point in the literature on attempted Bayesian treatments of transmission
failure. For an example, see Samir Okasha 2004. Jake Chandler 2010 attempts to improve on Okasha’s account.
Helpful further contributions on the issues are Moretti 2012/2011 and Pynn forthcoming.
456 CRISPIN WRIGHT
acquire a knowledgeable belief. Then it is very difficult to see how there can be such a
thing as transmission failure on a reliabilist account of knowledge. For valid deduction
from known premises is, par excellence, a reliable way of forming true beliefs.
One might demur that this won’t be so if the reliable method whereby the premises
came to be known somehow already involves a prior belief in the conclusion of the
argument in question, and if the reliability of that method depends on the conclusion’s
truth.6 In that case, carrying through the inference to the conclusion will not be any
way of forming a belief in that conclusion. Maybe. But this kind of play is no help in the
crucial range of cases where the belief-forming method that leads to acceptance of the
premises consists simply in the exercise of some non-inferential cognitive faculty—say,
perception. If the circumstances are such that just looking at animals in zoo pens and
cages is indeed a reliable way of forming beliefs about their species, then validly
deducing conclusions—any conclusions, including those about the absence of clever
disguises—from those beliefs will be a reliable way of forming beliefs in them. A simple
reliabilist conception of knowledge can thus, on the face of it, have no problem with
the knowledge-transmissiveness of Dretske’s famous inference.
It is similar with safety-based conceptions of knowledge.7 There are various con-
ceptions of safety. But the generic idea of a true belief that holds good in all reasonably
close worlds in which it is formed in the way it is actually formed is obviously
hereditary across valid inferences leading to beliefs in their conclusions. So again it
seems unintelligible, on any view whereby safety, broadly construed, suffices for
knowledge, how a valid argument might fail to be knowledge-transmissive.
It cannot of course be flatly excluded that more refined externalist conceptions of
knowledge may fare better. But that is not the direction that Zalabardo attempts. His
argument that externalist conceptions of knowledge may yet have something interest-
ing to say about Moorean inferences, is not that particular such conceptions can
accommodate transmission-failure. It takes a different, and surprising, turn.
Zalabardo formulates Transmission (for the single premise case) as follows:
T1: If P is a proposition that S believes and for which S has warrant (and Q is a proposition
for which S doesn’t have warrant), then by recognizing the validity of the inference from P to
Q, S will acquire warrant for Q.
6
The reader will easily foresee relevant kinds of case. Suppose, for instance, that I am a less than expert but
not totally incompetent bird watcher and that—to adapt an example of Austin’s—my method of deciding
whether or not a bird is a willow warbler is to judge on the basis of its size, colour, movement and location
provided I already believe it is not a chiff-chaff. If, on the other hand, I regard that issue as open, then I have no
method except to ask an expert. Let the region be such that there are no chiff-chaffs in the local bird
population—indeed, no warblers of any kind other than willow warblers. And suppose I believe this. Then
my method will be reliable and—some forms of reliabilism will grant—locally knowledge-productive. But
the use of this method, plus inference that a bird it identifies as a willow warbler is not a chiff-chaff, is not a
reliable way of forming the latter belief, since it is not a way of forming that belief at all.
7
That is, any view that holds that safety, suitably construed, is sufficient for knowledge.
REPLIES 457
His argument is that we can actually endorse this principle without restriction and still
quite coherently refuse to countenance that warrant can be acquired via the arguments
that we might be inclined—misguidedly—to classify as failures of transmission, includ-
ing Dretske’s Zebras and Moore’s ‘Proof ’ itself.
The key to this prima facie startling claim is provided by the following formulation
of Closure:
If P entails Q, and S has warrant for P and for the proposition that P entails Q, then S has
warrant for Q.
Suppose it is possible to have warrant for the proposition that P entails Q without
carrying through a deduction of Q from P and recognizing its validity. And suppose
someone has such a warrant, and also has warrant for P. Then it won’t be possible for
her, if she now recognizes the validity of the inference from P to Q, to acquire warrant
for Q thereby—since by Closure, formulated as by Zalabardo, she will have such
warrant already. Not that she will be a counterexample to transmission, however; rather,
she will fail to satisfy its antecedent, by failing to satisfy the portion in parentheses in T1.
For Q will not be a proposition for which, before she recognizes the validity of the
inference from P to Q, she lacks warrant.
In sum, given the way Zalabardo has formulated the two principles concerned, and
assuming Closure so formulated holds, warrant cannot be transmitted by inference in
any case where, in advance of inference, a thinker has warrant for the relevant premises
and knows of their entailment of the conclusion in question by some other means.
It’s hardly controversial that there are ways—testimony, for instance—of knowing
about entailments without articulating them inferentially. Zalabardo’s idea about Moor-
ean inferences is to try to exploit the more interesting kind of case where an entailment is,
as he puts it, epistemically transparent—that is, where any normal subject who possesses
the concepts that feature in the premises and conclusion thereby directly has warrant for
the entailment, without any inferential process. It seems correct that there are—and have
to be—entailments of this character: entailments which are, as we like to say, immediate,
and where just to understand the premises and conclusion is to have warrant for the belief
that the latter follows from the former. But granting that there are such arguments, then it
will follow—at least on Zalabardo’s formulations—not that they are counterexamples to
Transmission, but they are cases that do not fulfil the antecedent of Transmission, and are
thus, Zalabardo wants to conclude, cases where warrant can indeed not be acquired by
inference, where Transmission cannot be triggered.
Now, I think there is something right about this. One reaction one may well have,
confronting the claim that Moore’s Proof, for example, could give one warrant to adopt
the belief that there is an external world, is that there is no plausibly thinking oneself into a
frame of mind where, taking one’s sense experience as defeasible evidence for the
presence of one’s hand in front of one’s face, one then moves on to the conclusion that
there is an external material world. There is not enough of an intellectual gap, so to speak,
between the premise and the conclusion to make space for the possibility of a transition in
thought. But I believe Zalabardo mistakes the significance of this. The fact is that this is a
458 CRISPIN WRIGHT
feature of a very wide class of entailments which, subject to appropriate grounding for
their premises, are uncontroversially transmissive. It will hold, for example, for all entail-
ments that instantiate a primitively obvious introduction rule.8 Single steps of disjunction
introduction, or existential generalization, for instance, will thus be unsuitable to transmit
warrant on Zalabardo’s account. But it is evident that such simple inferences can and must
be transmissive of warrant in the general run of cases. So Zalabardo’s diagnosis over-
generalizes and is mistaken.
The problem is with Zalabardo’s characterizations of Transmission and Closure.
I think there are two things wrong. First, transmission should not, in the first place, be
formulated as a principle exclusively about the acquisition of warrant by inference.
Rather, we want a notion of transmission of warrant that is precisely suitable to
apply to cases where an entailment is epistemically transparent, and inference—taken
as an active intellectual transition—is, for the kind of reasons that Zalabardo gestures at,
not involved. Even in such cases, we still need a distinction between cases where
warrant for an entailed proposition can properly rest on the combination of warrants
for the premises and warrants for the entailment, both elements of the combination
being essential, and cases where it cannot. The latter is, in my view, the right
description of Zebras, even if it is conceded that the relevant entailment is epistemically
transparent:9 knowledge of the entailment, conferred (if it is) just in virtue of grasp of
the concepts involved, does not generate, or enhance, warrant for the conclusion when
adjoined to the (perceptual) warrant for the premise.
Second, the understanding of Zalabardo’s formulation of Closure that he needs to
drive his argument is in any case tendentious. For we have to understand the conse-
quent in such a way that “has warrant” contrasts with “is in position to acquire
warrant”—otherwise satisfaction of Closure won’t ensure non-satisfaction of the
antecedent of T1. But Closure is not believable when understood in that way. Any
competitive formulation of the principle has to proceed in terms of the availability of
warrant; but when “has warrant” is so understood, it doesn’t have the significant tense
exploited in Zalabardo’s formulation of Transmission.
I am inclined to think, then, that the upshot of Zalabardo’s interesting discussion is
to intensify, rather than alleviate, such doubts as there may be that there are plausible
externalism-friendly accounts to be given either of transmission failure or of what else is
going on in the arguments that have been canvassed as involving transmission failure.
Externalism, as an epistemologically fundamentalist rather than ecumenical position,
may be forced to deny the phenomenon. And to the extent that the phenomenon,
however best characterized, seems to be intuitively compelling, that is a problem for
such epistemological fundamentalism.10
8
I believe I mean “primitively obvious” in the sense of Christopher Peacocke.
9
Actually, I think it questionable whether the Zebras entailment is immediate.
10
But for further discussion of relevant issues, see Smith 2009, Kotzen 2012, and Pynn 2012.
REPLIES 459
When Does Warrant Transmit? Three Further Answers

There are at least three other answers to the question, When does warrant transmit?,
which are worth considering. Warrant transmits, it may be said,
(iv) When the argument concerned may be rationally deployed to overcome a prior
doubt about its conclusion, C.
This answer, in contrast with the previous three, takes the notion of warrant transmis-
sion into the territory of rational belief management. In effect, it introduces a new
relativity additional to those noted earlier: it views the transmissiveness of an argument
as relative to the background doxastic state of the subject who entertains it. Clearly the
strength of any prior doubt about C may be relevant: and here “strength” should be
understood to cover both the psychological intensity of the doubt and the strength of
the grounds for it.
I think it is clear, though, that this proposal is misconceived. For no matter what
argument we consider, it is almost always going to be a possibility that a given subject
have in advance a sufficiently strong doubt about its conclusion—in either or both of the
above senses of “strong”—to make it rational for them to look askance at the argument,
and specifically at the grounds for its premises. This possibility afflicts virtually any
argument. So answer (iv) threatens to obliterate the distinction between transmissive
and non-transmissive arguments unless some kind of controls are placed on the prove-
nance and strength of the doubt about C. But it is not clear how to go about that.
It would not help to finesse this problem if it were suggested instead that warrant
transmits
(v) When the argument may be rationally deployed to overcome agnosticism about C,
—not if agnosticism too is interpreted as a state of mind that allows of degrees of
strength. And it is so interpreted when it is viewed as itself a state of belief: the belief that
the question, whether or not C, is something that allows of no settlement (at least by
the kinds of considerations deployed in the argument in question). Strength, once
again, may admit of both rational and psychological interpretation. In either case,
though, a sufficiently strong prior agnosticism may dominate a subject’s response to the
details of the argument concerned, and hinder what is in fact a perfectly transmissive
argument from inducing any significant degree of belief in C.
We should conclude that if warrant transmission is to be characterized in anything
like this kind of way—as residing in the capacity of an argument to rationally overcome
a prior lack of commitment to its conclusion—we need to appeal to a version of the
latter that, in the nature of the case, does not allow of rational, or psychological,
entrenchment. The salient suggestion is therefore
(vi) When the argument may be used to rationally move past an open-minded stance
towards C.
460 CRISPIN WRIGHT
Open-mindedness, in its nature, does not allow of strength, or entrenchment: it involves

a readiness to respond to whatever can be said for or against C. If an argument, together
with a set of grounds for its premises, does indeed generate an all-things-considered
sufficient case for C, a rational but open-minded subject will be responsive to it.
At this point, the reader will likely be struck by the thought that answers (i) and (vi)
in effect converge. For rationally to move past open-mindedness about C in response
to a valid argument for it is to acquire a first-time warrant for C. And to acquire, for the
first time, a warrant for P is to be in a position when one should relinquish open-
mindedness about it. The outstanding issue is whether answer (ii) covers a significant
additional class of cases. But I will not here pursue that question further here. The
remarks on transmission failure to follow will be directed at the conception of
transmissiveness gestured at by answers (i) and (vi).
Transmission Failure (I)—The Background

Warrant Model
So, when and why does warrant fail to transmit? And what are the prospects for
developing criteria to identify warrant transmission failures, and so perhaps to rule on
controversial cases, like Moore’s Proof, McKinsey’s paradox, and Putnam’s Proof?
In his intricate and illuminating contribution, Jim Pryor proposes what he calls the
Background Warrant model of transmission failure. To fix ideas consider these two
catchy examples:11
FBI-1 Men dressed in blue “FBI” jackets are swarming around this intersection, posting
signs that say, “No cameras allowed!”
So, apparently:
FBI-2 The FBI has prohibited any filming from taking place here today
So:
FBI-3 What I’m witnessing is not part of the filming of a movie scene
Similarly:
CarTheft-1 I intend to walk to Lot 15, where I left my car, and drive away.
So, probably:
CarTheft-2 I will end up driving my car away from Lot 15.
So:
CarTheft-3 My car hasn’t been stolen or towed away; it will still be in Lot 15 when I get there.
In both cases the impression of transmission failure is powerful. That I see the scene
depicted in FBI-1 is good, but defeasible evidence for the existence of the prohibition
affirmed in FBI-2; but it is, intuitively, no kind of evidence, unless supplemented, that
the scene I witness is not a set-up for a film. That I have the intention, and memory,
11
They are Pryor’s, in correspondence.
REPLIES 461
described in CarTheft-1 is good, but defeasible evidence that I will do what CarTheft-
2 describes; but it is, intuitively, no kind of evidence, unless supplemented, that my car
hasn’t been stolen.
As I (perhaps wrongly) understand Pryor, what the Background Warrant model says,
very simply, is that warrant fails to transmit in these cases because the respective 3-
propositions are among a set of background premises that are required if the respective
1-propositions are to warrant acceptance of the respective 2-propositions. The charac-
teristic examples of non-transmissive inference are thus, according to the model,
merely enthymematic versions of larger, explicitly circular inferences, and are episte-
mically futile for the same reason and in the same way as the latter.
Pryor contends that the Background Warrant model is “the most plausible and
recognizable form of the phenomenon Wright has in mind”, and that “the only
models of transmission failure Wright successfully presents us with are [Background
Warrant] Models”. I think, to the contrary, that at least when understood as just
outlined, the Background Warrant model misrepresents the action in a wide class of
intuitive examples of transmission failure and I will offer some considerations in favor
of that assessment. At the least, it may be useful to note three hurdles that a defence of
the model will need to clear. One concerns cases manifesting an intuitive failure of
warrant transmission although no background premise seems to be needed to support
the argument concerned. The second concerns the strains placed on the notion of a
“background premise” if the model is to have the requisite generality, even when there
is intuitively a need for additional premises. The third concerns the presence among the
characteristic examples of non-transmissive arguments of cases where it seems there is
simply no room for background premises.
Let us consider an intuitive case of transmission failure where the Background
Warrant model seems exactly right.12 Suppose:
Testimony-1 I ask Smith the time and, glancing at his watch, he says, “Eight o’clock”
So, probably:
Testimony-2 It’s eight o’clock
So:
Testimony-3 Smith is telling the truth on this occasion.
Testimony-1 and Testimony-2 conjointly deductively entail Testimony-3. But Testi-

mony-1 and Testimony-3 likewise conjointly deductively entail Testimony-2. And it
is plausible enough to suggest that the justifiable conclusion of Testimony-2 from
Testimony-1 proceeds via an enthymematic inference that suppresses mention of
Testimony-3 as a second premise. However, on that construal of the argument,
12
Cognoscenti will recognize that the following is the pattern of the crucial sub-proof in Stewart Cohen’s
original formulation of ‘bootstrapping’ arguments. Cohen 2002.
462 CRISPIN WRIGHT
there is no question of running it on to acquire, or enhance, epistemic credentials for

Testimony-3: the resulting argument is question-begging in the classical sense of
(tacitly) assuming its conclusion as a premise.
Now, FBI-3 and CarTheft-3 are not background premises in that sense—premises
of which mention is suppressed but which need to be made explicit if the respective
arguments are to be valid. When they are entered as additional explicit premises, the
arguments to the respective 2-propositions are still ampliative and defeasible by
additional information. Still, presumably ampliative arguments can also have sup-
pressed premises. When is it reasonable to regard a proposition as a suppressed—a
“background”—premise in such an argument, given that its introduction will not
make the argument deductively valid?
One suggestion, as a first approximation,13 would be that something may properly
be described as a suppressed premise in any argument, ampliative or otherwise, if it
represents information without which the conclusion will be unjustified but with
which, together with that represented by the explicit premises, it will be justified—
albeit perhaps still defeasibly so. On this suggestion, FBI-3 and CarTheft-3 will count
as suppressed premises in their respective arguments just in case in conjunction with
their respective 1-propositions they provide adequate, though still defeasible justifica-
tion for accepting the respective 2-propositions, for which however the respective
1-propositions do not provide adequate justification on their own.
Is that so? Well, obviously enough, the question needs situating in some further
assumptions about the informational context. On the one hand, one might very well
be justified in believing FB1-2 purely on the basis of FBI-1—in an ordinary everyday
context where it was reasonable not to be bothered about possible movie-making. On
the other hand, one might very well not be justified in believing FBI-2 even on the basis
of FBI-1 and FBI-3—in a context, say, where there was a reasonable concern about the
possibility that criminals, posing as FBI agents, were setting up a bank robbery.
That brings out that the Background Warrant model should predict that transmis-
sion failure is going to be a highly contextual feature of an argument—specifically, that
it will have the effect, in the kind of simplified three-step inferences standardly used
as foci in the discussion, that whether or not there is transmission failure will depend
on whether or not the context—what is at stake and what is given by way of
collateral information—is such as to require the assurance of antecedent justification
of the 3-proposition in order for the ampliative inference from the 1-proposition to the
2-proposition to be reasonable. And that may seem to be as it should be.
But actually things do not seem to be as they should be. If the model of transmission
failure is to be that of a (suppressed) premise-circular argument, then where the context
is (suitably relaxed and therefore) such that a candidate suppressed premise is not
13
Only a first approximation, because we need to allow for the possibility that something features as a
background premise in a bad argument. A second approximation might suggest that the notion we need is
that of a thinker taking it that his conclusion needs, and has, the support of the premise in question, in addition
to that of those explicitly cited.
REPLIES 463
needed, no problem of transmission of warrant to it as a conclusion should be expected.

For instance, if the context is such that men dressed in FBI jackets can reasonably be
assumed to be FBI agents, about their lawful business, then—on the Background Warrant
model—there should be no problem about transmission of warrant from FBI-1 via FBI-
2 to FBI-3. But there manifestly is a problem: even in a context where it is independently
reasonable to take it that what I’m witnessing is not part of the filming of a movie scene, it
will be hardly part of my grounds for that conviction that men dressed in blue “FBI” jackets
are swarming around the intersection, posting signs that say, “No cameras allowed!”
Similarly for CarTheft: the context may be such that it is independently reasonable to
expect that my car hasn’t been stolen, or towed, and hence that I am going to be able to act
on my intention to drive my car away from where I left it; but it is hardly part of my
evidence that my car is where I left it that I have that particular intention.
In brief: whether an additional premise is needed in a putatively enthymematic,
ampliative inference varies as a function of what may reasonably be held to be open to
doubt in the context and the stakes—the importance of the inference’s leading to truth.
But impressions of transmission failure, though information-sensitive in the ways earlier
reviewed—sensitive, that is, to the character of the warrant for the premises and one’s
collateral information (recall the equine-loving zoo-owner of note 4)—do not seem to
vary in this way. FBI and CarTheft seem, on first take, to be non-transmissive whatever is
taken to be doubtful, and whatever the stakes. That is what makes them nice examples.
There is a second problem with the Background Warrant model. Pryor himself
notes that the model is not in fact the best of prima facie fits for some of the classic
putative examples of transmission failure, for instance Zebras:
Zebra-1 There’s a black-and-white striped, horse-shaped creature in this zoo pen, and the
pen is labelled “Zebra”
So, probably:
Zebra-2 That animal is a zebra
So:
Zebra-3 That animal is not a mule cleverly disguised to look like a zebra,
and Red Wall:

RedWall-1 That wall looks red
So probably
RedWall-2 That wall is red
So:
RedWall-3 That wall isn’t white but lit by tricky red lighting.
Pryor’s thought is that Zebra-3 (and, I would imagine, RedWall-3) are somehow
too specific, that what is wanted by way of a suppressed premise in each of the arguments
is something of the level of generality of, respectively,
464 CRISPIN WRIGHT
Zebra-3* Zoo pens don’t often contain animals disguised to look other than they are.
and
Red Wall-3* Surfaces are not often illuminated by coloured lighting in such a way as to
appear other than their actual colour even though the presence of the
coloured lighting is otherwise visually unapparent.
Maybe these two premises represent additional information which would—in a given
context—be both needed and adequate, in conjunction with Zebra-1 and RedWall-1
respectively, to justify the transitions to Zebra-2 and RedWall-2; and maybe the more
specific Zebra-3 and RedWall-3 are not so adequate: after all, it is perfectly consistent
with Zebra-3, but in tension with Zebra-3*, that the animal in question be a cleverly
disguised Shetland Pony, and perfectly consistent with Red Wall-3, but in tension with
Red Wall-3*, that the wall in question be a pale grey wall lit by tricky red lighting. But
what in that case is the explanation, according to the Background Warrant model, of
the failure of transmission to the original Zebra-3 and RedWall-3? The core idea of the
model is, simply, that warrant doesn’t transmit to a conclusion that is, in effect, a
suppressed premise in an enthymematic argument; such arguments are question-
begging. But Zebra-3 and RedWall-3 are not plausible candidates to be the relevant
suppressed premises. The plausible candidates are Zebra-3* and Red Wall-3*. So the
letter of the model does not apply. What, to repeat, is the story about the failure of
transmission to the original conclusions, Zebra-3 and Red Wall-3?
A proponent of the model might try saying that warrant doesn’t transmit to a
conclusion which is independently supported by a suppressed premise in the argument.
But that will not do. Something can be an independent (ampliative) consequence of such
a premise and at the same time have its credentials enhanced by the additional informa-
tion encoded in the explicit premises—so that intuitively an (additional) warrant for it is
provided when the argument to it is run from the full premise set. Consider, for example,
Mushroom-1 The white flesh of this mushroom reddens when bruised
So, probably:
Mushroom-2 This mushroom is a specimen of The Blusher (Amanita Rubescens)
So:
Mushroom-3 This mushroom belongs to the Amanita genus.14
As fungi enthusiasts will know, the inference from Mushroom-1 to Mushroom-

2 would be pretty chancy without collateral information; there are many white-fleshed
14
A genus that includes many of the most famous and stereotypical toadstools, including the red and
white spotted Fly Agaric, the Death Cap used by Agrippina to poison Claudius, and the brilliant white
Destroying Angel.
REPLIES 465
fungi that display a reddish bruising effect, including Agaricus Campestris, the common
Field Mushroom. But add the premise
Mushroom-1* This mushroom has the characteristic appearance of many examples of the
Amanita genus, including the bulbous-based stem, and spotty umbrella-shaped cap,
and the inference to Mushroom-2 becomes very strong (though still defeasible: there
are other, rarer, blushing Amanitas). And while Mushroom-1* independently supports
Mushroom-3, it is plausible that this support can be strengthened by adding Mush-
room-1—as it will be if false Amanita-lookalikes are more common than false Amani-
ta-lookalikes that blush, i.e. false Blushers.15
There is thus some doubt whether the Background Warrant model actually has
anything to say about Zebra, Red Wall, and others among the standard examples.
The impression that the Background Warrant model misunderstands the nature of
warrant transmission failure is strengthened further by the third of the considerations
I advertised: the fact that warrant transmission failure is a feature of arguments where
there is simply no room for the idea that additional suppressed premises are at work. All
the three-step illustrations that we have so far considered involve a defeasible inferential
step from a 1-proposition to a 2-proposition, and the idea that background premises
may be involved gets a grip only and purely in virtue of this feature—simply, additional
information may be needed to rationalize this defeasible inferential step. So, what if
there is no such inferential step? What if we start with the 2-proposition, thought of as
grounded not on the basis of ampliative inference from a 1-proposition but as the
output of some non-inferentially conceived capacity of cognition—perception, for
instance, or memory? After all, it would be philosophically controversial, at best, to
represent the epistemology of Zebra-2 as always inferential in the kind of way depicted
by representing it as arrived at via something like Zebra-1. Cannot one normally just
look and see? When you ask a small child, “And what are those animals, Jenny?”, you
are seeing whether she can recognize them by looking, not whether she can construct
an adequately supported ampliative inference. But an inference that starts with
Zebra-2 That animal is a zebra,
conceived as non-inferentially warranted by casual observation, and then proceeds to

Zebra-3 That animal is not a mule cleverly disguised to look like a zebra,
is intuitively no more warrant transmissive than the original, and for just the same
reason: the subject has no warrant for Zebra-3 based purely on the visual observation
of the animal; the relevant possibility of a cleverly disguised mule is precisely that
of an animal fashioned to be visually indistinguishable from a zebra. When, however the
inference starts with Zebra-2, it is too late (there is no need) for a suppressed premise.
15
It needs remark that the considerations raised by this kind of example, and by the preceding discussion
are effectively noted in section III of Pryor’s essay. I am not sure why, in the light of these “tricky details”, he
seemingly retains confidence that a successful charge of transmission failure waits upon the applicability of the
Background Warrant model.
466 CRISPIN WRIGHT
Transmission Failure (II)—Authenticity-Conditions

As Pryor notes, I have in previous work16 proposed two broad ‘templates’ for
the characterization and diagnosis of non-transmissive arguments. The first of these,
the information-dependence template,17 was designed for the case where the warrant
for the premises of an argument is itself grounded in an ampliative inference from a
defeasible evidential base. The second, the disjunctive template, was an attempt to cope
with the kind of apparent transmission failure cases, just remarked, where the premises
of the argument are not inferentially grounded at all but conceived as delivered directly
by the operation of some appropriate cognitive capacity. The information-dependence
template proposed, roughly, that warrant does not transmit to a conclusion which one
needs already to know, or justifiably believe, if one is to be justified in regarding one’s
evidence for its premises as adequate. The disjunctive template proposed, roughly, that
warrant does not transmit to a conclusion a suspicion of whose falsity would rationally
require doubt that the conditions are suitable for the effective working of the cognitive
faculties relied upon to accomplish warrant for the premises in the first place.18
Pryor suggests that the two templates can each be seen as instances of the Back-
ground Warrant model. I think he is right that it is desirable—and should be possible—
to subsume them under a single account (though not, if the reservations I have lodged
against it are sound, under the Background Warrant model). I’ll briefly indicate what
I think is the right direction of generalization.
First, let us introduce a very broad notion of a cognitive project. A cognitive project is
defined by a pair: a question, and a procedure one might competently execute in order to
answer it. Thus there is a cognitive project associated with the question “What’s the
weather like today?”, which one can execute by looking outside; another cognitive project
associated with the same question which one can execute in a windowless room by looking
at the forecast in yesterday’s newspaper. There is a cognitive project associated with the
question, “Where will you spend July?”, which one can execute by an operation of
whatever it is one does to know of one’s intentions. There is a cognitive project associated
16
Wright 2000, p. 143; Wright 2002, }IV; Wright 2003, p. 59; Wright 2004, p. 171.
17
Here is the characterization offered at Wright 2002, pp. 335–6:
A body of evidence, e, is an information-dependent warrant for a particular proposition, P, if regarding e as
warranting P rationally requires certain kinds of collateral information, I. Some examples of such e, P and I
have the feature that elements of the relevant I are themselves entailed by P, (together perhaps with other
warranted premises). In that case, any warrant supplied by e for P will not be transmissible to those elements of I.
Note the play in this formulation with what is required if one is to regard e as warranting P. Potential contrast:
what is required if e is to warrant acceptance of P. This will become important shortly.
18
Thus, a suspicion that the animal in question is a cleverly disguised mule would require doubt whether
the circumstances are suitable for identifying animal species simply by looking. The first formulation of the
disjunctive template is in Wright 2000. That was defective in a way noted by Brian McLaughlin, and
remedied in Wright 2003. Subsequent further vicissitudes of the disjunctive template are described in Wright
2011, pp. 88–92. The descendant of it put forward in that discussion was designed to unify its precursors with
the information-dependence template. I shall outline its general motivation below, but I will not in this
discussion be concerned with its detail.
REPLIES 467
with the question, “Are there planets associated with that star?”, which one can execute by
the operation of a radio telescope and a suitable interpretation of one’s findings.
Next, let an authenticity-condition19 for a given cognitive project be any condition
doubt about which would rationally require doubt about the efficacy of the proposed
method of executing the project, or about the significance of its result, irrespective of
what that result might be.20 That my visual system is working properly, or that
yesterday’s weather forecast is likely to be accurate, are among the conditions for the
authenticity of the two mentioned projects for determining the weather; that I am
likely to have a normally lucid awareness of my intentions is an authenticity-condition
of the project focused on my prospective whereabouts in July; and that my radio
telescope is functioning properly, as well as a whole load of theory about electromag-
netic radiation, are authenticity-conditions for the project about the distant star.
In general, authenticity-conditions thus include such things as: normal and proper
functioning of relevant cognitive faculties, the reliability of instruments utilized, the
amenability of the circumstances to the proposed method of investigation, the correct-
ness of relevant theory, the soundness of relevant principles of inference utilized in
developing and collating one’s results, the good standing of relevant concepts used in
any aspect of the enquiry, and so on. Such conditions are typically things one takes for
granted in cognitive projects ranging from the mundane, like glancing at a clock to see
if it is time to leave, to the more methodologically self-conscious, like a carefully
controlled scientific experiment. But for cognitive projects that involve ampliative
inference, authenticity-conditions will also encompass conditions that are necessary if
the evidence encoded in the premises of such inferences is to have its intended bearing
on the conclusions. It is, for example, rational to take the look of a wall as good
evidence for its colour only if one has no reason to doubt that the lighting conditions
are such as to allow the real colours of surfaces to be revealed in how they look.
Likewise, it is rational to take my intentions as good evidence of what I will end up
doing only if I have no reason to doubt that conditions are, in relevant respects, such as
not to obstruct my carrying out my intentions. It is this generality in the notion of an
authenticity-condition—encompassing both conditions for effective cognitive func-
tion and conditions of evidential relevance—that underwrites the kind of unifying
generalization of the two templates that I want to canvass.
My basic suggestion about transmission failure, in the most general (somewhat
overstated) terms, is this: that you cannot rationally transmit a claim to warrant for certain
19
Elsewhere I have more often used the term, “presupposition”, for this notion. But of course it is a major
question whether, or in what sense, the satisfaction of conditions of this kind is indeed in large measure
presupposed in ordinary cognitive commerce. So the more neutral “authenticity-condition” seems better.
20
As the attentive reader will have spotted, this qualification is crucial if we are to capture the intended
notion. If it were waived, a prior opinion about the matter under investigation might rationally lead one to
doubt the competence or significance of the project if that opinion happened to conflict with the upshot; and
that would then have the effect that the very proposition which the project found in favor of would rank as an
authenticity-condition of the project concerned.
468 CRISPIN WRIGHT
premises across an entailment to anything that is an authenticity-condition for (any among) the very
cognitive project(s) which purportedly generate(s) that claim to warrant. There had better be
nothing in general amiss with the idea that merely looking at an animal in a zoo
enclosure, or looking at a wall, can put one in a position to claim warrant that the
animal in question is a zebra, or that the wall is red. But you cannot, merely by looking at
the animal, or the wall, get in position rationally to claim that conditions are suitable for
the identification of animal species, or surface colours, just on the basis of unaided vision.
That conditions are so suitable is a general authenticity-condition for each of the two
projects concerned. But so are the more specific consequences of that condition detailed
in their respective 3-propositions. Neither the general condition, nor the specific forms
of it detailed in the relevant 3-propositions, are open to merely visual confirmation.21
Now, the general run of authenticity-conditions for a cognitive project that putatively
eventuates in a warrant for a statement S will be logically independent of S. That my eyes
are working properly is logically independent of the wall’s being red; that I am in a state of
normal cognitive lucidity sufficient to enable the identification of my own intentions is
logically independent of my being about to drive my car away from Lot 15. But Cohen’s
bootstrapping manoeuvre illustrates how one particular kind of authenticity-condition—
the reliability of a source on a particular occasion—becomes deductively accessible if one
pools the truth of S, = the deliverance of the source on that occasion, with the fact that
the source delivered it. And the type of case typified by Dretske’s examples illustrates a
different ploy (Sceptical Hypothesis Negating, or Skyping) to the same effect: whatever
S may be, it will always be straightforward to construct a specific sceptical hypothesis
whose obtaining would falsify an authenticity-condition for the achievement of warrant
for S by a particular kind of cognitive project and whose negation, which will thus also
qualify as an authenticity-condition for the project concerned,22 will actually be entailed
by S. Where M is the method, or combination of methods, constitutive of the project
concerned, we can construct such a sceptical hypothesis merely by formulating some-
thing that realizes the rubric: not-S is true in circumstances that render its truth undetect-
able by M. If the basic suggestion about transmission failure outlined is correct, we will
therefore return the same verdict about both kinds of case: bootstrapping arguments, no
less than skyping arguments, fall foul of transmission failure.23
But how might one argue that the basic suggestion is correct? A rigorous demonstra-
tion would be possible only within the compass of a developed theory of evidential
21
So, the basic suggestion I am canvassing comes close to Pryor’s Anti-underminer model of transmission
failure. See his essay at pp. 298-300. I will react to his counterexamples below.
22
It follows from the characterization of an authenticity-condition that the negation of any proposition,
Q, which is inconsistent with an authenticity-condition, A, for a given project is itself an authenticity-
condition for that project. For then to the extent that you doubt not-Q, you must raise your credence in Q,
and correspondingly in not-A. So doubt of not-Q must impact upon the good standing of the relevant
project to exactly the extent that doubt of A does.
23
—and do so at the stage of the sub-proof, before the accumulation of any putative ‘track-record’ of
success. Objections to the warrant-generative capacities of bootstrapping reasoning that focus on its inductive
aspect are therefore beside the (principal) point.
REPLIES 469
support and rational belief management whose basics would require, and ultimately
admit of no more than, intuitive motivation. So I think we have to settle, au fond, for
intuitive plausibility. But the tabled suggestion is, I think, very plausible when taken in
a certain kind of epistemically reflective setting. Suppose you are about to undertake an
investigation, using method M, into a particular question—let’s say: determining the
capacity of a garden water-tank by measuring its sides and multiplying. One authen-
ticity-condition for the effectiveness of M is the correct calibration of your tape
measure. Another is whether it can be depended on to be stable in length through
the successive measurements. A third is that the dimensions of the tank be, near
enough, regular (that each of its faces is, near enough, rectangular). A fourth is the
steadiness of your hand. And so on. You would presumably, if prompted, recognize
each of these conditions as germane and would have some significant prior degree of
confidence that they are likely to be satisfied if you remained willing to undertake the
investigation in the described way. Now, you might run the investigation and, as a
result, come rationally to have some increased measure of uncertainty about any of
these conditions. Perhaps your results on a single edge of the tank fluctuate. Perhaps
your results on corresponding edges diverge. So, running the investigation, you can get
into a position when you may justifiably come to increased doubt whether all the
conditions that need to be satisfied in order for your project to deliver reliable results
are simultaneously met. By contrast, there seems to be no question of running the
investigation and, as a result of what you find, rationally taking yourself to have
acquired grounds for increased confidence about that—increased confidence that every
one of the various relevant authenticity-conditions, a doubt about which would
occasion rational misgivings about the status of the results of the project, is satisfied.
This point needs careful statement. A project may develop in such a way that you do
get to refurbish your confidence, en route, in one of its authenticity-conditions or
another. You may get evidence that the tank is indeed regular, for example (and hence
that its capacity is determinable by a simple multiplication), by obtaining mutually
consonant results for each of its three relevant dimensions, even on a badly calibrated
tape measure. But you cannot even do that unless the tape measure is at least stable.
You may get evidence that the tape measure is stable—or stable enough—by obtaining
a series of mutually consonant results on a single edge. But you cannot do that unless
the edge itself is stable in length . . . The authenticity-conditions pop up like Hydra-
heads. Even if, in the course of the investigation, things occur that reinforce your
confidence in one or another amongst them, you will have to acknowledge that, in
taking that to be the correct assessment of the effect, you will be investing in a
significant prior degree of confidence in other such conditions about which nothing
has occurred that should bolster your confidence. And your prior level of confidence in
these conditions will set a limit on how confident you should rationally allow yourself
to be about the overall significance of the investigation, and the good standing of the
results that it returns. It would be quite irrational, for example, to invest more
470 CRISPIN WRIGHT
confidence in your finding about the stability of the tape measure through the relevant
measurements than you have in the stability of the measured edge.
The general point is that any cognitive project will have authenticity-conditions in
which nothing that happens in the course of the project should, if one is rational and
clear-headed, improve one’s prior credence. But now that limit will apply even if there is
some trick one can run—like bootstrapping, or skyping—to bring an authenticity-
condition within the deductive light-cone, as it were, of the very results of the investi-
gation. In particular, it would be absurd to profess oneself open-minded about some such
condition at the outset of the investigation and then to find reassurance in its entailment
by results whose good standing in turn one had recognized to depend upon its satisfac-
tion. If you had started out open-minded about whether the condition was met, you
should be open-minded about the standing of those results. And in that case, you should
be open-minded about whether you have any warrant to transmit to their consequences.
So here again is a rather more cautious statement of the general point. Any cognitive
project—any question and method pair, P? and M—will be associated with a more or less
open-ended set of authenticity-conditions. And some of these—call them M-inaccessible
authenticity-conditions—will be such that one can recognize in advance that the execu-
tion of the project will do nothing, except coincidentally, to rationally raise one’s credence
that they are satisfied. A proper subset of the latter may be included among the logical
consequences of P. Should the project find in favor of P, the claim to warrant for P
thereby afforded will not transmit to any among these M-inaccessible consequences of it.
I shall not here attempt to make any general proposal about how exactly it may be
settled that a particular authenticity-condition is or is not in this situation for given P?
and M. At the least, it is going to be an information-relative matter. For example,
background information may make it possible for the mere visual appearance of
Dretske’s zebras to contribute towards a warrant that they are indeed not cleverly
disguised mules. (See note 4.) The examples of transmission failure standardly canvassed
in the literature are all cases where the conclusion of the argument in question would
be expectably inaccessible by the method of investigation of the premises understood to
be in play, but where suitable additional information might change that. I leave it as an
exercise to the reader to devise informational settings that would turn the trick for
others of the standard examples.24
24
“Pryor’s counter examples to the Anti-Under minor model present the obverse case: examples where,
on tacitly understood assumptions of normality, warrant does transmit to an authenticity-condition but
where transmission failure can be introduced by tinkering with those assumptions. For example, in the
argument (this volume, p. 299):
Rat-1. [Visable experiences of rats]

Rat-2. I see rats in the corner
Rat-3. So there are some visable animals in the pen.
Rat-3 does indeed present an authenticity condition, as characterized earlier, for the evidential relevance of the
experience described in Rat-1 to the statement Rat-2. And, as Pryor reports, it does seen that in normal
REPLIES 471
The reader will have marked the shift over the preceding few paragraphs to a
preponderance of talk of the transmission of claim to warrant. This is of course fully
deliberate, and significant. Pryor opines at one point that there is not “enough in the
texts to support reading much in to Wright’s higher-order-ish talk”.25 I own that there
are elements of equivocation in my earlier discussions, but my considered view is
that insofar as a diagnosis of transmission failure is rendered intuitive by the basic
examples, the intuition is of a failure appreciated at the level of reflective epistemic
responsibility: thus, you couldn’t sensibly regard, or offer to others, your intention as
evidence that conditions obtain that will enable you to carry it out; you couldn’t
sensibly regard, or offer to others, the mere look of the stripy equines as evidence that
the stripes are not a sham; you couldn’t sensibly regard, or offer to others, the activity if
the FBI-jacket wearing men as evidence that they are not part of a scene for a film. The
examples that supposedly provide intuitive support for the reality of transmission failure
phenomena in the first place are all most naturally interpreted in this kind of way.
Further clarification is wanted, naturally, of what the higher-level notion comes to—
what it is to ‘be in position to claim warrant’. As I intend the notion, it demands a
potential for reflective awareness that one is warranted and hence, at least if it is allowed
that ordinary—first-level—warrant may be conferred by external circumstances beyond
reflective awareness, is not the same thing as having warrant for warrant.26 But I cannot
here encroach on the task of trying to provide a more satisfyingly explicit account.
Suffice it to say that unless we take the view—which some internalists may but many of
externalist sympathies will not—that to be in possession of warrant for accepting a certain
proposition always ensures being in a position to lay claim to warrant for accepting that
proposition, it is going to be an additional step to argue that the standard examples can
serve equally well as examples of warrant transmission failure at first-level.27
The matter is of some considerable significance. There is, in particular, a substantial
issue whether the debate between so called liberals and conservatives over perceptual
justification may not turn on an ambiguity of focus between issues concerning first-level
warrant and second-level warrant respectively, with the possible result that Moore’s
Proof, for instance, might indeed be serviceable for the provision of warrant to believe in
an external world while useless for underwriting claim to such warrant, so useless in the
context of epistemological debate, because non-transmissive at second-level.
circumstances the argument would transmit justification to Rat-3. But imagine circumstances in which we have
reason to suspect that our mule-disguising zookeeper has also installed hallucination-engendering radiation in
many of his empty pens, to give the impression of a well- and interestingly stocked zoo (the ‘rats’ are built in for
extra plausibility). Then we might want reassurance of Rat-3 enroute to investing in any confidence that the
experiences described by Rat-1 provide good evidence for the cognitive accomplishment marked by Rat-2.
25
This volume, p. 294.
26
For if warrant is nothing of which one need be capable of being aware, nor is warrant for warrant.
27
Let the reader beware of confusing issues concerning the transmission of (various form of ) higher-order
warrant with those raised by Pryor when he introduces consideration of what he calls the Higher-Order
Background Warrant Model (this volume, pp. 295–7). The latter is a model that proposes that there are
constraints on the transmission of first-order warrant that involves higher-order warrant for certain of the
background assumptions. That is an interesting idea, but it is not what I have been talking about in the
preceding paragraphs here.
472 CRISPIN WRIGHT
Liberals and Conservatives

We can characterize an inference as Moorean if its conclusion is an authenticity-
condition, of one kind or another, for the attainment of warrant in some envisaged
way for one of its premises or lemmas. Since the existence of an external material world
is an authenticity-condition both for the evidential relevance of one’s sensory experi-
ence to propositions about the characteristics of local material objects and for the
effective function of one’s perceptual faculties, Moore’s Proof —whether its premise,
“Here is one hand”, is taken as defeasibly evidentially grounded or as directly known—
is a Moorean inference par excellence.
Liberalism and Conservatism are best characterized with respect to quadruples
consisting of a target proposition, P; some envisaged form of evidence or cognitive
achievement, C; an authenticity-condition, A; and a species of epistemic warrant,
E. The Conservative view will be that the accomplishment by means of C of a warrant
of type E with respect to P will require as an enabling condition that one be in E (or
some other specified warrant state) with respect to A. The Liberal will demur, holding
that the attainment of E for P by means of C need rest on no positive form of warrant
for A—that it will suffice merely if the agent has no warrant for not-A.
Suppose, for example, that E is knowledge, and the first three elements in the
relevant quadruple are those in Moore’s Proof, considered as the transition:
Moore-1 (C) My experience is in all respects as if I was holding my hand in front of my face
So probably:
Moore-2 (P) Here is a hand So:
Moore-3 (A) There is an external material world.
Then Liberalism with respect to this particular quadruple says that, in order to
accomplish knowledge of the proposition P on the basis of the evidence, C, I do not
need to know, or independently have adequate grounds for accepting A, that there is
an external material world. It is enough merely that I have no antecedent reason to
doubt it. Conservatism, by contrast, will hold that it is only in an epistemic context
where I have a specified kind of prior warrant for acceptance that there is a material
world that my evidence C has any tendency to support P, that here is a hand.
Obviously there is going to be space for such a clash of attitudes with respect to the
quadruple associated with absolutely any Moorean inference. Consider again:
Testimony-1 (C) I ask Smith the time and, glancing at his watch, he says, “Eight o’clock”
So, probably:
Testimony-2 (P) It’s eight o’clock
So:
Testimony-3 (A) Smith is telling the truth on this occasion.
REPLIES 473
Let E again be knowledge. Then in order, the Conservative will say, to achieve
knowledge that P on the basis of S’s testimony C, I must independently know, or
have adequate grounds for accepting A, that S is telling the truth on this occasion. For
the Liberal, by contrast, knowledge that P can be achieved, in the best case, purely on
the basis of S’s testimony; no antecedent knowledge of or other form of warrant for S’s
truthfulness is needed.
The relativization of the debate to the appropriate kind of quadruples is important. For
one thing, it allows that even when C, P, and A are fixed, the merits of a conservative, or
liberal stance may vary as a function of E, the kind of epistemic warrant that is involved.
Knowledge, for example, may be taken to be more exigent than other forms of epistemic
warrant exactly in that it demands warrant for authenticity-conditions in particular cases
where they do not. (This is one version of the thought behind ‘Russellian Retreat’.) But
perhaps more important for my present purpose, it leaves provision for the possibility that
one may, with respect to a particular kind of quadruple, be conservative concerning the
conditions for higher-level warrant—claim to warrant—while liberal about the condi-
tions for the (mere) possession of warrant.
Clearly there can be no general opposition between “the liberals” and “the con-
servatives”. No one is going to be liberal right across the board, for all types of
proposition, all kinds of defeasible evidence or cognitive accomplishment that may
bear on them, all authenticity-conditions for that particular bearing, and all forms of
epistemic warrant. Such a view would condone an open flood of epistemic irresponsi-
bility. The interesting, disputed questions concern for which selections of the four
parameters conservatism is appropriate, and why, and for which selections the more
relaxed stance of liberalism is perfectly rational.28
At the time I write this, it is still true that most liberal–conservative debate has been
focused on the nature of the warrant provided by a perceptual experience for beliefs
about the local environment. But it is striking that the considerations which, respec-
tively, most powerfully motivate the opposed views do seem to belong at different
levels. The most powerful consideration on behalf of dogmatism—liberalism about basic
perceptual warrant—is that we do not wish to deny the title of warranted belief to
opinions that children, and others who are relatively epistemologically innocent, form
without considering, let alone marshalling evidence to discount the kind of possibility
typified by hallucination or the artful disguise of mules. So it seems we think that the
acquisition of perceptual knowledge, or other forms of perceptual warrant, doesn’t
28
Perhaps, indeed, a fifth parameter is called for: that of the context of interests in which the evidence in
question is assessed, or the relevant putative cognitive achievement takes place. Certainly, it is quite plausible
enough that, under the aegis of a broadly conservative view of the role of a particular authenticity-condition, A,
how much, and what quality, of independent evidence for A is required in order for belief in the target
proposition to be warranted may very well vary as a function of the interests of the believer (or the attributer, or
an assessor . . . ). It is not so clear, though, that such a change of context can transform a case where a liberal view
is appropriate, that is, when no independent warrant is required for A, to one where conservatism is appropriate
and independent warrant is required. But I shall not further consider this kind of complication here.
474 CRISPIN WRIGHT
require the kind of epistemological ‘policing’ of authenticity-conditions that conser-

vatism suggests. But it is different when we adopt a stance in which we undertake to
scrutinize our claims to perceptual knowledge, or warrant. Then we seem obliged
either to take a positive view of any authenticity-condition that may be entered into
the conversation, or to qualify our claim to warrant. Standing before the zoo pen and
inclined to claim knowledge that the animals before me are zebras, there is no ducking
the question, “So you are taking it that your visual system is functioning satisfactorily
today, and that those animals are not artfully disguised members of another species?”
And the intuitively needed answer if the claim to warrant is to be sustained will be not
that “I have no view about those matters, and am not required to have one. It’s enough
that I have no reason to doubt either condition”, but rather, “Of course”. So there
does seem to be an undeniable difference between what we want to say about the
conditions governing the acquisition of perceptual warrant and what we want to say
about the conditions under which such a warrant may rationally be claimed. And the
conservatives seem to have the better of the issue when what is in question is the latter:
one cannot rationally profess agnosticism about something one acknowledges to be an
authenticity-condition for a route to the acquisition of a warrant which one simulta-
neously enters a claim to have travelled.
However, the difficulty with conservatism, whether at first or second level, is that—
unless curtailed at some point—it threatens to set impossible standards for the acquisi-
tion of warrant, or the acquisition of the right to claim it. For if a subject is to be
required to possess independent warrant for the satisfaction of a relevant authenticity-
condition, or for the right to claim that it is satisfied, then that seems to demand the
satisfactory completion of a prior independent cognitive project. And that in turn will
have its own authenticity-conditions. If a conservative attitude is taken in turn towards
them, then do we not launch a regress of which the upshot must inevitably be (first- or
second-level) scepticism?
I don’t know if the predicament has often been expressed exactly along the
foregoing lines. But there is, in the current epistemological milieu, a widely supported
potential response. The response is that of externalism. If being in position to claim
warrant is, as I at least intend, a matter to be ascertained by an exercise of reflective
epistemic conscience, then provided warrant itself is internally construed—as a condi-
tion always apprehensible by, roughly, the exercise of psychological self-knowledge
and a priori reflection—a suitably conceptually endowed and reflective subject is going
to be able to get into a position to claim warrant just when she has one. And in that
case, to the extent that we regard the second-level, conservative intuitions as robust,
we will have to regard the first-level intuition—the relative ease with which perceptual
knowledge is available to the epistemologically unsophisticated—as misleading, or at
best in need of qualification. Children, we might then say, can at best get into
perceptual states where, had they the appropriate conceptual resources and were
they able to understand and police relevant authenticity-conditions, they would possess
perceptually warranted, or knowledgeable, beliefs. They can, as the matter is some-
REPLIES 475
times expressed, accomplish propositional warrant. But their perceptually based beliefs
would not be doxastically warranted. To describe their state as knowledgeable, in particu-
lar, would be a kind of (parental) indulgence. But the situation changes if perceptual
warrant is regarded as an external matter—a question of “situational provenance”, of
the nature of the circumstances and capacities drawn on in the formation of the
relevant belief. On such a view—perhaps a simple reliabilism, perhaps something
more elaborate—the conditions for the acquisition of perceptual warrant can be as
liberalism conceives them: there is no need for the policing of authenticity-conditions.
It is only when we go to the level of reflection and the vindication of claims that our
epistemic consciences demand that policing be reinstated, and the ante be upped. To
the extent that we are then persuaded that the demands of conscience are regressive and
insatiable, there is nothing for it but to draw the gloomy conclusion that full intellectual
conscientiousness—the taking of a thoroughgoing epistemic responsibility for all one’s
beliefs—is an unattainable ideal. One must merely hope for the best. Specifically, one
must hope that where, in the absence of demonstrated warrant for a recognized
authenticity-condition, the right to claim warrant has not been made out, the external
situational provenance of one’s belief, or beliefs, that have come under scrutiny is
indeed of a relevantly felicitous kind. And at that point there is little left to do but to
give the famous Strawsonian shrug.
The “Path of Entitlement”, as I conceive it, tries to do better than a Strawsonian
shrug. The basic idea is very straightforward. If an authenticity-condition for a cogni-
tive project, culminating in a particular belief that P, is something which one may
rationally take for granted—a circumstance which one may implicitly rationally trust to
obtain—then there would be a sense in which the belief that P can be warranted by the
outcome of the project even for a subject who does not, and perhaps could not,
consider the status of the condition in question. At first level, it will be provided for that
the subject may proceed to the acquisition of knowledge, or warrant exactly as would
be allowed—with no more cognitive work required than—if liberalism was true,
without giving any thought to the status of the relevant authenticity-conditions. But
at second level, at the level of laying claim to one’s epistemic accomplishments, so to
speak, one would have to take ownership of one’s commitment to the satisfaction of
the authenticity-condition in question, and that commitment would have to be in
good standing. The presupposition of the thought that connected conservatism with
scepticism was that this good standing would have to be a product of further enquiry.
But of course the point of the notion of entitlement is that that is not so: epistemic
entitlements are things that we may rationally trust to obtain without enquiry.
To render more vivid the way in which, as I would like to conceive it, the notion of
epistemic entitlement works to reconcile the apparently conflicting intuitions at first
and second level, it may be useful to make a comparison with the notion of a moral right.
An agent does not need to know her rights in order to have them. Indeed, she may
have no conception of a (particular kind of ) right. And when she acts in ways that her
rights mandate, her actions are in good standing even if she is unaware that they are so
476 CRISPIN WRIGHT
mandated or, though aware, unable to make out a cogent case that they are. To the
extent that they are conceived as having analogues of exactly these features in the
sphere of action constituted by the formation and management of belief, entitlements
may usefully be thought of as determining epistemic rights. In that respect, they contrast
with evidential warrants, which seem more naturally conceived as conferring epistemic
obligations: as determining what you ought to think. Thus, if there is indeed a general
entitlement to take it that, absent evidence to the contrary, one’s sensory faculties are
working normally in conditions broadly conducive to their effective operation, then a
young child, with no developed conception of sensory abnormality or illusion, who
forms beliefs spontaneously in response to the promptings of her sense experience, is
acting fully within her rights, and is, in that sense, justified in so doing, exactly as the
dogmatist first-level intuition requires. But a mature agent, challenged to produce a
warrant for so acting, will have to acknowledge a commitment to the obtaining of
whatever circumstances she can recognize as being such that their failure would
compromise the abilities that that the child is exercising. And a defence of the claim
that her epistemic actions are generating perceptual warrants will require defending
these commitments. That is the gist of the conservative second-level intuition. Scepti-
cism threatens when we see no other possibility than evidential justification for these
commitments in turn. But the point of the notion of entitlement is precisely that that is
not the only possibility.
A Third Way?
Here is an opportune place to consider Annalisa Coliva’s interesting suggestion that
liberalism and conservatism do not exhaust the possibilities. Liberalism holds that, for a
selected range of propositions, P, warrant achieved by a particular method for which
A is an authenticity-condition requires only that a thinker has no reason to doubt that
A is met, where “no reason to doubt” embraces the case, in particular, when he has no
information bearing on the matter one way or the other. No particular attitude to A is
needed. He need not consider A at all. Conservatism, by contrast, holds that in such a
case, the warrantability of the belief that P depends in addition on (some form of )
independent warrant to take it that A is met. Coliva’s suggestion is that, as against
liberalism, the agent does need to take a stance on A: he needs to assume that A is
satisfied. But this assumption is not something for which he requires warrant in turn. In
particular, he has no need for a “non-evidential warrant”.
I think there is a genuine additional possibility here, but it may be that the interpre-
tation of it that I will consider is not exactly what Coliva has in mind. For one thing, it
seems to me that Coliva’s suggestion is very much more easily received at second level
than at first level. That an unreflective thinker may acquire a perceptual warrant for a
particular belief just in virtue of the course of her perceptual experience, without any
consideration of authenticity-conditions and defeaters, is common ground both for the
dogmatist and for the conservative who regards the satisfaction of relevant authenticity-
REPLIES 477
conditions as a matter of entitlement. They can agree, in other words, about what the
subject has to do in order to obtain perceptual justification. Their disagreement is about
the supporting architecture of the perceptual justification thereby obtained: the dog-
matist holds that the warrant is conferred purely by the occurrence of the relevant
perceptual experience; the conservative holds that the perceptual experience confers
warrant only in a context in which there is either independent reason to believe that a
given authenticity-condition is satisfied, or a right to take it for granted—to act,
doxastically, in the manner which its acceptance rationalizes—even if the condition in
question is not explicitly considered, even if the thinker concerned is in no position to
consider it. In other words: although the entitlement-conservative augments his con-
ception of the justificational architecture of a perceptual belief with the thesis that
thinkers are rationally entitled to trust in satisfaction of the relevant authenticity-
condition, what he requires of the thinker if she is justifiably forms a belief on the basis
of her experience is exactly what the dogmatist requires. But on Coliva’s proposal, at
least on the natural understanding of “assume” as denoting a propositional attitude,
something more would seem to be required: the thinker will also have to make some
assumptions, whatever exactly ‘assuming’ is taken to consist in. Thus when interpreted
at first level, Coliva’s proposal can seem more demanding of the thinker than either
entitlement-conservatism or dogmatism—and consequently open to the children-and-
intelligent-animals kind of objection that moves the dogmatist in the first place.
I am not sure if it makes exactly the same point to observe that there seems to be no
way of taking the thesis that justification requires that a thinker make certain assumptions
as a thesis about propositional justification. A constraint that demands that the thinker be in
certain actual psychological states can only be a constraint on doxastic justification—on
the actual justification of actual belief. By contrast, as noted, dogmatism and entitlement-
conservatism are, at first level, theses about propositional justificational architecture.29
Matters are much more straightforward, though, if Coliva’s proposal is interpreted at
second level—at the level of claims. And that is how I am going to take it. The key
second-level intuition was that in order to stake a claim to warrant for a perceptual
belief, a fully reflective, explicit thinker does need to take ownership of anything she
recognizes as an authenticity-condition and, if she can muster no evidence on its behalf,
to acknowledge that she is taking its satisfaction on trust. I read Coliva as intending to
respect this intuition when she speaks of a requirement of “assumption” and accord-
ingly, thus far, as marching in step with the conservative. What sets the ‘third way’
apart is rather, at a first approximation, that—apart of course from the thinker’s having
the relevant perceptual experiences—that is all that needs to be in place. No further
29
The reader should note, however, that Coliva herself gives sympathetic mention to the children-and-
intelligent-animals point. (This volume, p. 325.) This suggests that her requirement that authenticity-
conditions be assumed is indeed intended to operate as a constraint on propositional justification, and
hence that she does not intend ‘assumption’ to denote an actual psychological attitude. But I won’t speculate
further on this here.
478 CRISPIN WRIGHT
condition has to be met—in particular, there is no call for some kind of further, non-
evidential warrant for the assumptions concerned.
Taken just this far, Coliva’s third way may be taken to coincide with a view
associated with Harman30 to the effect that all enquiry requires assumptions, and, in
the absence of specific evidence pro or anti, there is nothing rationally to choose
between one such set of assumptions and any other; so start where you like, as it were,
and trust to pragmatic providence to filter out the wheat from the chaff. But I do not
think that this is Coliva’s proposal, nor do I think it particularly plausible. If a form of
enquiry is proposed into some subject matter the credibility of whose results will
depend on certain ad hoc assumptions about which there is not a jot of evidence, for
or against, our intuitive sense in general will be that those results can acquire an at most
conditional credibility. If we are to get past this conditionality, the assumptions have to
be ones that we consider it reasonable to make. The Harmaniac idea fails to engage with
this point. The essentially situated—assumption-dependent—character of all enquiry
seems to enjoin that there has to be a kind of reasonableness which is not evidence-
dependent if we are ever to achieve an outcome which we may sensibly regard as more
than conditional.
So, isn’t that a reason to move to the entitlement-conservative conception of the
matter, and so to pursue the entitlement project? Coliva’s key thought, if I interpret her
correctly, is that it is not. It is a reason to accept that there is a distinction between
reasonable, or rational, assumption and unreasonable, or irrational, assumption that is
not to be drawn in terms of a relatively favorable evidential situation. But that is not to
say that the assumptions that it is reasonable to make without evidence on their behalf
are assumptions that possess “non-evidential warrant”. Rather, she suggests, there are
certain assumptions that are constitutive of rational empirical enquiry. It is not that
making them is sustained by certain special considerations that serve to explain why it is
rational so to do. Rather, rational empirical enquiry simply is an activity in which these
assumptions are made and allowed to govern the enquirer’s conception of the eviden-
tial significance of various types of occurrence. To ask why they are rational—Coliva
doesn’t say exactly this, but it would seem to be in keeping with what she does say—is
to ask a question incorporating a mistake very similar to that made by someone who
asks what it is in the nature of Chess that mandates playing it on a board of 64 squares,
8 8, alternating black and white.
Coliva cites a Wittgensteinian inspiration for this suggestion. And there are certainly
passages in his later writings, especially on mathematics, that seem to belong on this
page. But though the tendency is there, I am doubtful whether it can provide a full,
intellectually satisfying response to sceptical paradox. Suppose I ask, in a sceptical
moment, why is it rational to expect propensities exhibited in our experience hitherto
to be sustained into the future? And the reply comes: “Well, there is certainly no
30
See e.g. Harman 2003.
REPLIES 479
evidence you can marshal to justify that conviction without in effect presupposing that
inductive practice is rational (Hume’s insight). But the truth is that the rationality of
inductive inference is primitive, and in part definitional of our concept of rational
empirical enquiry. There is therefore insufficient distinction between the notion of
rational empirical enquiry in general and the notion of an enquiry incorporating
inductive methodology to make intelligible space for your question.” This is the
kind of thought that—returning to her main example—I take Coliva to be advancing
with respect to the notion of rational empirical enquiry and interpretation of percep-
tual experience based on the assumption that there is an external material world.
If this interpretation is broadly correct, then Coliva’s third way is, in fact, a fairly
familiar line of thought: the so called “paradigm case” response to scepticism that had
some currency in the middle of the twentieth century. And if so, then—in the limited
space at my disposal here—let me merely flag a couple of considerations which
disincline me, for one, to this kind of response.
The first is that if it really were constitutive of our conception of rational empirical
enquiry to assume that there is an external material world, then there should be a kind
of unintelligibility about a sceptical challenge to the rationality of this assumption
which would be at odds with the sense of paradox created by the best sceptical
arguments that challenge it. There is, it seems to me, an implicit tension in the very
notion that elements which are constitutive of a concept—which belong primitively to
its identity and are not sustained by other features of it—should be sufficiently opaque
to be controversial and apparently vulnerable to philosophical challenge. If free action,
to take a parallel example, is, conceptually constitutively, simply action performed with
a sense of freedom, for normal human reasons, without external force or duress, why
does anyone feel the familiar kind of challenge posed by determinism as any kind of
problem? The reply should leap to one’s lips: “Well, but you misunderstand what
freedom is. Mere antecedent causation of your action, even by events before your
birth, is in no sense in tension with its freedom.” Of course, some philosophers have
said exactly that kind of thing. But the point is that their response is controversial; others
(including the present author) have found it beside the point. How can a thesis about
what is primitively constitutive of a concept be controversial? And how, if it can, might
it be recognized to be correct?
In parallel, while a proponent of the third way and an entitlement theorist will agree
that what the best sceptical paradoxes show is that our title to these takings-for-granted,
trustings, assumptions, and expectations cannot in the end be evidentially earned,
shouldn’t the sceptic’s transition from that acknowledgement to the conclusion that
they cannot be rational at all scream out as an obvious non sequitur if their title as
‘rational’ was conceptually constitutive of our notion of rationality—if there were no
further good question about what makes them rational? A proponent of the third way
needs to explain how features that are constitutive of our concept of rational enquiry
can nevertheless be sufficiently opaque to those who have mastered the concept to be
prima facie coherently questionable.
480 CRISPIN WRIGHT
My second, related reservation has to do with the question of what fixes the identity
of concepts with the kind of normativity—I take it to be relevantly similar—exhibited
both by concepts of epistemic rationality and concepts of morality. It seems to me, as to
many, that it is possible in principle for cultures to have enormously divergent moral
codes, major discrepancies in the things that they are prepared to classify as good, or
obligatory, without raising any significant question whether all are exercising genuine,
shared concepts of the moral good and the morally obligatory. Moral concepts can permit
all kinds of divergent and outré applications without any questions being raised,
necessarily, whether it is indeed concepts of the morally good and obligatory that are
being applied. So my suspicion—and without much more argument, that’s all it can
be—is that such concepts have, in effect, no paradigms, no canonical in-rules, as it were.
What unifies morally evaluative concepts across communities whose fundamental
moral standards are radically different is rather—very roughly—a common conception
of the consequences of classifying a type of action as moral.
I want to suggest that the same model applies to epistemic evaluation. And if that is
correct, then the sceptical challenge is not to be silenced by the suggestion that the
rational can only be what we most fundamentally call ‘rational’. The model does not
imply that there will always be a good challenge to explain why a particular kind of
action in particular circumstances is morally good; or why a particular pattern of belief
formation is rational. But the challenge is at any rate not to be stifled by the assertion
that it enters primitively into our concepts of the good, or the rational, that they
respectively embrace that kind of action, and that pattern of belief-formation.
If this is right, and if concepts of epistemic rationality are similar to moral concepts in
this respect, then the basic claim of the third way about the constitution of our concept
of epistemic rationality should give way to a thesis about the epistemic value of a
trusting reliance on induction, or an uncritical acceptance of the existence of an
external material world. The argument should be, not that the rationality of such
acceptances is part of what we mean by “rational” but, substantively, that they are an
essential part of any form of enquiry that is harnessed to the essential goals of enquiry:
truth, knowledge, the avoidance of error, understanding, and the construction of an
integrated, systematic and powerfully predictive framework of belief. I’ll say a little
more about this at the end.
Entitlement and Scepticism

Michael Williams’ thoughtful and probing essay displays many points of agreement
with the work of mine, primarily Wright 2004, on which he mainly focuses, but some
fundamental differences too. We are in broad agreement about the importance of the
distinction between those doxastic merits that have to do with intellectual conscience
and the attempt to attain reflective epistemic responsibility, and those conferred by
“situational provenance”—reliable faculties, working in suitable conditions, for in-
stance—and we agree that the proper focus of the sceptical challenge is on the former.
REPLIES 481
It is thus a challenge that arises at the level of claims to warrant in my preferred

terminology, or at the level of “scepticism about philosophy” in his. We are also
agreed that the challenge can be met: no “post-modernist” conclusion is licensed by
the best—or worst—sceptical paradoxes. And, while disagreeing about how intuitive
the premises are which those arguments exploit, I think we agree that some notion of
epistemic entitlement has a key part to play in seeing the paradoxes off.
There are in my judgment just three major points of disagreement. First, Williams
doubts that the right to trust, when grounded in the kinds of considerations marshalled
in Wright 2004, can avoid being exceedingly permissive, providing pari passu not
merely for entitlements to theological postulates, but to all manner of eccentric and
‘left-field’ notions. Second, he rightly observes that to address the essentially a priori
challenge of scepticism, the considerations that ground entitlements must be a priori
too—and he is doubtful whether I can consistently operate under that constraint. And
finally, Williams is suspicious of the static epistemological architecture which, as he
views matters, conditions my handling of the entitlement project—the framework of
rigid, once-and-for-all distinctions between types of statement, determined by subject-
matter, and types of evidence for them. This rigid framework, in his view, is required
for the form of scepticism that I called ‘I-II-III’ scepticism, or Humean scepticism, and
that he nicely calls Locality scepticism. He regards Locality scepticism, with some justifi-
cation, as part of the pernicious legacy of foundationalism. Maybe nobody any longer
believes in the system of our warranted beliefs as a great pyramidical structure, stratified
by types of subject-matter, with experiential confirmation flowing in at the bottom
and suffusing upwards via deeper and deeper theoretical inference. But in Williams’
view, I have kept too much of the image of this structure, repudiating in effect only the
foundationalist conception of the flow of confirmation from experience. In Williams’
view, we have to drop the foundational structure too. And we have to do so not by
way of a shift to Coherentism, but by way of a realization of the radically contextual
character of epistemic projects, and the corresponding contextuality of the status,
inferential or non-inferential, of the beliefs that they aim to vindicate or refute.
Williams’ view is that ‘statements about other minds’, ‘statements about material
objects’, ‘statements about the past’, ‘statements about the future’, do not form
epistemological kinds, and cannot therefore be open to any sceptical problem that
presupposes that their justificational architecture can be captured by a fixed template.
He has of course famously argued for this claim, and for the associated epistemological
contextualism, at length elsewhere,31 but since he does not go into the details in his
present contribution, I should not take up the issue here. However, I actually doubt
that a great deal hangs on it in the present context. The significance of the distinction
between Cartesian scepticism and Locality scepticism is that they provide for
generalized forms of sceptical challenge that respectively apply to the acquisition of
31
Williams 1991.
482 CRISPIN WRIGHT
non-inferential warrant, in the case of Cartesian scepticism, and of defeasible inferential

warrant, in the case of Locality scepticism. As far as I can see, it makes no odds whether
inferentiality/non-inferentiality are absolute (subject-matter determined) or contextual
properties. The Cartesian and Humean arguments provide schematic recipes for
challenging claims to warrant of those respective kinds, whatever settles what kind of
warrant a statement is eligible for. It doesn’t help to address them if the type of warrant
for which a statement is deemed eligible is determined not once-and-for-all, by its
subject matter, but by variable elements of epistemic context.32
Williams thinks that Humean—Locality—scepticism is in the end untroublesome
because he doubts that there is any such thing as cognitive locality in the sense
presupposed by standard presentations of the sceptical argument. I reply that contextual
locality is enough to generate the problem. If, in a given context, the best evidential
grounds we can muster for a certain belief are ampliative-inferential, there will be
authenticity-conditions which, under challenge, we will need to endorse in order to
accredit the inferential grounds concerned. The question is then going to arise with
what right we take those conditions to be satisfied. The response can’t always appeal to
evidence—we won’t always have any relevant evidence. And if it is, instead—what, if
I interpret Williams correctly, he will want to say—that epistemic contexts come with
a ready-made entitlement to take for granted or trust in things that define them, that is
not a response to which I am unsympathetic. But it is a response that needs the support
of a theory of entitlements that explains their contextual character.33
My expectation, though, would be that some of the requisite takings-for-granted—
precisely the Moorean or metaphysically ‘heavyweight’ propositions34—are going to
have to enjoy an inter-contextual entitlement status, things we may rationally take for
granted across a huge sweep of investigative contexts. If Williams agrees that these
metaphysical heavyweights are indeed entitlements, and that their status as such needs
explanation, I do not see how he can avoid the kind of investigation attempted in the
work of mine to which he is principally responding.
32
Williams has an independent point to make against Cartesian scepticism. The Cartesian arguments
proceed from a premise to the effect that we have no warrant for discounting a certain scenario of relevant
cognitive disablement. Williams observes that we typically will have every reason to discount such a
scenario—provided we are allowed to draw on ordinary background empirical knowledge! The much-
derided pinch test, for example, can provide good evidence that one is not now dreaming if one is allowed to
draw on the background empirical knowledge that pain is never dreamt. Now, this point about dreams one has
been able to identify as such seems to owe nothing to any contextual conception of the distinction between non-
inferentially and only-inferentially warrantable claims. But—as I mean to indicate by the italics—it has only
limited force. Switch the disablement scenario to that of the malin génie, and it is hard to see how even to
formulate a parallel reservation: maybe we have lots of ordinary empirical knowledge about what
happens in dreams and in normal waking consciousness, but do we have comparable ordinary empirical
knowledge about significant contrasts in the kinds of experience suffered by normally situated folk and
genie-victims respectively?
33
Maybe my own sketchy thoughts about ‘entitlement of cognitive project’ in Wright 2004 point in a
direction that might be helpful for this.
34
John Hawthorne’s term.
REPLIES 483
I do not know how Williams himself would address the concern he voices about the
possible excessive leniency of entitlement—the worrying prospect that anything that
could be said to entitle us to matter, other minds, a substantial past, and a broadly
nomologically regular world will also give houseroom to gods, fairies, interfering sprits,
and mysterious rays. In any case, it seems to me that his contextualism faces an
analogous challenge. Either the hypothesis that leprechauns exist is available to frame
some, in context, legitimate cognitive project or other, or it is not. One hopes not. But
if so, presumably the distinctions, yet to be drawn, that explain why not are going to be
common ground.
There are a number of more specific issues raised by Williams’ discussion which it
would be good to try to take further. But I must defer that for another occasion. Let me
conclude by responding to his concern that the constraint that the considerations that
underwrite a particular entitlement utilize no non a priori resources, cannot be strictly
adhered to. Williams thinks this because he is thinking—with ample justification
provided by some elements in my discussions—of the non-evidential warrants that
are grounded in entitlements as pragmatic warrants, as pragmatic reasons to trust. And
pragmatic reasons are, of course, going to be contingent on our desires. So the reasons,
if any, issuing from the considerations that are supposed to ground entitlements cannot
be resolutely a priori. Where Hume postulated contingent if unalterable psychological
propensities and limitations, I—Williams thinks—must postulate contingent human
desires. The contingencies concerned may be unquestioned. But the fact will be that
empirical information will be being smuggled in to the grounding of entitlements; and
there will then be a question, in a context where other minds scepticism, for example,
is on the table, why that is not a foul.
It is a good objection, but it needs to be distinguished from another one which has
been lodged in the literature35 and which likewise tries to make out that the entitle-
ment project is spoiled as a response to scepticism by the pragmatic character of the
warrants in which it trades. I think it will be useful if, in responding to Williams, I take
the opportunity to outline a response to this latter objection first. The objection is in
essence that the non-evidential warrants offered up by entitlement theory are in all
respects comparable to the kind of reasons for believing in the Christian God high-
lighted by Pascal’s wager. Those are non-evidential reasons par excellence. Their thrust
is, not quite that believing is a dominant strategy for one’s future utility, but that the
utility of the belief, in the event that it’s true, sufficiently outweighs the marginal
disutility, should it be false, to make it a potentially rational one. Even if we granted
that there could be such pressing prudential reason to try to bring it about that one held
a certain belief, irrespective of its evidential status, the charge is that it is a point-missing
response to a challenge to produce epistemic reasons to accept the propositions that have
come under sceptical question.
35
See for instance Jenkins 2007 and Elstein and Jenkins forthcoming; also Pritchard 2005.
484 CRISPIN WRIGHT
There is a sneer associated with the term “pragmatic” in this context. It is good to shift
the example so that goes away. So suppose that instead of eternal bliss for myself, what is
in prospect, if I successfully undertake Pascal’s wager, is a very considerable alleviation of
Third World suffering, poverty, and disease. In that case, I think we would be happy to
say that there is a strong moral reason—admittedly an unusual one, but no less strong for
that—for undertaking the wager and trying to bring it about that I have the appropriate
belief. It is still a pragmatic reason. But the shift in the example brings out that pragmatic
reasons are not a special genre of reason, to be contrasted with e.g. epistemic, moral, and
prudential reasons. The thing about pragmatic reasons is that they are contingent on the
goals of the agent—the rational explanation of an action performed in the light of a
pragmatic reason will take the form of an ordinary practical syllogism involving belief
and desire. There is therefore no good cause to deny certain kinds of pragmatic reason
the title “epistemic”. This will be the case where, in the slot in the structure of the
reasons for an action that is to be filled by the desires of the agent, the relevant desires are
focused on epistemic goods and goals. In this sense, to show that certain trustings and
unevidenced acceptances further the attainment of epistemic goods—of truth, under-
standing, and the anticipation of future experience, for example—is to provide epistemic
reason for those acceptances. There is no point in denying the title.
The question that then immediately comes into focus is whether evidence—epistemic
reason proper, as critics may have it—is different in this respect. It may seem so. If you
have good evidence that P, and no defeaters, surely that gives you a reason to believe
P that is quite independent of your desires? And surely, conversely, it would be a
complete explanation of your actually believing P that you did have that evidence,
irrespective of your desires. By contrast, the explanation of a belief held on pragmatic
grounds will have to advert to the believer’s goals.
However I wonder if this is quite right. Suppose we replace the desire component by
one of valuing. One’s values need not be available to one in the way that the having of
desire, or the having of a certain goal, normally are. One does not have to appreciate
one’s values, in the way one arguably does have to appreciate one’s desires and goals, in
coming to rational practical decisions about what to do. Rather, one’s values shine
through in one’s natural predispositions of action. Don’t we prize evidence because we
consider it more likely that beliefs that are responsive to evidence will be true than
otherwise? Isn’t the reason giving force of evidence contingent on our valuing the truth?
The question is in effect whether evidence provides categorical reason to believe—
reason whose force is unconditional on any aspect of the psychology of the recipient.
Certainly, it seems intelligible that someone who—on a particular question at least—
has some kind of vested emotional interest in concealing the truth from himself should
fail to feel the motive force of good evidence. Of course we would still want to say that
he ought to feel its force. But is that because its force is categorical, or is it because we
think that he ought to value the truth? If the latter is the right account, then part of the
response to Williams’ concern about the a priority of entitlements might go like this.
Even if entitlement is construed as pragmatic (though still epistemic) reason to trust in
REPLIES 485
certain propositions, it does not follow that the reasons thereby supplied involve
psychological contingency in a way that violates the constraint that the considerations
which disclose entitlements should be available at a purely a priori reflective level of
consideration—the level “of philosophy”. For the motivational force of evidence too is
conditional on the agent’s possession of those same epistemic values. If, per impossibile,
someone was to produce, by way of a response to scepticism, a genuinely forceful
argument for believing that there is an external material world, it would be bizarre to
respond that the force of the argument was qualified by its dependence on the
contingencies of our epistemic values. Philosophy is already up to its neck in those
values. And nothing can be a forceful a priori argument for anything if the values
concerned are not taken for granted.
I outline this response because I think it is worth thinking about. But it is not my
preferred response. The basic insight behind the entitlement project—Wittgenstein’s
insight, I believe—is that all enquiry is essentially situated in acceptances, some general,
others specific to the particular context of enquiry, for which we lack evidence. This is
not a shortcoming, a lapse which, though unavoidable, is nevertheless regrettable. It is
in the nature of enquiry that this should be so. The accumulation of evidential reason to
believe is possible only within the framework set by these trusting acceptances.
Without them, we forfeit, in particular, our detailed conceptions of evidential rele-
vance—of why, for example, perceptual experience bears on how matters are config-
ured around us in the external world. The entitlement project is to characterize the
various roles of the acceptances, the specific aspects of our enquiries that they sustain
and rationalize. It remains to be seen in what degree of convincing detail the project
can be executed. But the basic point is that there is no way things could be otherwise:
without trusting, there is no rational enquiry.
A critic might say that, even if this is true, it doesn’t change matters: a rationale is
provided for the relevant trustings only if enquiry is valued. We have the option of
dropping out, or dropping dead. But I think that to accept this much contingency in
the offered account is quite consistent with it’s being properly philosophical. If I am
charged to give you a moral justification for a certain course of action, and proceed to
offer considerations which demonstrate, for example, the injustice of the only available
alternative, it would be strange to reply, “Well, that only works if you presume that
I am concerned about justice.” In the same way, I think we have given a properly
epistemic justification of certain aspects of our practices if it is demonstrated that they
serve certain epistemic values, or—the contention I have just been stressing—that they
are integral to enquiry itself. That demonstration, if it goes through, will be a priori and
philosophical. And it will be an epistemic justification. Whether we actually have, or
must have, the epistemic values to which enquiry is harnessed, or whether we value
enquiry itself, is not an issue which epistemic justification has to tackle. Nor does a moral
justification have somehow to negotiate the contingencies of our valuing the moral
goods to which it appeals.
486 CRISPIN WRIGHT
Bibliography
Chandler, J. 2010 “The Transmission of Support: A Bayesian Re-analysis”, Synthese 176/3,
pp. 333–43.
Cohen, S. 2002 “Basic Knowledge and the Problem of Easy Knowledge”, Philosophy and
Phenomenological Research, 65/2, pp. 309–29.
Elstein, D. and Jenkins, C. forthcoming “The Truth Fairy and Epistemic Consequentialism”, in
P. J. Graham and N. Pedersen (eds.) New Essays on Epistemic Entitlement.
Harman, G. 2003 “Skepticism and Foundations”, in S. Luper (ed.) The Skeptics: Contemporary
Essays, London, Ashgate, pp. 1–11.
Jenkins, C. 2007 “Entitlement and Rationality”, Synthese 157/1, pp. 25–45.
Kotzen, M. 2012 “Dragging and Confirming”, The Philosophical Review 121/1, pp. 55–9.
Moretti, L. 2012 “Wright, Okasha and Chandler on Transmission Failure”, Synthese 184 (3),
pp. 217–34.
Okasha, S. 2004 “Wright on the Transmission of Support: A Bayesian Analysis”, Analysis 64/2,
pp. 139–46.
Pritchard, D. 2005 “Wittgenstein’s On Certainty and Contemporary Anti-Scepticism”, in
D. Moyal-Sharrock and W. H. Brenner (eds.) Investigating On Certainty: Essays on Wittgen-
stein’s Last Work, Basingstoke, Palgrave Macmillan, pp. 189–224.
Pynn, G. 2011 “The Bayesian Explanation of Transmission Failure”, Synthese Online First.
Smith, M. 2009 “Transmission Failure Explained”, Philosophy and Phenomenological Research 79/1,
pp. 164–89.
Williams, M. 1991 Unnatural Doubts: Epistemological Realism and the Basis of Skepticism, Oxford,
Blackwell.
Wright, C. 2000 “Cogency and Question-Begging: Some Reflections on McKinsey’s Paradox
and Putnam’s Proof ”, Philosophical Issues 10, pp. 140–63.
Wright, C. 2003 “Some Reflections on the Acquisition of Warrant by Inference”, in
Wright, C. 2004 “On Epistemic Entitlement: Warrant for Nothing (and Foundations for
Free)?”, Proceedings of the Aristotelian Society Supp. Vol. LXXVIII, pp. 167–212.
Wright, C. 2011 “McKinsey One More Time”, in A. Hatzimoysis (ed.) Self-Knowledge, Oxford,
Index Nominum
Alston, W. 296, 299 Detlefsen, M. 222

Anscombe, G. E. M. 51, 64, 168 Dretske, F. 9, 271, 275, 278, 305, 310, 312, 314,
Audi, R. 296 318–19, 451, 454–7, 468, 470
Austin, J. L. 171, 372, 374, 456 Duhem, P. 228, 232
Ayer, A. J. 169, 189 Dummett, M. 103–4, 115–16, 134, 136, 138–40,
203, 214–15, 413–14, 418–20, 425, 428
Bach, K. 134
Backer, G. 347 Egan, A. 436
Baldwin, D. 156 Elga, A. 232
Bar-On, D. 6, 163–5, 166–9, 171, 176, Elstein D. 483
179–82, 185, 188–9, 402–3, 407–12 Evans, G. 180, 182, 388
Beebee, H. 324
Bennett, J. 98, 123, 198 Feldman, R. 275
Bergmann, M. 286 Feyerabend, P. 260
Bilgrami, A. 4–5, 96, 102–3, 108, 114–15, 117, Field, H. 7, 198–7, 209, 275, 419, 421, 423–4
122–3, 164, 179, 245, 379, 394–8, 404 Fine, K. 443
Blackburn, S. 7, 171, 195, 199, 200, 202, 205, Fodor, J. 37, 108, 153
330, 418, 420–5 Follesdal, D. 139
Boghossian, P. 3, 27–8, 34, 38, 44, 48–9, 92–3, Frege, G. 1, 4–6, 54, 104–6, 109–11, 115–16,
101, 179, 269, 314, 337, 347, 377, 379, 118–23, 147, 221–4, 377, 397, 422, 427
381–4, 386, 395, 405–6, 437, 441 Fricker, E. 179
Bolzano, B. 220–2, 224
Brandom, R. 79, 207, 375, 419, 424 Geach, P. T. 6, 163, 169–71, 173, 176,
Brenner, W. H. 324 186, 422
Broome, J. 285 George, A. 380
Brown, J. 308, 324 Gibbard, A. 49, 58, 81, 92, 171
Burge, T. 4, 103, 116, 152, 154, 221–2, 275, Goodman, N. 241, 260–1
277, 325, 340, 386 Grice, P. 97–102, 110, 116–19
Burgess, J. 211–12, 226
Hacker, P. M. S. 167, 347
Carnap, R. 199, 241, 260–1, 318 Harman, G. 432–3, 436–7, 477–8
Carroll, L. 41, 381 Hawthorne, J. 305, 436, 482
Carston, R. 134 Hempel, C. G. 223
Chandler, C. 455 Horwich, I. 3, 4, 7, 43, 48, 77, 87–9, 91–3,
Chomsky, N. 60, 83, 117, 136, 150–2, 380–1 198–9, 203–7, 209, 379, 381–2, 386–7,
Chrisman, M. 172, 174, 188–9 392–4, 398–400, 416–17, 419–21, 423–4
Christensen, D. 231 Hume, D. 189, 196, 329–30, 332, 346, 355,
Clifford, W. K. 366 360–1, 364–5, 368, 478, 483
Cogburn, J. 230
Cohen, S. 275, 461, 468 James, W. 366
Coliva, A. 12, 323, 325, 327–9, 333, 336, 338, Jenkins, C. 230, 236, 281, 334, 483
341, 345–8, 377, 451, 476–8 Jeshion, R. 222
Conee, E. 269, 275
Cruz, J. 46 Kant, I. 28, 66, 212–14, 221
Kitcher, P. 226
Davidson, D. 92, 103, 127, 134, 140–1, 144, Kotzen, M. 48, 269, 458
147, 154–5, 174, 209, 212, 245, 259–60, 329, Kölbel, M. 244, 247
384, 406, 413–14, 431 Kripke, S. 2, 27–8, 30–9, 43–7, 52, 54, 63, 77–9,
Davies, M. 269, 272, 275, 277, 288–9, 297, 299, 81–4, 86, 88, 96–7, 103, 105–6, 111, 116,
307, 310, 312, 317, 325–6, 331–2, 340–1 119–23, 129, 380–1, 385, 393–6
Dennett, D. 174 Kuhn, S. T. 241, 260
488 INDEX NOMINUM
Lance, M. 92 Shah, N. 93
Leibniz, G. 222 Shapiro, S. 7–8, 211, 233, 235,
Lewis, D. K. 66, 82, 98, 195, 446 418, 429–35
Shoemaker, S. 180
MacFarlane, J. 229, 247, 436–7, 440–1 Silcox, M. 230
Mackie, J. 172 Silins, N. 269–70, 272, 278, 324
Maddy, P. 222 Smith, B. C. 5, 127, 135, 138, 141, 146,
Malmgren, A. S. 92 148–9, 152, 154, 323, 402, 408,
Marconi, D. 346, 348 413–17, 458
Mates, B. 105 Soames, S. 43
McCullogh, G. 147, 152 Sperber, D. 134
McDowell, J. 9, 54, 67, 79, 92, 127, 138, 144–8, Stanley, J. 134, 388
152, 159, 179, 326, 346, 402–3, 414 Stevenson, C. 169
Meiland, J. 275 Strawson, G. 152
Moran, R. A. 179, 187 Strawson, P. F. 98, 199, 329–30, 345
Moretti, L. 455 Stroud, B. 348, 354, 364, 369
Moyal-Sharrock, D. 324, 347
Tanney, J. 187
Neta, R. 190 Taschek, W. 233, 235
Nozick, R. 9, 312, 455 Taylor, C. 187
Nuccetelli, S. 9–10 Tennant, N. 214
Travis, C. 108, 134
Okasha, S. 455
Oppenheim, P. 223 Velleman, D. 93
Vogel, J. 315
Pagin, P. 127, 136
Peacocke, C. 3, 38, 48–9, 53–4, 66, 68, 70–1, Waissman, F. 211–14
275, 324–5, 340, 379, 382, 386–93, 458 Weatherson, B. 436
Pettit, P. 33, 48 White, R. 275, 287, 324
Plantinga, A. 275 Williams, B. 260, 436
Pollock, J. 46, 275 Williams, M. 12–13, 341, 352, 367, 374, 419,
Price, H. 207 451, 480–4
Priest, G. 261 Williamson, T. 388, 419
Pritchard, D. 334, 483 Wilson, D. 134
Pryor, J. 10–12, 269, 275–7, 282, 286, 308, Wittgenstein, L. 2–6, 11–13, 27, 30, 39–43,
310–11, 316–17, 319, 323–8, 331, 344–5, 49–52, 57, 59, 63–5, 69–70, 77–80, 91,
358, 451, 460–1, 463, 466, 470 96–7, 100–1, 112, 114–16, 118, 128–31, 133,
Putnam, H. 9–10, 71, 146, 212, 307, 329, 139, 147, 152, 154, 157, 167, 171, 178, 180,
451, 460 196–7, 208–9, 243, 256, 324, 329–30, 332,
Pynn, G. 455, 458 345–8, 352, 357, 367, 374, 379–81, 384, 388,
393, 399, 402–3, 407–8, 412–14, 418, 484
Quine, W. V. O. 43, 134, 136–7, 139–40, 145, Wright, C. 1–13, 27–8, 30, 32–7, 46, 49–50, 52,
151, 212–14, 228, 232, 384, 413–14, 431 59, 63–70, 77–9, 83, 85, 96, 127,
129–32, 141–5, 147, 152, 157, 159,
Recanati, F. 134–5, 160 163–6, 168–81, 186–7, 190, 195, 197–8, 209,
Richard, M. 440–1 213–17, 219–20, 222–3, 225, 227–8, 230–6,
Rorty, R. 68–70, 166, 207, 347 239–41, 244–55, 258–9, 263, 269, 271–3,
Rosen, G. 211–12, 226 275, 277, 279–81, 283, 287–99, 304–5,
Rossberg, M. 228 307–8, 310–12, 315–17, 319–20, 323–36,
Rovane, C. 8–9, 123, 238, 243–7, 255, 258–9, 340–7, 352–4, 356–75, 377, 381, 394, 403,
262, 323, 418, 442–7, 449 407, 410–11, 421, 429, 432,
437–9, 470, 480, 482
Sainsbury, R. 212
Salmon, W. 224–5 Yablo, S. 219–20
Schiffer, S. 48, 98–9, 324
Seay, G. 10 Zalabardo, J. L. 10, 11, 275, 304, 315, 451,
Sellars, W. 182–3 453–4, 456–8
Index
Acceptance(s) 3, 11–13, 31–4, 39, 40–3, 46–7, Avowal(s) 5–6, 34, 131–2, 142–3, 163–83,
87–8, 92–3, 157, 206, 208–9, 260, 288, 185–9, 402–4, 407–13, 441
294–5, 328, 330–7, 339–41, 344, 346–7,
357, 359, 373, 385–7, 390, 392, 399, Behavior 2–4, 27–32, 34, 38, 43, 79–80, 91,
421, 439, 448, 456, 461, 466, 472, 476, 130, 133–4, 139–40, 142, 165, 167–8,
480, 484–5 182–4, 187, 216, 255, 356, 395, 398,
ACRE 411–12, 414, 420, 431
Model 384–7, 390. Linguistic vs. non-linguistic 108, 136–8,
Action(s) 35, 40–1, 45, 51–2, 56, 71–2, 78–80, 145–8, 151–2, 159, 400, 401, 406–7
96–7, 100, 113, 116, 163, 182–3, 187, Being ignorant vs. being irrational 5, 104–5,
189, 202, 206, 246, 248, 254–5, 294, 120, 397
335, 357, 381, 383, 386, 390, 391, 393, Belief(s)
395, 397, 398–400, 412, 453, 475–6, Entry-rational 384, 387–9
479, 480, 483–5 -Formation 330, 337–8, 365, 384–7
Rational 31, 386 390–2, 395, 430, 474–5, 480
Alethic pluralism 420–3, 425, 427 Irrational 30, 288, 337, 397
Alternative(s) 9, 11, 70, 89–90, 110, 128, 129, Rulish picture of 384–6
151–2, 163, 167, 169, 172, 174, 176, Bivalence 230, 234, 379, 420, 438, 439
178, 241–3, 249, 255–6, 258–61, 264–5, Bootstrapping 461, 468, 470
269, 277–8, 280–1, 292, 310, 324, 335,
345, 367, 369, 374–5, 381, 403–4, 419, Circularity 11, 81, 91, 302, 344–5, 349, 368, 373
424, 426, 442–6, 485 Cognitive
Anti-realism 6–9, 17, 19–20, 23, 160, 163, Access 141, 431
171–2, 174–7, 179, 186, 188, 190–2, Command 6–8, 69–70, 197, 198, 215,
214, 240, 244, 250–4, 261–4, 341, 227–9, 231–6, 410, 427, 429–31,
418–20, 426, 428–31, 433–6 434, 437, 439, 449
A priori 7, 22, 43, 69, 72–3, 81–3, 87, 89, 92, Function 467
142–3, 185, 199, 221, 228, 236, 274, Project 333, 357, 366, 373–4, 466, 468,
290, 302–3, 344, 364, 397, 401, 404–5, 470, 474, 482
410, 412, 430, 432, 439, 451, 474, 480, Shortcoming 7, 227–9, 232, 233–5, 250,
482, 483–5 410, 427, 429–30, 432, 439, 449
Areas of discourse Value 4
Objective vs. non-objective 6–8, 69–70, 162, Coherent(ism) 17, 18, 28, 213, 219, 221,
172, 178, 214–16, 219, 220, 223–8, 232, 238, 241, 243, 248–50, 261–2, 338,
235, 236, 352, 418, 424, 426–9, 433 354, 369, 373, 379, 384–5, 392,
Assertibility 412–13, 417, 436–7, 444, 457, 479, 481
Conditions 101, 104, 115–16, 118, Community 52, 59–60, 70, 85, 104, 133,
418–19, 447 156, 339, 346
Assumption(s) 30, 35–9, 41, 45, 87–90, 93, 96, Linguistic 4–6, 92, 128–9, 134, 148, 205
114, 119, 134, 144–5, 154–5, 157–9, Concept(s) 2–3, 11, 29, 49–75, 78, 82, 88, 101,
175, 217, 229–31, 233, 235, 239, 245–6, 122–3, 140, 143, 155, 177, 188–9, 199,
254, 277, 307–8, 317–18, 321, 323, 326, 212–14, 220, 313–15, 339, 354, 362, 382,
327, 331–2, 334, 336, 338–9, 343–4, 385–9, 392, 394, 398, 433, 449, 479
357, 359, 366, 375, 385, 391, 398, 400, Moral see moral
415, 453, 462, 476–9 Conditional 2, 6–7, 40, 42, 47, 84, 91–2, 115,
Authenticity-condition(s) 426, 466–77, 482 163, 169–70, 173, 189, 196, 285, 331,
Authority 33, 52, 132–4, 141–2, 153, 159, 342–3, 381, 383, 395, 397, 421, 478, 484
163–4, 166, 176–8, 186, 190, 191, 197, Conservatism 472–7
402–3, 405–7, 409, 411, 415 Constructivism 418
490 INDEX
Content(s) Ethics 172, 238, 285, 393, 420–1, 432, 435,

Assertoric 1, 6 437, 453
Conceptual 53–4, 59, 390 Evidence 11, 13, 30, 56–7, 70, 65, 69, 80, 82,
General conditional 40–2, 383, 395 85, 108, 131, 133–4, 136, 140, 143, 151,
Intentional 30, 46, 63, 64, 73 153, 166, 173, 177, 181, 214, 231–2,
Linguistic 2, 36, 380, 416 252, 272, 277, 279, 285–6, 290–1, 293,
Mental 2, 34, 36, 38–9, 43, 45–7 295, 298, 305, 315–21, 330–1, 337–9,
Propositional 78, 171, 173, 418, 420, 341, 345, 347–8, 355, 357–9, 362–3,
424–6, 440 365–7, 370–2, 374, 388, 399, 408,
Semantic 78, 178, 180, 183 419–20, 426, 428, 431–2, 435,
Truth-evaluable 6, 170–1, 412, 443 453–5, 457, 460–1, 463, 466–7, 469,
Context of assessment 247 471–3, 475, 477–8, 481–2, 484–5
Contextualism Experience
Epistemological 13, 341, 481 Perceptual 55–6, 156, 324, 326, 328, 331,
Nonindexical 440, 441, 447 334, 338–9, 389, 473, 476–7, 485
Contradiction Visual 56, 71–2, 75, 269–76, 279–82,
Principle of non- 8, 9, 240–4, 247, 249–51, 287–8, 298, 405, 453, 470
253, 254, 261–4, 437–9, 442, 444–6 Explanation(s) 3, 6–8, 31–2, 39, 51–7, 60–1,
Correct vs. incorrect 2, 30, 53, 65, 101, 119, 66–8, 70–5, 87–9, 94, 128, 142, 150,
128, 394, 398 155, 165, 168, 178–9, 182, 187, 203,
Credence(s) 453, 454, 468, 470 205, 211, 213–20, 222–7, 239, 262,
288, 306–7, 315, 319, 335–6, 370, 373,
Deduction 88, 224, 272, 451, 456–7 388–9, 392–3, 395, 399, 401, 404,
Default View 163, 176–9, 185, 187, 407–8 407–8, 432–4, 454–5, 464, 482–4
Deflationism 88–9, 195–200, 204, 206, 208–9, Explanatory order of things
420–4 Objective 433
Dialetheism 230, 438 Expressivism 6, 162–4, 167–77, 179, 190,
Dilemma 9, 88, 145, 207, 226, 242, 249, 256, 203, 196–7, 206–7, 209, 228, 403,
264–5, 321, 344, 382, 422–3, 428 407, 410–12, 421–3, 425, 435
Language First (Thought First) 404 Externalism 13, 360, 363–4, 367, 453–5,
Disagreement 7, 9, 82, 166, 174, 227, 230, 458, 474
239–42, 244–50, 252–5, 259, 261–2, Semantic 10, 138, 146, 307, 397
264, 429–31, 445, 449, 476, 480
Blameless 228–34, 253 Fact(s) 3–4, 9, 11, 34, 38, 45, 54, 56–7, 58–60,
Faultless 8, 230, 250, 251, 435, 437, 439, 68, 74, 77, 79, 82–4, 86–7, 90–3, 107,
441–2, 444, 446, 448 129–33, 135–41, 144–5, 148–9, 151–4,
Disjunctive template 289, 291–4, 296, 466 162, 169–70, 172–3, 175, 177–9, 186–8,
Disposition(s) 7, 33–4, 36, 43–7, 51, 55–6, 59, 200, 203, 208, 211, 214, 216–17, 222–3,
80, 83, 97, 130, 139, 164, 196, 200, 204, 226–8, 245, 247, 250, 252–4, 297, 299,
206–7, 388, 394, 396, 484 305, 308–9, 315, 323, 356, 359, 367, 371,
Austere view 392–3 374, 380, 393–5, 399, 405–6, 409, 420,
Dogmatism 8, 10, 260, 276–7, 280, 282–4, 289, 422, 427, 430, 433, 438, 442–3, 445
297, 311, 473, 477 False – falsity 4, 6, 10, 45, 47, 53, 65, 106, 112,
117–18, 165–7, 172, 181, 185–6, 188–9,
Egregious misunderstanding 444, 447–8. 199, 201–4, 212, 215, 217, 233–4, 242,
Entailment 9, 11, 285, 292, 299, 313–14, 247–9, 255, 257–8, 260, 263, 275, 277,
342–3, 345, 383, 391, 443, 457–8, 280–1, 283–4, 287, 292, 298–9, 310–11,
468, 470 316, 324–6, 331, 341, 348, 355, 358–9,
Entitlement 5, 12–13, 155, 157, 159, 252, 275, 362, 410, 421–2, 430, 432, 438,
277, 292, 294–6, 299, 305, 327, 331–6, 440–2, 446, 448–9, 465–6, 468, 483
340, 352–3, 356–64, 366–8, 371, 373–4, First person 5–6, 33–4, 50, 83, 129–30, 133,
414–15, 475–85 136–7, 144, 147, 152–5, 159, 163–5,
Epistemic security 6, 142, 164–6, 168, 175, 187. 174–5, 326, 406, 409, 411–12, 416
Epistemic responsibility 365, 367, 373–5, 471, Authority 132, 140–3, 166, 176–8, 186, 402
475, 480 Foundationalism 131, 348, 371–2, 381, 474
Epistemic Obligation(s) 475 Frege–Geach Problem 6, 163, 169–71, 173,
Epistemic value(s) 90, 480, 484–5 176, 186, 422
Error-theory 435–6 Fundamentalism 458
INDEX 491
Idealism 212, 230, 244 Law of the Excluded Middle 438

Linguistic 3, 50, 64–7 Liberalism see Dogmatism
Identity 57, 59, 81, 103–6, 111, 197, 199, Limitation clauses 10, 307–8, 312, 314
404, 479 Logic
Statements 4, 119–23 Classical 438
Idiolect 6, 37, 94, 137, 144, 159, 414–16 Revision of 250, 418,
Immediacy 6, 131, 159, 192, 280–1, 403, 409, Intuitionistic 233–4, 250–1, 438–9, 448–9
411, 414–16
Indeterminacy of translation 136, 431 Meaning
Induction 357, 391, 428, 480 Assumption 36–9, 45
Inference Awareness of 61–4, 73, 153, 400
Ampliative 356, 462–7, 482 Constitutiting 87–9, 93–4, 380, 387,
Information 73, 274 394–5, 399
Processing 383, 389, 392 Dispositional account of 393
Logical 11, 163, 168, 189, 391 Intentions 4–5, 63, 96–103, 106–7,
Moorean 306–17, 321, 456, 471–2 111–14, 116, 119, 394–8
Problem 385, 390–1, 395 Theory of 58, 104, 132, 200, 203–4, 418,
to the best explanation 3, 73, 217 428, 388
Information-depedence template 289–90, 466–7 Use 5
Intention(s) Metaphysics (al) (ally) 1, 9, 36, 44, 49,
Meaning-constituting (fixing) 93, 380, 394–5, 399 68–70, 77, 82, 93, 129, 136, 138,
View (model) 2–3, 27–8, 33–43, 46–7, 395 146, 155, 159, 165, 172, 175, 189,
Standard-determining (setting) 396–8 195–6, 212, 220, 222, 227, 232–3,
Interest-relativity 266, 327, 432–4 238, 240–1, 243–5, 247, 251–4, 256,
Internalism 363, 364, 368, 453 258–9, 262–5, 326–7, 338, 348, 404,
Intuitionism 251, 379 407, 413, 418, 420, 424, 426–7,
Investigation-indepedence 59, 64–5, 70, 380, 433–6, 443, 449, 482
387, 389 Minimalism 6–7, 171, 195, 209,
Introspection 83, 165, 167, 176, 179, 182, 410, 420–1
186, 404 Minor Premise(s)
Problem 385, 390
Judgement(s) Misspeak(ing) 109–13, 116, 119–23, 397
Dependence see response Modality 49, 420, 446
Propensities 3, 380 Modus ponens 3, 32, 41, 43, 45, 334, 336, 339
Entry-Rational 388–9 Model 2, 381, 383–5
Inferential 389, 403 Moore’s Proof of an External World 1, 10–12,
Non-inferential 388, 403–4 306–7, 323–4, 327, 329, 331, 339,
Justification 44, 49, 164, 221–3, 232, 275, 341–5, 451, 453–4, 457, 460, 471–2
293, 316, 327, 331–2, 335–6, 341, Moral 35, 171, 175, 195–6, 206, 214, 216,
348, 352–4, 358–60, 372, 374, 453, 410–11, 422, 423, 433–4, 436–7, 449,
462, 476, 477, 481, 483 475, 485
Epistemic 4, 85, 92–3, 326, 329, 361–6, Agency 393
368–71, 375, 485 Good 162, 453, 479–80
Perceptual 317, 391, 451, 471 Obligatory 479
Architecture of 12, 332, 356, 476–7 Multimundialism 9, 244, 256–8, 442–8
Knowledge Negation 6, 163, 170, 173, 233–4, 282, 292,

Ascriptions 162, 174, 436 296, 311, 315, 421, 454, 468
Authoritative 5, 34, 132–3 Norm(s) 29, 30, 49, 53–6, 58, 60, 65, 85, 90,
First-person(al) 5, 34, 83, 131–2, 136, 141, 92–3, 97, 101, 106–9, 112, 114–15, 129,
152, 416 154, 205–6, 294–5, 387, 392, 421, 448
Immediate 5, 83, 130–3, 395 Epistemic 3, 385, 387
Meta-linguistic 416–17 Normative insularity 243–4, 258, 261, 264–5,
Reliabilist account (conception) of 456 265, 443–4
Safety account (conception) of 408, 456 Normativity 2, 4–5, 83, 91–3, 96–7, 100–3,
Tacit 3, 53–62, 66–74, 135, 387–90, 392 108–9, 112, 114–15, 118–19, 195–6,
-Transmissive argument 455–6, 459 198, 209, 393–6, 398, 436, 479
492 INDEX
Objectivity 1, 6–8, 49–50, 60, 67–9, 115, Quandary 9, 241, 250–3, 255, 258
131–2, 137–8, 147, 162, 195, 211–20,
222–8, 230–4, 236, 405, 420–7, 429–31, Rationality
434–7, 439, 448–9 Epistemic 12–13, 330, 336, 338–41, 346–8,
Operational Shortcoming 430 479–80
Order of determination test 6–7 Pragmatic 12, 339
Reactions
Paradigm case 478 Human 3, 7, 64–7, 70
Paradox(es) 57, 261–2, 448 Realism 1, 49, 186, 190, 215, 230, 243, 341,
Kripkean 399 411, 418–20, 426–8, 430–2, 434–7
Moorean 410 Quasi- 422
Regressive 387 Reason(s)
Sceptical 364, 369, 381, 478–80, Categorical 484
Parity 439–42, 444–8 Epistemic 61, 74, 483
Perception 54–5, 66, 70–1, 74, 140, 149, 158, Moral 483
211, 215, 276–7, 281, 283, 294, 297, Reasoning 3, 45, 47, 119, 146, 200, 235, 242,
299, 316, 327, 356, 388, 456, 465 249, 271–3, 277–81, 284–5, 290–1, 295,
Platonism 2, 50, 64, 67–8, 144, 418 298–301, 306, 309–11, 313–15, 317,
Rampant 3 383–4, 421, 440, 451–2, 468
Practice(s) 2, 6, 30, 65–6, 79–80, 84, 90, 92–4, Moorean 273, 276, 278, 280, 282–3, 287–9,
128–9, 139, 147–8, 157, 177, 207, 215, 296–7, 300–1
270, 325, 330, 332–4, 337–41, 346–8, Reconciliation Problem 5, 133, 136, 138,
353–4, 372, 383–4, 386–7, 420, 428, 143–4, 147, 152, 159, 413–16
478, 485 Reference 3, 7, 52–3, 65–7, 105, 108, 115,
Inferential 263–4 119–20, 128, 136, 140, 148, 153, 169,
Linguistic 64, 85, 106–7, 121, 129, 133–4, 173, 203–8, 245, 425
138, 246, 379–81, 388–9, 394–5, Regress 181, 374–5, 434, 474
398–400, 403, 407, 436 Vicious 3, 41, 46, 85–6, 384–6, 390–1,
Presupposition 12, 58, 79, 157, 186, 260, 395, 406
293, 299, 301, 330, 332–4, 337–8, Regularities 4, 77, 82, 94, 106, 139, 204–5, 355,
340, 346–8, 352, 357, 361, 366, 368, 391, 395, 399–401, 416, 424
374–5, 430, 467, 475 Relations
Private language 2, 130, 402 Indication 7, 200–6
Privileged access 27, 178, 181, 307, 402–5, 407, Causal 7, 31, 203, 206
411, 413, 415–16 Logical 9, 89, 242–3, 256–8, 264
Projectivism 228 Relativism 9, 196, 230, 238–45, 250–4, 256,
Proposition(s) 7–8, 11, 42–4, 47, 50, 54, 61, 87, 258, 260–2, 264–5, 367, 439, 442–3,
172, 182–5, 188–9, 195, 197–200, 203, 447, 448
208–9, 214, 217, 219, 221–3, 227, 229, Assessment- 247–9, 440–1, 447, 449
236, 238, 240, 242, 244–7, 262, 270, New Age 436–7
280, 285–6, 293, 305, 311–20, Semantic(al) 241, 244, 245, 247, 250, 253–4
325–8, 331–3, 339–40, 342, 345, 347, Truth- 8, 435–6, 438, 444
359–61, 404, 415–16, 426, 436, 441, Representation(al) 7, 40, 54, 55, 57, 68, 74, 105,
445–6, 454, 456–8, 461–2, 465–8, 135, 150, 182, 196–8, 200, 207, 209,
471–3, 476, 483–4 211–12, 215, 219–20, 391, 410, 422–4,
Arithmetical 379 427, 429–31, 438, 448
Cornerstone(s) 12–13, 157, 333, 354–6, Response
359–63, 365–71, 373–5 Dependence 405, 407–8
Metaphysical Heavyweight 305–10, 312, Rule-informed behaviour 4, 382
314–15, 321, 482 Rule(s)
Normative 29, 31 Epistemic 35, 384–5
Propositional surface 421–4 Fundamental reference of - 3, 53–63,
Psychological Self-ascriptions of propositional 66–75
attitudes 5, 143, 165–6, 173, 177, 181– Supertask 385, 390
2, 185, 408 Rule-following,
First person(al) 164, 180, 409 Blind 2–3, 42–3, 47, 51–2, 56, 64
Present-tense 6, 141, 164, 172–3, 179 Considerations 2, 5, 46–7, 52, 65, 68–70,
Third-person(al) 164, 180 380, 385
INDEX 493
Dispositional account of 2, 37, 43–5, 47, 83, Information 387, 390–2, 394
388–90, 392 Information-bearing 388
Explicit 32, 78–9, 83–7, 386–7, 400 Intentional 4, 27–8, 33–4, 36–7, 42–3,
Implicit 3–4, 78–88, 90–1, 94, 386–7, 400 46–7, 90, 96–7, 127, 142, 395–6,
398–9, 416, 431
Scepticism Normativity of 97, 100
About Philosophy 13, 353, 364–5, Mental 1, 36, 40, 71, 130, 137, 139–43,
372–3, 480 162–3, 165–70, 172–9, 183–5, 187–9,
Cartesian 11, 306, 315, 319, 329, 354, 357, 326, 402–4, 406–9, 411–13
370, 481 First-order 5–6
Content 387, 389 Self-ascriptions of 163, 166, 168–70, 175,
Direct vs. Indirect response to 333 180–3, 186
Humean 11–12, 329, 334, 339–40, 481 Registration 390
First order vs. Second order 297, 474 Subpersonal 42, 71, 73–4, 389–90
Internal World 167 Superassertibility 7, 477
Locality 355, 357, 370–1, 481–2 Sustainability 438–9, 442, 446
Meaning 128, 136, 387, 396
Philosophical 13, 353, 363–5, 372 Testimony 173, 257, 457, 461–2, 472
Rule-following 42, 46–7 Thought(s)
Sceptical Self-directed 403
Argument(s) 63, 284, 290, 305, 319, 346, Transparency 105, 109, 165, 403, 405–406, 409,
352, 354, 358–9, 361, 363, 369, 372–3, 411, 414–15
389, 393–4, 396, 398, 479, 482 of Sinn 5.
Anti- 324, 352–3, 360 Trigger-condition 383, 385–6, 392
Hypothesis 11, 311, 315, 319, 468 Truth
Paradox(es) 364, 369, 381, 478–80 -Apt(itude) 162, 171, 215, 410, 421–4
Kripke(an) 27, 28, 33, 82, 394 Minimally 172, 410, 431
Solution 52, 333, 393–4 Discourse(s) 7, 437, 448–9
Scenario(s) 11–12, 354, 369–70 Concept(ion) of
Seeming(s) 68, 149, 311 Deflationary vs. non-deflationary 88, 423–5.
Doxastic. vs. non-doxastic 404 -Conditions 4, 7, 77, 94, 99, 101–4, 107–10,
Perceptual 404 112–14, 116, 118–20, 122, 135, 170–1,
Self-knowledge 176–8, 183, 186, 195, 200–1, 203, 205,
Constitutive account of 5, 6, 142, 176, 404–8 388, 408, 418–19, 428
Immediate 131–3, 394–5 Degree of 443
Neo-expressivist account of 6, 163, 179–80, Disquotational- schema(s) 7, 189, 230, 420–2
182, 185, 186–8 Evidence-transcendent 214, 419–20,
meaning intentions 5 426–8, 431
meanings 5, 413, 417 Verificationist rejection of 435
Semantic(s) 60, 72, 117–18, 134–5, 163, 186, -Makers 186, 189, 195–6, 207, 297, 423,
199, 204, 229, 436, 443 425, 443, 449
Theory 388, 419, 425–6, 437 Mathematical 419
Inferential-role 419, 424 Minimal(ist) 1, 6, 410
Conceptual 419, 424 Predicate 7, 196–9, 422–3, 429, 443, 448
Content 78, 178, 180, 183 Property of 198, 421–2, 425
Continuity(ies) 163, 168–9, 173, 177, Relativization of 438
179, 189 Robust 7, 196–7, 199, 446
Externalism 10, 146, 307, 397 -Value(s) 2, 229, 245, 379–80, 404, 406, 419,
Intentions 112–13, 119, 121–3 423, 426, 438, 440–1, 444, 446
Referential 424–5 -Bearers 242–3, 256–8, 263, 423
Truth-conditional 419
Value 53, 55 Underdetermination 406, 431
Skyping argument(s) 468, 470 Understanding
State(s) Acquisition 419, 428
Attitudinal 217 -Based Rational Application 50, 52, 57–8, 71
Computational 3, 74 Immediate 6, 414
Contentful 3–4, 37 Manifestation 419, 428
494 INDEX
Warrant(s) Transmission 296, 305–10, 312–13, 316, 451,

Acquisition of 10, 276, 279, 281, 284, 293–4, 458–9, 464
296, 305, 314–16, 328, 334, 340, 355, Failure 9–12, 272, 276, 279–80, 282,
374, 451, 474, 476, 481 288–92, 294–5, 297–301, 317–21,
By inference 306–12, 359, 456–8 341, 343–5, 358, 454–7, 460–3,
Closure of 9, 11, 216, 271–2, 291–2, 312, 465–8, 470–1
314, 327, 345, 451, 455, 457–8 Background Warrant model of 273, 276,
Doxastic 269, 272, 286, 326, 474 278, 280, 283
Epistemic 295, 329, 330, 333–4, 337, 340, Reliabilist account of 11, 277, 309,
346, 358, 362–4, 366–7, 373, 451, 453, 311, 363
472–3 Testimonial 452
Empirical 13, 340, 342, 345, Unearned 13, 275, 277, 281, 290–1, 294,
369, 452 353, 356–7, 359–60, 366, 375
Architecture of 12, 332 Width of cosmological role 6, 7, 197–8,
Evidential 11–12, 332–7, 339–40, 453, 475, 215–25, 227, 231, 427, 432–3
477, 483 World(s),
First-time 452, 460 As one (oneness of) vs. many 9, 243–4,
First-level 471, 474–7 256–8, 443–6
Non-evidential 12, 305, 332–6, 340, 476, Inner 130, 137
478, 483 Mental 167
Pragmatic 359–60, 366–7, 483 Mind-independent 213, 218, 230
Propositional 269, 272, 326–7, 474 Outer 147–8
Second-level 474–7 Public 133

Annalisa Coliva - Mind, Meaning, and Knowledge - Themes From The Philosophy of Crispin Wright (2012, Oxford University Press) PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Annalisa Coliva - Mind, Meaning, and Knowledge - Themes From The Philosophy of Crispin Wright (2012, Oxford University Press) PDF

Hochgeladen von

Copyright:

Verfügbare Formate

Mind, Meaning, and Knowledge

This page intentionally left blank

Edited by Annalisa Coliva

Notes on the Contributors vii

Section II. Knowledge of Our Own Minds and Meanings

Section III. Truth, Objectivity, and Relativism

Section IV. Warrant, Transmission Failure, and Scepticism

Index Nominum 487

Dorit Bar-On is Professor of Philosophy at the University of North Carolina at

Section I: Rule-Following and the

Section II: Knowledge of Our Own

entitled to do so because otherwise the communicative practice couldn’t get started.

Section III: Truth, Objectivity and Relativism

“cognitive command”, “width of cosmological role”, and the “order of determination

shortcoming, varying from ignorance or error, to inattention and distraction, up to

attempt to formulate relativism in terms of irresoluble conﬂict. Either both parties to

Section IV: Warrant, Transmission Failure

strategy of response to both forms of scepticism.21 Such a strategy grants to sceptics of

1980 Wittgenstein on the Foundations of Mathematics, London, Duckworth.

Articles and Critical Studies

1998 “Euthyphronism and the Physicality of Colour”, in R. Casati and C. Tappolet

1985 “In Defence of the Conventional Wisdom”, in I. Hacking (ed.) Exercises in

Forthcoming “An Entitlement to Hume’s Principle?”, in P. Ebert and M. Rossberg

Wittgenstein, Mind, Rules and Privacy

1981 “Rule-following, Objectivity and the Theory of Meaning”, in C. Leich and

1998 “Self-knowledge: The Wittgensteinian Legacy”, in C. Wright, B. Smith and

2. What is a Rule? What is Rule-Following?

(Castle) If the conﬁguration is C, you may castle.

Whether or not this can be attributed to Wright, it is worthwhile seeing what is

Let us take a clear case of personal-level rule-following. Suppose I receive an email

I learned—and internalized instructions for—a rule, which determines how addition is to be

(Acceptance) If S is following rule R (‘If C, do A’), then S has somehow accepted R.

3. Acceptance, Intention, and Wright’s Problem

This is the position that I earlier called the Intention View.11

4. Problems for the Intention View

The ﬁrst reason would be provided by the assumption of Naturalism. A Naturalist

5. Are there Solutions to these Problems

must be at bottom non-intentional, because even intentional forms of rule-following

For meaning to be a matter of rule-following in the sense presupposed by the

6. Can the Intention View be Saved?

But what Wittgenstein says in reply is rather this:

Premise: This is an email that I have just received.

Conclusion: Answer it immediately!

(MP*) From ‘If C, do A’ and C, conclude ‘do A’!

7. Intentions and Intentional States

Even without assuming Naturalism as an a priori constraint on the acceptability of a

8. Rule-Following without Intentionality: Dispositions

9. Is Going Subpersonal the Solution?

1. The Intuitive Conception

2. A Positive Account of Understanding-Based

judgment rationally on the basis of a reason-giving state that is recognized as such in

3. A Comparison with Wright’s Wittgenstein

(ii) Patterns and Rails to Inﬁnity

(iii) Individual Understanding and its Objectivity

about attributions to others are facilitative of our making correct attributions of

(iv) Knowing Your Own Meaning: the Reasons-Awareness Account

4. Between Linguistic Idealism and Extreme Platonism

It may be tempting to hold that the rule-following considerations, under a Wright-

Few contemporary philosophers would endorse an unrestricted linguistic idealism.

always, on the explanatory power of that attribution in explaining the thinker’s

5. The Possibility of Substantive Philosophy

6. Are These Explanations in Terms of Tacit

A ﬁrst doubt may be that computational explanation by states of tacit knowledge is