KNOWLEDGE, TRUTH, AND DUTY

This page intentionally left blank

KNOWLEDGE, TRUTH,
AND DUTY
ESSAYS ON EPISTEMIC JUSTIFICATION,
RESPONSIBILITY, AND VIRTUE

EDITED BY
Matthias Steup

OXJORD
UNIVERSITY PRESS

2001
OXTORD
UNIVERSITY PRESS

Oxford New York

Athens Auckland Bangkok Bogota Buenos Aires Calcutta
Cape Town Chennai Dar es Salaam Delhi Florence Hong Kong Istanbul
Karachi Kuala Lumpur Madrid Melbourne Mexico City Mumbai
Nairobi Paris Sao Paulo Shanghai Singapore Taipei Tokyo Toronto Warsaw
and associated companies in
Berlin Ibadan

Copyright © 2001 by Oxford University Press, Inc.

Published by Oxford University Press, Inc.
198 Madison Avenue, New York, New York 10016
Oxford is a registered trademark of Oxford University Press
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,
electronic, mechanical, photocopying, recording or otherwise,
without the prior permission of Oxford University Press.

Library of Congress Cataloging-in-Publication Data

Knowledge, truth, and duty : essays on epistemic justification, responsibility, and virtue /
edited by Matthias Steup.
p. cm
Includes bibliographical references and index.
ISBN 0-19-512892-3
1. Justification (Theory of knowledge) 2. Duty. I. Steup, Matthias.
BD212.K57 2000
121—dc2l 00-021215

1 3 5 7 9 8 6 4 2
Printed in the United States of America
on acid-free paper
 Acknowledgments

The following essays have been published previously:

Essay 1, Susan Haack, '"The Ethics of Belief Reconsidered," in The Philosophy
of Roderick M. Chisholm, edited by Lewis Edwin Hahn, vol. xxv of The Library of
Living Philosophers (Chicago and LaSalle, 111.: Open Court, 1997). Reprinted by per-
mission of the Open Court Publishing Company.
Essay 7, Alvin Goldman, "Internalism Exposed," The Journal of Philosophy XCVI,
no. 6 (1999): 271-93. Reprinted by permission of The Journal of Philosophy.
Essay 11, Ernest Sosa, "Reflective Knowledge in the Best Circles," The Journal
of Philosophy, XCIV, no. 8 (1997): 410-30. Reprinted by permission of the Journal
of Philosophy.
This page intentionally left blank
 Contents

Contributors ix

Introduction 3

Part I: Epistemic Duty and the Normativity of Justification

1. "The Ethics of Belief" Reconsidered 21

S U S A N HAACK

2. Epistemic and Moral Duty 34

BRUCE RUSSELL

3. Epistemic Justification and Normativity 49

RICHARD FUMERTON

Part II: Epistemic Deontology and Doxastic Voluntarism

4. Deciding to Believe 63
C A R L GINET

5. Voluntary Belief and Epistemic Evaluation 77

RICHARD FELDMAN
viii Contents

6. Doxastic Voluntarism and the Ethics of Belief 93

R O B E R T AUDI

Part III: Epistemic Deontology and the Internality of Justification

7. Internalism Exposed 115

ALVIN GOLDMAN

8. Epistemic Duty, Evidence, and Internality 134

MATTHIAS S T E U P

Part IV: Justification and Truth

9. Truth as the Epistemic Goal 151

M A R I A N DAVID

10. Value Monism in Epistemology 170

M I C H A E L R. D E P A U L

Part V: Epistemic Virtue and Criteria of Justified Belief

11. Reflective Knowledge in the Best Circles 187

ERNEST SOSA

12. Commonsensism in Ethics and Epistemology 204

NOAH LEMOS

Part VI: Beyond Deontology

13. Knowing People 221

V R I N D A DALMIYA

14. Recovering Understanding 235

LINDA Z A G Z E B S K I

Index 253
 Contributors

Robert Audi Alvin Goldman

University of Nebraska University of Arizona

Vrinda Dalmiya Susan Haack

University of Hawaii University of Miami

Marian David Noah Lemos

University of Notre Dame De Pauw University

Michael R. DePaul Bruce Russell

University of Notre Dame Wayne State University

Richard Feldman Ernest Sosa

University of Rochester Brown University

Richard Fumerton Matthias Steup

University of Iowa St. Cloud State University

Carl Ginet Linda Zagzebski

Cornell University University of Oklahoma

ix
This page intentionally left blank
KNOWLEDGE, TRUTH, AND DUTY
This page intentionally left blank
 Introduction

I
This volume is divided into six parts, all of which bear, more or less directly, on
issues having to do with epistemic justification, responsibility, and virtue. The first
two essays in part I, written by Susan Haack and Bruce Russell, examine the relation
between epistemic and moral duty. In the third essay of this part, Richard Fumerton
examines the question of whether relating epistemic justification to an epistemic
'ought' warrants the widespread assumption that epistemic justification is a norma-
tive matter.
According to what Haack calls the special-case thesis, epistemic appraisal is merely
an instance of moral appraisal. Haack argues that we should reject this thesis as well
as the correlation thesis, according to which moral and epistemic appraisal are cor-
related such that whenever a positive (negative) epistemic appraisal of a belief is
appropriate, a positive (negative) moral appraisal of it is also appropriate. The kind
of case that, according to Haack, refutes these theses is of the following kind: (i) The
subject believes that p without having adequate evidence for p; (ii) the reason why
the subject believes thatp lies in personal or cultural cognitive inadequacy, a kind of
inadequacy for which the subject bears no responsibility. Because of (i), the subject's
belief is epistemically unjustified. Because of (ii), however, a negative moral appraisal
would be inappropriate. Thus not all cases of epistemically unjustified belief are cases
in which an unfavorable moral judgment is appropriate.
So epistemic appraisal is neither a special case of, nor correlated with, moral ap-
praisal. However, Haack thinks that there is overlap between them and thus endorses
the overlap thesis: An unfavorable moral appraisal of epistemically unjustified be-

3
4 Introduction

lief is sometimes appropriate. Clifford's famous example is a case in point: The

shipowner's belief in the seaworthiness of his vessel led to great harm and thus was
not merely epistemically but also morally unjustified. Regarding the Clifford-James
controversy, Haack argues that it is always epistemically wrong to believe without
sufficient evidence and in this respect sides with Clifford. But she also sides with
James inasmuch as she rejects the Cliffordian claim that it is always morally wrong
to believe on insufficient evidence.
Russell's essay begins with the distinction between subjective and objective
moral justification. An action is an instance of subjective moral justification, if,
and only if, the agent has no adequate reason to think it is wrong, or has a legiti-
mate excuse for it. An action is an instance of objective moral justification if it is
not forbidden. For example, a doctor who unintentionally kills his patient is sub-
jectively justified, objectively unjustified. A killer who intends to poison his vic-
tim but actually cures her of a fatal disease is objectively justified, subjectively
unjustified. Likewise, there is subjective and objective epistemic justification. A
belief is subjectively justified if and only if it is not epistemically blameworthy.
What determines whether a belief is objectively justified is whether it is supported
by adequate evidence. Consider a belief that results from wishful thinking. Under
normal circumstances, such a belief is both subjectively and objectively unjusti-
fied. However, if the subject lives in a society in which wishful thinking is accepted
as a common practice, he cannot be blamed for engaging in wishful thinking. In
that case, if the subject believes something because of wishful thinking, his belief
is subjectively, but not objectively, justified.
Russell's main thesis is that knowledge requires the combination of both subjec-
tive and objective justification. This allows him to rebut certain counterexamples by
Plantinga, the purpose of which is to challenge the necessity of justification for knowl-
edge. Furthermore, Russell also replies to Plantinga's London bus example, suggesting
that it involves mistaken assumptions about the nature of subjective and objective
justification.
In the third essay of part I, Richard Fumerton examines the claim, shared by many
contemporary epistemologists, that epistemic justification is a normative matter.1
Fumerton argues that it is by no means clear what this claim amounts to. He con-
siders various candidates for explaining normativity and finds each one of them prob-
lematic. First, we might say that the epistemic 'ought' functions in analogy with the
moral 'ought.' But in which sense can we take the moral 'ought' to be normative if
we subscribe to cognitivism; the view that epistemic judgments have truth values?
For according to cognitivism, true judgments involving the moral 'ought' describe
objective reality, just as true judgments that do not involve an 'ought' do. Second,
we might want to explicate epistemic normativity via an appeal to rules. However, is
all moral evaluation reducible to rules? And does it make sense to ground normativity
in rule-conformance if we are cognitivists? Third, it might be suggested to ground
epistemic normativity in its relation to the goals that are appropriate to cognitive
conduct. Here the problem is to avoid the circularity that results from analyzing
epi stemically justified belief in terms of the goal of believing what we would be jus-
tified in believing. Finally, Fumerton considers the proposal to analyze epistemic
normativity in terms of epistemic blame and praise. His objection to this approach is
 Introduction 5

that the connection between epistemic justification and praise/blame is a tenuous one.
The gist of Fumerton's arguments, then, is that it is not all that easy to pin down exactly
what the claim that epistemic justification is normative amounts to. The chief prob-
lem would appear to be that cognitivism, the prevalent view among epistemologists
that epistemic judgments describe objective features of the world, makes it difficult
to explain why epistemic judgments should be any different from other kinds of judg-
ments that describe objective reality.

II

In his influential essay, "Concepts of Epistemic Justification," William Alston has

stated an objection to epistemic deontology that is commonly referred to as the "ar-
gument from doxastic involuntarism."2 According to Alston, there can be an obliga-
tion to believe that p, or an obligation to refrain from believing that p, only if we
have direct voluntary control over our beliefs. But according to Alston, we do not.
As he succinctly puts the matter: "When I see a car coming down the street I am not
capable of believing or disbelieving this at will."3 This point—that we cannot choose
at will between believing and not believing—applies, Alston argues, to the vast
majority of our beliefs. Alston's argument raises two critical issues: (i) Is it indeed
correct that we cannot exert the kind of voluntary control over our beliefs that deon-
tology requires? (ii) Does the application of deontology to belief require in the first
place that we can exert voluntary control over our beliefs? The three essays of part II
address these issues.
In his contribution to this volume, Richard Feldman argues that doxastic involun-
tarism—the thesis that our beliefs are completely beyond our control—is false. How-
ever, although we have some kind of control over our beliefs, it must be conceded
that it is not of the right kind. Since it is not, the argument from doxastic involuntarism
can easily by modified to yield the intended conclusion. Next, Feldman discusses an
argument of mine for the conclusion that we have control over belief after all. Ac-
cording to this argument, we can make doxastic decisions because we can deliber-
ate. Feldman replies that a belief can be involuntary even if it results from delibera-
tion. What is needed to make belief voluntary is intentionality. But beliefs that result
from deliberation are not beliefs that exemplify intentionality. Consequently, from
the fact that we can deliberate about what to believe, it does not follow that we have
voluntary control over what to believe.
Feldman, then, accepts that we do not have the kind of control over beliefs that
we have over actions, but rejects the second premise of the argument from doxastic
involuntarism: the claim that such control is necessary. The epistemic 'ought,' ac-
cording to Feldman, is a role 'ought.' Judgments involving such an 'ought' tell us
how we ought to act if we play a certain role. Epistemic evaluations tell us that, as
beings who form beliefs, we ought to follow our evidence. This is something we ought
to do even though we do not have control over whether we do so or not.
In marked contrast to Alston and Feldman, Carl Ginet makes a case for the thesis
that we are sometimes in the sate of believing that p simply because we have decided
to believe that/). He proposes an analysis of the following locution:
6 Introduction

In deciding to A, S decided to believe that p if and only if in deciding to A S

decided to count on its being the case that p.

To count on its being the case that p is to dismiss the possibility of not-/), to be sur-
prised should it turn out that not-/?. What comes into being when one decides to count
on its being the case that p is the disposition to count on the truth of that statement in
closely similar situations. But that we can acquire a disposition by deciding to have it
should be unproblematic. Consequently, it should be unproblematic to assume that a
belief that p is the sort of thing we can acquire by deciding to acquire it. However,
Ginet points out that, although deciding what to believe is something we can do, it is
something we do only rarely. In the last two sections of his essay, he discusses and
rejects arguments for doxastic involuntarism by Bertrand Williams and William Alston.
In the third essay of this part, Robert Audi attempts to find a place for an ethics of
belief although he rejects doxastic voluntarism, which he takes to be the thesis that
belief is sometimes under direct control. He distinguishes between a behavioral and
a genetic version of voluntarism. According to the behavioral version, believing it-
self is sometimes an action-type; according to the genetic version, the formation of
a belief is sometimes an action-type. Audi reject the behavioral version on the ground
that actions are events, beliefs are not. Thus if volitions are actions of the will, then
beliefs cannot be volitions. He rejects the genetic version as well. Distinguishing
between causing oneself to believe something, and a belief's forming, he likens be-
lief formation to the forming of a damp ring on a glass of water. Belief formation,
then, is not causing oneself to believe something, but simply a belief's forming. But
if this is what belief is like, in which sense can we speak of an ethics of belief? At the
end of his essay, Audi suggests that there are various ways in which an ethics of belief
can emerge. For example, we may say that it is morally objectionable to conduct one's
intellectual life sloppily. Furthermore, we may accept an epistemic obligation to be
attentive to one's evidence, from where may be led to an ethics of inquiry.

Ill

The two essays of the third part are about the relation between epistemic deontology
and the internality of justification. In the first essay, Alvin Goldman subjects
internalism—the view that the things that determine whether a belief is justified must
be suitably accessible to the subject—to a penetrating critique. In the second essay,
I defend internalism against Goldman's objections.
Goldman begins his critique of internalism by reconstructing the rationale for it,
which he takes to be derived from what he refers to as the "guidance deontological"
conception of justification. This conception imposes an accessibility constraint on
justifiers: If they are to guide the subject in meeting their epistemic duties, then they
must be readily knowable. The rationale derived from this conception, however, pre-
sents the internalist with a dilemma. The desideratum is a conception of internality that
identifies reliabilism as an externalist theory. The dilemma arises because the accessi-
bility constraint as derived from the guidance-deontological conception of justifica-
tion does not meet this desideratum, whereas the kind of accessibility constraint that
 Introduction 7

does meet the desideratum cannot be derived from the guidance deontological con-
ception. Moreover, internalism faces numerous problems when it comes to avoiding a
collapse into skepticism. Some justified beliefs are justified by evidence the subject
has forgotten. Goldman argues that internalism lacks the resources to identify such
beliefs as justified ones and thus leads to skepticism. Furthermore, it does not allow for
logical and probabilistic relations, nor for epistemic principles, to play the role of
justifiers and thus leads to skepticism once again. And since internalism demands that
justifiers be accessible, it faces a problem with regard to the availability of the methods
that must be employed to recognize justifiers, and thus runs again into the problem of
skepticism. At the end of his paper, Goldman objects to internalism on the ground that
the a priori methodology with which it is associated is problematic. For as research in
cognitive science has shown, which characteristics determine whether, for example,
memorial beliefs are justified is a question that must be answered empirically.
In my response to Goldman, I propose an alternative rationale for internalism. I
agree with Goldman that the accessibility constraint does not give internalists what
they want: a conception of internality according to which reliability cannot play the
role of a justifier. However, suppose we replace the accessibility constraint with an
evidential states constraint, according to which only evidential states qualify as
justifiers. This constraint effectively bars reliability from the club of justifiers and
thus gives internalists the desired notion of internality. It would be a mistake, how-
ever, to think that what internalists want is to denounce reliability as altogether irrele-
vant. As I argue in this essay, internalism need not be construed as a view according
to which reliability is altogether without epistemological significance. Rather, what
internalists and externalists disagree about might merely be the proper role of reli-
ability within the analysis of knowledge. And its proper role is, as I would suggest,
that of an ingredient in the fourth condition of knowledge, the job of which is to make
the analysis of knowledge immune to Gettier cases.
So one can motivate internalism without having to maneuver oneself into a cor-
ner from where a proper appreciation of what externalism has to offer is not pos-
sible. However, can internalism be defended against Goldman's charge that, for a
plethora of reasons, it has skepticism-breeding consequences? I argue that it can. What
is common to the battery of Goldman's objections is the assumption that, if we under-
stand justification deontologically, we cannot have justified beliefs unless we also
form (or are at least capable of forming) correct beliefs about which justifiers justify
our beliefs, about the logical/probabilistic relations by virtue of which our beliefs
are justified, and about the epistemic principles that license our beliefs. I reject this
assumption, for it seems to me that, to have justified beliefs, we must meet one, and
only one, condition: We must have adequate evidence. But if having adequate evi-
dence is indeed sufficient for deontological justification, then internalism does not
have the skeptical consequences Goldman attributes to it.

IV

What is distinctive about epistemic justification? After all, our beliefs can also be
justified morally or prudentially. Clearly, though, these types of justification are to
8 Introduction

be distinguished from epistemic justification. Exactly how is the difference to be

explained? As Marian David points out in his essay "Truth as the Epistemic Goal,"
to explain the nature of epistemic justification, epistemologists, regardless of what
particular approach they favor, tend to invoke the idea that what makes justification
epistemic is its relation to what David calls the truth-goal: roughly, the goal of be-
lieving what's true and not believing what's false. However, invoking the truth-goal
is anything but unproblematic. To begin with, there is the problem of whether as-
cribing the truth-goal to us gets our psychology right. Furthermore, even supposing
that most of us do have the truth-goal, it is not clear that having justified/unjustified
beliefs requires having the truth-goal, for it seems plausible to say that those who do
not have the truth-goal have nevertheless justified/unjustified beliefs. A different set
of problems has to do with the question of how the truth-goal ought to be expressed.
Is it the goal with regard to every proposition p to believe p if, and only if, p is true?
Thus understood, the goal seems excessively demanding. However, after reviewing
alternatives and indicating their respective problems, David argues that we might as
well stick to our original version of the truth-goal.
A further set of problems arises when we attempt to spell out exactly how indivi-
dual justified/unjustified beliefs are related to the truth-goal. Exactly what makes
justified/unjustified beliefs good/bad in relation to that goal? One answer is to say
that the relation is that of a causal ends-means relationship. David argues that such a
diachronic conception of justification is unacceptable, for future causal consequences
of a belief do not affect its justificational status in the present. Another answer is that
the relation is one of constitution: having a justified belief is part of what's involved
in reaching the truth-goal. However, David argues that this approach collapses justi-
fied belief into true belief. As a constitutive part of the truth-goal, a true belief is a
good thing even if it is unjustified, and a false belief is a bad thing even if it is justified.
How can this collapse be blocked? One option is to abandon the truth-goal, that
is, to view the truth-condition and the justification-condition in the analysis of
knowledge as conditions that are independent of each other. However, if that op-
tion is exercised, the problem of explaining the nature of justification might re-
main unsolved. Another option is to adopt two goals: truth and justification. Un-
fortunately, this approach does not allow for justified false beliefs, for in relation
to the truth-goal, having a false belief is a bad thing, even if that belief is justified
and thus meets the other goal. A third option is to conceive of the relation in the
way Chisholm suggested: If I aim at the truth-goal, it is rational to prefer justified
to unjustified beliefs. But how is the relevant notion of rationality to be understood?
The attempt to answer this question will run into either a circularity problem, or
the difficulties of the ends-means-relationship approach, or the problem of aban-
doning the truth-goal altogether. The final option David considers is reliabilism,
which blocks the collapse because, according to it, justified beliefs are related to
the truth-goal since they are produced by processes that produce mostly true be-
liefs. However, David argues that reliabilism is implausible as long as it remains a
diachronic theory: a theory that makes justification dependent on the performance
of cognitive processes in the past and future. He concludes the essay by consider-
ing an alternative version of reliabilism, one that involves a subjunctive truth-goal
and a subjunctive conception of reliability.
 Introduction 9

In the second essay of this part, Michael DePaul argues against what he calls value
monism in epistemology: the view that truth is the only intrinsic epistemic good. From
the point of view of monism, justification is valuable only as a means to true belief.
However, value monism is dominant only within epistemology proper, the project
of analyzing the concept of knowledge. It is not dominant in broad epistemology,
where we find a pluralism of epistemic values. For example, Foley-rationality is an
epistemic good, but it is neither necessary nor sufficient for turning true belief into
knowledge.4 Consequently, Foley-rationality is an epistemic value within broad epis-
temology but not within epistemology proper. Within epistemology proper, as DePaul
claims, value monism is the dominant view. And DePaul rejects value monism be-
cause it is inconsistent with what he takes to be an obvious truth: Knowledge is bet-
ter than mere true belief. To see what's wrong with value monism, we need to con-
sider the following two propositions:

(1) Truth is the only intrinsic epistemic good,

and

(2) Knowledge is better than mere true belief.

DePaul argues that (1) and (2) cannot both be true. Since he takes (2) to be an obvi-
ous truth, he concludes that (1) is false. What, however, is the alternative to value
monism in epistemology? According to DePaul, the alternative is value pluralism,
the view that in addition to truth there are further intrinsic epistemic goods, such as
knowledge and justification. However, as DePaul points out in the concluding para-
graph of his essay, from this pluralist point of view, justification must be good inde-
pendently of any connection between justification and truth.

The essays in this part address issues having to do with the question of how, in the
attempt to analyze knowledge, we can choose philosophical starting points without
falling into vicious circularity. According to both Ernest Sosa and Noah Lemos, our
starting points are reasonable and can be chosen without vicious circularity if they
arise from the exercise of virtuous cognitive faculties. What makes cognitive facul-
ties virtues is, however, an external matter that goes beyond the fulfillment of epis-
temic duties. On the other hand, neither Sosa nor Lemos mean to propose what could
accurately be labeled an externalist account of epistemic justification. Their papers,
then, illustrate ways in which external elements can be appreciated even by those
who think that knowledge involves an internal dimension as well.
According to skeptics, we cannot know how we know, for we cannot argue for
the reliability of our faculties without vicious circularity. Sosa argues that externalist
virtue epistemology affords us a way to stay clear of skepticism. The externalist ele-
ment of Sosa's virtue epistemology is motivated by the circularity problems we can
find in Descartes and Moore. How can Descartes know he is thinking without know-
10 Introduction

ing in the first place that clear and distinct perceptions are perfectly reliable? How
can Moore have knowledge of his hand without knowing he is not dreaming in the
first place? In response to these questions, Sosa recommends externalism: For the
strategy employed by Descartes and Moore to work, it merely must be the case that
they know their starting points; it is not necessary they know how they know their
starting points. Do they, however, know their starting points? According to Sosa,
they do, if their starting points arise from virtuous, or apt, faculties. The key idea
here is that a belief amounts to knowledge—exemplifies epistemic virtue—if its truth
is not an accident.
This account faces an important objection: If externalist virtue epistemology were
right, we could justify using a crystal ball, relying on the crystal ball's own testi-
mony as to its reliability. Sosa replies that even if the users of crystal balls can attain
justification just as well as ordinary perceivers—assuming, of course, things are
equal with regard to internal coherence—a crucial difference emerges, neverthe-
less, when we consider knowledge: While the users of crystal balls do not have
knowledge, ordinary perceivers do, for unlike crystal ball gazing, perception is
reliable. Sosa's account, then, is externalist, for knowledge requires no more than
the exercise of apt faculties. But the kind of knowledge thus attained is mere ani-
mal knowledge, not reflective knowledge. How, though, can the latter kind of
knowledge be achieved? Sosa proposes to follow the lead of Descartes and Moore.
We can begin with instances of animal knowledge as starting points and use these
to advance explanations of how we know, thus giving our account explanatory
coherence. Sosa, then, makes a case for virtue perspectivism, according to which
reflective knowledge—first-order knowledge plus knowledge of how we know—
involves two essential components: external aptness as well as coherence internal
to the subject's epistemic perspective.
Like Sosa's, Lemos's essay is concerned with the question of how we can, with-
out falling into circularity, derive criteria of justification and knowledge. The strat-
egy Lemos prefers is to appeal to common sense beliefs, among which he includes
epistemic beliefs such as "I know that I have two hands." What makes it reasonable
of us to appeal to such beliefs is that they have a positive epistemic status. But wouldn't
we first have to show that such beliefs have a positive epistemic status? Lemos
argues that this objection presupposes an unacceptable theory of justification.
Next, Lemos examines the objections Brandt and Hare have directed against com-
mon sensism in ethics. They have argued that the appeal to common sense beliefs is
parochial—it merely shuffles our own, culturally inherited biases around—and more-
over involves vicious circularity. To the first of these objections, Lemos replies that
the methods Brandt and Hare prefer, to go outside of ethics and appeal to consider-
ations of logic and rationality, are unlikely to be any less parochial. In response to
the second objection, he proposes that the attribution of reliability to one's own cog-
nitive faculties may be derived from the attribution of knowledge to oneself, that is,
from an epistemic common sense belief one accepts to begin with. Such a procedure
faces the objection that it is as well available to madmen and crystal-ball users. In
reply to this worry, Lemos endorses Sosa's strategy. The relevant difference is that,
whereas our common sense beliefs have epistemic virtues as their source, the beliefs
of madmen and crystal-ball gazers do not.
 Introduction 11

VI

The duty-based approach to understanding epistemic justification is challenged not

only by externalists such as Goldman, but also by those philosophers who favor a
virtue-based approach to epistemology. According to Linda Zagzebski, virtue epis-
temology ought to be construed in analogy to the virtue approach in ethics and thus
ought to be not act-based but agent-based. According to her, the evaluation of par-
ticular beliefs must be derived from an evaluation of the agent.5 The essays in the
last part of this volume represent this approach.
In the first essay of this part, Dalmyia makes a case for redirecting epistemology
toward a new paradigm, the two most important elements of which are virtue-ethics
and the method of care. According to virtue epistemology, what confers epistemic
value are properties of the epistemic subject: her epistemic character, belief-forming
habits, or cognitive dispositions. The method of care is a complex, interactive pro-
cess of acquiring justified beliefs or knowledge, a process that integrates the subject
into a social and ethical context. Starting out with a discussion of knowledge of other
minds (selves), Dalmyia moves on to an examination of the role the knowing self
plays within the kind of epistemology she has in mind. Epistemic responsibility is
one important element of that kind of epistemology. It cannot, however, be reduced
to epistemic duty-fulfillment. Rather, what matters within care-based epistemology
is to cultivate and reinforce attitudes that are deemed admirable in the epistemic
community. Dalmyia, then, argues for a departure from epistemic deontology and
the internalism that is connected with it.
In the concluding essay of this part and this volume, Zagzebski proposes an analysis
of the concept of understanding, which, as she argues, epistemologists have neglected.
According to Zagzebski, there are three important strands that can be found in Plato
and Aristotle's thinking about understanding: (i) understanding arises from techne,
that is, from practical activities that are not purely cognitive; (ii) understanding has
to do with mastering an entire field of interrelating parts; (iii) understanding involves
representing the world nonpropositionally, for example, through visualization or
diagrams. Taking (iii) to be the defining characteristic, Zagzebski proposes that under-
standing is a state of comprehending nonpropositional structures of reality, such as
automobiles, pieces of music or art, the character of a person, or a causal nexus. Thus
Zagzebski takes understanding to be something that is essentially different from
knowledge, which, unlike understanding, has a prepositional object. Consequently,
she argues that virtue epistemology is better suited than non-virtue epistemology to
provide an adequate analysis of understanding. For unlike the theories from which it
departs, virtue epistemology takes the objects of valuable epistemic states to consist
of not only propositional but nonpropositional objects as well.

VII

In his contribution to this volume, Bruce Russell refers to and endorses my view on
the issue of whether we have the kind of control over our beliefs that epistemic de-
ontology requires, and in his essay, "Voluntary Belief and Epistemic Evaluation,"
12 Introduction

Feldman discusses and rejects my view. It seems to me, therefore, that I should pro-
vide a brief account of my view here.
Let us call the control thesis the claim that we have over our beliefs the kind of
control that epistemic deontology—the evaluation of beliefs in terms of epistemic
duty, or an epistemic 'ought'—requires. Suppose we argue against the control thesis
by comparing a typical belief, say the belief that it is raining when you see that it is,
with a typical action such as taking a walk. This comparison seems to suggest that
belief and action differ with regard to our control over them. When you perceive that
it is raining you cannot refrain from believing that it is; that is, your belief that it is
raining is forced on you. However, taking a walk is a different matter. Few people
take walks because of internal compulsion or external force. Rather, whether or not
to take a walk is typically a matter of choice/This comparison lends itself to gener-
alization. Perceptual beliefs about our environments and introspective beliefs about
our mental states are hardly a matter of choice. The same would apply to memorial
beliefs and beliefs about logical and mathematical matters. Actions, however, seem
different. True, some people act sometimes out of compulsion. However, ordinarily
we do enjoy choice with regard with what we do. And so the conclusion seems ines-
capable that, while we have (at least normally) control over our actions, beliefs are,
as Alston suggests, just as much beyond our control as are cell metabolism and the
secretion of gastric juices.6
The chief problem with this argument is that it bypasses the question of whether
we really have control over our actions in the first place. After all, there is such a
thing as determinism, a view that cannot be dismissed lightly. So anybody who thinks
we have control over our actions would have to take a stand in response to the fol-
lowing questions: Is determinism true? If it is, how is control over our actions pos-
sible? And if determinism is not true, how is control over our actions possible in that
case?7 An argument for or against the control thesis doesn't get to the bottom of the
issue unless these questions are answered in the first place. And when they do get
answered, we get, it seems to me, the following result. If we take determinism to
be true and nevertheless think we have control over our actions, that is, if we are
compatibilists, then it turns out that we have plenty of control over our beliefs and
certainly as much as is needed for deontology. If, on the other hand, we are in-
compatibilists and take determinism to be false, we do get the result that we can control
what to believe only to a very limited extent. However, as I will argue below, it will
in that case turn out that if we are prepared to apply deontological evaluation to ac-
tions, there is no reason not to apply such evaluation to beliefs.
Suppose, then, we take compatibilism to be true. As compatibilists we accept
determinism, the view that every event and every state has a sufficient causal condi-
tion. But if determinism is true, then any action of mine is such that, given ante-
cedent history and the laws of nature, it had to occur, which is to say that no alterna-
tive path of action was open to me at the time I performed that action. If this is so, the
question arises of how I can have control over my actions at all. There are many moves
that, as compatibilists, we can employ for the purpose of attempting to reconcile de-
terminism and control. Let us consider just two. First, we might employ suitable con-
ditionals. We might say, for example, that my doing x was something over which I
had control—despite the fact that my doing x was causally determined—because if
 Introduction 13

I had decided not to do x, I would not have done x. It is hard to see how the claim that
we do not have control over our beliefs can be sustained if we endorse this sort of
compatibilism. For if what puts us in control of our actions is the truth of the sort of
conditional we just considered, then by far most of our beliefs turn out to be under
our control as well.
Consider perceptual beliefs. While taking a walk in the park you see a dog on the
lawn. So you believe there is a dog on the lawn. But suppose you assess your overall
evidence and conclude (sincerely, that is) that you are hallucinating. What reason is
there to deny that you would then not believe that there is a dog on the lawn?8 In
general, what reason is there to suppose that the relevant conditionals about actions
are true but that the relevant conditionals about beliefs are false? It is true that, since
our perceptual evidence typically supplies us with decisive evidence, perceptual
beliefs are typically forced on us. But they are no more forced on us than those of
our actions and omissions that are supported by decisive reasons, such as stepping
on the brakes when approaching a busy intersection or not throwing oneself in front
of an oncoming bus. According to the kind of compatibilism under consideration
now, we have control over such actions and omissions, although they are the result
of causal determination, because had our reasons been different, we would have acted
differently. The parallel move works just as well for perceptual beliefs: Neglecting
exceptions, for any such belief we can say that, had your total body of evidence been
different (e.g., had you had reasons for doubting the reliability of your perceptual
faculties), you would have refrained from acquiring (or sustaining) it. The same can
be said for memorial and introspective beliefs.9 It seems appropriate, then, to con-
clude that the version of compatibilism we just considered does not lend itself very
well for claiming that we can control what we do but not what we believe.10
Alternatively, compatibilists might say that we enjoy control over those of our
actions that have the right kind of causal history and have control over those states
of affairs that bear the right kind of causal relation to us. But surely, with regard to
this criterion, our beliefs are on a par, not with cell metabolism and the secretion of
gastric juices, but rather with our actions. Unlike those processes, our beliefs are
sensitive to our evidence and respond to reasoning. There are neural pathways lead-
ing from evidential states and the activity of reasoning to the formation of doxastic
states: states of believing, not believing, and disbelieving, just as there are neural
pathways that lead from practical reasons to actions. If the latter connection brings
our actions under our control, why should the former connection not bring out be-
liefs under our control? Of course, neural pathways and other requisite connections
are sometimes blocked so as to render control impossible. Sometimes beliefs result
from fears, desires, neuroses, or paranoia. In such cases, their causal origin puts them
beyond the reach of reason and thus beyond our control. It would be exceedingly
implausible, however, to suggest that all of our beliefs are like that. Rather, it seems
to me that if we endorse the causal history criterion, the vast majority of our beliefs
turn out to be no less under our control than our actions are.
So if involuntarists accept compatibilism, they face the following general prob-
lem. The starting point they accept, determinism, raises the question of how we can
have control over anything at all. To solve this problem, compatibilists employ de-
vices that allow them to reconcile control with causal determination. These recon-
14 Introduction

ciliation devices, however, are equal opportunity devices. Unless a special case is
made, we cannot reasonably take them to discriminate between beliefs and actions.
Of course, I have not even begun to rummage in earnest through all the devices that
can be found in the compatibilist' s treasure box. My point, then, can be understood
as a request for clarification. If you are a compatibilist, and if you think that whereas
our actions are under our control, our beliefs are not, then you had better come up
with a special explanation of why the compatibilists' reconciliation devices bring
under our control actions but not beliefs.
Alternatively, the involuntarist might favor libertarianism. There are, however,
different versions of libertarianism, and thus we cannot simply identify one standard
way in which a libertarian might make a case to the effect that unlike actions beliefs
are beyond our control. I will, therefore, merely indicate how I would approach this
matter. I agree with Peter Van Inwagen that, whereas compatibilism makes control
over action easy to come by, libertarianism makes this rather hard.11 Ironically, then,
libertarianism has the consequence of significantly shrinking the range of things over
which we enjoy control. This is so because it would be rather implausible of the liber-
tarian to assert that we can always freely choose between doing x and not doing x.
Rather, such a choice is possible only when decisive reasons either for or against
doing x are absent, which is sometimes the case but not very frequently.12 The same
applies to beliefs. Though certainly not all of them, by far most of our beliefs are
supported by decisive evidence.13 If, as libertarians do, we identify control over be-
liefs with the dual ability to believe that/? and not to believe thatp under exactly the
same circumstances, then the majority of our beliefs lies beyond our control. The
same applies to our actions and omissions: To the extent that they are supported by
decisive reasons, they also lie beyond our control.
Consider approaching a red light at a busy intersection. Decisive moral and pru-
dential reasons demand that you step on the brakes. It seems to me that libertarians
should admit that you have no choice here: It is not within your power to refrain from
doing this. The same applies to beliefs that are supported by decisive reasons. For
example, your perceptual beliefs about your environment are (typically) supported
by decisive evidence. If we are libertarians, then I think we should admit that it is not
within our power to reject these beliefs. Note, however, that we do not hesitate to
say that, when approaching a busy intersection, you ought to step on the brakes. It is
your moral and prudential duty to do so. The fact that you cannot do otherwise is no
obstacle to this verdict. But if that much is agreed on, then libertarians can with equal
right apply deontological evaluation to beliefs. The perceptual beliefs you have about
your environment are the beliefs you ought to have, notwithstanding the fact that it
is not within your power to get rid of these beliefs as long as you have decisive rea-
sons for them.
Of course, libertarians believe that at least sometimes you can, under exactly the
same circumstances, choose between doing and not doing x, as well as choose be-
tween believing and not believing that/?.14 However, since this is only sometimes
so, libertarians can sustain the broad range of deontological judgments that we ordi-
narily make, only if they reject the principle that we are responsible at a time t for a
particular action or belief only if we have at t libertarian, dual control over that ac-
tion or belief. On the other hand, what motivates libertarianism to begin with is the
 Introduction 15

thought that without control there cannot be any responsibility at all. Thus one of the
problems libertarians face is to reconcile this thought with the denial of a straight-
forwardly direct connection between responsibility and control.
Let me sum up. Compatibilism makes control easy to come by. Consequently, it
does not lend itself very well for making a case for the claim that we lack control
over beliefs. Unless a special explanation is offered, we have no reason for thinking
the compatibilist reconciliation devices do not equally apply to beliefs. Libertarian-
ism, in contrast, makes control hard to come by. Consequently (and perhaps ironi-
cally) libertarianism lends itself well for making a claim for quasi-involuntarism: the
view that although what to believe is sometimes under our control, most of the time
it is not. However, the same quasi-involuntarism applies to actions and omissions:
Many, if not most of them, are beyond our control. I doubt libertarians favor limiting
the scope of deontological evaluation accordingly. And thus I conclude that if we
are libertarians and consider deontological evaluation of actions and omissions as
legitimate although so many of them are not under our control, then we must be pre-
pared to accept a parallel conclusion for the application of deontological evaluation
to beliefs.
The difference between the way compatibilists and libertarians must view the
issue of belief and control presents itself clearly when we ask how they ought to re-
spond to the following statement by H. H. Price: "If you are in a reasonable frame of
mind, you cannot help preferring the proposition which the evidence favors, much
as you may wish you could."15 Let us begin with compatibilism. Note that com-
patibilists do not merely say that causal determination is no obstacle to control and
responsibility but rather say that control and responsibility even require causal de-
termination. If your reasons, together with your attitudes, preferences, character traits,
and so on, did not causally determine your actions, in which sense could you then be
said in control of them? Thus, if an action is under our control, is must be the result
of the right kind of causal determination; that is, it must be the result not of causal
determination involving things like neuroses or paranoia but rather of causal pro-
cesses in which reasons figure appropriately. But why should this point not apply
with equal force to beliefs that are produced by causal processes in which epistemic
reasons figure appropriately? It seems to me that if epistemic reasons figure appro-
priately in a belief's causal history, then by parity of argument we should conclude
that they are just as much under our control as actions in whose causal history prac-
tical reasons figure appropriately.16 I do not think, therefore, that compatibilists are
in a position to appeal to Price's point as an argument for denying that we have con-
trol over our beliefs. Rather, they should say exactly the opposite: When the percep-
tual experience of seeing that it is raining "causes" us to believe that it is, then that
belief is within the scope of our control precisely because what causes it is a good
reason, and not something like an irrational fear or a neurotic compulsion.
The libertarian response to Price's point, on the other hand, looks rather differ-
ent. Libertarians would have to argue that, if a decision you made (practical or
doxastic) was a matter of causal determination, where the strength of your reasons
was one element of what determined your decision, then you were not in control.
However, the same point would have to be made for actions as well. And thus, if we
are libertarians, Price's observation poses no special problem for beliefs. Rather, if
16 Introduction

we are libertarians, then, given the fact that practical as well as epistemic reasons
can be compelling, that is, add up to causal determination, we face the task of having
to explain why deontological evaluation can be applied to both beliefs and actions
even on those occasion when they are causally determined.
To conclude, let us return to the comparison with which we began. We compared
our being unable to refrain from believing that it is raining when we perceive that it
is with our being able to choose between taking and not taking a walk. Since similar
comparisons can be listed ad nauseam, opponents of the control thesis might argue
that in general we have control over what we do but not over what we believe. But
from the compatibilist point of view, is there really a difference here? Suppose then
you did take a walk, and we say that while that is something over which you had
control because you could have refrained from doing it, your belief that it is raining
is not because you cannot refrain from believing this. However, if we are com-
patibilists, we must acknowledge that your belief and your walk are both the result
of causal determination. Yet compatibilists would say that if you took the walk not
because of compulsion but for a good reason, then you had control over it. But
shouldn't we then say the same about your belief that it is raining? After all, this
belief is not the result of compulsion. You wouldn't believe it's raining no matter
what your evidence is. Rather, you believe it's raining because your belief is sup-
ported by excellent evidence. Where, then, is the difference that would justify
compatibilists in saying that, whereas the walk was under your control, your belief
was not?
The matter looks different for libertarians. If nothing much is at stake with regard
to the question of whether to take a walk, then, so libertarians would argue, we can
indeed do both under exactly the same circumstances: decide to take a walk and decide
not to take a walk.17 Obviously, this is not the case with regard to deciding to believe
that it is not raining when we see that it is. That's a decision we cannot make. But
surely it would be wrong to conclude that our comparison establishes that belief and
action differ with regard to control. For, as libertarians should argue, the comparison
is not fair. It is misleading to compare a doxastic attitude that is decisively opposed
by our epistemic reasons with an action that is neither opposed nor demanded by
decisive practical reasons. Rather, what we should compare the belief case with is an
action that is decisively opposed by practical reasons, such as giving the first person
you see on the street a hefty blow to the nose. Once the comparison is carried out in
this way, what emerges is that we cannot decide to do either of these things. The
difference the comparison was meant to establish no longer exists.
Note that I have not set forth a straightforward defense of the thesis that we have
voluntary control over our beliefs. Rather, I have argued for the following two claims.
First, if we endorse compatibilism, then critics of epistemic deontology need an ex-
planation of why we have less control over our beliefs than over our actions. I do not
think there is such an explanation, but others disagree. Feldman, for example,
argues in his contribution to this volume that there is such an explanation. Second, if
we endorse libertarianism, then critics of epistemic deontology need again an expla-
nation of why beliefs are less suitable for deontological evaluation than actions are.
Again, I do not think that there is such an explanation, but doubtless there will be
others who disagree with that verdict.18
 Introduction 17

Notes

1. For statements to the effect that epistemic justification is normative, see, for example,
Roderick Chisholm, Theory of Knowledge, 3d ed. (Englewood Cliffs, N.J.: Prentice Hall,
1989); John Pollock, Contemporary Theories of Knowledge (Totowa, N.J.: Rowman and
Littlefield, 1986), 7; William Alston, "Concepts of Epistemic Justification," in Epistemic
Justification: Essays in the Theory of Knowledge (Ithaca, N.Y.: Cornell University Press,
1989), 83.
2. See "Concepts of Epistemic Justification," The Monist 68 (1985), as well as Alston's
paper, "The Deontological Concept of Epistemic Justification," Philosophical Perspectives 2
(1988). Both papers are reprinted in Alston's Epistemic Justification: Essays in the Theory of
Knowledge (Ithaca, N.Y.: Cornell University Press, 1989), 81-114 and 115-52.
3. See Alston (1989), 92.
4. See Richard Foley, Working Without a Net (New York: Oxford University Press, 1993).
5. See Linda Trinkaus Zagzebski, Virtues of the Mind: An Inquiry into the Nature of Vir-
tue and the Ethical Foundations of Knowledge (Cambridge: Cambridge University Press,
1996).
6. See Alston (1989), 122.
7. The problem is that it is hard to see how we can exert control over our actions, and thus
be responsible for them, if our reasons, choices, preferences, decisions, and the like are not
causes of our actions. In short, the point is that control and responsibility seems to require
causal determination.
8. Of course, you might still believe it because of cognitive dysfunction. For that kind of
case, I would say that your belief is beyond your control.
9. It is, of course, difficult to see how one could have countervailing evidence with re-
gard to many introspective beliefs or beliefs about logical or mathematical matters. Never-
theless, the point remains: If such evidence were to be had and was strong enough, we would
believe differently.
10. As an explanation of the relevant difference, it might be suggested that in response to
a practical reason I can do, or not do, a certain thing, but I cannot believe, or not believe, a
certain proposition. The problem with this argument is its bias in favor of practical reasons.
Why must it be the case that, for us to have control over our beliefs, we must be able to adopt
doxastic attitudes in response to practical reasons? Why doesn't it suffice that we can adopt
doxastic attitudes in response to epistemic reasons?
11. See Peter Van Inwagen, "When Is the Will Free?" in Timothy O'Connor, ed., Agents,
Causes, and Events: Essays on Indeterminism and Free Will (Oxford: Oxford University Press,
1995), 219-38.
12. Alternatively, libertarians might say that, while none of our actions are causally de-
termined, our beliefs are. In that case, I would want to know what the reason is for saying that
our actions, but not our beliefs, are under control. Why is it that, whereas decisive reason for
an action (like the ones you have for stepping on the brakes when approaching a red light at
a busy intersection) do not (in conjunction with other relevant factors) add up to causal deter-
mination, decisive reasons for a belief (say the reasons you have for believing that cats are
mammals) do? Libertarians might also say that nothing is causally determined. But if my not
believing that cats are insects does not have a sufficient causal condition, why is it neverthe-
less not within my power to decide in favor of a different doxastic attitude, that is, to believe
that cats are insects?
13. If this were otherwise, so much the better for doxastic voluntarism.
14. For example, supposing libertarianism is true, it seems plausible to me that you can
both decide to do x and decide not to do x when you are completely indifferent with regard to
the question of whether to do x or not. And, again supposing libertarianism to be true, I see
no objection to saying that, for example, if your reasons against believing that p are opposed
by a powerful desire to believe that p, deciding to believe that p and deciding not to believe
that p is both within your power.
18 Introduction

15. H. H. Price, "Belief and the Will," Proceedings of the Aristotelian Society, suppl. 28
(1954): 1-27.
16. Here it might be objected that reasons simply do not figure appropriately in the causal
history of our beliefs. But that assertion flies in the face of the evidence. Obviously reasons
are causally efficacious not only in the domain of action, but in the domain of belief as well.
In fact, I would say that the very reason why we seem not to be in control of our beliefs lies
in the fact that, for example, perceptual evidence is so remarkably effective in triggering the
beliefs it supports.
17. Alternatively, we could consider a case where you have decisive reasons for taking a
walk. In that case, the intended contrast between the action and the belief would immediately
disappear.
18. I defend an expanded version of the argument presented here in "Doxastic Voluntarism
and Epistemic Deontology,"Acto Analytica 15 (2000): 25-56. This article is also available
on my web-site (http://condor.stcloudstate.edu/~steup/).
Part I

EPISTEMIC DUTY A N D T H E
NORMATIVITY OF JUSTIFICATION
This page intentionally left blank
 1

\\
The Ethics of Belief" Reconsidered

SUSAN HAACK

What is the relation of epistemic to ethical appraisal? Possible answers include:

(1) that epistemic appraisal is a subspecies of ethical appraisal—henceforth, for short,
the special-case thesis',
(2) that positive/negative epistemic appraisal is distinct from, but invariably associ-
ated with, positive/negative ethical appraisal—the correlation thesis;
(3) that there is, not invariable correlation, but partial overlap, where positive/nega-
tive epistemic appraisal is associated with positive/negative ethical appraisal—the
overlap thesis',
(4) that ethical appraisal is inapplicable where epistemological appraisal is relevant—
the independence thesis;
(5) that epistemic appraisal is distinct from, but analogous to, ethical appraisal—the
analogy thesis.
I hope this list exhausts the serious options.1 But refinements will be needed to
take account of the fact that each of the positions listed has both a completely gen-
eral form ("for every dimension of epistemic appraisal"), and a variety of specific
forms (e.g., "where epistemic appraisal of someone as completely, or to some de-
gree, justified, or as unjustified, in believing that . . . is concerned"). The correct
account may be different with respect to different dimensions of epistemic appraisal.
But the logical relations among the positions listed are the same whether one consid-
ers each in its general form, or each in the same specific form. The special-case the-
sis is incompatible with any of the others. The correlation thesis is incompatible with
the overlap thesis and with the independence thesis. The analogy thesis, however,
though incompatible with the special-case thesis, is compatible with the correlation
thesis, with the overlap thesis, and even with the independence thesis.

21
22 Susan Haack

In "Firth and the Ethics of Belief," published in 1991, Chisholm writes that since
1938, when he and Firth both enrolled in Ralph Barton Perry's seminar on value
theory, "Firth's inclination was to say that [epistemic justification] is merely an ana-
logue [of ethical justification]; and my inclination was to say that it is a subspecies....
I still find myself inclined to accept the original view."2 For most of this paper, I
shall concentrate, like Chisholm and Firth, specifically on the relation of epistemic
to ethical justification. Here, as I see it, the special-case thesis is too strong, the anal-
ogy thesis (not false but) too weak; the relation of epistemic to ethical justification is
as stated in the overlap thesis: less intimate than partial identity, more intimate than
analogy.
Interwoven with my arguments for this specific version of the overlap thesis will
be some speculations of a more historical character: that in the celebrated debate
between Clifford and James, it is their shared failure to distinguish epistemological
from ethical justification which creates the false impression that one must choose
either the morally over-demanding account proposed in "The Ethics of Belief" or
the epistemologically over-permissive account proposed in "The Will to Believe." And
interwoven with these speculations will be an argument that locates Chisholm closer
to the Jamesian side of that debate, significantly more permissive epistemologically
than Clifford—and somewhat more permissive epistemologically than myself.
Finally, I shall turn my attention briefly to the relation of epistemic to ethical
appraisals of character, which, I shall suggest, seems more intimate than the relation
of epistemic to ethical justification. This will suggest a friendly, if revisionary, rein-
terpretation of what is plausible in Clifford's, and Chisholm's, talk of "the ethics of
belief."

Like Chisholm, I take for granted the essentially evaluative character of epistemo-
logical concerns, the focus on what makes evidence better or worse, what determines
to what degree a person \sjustified in a belief, how inquiry should be or is best con-
ducted. But when one thinks about the different ends on which epistemological and
ethical appraisal are focused, it looks likely that their relation is to be expected to be
at least as complex and oblique as the relation of knowledge to human flourishing.
This expectation is confirmed by reflection on such questions as: Is all knowledge
conducive to human flourishing, or is there some knowledge we should be better off
without? Is it always morally, as it is epistemologically, best to seek out all available
evidence, or are some means of obtaining evidence unethical? Is it always harmful
to believe unjustifiedly, or is it sometimes harmless or even beneficial?3
If it is possible that there should be cases where a person believes unjustifiedly,
but where the appropriate moral appraisal is favorable or indifferent, the claim that
to say that a person believes unjustifiedly is eo ipso to make an unfavorable moral
appraisal—henceforth, the special-case thesisj—is false. Unless, therefore, it is in-
coherent to claim that believing unjustifiedly is sometimes simply harmless, or, as
some philosophers have done, that there is moral merit in faith, or in a husband's
believing that his wife is faithful even if the evidence indicates otherwise, the special-
case thesis; is false. And, whether or not they are true, such claims are surely coherent.
This argument is not quite conclusive, however, for a defender of the special-case
thesisj might reply that the coherence of these descriptions is insufficient to refute
 "The Ethics of Belief" Reconsidered 23

his thesis; in the cases described, he might argue, there is a prima facie moral failing
(believing unjustifiedly), but it is so slight as to be negligible, or has been overrid-
den by weightier considerations (the moral value of trust between husband and wife,
for example).
Another argument against the special-case thesisj appeals to the fact that "mor-
ally ought" implies "can," and hence that "epistemically ought" cannot be a subspe-
cies of "morally ought," since it does not imply "can"; for believing, and hence be-
lieving unjustifiedly, is not in any straightforward sense voluntary.
If this argument were conclusive, it would rule out not only the special-case thesisj,
but also the correlation thesisj and the overlap thesisj. But it is not conclusive.
Chisholm observes that, though indeed one cannot stop believing that p or start be-
lieving that/? now, no more can one fulfill all one's (as the special-case thesisj would
have it, all one's other) moral obligations now; what is required is only that one can
in due course.4 Of course, the sense in which one can't just stop believing or start
believing that/? now is quite unlike the sense in which one can't, say, answer all one's
correspondence now; the difficulty isn't that one hasn't time just now to stop or start
believing that/;, but that one can't simply stop or start believing that/? at any time.
Believing that p is a condition one finds oneself in, not something one does. How-
ever, as Chisholm pointed out in an earlier discussion of this issue,5 one can some-
times bring it about that in due course one believes . . . ; one can sometimes induce
a belief, by bringing about the circumstances in which that condition is likely to arise.
One cannot believe at will; nevertheless, sometimes the wish is father to the thought,6
and this may be enough for moral appraisal to be applicable.
A better argument against the special-case thesis, is this. A person is epistemically
unjustified in believing that p just in case his evidence isn't good enough. But he
can't be morally at fault in believing that/? unless his belief is willfully induced. And
his evidence may not be good enough even in cases where his belief is not willfully
induced. So it is possible that there should be cases where a person is epistemically
unjustified but not morally at fault; and the special-case thesisj is false.
Before I turn to the correlation thesis:, however, I need to consider a reinterpreta-
tion of the special-case thesis: recently suggested by Chisholm. "The distinguishing
feature of ethical duty," Chisholm writes in a paper published in 1991, "is not to be
found in the considerations that impose that duty. Rather, an ethical duty is simply a
requirement that is not overridden by any other requirement."7 And so, he argues,
when an epistemic requirement is not overridden by any other requirement, it is one's
ethical duty. Even if this account of what it is to be an ethical requirement were ac-
ceptable, this would be insufficient to establish the special-case thesisj; for it would
show only that some epistemic requirements—those which are not overridden by other
requirements—are ethical. And, in any case, it seems that "ethical" has been persua-
sively redefined, as "any normative requirement not overridden by some other re-
quirement"; that this is a re-definition becomes apparent when one considers that it
implies, for example, that any requirement, of prudence, say, or of aesthetics, would
thereby be classified as ethical provided only that it is not overridden by any other
requirement.8
According to the correlation thesisj, although to say that a person believes
unjustifiedly is not eo ipso to say that he is morally at fault, nevertheless, whenever
24 Susan Haack

a person believes unjustifiedly, he is morally as well as epistemologically at fault.

Two arguments against this thesis immediately suggest themselves. If it is ever true
(not merely, as the first argument considered against the special-case thesis required,
possible) that believing unjustifiedly is beneficial or harmless, or if it is ever false
(not necessarily, as the second argument considered against the special-case thesis
required, always) that a person is responsible for believing unjustifiedly, then the
correlation thesis is false.
There are cases in which a person's believing unjustifiedly is harmless or even
beneficial. My believing, on inadequate evidence, that the apples I just selected are
the best in the supermarket, is, like many inconsequential beliefs, harmless. Again,
if a patient's believing, on inadequate evidence, that he will recover from his illness
significantly improves the chances that he will recover, then he may properly be
appraised neutrally from a moral point of view.9
Cases like this are sufficient to show it false that, whenever a person believes
unjustifiedly, his so believing is always also subject, all things considered, to unfa-
vorable moral appraisal. They are not sufficient, however, to show it false that, when-
ever a person believes unjustifiedly, his so believing is always also subject to un-
favorable moral appraisal prima facie. But if a subject is not always responsible for
believing unjustifiedly, even a prima facie correlation thesisj is false.
Possible explanations of someone's believing unjustifiedly are: negligent incon-
tinence10—he has been careless or perfunctory in inquiry, but, jumping to conclu-
sions, has formed a belief anyway; self-deception—self-interest has skewed his per-
ception of the weight or relevance of this or that evidence; or cognitive inadequacy
—he has done his best, but on this matter his best cognitive effort isn't good enough,
and has resulted in an unjustified belief. (The first two kinds of explanation are not
really so distinct as this rather crude list makes them appear, since a sort of one-sided
carelessness in inquiry is one of the forms in which self-deception manifests itself;
but, though they may be, negligence in inquiry and incontinence in belief-formation
need not be self-interested.)
One may distinguish two kinds of cognitive inadequacy: the personal—an in-
dividual's good-faith misjudgment of the weight of complex evidence—and the cul-
tural. The latter arises because of the perspectival character of judgments of
relevance, their dependence on background beliefs. Sometimes the explanation of
someone's believing on skimpy evidence is that he doesn't realize that certain
relevant evidence is relevant, because the background beliefs which determine what
evidence he perceives as relevant are mistaken—background beliefs which are taken
for known facts in his epistemic community, and which he may have no way of know-
ing are not so.
Where there has been no negligence and no covert operation of wishes or fears,
where the explanation of the person's believing unjustifiedly is cognitive inadequacy,
personal or cultural, unfavorable moral appraisal is inappropriate even if the belief
is harmful. 11
Perhaps it will be argued in defense of a prima facie correlation thesis; that, even
in cases of unjustified believing explicable by cognitive inadequacy, the subject is
still morally culpable in an indirect way; culpable, that is, not directly for believing
unjustifiedly (ex hypothesi, that represents his best cognitive effort at the time), but
 "The Ethics of Belief" Reconsidered 25

indirectly, for not having cultivated better judgment. There are circumstances where
this is appropriate—for example, in some cases where it is this person's (this doctor's,
this lawyer's, this juror's, this academic's) particular responsibility to know about
the matter at hand; but the correlation thesisj requires that it always be so. And this
is not true. Even if one were morally required to cultivate one's capacity to judge
evidence to the very best of which one is capable (a very demanding assumption),
still, for any person, there would be some degree of finesse which he could not, by
even the most strenuous mental discipline, surpass.
Or perhaps it will be argued in defense of the correlation thesisT that, even in cases
of unjustified believing explicable by cognitive inadequacy, the subject is still mor-
ally culpable by omission: he morally ought to be aware of his cognitive limitations.
There are circumstances where this observation is appropriate too; but, again, the
correlation thesisj requires that it always be so. And this is not true either. Even if
one were morally required to be as aware as possible of one's cognitive limitations
(again, a very demanding assumption), a complete grasp of those limitations may be
beyond one's cognitive powers. If a person has done the best he can, not only to find
out whetherp, but also to determine that he is competent to find out whether;?, he is
not morally culpable even if his belief in his competence and his belief that p are, by
reason of cognitive inadequacy, unjustified.
If these arguments are correct, the correlation thesisj, even in its weaker, prima
facie, form, is false.
Unlike the correlation thesisj, which requires that unjustified believing be always
(at least prima facie) harmful and always something for which the subject may prop-
erly be held responsible, the overlap thesisj requires only that unjustified believing
sometimes cause (at least prima facie) harm and sometimes be something for which
the subject may properly be held responsible.
And this is so. Acting on false beliefs sometimes causes either actual harm or, at
least, unacceptable risk of harm. Justified beliefs may be false, and unjustified be-
liefs may be true; nevertheless if, as we hope and believe they are, our criteria of
justification are indicative of truth, justified beliefs are likely to be true, and unjusti-
fied beliefs are likely to be false. And so, acting on unjustified beliefs is also (though
less) likely to cause harm or, at least, unacceptable risk of harm.
And when unjustified believing is the result either of negligence or of self-
deception, though it is not belief at will, it is willful—it is, as we say, a kind of "will-
ful ignorance." One might reasonably feel that a person who knowingly causes harm
reveals himself to be a more hardened character than a person who induces himself
to believe, unjustifiedly, that his action will not be harmful. Nevertheless, the quasi-
voluntary nature of willful ignorance seems to suffice, at least sometimes, for the
ascription of responsibility.
In other words, believing unjustifiedly is sometimes a form of morally culpable
ignorance. It is not, of course, the only form. Ignorance comes in at least three vari-
eties: one may fail to know because one has no belief on the matter at hand (agnos-
ticism), or because the belief one has is false (mis-belief), or because the belief one
has is unjustified (over-belief). Agnosticism, in turn, comes in at least three sub-
varieties: one may have no belief because one hasn't investigated and has no evi-
dence either way (plain agnosticism); because, though one has investigated and has
26 Susan Haack

evidence, that evidence seems insufficient to settle the matter (can't-tell agnosticism);
or because one has failed to draw a conclusion which the evidence would support
(under-belief). There are epistemologically and psychologically interesting similari-
ties between the phenomena of under-belief and over-belief;12 but it is the latter that
concerns me here. Over-belief, unjustified believing, constitutes culpable ignorance
when, as it sometimes but not invariably is, it is both harmful and peccable.
If these arguments are correct, the overlap thesisj is true.13

The arguments thus far put the strategy of W. K. Clifford's celebrated essay, "The
Ethics of Belief,"14 in a new perspective. The main thesis of that essay is that "it is
wrong always, everywhere, and for anyone, to believe anything upon insufficient
evidence" (p. 77). Neither here nor elsewhere in the essay does Clifford ever distin-
guish "it is epistemologically wrong" from "it is morally wrong."15 But he offers no
arguments for identifying the two, or even for the special-case thesisj, that the former
is a subspecies of the latter. Instead, extrapolating from a striking case where unjus-
tified believing is culpable ignorance, he tries to persuade one that all cases of un-
justified believing are, in some measure, both harmful and willful. He offers, in other
words, only arguments that could, at most, establish the correlation thesisj. It is illu-
minating—as a further test of the claim that the correlation thesisj is not true, although
the overlap thesisj is—to show how Clifford's attempted extrapolation fails.
In the vivid case with which Clifford's essay opens, we are to imagine a shipowner
who "knowingly and willingly" suppresses his doubts, doesn't check, manages sin-
cerely to believe that his vessel is seaworthy, and allows the ship to depart. He "had no
right to believe on such evidence as was before him" Clifford observes; and he "is
verily guilty" of the deaths of passengers and crew when the ship goes down (p. 70).
The description of the ship-owner's self-deception as "knowingly and willingly" un-
dertaken (p. 71) is a bit lacking in subtlety; and it would have been desk-able that Clifford
say explicitly that it is the element of willfulness that justifies an unfavorable moral
appraisal in this case, as in more-straightforward cases where harm is knowingly
caused—for example, if the ship-owner knew full well that the vessel was unseawor-
thy and allowed it to depart anyway. Nevertheless, Clifford's judgment of this case
seems correct: it is a case of morally culpable ignorance, of failure in a duty to know.
But the case has a number of features which are not invariably found whenever
someone believes unjustifiedly, and some of which are essential to the unfavorable
moral appraisal appropriate here. The unjustified belief is false; the proposition con-
cerned is of great practical importance; the person concerned is in a position of spe-
cial responsibility; the false belief leads to dramatically harmful consequences; and
the belief is willfully self-induced. The correlation thesisj is false unless the igno-
rance would still be morally culpable even if all these features were absent.
Clifford is aware that a belief held on insufficient evidence may be true, and clear
that it is the belief's being unjustified, not its being false, that matters. There are two
points to consider here, only one of which Clifford raises. The first, which he does
not mention, concerns cases of false but justified belief. If the ship-owner had inves-
tigated carefully and honestly, and had been justified in believing the vessel seawor-
thy, but his justified belief had been false, and the ship went down, the appropriate
verdict from a moral point of view would surely be that he was not to blame for the
 "The Ethics of Belief" Reconsidered 27

false belief, nor, therefore, for what one would be inclined to describe as a tragic
accident. The second point, which Clifford does discuss, concerns cases of unjusti-
fied but true belief. He first remarks that the ship-owner would still be morally re-
sponsible even if his belief that the vessel was seaworthy was true, because "he had
no right to believe on such evidence as was before him." This trades on his failure to
distinguish epistemic from ethical justification, and hence fails. Later, however (p. 72),
Clifford comes up with a better argument: by failing to investigate properly and in-
ducing himself to believe on inadequate evidence, the ship-owner would have taken
an unacceptable risk of causing harm. That seems correct. So far, so good, for Clifford
and for the correlation thesisj.
But what if the proposition concerned were not, as in the original case, a conse-
quential one, or if the person with the unjustified belief were not the person respon-
sible for deciding whether the ship is to be allowed to make the voyage? Where, then,
is the harm, or, if the unjustified belief happens to be true, the risk of harm? Clifford
offers what are in effect two kinds of answes to these questions about apparently
harmless unjustified belief. The first is—urging that a belief must be connected to
action somehow, however indirectly, to count as a belief at all—that no belief is re-
ally altogether inconsequential; there is always at least the potential that action might
be based on it, and might prove harmful (p. 73). The second is to suggest that unjus-
tified believing always discourages scrupulous inquiry and strengthens the habit of
"credulity"; it weakens the epistemic fiber, one might say, and hence carries, if not
invariably a risk of harm, a risk of risk of harm (p. 76).
Clifford's responses depend on two false assumptions: that mere potential for harm,
however remote, is sufficient for unfavorable moral appraisal (provided the subject
is responsible for the unjustified belief); and that a subject is always responsible for
unjustified believing. But remote potential for harm is not sufficient; if it were, not
only drunken driving, but owning a car, would be morally culpable. And a subject is
not always responsible for believing unjustifiedly; the cause, sometimes, is cogni-
tive inadequacy.
Matters are confused by the way Clifford combines the two responses in the ar-
gument that unjustified believing encourages "credulity," and thereby carries poten-
tial for harm or risk of harm. It is true that sloppy inquiry, jumping to conclusions,
and wishful thinking manifest undesirable dispositions—dispositions to which, no
doubt, some people are temperamentally more inclined than others, but dispositions
which one can either check and discourage in oneself, or allow to operate unchecked,
and, by unchecked indulgence, encourage. They are bad habits which may, if un-
checked, become inveterate. (It is not clear, however, that Clifford is right to suggest
that any individual's indulgence in such habits is bound to encourage them in oth-
ers.) But it is not true that unjustified believing is always the result of self-deception
or negligence; so Clifford's oblique argument that unjustified believing is always
harmful also fails.
These disagreements with Clifford by no means imply agreement with his most
famous critic, William James.16 Clifford holds that it is always wrong to believe on
insufficient evidence. I have pointed out that Clifford fails to distinguish "epistemo-
logically wrong" from "morally wrong," and argued that his thesis is not true if in-
terpreted as an ethical claim. James holds that it is not always wrong to believe on
28 Susan Haack

insufficient evidence. He would be correct, therefore, if by "wrong" he meant "mor-

ally wrong" only; but this is not what he means.
Like Clifford, James never distinguishes these two possible ways of taking "ought,"
"justified," "our duty in the matter of opinion," and so on. Some of the arguments in
"The Will to Believe" seem to be intended as epistemological: that knowing the truth
is no less valuable than avoiding error (pp. 17ff); that believing that p sometimes
contributes to bringing it about that;? is true (pp. 23-24). But others seem to be of an
ethical character: that we should not condemn those who have faith for believing
without adequate evidence, but should "respect one another's mental freedom" (p. 30);
and the quotation from Fitz-James Stephen at the close of the essay, urging that we
have faith because "if death ends all, we cannot meet death better" (p. 31). This sug-
gests that the best way to read James is as holding that it is not always wrong either
epistemologically or morally to believe on insufficient evidence.17
James's argument about respect for others' mental freedom deserves special com-
ment. If, like James, one fails to distinguish epistemic from ethical justification, one
can make room for (moral) tolerance of others' unjustified opinions only as James
seems to, by weakening one's standards of epistemic justification. But if one distin-
guishes the two, one has no need of any such radical epistemological measures. In
any case, one's judgment that another's belief is unjustified must, because of the
perspectival character of judgments of justification, their dependence on one's back-
ground beliefs, be acknowledged to be thoroughly fallible. And, most to the present
point, unjustified believing is not morally culpable if it results from cognitive inad-
equacy, whether personal or cultural.
Unlike both James and Clifford, I distinguish epistemological from ethical justi-
fication. Like James and unlike Clifford, I do not think it always morally wrong to
believe on inadequate evidence. Clifford's position is over-demanding morally. Like
Clifford and unlike James, however, I think it is always epistemologically wrong to
believe on inadequate evidence—in the sense that believing on inadequate evidence
is always epistemologically unjustified belief. James's position is over-permissive
epistemologically.
Perhaps it will be objected that sometimes it is all to the good—epistemologically
all to the good—that a person believe something even though his evidence is inad-
equate; for example, the scientist whose faith in an as yet inadequately supported
theory motivates him to develop, articulate, and test it, and thus advances inquiry.
This objection is focused, not on the concept of epistemic justification, but on
questions about the conduct of inquiry.18 It is irrelevant to the claim that believing
on inadequate evidence is always believing unjustifiedly; it argues, rather, that be-
lieving unjustifiedly is not always damaging, and may even be helpful, to the progress
of inquiry. This, I think, is true. Not that over-belief is ever an optimal condition for
the conduct of inquiry; the ideal, I take it, would be, not for our hypothetical scien-
tist to have faith in the theory's truth, but for him to recognize it as, though thus far
unworthy of belief, nevertheless promising enough to be worthy of serious further
investigation.19 Still, given human inquirers' inevitable frailties, a scientific commu-
nity in which some are disposed to over-belief and others to under-belief may, by
virtue of individuals' epistemic imperfections serendipitously compensating for each
other, be a reasonable ersatz of a community of inquirers who conform to the episte-
 "The Ethics of Belief" Reconsidered 29

mological ideal. So, although over-belief is always epistemologically wrong in the

sense both of "epistemologically unjustified" and in the sense of "not the ideal with
respect to the conduct of inquiry," it is not always epistemologically wrong in the
sense of "damaging to the conduct of inquiry."
So I do not mean to deny that, as James observes, "science would be much less
advanced than she is if the passionate desires of individuals to get their own faiths
confirmed had been kept out of the game" (p. 21). The point is, rather, that because
James fails to distinguish the question whether believing on inadequate evidence is
always unjustified belief, from the question whether believing on inadequate evi-
dence is always damaging to the conduct of inquiry, he runs together a correct nega-
tive answer to the latter with an incorrect—over-permissive—negative answer to the
former.20
Clifford and James simply fail to distinguish epistemic from ethical justification;
Chisholm explicitly maintains that epistemic justification is a subspecies of ethical
justification. James claims that it is sometimes legitimate to believe on insufficient
evidence, suggesting that a man who has a moral duty to believe that p may thereby
be epistemologically justified in so believing; Chisholm explicitly denies this, but
he also protests that Clifford's "rigid evidentialism" is epistemologically over-
demanding, suggesting, instead, that a belief is epistemologically "innocent until
proven guilty."21
So Chisholm's position is further from Clifford's and closer to James's than his
borrowing Clifford's title for part I of Perceiving might have led one to expect. And
so my disagreement with Chisholm, like my disagreement with James, extends be-
yond the matter of the distinctness of epistemological from ethical justification to a
more strictly epistemological issue. For, where the question of epistemic justifica-
tion is concerned, my position is closer to Clifford's than to James's. Closer, but not
identical; for I think it vital to acknowledge the gradational character of epistemic
justification:22 whether, or to what degree, a person is justified in a belief depends on
how good—how supportive, how comprehensive, and how independently secure—
his evidence with respect to that belief is.23 Ideally, I should prefer to put this in terms
which also acknowledge that belief, as well as justification, comes in degrees. But
the point on which I am presently taking issue with Chisholm doesn't depend on these
subtleties; it is that, by my lights, one believes that p unjustifiedly, even if one's
evidence supports p over not-/?, unless one's evidence includes enough of the rele-
vant evidence.24
The goal of inquiry is substantial truth. When one focuses on guidelines for the
conduct of inquiry, one must concern oneself with substance as well as truth. But
when one focuses on criteria of justification, one is ipso facto restricting oneself to
the dimension of truth; for truth-indicativeness is the characteristic virtue of criteria
of justification. Chisholm, noting, correctly, as James does that "playing it safe" is
not always the most successful course in inquiry, then suggests, incorrectly, as James
does, that this motivates less demanding criteria of justification.25

Complex as this has been, it has been, thus far, focused quite narrowly, on the ques-
tion of the relation of epistemic to ethical justification only—on which I find myself
in disagreement with Chisholm. I want, by way of conclusion, to offer some more
30 Susan Haack

positive thoughts with regard to a different dimension of epistemic appraisal—the

appraisal of a person qua inquirer or cognizer.26
Our vocabulary for epistemic appraisals of character is varied and subtle ("me-
ticulous," "sloppy," "imaginative," "closed-minded," "brilliant," "obtuse," etc.). It
is striking that a significant subclass of this vocabulary is shared with ethics: "hon-
est," "responsible," "negligent," etc., come immediately to mind. And I am not sure
but that here the relation of epistemic to ethical appraisal may be as intimate as the
special-case thesis maintains; perhaps, at least without an "otherwise," "he is a good
man but intellectually dishonest" really does have the authentic ring of oxymoron.
Recall that, if my earlier arguments are correct, it is precisely when a person's
unjustified believing stems, not from cognitive inadequacy, but from self-decep-
tion or negligent incontinence—from a lack of intellectual integrity27 on his part—
that we hold him responsible for his belief. This suggests a friendly reinterpreta-
tion of what is most plausible in Clifford's condemnation of "the habit of credulity,"
and Chisholm's defense of the special-case thesis, as pointing to the moral impor-
tance of intellectual integrity.
Which prompts the following concluding observation: at the price of a little over-
simplification, one might say that, as courage is the soldier's virtue par excellence,
so intellectual integrity is the academic's. (The over-simplification is that intellec-
tual integrity itself requires a kind of courage, the hardihood called for in relinquish-
ing dearly held beliefs, or in resisting some conventional wisdom or fashionable shib-
boleth.) As C. I. Lewis writes, more eloquently than I could: "Almost we may say
that one who presents argument is worthy of confidence only if he be first a moral
man, a man of integrity.... [W]e presume, on the part of those who follow any sci-
entific vocation, . . . a sort of tacit oath never to subordinate the motive of objective
truth-seeking to any subjective preference or inclination or any expediency or op-
portunistic consideration."28

Notes
This paper is reprinted, with minor modifications, from Lewis Hahn, ed., The Philosophy of
Roderick ML. Chisholm (Library of Living Philosophers, vol. 25, La Salle, 111.: Open Court,
1997), 129^14, by permission of the publishers. I would to thank Richard Brandt, Peter Hare,
Mark Migotti, Sidney Ratner, Harvey Siegel, David Stove, and Joanne Waugh for helpful
comments on a draft of this paper; Howard Burdick and Risto Hilpinen for helpful conversa-
tions; and the audiences at the universities (Washington State, Utah, Saint Cloud State, Michi-
gan, Toronto, Gothenburg, Helsinki, Granada, the University of the Basque Country in San
Sebastian) and conferences ("A Hundred Years of American Philosophy" in Cerisy La Salle,
and the Mid-South Philosophical Association at the University of Memphis) where this paper
was presented.

1. In principle, there are two other possibilities: that epistemic appraisal and ethical ap-
praisal are identical, and that ethical appraisal is a special case of epistemic appraisal. I shall
not consider either here. The first seems too obviously false to consider; the latter, with its
Platonic overtones, would require a paper of its own.
2. R. M. Chisholm, "Firth and the Ethics of Belief," Philosophy and Phenomenological
Research, 51.1 (1991): 119-28 (the quotation is from p. 119). See also R. M. Chisholm,
"Epistemic Statements and the Ethics of Belief," Philosophy and Phenomenological Research
16 (1956): 447-60; R. M. Chisholm, Perceiving: A Philosophical Study (Ithaca, N.Y.: Cornell
 "The Ethics of Belief" Reconsidered 31

University Press), 1957; R. Firth, "Chisholm and the Ethics of Belief," Philosophical Review
68 (1959): 493-506; R. M. Chisholm, "'Appear,' Take,' and 'Evident'," Journal of Phi-
losophy 53.23 (1956): 722-31; R. Firth, "Ultimate Evidence," Journal of Philosophy 53.23
(1956): 732-39; R. M. Chisholm, "Evidence as Justification," Journal of Philosophy 58 (1961):
739^8; R. M. Chisholm, Theory of Knowledge (Englewood Cliffs, N.J.: Prentice Hall, 1966;
2d ed., 1977; 3d ed., 1989); R. M. Chisholm, "Lewis' Ethics of Belief," in The Philosophy of
C. 1. Lewis, P. A. Schilpp, ed. (La Salle, 111.: Open Court, 1968), 223-42; R. Firth, "Are
Epistemic Concepts Reducible to Ethical Concepts?" in Values and Morals, eds. A. I. Goldman
and J. Kim, (Dordrecht, the Netherlands: Reidel, 1978), 215-30; R. M. Chisholm, "Self-
Profile," in Roderick M. Chisholm, ed. R. J. Bogdan, (Dordrecht, The Netherlands: Reidel,
1986), 3-77.
From time to time Chisholm writes of "analogies" between ethics and epistemology: see,
e.g., Perceiving, pp. 12, 13, 18, 30; "'Appear,' 'Take,' and 'Evident'," pp. 723ff; Theory of
Knowledge, first, 1966, edition (1) and third, 1989, edition (57-8); "Epistemic Reasoning
and the Logic of Epistemic Concepts," in Logic and Philosophy, ed. G. H. Von Wright, (The
Hague, the Netherlands: Nijhoff, 1980), 71-78. If, as it seems, his point is that there are struc-
tural analogies between the overriding of one moral requirement by another, and the induc-
tive overriding of certain body of evidence by further evidence, this is quite compatible with
his commitment to the special-case thesisj.
On p. 54 of his "Self-Profile" Chisholm writes that "epistemic concepts are not moral
concepts"; by the final sentence of the section, however (56), he writes that the concepts of
epistemology are reducible to the concepts of ethics.
3. If, as I believe, the answer to the next-to-last of these questions is clearly "yes," this is
sufficient to show the special-case thesis false in its most general form.
4. Chisholm, "Firth and the Ethics of Belief," 125-27.
5. Chisholm, "Lewis' Ethics of Belief," 223-24.
6. A phrase of which F. C. S. Schiller reminds us in his commentary on James's "The
Will to Believe," Problems of Belief (London: Hodder and Stoughton, n.d.), 111. See, be-
sides Chisholm's discussion of the quasi-voluntary nature of belief referred to above, H. H.
Price's, in "Belief and the Will," Proceedings of the Aristotelian Society, suppl., 28 (1954):
1-27.
7. Chisholm, "Firth and the Ethics of Belief," third, 1989, edition 127; cf. Theory of
Knowledge, 58-59.
8. In his "Self-Profile," Chisholm suggests two arguments for the general form of the
special-case thesis. The concept of requirement, Chisholm says, is central to ethics, and the
concept of epistemic preferability can be defined in terms of requirement; to reach the con-
clusion that the concept of epistemic preferability is reducible to ethical concepts, however,
one needs the stronger premise that the concept of requirement is uniquely ethical. Knowl-
edge, Chisholm says, is, as Aristotle thought, intrinsically valuable; to reach the conclusion
that epistemic concepts are reducible to ethical concepts, however, one needs the stronger
premise that knowledge is intrinsically morally valuable.
9. Perhaps, if surviving his illness enables him to continue his morally admirable work
or to meet his obligations to others, a favorable moral appraisal is in order; but that issue
need not be decided here.
In the case described, the person's believing thatp makes it more likely thatp will turn
out true, but the point does not depend on that. Think of the kind of the case Peirce envisages
when he observes that he could not condemn a man who, having lost his wife, induces him-
self to believe in an afterlife in which they will be reunited, even though the belief is unjus-
tified, if, without it, "his usefulness would be at an end"; see C. S. Peirce, Collected Papers,
eds. C. Hartshorne, P. Weiss, and A. Burks (Cambridge, Mass.: Harvard University Press,
1931-58), 5.583, 1898.
10. A phrase adapted from J. Heil, "Doxastic Incontinence," Mind 93 (1984) 56-70.
11. This comports with the attractive conjecture (proposed by J. Shelton, "Contextualism:
A Right Answer to the Wrong Question," Southwest Philosophical Studies 9.2 (1983): 117-
124), that the appeal of contextualist theories of epistemic justification may arise in part from
32 Susan Haack

a confusion of epistemological with ethical justification. The same conjecture might also serve
to explain Goldman's claim that there are two concepts of epistemic justification, one objec-
tive and reliabilist, the other context-relative; see A. I. Goldman, "Strong and Weak Justifi-
cation," in Philosophical Perspectives, 2: Epistemology, ed. J. Tomberlin, (Atascadero, Ca-
lif.: Ridgeview, 1988), 51-70.
12. See Chisholm, Perceiving, 14.
13. My arguments against the special-case thesisj and the correlation thesisj presuppose
that harmfulness and responsibility are necessary for unfavorable moral appraisal; my argu-
ments for the overlap thesisj presuppose that they are sufficient. These assumptions, though
fairly weak, are, of course, not vacuous. For example, as the argument against the correlation
thesisj revealed, someone who maintained that one has a moral obligation to develop one's
capacities, generally, or one's capacity to judge evidence, specifically, would reject the former
presupposition.
14. W. K. Clifford, "The Ethics of Belief (1877), in The Ethics of Belief and Other Es-
says (London: Watts and Co., 1947), 70-96.
15. Richard Gale, "William James and the Ethics of Belief," American Philosophical
Quarterly 17.1 (1980): 1-14, claims (p. 1) that Clifford has to be read as proposing the ethi-
cal thesis, that it is always morally wrong to believe on insufficient evidence; he observes in
a footnote, however, that Clifford's words also bear another interpretation, that it is always
epistemologically wrong to believe on insufficient evidence.
16. W. James, "The Will to Believe" (1896), in The Will to Believe and Other Essays in
Popular Philosophy (1897), (New York: Dover, 1956): 1-31.
17. Cf. Jack W. Meiland, "What Ought We to Believe? or, The Ethics of Belief Revis-
ited," American Philosophical Quarterly 17.1 (1980): 15-24, which precisely, but more ex-
plicitly, follows James in this regard.
18. The distinction is articulated in more detail in my Evidence and Inquiry: Towards
Reconstruction in Epistemology (Oxford: Blackwell, 1993), chap. 10.
19. The Will to Believe is dedicated "To My Old Friend, CHARLES SANDERS PEIRCE, to whose
philosophic comradeship in old times I owe more incitement and help than I can express or
repay." In a letter of thanks, Peirce writes to James that in practical affairs, "'Faith,' in the
sense that one will adhere consistently to a given line of conduct, is highly necessary.. .. But
if it means that you are not going to be alert for indications that the moment has come to
change your tactics, I think it ruinous in practice" (Collected Papers, 8.251, 1897). The next
year one finds Peirce writing of the "Will to Learn" (5.583), and commenting that, where
science is concerned, "full belief is willingness to act upon . .. the proposition . . . [The] ac-
cepted propositions [of science] are but opinions at most; and the whole list is provisional"
(1.635).
20. The argument here raises an awkward question about the intended scope of James's
Will to Believe doctrine. His initial statement, that "our passional nature lawfully may de-
cide" any genuine option "that cannot by its nature be decided on intellectual grounds," strongly
suggests that the doctrine is to apply only to hypotheses, for example, of a religious nature,
which are in principle undecidable by evidence (which, however, raises the further awkward
question, whether such hypotheses would qualify as meaningful by the standards of the Prag-
matic Maxim). James's later reference to the role of "faith" in scientific inquiry, however,
suggests that the scope of the doctrine is intended to be much broader, applying also to hy-
potheses with respect to which we merely happen, thus far, to lack sufficient evidence.
21. Chisholm, Perceiving, 9, 11, 100; Theory of Knowledge (Englewood Cliffs, N.J.:
Prentice Hall, 1966), 18-19. (The reference to Clifford is, however, missing from the second
and third editions of Theory of Knowledge.) Chisholm's disagreement with Clifford on this
matter seems to have escaped the attention of some commentators; see, for example, L. Pojman,
"The Ethics of Belief," Southwest Philosophical Studies 9.2 (1983): 85-92, who describes
Chisholm as subscribing to "rigid evidentialism," according to which "one ought to believe
propositions if and only if they are backed by sufficient evidence." Pojman attributes this
account of Chisholm's position to Meiland, "What Ought We to Believe?"; but the attribu-
tion is incorrect, since Meiland is careful to distinguish a stronger evidentialism (one has a
 "The Ethics of Belief" Reconsidered 33

right to believe that p only if the evidence is sufficient) from a weaker (one has a right to
believe that p provided one does not have sufficient evidence for not-p), and does not say
which, if either, he takes Chisholm to hold.
22. Chisholm too seems to acknowledge the gradational character of epistemic justifica-
tion, most clearly in the third edition of Theory of Knowledge. But the fact that epistemic
justification comes in degrees, whereas (I take it) ethical justification does not, suggests a
further argument against the special-case thesisj.
23. It is because I take comprehensiveness to be only one of three determinants of degree
of justification that I shifted, above, from Clifford's favored expression, "insufficient evi-
dence," to writing of "inadequate evidence," which is, I hope, less likely to suggest failure of
comprehensiveness alone.
24. My account of the determinants of degree of epistemic justification, one of which is
how much of the relevant evidence the subject's evidence includes, is spelled out in detail in
Evidence and Inquiry, chap. 4.
My comprehensiveness requirement is motivated in part by an analogy between the struc-
ture of empirical justification and a crossword puzzle; as the reasonableness of one's confi-
dence that a crossword entry is correct depends in part on how many of the intersecting en-
tries one has completed, so one's degree of justification in a belief depends in part on how
much of the relevant evidence one's evidence includes. So my neglect of the analogy thesis;
does not stem from any prejudice against analogies, nor, I should add, from the belief that
there are no interesting analogies between meta-epistemology and meta-ethics. For explora-
tions of such analogies, see (besides the papers of Firth's referred to above) R. B. Brandt,
"Epistemology and Ethics, Parallels Between," in Paul Edwards, ed., the Encyclopedia of
Philosophy (New York: Macmillan, 1967) "The Concept of Rational Belief," The Monist 68.1
(1985), 3-23; and W. P. Alston, "Meta-Ethics and Meta-Epistemology," in Goldman and Kim,
eds., Values and Morals, 275-98.
25. Chisholm, Perceiving, 22; Theory of Knowledge, 3d ed., 13-14.
26. There remain, of course, many other important questions which I shall have to put
aside: for example, whether Chisholm and Firth are correct in supposing that justification is
as central a concept in ethics as, I agree, it is in epistemology.
27. An expression that comports with the plausible idea that thinking is well construed as
inner dialogue and self-deception as involving distracting one's own attention from inconve-
nient evidence, as the deception of another involves distracting his attention. Cf. Peirce, Col-
lected Papers, 5.421, 1905.
28. C. I. Lewis, The Ground and Nature of the Right (New York: Columbia University
Press, 1955): 34. Of course, Lewis is using "scientific" in a broad sense, equivalent to "intel-
lectual." The reference to a "tacit oath," by the way, suggests that the special-case thesisj may
seem more plausible than it really is to those who are bound by such an oath and thus have a
special moral duty to objective truth-seeking.
 2

Epistemic and Moral Duty

BRUCE RUSSELL

In this discussion, I will explore the similarities between epistemic and moral duty,
especially regarding the distinctions between subjective and objective duties and
between acting in accordance with and acting from duty. I will also discuss what seems
to be the most noteworthy difference between epistemic and moral duty, namely,
that moral duties apply to actions that are within our control while if there are any
epistemic duties they apply to beliefs that are not within our control.

I. Moral Duty

There are two senses in which a person can have a moral duty. In one sense, a person
has a moral duty if she is blameworthy if she does not fulfill it, but there is another
more fundamental sense in which a person can have a duty. To illustrate the differ-
ence, consider a physician who is about to treat a comatose individual. The physi-
cian does all that he can to determine whether the patient is allergic to the drug he is
about to administer. He finds no evidence that he is and, because it's an emergency,
cannot wait until the patient regains consciousness. He administers the drug, and the
patient has an allergic reaction and dies. In one sense, the physician did not do what
he should since he should not have given the patient the drug. In another sense, he
did, since he did what it was reasonable of him to think was best for his patient, and
he was blameless even though what he did killed the patient. We might say that the
physician did not fulfill his objective, but did fulfill his subjective, duty. Duty can be
defined in terms of wrongness: a person has a duty to do something if it is not wrong
of him to do it and wrong of him to fail to do it. So I will take wrongness to be the

34
 Epistemic and Moral Duty 35

primitive concept and discuss what objective and subjective wrongness are, keeping
in mind that this will then allow us to define objective and subjective duty.
It is tempting to define objective wrongness in terms of blame, namely, as what a
person would be blameworthy for doing if he had no legitimate excuse. But examples
can be given where a person is blameworthy and has no legitimate excuse but does
what is objectively right. Suppose, for instance, that I have every reason to believe
that doing something is objectively wrong, say, giving an innocent person something
I think is poison, even though, as it turns out, it is a substance that will cure the per-
son of her fatal disease. Then, whether intentionally or out of laziness or inattention,
if I give the person what is in the bottle marked "poison" but is really the medicine
that will cure her, I will do what is objectively right, even though I will be blame-
worthy and have no legitimate excuse for doing it.
So objective wrongness will consist in doing something that is morally forbid-
den. A person will be blameworthy just in case she has adequate reason to believe
that what she did was objectively wrong and has no legitimate excuse for doing it. It
does not matter whether it was in fact objectively wrong, as the above example shows,
since the person is blameworthy even though she does what is objectively right. She
will be blameless just in case she has no adequate reason to believe that what she did
was objectively wrong or has a legitimate excuse for doing it. Again, it does not matter
whether it was in fact objectively wrong, as the example of the physician shows, since
he is blameless even though he does what is objectively wrong.
The following matrix indicates the four possible cases and specifies the conditions
that determine why a given case will fall in some particular cell of the matrix:

blameless blameworthy
objectively right no adequate reason to think adequate reason to think
it's objectively wrong or a it's objectively wrong &
legitimate excuse no legitimate excuse
objectively wrong no adequate reason to think adequate reason to think
it's objectively wrong or a it's objectively wrong &
legitimate excuse no legitimate excuse

Now, a full account of what objective rightness/wrongness and blamelessness/blame-

worthiness are would require an account of what a legitimate excuse is and what
objective Tightness or wrongness is. However, all I want to do here is argue that ig-
norance of what the correct principles are that determine objective Tightness and
wrongness can serve as a legitimate excuse. Imagine people who grow up thinking
that certain fairly mild forms of slavery, the subjugation of women or the cruel treat-
ment of animals are morally permissible even though they are objectively morally
wrong. I think that in certain circumstances such people would be blameless in sup-
porting slavery or practices that subjugate women or in treating animals cruelly,
despite the fact that what they do is objectively wrong.
The second distinction I want to draw is between acting in accordance with, and
acting from, duty. Perhaps the best illustration of this difference comes from Kant's
36 Bruce Russell

example of the two grocers. The one grocer does not cheat his customers because he
follows the prudential principle that honesty is the best policy. However, if he thought
he could get away with cheating them he would. The other grocer does not cheat his
customers because he thinks it is the wrong thing to do. Both grocers act in accor-
dance with duty, but only the second acts from duty; that is, only he is motivated to
do what he does because he believes that what he does is morally required.

II. Epistemic Duty

There is a distinction in epistemology that corresponds to the difference between

objective wrongness and blameworthiness. Someone who grows up in a religious
society and is taught to listen to the deliverances of an oracle can be epistemically
blameless in believing those deliverances even though her belief may not really be
supported by the evidence and so is objectively iwijustified. Someone who grows up
in a society where everyone forms hasty generalizations or engages in wishful thinking
would be in the same boat. From the standpoint of blame, these people are like the
physician who gives a drug that kills, having every reason to think it would save.
Contrast this with a person who is epistemically blameworthy for believing some-
thing on the basis of wishful thinking (say, that he will win some prestigious award)
even though relative to the evidence he has he is objectively justified in believing
what he does. This person is like the person who gives someone else the contents of
a bottle marked "poison" even though it luckily turns out that the bottle contains a
lifesaving drug.
Of course, there are the standard cases where one is both objectively justified and
blameless in believing what one does, say, your believing that there is a piece of paper
in front of you, and cases where one is both objectively unjustified in believing some-
thing and blameworthy for believing it, for example, your now believing that there
is an elephant in front of you. The following matrix illustrates the four possible cases
involving belief and where the cases fit in that matrix.

blameless blameworthy
objectively justified piece of paper the award winner
not objectively justified society of hasty generalization elephant
or wishful thinking

A difference between epistemic and moral duty is that one must either believe a
proposition, disbelieve it, or suspend judgment, while sometimes one can permissi-
bly either perform or not perform some action.1 There is no epistemic attitude that it
is merely permissible to have, while it can be merely permissible to perform some
action. If my evidence for and against extraterrestrial intelligence has equal weight,
I must suspend judgment; I cannot suspend judgment, believe, or disbelieve. How-
ever, assuming that the moral reasons for and against my now getting up and getting
a drink of water are equal, it is permissible for me to do either. It is not the case that
I must do one or the other.
 Epistemic and Moral Duty 37

The cases I have given illustrate the two pairs of epistemic differences: objectively
justified/objectively unjustified and epistemic ally blameless/epistemically blamewor-
thy. I now want to argue that epistemic distinctions that others have made do not
capture the distinctions that the examples illustrate.
In "Strong and Weak Justification," Alvin Goldman initially thinks of weak jus-
tification as "faultless, blameless, or non-culpable belief" and illustrates the sort of
justification he has in mind with an example of a scientifically benighted culture whose
members use methods for forming beliefs that "appeal to the doctrine of signatures,
to astrology, and to oracles."2 However, because he wants strong justification to be
opposed to weak justification, so that no one could be both strongly and weakly jus-
tified in believing something, he requires that weak justification be "ill-formed-but-
blameless belief" (p. 56), where by "ill-formed" he means "not reliably produced."
Still, according to Goldman, a person can be blameless in holding such an ill-formed
belief if she does not think it was not reliably produced and she does not believe there
are, and there in fact are no, reliable processes or methods available to her through
which she could discover the belief' s unreliable origins. But ultimately Goldman does
not mean by "weak justification" merely epistemically blameless belief, since he also
requires that such beliefs be ill-formed. So, for instance, your belief that you are now
looking at a piece of paper would not be weakly justified on Goldman's account,
even though you would be epistemically blameless in holding it, if it were the result
of a reliable belief-producing process or mechanism.
His account of strong justification is even further from what I have in mind with
the notion of objective justification than his account of weak justification is from
epistemic blamelessness, since weak justification is at least a species of blameless
belief. Roughly, for Goldman, a belief is strongly justified if it is the result of a reli-
able cognitive process or method, and its reliability is not undermined by the subject's
cognitive state (e.g., his believing that the process or mechanism that produced the
belief is not reliable).3 On the basis of this account, Goldman concludes that a cog-
nizer in a Cartesian demon world is not strongly (though he is weakly) justified in
believing what he does, since his beliefs are not the result of a reliable cognitive pro-
cess or method (though he is blameless in holding them).4
But Goldman's account of strong and weak justification does not allow him to
contrast the cognizer in the demon world with the subjects in the religious society.
On Goldman's account, both are weakly, but not strongly, justified in believing what
they do, since (1) their beliefs are the result of unreliable cognitive processes or
methods that they do not believe are unreliable; (2) they do not believe that there is
a reliable method that they possess, or that is available to them, that would enable
them to discover that unreliability; and (3) there is not. But surely from the stand-
point of justification, the person in the demon world is more like his counterpart in
the real world than like members of the religious society. In both those worlds, there
really is evidence for the existence of the external world, while in the religious society,
the deliverances of the oracle are not founded on good evidence. Because of this
difference, people in the demon and the nondemon worlds are equally objectively
justified in believing what they do about the world, while the people in the religious
society are not objectively justified in believing the deliverances of the oracle. In
other words, Goldman is forced to find a contrast in justification between demon and
38 Bruce Russell

nondemon world subjects where there is none (they are both objectively justified)
and cannot account for the contrast between people in the demon world and those in
the religious society (the former are objectively justified; the latter, only subjectively
justified). Of course, people in the demon and nondemon worlds and those in the
religious society are all epistemically blameless in believing what they do.
The grounds for saying that people in the demon and the nondemon worlds are,
from the standpoint of justification, in exactly the same situation is not, as Goldman
thinks, that if we did not say this we would be allowing luck to play a role in justifi-
cation that it should not. (One might think that the people in the nondemon world are
lucky to have true beliefs, since if they were in the demon world, they would have
false beliefs based on the same evidence they have.)5 Rather, it is the intuition that
each has equally good evidence for what he believes—from which it follows, of
course, that the goodness of evidence is independent of its tendency to produce true
beliefs. The contrast with the example of the religious society shows that questions
of epistemic blamelessness are separate from questions of the goodness of evidence,
and both are distinct from questions of reliability. Again, Goldman's distinction
between weak and strong justification forces us to draw a distinction regarding jus-
tification between people in the nondemon and demon worlds that is not there, and
it fails to account for the distinction between people in the demon world and those in
the religious society.
I now want to consider Richard Feldman's distinction between various forms of
subjective justification and objective justification to see if they can better account
for the intuitive similarities and differences in these three cases. In the course of dis-
cussing John Pollock's way of distinguishing subjective and objective justification,
Feldman draws a distinction between what he calls radical subjective epistemic jus-
tification, which only requires that a subject believe he has some reason to believe
some proposition, and what he calls moderate subjective epistemic justification, which
requires that a subject have good reason to think she has good reason to believe some
proposition.6 He goes on to argue that a person is not justified in any sense merely
because she is radically subjectively justified in believing something. Surely the mere
fact that I believe I have good reason to believe I will win the lottery does not in any
sense justify me in believing I will. In any case, being merely radically subjectively
justified does not imply that I am epistemically blameless in believing what I do, as
the lottery example also shows, and so does not imply what I have called subjective
epistemic justification.
Feldman argues that moderate subjective justification implies objective justifica-
tion, which means that he thinks that whenever a person has good reason to think he
has good reason to believe p, he does have good reason to believe p. Whether it is
reasonable to think that this is necessarily true will depend, I think, on what one says
about arguments that involve subtle fallacies of equivocation. Suppose a person is
justified in believing the premises of such an argument, and the appearance of va-
lidity gives him good reason to believe that the argument is valid. Does that appear-
ance of validity give him good reason to believe the conclusion or must the premises
really support the conclusion for that to be true? If the appearance is enough, then
the example does not count against Feldman's view, which implies that the person
has objectively good reason to believe the conclusion. But if the appearance of va-
 Epistemic and Moral Duty 39

lidity does not give the person good reason to believe the conclusion, this is a counter-
example to Feldman's view that moderate subjective epistemic justification implies
objective epistemic justification, since the person would have good reason to think
he has good reason to believe the conclusion even though he does not (since the pre-
mises do not really support it).
In any case, moderate subjective justification is not equivalent to what I have been
calling subjective epistemic justification and that I have identified with epistemic
blamelessness. On the basis of his professor's arguments, a student might have good
reason to believe that he has good reason to believe, say, that God does not exist and
yet still in fact believe that he does not have such reason. Suppose as the result of the
emotional pressure of his atheist friends, the student ends up disbelieving in God.
Then he will be epistemically blameworthy for believing contrary to his epistemic
conscience even though he really is moderately subjectively justified in his belief.
Conversely, a person might be epistemically blameless in her belief but be neither
moderately nor objectively justified in believing what she does. Suppose everyone
in her society constantly makes hasty generalizations. She then makes a hasty gener-
alization herself about a particular matter, say, that there is a higher rate of mental
retardation among a certain group of people than among other groups. In that case,
she would neither have good reason to believe nor good reason to think she has good
reason to believe what she does. In other words, she would be epistemically blame-
less but neither objectively nor moderately subjectively justified in believing what
she does.
So what Feldman considers under the category of subjective epistemic justifica-
tion, namely, radical and moderate subjective justification, are not equivalent to
epistemic blamelessness. Yet an epistemically blameless belief need not be one a
person is either objectively or moderately subjectively justified in holding—as the
examples of the religious society and the one involving hasty generalization show.
So Feldman's distinctions do not allow him to account for the contrast between the
cases that I am interested in here.
I propose that we adopt Feldman's account of an objectively justified belief, namely,
a belief that a person has good reason to believe, and let a subjectively justified belief
be one that the person is epistemically blameless in holding. She will be blameless in
holding it if the belief is objectively justified for her and she has no adequate reason to
think otherwise, or if the belief is not objectively justified for her but she has an excuse
for holding it, as with the oracle-based beliefs in the religious society or the belief of
the woman in the society where hasty generalizations are rampant.
This account of objective and subjective justification can make sense of the dif-
ferences and similarities between the examples in a way that neither Goldman's nor
Feldman's accounts can when taken alone. Like Feldman's account and unlike
Goldman's, it does not distinguish between people in a demon and a nondemon world
but implies that they are both objectively justified. Like Goldman's initial account
of weak justification (as blameless, nonculpable, faultless belief), it implies that people
in the demon and nondemon worlds and people in the religious society are blame-
less in believing what they do. Unlike Goldman's account, it allows us to draw a
distinction between the justification of people in a demon world and those in the
religious society. Feldman's account will imply that people in the religious society
40 Bruce Russell

are objectively justified in believing the pronouncements of the oracle if they are
objectively justified in trusting their elders and teachers, and the same results will
follow for those in the society where hasty generalization and wishful thinking are
the rule. Since he does not discuss the notion of epistemically blameless belief, we
cannot tell whether he would allow that there could be a society where people are
not objectively justified in trusting their elders and teachers yet they are epistemically
blameless in accepting what they are not objectively justified in believing.
While the distinction between objective and subjective epistemic justification
parallels a similar distinction in ethics, the similarity is not exact. As Feldman
argues, "objective justification in ethics is taken to be independent of beliefs or cog-
nitive states of the agent. It in no way depends on the agent's perspective."7 But an
account of objective epistemic justification that says that a person is objectively jus-
tified if and only if what she believes is true "amounts to identifying a kind of epistemic
justification with truth and that is surely a counter-intuitive identification."8 Feldman
goes on to say that both objective ethical and epistemic justification are evaluative
but "objective ethical justification turns on what really has value, while its epistemic
counterpart turns on what really are good reasons to believe."9 Good reasons are
relative to a person's cognitive states so, as Feldman says, objective epistemic justi-
fication is perspectival in a way that objective ethical justification is not.
What about the other distinction in ethics between acting from and acting in ac-
cordance with duty? There seems to be a similar distinction in epistemology. Con-
sider Keith Lehrer's example of Mr. Raco, who originally, and out of prejudice,
believes that only members of a particular race contract a certain disease but who
then learns in medical school that this is true.10 Suppose that at some point Mr. Raco
has the medical evidence but does not hold his belief on the basis o/that evidence.
Then, since he is aware of the evidence, he is objectively and subjectively justified
in believing what he does even though his belief is not based on the evidence. Here
we might say that Mr. Raco acts in accordance with his epistemic duty but not from
it and so is like Kant's grocer who gives the correct change because doing so is good
for business.
However, earlier I said that someone who has sufficient evidence to believe he will
win some coveted prize but bases his belief on wishful thinking is epistemically blame-
worthy in what he believes. Both this person and Mr. Raco believe in accord with, but
not from, the good evidence they have for their beliefs and so are objectively justified
in believing what they do. But I said the prize winner is epistemically blameworthy for
believing what he does but Mr. Raco is not. What is the difference between the two?
Neither of them bases his belief on the evidence he has; both of them have good evi-
dence for their beliefs. Why is one blameless and the other blameworthy?
I think the difference is that the person who bases his belief on wishful thinking
must think that he does not have sufficient evidence to support his belief and so be-
lieves against reason. However, after Mr. Raco attends medical school, he does not
believe he lacks sufficient evidence and so does not believe against reason. The first
person goes against his epistemic conscience but Mr. Raco does not. The difference
between the two is like the difference between someone who thinks that what he is
doing is wrong but goes ahead and does it anyway and someone who does not think
that what he is doing is wrong and does the same thing. Even if both do what they
 Epistemic and Moral Duty 41

have reason to believe is objectively permissible, the one is morally blameworthy

and the other is not. In other words, there are four possible epistemic cases:

bases belief on good does not base belief on

evidence he has good evidence he has
believes he does not have (1) (2)
good evidence
does not believe he does (3) (4)
not have good evidence

The case of wishful thinking falls in cell (2); the case of Mr. Raco in cell (4). Stan-
dard cases of both objectively and subjectively justified belief will fall in (3), say,
your belief that you are looking at a piece of paper.
A case that seems to fit in cell (1) is the following: I have good evidence that my
wife has been unfaithful to me. However, I have been going to a psychiatrist to deal
with my excessive jealousy and learn that out of jealousy I often jump to conclu-
sions when they are not supported by the evidence. Hence, I believe that the evi-
dence I now have that my wife has been unfaithful is not good evidence—though in
fact it is. Nevertheless, I find myself believing that she has been unfaithful and, in
fact, the basis of my belief is the evidence of her unfaithfulness that I have (e.g., her
staying late at the office, phone records of calls to a male colleague, a phone mes-
sage from that colleague filled with sexual innuendo and overtones left on our an-
swering machine, her reawakened interest in her appearance, and hints dropped by
her secretary that she is having an affair.)11
In both this case and the case of wishful thinking, the person is epistemically blame-
worthy in going against his conscience. In cases that fall in the second row of the
matrix, people are not epistemically blameworthy whether or not they base their
beliefs on the good evidence they have. This shows that epistemic blameworthiness
is not a function of what a person bases his belief on. Rather, it is a function of what
the person believes about the merits of the evidence he has or, to allow for negli-
gence, of what he should believe about that evidence. This parallels ethics, since a
person's moral blameworthiness is a function of what he believes or should believe
about the moral status of the action he performs.
Mr. Raco and Kant's grocer who gives the correct change because it's good for
business show that someone can be blameless even if they lack virtue, epistemic and
moral, respectively. And someone who has epistemic or moral virtue can on occasion
do something wrong or hold a belief for which he is epistemically blameworthy.
So there are many similarities between moral and epistemic duty, with the main
difference so far being that objective epistemic justification is perspectival, that is,
necessarily depends on the internal psychological states of the subject, while objec-
tive ethical justification need not. But you might be wondering whether there isn't
an even more crucial difference, namely, that actions are at least sometimes within
our control, while beliefs never are. And if beliefs never are, how can it make sense
to talk of epistemic duty and blame?
42 Bruce Russell

There are several types of reply that have been given. Richard Feldman has ar-
gued that people have certain types of obligations even when they cannot do what
they are obligated to do. He gives examples of academic, financial and legal obliga-
tions where a person is still obligated to do something, say, pay his mortgage, even
if he cannot do it or even if, for some peculiar reason, he is not able to avoid doing it.
And he argues that epistemic obligations are like those, in that they remain even if
the person cannot meet them or cannot fail to meet them.12
The problem is that we do not think it appropriate to blame people if they cannot
fulfill the obligations Feldman cites because fulfilling them is beyond their control.
However, talk of epistemic blame seems appropriate, for example, where you be-
lieve against your epistemic conscience, even if you have good evidence to support
your belief. So it seems necessary to argue that what doxastic attitude we take is, in
some sense, within our control.13
Some try to get around the problem by granting that beliefs are not in our control
but that what we accept is. We can evaluate our beliefs relative to a background sys-
tem and then either accept or reject them. Suppose I involuntarily believe that the
wall is red. Still, I can ask myself, say, whether I have any reason to think there is a
red light shining on this wall, which is really white, and, if I have, refrain from be-
lieving that the wall is red. On this view, beliefs are justified insofar as they would
be accepted by the person who holds them if he were seeking the truth and using his
background system of acceptances to evaluate the truth of his beliefs.14 It does not
matter whether the beliefs themselves are within our control as long as what we ac-
cept is.
The problem with this solution is that it seems to change the subject. Why should
people be epistemically blameworthy for what they believe if beliefs are not in their
voluntary control even if what they accept is? Suppose it were granted that delibera-
tion is within a person's voluntary control but that actions are not. We would then
want to ask a parallel question: why is a person blameworthy for his actions if they
are not in his voluntary control even if what he deliberates about is? So talk of what
a person accepts does not seem to face the issue.
The best solution, I think, is one that has been offered by Matthias Steup.15 His
basic idea is that beliefs are not different from actions with respect to voluntary con-
trol. It's true that I cannot decide whether to believe a car is coming at me when I see
one bearing down on me or whether I have hands, arms, and legs. But neither can I
decide to stick a knife in my eye for no reason, nor to kill a baby or rob a bank. In
general, I can neither decide to believe nor do what I think there is decisive reason
against, regardless of whether the reason is epistemic or practical.
Furthermore, I have a moral duty not to kill a baby or rob a bank even if doing
those things is not, in some sense, within my voluntary control. So why can't I have
an epistemic duty to believe there is a car coming at me, or that I have a hand, and so
forth, even if believing those things is not within my voluntary control? Perhaps
someone will respond that even if I have a duty not to kill a baby, if refraining from
killing a baby is not within my voluntary control, then I should not be blamed if I kill
one. Similarly, if what I believe is not within my voluntary control, I should not be
blamed if I do not believe it.
 Epistemic and Moral Duty 43

But this response elicits two replies: First, it does not defeat Steup's point that
actions and beliefs are on a par from the standpoint of duty and blame and, second,
suggests that some sort of compatibilist notion of voluntary control is sufficient for
ascribing blame. For instance, one might say that a person is blameworthy for doing
what he does if he does what, were he perfectly rational, he would not want to do
and he has no legitimate excuse (e.g., stemming from ignorance or his psychologi-
cal state) for doing it. Similarly, one might say that a person is blameworthy for the
epistemic attitude he has taken (belief, suspension of belief, disbelief) if he has an
epistemic attitude that he would not have were he perfectly rational and he has no
legitimate excuse (e.g., stemming from ignorance or his psychological state) for
having it.
The unfair contrast between the voluntariness of belief and action comes from
illegitimate comparisons where, say, epistemic reason requires me to believe some-
thing (e.g., that there is a piece of paper in front of me), while practical reason per-
mits either of two actions (e.g., my raising or not raising my arm). Where both
epistemic and practical reason require or prohibit something, the contrast dis-
appears. In normal situations, where a person's decisions are not controlled by
madness or outside forces, both belief and action are under the person's voluntary
control, in a sense that is compatible with determinism, and neither are under
his voluntary control in a sense that is incompatible with determinism. Hence,
epistemic duties and blame are possible if and only if ethical duties and blame are
possible.

III. The Significance of the Epistemic Distinctions

The distinctions between objective and subjective epistemic duty and between act-
ing from and acting in accord with such duty are important because I believe that a
person has knowledge only if she bases her belief on the evidence and believes in
accordance with both her objective and subjective epistemic duties. Mr. Raco does
not know that only members of the relevant race get the disease because knowledge
requires that the person base his belief on the evidence.
Alvin Plantinga has given many examples of people who he says lack knowl-
edge but have justified beliefs, and who suffer from brain lesions or whose beliefs
have been manipulated by demons or Alpha Centaurian scientists. For instance,
there is Paul who, when he hears the sound of church bells, believes something
orange is making that sound. Plantinga says that Paul has a "nearly ineluctable
tendency or impulse" to believe that there is something orange making the church-
bell sounds when he hears them. Plantinga asks us to assume that this tendency
has been caused by demons, Alpha Centaurian cognitive scientists, or a brain le-
sion, and that "those around him suffer from a similar epistemic deficiency." Plantinga
concludes that surely "Paul is doing his epistemic duty in excelsis" but, even if
what he believes is true, he lacks knowledge. Plantinga gives another example in
which he suffers from a brain lesion that causes his belief that he will be the next
president of the United States to seem as obvious to him as, say, 1 + 1 = 2 seems to
44 Bruce Russell

us. Here, too, Plantinga thinks it obvious that he is doing his epistemic duty in
believing he will be the next president but lacks knowledge even if it is true that he
will be the next president.16 However, these examples only show that a person can
be subjectively justified in believing something that is true, that is, can be epistemically
blameless in believing it because of a legitimate excuse (e.g., a brain lesion,
demons, Alpha Centaurian scientists), yet lack knowledge even where Gettier prob-
lems are absent. Still leaving Gettier problems aside, they do not show that objec-
tively justified true belief is not knowledge because they are not cases where ob-
jective justification is present.
However, the suggestion that knowledge is objectively justified true belief (plus
some condition to handle Gettier cases) also seems mistaken. In the example given
above of my believing that my wife has been unfaithful on the basis of what is in
fact good evidence, I am objectively justified in believing what I do and it is true,
but I lack knowledge. The reason I lack knowledge is that I am epistemically blame-
worthy in believing what I do, since I do not think I have good evidence for my
belief. Philippa Foot says the following of a person who acts against his moral
conscience:
One might say that there could not be a more radical moral defect than that of being
prepared to do what one believes to be wrong. A man of whom this is true is like an
archer who does not even aim at the target, it will be the merest chance if he does what
is good when he is doing what he sees as bad.17
People who act contrary to their epistemic conscience are prepared to believe what
they see as supported by bad evidence, and so it will be "the merest chance" if they
believe what is true. It is this element of chance that is incompatible with their hav-
ing knowledge.
So my proposal is that both objective and subjective justification are needed for
knowledge—Plantinga's examples showing that, leaving Gettier problems aside,
subjectively justified true belief is not enough and the last example showing that
objectively justified true belief based on the evidence is not enough either.
However, Plantinga has another example that seems to show that justification
is not necessary for knowledge, since a person can have knowledge and yet, in some
sense, violate his epistemic duty. In Plantinga's example, a big red London bus is
heading toward a subject who has been given reason to believe that if he believes
what his senses incline him to believe he will end up believing many falsehoods.18
In this example, the epistemic subject thinks that he has an epistemic duty to be-
lieve what is true and avoid believing what is false, which is understood to mean
that he has a duty to hold more true than false beliefs. That person also nonculpably
and truly believes that certain Alpha Centaurian conquerors will bring it about that
in the future he will have a set of absurdly false beliefs if he believes he is seeing
something red when he is appeared to redly. So he makes a great effort not to be-
lieve he is seeing something red even when it seems he is. Finally, the big red
London bus comes driving by, and he can't help himself. He forms the belief that
there is something red in front of him. Plantinga says that he violates his epistemic
duty and so is not justified in believing there is something red in front of him. Yet
 Epistemic and Moral Duty 45

we can imagine that, at least in certain circumstances of this sort, he knows there is
a red bus in front of him. Plantinga concludes justification is not necessary for
knowledge.
In what sense does this person violate his epistemic duty? Well, he would be blame-
worthy for going against his epistemic conscience in believing there is a big red bus
in front of him if he did not have an excuse. But doesn't he? I would think he does,
for he is like the drug addict who thinks that he should stay clean but succumbs to
temptation when his buddies start shooting up in front of him. The red, bus-like sen-
sations, the sound of the engine, the smell of the burned gasoline are just too much
for our epistemic subject. Though he thinks he shouldn't, he nevertheless gives in
and believes there is a big red bus in front of him. So he is epistemically blameless in
believing what he believes.
Does he violate an objective epistemic duty? Well, he would if he had an objec-
tive duty to believe something if believing that thing would mean that he would have
more (perhaps, important) true than false beliefs throughout the course of his life.
And then assuming, as I do, that the subject knows there is a big red bus in front of
him, this would be a counterexample to my claim that knowledge requires objective
justification, since the subject would have knowledge despite violating his objective
epistemic obligation.
However, I take this example to show that a person's objective epistemic obliga-
tion is not to believe something if and only if believing it will result in her believing
more true than false (important) propositions in the future course of her life. Perhaps
that is the appropriate epistemic goal, just as maximizing happiness might be the
appropriate moral goal. But just as plausibly, one might contend that the appropriate
epistemic goal is to believe some proposition if and only if it is true.19 In that case, if
the appropriate goal directly determines the duty, the subject in the example would
have a duty to believe that there was a big red bus in front of him, and so he would be
objectively justified in believing what he does.
Furthermore, if truth is the appropriate epistemic goal, it does not directly de-
termine objective epistemic duty as the above discussion of the contrast between
epistemic and moral duty showed. Objective epistemic duty is relative to what a
person has good reason to believe, and good reason does not imply truth or even
objective (or factual) probability of truth—as demon world examples show. As we
have seen, objective ethical duty is not always perspectival in this way, at least
sometimes being relative to what is in fact the morally best course of action inde-
pendently of what the agent believes. Here there is a danger that the analogy with
ethics will mislead.
Of course, if reliabilists were right in thinking the reliably produced true belief is
knowledge, then no sort of justification would be necessary for knowledge. But the
reliabilists proposal seems wrong even when it is coupled with Plantinga's idea that
the reliably produced true belief must also result from "cognitive faculties function-
ing properly in an appropriate environment according to a design plan successfully
aimed at truth."20 Ironically, Plantinga offers an example that seems to show that these
conditions are not sufficient for knowledge. He asks us to suppose God (or evolu-
tion) designed human beings so that
46 Bruce Russell

Just as we are by nature such that when appeared to by something that is red, we form
the belief that we are appeared to in that way by something that is red, so these crea-
tures are by nature such that when appeared to in the church-bell fashion, they form
the belief that they are appeared to that way by something that is orange.21
Someone is appeared to in the church-bell fashion just in case he seems to hear the
sound of church bells. We are to assume that "there is a common but rarely visible
[orange] bird that makes the church-bell sound" and that most things that appear
in the church-bell fashion on this planet really are orange, though atmospheric con-
ditions make it nearly impossible for the people on the planet to confirm this con-
nection between sound and color. He then asks rhetorically why the inhabitants
wouldn't know that this orange bird that makes the church-bell sound is orange
simply on the basis of hearing (but never seeing) the bird. Of course, the belief
would be the result of cognitive faculties functioning properly in an appropriate
environment according to a design plan successfully aimed at truth and so, according
to Plantinga's theory, the people would know that the bird is orange.
But despite Plantinga's rhetorical question, I would think that nearly everyone
would agree that the inhabitants of the planet do not know that the bird is orange.
Epistemically, they are in no better position than Paul of a few pages earlier who
believed on no evidence that things that make a church-bell sound are orange. The
inhabitants of the planet have no grounds for thinking that the bird, or any of the
other things that make the church-bell sound, are orange. They have never con-
firmed that things that make that sound are orange. And even if grounds are not
needed for knowledge in some cases, say, when I believe the corresponding con-
ditional of modus ponens or that it is wrong to lie about my colleagues in order
to advance my career,22 surely they are needed here. If epistemic subjects lack the
appropriate evidence for their beliefs, reliabilities that are hidden from them
won't turn non-knowers into knowers. Neither Paul nor his doppleganger on the
strange planet will know that something that makes the church-bell sound is or-
ange.23 Plantinga's diagnosis of why Paul lacks knowledge is mistaken. He does
not lack knowledge because he is not functioning properly; he lacks it because he
lacks the appropriate evidence. And the same is true of his counterparts on the other
planet.
If reliably produced true belief in an organism that is functioning properly is not
sufficient for knowledge, how is it possible for young children and animals to have
knowledge? Doesn't my account imply that they do not, since they lack the concepts
of justification and so cannot have a justified true belief? My answer is that it allows
for children and animals to have knowledge, since one can be epistemically justified
in believing something without having the concept of justification. To be subjectively
justified, one must be epistemically blameless, but children and animals will be
epistemically blameless, just as they are morally blameless, just because they lack
the concept of justification and so have a legitimate excuse for believing or doing
whatever they do. They can be objectively justified because they can be in appropri-
ate internal states that provide evidence for what they believe, even though they do
not realize that they do. Furthermore, putting Gettier problems aside, if children or
animals are objectively justified in believing what they do and their beliefs are caused
by the evidence they have, then they also have knowledge. So my account of knowl-
 Epistemic and Moral Duty 47

edge as subjectively and objectively justified true belief (with some added condi-
tions to handle Gettier problems) will offer a unified account of knowledge that ap-
plies to normal adult human beings as well as to children and animals.24

Notes
1. Richard Feldman, "Epistemic Obligations," in Philosophical Perspectives, 2: Episte-
mology, James E. Tomberlin, ed. (Atascadero, Calif.: Ridgeview, 1988), 235.
2. Alvin Goldman, "Strong and Weak Justification," Philosophical Perspectives, 2: Epis-
temology, 51-69; the quotes are from 53 and 51, respectively.
3. Ibid., 58.
4. Ibid., 59.
5. Ibid., 64-65.
6. Feldman, 411 and 414, respectively.
7. Richard Feldman, "Subjective and Objective Justification in Ethics and Epistemology,"
The Monist 71 (1988): 408. It is important to note that objective ethical justification need not
be independent of everyone's cognitive or psychological states; otherwise, there could be no
objective duties for utilitarians. And requiring that they be independent of the agent's cogni-
tive states will imply that we will not be able to have an objective Kantian duty not to treat
people as mere means, since whether we treat them in that way is at least partly a function of
our beliefs and intentions. Perhaps the best way to contrast objective ethical and epistemic
justification is to say that necessarily the latter is a function of what is internal to the person,
viz., his beliefs and sensations, while the former need not be (as in the case of the physician
who will kill his patient if he does what he has every reason to think is best for the patient).
Remarks made by my colleague Stephen Kershnar drew my attention to this problem with
Feldman's account of objective ethical justification.
8. Ibid., 410.
9. Ibid., 411.
10. Keith Lehrer, Theory of Knowledge (Boulder, Colo.: Westview, 1990), 172.
11. I owe this example to my colleague, Herbert Granger.
12. Feldman, "Epistemic Obligations," 240-43.
13. See Matthias Steup, An Introduction to Contemporary Epistemology (Upper Saddle River,
N.J.: Prentice Hall, 1996), 76-77, where he writes that, "Feldman's solution will be unappealing
to those who find the voluntariness principle intuitively plausible." That principle says, roughly,
that if I have an epistemic duty to believe p, then my believing p must be voluntary.
14. Keith Lehrer gives an account of knowledge in terms of what one is justified in ac-
cepting, not in terms of what one is justified in believing. See his Theory of Knowledge, esp.,
10-11, 20-21, 26-27, 113-114; and "Proper Function versus Systematic Coherence," in
Warrant in Contemporary Epistemology, Jonathan L. Kvanvig, ed., (Lanham, Md.: Rowman
& Littlefield, 1996), 25^t5, especially, 32-33, 35, 37, 40. Feldman discusses Lehrer's posi-
tion in "Epistemic Obligations," 240, but prefers his solution that recognizes that we can have
obligations with respect to involuntary behavior.
15. In his Introduction to Contemporary Epistemology but more fully in "Doxastic
(In)Voluntarism and Epistemic Deontology," presented to the Department of Philosophy,
Wayne State University, Dec. 11, 1997. Steup's discussion is much more complicated than I
indicate. He distinguishes between hypothetical and categorical voluntary control of actions
and beliefs and discusses whether each is compatible with determinism. He argues that we
have a great deal of hypothetical voluntary control, and little categorical voluntary control,
over both beliefs and actions, and that moral duties do not require categorical voluntary con-
trol, only hypothetical and compatibilist voluntary control. So there is no reason to think that
epistemic duties require more. I summarize all this by saying that he sees actions and beliefs
on a par with respect to voluntary control.
16. See Alvin Plantinga's Warrant: The Current Debate (New York: Oxford University
Press, 1993), 42 for the example of Paul; p. 44 for the example involving Plantinga himself.
48 Bruce Russell

17. Philippa Foot, "Moral Relativism," reprinted in Relativism: Cognitive and Moral, Jack
W. Meiland and Michael Krausz, eds. (Notre Dame, Ind.: University of Notre Dame Press,
1982), 159.
18. See Plantinga's Warrant: The Current Debate, 45.
19. In "The Coherence Theory of Knowledge," Philosophical Topics 14, no. 1 (Spring
1986), 6-7, Keith Lehrer writes:
It is acceptance of something for the purposes of obtaining truth and eschewing error
with respect to just the thing one accepts that is a condition of knowledge. . . . I add
the qualification "with respect to just the thing one accepts" because it is possible that
accepting something that the evidence indicates is false [e.g., that there is nothing red
in front of one] might serve the general purpose of accepting as much as one can of
what is true and accepting as little as one can of what is false. Accepting the one false-
hood might be bountifully fecund with respect to accepting other truths. It is clear,
however, that accepting something that the evidence indicates is false for such a gen-
erally worthwhile purpose is not the sort of acceptance that is required to know that
the thing in question is true. (My addition in brackets.)
Lehrer says that the acceptance of a "bountifully fecund" falsehood is not required for knowl-
edge. My point is that it, together with true belief and a condition to handle Gettier cases, is
not sufficient for knowledge either. Our epistemic duties not only do not require, they actu-
ally prohibit, our accepting what we have sufficient reason to believe are "bountifully fe-
cund" falsehoods. Feldman distinguishes between goals and obligations in "Epistemic Obli-
gations," pp. 246^48, 254-55 and in "Subjective and Objective Justification" p. 410.
20. Alvin Plantinga, "Why We Need Proper Function," Nous 27, no. 1 (1993), 73.
21. Platinga, Warrant: The Current Debate, 62; his description of the case continues
on 63.
22. Plantinga, "Respondeo," Warrant in Contemporary Epistemology, 332, gives these
examples in response to William Lycan's paper.
23. The same considerations apply to Laurence BonJour's well-known example of Norman,
who one day, on the basis of clairvoyance but for no reason, comes to believe that the Presi-
dent is in New York. It is assumed that clairvoyance is a reliable belief producing mecha-
nism, though Norman has no reason to think it is. Hence the relevant reliability is "hidden"
from Norman. See Bonjour's The Structure of Empirical Knowledge (Cambridge, Mass.:
Harvard University Press, 1985), 41-42.
24. I want to thank my colleagues at Wayne State University, especially Herbert Granger,
for their comments on an earlier draft of this paper that I delivered at a departmental collo-
quium. I am especially grateful to Sharon Ryan and Matthias Steup who gave me detailed
written comments on an earlier version of the paper. I also want to thank Matthias for his
photograph of what certainly looks to be a barn.
 3

Epistemic Justification and Normativity

RICHARD FUMERTON

It is plausible to argue that the concept of epistemic justification is the most funda-
mental concept in epistemology. The so-called traditional account of knowledge takes
justified belief to be a constituent of knowledge.1 Furthermore, on many accounts of
knowledge, the conditions for knowledge that go beyond having justified belief, for
example, the truth condition and conditions designed to "Gettier-proof" the analy-
sis, seem to be less interesting to the philosopher seeking assurance of truth from the
first-person perspective. There is a sense in which the best one can do through philo-
sophical reflection is assure oneself that one has a justified belief—whether or not
one has knowledge as well is a matter of "luck," is a matter of whether the world
cooperates so as to reward justified belief with truth.
It is an understatement to suggest that there is no agreement among epistemolo-
gists as to how to analyze the concept of epistemic justification. But a surprising
number of philosophers with radically different approaches to analyzing justified
belief seem to agree that the concept of epistemic justification is in some sense a
normative concept.2 The issue is potentially significant because the alleged norma-
tivity of epistemic justification has been used to attack prominent analyses of justi-
fied belief. Ironically, many of these attacks have focused on externalism. The irony
lies in the fact that the most prominent externalist, Alvin Goldman (1979 and 1986),
explicitly endorses the claim that the concept of epistemic justification is a norma-
tive concept and denies for that reason that he is proposing a meaning analysis of
epistemic justification. Rather he proposes to identify the nonnormative (necessary
and sufficient) conditions on which epistemic justification supervenes. But whether
he was proposing a meaning analysis or identifying synthetic necessary and suffi-
cient conditions for the application of concept, a number of his critics have com-

49
50 Richard Fumerton

plained that one can have a belief that results from an unreliable process, even though
it would be quite inappropriate to blame the person for having the belief or criticize
the person for the way in which the belief was formed.3 If the concept of epistemic
justification is genuinely normative, how can we describe such a belief as unjusti-
fied? How can we characterize the victims of demonic machination as having unjus-
tified beliefs when such victims are believing precisely what they should believe given
the available subjective evidence (evidence that is phenomenologically indistinguish-
able from the evidence you and 1 use to reach our conclusions about the physical
world)?
The above objection may well confuse evaluation of a subject with evaluation of
a subject's belief, but it may be enough to motivate a more-detailed examination of
the question of whether and in what sense epistemic justification is usefully thought
of as normative. In examining this question, we must get clear about what makes a
concept or judgment normative.

Epistemic Judgments and Value Terms

One might begin to suspect that a judgment is normative if it is equivalent in mean-

ing to a conjunction of statements that include paradigmatically normative terms. This
approach would seem to require that we give some characterization of what makes a
term normative, but we might try to side-step this problem initially by simply listing
some paradigmatic normative expressions and characterizing as derivatively norma-
tive other expressions whose meaning can be partially explicated using these. Our
list of paradigm normative expressions might be long or short depending on whether
we are reductionists with respect to the content of various sorts of normative judg-
ments. Thus, if one is a consequentialist of some kind who thinks that all ethical judg-
ments are ultimately judgments about the ways in which actions produce things of
intrinsic value, one might get by with "intrinsically good/bad" as the fundamental
normative terms—all other normative terms will be derivatively normative because
an explication of their meaning will inevitably involve reference to intrinsic good-
ness/badness. But so as not to prejudice such issues, one might make the initial list
relatively long and include such terms as "good," "ought," "should," "right," "per-
missible," "obligatory," and their opposites.
If we proceed in this fashion, it seems undeniable that the concept of epistemic
justification looks suspiciously like a normative concept. As Plantinga (1992) has
effectively reminded us, the etymology of the word "justification" certainly suggests
that we are dealing with a value term. And epistemologists often seem quite com-
fortable interchanging questions about whether evidence £ justifies one in believing
P with questions about whether or not one should believe P on the basis of E. In
what is often taken to be one of the earliest statements of a justified true belief ac-
count of knowledge, Ayer (1956) described knowledge as true belief where one had
the right to be sure. So again, the idea that the concept of justification is normative is
at least prima facie plausible. But we must surely proceed more cautiously than this.
While it may be all right to begin by listing paradigm normative expressions and
 Epistemic Justification and Normativity 51

characterizing judgments as normative whose meaning can be explicated (in part)

through the use of these expressions, it doesn't require much reflection to convince
us that expressions like "right" and "should" are importantly ambiguous. When we
talk about whether someone should do X, we might be talking about what that per-
son morally should do, prudentially should do, legally should do, should do given
the rules of etiquette, should do given that the person has certain goals or ends, and
so on. If we add to the mix judgments about what someone should believe, it seems
that we must add to the list of "shoulds" the epistemic "should." If it makes sense to
treat belief as something one can do (and be held responsible for), then it seems
obvious that we must carefully distinguish our moral obligations with respect to what
we should believe, what prudence dictates, and what it is epistemically rational to
believe.4 Thus it has been argued that a husband might have a special moral obliga-
tion to believe in his wife's innocence even in the face of rather strong evidence that
she is guilty of infidelity. It might also be the prudent thing to do in the sense that his
subjective goals or ends might be more effectively satisfied by trusting his wife. But
at the same time it might be wildly irrational epistemically to believe in his wife's
innocence.
There has been a great deal of literature attempting to cast doubt on the intelligibil-
ity of treating believing as an action, as something one chooses to do. One doesn't just
decide to believe something the way one decides to go to the store. Many of our beliefs
might seem to be forced on us in a way that makes inappropriate questions about whether
we should have the beliefs in question.5 At the same time, it is hard to deny that one
can indirectly influence one's beliefs. If one concludes that one would be happier if
one believed in an afterlife and that it would be advantageous to have such a belief,
there are certainly things one can do that will increase the probability of bringing about
the belief. In any event, I am not concerned in this essay with the question of whether
it makes sense to talk about what a person ought to believe. I presuppose the intelligi-
bility of such judgments but insist that we make the relevant distinctions between kinds
of judgments we can make about what we ought to believe.

Ethical Judgments as the Paradigm of Normativity

If we recognize the ambiguity inherent in judgments about what one ought to be-
lieve, then one must decide whether it is all or only some of these "oughts" that
indicate the normativity of judgments that employ them. One approach is to begin
by simply stipulating that the moral "ought" is the example of a normative expres-
sion, par excellence, and the question of whether the epistemic "ought" is normative
rests on how close its meaning is to the moral "ought." But if this is the approach we
take then to investigate the relevant similarities, we will still need to characterize what
it is about moral judgments that makes them normative.
At this point, our investigation into the alleged normativity of epistemic judgments
seems headed into a morass of issues involving metaethics. There is no agreement
among ethical philosophers about what makes moral judgments distinctively nor-
mative, nor indeed what the relevant contrast is supposed to be between the norma-
52 Richard Fumerton

live and the nonnormative. For many, the relevant contrast is between descriptive
judgments (concepts, terms) and prescriptive judgments (concepts, terms). Accord-
ing to many of the classic noncognitivists, the normativity of ethical judgments con-
sists specifically in the fact that their primary function is not to describe some state
of affairs but rather to recommend or prescribe some specific action or action kind.
The most straightforward version of this view is Hare's claim that moral judgments
are grammatically disguised (universalizable) imperatives.6 Frankly, I don't know
of any prominent epistemologists who endorse the idea that epistemic judgments are
normative and who explicitly intend thereby to contrast them with descriptive judg-
ments that have a truth value.7 We can put the conclusion conditionally. If moral
judgments are disguised imperatives lacking truth value, and if one is a cognitivist
with respect to epistemic judgments, then one must surely hesitate before reaching
the conclusion that epistemic judgments are in some important sense normative.
Of course, not all ethical philosophers are noncognitivists. Indeed, noncognitivism
is very much a product of twentieth-century philosophy. If one holds that there are
genuine moral properties and that moral judgments typically describe their exempli-
fication by things, people, or actions, what would the relevant contrast be between
the way in which these judgments are normative and the way in which other descrip-
tive claims are not normative? One can, of course, simply stipulate that a judgment
is normative if and only if it refers directly or indirectly to these distinctively moral
properties. But if we take this approach, then after we distinguish the epistemic "ought"
from the moral "ought," there isn't even a prima facie reason to suppose that epistemic
judgments are normative in this sense. If referring to moral properties is a necessary
condition for a judgment's being normative and we reject any reduction, in part or in
whole, of epistemic judgments to moral judgments, then we will have removed
epistemic judgments from the class of normative judgments.
One can try to combine one's descriptivism in ethics with an acknowledgment of
the claim that morality necessarily motivates rational people. And one could go on
to describe the normative character of moral judgments as consisting precisely in
this "pull" that moral judgments have. Just as one cannot recognize that one ought to
take some action X without being "moved" to do X, so one cannot recognize that one
epistemically ought to believe P without being at least moved to believe P. But it is
precisely the acknowledgment of this special character that moral judgments are
supposed to have that leads so many philosophers either to abandon descriptivism in
ethics or combine it with some version of subjectivism. If moral judgments describe
objective properties, it is more than a little difficult to see how the mere belief that
something has the property can in itself necessarily motivate the person to pursue
that thing.8 If the connection is only contingent, then the claim that it exists might be
philosophically unproblematic (though empirically suspect). I certainly have no in-
terest in denying that when one decides that it is epistemically rational to believe P,
one sometimes (or even usually) ends up believing P as a result, and if the existence
of a propensity to believe what one judges epistemically rational to believe is all that
is meant by claiming that epistemic judgments have normative force, I concede that
they might well be normative in this sense (though again, the normative character of
a judgment is now a matter for empirical investigation).
 Epistemic Justification and Normativity 53

Normativity and Rules

Without identifying normative judgments with prescriptive judgments, one might still
suppose that Hare was on to something in his attempt to characterize what makes
normative judgments special. A great many philosophers concerned with metaethics
have sought to tie the meaning of ethical judgments to rules. Hare thought of those
rules as universalizable imperatives, but one needn't go that far to embrace the con-
clusion that moral judgments always involve at least implicit reference to rules. To
judge that one ought to do X is to judge that the relevant rules of morality require one
to do X. To judge that it is morally permissible that one do X is to judge that the
relevant rules of morality do not prohibit one from doing X. And to judge that it would
be wrong for one to do X is to judge that the relevant rules of morality do prohibit
one from doing X. If we turn to judgments about what one is legally required, per-
mitted, or prohibited from doing, one might suppose that there, too, the relevant con-
cepts are to be defined by reference to rules, this time the rules of law. Legally pro-
hibited actions are those the rules of law prohibit. Legally permitted actions are those
the rules of law do not prohibit. Even etiquette has its "rules," and one can easily
follow the model to define the relevant judgments concerning what one ought to do
from the perspective of etiquette. Perhaps, then, we should view normative judgments
as those that make implicit reference to rules that prescribe, permit, and prohibit certain
actions or moves, and epistemic judgments might be viewed as paradigmatically
normative because there are certain rules of inference that tell us when we must be-
lieve, are permitted to believe, or are prohibited from believing certain propositions,
given that we believe certain others or are in certain nondoxastic states (in the case
of noninferentially justified belief).
While the above might seem initially promising, it is clear that we must proceed
more carefully lest we overlook important distinctions between the kinds of rules to
which judgments might make implicit or explicit reference. In metaethics, there is
again no consensus on whether the content of moral judgments does always involve
reference to rules or, if they do, how we should understand those rules. It is useful,
however, to distinguish two importantly different kinds of rules. Some rules, for
example, the rules that a rule utilitarian has in mind in analyzing the content of moral
judgments, can themselves be thought of as propositions that have a truth value. Thus,
according to some rule utilitarians, the relevant rules take the form: It is always (prima
facie) right (wrong) to take some action of kind X. The statement of the rule will be
true if a certain proposition describing the consequences of people following that
rule compared to the consequences of their following alternative rules is true.9 The
rules of law, the rules of a game, or the rules of etiquette, might be better thought of
as imperatives that are neither true nor false. Propositions describing particular ac-
tions as permissible or impermissible relative to the rules are true or false but are so
because they report what the relevant rules prescribe and prohibit. One can, of course,
take precisely the same approach with respect to moral judgments, but as I indicated,
one certainly need not.
If epistemic judgments involve implicit reference to rules, how should we think
of those rules? Again, one could be a noncognitivist with respect to the relevant rules.
54 Richard Fumerton

One could think that rules of nondeductive inference, for example, are imperatives
that are neither true nor false. Individual epistemic judgments are either true or false
but only because they report what the relevant epistemic rules require, permit, and
prohibit. But I daresay most epistemologists would resist this suggestion. The rele-
vant generalized rules of epistemology will take the form of propositions that assert
that one is justified in believing certain propositions relative to one's justifiably be-
lieving certain others or relative to one's being in certain nondoxastic states. It doesn' t
hurt to characterize these propositions as rules, if one likes, but if the "rules" them-
selves have a truth value, then it is not clear to me that we have uncovered an inter-
esting sense in which epistemic judgments are normative. Epistemic judgments are
no more normative than judgments about lawful necessity and possibility are nor-
mative. Such judgments also implicitly involve reference to general propositions. To
claim that it is lawfully possible that X is probably just to claim that the conjunction
of the laws of nature, L, is logically consistent with the proposition describing the
occurrence of X. Events "obey" laws in the sense that we can usefully generalize over
kinds of events that always occur. In the same sense, individual beliefs are justified
or not in virtue of exemplifying certain general properties, where we think of the
"rules" of epistemology as generalizations describing the kinds of conditions under
which beliefs are justified.

Normativity and Goals or Ends

Richard Foley (1987) and others have suggested that we might profitably view the
different "oughts" as species of a common genera. Crudely put, Foley's idea is that
normative judgments all assess the efficacy of achieving goals or ends. In a sense, all
normative judgments are species of judgments concerning practical rationality. There
are different kinds of normative judgments concerning what we ought to do and what
we ought to believe because there are different goals or ends that we are concerned to
emphasize. Thus when we are talking about morally justified action (what we morally
ought to do), the relevant goal might be something like producing moral goodness
(avoiding evil), and the actions that we ought to perform are those that are conducive
to the goal of producing the morally best world. When we are concerned with what
prudence dictates, however, the relevant goals or ends to be considered expand, per-
haps to include everything that is desired intrinsically, for example. On one (rather crude)
view, what one prudentially ought to do is what maximizes satisfaction of one's de-
sires. What one ought to do legally or what one is legally justified in doing is a func-
tion of the extent to which an action satisfies the goal of following the law. What one
ought to do from the standpoint of etiquette is a function of following the goals or ends
set down by the "experts" who worry about such things. So all one has to do to fit the
epistemic "ought" into this framework (and thus classify usefully the kind of normativity
epistemic judgments have) is delineate the relevant goals or ends that define what one
epistemically ought to believe. And the obvious candidates are the dual goals of be-
lieving what is true and avoiding belief in what is false.
If Pascal were right about his famous wager, belief in God might be the path one
prudentially ought to follow, focusing on such goals as avoiding pain and seeking
 Epistemic Justification and Normativity 55

comfort. If you have promised your parents to believe in God, if it is good to keep a
promise, and if there are no other good or bad effects of such a belief to consider, it
might follow that prima facie you morally ought to believe in the existence of God.
But neither of these normative judgments is relevant to whether you epistemically
ought to believe in the existence of God. The only consideration relevant to this
normative judgment is the efficacy with which such a belief contributes to the goals
of believing what is true and avoiding belief in what is false.
Now as plausible and potentially illuminating as this account might seem initially,
it is, I think, fatally flawed. In the first place, it must be immediately qualified to ac-
commodate certain obvious objections. Suppose, for example, that I am a scientist in-
terested in getting a grant from a religious organization. Although I think that belief in
the existence of God is manifestly irrational (from the epistemic perspective), I dis-
cover that this organization will give me the grant only if it concludes that I am reli-
gious. I further have reason to believe that I am such a terrible liar that unless I actually
get myself to believe in the existence of God, they will discover that I am an atheist.
Given all this and my desire to pursue truth and avoid falsehood, which I am convinced
the grant will greatly enable me to satisfy, I may conclude that I ought to believe in the
existence of God (or do what I can to bring it about that I believe in the existence of
God). Yet, by hypothesis, this belief is one that I viewed as epistemically irrational.
We cannot understand epistemic rationality simply in terms of actions designed to sat-
isfy the goals of believing what is true and avoiding belief in what is false.
How might one modify the account to circumvent this difficulty? Foley suggests
restricting the relevant epistemic goal to that of now believing what is true and now
avoiding belief in what is false.10 Even this, however, will fall prey to a revised (al-
beit more farfetched) version of the objection presented above. Suppose, to make it
simple, that belief is under one's voluntary control and that I know that there is an
all-powerful being who will immediately cause me to believe massive falsehood now
unless I accept the epistemically irrational conclusion that there are unicorns. It would
seem that to accomplish the goal of believing what is true and avoiding belief in what
is false now, I must again adopt an epistemically irrational belief.
The obvious solution at this point is to restrict the relevant goal that defines the
epistemic "ought" to that of believing what is true now with respect to a given propo-
sition. If I epistemically ought to believe that there is a God, the only relevant goal is
that of believing what is true with respect to the question of whether there is or is not
a God. If we say this, however, we must be very careful lest our account collapse the
distinction between true belief and epistemically justified or rational belief. If we
are actual consequence consequentialists11 and we take what we ought to do or be-
lieve to be a function of the extent to which our actions and beliefs actually satisfy
the relevant goals, then trivially we epistemically ought to believe in God when there
is a God, and we epistemically ought not believe in God when there is no God. Foley
suggests at this point that it is something about beliefs an agent has (or more pre-
cisely would have after a certain process of reflection) about the efficacy of achiev-
ing the epistemic goals that is relevant to evaluating what one epistemically ought to
believe. But there is a much more natural way of explicating the relationship between
epistemic goals and what a person ought to believe, just as there is a more natural
way of explicating the relevant relation that holds between a person's moral goals
56 Richard Fumerton

and what a person morally ought to do and a person's prudential goals and what a
person prudentially ought to do. The obvious move is to say simply that what a per-
son ought to believe is a function of what that person is justified in believing would
accomplish the goal of believing now what is true with respect to a given proposi-
tion. But that is, of course, a convoluted way of saying that what a person is justified
in believing is what a person is justified in believing, an account entirely plausible
but less than enlightening.
Notice too that on many standard consequentialist accounts of morality or practi-
cal rationality, it is also crucial to introduce epistemic concepts into the analyses of
what one morally or prudentially ought to do. I have argued in some detail that the
concepts of what one morally ought to do and what one rationally ought to do are
extraordinarily ambiguous.12 Although there are actual consequence consequentialist
analyses of what one morally or rationally ought to do that find occasional expres-
sion in ordinary discourse, they are far from dominant.13 Consider the sadist who
kills for pleasure the pedestrian in the mall when that pedestrian (unbeknownst to
the sadist) was a terrorist about to blow up the city. There is surely a clear sense in
which the sadist did not behave as he morally ought to have behaved.14 The conven-
tional poker wisdom that one should not draw to fill an inside straight is not falsified
by the fact that this person would have filled the straight and won a great deal of
money. How can we acknowledge that a person did what he ought to have done even
when the consequences are much worse than would have resulted from an alterna-
tive? How can we acknowledge that a person behaved as he should not have behaved
even when the consequences are far better than would have resulted from some
alternative? The answer seems obvious. We must recognize the relevance of the
epistemic perspective of the agent.
To determine what someone (morally or prudentially) ought to have done, we must
consider what that person was epistemically justified in believing the probable and
possible consequences of the action to be. Indeed, I have argued that there are liter-
ally indefinitely many derivative concepts of morality and rationality that also take
into account what a person was epistemically justified in believing about the moral-
ity or rationality of actions, given more fundamental concepts of morality and ratio-
nality.15 But if the analysis of familiar concepts of what a person ought to do must
take into account the epistemic situation of the agent, it is simply a mistake to try to
assimilate the epistemic "ought" to the "ought" of morality or practical rationality.
In fact, an understanding of the "oughts" of morality and practical rationality is para-
sitic on an understanding of rational or justified belief. It would be folly, needless to
say, to try to understand fundamental epistemic concepts in terms of what the agent
was epistemically justified in believing about the probable and possible consequences
of having a certain belief. Even philosophers who do not mind "big" circles in their
philosophical theories will get dizzy traveling the circumference of this one.

Normativity, Praise, and Criticism

So far the only sense in which we have acknowledged that epistemic judgments are
normative is that they are sometimes expressed using an "ought." That "ought" has
 Epistemic Justification and Normativity 57

been shown not only to be distinct from other "oughts" used in the expression of
paradigm value judgments, but it has been shown to be fundamentally different.
Nevertheless, we have not yet exhausted attempts to explicate normativity in a way
that allows us to fit both epistemic judgments and our paradigm normative moral
judgments under the same umbrella. It is sometimes claimed that our epistemic judg-
ments are normative in that they implicitly involve praise or blame and criticism.
Should we construe this as the relevant mark of normativity? Almost surely not. The
problems with doing so are enormous. For one thing, however we define normativity,
we want our paradigm of normative judgments, moral judgments, to fall under the
concept. But it is far from clear what the relationship is between judging that some-
one did not do what he or she ought to have done and blaming or criticizing that
person.
If you see a fire in the house next door and heroically attempt to save the people
inside, I may conclude that you ought to have called the fire department instead of
trying to solve the problem on your own. At the same time I might not blame you for
failing to make the call. I might decide that under the circumstances it is perfectly
natural for a person to panic and fail to do the rational thing. I might also think that
you are just too stupid to figure out what you ought to do, and indeed, I might sel-
dom blame you for the many idiotic things you do that you should not do. In short,
there seems to be no conceptual connection between the evaluation of an agent's action
and the praise or blame of the agent who acted that way. And if this seems right con-
cerning the evaluation of what a person ought to have done, it seems even more
obvious in the epistemic evaluation of a person's belief. Do we blame or criticize
very stupid people for believing what they have no good epistemic reason to believe?16
At the very least, logic does not require us to blame people for believing what it is
epistemically irrational for them to believe.
It might be argued, however, that I am confusing the praise or blame of an agent
with the positive evaluation or criticism of the agent's action or belief. "I am not
criticizing you," someone might say, "I am criticizing what you did." And there surely
does seem to be some sense in which when one's beliefs are called unjustified or
irrational, one takes those beliefs to have been criticized. Shall we say that judgments
about the epistemic justifiability or rationality of a belief are normative in that they
imply praise or criticism of the belief (as opposed to the subject who has that belief)?
This is not helpful for two reasons. First, the notion of implying praise or criti-
cism is simply too vague. When I tell the store owner that the knife I bought is ex-
tremely dull, there is surely a sense in which I am criticizing the knife (or implying
criticism). When after test driving the car, I complain that it accelerates very slowly
and pulls to the left, I am in some sense criticizing the car. But does that make "dull,"
"accelerating slowly," and "pulling to the left" normative expressions? Surely not.
But why? One answer might be that there is no conceptual connection between judging
that something has these characteristics and criticism. I might have wanted a dull
knife to minimize the possibility of accident, for example. Now is there any concep-
tual connection between judging of a belief that it is epistemically irrational and criti-
cizing the belief? Can we not imagine societies in which one values a kind of irratio-
nality much the way a few people value dull knives? Indeed, I can think of a few
philosophical movements that for all the world seem to place a premium on the inco-
58 Richard Fumerton

herence of belief systems. And if that suggestion seems a little snide, can we not at
least find some subculture of poets who explicitly disdain the confines of epistemically
rational belief systems, the pursuit of truth, and so on? I have already agreed, of course,
that there is a sense of "ought" that is customarily used in describing beliefs that a
person is justified or rational in holding. And one can claim that if a belief is judged
to be irrational, it is being implicitly criticized as one that the subject ought not to
have, but this will now take us full circle to the earlier problematic attempt to char-
acterize the normativity of the epistemic "ought."

Conclusion

We have explored a number of different ways in which we might interpret the claim
that epistemic judgments are normative. But after we have carefully distinguished
the epistemic judgments we make about beliefs from the other ways in which we
might evaluate beliefs, it is not clear to me that there is really any interesting sense in
which epistemic judgments are normative. Indeed, it is not clear that we can really
develop any philosophically interesting sense of normativity that does not itself pre-
suppose highly controversial views. If any judgments are normative, it is ethical judg-
ments, but unless some version of noncognitivism is true, ethical judgments describe
some feature of the world in precisely the same sense in which other judgments de-
scribe some feature of the world. We explored the idea that the relevant feature of
the world might be the existence of rules that lack a truth value, and that this might
be the essence of their normativity, but we saw that a great many moral philosophers
would deny that the relevant moral rules lack truth value, and an even greater num-
ber of epistemologists would resist the analogous suggestion that the relevant
epistemic rules lack truth value. The idea that normative judgments all make implicit
reference to goals or ends gave little comfort to the idea that epistemic judgments
are normative, for on reflection, the way in which other normative judgments involve
reference to goals or ends seems to presuppose a prior understanding of epistemic
probability. It seems even more hopeless to claim that there is a conceptual connec-
tion between judgments about epistemic justification and praise and blame.
If the above is correct, then some epistemic internalists may be off target in their
criticisms of externalism. As we noted at the beginning of the essay, many would
argue that externalist epistemologies are implausible precisely because they fail to
capture some alleged normativity of moral judgments. Although I believe there are
fatal objections to the externalist's approach to understanding epistemic concepts,
I'm not convinced that this is one of them.17

Notes
1. Butchvarov (1970) has argued, somewhat persuasively, that the "traditional" account
of knowledge is remarkably hard to find in the history of philosophy.
2. Chisholm has flirted off and on with attempts to reduce epistemic concepts to norma-
tive concepts ever since he first toyed with the idea in Chisholm (1957). Goldman (1979) and
(1986), Foley (1987), Sosa (1991), Kim (1988), Hookway (1994), Plantinga (1992) and (1993),
and Steup (1988), among many others, have all stressed the normative dimension of epistemic
concepts.
 Epistemic Justification and Normativity 59

3. Goldman himself became so sensitive to this objection that he eventually introduced a

second (nonreliabilist) conception of justification to accommodate it—see Goldman (1988).
See also Foley (1985) for a clear presentation of the objection.
4. We could even imagine a society odd enough that it tries to legislate over matters of
belief, thus creating legal obligations to believe and refrain from believing certain proposi-
tions.
5. See, for example, Alston (1988).
6. The emotivists Ayer and Stevenson also emphasize the "quasi-imperative" character
of moral judgments. See Ayer (1952) and Stevenson (1944).
7. My colleague Laird Addis, whose area of specialization is not epistemology, would
endorse the idea that epistemic judgments lack truth value, but as I say, he is surely the
exception.
8. One of the fundamental objections to objectivism first raised by Hume (1988) and de-
veloped by many others, perhaps most vigorously by Mackie (1977).
9. This is, of course, a crude statement of rule utilitarianism. There are all kinds of subtle
variations on the view designed to circumvent objections.
10. Foley (1987), 8.
11. For a detailed discussion of what constitutes actual consequence consequentialism
and what differentiates it from other versions of consequentialism, see Fumerton (1990).
12. Fumerton (1990), chap. 4.
13. We do sometimes seem to tie our evaluation of an agent's action to the actual conse-
quences of that action. The child playing catch in the living room who breaks the picture
window gets accused of a far greater wrongdoing than the child playing that same game of
catch who makes a luck stab at the ball deflecting it just before the window breaks. I'm in-
clined to think that appropriate philosophical reflection should lead one to reject an analysis
of wrongdoing that makes it dependent on actual consequences, but in the end I'm content to
argue that there are more interesting and fundamental concepts of what someone ought to do
that must take into account epistemic perspective.
14. A sense that is still distinct from our evaluation of the moral character of the agent.
15. For a detailed discussion of these important derivative concepts of morality and ra-
tionality, see, again, Fumerton (1990), chap. 4 and Foley (1990).
16. For a useful critical evaluation of possible conceptual connections between epistemic
evaluation and moral evaluation, see Alston (1988), Plantinga (1988), and Feldman (1988).
17. I owe special thanks to Matthias Steup. Through extensive e-mail correspondence,
he helped me better understand many of these issues.

References
Alston, William. 1988. The Deontological Conception of Epistemic Justification. Vol. 2 of
Philosophical perspectives, edited by James Toberlin. Atascadero, Calif.: Ridgeview.
Ayer, A. J. 1952. Language, Truth, and Logic. New York: Dover.
. 1956. The Problem of Knowledge. Edinburgh, Scotland: Penguin.
Butchvarov, Panayot. 1970. The Concept of Knowledge. Evanston, 111.: Northwestern Uni-
versity Press.
Chisholm, Roderick. 1957. Perceiving: A Philosophical Study. Ithaca, N.Y.: Cornell Uni-
versity Press.
Feldman, Richard. 1988. Epistemic Obligations. Vol. 2 of Philosophical Perspectives, edited
by James Tomberlin. Atascadero, Calif.: Ridgeview.
Foley, Richard. 1985. What's Wrong with Reliabilism? The Monist 68 (April): 188-202.
. 1987. The Theory of Epistemic Rationality. Cambridge, Mass.: Harvard University
Press.
. 1990. Fumerton's Puzzle. Journal of Philosophical Research 15: 109-13.
Fumerton, Richard. 1990. Reason and Morality. Ithaca, N.Y.: Cornell University Press.
Goldman, Alvin. 1979. What is Justified Belief? In Justification and Knowledge, edited by
George Pappas. Dordrecht, The Netherlands: Reidel.
60 Richard Fumerton

. 1986. Epistemology and Cognition. Cambridge, Mass.: Harvard University Press.

. 1988. Strong and Weak Justification. Vol. 2 of Philosophical Perspectives, edited
by James Toberlin. Atascadero, Calif.: Ridgeview.
Hookway, Christopher. 1994. Cognitive Virtues and Epistemic Evaluations, International
Journal of Philosophical Studies (2): 211—27.
Hume, David. 1888. A Treatise of Human Nature. Edited by L. A. Selby-Bigge. London:
Oxford University Press.
Kim, Jaegwon. 1988. What is "Naturalized Epistemology"? Vol. 2 of Philosophical Perspec-
tives, edited by James Tomberlin. Ataradero, Calif.: Ridgeview.
Mackie, J. L. 1977. Ethics: Inventing Right and Wrong. New York: Penguin.
Plantinga, Alvin. 1988. Positive Epistemic Status and Proper Function. Vol. 2 of Philosophi-
cal Perspectives, edited by James Toberlin. Atascadero, Calif.: Ridgeview.
. 1992. Justification in the 20th Century. Rationality in Epistemology. Vol. 2 of Philo-
sophical Issues, edited by Enrique Villanueva. Atascadero, Calif.: Ridgeview.
1993. Warrant: The Current Debate. New York: Oxford University Press.
Sosa, Ernest. 1991. Knowledge in Perspective. Cambridge: Cambridge University Press.
Steup, Matthias. 1988. The Deontic Conception of Epistemic Justification, Philosophical
Studies 58: 65-84.
Stevenson, C. L. 1944. Ethics and Language. New Haven: Yale University Press.
Part II

EPISTEMIC DEONTOLOGY AND

DOXASTIC VOLUNTARISM
This page intentionally left blank
 4

Deciding to Believe

CARL GINET

Sometimes we judge that a person was not justified in coming to have a particular
belief. We think that, given what she was aware of at the time, she ought not to have
adopted the belief. For example: "You ought not to have leaped to the conclusion
that he was bored just because he looked at his watch." "You ought not to have con-
cluded she was still at home just because her phone was busy." "The jurors ought
not to have believed his mother's statement (that at the time of the murder he was
with her two miles away from the scene of the crime), since they knew she had a
strong motive to lie about the matter."
It seems that in making such judgments we presuppose that the person could in
the circumstances have not come to have the belief in question, that she chose or
decided to believe that which we fault her for believing. Or else (if we are unsure of,
or disbelieve, this presupposition, as is often the case) our judgment is tacitly condi-
tional: We mean that, if she could have decided otherwise, then she ought not to have
decided as she did. If this is right, then our judgment makes sense only if it makes
sense to suppose that a person might come to believe something simply by deciding
to do so.
Some, however, have thought that this is ruled out by the very concept of what it
is to believe a truth-valued proposition.1 Others have thought that, though it is con-
ceptually possible, deciding to believe is never psychologically possible and, if it
did occur, it would be quite irrational.2 Against these views, I wish to defend the naive
intuition that coming to believe something just by deciding to do so is possible, that
it sometimes seems to us that we do this, and that our doing so need not offend against
epistemic reason. My hope is to make it plausible that there is a sort of state that counts
as a state of believing a proposition, which state is such that it is clear that one could

63
64 Carl Ginet

come to be in such a state simply by deciding to do so, and clear, moreover, that such
decisions can (though they need not) be perfectly rational and motivated entirely by
one's appraisal of the available evidence and one's general desire that one's beliefs
be true. (Later I will consider the reasons some have offered for denying that decid-
ing to believe is conceptually possible or for denying that it is psychologically
possible.)

Let me start by giving examples of the sorts of cases that seem to me good candi-
dates for being described as someone's deciding to believe something. Then I will
indicate what it is in these cases that (as it seems to me) both constitutes the subject's
believing a proposition and is such that it came into existence directly by the subject's
deciding that it would.

1. Sam is on a jury deliberating whether to find the defendant guilty as charged; if cer-
tain statements of a certain witness in the trial are true, then the defendant cannot have
done what he is charged with; Sam deliberates whether to believe those statements,
to believe the prosecutor's insinuations that the witness lied, or to withhold belief on
the matter altogether. He decides to believe the witness and votes to acquit.
2. Sue is in a poker game of seven-card stud. After all cards have been dealt, everyone
folds except Sue and Hank. Sue has three aces showing and two kings in the hole.
The fourth ace was seen in an already folded hand. Hank has three jacks showing. If
he has a fourth jack in the hole, he beats Sue's full house. Hank raises by a fairly
substantial sum. Sue asks herself, "Shall I see him or fold? Does he have a Jack in the
hole or is he bluffing? I detect a certain subtle nervousness in Hank's manner. Of
course, he could be pretending to be worried so as to lure me into betting, but I've
played with him a number of times and don't recall his having tried that sort of pre-
tence before, and he is generally such an open book that I doubt very much he could
pull it off." So she decides to dismiss that possibility, to read Hank as being anxious
about bluffing; she decides to believe that he does not have a jack in the hole and
meets his raise while beginning to think about what she will do with the winnings.
She is greatly surprised when Hank reveals that he does have the fourth jack.
3. Before Sam left for his office this morning, Sue asked him to bring from his office,
when he comes back, a particular book that she needs to use in preparing for her
lecture the next day. Later Sue wonders whether Sam will remember to bring the
book. She recalls that he has sometimes, though not often, forgotten such things,
but, given the inconvenience of getting in touch with him and interrupting his work
and the thought that her continuing to wonder whether he'll remember it will make
her anxious all day, she decides to stop fretting and believe that he will remember
to bring it.
4. We have started on a trip by car, and 50 miles from home my wife asks me if I
locked the front door. I seem to remember that I did, but I don't have a clear, de-
tailed, confident memory impression of locking that door (and I am aware that my
unclear, unconfident memory impressions have sometimes been mistaken). But,
given the great inconvenience of turning back to make sure and the undesirability
of worrying about it while continuing on, I decide to continue on and believe that
I did lock it.
 Deciding to Believe 65

In these examples, the subject S decided to believe a certain proposition p. S did

this in deciding to act, or not to act, in a certain way. In deciding to vote for acquittal,
Sam decided to believe the statement of the witness. In deciding to meet Hank's raise,
Sue decides to believe that Hank is bluffing. In deciding not to remind Sam to bring
the book she needed, Sally decided to believe that he would remember to bring it. In
deciding to continue on down the road without worrying about it, I decided to be-
lieve that I'd locked the door. What makes true a statement of the form "In deciding
to A, S decided to believe that/?"?
One thing required is surely this: In deciding to A, S staked something on its
being the case that p. What is it to do that? I suggest the following explication:

In deciding to A, S staked something on its being the case that p iff when de-
ciding to A, S believed that A-ing was (all things considered) at least as good
as other options open to her iff p (equivalently: that no other option open to
her was preferable to A-ing iff p; or, for short, that A-ing was optimal iff p).

On this definition, staking something on its being the case that/? is not sufficient for
believing that/?: The definiens says only that S had a certain belief about A-ing when
she decided to A, and the content of that belief does not entail that p.3 This is as it
should be, of course. When I bet on a coin-flip landing heads, I stake something on
the coin's landing heads; and in doing this, though I may hope that it will land heads,
I need not believe that it will. What more would I need to do to believe this? I would,
I think, need to count on its being the case that the coin will land heads.
To count on its being the case that p is, in addition to staking something on p, to
adopt a dismissive or complacent attitude toward the possibility of losing what one
has staked on p because of its turning out that not-/), an attitude of a sort that the
mere gambler on p does not adopt. It is to not prepare oneself for the possibility of
not-/?. (Obviously, whether one prepares oneself for the possibility of losing what
one has staked on p should it turn out that not-/? is a question that arises only if one
has staked something on/?.) If I merely staked something on the coin's landing heads
but did not count on it, then I was prepared for the possibility of its not landing heads
(whether or not I was concerned or anxious about that possibility, which would de-
pend on how much I valued what I staked on its not happening). But if I counted on
its landing heads (perhaps I thought the coin was biased, or perhaps T committed the
gambler's fallacy after a string of tails), then I did not prepare myself for that possi-
bility, at any rate not as much as I otherwise would have. To count on p is to stake
something on p with this sort of dismissive or unconcerned or unready attitude to-
ward the possibility of not-/?.
In example 1, Sam might, of course, have still decided to vote for acquittal, even
if he had not decided to believe the witness's exonerating testimony, on grounds
that the fact that there was such testimony gives him some reason to withhold be-
lief in the guilt of the accused; but in doing that, he would have had to prepare him-
self for the possibility that the exonerating testimony was false—for example, to take
the attitude that he will have no regrets should it turn out that the witness was lying and
the accused was guilty—whereas in deciding to count on the truth of the witness's
testimony, he chose not to take up any such protective attitude toward the possibility
66 Carl Ginet

of its falsity. In example 2, Sue might not have decided to believe that Hank was
bluffing—she might have withheld belief on that question—and yet still decided to
risk a further bet; if she had done that, she would have been more ready than she was
to find that Hank was not bluffing. In example 3, Sue might have decided not to re-
mind Sam about the book but also not to count on his remembering it and to think
about what to do should he forget it. In example 4,1 might have decided to continue
on without believing that I locked the door (only hoping that I did), but that would
mean that I would continue to worry about the possibility that I did not lock it or at
least be prepared to find it unlocked when I returned.
To not prepare oneself for dealing with the possibility of not-/? is to not think about
the possibility of not-/? or at least not to give any consideration of what to do if not-
p. And in fact, since to not prepare for the possibility of not-/? is itself to stake some-
thing on p (assuming that one believes that not preparing for the possibility of not-/?
is optimal iff/?), choosing to not prepare oneself for the possibility of not-/?, resisting
an impulse to do so, can be the choice such that it is in making it that one decides to
count on p.
In the right circumstances, it can take effort to avoid preparing oneself for the
possibility that not-/?. Such efforts, to suppress considering that possibility and what
to do if it is realized, are what make it apt sometimes to describe oneself as making
oneself believe (or trying to make oneself believe) something. S receives a telephone
call from the police saying that his wife has been involved in a car accident and that
she wishes him to come to the scene. The police caller says that she believes there
are no serious injuries. While S hastens to the scene, S is, as it seems natural to put it,
making himself believe that his wife has not been seriously injured. Part of what S
does that seems to deserve that description is to suppress all impulse to imagine what
his wife's injuries might be or to consider how he will handle it if he does find her
seriously injured. (Another part is his repeatedly reminding himself of the evidence
he has that she had not been seriously injured, namely, that the police caller would
very likely have known and told him if there had been serious injuries.)
Counting on/? typically leads to feeling surprise on learning that not-/?. It is a symp-
tom of Sue's having not prepared herself for the possibility that Hank was not bluffing
that she felt surprise on learning that he was not. But it is implausible to suppose that
not preparing oneself for the possibility of not-/? entails that one will have & feeling of
surprise if it turns out that not-/?. A very experienced and cool poker player might suf-
fer no emotional reaction at all at having such a belief falsified. Of course, there is a
sense of "S was surprised that not-/?," which means nothing more than that S learned
that not-/? when she had believed that/? and does not imply any feeling or emotional
reaction on S's part to learning that not-/?. Counting on/? does entail being surprised in
that sense should one learn that not-/), since it entails believing that /?.
If counting on /? is what F ve said it is, then it is possible in the right circumstances
to decide to count on /? and thus to decide to believe that /?. Of course, it can also
happen, and very frequently does, that one comes to count on (believe) something
without having decided to do so. Often one's perceptual experience or memory or
beliefs about the evidence simply compel one's counting on/?. But where, as occa-
sionally happens, such things do not compel belief that p but do provide some evi-
dence of the truth of/?, the opportunity may open for the subject to decide whether to
 Deciding to Believe 67

believe p. (In such cases, the subject's decision may be influenced by reasons she
has for wanting it to be the case that/?, as is illustrated in examples 3 and 4 above and
of which I'll say more later.)
So the explication of "In deciding to A, S decided to believe that p" that I am
suggesting is the following:

In deciding to A, S decided to believe that p iff in deciding to A, S decided to

count on its being the case that p.

where the notion of counting on its being the case that p is understood, as I've
explained.

II

Belief is commonly thought to be a dispositional state. How do dispositions get into

the picture in cases of deciding to believe like those illustrated in my examples? Well,
in such cases, the subject S, in deciding to believe that/?, simultaneously adopts and
manifests the following disposition: To count on/? in deciding to act in a certain way
when presented with an opportunity to do so relevantly like the one she was pre-
sented with.4 (What is meant by "relevantly like" is that S has a similar menu of options
and S's other beliefs, desires, preferences, evaluations, and so on that bear on the
question of which option will seem to her the best one to choose are similar.) What
comes into being as a result of S's deciding in a particular situation to count on p is
a disposition to count on p in closely similar situations. That disposition is a dispo-
sitional belief and S decided to adopt the disposition in the very decision that first
manifested it. (If one already has such a dispositional belief that p, then when one
encounters a relevantly similar choice-situation and chooses to act in a way that is
optimal iff p, one does not therein decide anew to believe that p; rather, one already
regards the action's being optimal iff p as a reason for choosing it.)
The notion of coming to have a disposition (to act in certain ways in certain sorts
of situations, perhaps with a certain sort of attitude) by deciding to have it should not
be problematic. This is something we often do, in deciding to have a standing policy
for action, in adopting a general conditional intention to act in such-and-such a way
in such-and-such circumstances.
The dispositional belief that one both manifests and decides to have, in deciding
on a particular occasion to count on/?, could be very short-lived; but normally it will
last awhile, without having to be readopted by a new decision or brought about anew
in some other way, though eventually something may happen to change S's mind
(typically by providing her with a reason to decide to change it), or S may lose the
belief through forgetting it.
It is possible, of course, to decide to adopt a belief-disposition, to decide to count
on the truth of a certain proposition when in a certain sort of situation, in the absence
of that sort situation, that is, without at the same time manifesting the disposition.
That would be a case of deciding to believe that p that would not be a case of decid-
ing to believe that p in deciding to act in a certain way.
68 Carl Ginet

The disposition we've talked about so far is rather specific: to count onp in choos-
ing a particular sort of action in a particular sort of situation. And it could be that the
subject never has any more general disposition to count on p. Typically, though, a
disposition acquired in certain specific circumstances will generalize at least to some
extent—it will be a disposition to count on p in choosing other sorts of actions in
other sorts of situations.
If belief that/? is the sort of disposition I've been talking about, then it may ex-
hibit a certain sort of context-relativity. Thus, for example, Sue might be disposed to
count on its being the case (i.e., believe) that Hank is bluffing when the issue is whether
to meet his raise of $20, but not disposed to count on it when the issue is whether to
bet the farm. I might be disposed to count on its being the case that the temperature
won't go below freezing during the next week, when the issue is whether to leave all
those potted plants outside while we're away, but not disposed to count on it, when
the issue is whether to leave the furnace on (with the thermostat turned down). I might
be disposed to count on its being the case that this large dog won't bite someone
approaching it closely when the issue is whether to assert that it won't (and thereby
stake my reputation for veracity on that's being the case), but not disposed to count
on the dog's not biting when the issue is whether to approach it closely myself.
That is, one might be disposed to count on p in choosing some sorts of actions in
some sorts of situations but not disposed to count on p in choosing other sorts of
actions in those or other sorts of situations where the value staked onp in so choos-
ing would be different and greater. This is to believe that;? relative to some contexts
and not to believe it relative to other contexts. We should not insist that believing
that/? requires being disposed to count on p in every sort of situation. There is, to be
sure, something we can call "acceptance" that is not belief. For example, it could
have been that Sue merely staked something on the proposition that Hank is bluffing
without counting on it: She merely took that proposition as her "betting hypothesis."
But if the subject does not merely stake something on p but also counts on p in mak-
ing the decision she does, then she therein manifests belief that/?.
In the examples given in the paragraph before last, what was at stake for the sub-
ject in the context where the subject was disposed to count on p was less than what
was at stake in the context where the subject was not disposed to count on p. This
permits us to say that the subject believes that/? only at a certain level of confidence
(only to a certain degree) and not at a higher level (not to a greater degree), provided
that the same relation holds between every pair of contexts such that the subject is
disposed to count onp in one and not disposed to count onp in the other. And it may
be that only when that relation holds is it fully rational to count on p relative to one
context and not to count on it relative to another. But it surely can happen that a sub-
ject is disposed to count on p in one context but not count on it in another where
what would be at stake in counting on it in the latter is not greater than what would
be at stake in counting on it in the former.
George, let us suppose, is disposed to count on the rickety old footbridge's hold-
ing the weight of a large man if the issue were whether to cross the bridge himself
(when he wants to get to the other side of the ravine in a hurry) but not disposed to
count on it if the issue were whether to agree with his father's assertion that it will
hold such a weight while discussing whether his father should have the bridge re-
 Deciding to Believe 69

built (where no one's bodily safety would be staked on his act of asserting agree-
ment). In having the first disposition while lacking the second, George may be less
than fully rational, but that is not a good reason to deny that George does believe the
bridge will hold his weight relative to the first context and does not believe it rela-
tive to the second context. Here we can't say that George believes the proposition to
a certain degree and not to any greater degree. We can say only that he believes it
relative to one context and not relative to another or that "in a way" he believes it
and in a way he doesn't.5'6
L. J. Cohen has offered the following account of belief:

belief that p is a disposition, when one is attending to issues raised, or items referred
to, by the proposition that p, normally to feel it true that p and false that non-p, whether
or not one is willing to act, speak, or reason accordingly. . . . The standard way to dis-
cover whether you yourself believe that p is by introspecting whether you are normally
disposed to feel that p when you consider the issue.7

This, if I understand it right, entails that belief is not relative to context, to types of
choice situation, in the way I've described. But I find it implausible to suppose that
one has a special feeling, literally so called, toward a proposition one believes when-
ever one considers it. (The attitude toward p of counting on it, of not taking seri-
ously the possibility that not-;?, on which I place so much weight, is not a feeling,
and it is possible only when one has staked something on p.)
Cohen's theory seems prompted, at least partly, by the fact that we are quite good
at distinguishing among the propositions we entertain between those we currently
believe and those we do not currently believe, plus the thought that this cannot be
explained if belief is the kind of disposition to action that a pragmatic account says
it is. But we need not accept this latter thought: Insofar as we have direct knowledge
of what our beliefs are, when we are not manifesting them, it is of the same sort as
we have of our conditional intentions. I have a kind of privileged access to my own
conditional intentions—to the fact that, unless my intentions were to change, I would
act in such-and-such ways in such-and-such circumstances. I have the same kind of
privileged access to at least some of my beliefs, to at least some of the facts about me
of the form "Unless my beliefs were to change, I would count on/? in such-and-such
circumstances."
Of course, we are not infallible about facts of this sort about ourselves. I might be
aware of the fact that I would count on this dog's being harmless in circumstances
where, by asserting that this dog is harmless, I would be staking my testimonial ve-
racity on its being harmless, but unaware of the fact that I would not count on its
being harmless in other circumstances where, by approaching the dog closely, I would
be staking my bodily safety on its being harmless.
Though one can come to have a belief, a disposition to count on the truth of a
proposition, by deciding to have it, one can, of course, and often does, come to have
such a disposition in other ways, involuntarily. For instance, in the normal case, one's
perceptual beliefs—one's beliefs as to what one is now seeing, hearing, feeling, and
so on (at least insofar as the content has to do with the more superficial, perceivable
features of what one perceives)—will be involuntary. When I see a tree or a dog before
me, a car coming down the street, snow on the ground, and the like, and as a result
70 Carl Ginet

come to believe that I see such a thing, I do not (usually) experience the coming to
be of that belief as something I have a choice about and decide to make happen. There
could, however, be an atypical situation where I have some reason, though not a
conclusive reason, to think that my senses might currently be deceiving me; then I
might be in a position to decide whether to believe what my senses deliver.
Similarly, most of my beliefs based on memory, where I believe something be-
cause it seems to me I remember witnessing it or learning it, are not voluntary. Un-
like my example where I decide to trust my memory impression that I did lock the
door, I do not decide to trust my memory impression that I ate breakfast at home this
morning, that I had lunch with S yesterday, that I have been in New Zealand, and so
on. In most cases where I believe because I remember, my memory does not leave
me free to decide whether to trust it. It is fairly clear that in the large mass of beliefs
held by any normal person at a given time, the overwhelmingly major part will have
come about involuntarily and only a small portion will have been adopted voluntar-
ily (by decision). Coming to believe by deciding to believe (or seeming to do so) is
undoubtedly a rare phenomenon in that sense. But it is nevertheless a phenomenon
that we are familiar with—it may happen every day in the lives of some of us—and
that is why we find it intelligible, when judging whether a particular belief was jus-
tified, to think of ourselves as judging whether the believer ought to have chosen to
adopt the belief or ought to have done so if she could have done otherwise.
I do not wish to claim that every dispositional belief must consist of the sort of
disposition that I have specified. I claim only that having such a disposition with
respect to a certain proposition suffices for believing that proposition. There may be
phenomena that we are happy to call beliefs that do not fit this analysis. For example,
there can be belief in a proposition p such that there could be for the believer no such
thing as its turning out that not-/? (no such thing, that is, as her coming to have conclu-
sive or overwhelming evidence that not-p) and thus no such thing as her preparing
herself for that possibility or her staking something on its being the case that p. That
there is life elsewhere in our galaxy might, for instance, be such a proposition. (I'm
inclined to think that such propositions can be believed only in an attenuated sense.)

Ill

Where believing that p is the sort of disposition I have been talking about, a reason
for believing that/? will be a reason for being disposed to act in such a way that in so
acting, one counts on its being the case that p. The reasons one can have for doing
this fall into two mutually exclusive classes: the interested and the disinterested. An
interested reason is a reason for wanting it to be the case that/?. Thus I had interested
reasons for believing that I locked the door, and S for believing that his wife was not
seriously injured in the car accident. A disinterested reason is one that is not inter-
ested. Disinterested reasons will include having evidence that/? or having percep-
tual experience whose content entails that/? (while having no reason to mistrust one's
senses in that instance).
How can there be an interested reason for believing that/?? How can wanting it to
be the case that p be a reason for believing that /?? Can wanting it to be the case that
 Deciding to Believe 71

p be a reason for staking something on pi Clearly not when one believes that not-/?.
But suppose one has no belief either way. Then I think a desire that p can motivate,
further, one's staking something on p: It can motivate one's not preparing oneself
for the possibility of not-p, that is, one's counting on p. (To not prepare oneself for
the possibility of not-/?, to count on p, can itself be to stake something on p; that is,
it can be that one's counting on p is optimal iff/?.) For example, suppose that S's
husband was scheduled to be on a flight that, S has learned, has crashed; but he tele-
phoned S an hour before the flight was to leave to say that he thought he would prob-
ably not make that flight and would come on a later one. S's desire that her husband
was not on the ill-fated flight is about as strong as any desire she's ever had. In the
circumstances, it would not be surprising if, because of it, she began to count on its
being the case that he was not on that flight, began to believe that he was not; and her
doing so would not even be, all things considered, irrational.
Or suppose a subject S has decided to stake on the truth of a certain proposition p
something other than not being prepared for the possibility of not-/? (e.g., a large sum
of money). Then S's having done this gives S a reason for wanting it to be the case
that/?, and it is intelligible that this desire might motivate S to take the further step
involved in counting on (and thus believing) p, viz., to not prepare herself for the
possibility of not-/?, to not contemplate the costs should it turn out that not-/?. One
would expect this desire actually to lead a normally rational person to count on p
only when the stake is rather significant and the person has some evidence that/7 and
no compelling evidence that not-/?. Suppose, for example, that S has a life-threaten-
ing disease. There are two treatments available, for each of which there is some rea-
son to think it might bring about a cure in S's case: Each has worked in some cases
and failed to work in some. But the treatments are mutually exclusive: Pursuing ei-
ther makes it impossible to pursue the other. S decides to pursue treatment A. Now
S has reason for wanting it to be the case that treatment A is the better cure for her
case or at least that treatment A is as likely to cure her as treatment B would have
been. It would not be surprising if this desire led S to believe this, to count on its
being so, to avoid the psychic cost of taking seriously the possibility that it is not so.
Coming to believe that p by deciding to do so seems more psychologically pos-
sible, and less irrational (all things considered), when the subject has some evidence
that p (but it is not compelling evidence, or she also has some noncompelling evi-
dence that not /?) and also has significant reason for wanting it to be the case that /?.

IV

But some philosophers have thought that deciding to believe is not so much as con-
ceptually possible, let alone psychologically possible. According to Bernard Will-
iams, "it is not a contingent fact that I cannot bring it about, just like that, that I be-
lieve something . . ." He offers two reasons for this claim:
One reason is connected with the characteristic of beliefs that they aim at truth. If I
could acquire a belief at will, I could acquire it whether it was true or not; moreover I
would know that I could acquire it whether it was true or not. If in full consciousness
I could will to acquire a 'belief irrespective of its truth, it is unclear that before the
72 Carl Ginet

event I could seriously think of it as a belief, i.e., as something purporting to represent

reality. At the very least, there must be a restriction on what is the case after the event;
since I could not then, in full consciousness, regard this as a belief of mine, i.e., some-
thing I take to be true, and also know that I acquired it at will. (1973, p. 148)
Williams says that it is a characteristic of beliefs, presumably a characteristic that
is essential to their nature, that they aim at truth. And he asserts that, if I could ac-
quire a belief at will, I could acquire it whether it was true or not, which would be
contrary to the nature of belief. What does Williams mean by "I could acquire it
whether it was true or not"? (He can't mean that I could acquire a belief that is not
true. Entailing that possibility would be nothing against acquiring beliefs at will, since
we do actually acquire beliefs that are not true.) What does he mean by "beliefs aim
at truth"? One might naturally take this to mean that, necessarily, one wants one's
beliefs to be true, and thus "I could acquire it whether it was true or not" would mean
that I could acquire it without caring at all whether it was true or not, without having
any desire that it be true. If its consequent is construed this way, then there is no rea-
son to accept Williams' conditional, "if I could acquire a belief at will, I could ac-
quire it whether it was true or not."
- Suppose it is conceptually true that one cannot come to believe something while
lacking any desire that one's belief be true. This is quite compatible with its being the
case that one's reason for deciding to believe that;? is a desire that;? be true—a desire
had independently of having evidence that/) is true. Suppose that S has some evidence
that her husband was on the plane that recently crashed and some evidence that he was
not; but instead of withholding belief, S believes that her husband was not on the plane,
because she wants it very much to be true that he wasn't. This desire is no sort of evi-
dence that he wasn't on the plane, but it does have to do with the truth of that proposi-
tion. It is precisely because she does want so much that the proposition be true that she
is motivated to believe it. So it is obviously not the case that deciding to believe, for an
interested reason, entails not caring whether one's belief is true or not.
There is another way of construing "beliefs aim at truth." One might say that one
cannot believe that/? unless one's counting on;? is motivated by a desire to have true
beliefs (a desire to count on a proposition's being true only when it is true) that is
independent of any desire one has regarding any particular proposition that it be true,8
or, in other words, unless one has disinterested reason for believing it. This is per-
haps the most plausible construal of Williams' meaning and of the meaning of other
writers who have invoked this dictum about belief. (Given that we must have a de-
sire for true beliefs in general that is independent of the content of any particular
belief, there can arise cases where one has a belief that, qua belief, one of course
wants to be true, but that, qua having the particular content it does, one wants to be
false. If the wife in our recent example, after acquiring compelling new evidence,
came to believe that her husband was on the plane that crashed, then she would be in
such a position.)
If "beliefs aim at truth" is construed in this way, then, although I see no reason to
think that it captures a conceptual truth about belief as such, it may well express a
conceptual truth about rational belief. And we may grant further that it is psycho-
logically impossible to believe a proposition, to count on its truth, without having
some disinterested reason for doing so. But from its being psychologically impos-
 Deciding to Believe 73

sible to believe without disinterested reasons, it does not follow that it is psychologi-
cally impossible to decide to believe for such reasons.
And from its being conceptually impossible to believe rationally without disin-
terested reasons, it does not follow that it is conceptually impossible to decide to
believe for such reasons. It may be that cases where some of S's reasons for decid-
ing to believe that/? are interested ones should be said to exhibit a degree of epistemic
irrationality, by which we would mean simply that some of S's reasons are non-
evidential, are not such as to increase the probability that/) is true. But cases of de-
ciding to believe need not all be cases where interested reasons are involved; some
cases where none are involved might be cases where the decision to believe that/7 is,
even epistemically, fully rational, where the subject's disinterested reasons do jus-
tify counting on/7 in the context. In example 1 above, Sam's disinterested reasons
might be good enough to justify his counting on the truth of the exonerating witness's
testimony, even though they do not compel him to do so: Whether having certain
evidence compels a belief is not the criterion of whether having that evidence justi-
fies the belief.
Another reason Williams offers (1973, pp. 148-49), for his claim that it is non-
contingently true that I cannot come to believe just by deciding to do so,

stems from our considerations about perceptual belief: a very central idea with regard
to empirical belief is that of coming to believe that p because it is so, that is, the rela-
tion between a man's perceptual environment, his perceptions, and the beliefs that re-
sult. Unless a concept satisfies the demands of that notion, namely, that we can under-
stand the idea that he comes to believe that p because it is so and because his perceptual
organs are working, it will not be the concept of empirical belief . . . But a state that
could be produced at will would not satisfy these demands, because there would be no
regular connexion between the environment, the perceptions and what the man came
out with, which is a necessary condition of a belief. . . .

Williams here seems to claim that, from the proposition that S's belief that p is a
correct perceptual belief, it follows that S's belief was caused by the fact that/7 in a
way that is incompatible with the supposition that S came to believe that p by decid-
ing to believe it. I am unable to understand how this is so.
Suppose that S's eyes are directed toward a square red patch on a white wall. The
light is good and S's visual system is in good working order, so she is caused to have
a visual experience as if seeing a square red patch on a white surface, and in fact she
sees a square red patch on a white surface. Suppose, however, that she hesitates for
a moment to believe that she sees such a thing because she has some slight reason to
think that she might be hallucinating the red patch; but then she decides to cast doubt
aside and trust her vision: She decides to believe that (R) there is before her a red
patch on a white surface.
Here we can certainly say that she came to her belief that R because it was a fact
that R and her vision was in good working order: Had these things not been the case,
she would not have had the visual experience, she had, and had she not had that ex-
perience she would not have believed that R. We cannot, it is true, say that the fact
that R together with the fact that her vision was working properly causally necessi-
tated her believing that R. But we should not, in any case, want to say that. Some-
74 Carl Ginet

times people do mistrust their senses when they are in fact working properly. So I
see no reason to accept Williams' claim that, if a person with properly working per-
ceptual organs perceives a certain external scene and as a result has a perceptual belief
that corresponds to her perceptual experience, then it could not be that she decided
whether to adopt that belief.9

William P. Alston (1988) attacks the thesis that it is psychologically possible for one
to "take up at will whatever propositional attitude one chooses" (p. 122). This is not
a thesis I have asserted or wish to assert. Obviously, there are a great many proposi-
tions I do not now believe that I cannot come to believe just by deciding to believe
them. For example, it is not in my power now to decide to believe that I am now
playing squash or that I'm now immersed in water. Obviously, there are also a great
many propositions I now believe that I cannot cease to believe just by deciding to do
so. For example, I cannot now decide not to believe that the current year is 1999 or
that I was born in Wyoming. (But note that it is also not in my power now to acquire,
just by deciding, the intention to run over pedestrians with my car at the next oppor-
tunity and not in my power to abandon my intention always to avoid driving my car
into the path of onrushing vehicles; and intention is unquestionably a sort of mental
state that a subject sometimes comes to have just by deciding to do so.)
But in the course of arguing against this obviously false thesis, Alston makes plain
that he also rejects a weaker thesis that I do wish to assert, namely, that it is psycho-
logically possible, in the right circumstances, for a subject to come to believe some-
thing just by deciding to believe it, where the subject has it open to her also to not
come to believe it. He is convinced that, in any actual case that anyone might be
tempted to describe by saying that a person came to believe that p just by deciding to
do so, what really happened must be something else. It must be either that its seem-
ing to the subject that it is highly likely that/) compelled the subject's belief (leaving
her no choice in the matter) or that what the subject decided to do was, not to believe
that p, but to proceed on the assumption that p.
It is certainly true that there are various cases properly described as deciding to
proceed on the assumption that/? that are not cases of deciding to believe that/?. One
of them is the case of staking something on its being the case that/) without counting
on it (in my sense)—as in ordinary gambling or in Alston's example (1998, p. 126)
of the military commander who says, "I don't know what the disposition of enemy
forces is; I don't even have enough information to make an educated guess. But I
have to proceed on some basis or other, so I'll just assume that it is H and make my
plans accordingly." Another is the case of assuming that p as a hypothesis for the
purpose of testing it or of reducing it to absurdity, where one does not even stake
anything on the truth of the hypothesis.
But the cases I want to describe as deciding to believe are not of those sorts. In
deciding to believe the exonerating witness's testimony and vote for acquittal, Sam
decided not merely to proceed on the assumption that the witness's testimony was
 Deciding to Believe 75

true, but to count on its truth. When the question arose as to whether I locked the
door, I might have decided merely to stake something on the hypothesis that I did,
merely to proceed on that assumption, but I decided to do more, to believe that I locked
it, to adopt an attitude toward that proposition that is not implied by my merely stak-
ing something on its truth.
If they are not cases of merely proceeding on the assumption that p, must they
then be cases of the other sort Alston allows, where the subject did come to believe
that p but, despite the subject's impression of deciding to do so, she actually had no
choice about the matter and was compelled to believe (by, say, her probability esti-
mate)? I suppose that it is possible that I and others are always victims of a kind of
illusion about what is going on in our minds here—I do not want to claim that would
be conceptually or psychologically impossible—but we have not been offered any
good reason to think it is so. Until we are offered such a reason, we are entitled to
continue taking our occasional impressions, that we come to believe something just
by deciding to do so, at face value. And that means, concerning the concept of
epistemic justification, that we can continue, in good conscience, to understand it in
a deontic way: We can continue to take the judgment that a belief is not justified as
implying that the subject ought not to have adopted the belief or ought not to have
done so if she could have avoided it.10

Notes
1. See, for instance, Hampshire (1971) and Williams (1973).
2. See Alston (1988).
3. But the definiens, in ascribing the belief that A-ing is optimal iff p, does entail that the
subject understands the proposition that p.
4. The idea that belief is a disposition to act in certain ways is commonly labeled the
"pragmatic" account of belief. See, for example, Braithwaite (1932) and Stalnaker (1984).
5. I think that Michael Bratman is not disagreeing with what I say here when he says:
Reasonable belief is, in an important way, context independent: at any one time a rea-
sonable agent normally either believes something (to degree n) or does not believe it
(to that degree). She does not at the same time believe that p relative to one context but
not relative to another. (Bratman 1992, 3)
The last sentence here does not disagree with what I've said if it means that a fully rational
person does not at the same time believe that p to a certain degree relative to one context but
does not believe it to that same degree relative to another.
6. Robert Stalnaker (1984) holds that belief is a species of a broader genus he calls accep-
tance and suggests that context-relativity is unproblematic only for acceptance that is not belief.
He says (p. 81), "A person may accept something in one context, while rejecting it or sus-
pending judgment in another. There need be no conflict that must be resolved when the dif-
ference is noticed, and he need not change his mind when he moves from one context to the
other. But something is wrong if I have separate incompatible sets of beliefs for different
circumstances." This last assertion seems right only if "incompatible beliefs" means (as I trust
Stalnaker does mean) beliefs with incompatible contents. I don't see that there is anything
wrong or irrational about believing something relative to one context and not believing it (as
distinct from believing it false) relative to another.
7. Cohen (1992), 4.
8. Something like this construal of "beliefs aim at truth" is suggested by David Velleman
in "On the Aim of Belief" (unpublished ms.).
76 Carl Ginet

9. For a very different, but very interesting, critique of Williams's argument, see Bennett
(1990).
10. I am grateful to Matthias Steup for helpful comments on an earlier version of this
paper.

References
Alston, William P. 1988. The Deontological Conception of Justification. Vol. 2 of Philosophi-
cal Perspectives, edited by James Tomberlin. Atascadero, Calif.: Ridgeview.
. 1989. Epistemic Justification. Ithaca, N.Y.: Cornell University Press.
Bennett, Jonathan. 1990. "Why Is Belief Involuntary?," Analysis 50: 87-107.
Braithwaite, R. B. 1932. "The Nature of Believing," Proceedings of the Aristotelian Society
33: 129-46.
Bratman, Michael. 1992. "Practical Reasoning and Acceptance in a Context." Mind (101):
1-15.
Cohen, L. Jonathan. 1992. An Essay on Belief and Acceptance. Oxford, U.K. Clarendon.
Griffiths, A. P., ed. 1967. Knowledge and Belief. Oxford: Oxford University Press.
Hampshire, Stuart. 1971. "Freedom of Mind." In Freedom of Mind and Other Essays. Princeton,
N.J.: Princeton University Press.
Kiefer, H. and Munitz, M., eds. 1970. Language, Belief, and Metaphysics. Albany, New York:
State University of New York Press.
Stalnaker, Robert. 1984. Inquiry. Cambridge, Mass.: MIT Press.
Williams, Bernard. 1970. "Deciding to Believe," in Language, Belief, and Metaphysics,
edited by H. Kiefer and M. Munitz.
. 1973. Problems of the Self. Cambridge: Cambridge University Press.
 5

Voluntary Belief and

Epistemic Evaluation

RICHARD FELDMAN

I. Introduction

Our talk about epistemic matters parallels our talk about ethical matters in noteworthy
ways. Among the ethical judgments we make are judgments that a person ought to
perform a certain action, that someone should not do a certain thing, that people have
obligations to act in some ways, that they are permitted or required to do certain things,
that they have a right to do one thing and a duty to do another, and that sometimes they
deserve praise or blame for what they have done. We make seemingly analogous judg-
ments about beliefs and believers. We say that a typical well-informed contemporary
American ought to believe that the Earth revolves around the Sun and should not be-
lieve that the Earth is flat. A person wrongly accused of a crime might say his accusers
have no right to believe that he's guilty, since no evidence of his wrongdoing has been
brought forth. In such a case, we might say that believing the person is not guilty is
permitted or perhaps even required. We sometimes praise those who believe the things
they should, and we criticize those who fail in their believings. We can describe all
these judgments as deontological judgments about beliefs.
Much work in recent epistemology has been about epistemic justification. As
William Alston says, it is "natural" to understand epistemic justification in a "deonto-
logical" way. By this he means that it is to be understood in terms of "obligation,
permission, requirement, blame, and the like."1 Apparent examples of deontological
analyses include one derived from A. J. Ayer' s proposal that one has knowledge when
one has a true belief and one has a "right to be sure."2 On this view, one is justified
in believing a proposition when one has a right to be sure that it is true. Carl Ginet
has proposed that

77
78 Richard Feldman

One is justified in being confident thatp if and only if it is not the case that one ought
not be confident thatp; one could not be justly reproached for being confident that p.3
Roderick Chisholm appeals to the notion of an "intellectual requirement" in an ef-
fort to clarify the key concepts of epistemological theory.4 Analyses of epistemic
justification in deontological terms are common.5
Recently, Alvin Plantinga and Alvin Goldman have independently argued that the
viability of a deontological conception of epistemic justification is crucial to the debate
between internalists and externalists about epistemic justification. Goldman thinks
that a central, but mistaken, line of support for internalist theories begins with the
assumption of a deontological account of justification.6 Plantinga also argues that
internalism derives much of its support from a deontological view of justification.7
Whether deontological judgments about beliefs are ever true thus is of considerable
epistemological significance.
A central problem that both Plantinga and Alston find with deontological judg-
ments about beliefs is that they presuppose that we have voluntary control over what
we believe. Yet, reflection on our mental lives suggests that we have no such con-
trol. Alston says:
this conception of epistemic justification is viable only if beliefs are sufficiently under
voluntary control to render such concepts as requirement, permission, obligation, re-
proach, and blame applicable to them. By the time honored principle that "Ought im-
plies can," one can be obliged to do A only if one has an effective choice as to whether
to do A.8
He goes on to argue that we don't have an effective choice over what we believe. In
the process of objecting to Chisholm's views about justification, Plantinga says of a
particular proposition that "whether or not I accept it is simply not up to me; but then
accepting this proposition cannot be a way in which I can fulfill my obligation to the
truth, or, indeed, any obligation. . . ."9 Thus, according to Plantinga, our lack of con-
trol over beliefs implies that they are not the sort of thing that can be a matter of
obligation, and this undermines Chisholm's deontological conception of epistemic
justification. Matthias Steup presents a similar argument, though he goes on to de-
fend the deontological conception on the grounds that belief is voluntary.10
For the purposes of the discussion that follows, it will be helpful to distinguish
two steps of the arguments just presented. Their target is a deontological conception
of epistemic justification, a conception according to which epistemic justification is
to be understood or analyzed in terms of the deontological concepts of obligation,
requirement, and the like. This conception is "viable," in Alston's terms, only if belief
is sufficiently under our voluntary control. Presumably, the deontological concep-
tion is viable only if it can be true that we are required to believe things, that we ought
not believe other things, and so on. That is, the deontological conception of epistemic
justification is "viable" only if deontological judgments about beliefs are sometimes
true. Thus, the argument against epistemic deontologism begins with the claim that
believing is not a voluntary activity. This is used to support the proposition that
deontological judgments about beliefs cannot be true. And from this it is inferred
that the deontological conception of epistemic justification is not viable. My con-
cern in this essay will be primarily with the first inference, the move from invol-
 Voluntary Belief and Epistemic Evaluation 79

untarism about belief to the falsity of all deontological judgments about beliefs. My
goal in what follows is to argue that deontological judgments about beliefs can be
true even if beliefs are involuntary. I will not be defending deontological analyses of
epistemic justification.
We can formulate the issue in terms of the following argument:

The Voluntarism Argument

1. People do not have voluntary control over their beliefs.
2. If deontological judgments about beliefs are sometimes true, then people
have voluntary control over their beliefs.
3. Deontological judgments about beliefs are not sometimes true.

Epistemologists have three kinds of response to this argument open to them: (i) They
can argue that we do have the requisite sort of control over our beliefs, thereby re-
jecting premise (1); (ii) they can argue that the deontological judgments do not have
voluntarist implications, thereby rejecting premise (2); or (iii) they can accept the
argument and admit that the familiar deontological terms of epistemic appraisal re-
ally are inapplicable.11 This in itself is a surprising conclusion, whatever the merits
of deontological analyses of epistemic justification.
I will discuss response (i) in section II and response (ii) in section III. I will not
discuss response (iii) except in passing and by implication.

II. Voluntarism About Belief

In this section, I will discuss premise (1) of The Voluntarism Argument. I will even-
tually argue that we do have voluntary control over some of our beliefs, but I will
also argue that this fact is of absolutely no epistemological significance and that it
does nothing to help resolve the real puzzle concerning voluntarism and deontological
judgments about beliefs.

Arguments for the Conceptual Impossibility

of Voluntary Belief
The philosophical literature contains two quite different kinds of defenses of the claim
that people do not have voluntary control over their beliefs. One defense relies on
the idea that voluntarily formed beliefs are a conceptual impossibility, and the other
claims merely that it is a contingent fact that we are unable to believe voluntarily. I
will discuss each sort of defense, but most of my attention will be directed on the
second one.

The Conceptual Impossibility Thesis According to the conceptual impossibility the-

sis, the nature of belief somehow rules out the possibility that it is a state voluntarily
entered into. Arguments for this surprising thesis have found their supporters, nota-
bly Bernard Williams12 and more recently, Dion Scott-Kakures.13 Jonathan Bennett
80 Richard Feldman

has also expressed support for the thesis, though, in a thoroughly admirable confes-
sion he admits that the arguments for it are all no good.14 The heart of the argument
is that belief is by its nature directed at the truth. If beliefs could be acquired at will,
they could be acquired independently of all truth considerations. But any state en-
tered into independently of truth considerations is by definition not a belief.
The argument is far more complex than the sketch just given. However, I won't
pursue it here. I think Williams's argument has been effectively criticized by a num-
ber of writers, as has the more recent variation offered by Scott-Kakures.15 Alston
remarks that he "cannot see any sufficient reason for the" conceptual impossibility
claim.16 Neither can I. It seems to me that someone, or something, could in principle
will himself or herself into a state that would share many features of beliefs, includ-
ing the state of affirming the truth of the proposition. Obviously, if we don't typi-
cally just will our beliefs, then willed beliefs would not be caused in the same way
our beliefs typically are. So much the worse for the idea that to be a belief, a state has
to be caused in the way our beliefs typically are.
Furthermore, I will argue below that we do have voluntary control over at least
some of our beliefs. Since we in fact have control over some beliefs, it is possible to
have control. So, the conceptual impossibility thesis is false.

The Contingent Inability Thesis Alston has given the most thorough defense of the
contingent inability thesis, the thesis that as a contingent matter of fact, people are
not able to acquire beliefs voluntarily. Alston's paper includes an excellent survey
of a variety of notions of voluntary control. For each type except one, he argues that
we lack that sort of control over beliefs. Alston admits that there is one very weak
notion of control that does apply to belief. But he contends that this sort of control
does not provide the basis for a good response to the Voluntarism Argument.
Alston begins by discussing basic voluntary control^1 We have basic volun-
tary control over those actions that we can "just do." Simple bodily motions are
the prime examples. I can just raise my hand, close my eyes, and bend my knee.
Some people, but not I, can wriggle their ears and curl their tongue. Alston cor-
rectly says that forming a belief is not like that. We can't just do it at will. When
we have decisive evidence for a proposition, we typically can't help but believe it.
When we have decisive evidence against a proposition, we typically can't believe
it. When our evidence is not conclusive, we typically can't help but believe what
our evidence seems to support. I think that Alston is entirely right about this. Of
course, we don't always follow our evidence in the way just described. Sometimes
our hopes or our fears get the best of us and we believe contrary to our evidence.
But we don't believe at will in these cases either. The contrast between simple ac-
tions and beliefs is striking. In my current, relatively normal circumstances, if I
want to raise my hand, I can just do it. In contrast, if I want to believe that it will
start raining shortly, I cannot just do it. The same is true of other potential beliefs.
I assume that I am not unusual in this respect. We don't have basic voluntary con-
trol over what we believe.
Alston turns next to nonbasic immediate voluntary control}* One has this sort of
control over the things one can do right away by doing something else, typically
something over which one has basic voluntary control. Standard examples are open-
 Voluntary Belief and Epistemic Evaluation 81

ing doors and turning on lights. We can, in typical circumstances, do these things
simply by moving our bodies in the appropriate ways. There's vagueness here con-
cerning what counts as "right away" but that vagueness is in no way problematic.
This is because the boundary between nonbasic immediate voluntary control and the
next weaker kind of control, long-range voluntary control, is acceptably imprecise.19
The sorts of things over which we have long range voluntary control are the sorts of
things we can do over time by doing other things. Perhaps painting my house is an
example. Or, more precisely, I have long range voluntary control over what color
my house is because I can do things like paint it.
Finally, there is indirect voluntary influence.20 This is the kind of control we have
when we can undertake a course of action that may affect some condition over the
long term. Perhaps a person has indirect voluntary influence over the condition of
her heart, since diet and exercise, courses of action she can more directly control,
can affect it.
Consider my belief that the Earth is not flat. Suppose that there were some reason
why I'd prefer not to have this belief. There's nothing I can do to rid myself of this
belief right away. It's not like moving my hand or turning out the lights. I don't have
basic or nonbasic immediate voluntary control over whether I have this belief. And
there's nothing much I can do long range to control it either. Changing my belief is
not relevantly like painting my house. I can't simply set out on a course of action
that will almost surely result in my belief being changed. I might enroll in the Flat
Earth Society, read conspiracy literature asserting that satellite photos are all phony,
and so on. Perhaps this will help rid me of my belief. Alston would agree that it might.
But this gets us, at most, indirect voluntary influence, and this is not the sort of effec-
tive voluntary control required to refute The Voluntarism Argument.21
I believe that, for most part, what Alston says is right. However, I will argue that
we have considerably more control over some of our beliefs than Alston acknowl-
edges. Still, this control does not undermine the basic idea behind The Voluntarism
Argument, although it may show that the argument needs reformulation.

An Argument for Voluntarism About Belief

My argument for voluntarism begins with the assumption that there are states of the
world over which people have nonbasic voluntary control. For example, I have
nonbasic voluntary control over whether the lights in my office are on or off. All I
have to do is move in a certain way to get the lights to be on or off. And I can do this.
I assume that the proposition that I have control is not refuted either by the fact that
the power could fail so that the lights don't come on when I turn the switch or by the
fact that the switch could fail keeping them on no matter what I do. These mere pos-
sibilities don't undermine control, though if they were actualities perhaps they would.
The next step of the argument notes that my belief about whether the lights are on
tracks the actual state of the lights almost perfectly, As a result, I have control over
whether I believe that the lights are on. All I have to do is move a certain way, and
then I'll have the relevant belief. More generally, when I have control over a state of
the world and my beliefs about that state track that state, then I have just as much
control over my belief about the state as I have over the state itself. Thus, we have
82 Richard Feldman

nonbasic immediate voluntary control over our beliefs about states of the world over
which we have control, provided our beliefs are responsive to those states.
Here's the argument spelled out in detail as it applies to a single case.

An Argument for Voluntarism About Belief

1. You can causally determine whether the lights are on by moving or failing
to move your hand in a certain way.
2. You can similarly causally determine whether you believe that the lights
are on by moving or failing to move your hand in that same way.
3. If the truth of (1) is sufficient for you to have nonbasic voluntary control
over whether the lights are on, then the truth of (2) is sufficient for you to
have nonbasic voluntary control over whether you believe that the lights
are on.
4. The truth of (1) is sufficient for you to have nonbasic voluntary control over
whether the lights are on.
5. The truth of (2) is sufficient for you to have nonbasic voluntary control over
whether you believe that the lights are on. (3), (4)
6. You have nonbasic voluntary control over whether you believe that the lights
are on. (2), (5)22

I think that this is a sound argument. I also think that similar arguments establish that
each person has voluntary control over lots of beliefs. Roughly, we have nonbasic
immediate voluntary control over our beliefs about states of the world over which
we have control, provided our beliefs are responsive to those states. Furthermore, if
we know that we will respond in some mistaken way to some state of the world over
which we have control, we also have control over the resulting (erroneous) belief.
So, voluntarism is true after all.
I believe that the existence of nonbasic voluntary control over beliefs can have
prudential and moral significance. If the department chair announces that she'll give
a raise to all and only those members of the department who in 30 seconds believe
that the lights in their office are on, then it is prudent for me to head for the light
switch to make sure I have the desired belief. If the chair perversely announces that
the graduate students will be mercilessly tortured—say, by being forced to take ad-
ditional prelims—unless in 30 seconds I believe that my lights are on, then I'd better
make sure I have that belief. Excuses about the involuntariness of belief would fail
here. I'm in control. Thus, we do have control over many of our beliefs. Premise (1)
is false.

The Voluntarism Argument Revised

The existence of the sort of voluntary control over our beliefs described here does
nothing to alleviate the worries about the viability of deontological epistemic judg-
ments. There are many beliefs that are not about states of the world that we can con-
trol, so no argument like mine can support the conclusion that we have control over
them. Yet we make deontological epistemic judgments about beliefs that we can't
 Voluntary Belief and Epistemic Evaluation 83

control, and these judgments are as routine and commonplace as are judgments about
beliefs that we can control. Thus, our ability to control what we believe in the way
described here is epistemically insignificant.
We can take this point into account by reformulating The Voluntarism Argument:

The Voluntarism Argument (Revised)

\. People do not have voluntary control over beliefs concerning states of the
world they can't control.
2. If deontological epistemic judgments about those beliefs are true, then people
have voluntary control over those beliefs.
3. Deontological epistemic judgments about those beliefs concerning are not
true.

This revised argument is equally troubling. It implies that an enormous number of

the deontological epistemic judgments we routinely make cannot be true. Further-
more, the argument might be extended to beliefs that we can control but have not
formed in this voluntary way. Thus, I don't think that the fact that we have nonbasic
immediate voluntary control over some beliefs provides the basis for an effective
defense of epistemic deontologism.
There's another reason that anyone who is troubled by The Voluntarism Argu-
ment is unlikely to think that the control we have over our beliefs is sufficient to make
sense of deontological epistemic judgments about the beliefs that we can control. In
other words, if (3) of the revised argument is true, then deontological epistemic judg-
ments about beliefs we can control are not true either. This is because the control
that we do have comes at the wrong point in the belief-forming process. Favorable
epistemic judgments are made when a person believes the right thing given the evi-
dence the person has, and unfavorable judgments are made when a person believes
something not supported by the person's evidence. What my argument shows is that
we can control our evidence and thereby control our beliefs. The argument does not
contest the proposition that we are at the mercy of our evidence (or perhaps our evi-
dence combined with whatever other nonevidential facts cause beliefs). But it is this
fact that is, I think, the real heart of the worry about epistemic deontologism. The
worry isn't simply that we can't voluntarily control our beliefs. We can voluntarily
control them, at least in some cases. The real worry is that epistemic evaluations have
to do with how we respond to evidence, and we don't have voluntary control over
that.231 turn next to a defense of voluntarism about belief that addresses this point.

A More Robust Form of Voluntarism

Compatibilists in the free will debate contend that we voluntarily perform an action
when the action has the right sort of cause. A defender of voluntarism about belief
can argue that, analogously, we believe voluntarily when we believe as a result of
the right sort of causal process. Roughly, the idea is that when an unconstrained
consideration of evidence leads to belief, we believe voluntarily.24 Matthias Steup
has presented the most fully developed version of this interesting line of thought. It
84 Richard Feldman

will be difficult to come to any definite conclusion about its merits because we lack
a fully adequate understanding of what counts as the right sort of causal process. Still,
I think that there are good grounds to reject the idea that the deliberative process that
leads to belief is one that makes it appropriate to conclude that we believe things
voluntarily.
There are indeed a number of parallels between acting and believing. In particu-
lar, various conditionals seem to be true: If I had evaluated my evidence differently
(or if I had different evidence), then I would have believed differently. This is analo-
gous to what is true in the case of action: If I had evaluated my alternatives differ-
ently (or if I had different preferences), then I would have acted differently. Thus,
those who say that our beliefs are not voluntarily formed are faced with a challenge.
What is the relevant difference between the process that leads to action and the pro-
cess that leads to belief in virtue of which we count actions as voluntary and beliefs
as involuntary? For compatibilists, that the belief-forming process is deterministic
can't be the answer.
Compatibilists about action need not say that it is the truth of conditionals such as
the ones just mentioned that makes actions voluntary. Instead, they can say that an
action is voluntary when it is caused by the appropriate sort of internal process. In a
paradigm case of voluntary action (ordering from a menu), people deliberate about
their options and then select one. In this case, deliberation is a key aspect of the pro-
cess that leads to voluntary action. Some cases of forming a belief are at least super-
ficially similar. People weigh their evidence and come to a conclusion as a result.
Steup describes this sort of process as a "doxastic decision": the decision to believe
or not believe something. His suggestion is that beliefs are voluntary when they re-
sult from this sort of process.
It is true that we can think about what our evidence supports and form a belief as
a result. And it is true that this is something like making a decision about what to
believe. But I think that it is a mistake to conclude that the existence of this sort of
process shows that we have voluntary control over our beliefs. An example will bring
out the reason for this. Steup describes a case in which the author Vincent Bugliosi
struggled with what to believe about OJ Simpson's guilt.25 Steup says that for a long
time, Bugliosi was baffled by the case and suspended judgment. But eventually the
evidence became stronger and:

1. Bugliosi decided to believe that Simpson was guilty.

Now, you might think that he didn't really decide to believe this, and argue that (1)
is just an imprecise way of saying that

2. Bugliosi decided that OJ is guilty.

Steup responds that it's not up to Bugliosi whether OJ is guilty. That's not some-
thing he can decide. So (2) isn't true. While that's correct, I think that what's actu-
ally true in a case like this is best expressed by:

3. Bugliosi came to the conclusion that OJ is guilty.

 Voluntary Belief and Epistemic Evaluation 85

If we are to speak carefully, it would be best to say that (3) is the sort of thing that
is true in this sort of case. Either (1) isn't strictly true, or it just means something
like (3). We come to a conclusion as a result of weighing evidence. We don't de-
cide what to believe on the basis of that evidence, and we don't decide what that
evidence supports.
What is involved in genuine and paradigmatic decision making is the formation
of an intention. The most central use of the concept of a decision occurs in sentences
such as:

4. Bugliosi decided to write a book arguing that OJ is guilty.

And this implies:

5. Bugliosi formed an intention to write a book arguing that OJ is guilty.

Now, when (5) is true and he does write the intended book as a result, and perhaps
some other things are true, then Bugliosi voluntarily wrote the book. That is, on the
compatibilist view, to be a voluntary action, an action must be caused by an appro-
priate intention to perform that action.
We typically don't form intentions to form beliefs and form them as a result.26
After weighing the evidence, Bugliosi did come to his conclusion. But he didn't in-
tend to come to that conclusion and then come to it as a result. This can be made
more obvious by considering again the examples I gave earlier to illustrate the pos-
sibility of voluntarily believing something. In those cases I do form just such an in-
tention to believe something. I do act with the purpose of bringing it about that I have
a particular belief. The absence of this intention in all typical cases of deliberating
about what to believe is what makes the process one that does not give us voluntary
control over what we believe. We do control what we believe, in much the way we
control other processes in us. That is, we, or processes in us, control such things as
blood pressure and temperature. But this is not voluntary control. Epistemic delib-
eration does not result in effective intentions to believe. Except in rare cases, we don't
form intentions to believe. But such intentions are essential to voluntary control.
The cases of voluntary belief formation that I described earlier make it easy to see
what it would be like to decide to believe something and then to believe it as a result.
Consider again the case in which the department chair says that the fate of graduate
students depends on what I believe about the conditions of my lights in 30 seconds.
I might briefly consider the matter, weigh the plight of the students against the in-
convenience of getting up and flipping the switch, and decide to believe what is in
their best interests. I could make such a decision, carry it out, voluntarily believe the
right thing, and thereby win the loyal support of the students. That's what deciding
to believe, and voluntarily believing, would be like.
Furthermore, if the fact that belief is the outcome of deliberation were sufficient
to make belief voluntary, then many other plainly involuntary behaviors would be
voluntary. Deliberating about something can result in states other than belief. It can
result in desires or in panic. Thus, for example, I might think for a while about vari-
ous flavors of ice cream and end up with a desire for a scoop of mocha chip. Or, I
86 Richard Feldman

might think about the financial costs associated with retirement and end up in a panic.
Deliberation results in these outcomes. Had the deliberation gone differently in cer-
tain respects, then there would have been different outcomes. If we say that belief is
voluntary simply because it is the outcome of deliberation about evidence, then it is
hard to see why we shouldn't say that these other outcomes of deliberation are also
voluntary. Yet it is clear that they are not. Again, what seems clearly missing is the
right sort of intention. I never intend to desire mocha chip ice cream nor do I intend
to panic about retirement costs.27 Similarly, I don't (except in odd cases like those
presented above) intend to form a belief and then form one as a result. Believing may
be the consequence of a deliberative process, but it is not voluntary behavior.
Thus, I reject the more robust form of voluntarism for which Steup argues. There
is good reason to say that the deterministic processes that typically lead to action
render those actions voluntary. The process leads to effective intentions to act. De-
liberating about evidence does lead to belief, but not via such intentions. Such delib-
eration does not make belief voluntary. We need a better response to The Voluntarism
Argument.

III. Ought and Can

The second main sort of response to the voluntarism puzzle denies that the predi-
cates in deontological claims about beliefs imply that those beliefs are voluntarily
adopted. It denies premise (2) of both versions of The Voluntarism Argument. It may
be that there are differences among the various assertions that I've described as
"deontological." I'll focus first on judgments about what one is obligated to believe
and what one ought to believe and then discuss other judgments later in the section.
One way to defend deontological epistemic judgments in the light of involuntarism
is to argue that we can have epistemic obligations even though we can't fulfill them
(or even if we can help but fulfill them). This is to deny that "epistemically obligated
to" or "epistemically ought to" implies "can," no matter what the time-honored prin-
ciples imply. One way to make this denial plausible is to show that there are other
kinds of ought statements that don't imply voluntary control. I will consider several
candidates for this other kind of ought.

Contractual Obligations
You can have an obligation to pay your mortgage even if you don't have the money
to do so. Perhaps students in a class have an obligation to do the course work even if
they are incapable Of doing it. Other examples of this sort are rather easy to con-
struct. Perhaps epistemic obligations are analogous to these financial and academic
obligations, as I once suggested.28
I now think that it is an implausible to think that epistemic obligations are similar
to the obligations just described. The obligation to pay one's mortgage and the obli-
gation to do one's course work are contractual obligations, although in the latter case
the contract is in some sense implicit. It's difficult to see any basis for saying that we
all have a contractual obligation to believe things. Surely no such contract is explicit,
 Voluntary Belief and Epistemic Evaluation 87

and nothing analogous to enrolling in a course establishes an implicit contract. Fur-

thermore, these contractual obligations are obligations to someone else, but it is dif-
ficult to see to whom one could have an epistemic obligation.

Paradigm Obligations
In a recent paper, Nicholas Wolterstorff says that there are two kinds of obliga-
tions, paradigm obligations and responsibility obligations.29 Only obligations of
the latter sort are associated with voluntariness. As examples of paradigm obliga-
tions, he presents:
1. "You ought to be walking on it in two weeks"—said by a physician as he finishes
binding up a person's sprained ankle.
2. "That's strange; you ought to be seeing double"—said by a psychologist to his
subject while conducting an experiment in perception.
Wolterstorff suggests that epistemic obligations are similar to the obligations described
by these sentences. They lack any implication of voluntary control.
No friend of epistemic deontologism should be comforted by the idea that epistemic
obligations are like the obligations described by (1) and (2). This is because there
are no obligations described by (1) and (2). Sentences (1) and (2) are "ought" sen-
tences; they are not obligation sentences, and they cannot be paraphrased in any
straightforward way into obligation sentences. Your ankle has no obligation of any
sort to heal; you have no "perceptual obligation" to see double. So, if there are
epistemic obligations, they are not like the obligations described here, since there
are no obligations described here.
If there is a difference between sentences describing what we ought to do and
sentences describing what we are obligated to do, then perhaps Wolterstorff is guilty
only of failing to take note of this distinction. Perhaps he has given a good account
of epistemic oughts: they are "paradigm oughts," describing paradigmatic behavior.
(1) and (2) describe normal behavior, or at least normal behavior when there is no
interference. Thus, for example, (1) says that barring unforeseen and unlikely devel-
opments, your ankle will heal in two weeks. But epistemic oughts do not describe
normal function. Some researchers report that people typically make various unjus-
tified inferences and predictably form unreasonable or erroneous beliefs.30 A re-
searcher might say, while awaiting the subject's reply to a question that nearly every-
one misses, "He ought to make the wrong inference here." This use of "ought" is
Wolterstorff s paradigm ought; it is perfectly analogous to the use of "ought" in (2).
But to describe the inference as "wrong" is to say that the person epistemically ought
not do what the researcher expects him to do. It may be, then, that we epistemically
ought not do what we normally do. Epistemic oughts are not paradigm oughts.

Role Oughts
There are oughts that result from one's playing a certain role or having a certain
position. Teachers ought to explain things clearly. Parents ought to take care of their
kids. Cyclists ought to move in various ways. Incompetent teachers, incapable par-
88 Richard Feldman

ents, and untrained cyclists may be unable to do what they ought to do. Similarly, I'd
say, forming beliefs is something people do. That is, we form beliefs in response to
our experiences in the world. Anyone engaged in this activity ought to do it right. In
my view, what they ought to do is to follow their evidence (rather than their wishes
or fears). I suggest that epistemic oughts are of this sort—they describe the right way
to play a certain role. Unlike Wolterstorff s paradigm oughts, these oughts are not
based on what's normal or expected. They are based on what's good performance.
Furthermore, it is plausible to say that the role of a believer is not one that we have
any real choice about taking on. It differs in this way from the other roles mentioned.
It is our plight to be believers. We ought to do it right. It doesn't matter that in some
cases we are unable to do so. Thus, I reject the second premise of both versions of The
Voluntarism Argument. Even in cases in which a believer has no control at all, it makes
sense to speak of what he ought to believe and ought not to believe.
What counts as good performance in a role, and thus determines how a role ought
to be carried out, may be dependent in certain ways on what people are generally
able to do. Consider, for example, the claim that teachers ought to explain things
clearly. Arguably, what counts as a clear explanation is dependent at least in part on
what people are able to say and what people are able to understand. One could imag-
ine standards for a clear explanation that are so demanding that no one could ever
meet them. It is not true that teachers ought to explain things that clearly. Similarly,
the standards of good parenting or good cycling that apply to us are not at super-
human levels. It's not true that parents or cyclists ought to do things that would require
them to exceed the sorts of capacities people have. It is consistent with this, how-
ever, that an individual ought to do things that he or she is not able to do. An inar-
ticulate teacher may simply be unable to explain things as clearly as he ought, and he
may not have the capability of learning to explain things clearly. Thus, even if the
standards for good performance in a role are in some way limited by the capacities
of those who fill the role, it is not the case that the existence of those standards im-
plies that individuals must have basic or nonbasic voluntary control over that behav-
ior that is judged by those standards.
This reply to The Voluntarism Argument does not depend on any specific view
about what we ought to believe or what counts as the relevant kind of good perfor-
mance as a believer. However, a brief description of one view about the substance of
our epistemic requirements may help make the reply more plausible. In my view,
the right way to carry out one's role as a believer is to form beliefs that are supported
by one's current evidence. That is, if one is considering a proposition, then one ought
to believe it if one's evidence supports it, ought to disbelieve it if one's evidence
goes against it, and ought to suspend judgment if one's evidence is neutral.31 It's
consistent with it being the case that one ought to believe a particular proposition
that one is unable to believe it, perhaps because it is psychologically too troubling or
for other reasons unrelated to the merits of one's evidence. In other cases, the pos-
session of good evidence for a proposition may make believing that proposition
unavoidable. Nevertheless, it is true that one ought to believe it. Epistemic oughts
don't imply voluntarism about beliefs.
I think that most, possibly all, of the deontological terms we ordinarily use to
evaluate beliefs can be explained in similar evidential terms. Whenever one's evi-
 Voluntary Belief and Epistemic Evaluation 89

dence on balance supports a proposition, then one has an epistemic right to believe
that proposition. Believing it is permitted. Believing anything for which one lacks
good reasons is epistemically prohibited. If there are cases in which one's evidence
supports either of two attitudes equally well, then it will be true that one has a right
to either of those attitudes, and that either of them is permitted.32 It seems to me rea-
sonable to say that when only one attitude toward a proposition is permitted, then
one has an epistemic obligation to have that attitude, if one has any attitude toward
the proposition at all. It may be, of course, that one has no obligation to take any
attitude at all toward the proposition. None of these assertions implies voluntarism
about beliefs.
It is less clear what to make of judgments in which people are praised or blamed
for their beliefs. It may be that these terms are reserved for voluntary behavior. Even
here, the case is less than perfectly clear. Consider again Ginet's proposal that one is
justified in being confident that p if and only if one could not be justly reproached
for being confident that p. I think that to be reproached is to be blamed, yet it is not
clear that people can be justly reproached only for what they have done voluntarily.
Whatever the verdict on praise and blame, the case for the other deontological claims
is clear.
I conclude that deontological epistemic judgments can be true even if people lack
effective voluntary control over their beliefs. The Voluntarism Argument fails.

IV. Is This Epistemic Deontologism?

In his influential paper on epistemic deontologism, Alston begins by saying that it is

"natural" to understand epistemological terms in a "deontological" way, having to
do with "obligation, permission, requirement, blame, and the like."33 Alston objects
to views of this sort on the grounds the properties expressed by the deontological
terms apply only if beliefs are voluntarily adopted. My discussion has been focused
on this argument, and I've argued that the various deontological terms can apply even
if believing is not voluntary behavior. By the end of his paper, Alston seems to be
directing his attack more narrowly. In his concluding paragraph, he characterizes
epistemic deontologism as the view that analyzes "epistemic justification in terms of
freedom from blame for believing."34 I haven't attempted to defend this sort of
deontologism here.
Moreover, I don't think that this more narrowly defined sort of deontologism is
so natural or common. What is natural and common, I think, is to make claims like
those with which Alston began his paper, claims that imply that people have epistemic
rights, duties, and permissions, and that they ought to believe one thing rather than
another. They can, no matter what the truth about doxastic voluntarism is.
It will be instructive to conclude by examining one passage from John Locke that
Plantinga cites in his discussion of epistemic deontologism:
Faith is nothing but a firm assent of the mind: which if it be regulated, as is our duty,
cannot be afforded to anything, but upon good reason; and so cannot be opposite to it.
He that believes, without having any reason for believing, may be in love with his own
fancies; but neither seeks truth as he ought, nor pays the obedience due his maker, who
90 Richard Feldman

would have him use those discerning faculties he has given him, to keep him out of mis-
take and error. He that does not do this to the best of his power, however he sometimes
lights on truth, is in the right but by chance; and I know not whether the luckiness of the
accident will excuse the irregularity of his proceeding. This at least is certain, that he
must be accountable for whatever mistakes he runs into: whereas he that makes use of
the light and faculties God has given him, and seeks sincerely to discover truth, by those
helps and abilities he has, may have this satisfaction in doing his duty as a rational crea-
ture, that though he should miss truth, he will not miss the reward of it. F^or he governs
his assent right, and places it as he should, who in any case or matter whatsoever, be-
lieves or disbelieves, according as reason directs him. He that does otherwise, transgresses
against his own light, and misuses those faculties, which were given him.35

There are two lines of thought interwoven in this passage from Locke. One line of
thought is thoroughly evidentialist, and it expresses what I take to be the central
epistemological claims in the passage. These are the claims that it is our duty to as-
sent to things only on the basis of good reason, that to believe without good reason
is not to believe as one ought, and that one who places one's assent right "believes or
disbelieves, according as reason directs him." These assertions make use of terms
such as those that appear on Alston's initial list of deontological terms. This alone
provides grounds for regarding Locke as endorsing a version of deontologism about
epistemology or, at the very least, as holding that deontological judgments about
beliefs can be true.
No doubt there is a second strand to the passage from Locke, a strand that clearly
displays a voluntarist tone. It may well be that Locke did ascribe to some version of
voluntarism about belief. He does speak of seeking the truth "to the best of one's
power," for example. But a striking thing about the passage is that the central episte-
mological claims make good sense without the assumption of voluntarism. One can,
if one likes, insist that to be an epistemic deontologist, one must endorse the volun-
tarist strand in Locke's thought as well as the first. Nevertheless, it remains true that
the deontological judgments about belief with which we began do not imply that belief
is voluntary.36

Notes
This essay is an expanded version of section I of my paper, "The Ethics of Belief," Philoso-
phy and Phenomenological Research 60 (2000): 667-95.

1. "The Deontological Conception of Epistemic Justification," Philosophical Perspec-

tives 2 (1988): 257-99. The quotation is from p. 258. This paper is reprinted in William Alston,
Epistemic Justification (New York: Cornell University Press, 1989), 115-52.
2. A. J. Ayer, The Problem of Knowledge (Middlesex, U.K.: Penguin Books, 1956), 35.
3. Carl Ginet, Knowledge, Perception, and Memory (Dordrecht, The Netherlands: Reidel,
1975), 28.
4. Roderick M. Chisholm, Theory of Knowledge 2d ed. (Englewood Cliffs, N.J.: Prentice
Hall, 1977), 14.
5. For additional references, see "The Deontological Conception of Epistemic Justifica-
tion," footnote 5.
6. Alvin Goldman, "Internalism Exposed," Journal of Philosophy 96 (1999): 271-93.
7. Alvin Plantinga, Warrant: The Current Debate (New York: Oxford University Press,
1993), chap. 1.
 Voluntary Belief and Epistemic Evaluation 91

8. "The Deontological Conception of Epistemic Justification," 259.

9. Platinga, Warrant: The Current Debate, 38.
10. Matthias Steup, "The Concept of Epistemic Justification," An Introduction to Con-
temporary Epistemology (Upper Saddle River, N.J.: Prentice Hall, 1996), chap. 4.
11. This last alternative implies that either epistemic justification is not to be analyzed in
deontological terms or else that epistemic justification is itself inapplicable to our beliefs,
just like the other deontological epistemic terms.
12. Bernard Williams, "Deciding to Believe," Problems of the 5e//(New York: Cambridge
University Press, 1973).
13. Dion Scott-Kakures, "On Belief and the Captivity of the Will," Philosophy and Phe-
nomenological Research 53 (1993).
14. Jonathan Bennett, "Why Is Belief Involuntary?," Analysis 50 (1990).
15. For a discussion of Scott-Kakures's article, see Dana Radcliffe, "Scott-Kakures on
Believing at Will," Philosophy and Phenomenological Research 57, no. 1 (1997). For dis-
cussion of Williams, see Barbara Winters, "Believing at Will," Journal of Philosophy 76
(1979).
16. Alston, "The Deontological Conception of Epistemic Justification," 263.
17. Ibid., section III, 263-68.
18. Ibid., section IV, 268-74.
19. Ibid., section V, 274-77.
20. Ibid., section VI, 277-83.
21. Perhaps this point suggests that both (1) and the consequent of (2) should be about
effective voluntary control.
22. You can read premise (3) of this argument as asserting the following conditional: If
(if (1) then you have voluntary control over whether the lights are on) then (if (2) then you
have voluntary control over whether you believe that the lights are on). Premise (4) asserts
the truth of the antecedent of (3). Strictly speaking, (1) isn't needed in the argument. (2) is
needed, and (1) serves the role of warming up the reader to (2).
23. The argument of this paragraph relies on an evidentialist view about epistemic evalu-
ations. But it is likely that the same argument could be made using a different view about the
basis of epistemic evaluations.
24. For a defense of this view, see Matthias Steup, "Doxastic Voluntarism and Epistemic
Deontology," Acta Analytica 15 (2000): 25-56. Steup defends a similar view in chap. 4 of
An Introduction to Contemporary Epistemology. For a similar view, see James Montmarquet,
"The Voluntariness of Belief," Analysis 46 (1986). Bruce Russell endorses Steup's view in
"Epistemic and Moral Duty" this volume, 34^48.
25. Steup, "Doxastic Voluntarism and Epistemic Deontology," 7-10.
26. Perhaps the process in which by which formulate intentions and act on their basis can
become automated in certain ways, so that voluntary actions need not involve explicit con-
scious formulations of intentions. Belief formation is not like that either. It does not typically
involve the formation of intentions to believe at all.
27. One could imagine cases in which one does have these intentions. But they would be
as odd as the intentions in my examples of voluntary belief.
28. "Epistemic Obligations," Philosophical Perspectives 2 (1988), 240-43.
29. Nicholas Wolterstorff, "Obligations of Belief: Two Concepts," in The Philosophy of
Roderick Chisholm, ed. Lewis E. Hahn, (LaSalle, 111.: Open Court, 1997).
30. For a discussion of many of the examples allegedly showing that people are irrational
and an examination of their philosophical implications, see Edward Stein, Without Good
Reason: The Rationality Debate in Philosophy and Cognitive Science (Oxford: Clarendon,
1996).
31.1 expand upon this idea in section II of "The Ethics of Belief," Philosophy and Phe-
nomenological Research, 60 (2000).
32. In "The Ethics of Belief I argue that there are no cases in which each of two attitudes
toward one proposition are epistemically permissible.
33. "The Deontological Conception of Epistemic Justification," p. 257.
92 Richard Feldman

34. Ibid., p. 294.

35. An Essay Concerning Human Understanding, A. C. Fraser, ed. (New York: Dover,
1959), IV, xvii, 24, pp. 413-414. This passage is quoted by Plantinga in Warrant: The Cur-
rent Debate, p. 13.
36. In working on this essay I've benefited greatly from numerous discussions with Earl
Conee, John Bennett, Stewart Cohen, Jonathan Vogel, and all the students in my fall 1998
epistemology seminar. I'm also grateful to John Greco and Matthias Steup for helpful com-
ments. Earlier versions of the essay have been presented at the University of Rochester and at
Rutgers University. I'm grateful to the audiences on both occasions for helpful discussion
and to Keith DeRose for his provocative comments on the latter occasion.
 6

Doxastic Voluntarism and

the Ethics of Belief

ROBERT AUDI

Belief is profoundly analogous to action. Both are commonly grounded in reasons;

both are a basis for praising or blaming the subject; both are sensitive to changes in
one's environment; both can appropriately be described as objects of decision and
deliberation, and beliefs can appear quite action-like when conceived as formed by
assent or by acceptance. These similarities can make it plausible to think of belief as
sometimes (directly) voluntary in the sense that, like raising our hands, believing is
sometimes done "at will." This voluntarist idea has received considerable scrutiny,1
but there remains a need to evaluate it from a theoretical perspective rather different
from any so far brought to bear on it. Indeed, I believe that unless we appreciate the
bearing of a new theoretical perspective on the issue we cannot properly appraise
the question of the voluntariness of belief and, correspondingly, the connection
between justified beliefs and what are called epistemic duties cannot be properly
appraised. I begin with some of the crucial data favoring voluntarism, then proceed
to an analysis of some representative cases. In the light of that analysis, the theoretical
perspective I want to develop can be articulated and applied. Finally, with that per-
spective in view, we can see how best to conceive the ethics of belief.

I. The Case for Doxastic Voluntarism

There are many forms of doxastic voluntarism, some much stronger than others. For
my purposes, the view may be taken to be the thesis that belief is sometimes under
direct voluntary control. (That it is sometimes, in some way, under indirect volun-
tary control is uncontroversial.) Even this formulation admits of stronger and weaker

93
94 Robert Audi

interpretations. On a strong interpretation, it expresses the view that believing itself

is an action-type having some tokens that are directly voluntary. On a weaker inter-
pretation, it expresses the view that forming a belief is sometimes such an action-
type. Call the first view the behavioral version of doxastic voluntarism and the sec-
ond the genetic version.
The similarities between belief and action would probably not sustain the stron-
ger, behavioral version of (doxastic) voluntarism if there were not a number of locu-
tions that seem best understood on the assumption of its truth. To the question "What
did she do that so upset you?", the reply, "She believed her lying husband's fabrica-
tion" is an admissible answer. In the same context, another person might say, "For
my part, I can't decide whether to believe him or not; I'm still deliberating about
whether his story is true." Still another might say, "I rejected his testimony as fast as
she accepted it." And Descartes famously spoke of assenting to a proposition as a
kind of thing we do (presumably at will) that entails coming to believe the proposi-
tion in question.2
There is no need here to do a full-scale appraisal of behavioral voluntarism, even
for these special cases that seem to support it. It is enough to indicate an alternative
account of them that accommodates what is sound in voluntarism but also paves the
way for an overall conception of belief and action that yields better results in both
philosophy of mind and epistemology.
I want to start with a distinction between positive and negative (direct) control
over beliefs: between the ability to believe, or at least to form beliefs, at will and the
ability to prevent their formation at will. The two kinds of control might seem to be
on a par, each equally supporting voluntarism. They are surely not: That one can
bring something about at will is at least some reason to consider it an action,3 but the
capacity to prevent something at will need not be such a reason. Negative control
does not imply positive control. Consider sneezing, twitching, or even breathing, in
the sense in which breathing is nonvoluntary and automatic. One might learn to sup-
press these at will. But they are clearly not actions.4 Suppressibility, even at will, is
not unique to actions.
To be sure, suppressing an event such as a sneeze or a twitch or a breath is action.
Abstaining from such suppression is also action. But these events themselves are not
action, and the first two are surely the wrong kinds of events to be even candidates
for action. We can, however, bring about belief formation by abstaining from sup-
pressing it, as in cases in which, as a child gives us implausible excuses, we let our
defenses down. But this is only a case of indirectly voluntary belief formation: It is
by doing something else (e.g., putting out of mind our thoughts of the child's past
falsified excuses), that we make way for, and thereby complete the sufficient condi-
tions for, belief formation. Similarly, that we can sometimes prevent belief forma-
tion in what seems a directly voluntary way—as where we withhold belief of A's
testimony until we hear B's—does not imply that believing is under direct positive
voluntary control when we finally do come to believe A's testimony, or even that it
is under direct negative voluntary control.
One might think that even countenancing, for certain cases, indirect positive vol-
untary control of belief grants too much to voluntarism. But the existence of such
indirect control does not imply that believing is action and certainly does not require
 Doxastic Voluntarism and the Ethics of Belief 95

us to go on to countenance direct voluntary control of belief. A parallel case is pre-

vention of various nonactional events indirectly, say, preventing a twitch by tying
down the offending arm. Here, too, the corresponding direct negative control is not
implied, though in this case I have acknowledged that it apparently exists in some
instances.
It may also seem that we should not grant voluntarists that there is even direct
negative control of belief. Perhaps what appears to be direct withholding of belief is
really accomplished by doing something as a means, such as issuing inner commands,
and hence is really indirect. I doubt that instrumental action is always necessary here.
Granted, there may be a sense in which one says "no" to the impending event, such
as belief formation; but this may be a description of the exercise of one's direct nega-
tive control over the event. It need be no more a means to its prevention than "will-
ing" to raise one's arm is to the action of raising it.5
If negative direct voluntary control of belief does not imply the corresponding
positive control, there may yet be such positive control. Let us consider our putative
examples, beginning with the first, in which our agent's believing her husband's story
is said to be what she "did" that upset someone. Perhaps the genuine action here is
her not withholding belief. In that case, her belief formation is only indirectly volun-
tary: produced by her abstention from withholding belief. In any event, one could be
equally upset by someone's sneezing during the final prayer at a funeral. This might
be described as an unfortunate thing to have done; but "do" is a very wide term, and
the admissibility of such a description is not sufficient to imply that sneezing is an
action. It certainly does not imply that sneezing is directly (positively) voluntary;
indeed, presumably only actions are even candidates to be directly voluntary.
The case of "deciding to believe" is more difficult to deal with. Why should we
say we can't decide whether to believe someone or are deliberating about whether a
proposition is true unless we take it that when we do decide, we will be (voluntarily)
acting in fulfillment of our decision or deliberation and thereby believing, or at least
coming to believe, the proposition in question? And don't we weigh reasons pro and
con or scrutinize a speaker for signs of credibility or deceitfulness, just as we might
do where our concern is (say) whether to make a statement?
There is no question that we weigh and respond to reasons in both doxastic and
behavioral cases. But it is a further step to say that our cognitive response to the rea-
sons for belief is an action and still another step to maintain that it is (directly) vol-
untary. Saying that, after considering the reasons pro and con, I decided to believe
him, could embody a use of "decide" in which it designates not the voluntary pro-
duction of belief or any kind of "act of believing," but a manifestation of belief for-
mation in response to the preponderance of the reasons.6 Perhaps in an instant I notice
his eyes watering and hear a quaver in his voice; he is not acting, but genuinely hurt
to be doubted and sad in reporting what he does. Here "deciding to believe him"
indicates my cognitive resolution of the matter, as "deciding to try the left fork in the
road first" can indicate simply that I took that route and not that I experienced a sepa-
rate volitive thrust favoring it. The attractiveness of one route may simply have be-
come greater than that of the other as I looked at their vegetation, without my engag-
ing in any separate voluntary process of deciding on the route. Anthropomorphists
that we sometimes are, perhaps we homuncularize a bit: Picturing ourselves as judges
96 Robert Audi

hearing evidence, we imagine belief formation as giving or at least arising from a

verdict. And it does sometimes represent our coming down on one side of an issue
rather than the other. The question is whether we take that side at will or are instead
drawn to it by a nonvoluntary process of rising conviction.7
In support of their answer to this question, voluntarists can note that in forming a
belief of what someone says, we are taking a stance, which is surely doing some-
thing. Granted, if the matter is worth deliberating about or even calls for making a
decision, the idea of taking a stance can apply. But although "taking a stance" can
designate a directly voluntary action, it may also refer to coming to occupy the rele-
vant position as a result of belief formation that is plainly not due to decision. Thus,
"What is your stance on his competence to refuse life support?" may elicit a detailed
answer that simply lays out beliefs one formed about his illness the moment one read
a medical history. There need have been no question of what to believe; the apparent
truth may have been obvious.
A voluntarist might reply that this is only a case of coming to occupy a stance
without taking one. I grant the distinction, but it is not sharp. It is easy to imagine
cases in which a person is properly said to take a stance against life support, yet doing
so is simply a spontaneous and inevitable result of the interaction of standing biases
with the new information that the person immediately casts into an awaiting mold.
The naturalness of using decision-talk and deliberation-talk in relation to belief must
be granted. But it may well have more to do with a stance (or a cognitive position
something like it) resulting as a response to reasons than with the way it results from
them. Perhaps we speak of deliberating about what to believe because we take it that
our belief should be supported by reasons, rather as a deliberate action should be,
and not because the belief will be action or, as it were, enacted.
These considerations about belief formation are suggestive concerning the analy-
sis of acceptance. To be sure, when acceptance is contrasted with rejection, it typi-
cally is a case of action. But we can describe someone, especially a child, as having
accepted a person's story lock, stock, and barrel where there is no question of any
action: Beliefs were formed as fast as the child heard the testimony and with no con-
sideration of alternatives, much less any thought of rejection.8 Moreover, what we
have accepted when hearing it we may continue to accept when our attention turns
elsewhere. There are, then, states of acceptance as well as events of acceptance. The
states need not result from any action properly so called, and the events need not be
actions.
Suppose, however, that there is an event of acceptance of a proposition. Is it an
action of, or entailing, belief formation and under voluntary control? Imagine that
someone I respect makes a claim I know is controversial and that, before accepting
it, I think about it a moment. What have I "done"? One thing I may have done is
abstain from withholding belief formation, something I would have done if the claim
did not seem so plausible. The formative role of this withholding would not entail
that belief formation is positively voluntary or even an action. Recall sneezing; what
is withholdable at will need not be action. Moreover, suppose it is by not with-
holding belief of p that we come to believe it; this would be a case of indirect
voluntariness of belief formation, which is not controversial in such cases. Perhaps
we use "acceptance" here to designate passing from consideration of the proposition
 Doxastic Voluntarism and the Ethics of Belief 97

to believing it. This change of state need not be or represent an action; "acceptance"
may designate manifestation, rather than production, of belief formation, or perhaps
the nonvoluntarily formed belief itself: the product as opposed to the process, one
might say. Calling such events acceptance is at least as plausibly accounted for on
the hypothesis that a state of acceptance is a nonbehavioral product of the events that
occur in considering evidence as it is on the hypothesis that acceptance is a directly
voluntary action of belief formation.
Indeed, is there anything one can do at will that entails the formation of a belief?
This is what "acceptance" conceived as an act of belief formation should do.9 But if
we note that one can accept a proposition for the sake of argument without believing
it, and that, at least through self-deception,10 people can (sincerely) verbally accept
a proposition without believing it, there is room for acts of acceptance whose per-
formance does not entail belief formation and hence does not support doxastic
voluntarism in the way it may be claimed to. With propositions, as with people,
acceptance is possible without embrace, and embrace is possible without prior
acceptance. But if acceptance does not entail embrace, its occurrence does not sig-
nificantly support voluntarism. Granted, in accepting a proposition for the sake of
argument or a hypothesis for the sake of seeing what one learns from testing it, we
pass into a state of resolution to do certain things, such as to draw inferences, that
also characterize believing. But drawing inferences in this way does not entail be-
lieving, and it is a mistake to attribute beliefs on the basis of the results of acceptance
where this is the kind of action we can perform at will.
If assent (to) is contrasted with dissent (from), some similar points hold. But as-
sent differs from acceptance. There is no dispositional state of assenting, whereas
there is such a state of accepting: The term "assent" designates a kind of action—or
at least a kind of event. Perhaps, then, assenting to a proposition is a candidate for
voluntarily—and directly—forming a belief. By contrast with accepting, however,
assenting is more behavioral, more a matter of something like saying "yes" than of
changing in cognitive condition, though it may immediately cause such change. If
that is so, there is no difficulty in regarding assent as action under voluntary control;
but there is also little plausibility in taking it to entail belief formation.
There may be cases in which withholding assent suffices to prevent belief forma-
tion, and these are significant for such Cartesian purposes as preventing the forma-
tion of false beliefs. But that is a different matter, and the point lends no support to
a conception of assent that can sustain voluntarism. There may be times, as where
one is deliberating about evidence, when one is in a withholding mode and will not
believe the proposition in question unless one does assent to it; here, assenting is
necessary for belief formation. But it does follow (and seems false) that its occur-
rence entails belief formation. Opening a gate need not bring anything in; and if it
does, the entry is not accomplished at will but as a result of the forces that move the
entrant.
If we think of positive cases, such as those in which, having considered the evi-
dence, we find ourselves convinced that a proposition is true, the naturalness of call-
ing these instances of assenting—for instance, of deciding that p is, after all, true—
is readily explained as a manifestation of passing into a state that has as one of its
main characteristics a tendency to assent to p if the question whether/) is true comes
98 Robert Audi

up. We perhaps picture ourselves as agents of belief formation when what we have
really done is to create (or enter) circumstances in which it occurs as a nonvoluntary
response to a pattern of evidence. We have accepted />, but not because we assented
to it or performed an act of acceptance. Rather, the pattern of evidence produced the
belief; the belief is more like a response to external grounds than a result of an inter-
nal volitive thrust. The belief formation is like the conviction that underlies a com-
petent judge's verdict; it is not like the giving of the verdict: The latter is volitional,
the conviction is prevolitional.
If what I have said is correct, there is reason to think that although belief forma-
tion is in some instances under indirect voluntary control, both positive and nega-
tive, and may be in special cases under direct negative control, it is not under direct
positive control. I make no claim for conclusiveness in the nonvoluntarist (partial)
account of belief formation presented in this section to deal with the kinds of cases
that best support doxastic voluntarism. Indeed, I grant that despite all I have said there
is still a tendency to think that in the context of our willfully concentrating on evi-
dence, particularly when it seems to us equally divided, the cognitive upshot of the
process, namely belief formation, looks much like an action we might take on the
basis of practical reasoning.11 Why not say that for certain cases, especially those in
which the reasons leave us free to act or not and to believe or not, it is by an exercise
of will that we act or believe, and here believing is voluntary in much the same way
as action?12 At this point, we can best progress by framing some guiding theory.

II. Intellect and Will, Belief and Action

Recall the very old idea that the function of the intellect is to pursue truth, and the
function of the will is to pursue the good.13 In parts of Plato and Aristotle and later
in Aquinas, intellect and will may at times have been conceived, in somewhat
homuncular fashion, as faculties capable of a kind of action (pursuit of their proper
ends), but the basic idea does not require hypostatization and can be preserved by
simply thinking of different aspects of our intellectual and behavioral capacities and
of differing standards appropriate to each.14 There have been many versions of this
idea, and there are numerous views about how intellect and will are related, includ-
ing the famous Socratic doctrine that (baldly stated) knowledge of the good implies
willing its realization.15
In contemporary philosophy, the idea seems to have appeared (somewhat attenu-
ated) in the notion that belief and desire (and more generally the theoretical and prac-
tical attitudes) have different directions of fit to the world.16 Roughly, beliefs and
other theoretical attitudes (those with truth-valued objects) succeed (objectively speak-
ing) when they fit the world; desires and other practical attitudes (those with the kinds
of objects intentions have) succeed when the world, chiefly the future world, fits them.
This contrast between the two directions of fit is not equivalent to the related view
that belief "aims" at truth or that desire "aims" at the good. The point is that different
standards of appraisal are appropriate. The point does not commit us to attributing
any sort of content to either kind of attitude, nor even to believers' ascribing truth to
every proposition they believe, though it must be granted that the metaphorical con-
 Doxastic Voluntarism and the Ethics of Belief 99

ception of the truth of a belief as a kind of success invites the idea that belief aims at
truth in the first place.17
To be sure, true but unjustified beliefs, and similarly, realized but irrational de-
sires, are not unqualifiedly successful; one might call them internally unsuccessful
insofar as one thinks that despite their positive status considered from the agent's
perspective, they violate standards that the person accepts or would, if sufficiently
rational and adequately reflective, accept. Still, we may at least say that being false
is clearly a kind of defect in beliefs, whereas being unrealized (and thereby arguably
having a counterpart of falsehood) is not a defect in desires. Far from it: An unreal-
ized desire is often a spur that plays a major role in one's realization of its object.
What has generally not been noticed is that there is apparently a connection be-
tween the kinds of contents (or objects, in a different terminology) appropriate to the
theoretical and practical attitudes and, on the other hand, their different directions of
fit and, more broadly, their different functions as basic elements of intellect and will,
respectively. Despite these differences between beliefs and desires in function and
directions of fit, a number of philosophers have taken them to have the same kinds
of contents—truth-valued ones—and many other philosophers have been at least
casual about distinguishing these kinds of contents.18 Perhaps it need not be argued
that (de dicto) beliefs have prepositional contents, in the sense of items that are true
or false (we can leave open whether some kind of linguistic entity could serve here).
But something must be said about contents of desires, intentions, and other practical
attitudes.19
The first thing to be stressed here is that truth value may always be sensibly as-
cribed to beliefs, but never to desires. Connected with this, we express the content of
desires using (primarily) infinitives or subjunctives, not indicatives: I can want to
talk with you, or that we talk, but I cannot want that I am talking with you, that I will
talk with you, or even that it is true that I am talking with you. Similarly for intend-
ing (though here the content is arguably always first-personal): I intend to talk with
you (myself); and if I intend that we talk, this is apparently a matter of intending to
do things (myself) such as invite you to talk.
There is a parallel set of data regarding theoretical and practical reasons, roughly
reasons for believing and reasons for action (or at least for desire). A reason for ac-
tion is always expressible in a phrase of the form of "in order to A," where "A" ranges
over action-types, though a reason for action may also be expressed in other ways.20
A reason for believing is never so expressible. A reason why one believes might be
claimed to be expressible in that way; a voluntarist might say, for instance, that it
was in order to save his life that James believed he could jump across a precipice.
But this apparently makes good sense only where we think of the "in order to" as
expressing a reason to cause or sustain the belief. That is a reason for action.
The positive side of reasons for believing is more complex. If a reason for believ-
ing is conceived strictly, as expressible in a sentence of the form of "My reason for
believing that p is that q," then such a reason is always a proposition (or similar en-
tity, such as, on one conception, a fact). This is because q must express something
believed by the speaker (assuming the overall statement is true). Moreover, if what
is cited as a reason for believing p really is a reason for believing it, one can at least
sensibly ask whether it supports (e.g., entails, explains, or probabilistically implies)
100 Robert Audi

p. There are, however, grounds for believing that are not reasons in the strict sense
(though I do not mean to overstress this distinction, since the proposition that a ground
obtains is a reason). Simply seeing a tornado is a ground, not (I think) strictly a rea-
son, for believing there really is one. We can call seeing it a reason, particularly if
we are thinking of the point that once we believe we see it, we do have a reason, in
the strict sense, for believing there is a tornado: namely, the proposition that we see
it.21 In fact, we would not normally cite just our seeing a tornado as a reason for
believing there is one, but rather that we see it. Nonpropositional grounds do not seem
to serve as reasons for believing (except in the sense of reasons why it is that one
believes) until they are believed to obtain. They are evidences; reasons for believing
are prepositional expressions of evidence.
In any event, we may say of both reasons proper and of grounds that when they
figure as reasons for believing, it is as true or false or, in the case of nonpropositional
grounds, as obtaining states of affairs of the kind a true proposition expresses. Rea-
sons for believing (and for other theoretical attitudes) are in this way unlike practical
reasons, which are not truth-valued though they express a content—for instance that
one's friend receive an honor—that may be said to be realized or unrealized.

III. Practical and Theoretical Reasons

With this much theory in view, we can see something important about voluntarism.
But I must approach the point gradually, first by concentrating on the will and then
by a parallel consideration of the intellect.
Suppose we think of the will as functioning to change the world in the direction
of either some good (whether objective or subjective does not matter at this level of
generality) or at least some desired state of affairs. And suppose we also conceive of
the will as guided in exercising this function by practical reasons and as successful
when its functioning (by its acts, we might say) produces the changes. If the guiding
reasons are practical, they must point to acts. This is exactly what practical reasons
do. In the case of intention, and equally for action-desires, the content of practical
reasons is expressible by an infinitive clause designating an action-type, say "to sig-
nal my vote." This clause expresses my reason for raising my hand, my will pro-
duces that action, and the action changes the world.
By contrast, think of the intellect as functioning to provide a true representation
of the world, or the part of it relevant to the subject, at least to the subject's survival.
The intellect is properly guided by theoretical reasons and is (externally and objec-
tively) successful insofar as it represents the world correctly. If the guiding reasons
are theoretical, they must point to true propositions. (Whether the pointing must be
conceived in terms of epistemically internalist or externalist standards may be left
open here.) If their content is propositional, they can do this; if it is not (at least in the
indirect way a ground is, since it also serves as evidence for a proposition)—if, for
example, it is practical, as where "to make life bearable" expresses a presumptive
reason for believing one's friend is honest—they cannot. Thus, suppose my reason
for believing there is a tornado nearby is that a siren has sounded. This clause ex-
presses a proposition that constitutes a reason for my belief; my coming to believe
 Doxastic Voluntarism and the Ethics of Belief 101

that evidential proposition produces, and my believing it causally grounds, my be-

lieving that there is a tornado; and my intellectual state changes to reflect the world.
The first thing that may come to mind in the comparison is the different direc-
tions of causation: from the will to the world in the first case, and from the world to
the intellect in the second. This is as it should be: If the will succeeds when it changes
the world, it must produce a causal condition sufficient for the change; and if the
intellect succeeds when it truly represents the world, it must surely achieve such a
representation through a causal connection running from the world to it.22 If, how-
ever, the broad perspective on intellect and will so far presented is correct, there is
something more. It bears on the causal contrast just drawn but has more to do with
the ontology of intentional attitudes themselves.
Actions are events, in the ordinary sense in which the occurrence of an event en-
tails that of a change. Beliefs are not events (though their formation is). To believe is
not to do something or to change anything; nor does having a belief over time entail
changing over the time in question in any way related to the belief. Beliefs, then, are
not actions. If not, then they are not actions of will, as volitions (on the most com-
mon conception) are, nor are beliefs even basic actions of any sort, as elementary
bodily actions are. One form of voluntarism, then—behavioral voluntarism—is a clear
failure: It treats believing as action, when believing is not even of the right category
to be action.
There is a further implication of this contrast between action and belief, on the
plausible (though controversial) assumption that causes of events are themselves
events.23 If this assumption is true, and if the "acts" of the will are practical interven-
tions in the world that can change it in the desired direction, then believing is again
the wrong sort of thing to be an act of will or even a basic action performed "at will."
Not being an event or change, believing is not a candidate to change the world.24 To
see the force of this, suppose the will could produce nothing but beliefs. It would
have no impact on the external world and could at best pursue only cognitive goods,
dispositional states that, by themselves, apart from events that activate them, do not
produce actions or other events. It might provide one with an excellent map of the
world; but a map, however well it may guide action, cannot produce it.
A voluntarist might grant that beliefs themselves cannot be acts of will or basic
acts, but go on to point out that in connection with voluntarism, we have spoken of
the formation of belief as well as of believing. Despite all the references to deciding
to believe and similar phenomena, perhaps the voluntarist's main idea might be said
to be that only formation of belief need be possible at will, and this is an action. This
idea is central for genetic voluntarism, which we must explore more closely.
Let us first disambiguate: There is causing oneself to believe something, and there
is simply coming to believe, in the sense of a belief's forming, as where one sees the
approaching tornado. Neither of these is an action of belief formation. Causing one-
self to believe is a type of action that entails belief formation, but the belief forma-
tion is not a further action, any more than the rising of one's arm is an action beyond
one's raising it. Here the belief formation is the result of an action, not, as the volun-
tarist would have it, something done at will. Moreover, whereas the arm's rising is
the result of something done at will, the belief formation is at least normally the re-
sult only of something done by doing something else, say of inclining oneself to
102 Robert Audi

believe the proposition by turning one's attention to a proponent of it. Tn other cases,
there may seem to be an action of forming the belief that p; for just as one can, as it
were, say "Arm, rise!" and raise one's arm at will, it may seem that one can (in
favorable circumstances) say to oneself, or to one's "intellect," "Believe/?!" and
thereby believe/) at will. But if believing is neither action nor event, this description
does not have a clear sense; using the phrase "form a belief may disguise this point,
but we must not be misled by the phrase. If forming a belief comes to no more than
a belief s forming, there is no reason to consider it an action. If it does come to more,
the only clear behavioral addition plausible in the light of the data invites us to take
it as equivalent to causing oneself to believe. We can understand a number of kinds
of actions of causing oneself to form a belief, and we understand the event of a belief s
forming; but the idea of an action of forming a belief, construed as distinct from both
of these cases, invites confusion as to what belief is.
Here is a different way to see the point. If this supposed action of belief formation
at will is distinct from causing oneself to believe p, there should be a distinction
analogous to that between directly raising one's arm (raising it at will) and indirectly
causing its rising, as with a pulley. Might we say that the difference is between sim-
ply forming a belief at will and, for example, causing its formation by exposing one-
self to a hypnotist who will induce it? There is a significant disanalogy: In the first
(behavioral) case, we have a distinction between directly willed action, the arm rais-
ing, and indirectly willed nonaction, namely the arm's rising; but in the second
(doxastic) case, there is only a difference in how the same kind of event—belief
formation—is produced by the will. The kind of production in question is causing
oneself to believe. The result is not action in either instance. Again, in thinking of
believing as directly voluntary, the voluntarist is apparently imagining belief, or at
least its formation, as an action, perhaps of intellect, in the way that a volition (or
even a basic movement) may be an act of will. This either puts belief in the wrong
category, that of action, or simply begs the question as to whether its formation can
be an action, a question I want now to pursue in another dimension.
The difference between intellect and will (and more broadly, between theoretical
and practical reason) bears on the issue in another way. If belief formation, as dis-
tinct from causing oneself to believe, is an action, what would be the form of a prac-
tical reason for it? (I assume that, for any action, there at least can be a sensible ex-
pression of a practical reason.) The form could not be that of "to believe q" where q
is some proposition the person conceives as somehow related to (perhaps supported
by).p; for this is neither an act-description nor even an event-reporting phrase. It could
not be something like "to cause oneself to be more confident of jumping a preci-
pice"; for although this expresses a reason for causing oneself to believe p, it does
not in any clear sense express a reason for forming the belief that p if that is any
different (e.g., it does not indicate support for the truth of p). The form could not be
"that/? seems true" (or, say, "that/? is well evidenced"); for this does not express an
appropriate content for a practical reason. Given a desire to believe truths, the point
that p seems true would imply that there is a reason to cause oneself to believe p if
one does not already believe it; but then we are back to what no one doubts to be a
kind of action. If, as it appears, there is no plausible candidate for a practical reason
 Doxastic Voluntarism and the Ethics of Belief 103

for forming a belief, as distinct from causing oneself to believe, then we have good
ground for concluding that there is no distinct act of forming a belief at will.
There is one further support for voluntarism that must be addressed here. How
are we to explain the parallel between freedom to act and freedom to believe? I am
free to go to the library and also to work in my study instead; and, after hearing a
student's plausible account of apparent plagiarism, I may be free to believe the stu-
dent or not. Should we not say that in the second case, as in the first, I may do some-
thing—namely, form a belief?251 have two points here, one concerning reasons for
action in contrast with reasons for belief, the other concerning reasons for not with-
holding belief, as opposed to reasons for believing. I take these in turn.
First, the freedom in the case of alternative actions between which one has a genuine
choice is a broadly causal matter—there is no compelling cause operating on one
toward the first action or toward the second (we may leave open whether there must
be a compulsion if determinism is true); by contrast, the freedom in the second case
is evidential—there is no compelling reason for one to believe the proposition or its
negation. It is not behavioral freedom, but a normative kind, determined by the qual-
ity of the evidence. This is not to deny that there is a kind of normative freedom for
actions; but the relevant comparison is with behavioral freedom (the kind realizable
by exercises of will), and the point is that there is no clear doxastic analog of this. (I
can imagine a case in which, after I have had neurologists inducing beliefs in me, I
am told that I am now free to believe what I will; but here "free to believe" indicates
removal of an abnormal cause, not recovery of a volitional ability.)
The contrast I am drawing between freedom to act and to believe is not merely
verbal. It is what we should expect if, as I have argued, action and belief formation
are responses to different kinds of reasons: Those appropriate to belief are not rea-
sons for action at all, and this point concerns their evidential support of a proposi-
tion one might believe, not the absence of a causal power on the part of evidence.
Indeed, it is perfectly acceptable to be causally compelled by evidence, in a way it
is not acceptable to be causally compelled by factors (e.g., threats) producing ac-
tion. Doxastic compulsion by evidence conduces to a true representation of the
world; behavioral compulsion does not in general conduce to realizing anything
good.
Secondly, insofar as the evidential ground for my freedom with respect to believ-
ing the student is indirectly a ground for, as opposed to a cause of, action, the action
may be one of not withholding belief of the proposition I am inclined toward after
hearing the story. If the evidence on both sides is equally good (and not too meager),
I may have reason to cease withholding the belief I prefer, say that there is no plagia-
rism. There is, however, no good reason to think the action I might take here is either
believing or directly forming a belief; it is at best one that causes a belief to form
from my antecedent inclinations toward it. The data in question, then, are readily
(and I think best) explained without assimilating the relevant intellectual freedom to
practical freedom or endorsing any direct version of voluntarism.
In criticizing voluntarism in both its behavioral and its genetic forms, I may have
appeared to deny too much. It should help in concluding this section to indicate some
important points that I grant voluntarism. I recognize the possibility that an act of
104 Robert Audi

doxastic command, analogous to willing to raise one's arm, can cause belief forma-
tion. This is a contingent matter. I also accept (and have elsewhere explored in de-
tail26) a profound parallel between belief and (intentional) action: Both action and
belief formation are normally responses to reasons or grounds—a point crucial in
understanding how we can be responsible for (and sometimes criticizable for) both.
This is not a contingent matter. I would grant, too, that just as there can be weakness
of will, there can be weakness of intellect, in the sense that one can consider the
evidence for p conclusive and still fail to believe p. Indeed, it may be that just as,
when one fears suffering weakness of will, it is appropriate to will to do the deed
favored by one's practical judgment, so when one fears weakness of intellect, it is
appropriate to "will to believe" the proposition favored by the evidence. This anal-
ogy, however, supports at most a kind of indirect voluntarism: the view that one can
sometimes cause belief formation by a mental directive.
It is noteworthy that it is in cases in which one, as it were, needs help in doing a
deed or in believing a proposition that it is most natural to speak of volition or some
similar act of will. For various theoretical reasons, however, the volitionalist tradi-
tion in the theory of action posits volition or trying or some such act of will in the
genesis of every intentional action. Philosophers in this tradition who quite reason-
ably take seriously the analogy between action and belief as responses to reasons
will naturally believe the analogy extends to volitional production and control. If, as
I maintain, volitionalism is mistaken in the theory of action, there is no prima facie
case for positing volitional phenomena as needed to account for how reasons for
believing a proposition yield actual belief of it.27 If reasons for action can yield ac-
tion without a specific volitive thrust, then the analogy between action and belief
should not lead to a search for a counterpart role for volition in belief formation. In
special cases, as where one is deeply resistant to believing ill of someone despite
powerful evidence of wrongdoing, volitional elements may play a role, as they do in
resisting shameful actions. But there is still no reason to consider believing, or even
belief formation, to be an action. That believing is not an action, and that the kinds
of reasons that support beliefs (theoretical reasons) are different from practical rea-
sons, is not a contingent matter.
To be sure, the possibility of causing oneself to form beliefs is important, and
there is no need to deny that there can be good reasons to cause oneself to believe
something. Perhaps in special cases this can even be done at will. But then there
will be some basic action, such as contemplating the relevant proposition in a cer-
tain positive way, that produces the desired state of affairs: believing it. This ac-
tion will cause belief formation but is not an act of belief formation. It is not, then,
what doxastic voluntarists have wanted. They have wanted, minimally, that form-
ing beliefs, if not believing itself, be under direct voluntary control.28 But what is
under direct voluntary control must be an action and as such should admit of prac-
tical reasons for its performance. Belief, which is not even an event, is not an ac-
tion; belief formation is an event but, as distinct from causing oneself to belief, is
not an action. It remains true that there are major similarities between practical and
theoretical reasoning; but the kinds of reasons appropriate to each, and to action
and belief, are different. They have different kinds of content and go with differ-
ent aspects of our rational constitution.
 Doxastic Voluntarism and the Ethics of Belief 105

IV. The Ethics of Belief

It might seem that if I have been right, and neither believing nor forming beliefs is a
case of action, then there is little or no place for an ethics of belief. That is not so. If
we conceive ethics as concerned with proper standards of conduct, and if we allow
that conducting one's intellectual life sloppily may be in a broad sense morally ob-
jectionable, we can speak of an ethics of belief. (Perhaps there can be an "ethics" of
belief even if the sense of "ought" in which we ought to regulate our beliefs is not
strictly moral.) In any case, the results of this paper do not prevent our sustaining a
deontic version of an ethics of belief, specifically, an epistemic deontologism con-
ceived as the view that to be justified in believing that p is for believing p to be
(epistemically) permissible for one. We can say, for instance, that when a proposi-
tion meets certain standards of evidence, we may believe it without violating any
epistemic duties, that is (roughly), without liability to epistemic criticism, and that
there are certain standards of evidence such that unless a proposition meets them,
then, on pain of liability to epistemic criticism, one should not believe it. On this
view, withholding belief at will might sometimes be possible; doing that does not
imply that believing is an action. There can, moreover, be practical reasons for with-
holding a proposition. We can also say that when a proposition is supported by cer-
tain kinds of grounds or meets certain standards, one should try to cause oneself to
believe it if one does not believe it and try to retain belief of it if one does. These are
among the sorts of things one can say in developing an ethics of belief.29
It is important to see, however, that an ethics of belief is possible even for those
who reject epistemic deontologism. One need not think that justification for believ-
ing is any kind of deontic status to hold that there are ethical standards applicable to
regulating both what we believe and how strongly we believe it. One may also hold
(as I am inclined to) that even if a deep and illuminating analysis of justification in
terms of duty is not possible, there may be an instructive equivalence between the
notion of justification for believing and that of epistemic permissibility. It could turn
out, for instance, that our epistemic obligations are (in broad terms) to be properly
attentive both to logic (broadly conceived) and to the evidences of our senses, memory,
consciousness, and reflection, and that what constitutes justification is a kind of con-
formity to these sources.
This attentiveness conception of epistemic obligation suggests that we may also
speak of an ethics of inquiry as well as of appropriate—and significantly overlap-
ping—rational standards determining when belief is acceptable. One demand of both
ethics and rationality would be to seek cogent evidence for (and even sometimes
against) certain of one's beliefs and to try to achieve and maintain confidence levels
appropriate to the evidence one has. A confidence level regarding p may be unjusti-
fied (say, unwarrantedly high) even if there is ample justification for guarded belief
(or some strength of belief) of p. We can thus accommodate degrees of conviction
(and can posit "partial" as well as "full" belief), and a person can be criticizable for
having an inappropriate degree of confidence, say, for high confidence where the
relevant probability is low as well as for unqualifiedly believing something unsup-
ported by grounds or for not believing something supported by excellent undefeated
evidence. None of these standards requires a commitment to voluntarism.30
106 Robert Audi

Given what has emerged in this paper, at least three basic points about the ethics
of belief are appropriate. First, such cognitive regulation as the ethics of belief re-
quires is normally not possible by mere acts of will aimed at producing belief; nor-
mally we achieve it by actions, or (especially) activities, that indirectly affect belief.
For instance, we properly attend to relevant evidence, to which we then respond in
the ways appropriate to rational persons considering grounds for or against various
propositions of concern to them in the context.31 Second, normally, a sound ethics of
belief requires that we try (1) to develop dispositions that lead us to seek (evidential)
grounds in certain matters (say, questions on which people's welfare turns), (2) to
form beliefs when and only when our grounds adequately support them,32 and (3) to
avoid having a degree of conviction at variance with the strength of our grounds.
Third, for many of our beliefs we do not need to seek grounds. Some beliefs are, for
instance, obviously true, and some not obviously true are plainly unimportant.
If we are rational beings, then some of the principles appropriate to a plausible
ethics of belief may correspond to our natural tendencies. Here is a hypothesis about
reasons for believing that seems at once to have normative implications and to be
psychologically plausible: Where I have a (supporting) reason for believing p, then,
as a rational person, if I become aware of its supporting p (say rendering p highly
probable), (1) I will have some tendency to believe/?; (2) if I come to believe/? while
(occurrently) aware of this supporting reason so conceived, I will tend to believe it
on the basis of the reason, even if I also believe it on the basis of some other reason;
and (3) the more rational I am, the stronger these tendencies are. Call this (and its
counterpart for action and for the other propositional attitudes) the responsiveness
presumption: It says in effect that (under certain common conditions) rational per-
sons tend, in proportion to their rationality, to respond both in action and in attitude
(and again in a roughly proportional fashion) to reasons they have for these actions
or attitudes. The tendency can be weak—and a tendency generated by one set of
reasons can, of course, be overridden by one generated by another set. But even a
weak tendency can be unopposed and can be significant even when it is overridden.
As applied to belief, this presumption seems to be one of the truths underlying the
overstated idea (expressed by W. K. Clifford and other proponents of the ethics of
belief) that one should proportion one's conviction to the evidence.33 This is not to
imply that the normative standard associated with the presumption, a standard on
which one is criticizable as insufficiently rational when one's response to one's rea-
sons is inadequate in the indicated way, is itself ethical. The point is rather that the
ethics of belief can express a prima facie moral obligation to abide by a rational stan-
dard. Unlike the idea associated with Clifford, however, the standard in question does
not invite either the view that one has direct voluntary control of one's beliefs or the
assumption that the cogency of evidence is quantifiable in the relevant way or even
that one has direct voluntary control of the degree of conviction one has regarding
propositions one does believe. It does imply, however, that if one fails to tend to
believe p for a reason one (sincerely) offers in support of/?, there should be an expla-
nation for this, such as a momentary lapse in reasoning or the influence of a preju-
dice. A conscientious person aware of this anomaly should be disposed to wonder
why it obtains, for instance, whether it is because one is self-deceived in offering the
reason as such or is merely offering it as a rationalization, or both.34
 Doxastic Voluntarism and the Ethics of Belief 107

If the ethics of belief concerns the morality of actions, and if beliefs are not ac-
tions, then a proponent of the relevant ethical standards need not even posit moral or
immoral beliefs, much less construe justified and unjustified beliefs as morally jus-
tified or unjustified. An ethics of belief may, however, countenance, in special cases,
a moral obligation to induce in oneself certain false or unjustified beliefs or to unseat
certain true or justified ones. The precipice example (from William James) is a case
in point: Even if the evidence is against one's being able to jump across a precipice,
it could be both morally obligatory and behaviorally rational to cause oneself to
believe one can do so. Many people's lives might depend on one's doing it, and
believing one can do it might make the achievement much more likely. The rational-
ity of such self-manipulation is clearest where one would be justified in the belief
when coming to hold it, as where the resulting confidence adds enough strength to
one grounds for the belief to justify holding it. But even if the optimistic belief is not
justified for one before or after one holds it, the action of causing oneself to hold it
may still be rational.
It is important not to catapult the practical justification for inducing a belief into
a sufficient condition for theoretical ("epistemic") justification. Even where one is
obligated to induce a belief in oneself and rationally does so, it does not follow that
the belief itself is justified, and it surely need not be, on the normal presupposition
that what justifies a belief supports its truth (in some way). One could reply that
holding the belief is justified, but this is ambiguous between saying that the belief is
justified, which is false, and saying that causing oneself to hold it is justified, which
is true. Supposing, then, that one could form a belief at will, the reasons for doing so
would not transfer to supporting the belief itself. Practical reasons would not here
double as theoretical ones. A practical reason for inducing a belief is not a support-
ing reason for holding that belief.
If my points about the ethics of belief are sound, they can help to clarify the no-
tion of an epistemically responsible agent. On one plausible approach, such an agent
is roughly equivalent to one who adheres, in an appropriately scrupulous way, to a
sound ethics of belief. If enough is built into the notion of a sound ethics of belief,
one might go on to argue, on analogy with a strong version of virtue ethics, that a
justified belief is one appropriate to an epistemically rational agent, much as a mor-
ally permissible action is one appropriate to a morally virtuous person. My position
in this essay leaves room for an equivalence here. But it would be a mistake to think
we have a notion of a sound ethics of belief, or of an epistemically responsible agent,
that is independent of more basic standards for justified belief, standards tied to a
conception of theoretical reasons for belief, such as those we find in the domain of
sense experience or conceptual reflection or inductive reasoning. Epistemic respon-
sibility depends on, and does not create, standards for justified belief. I take this
point to be neutral among different approaches to analyzing justified belief, say,
foundationalist, coherentist, internalist, externalist, or Bayesian.35
The profound analogies between intellect and will and between theoretical and
practical reason are left intact by the work of this paper. But doxastic voluntarism, in
the sense in which it implies that believing is sometimes an act of will, or even that
forming beliefs is sometimes such an act distinct from causing oneself to believe, is
not. There is a significant role played by the will in much belief formation, and there
108 Robert Audi

are broadly ethical standards, or certainly standards of rationality, appropriate to that

role. When this is understood, we can see why we need not be passive in our intel-
lectual lives: We sometimes ought to believe something we do not or ought not to
believe something we do, and living up to the standards in question may require much
belief-related activity. But intellect and will are very different in their functions and
in the criteria for their successful operation. If the intellect, and with it theoretical
reasons, were not independent of the will, and even of practical reasons, in the way
they are, we as rational agents could not do as well as we do. To change the world in
pursuit of our ends, we need a realistic map of it, and as a guide to practical reason
that map cannot be drawn at our pleasure.36

Notes
1. For valuable discussions of the voluntariness of belief, see Bernard Williams, "Decid-
ing to Believe," Problems of the Self (Cambridge: Cambridge University Press, 1973), 36-
51; Louis P. Pojman, Religious Belief and the Will (London: Routledge, 1986); William P.
Alston, "The Deontological Conception of Epistemic Justification," Epistemic Justification
(Ithaca, N.Y.: Cornell University Press, 1989), 115-52; James Montmarquet, Epistemic Vir-
tue and Doxastic Responsibility (Lanham, Md.: Rowman and Littlefield, 1993); Alvin
Plantinga's treatment of the deontological concept of justification in Warrant: the Current
Debate (New York: Oxford University Press, 1993); and Matthias Steup, "Epistemic Deon-
tology and the Voluntariness of Belief," Acta Analytica 15 (2000): 25-56.
2. His overall conception seems to be one on which the will may in certain cases directly
bring about—or withhold—belief. In Meditation I, he says, "I ought no less carefully to with-
hold my assent from matters which are not entirely certain and indubitable than from those
which appear to me manifestly to be false," and "I ought not the less carefully to refrain from
giving credence to these opinions than to that which is manifestly false," where, for him,
"giving credence" seems roughly equivalent to "assenting" (trans. Haldane and Ross). In Medi-
tation IV, we find him concluding that "as often as I so restrain my will within the limits of
my knowledge that it forms no judgment except on matters which are clearly and distinctly
represented to it by the understanding, I can never be deceived." It is interesting that here he
seems to conceive the will as needing to produce assent (or credence) even where there is a
clear and distinct representation. That is not a commitment of doxastic voluntarism in gen-
eral, if indeed it represents even Descartes's overall view, which, for the most part, attributes
to the will only negative control over belief formation.
3. When I raise my hand at will, do I also bring about its rising at will? I am inclined to
say no, on the ground that I bring it about by doing something, whereas that does not apply to
my raising my hand when I do so at will. If a negative answer is correct here, then that one
can bring about an event at will is at least a very good reason to consider it an action.
4. To be sure, it may be by tightening muscles that one does such things, but once the
control becomes automatic it is like many other basic action-types: It conies to be basic when
one no longer needs to do the deed by way of doing something else. The muscles tighten
when one wants to stop breathing directly on the latter volitive "command."
5. Supporting arguments are given in my "Volition and Agency," Action, Intention, and
Reason (Ithaca, N.Y.: Cornell University Press, 1993), which shows how basic action is pos-
sible without distinct internal actions of volition or other such executive thrusts.
6. There are uses of "decide" in which deciding to believe is a way of coming to believe
that is at least close to the way sketched here. In describing a traditional view in the philoso-
phy of science, for instance, Bas C. van Fraassen speaks of "what we may rationally decide
or come to believe." See "Belief and the Will," Journal of Philosophy 81, no. 5 (1984): 235-
56, 255 (italics added). Given how he uses the terminology of deciding to believe and that his
reason for calling his view in the paper voluntarist is that "it makes judgment in general and
subjective probability in particular, a matter of cognitive commitment, intention, engagement"
 Doxastic Voluntarism and the Ethics of Belief 109

(p. 256), I doubt that his conception of deciding to believe (or of belief) requires (direct)
voluntarism. The relevant kind of engagement is accommodated by a number of points in
section IV.
7. Here it may be instructive to note that just as the reasons for a decision to do some-
thing are characteristically (and perhaps necessarily) the reasons for doing it and not rea-
sons for the higher order act of deciding, reasons for deciding to believe, in the standard
cases commonly so-called, are reasons supporting the proposition believed and not—un-
less perhaps this trivially follows—for an action of deciding to believe that proposition.
The former point about decision is argued by Thomas Pink in "Reason and Agency," Pro-
ceedings of the Aristotelian Society (1997): 263-80. Related points are made in sections II
and III.
8. Where the testimony is by someone sufficiently trusted, this is common even for ma-
ture adults, but my case does not depend on that point. It is defended in my "The Place of
Testimony in the Fabric of Knowledge and Justification," American Philosophical Quarterly
34 (1997): 405-22. For a different view on the kind of scrutiny appropriate to believing on
testimony, see Elizabeth Fricker, "Against Gullibility," in Knowing from Words, B. K. Matilal
and A. Chakrabarti, eds. (Dordrecht, The Netherlands: Kluwer, 1994).
9. For an account of acceptance apparently intended to fill this bill, see William P. Alston,
"Belief, Acceptance, and Religious Faith," in Faith, Freedom, and Rationality, Jeff Jordan and
Daniel Howard-Synder, eds. (Lanham, Md.: Rownian and Littlefield, 1996). Another useful
discussion is provided by Raimo Tuomela, "A Note on Acceptance and Belief" (forthcoming).
10. An account of self-deception that supports this is given in my "Self-Deception and
Practical Reasoning," Canadian Journal of Philosophy 19 (1989): 246-66.
11. Carl Ginet, for example, notes the apparent voluntariness of belief where conflicting
evidence or one's own uncertainty. "Suppose that, as I am deliberating about whether to trust
Sally's memory or mine or neither regarding the population of Syracuse, she offers to bring
me breakfast in bed if I decide to trust her memory . . . Might that not help me to decide to
believe the figure she remembers?" See "Contra Reliabilism," The Monist 65, no. 2 (1985):
156-87.
12. Montmarquet, Epistemic Virtue and Doxastic Responsibility, is among those who have
stressed this kind of analogy.
13. As Aquinas puts it at one point, "The object of the reason is the true . . . the object of
the appetitive power is the appetitive good." Summa Theologiae, question 60, article 1.
14. One pertinent thing Aquinas held is that "An act is voluntary when it is an operation
of reason" (Summa Theologiae la 2ae q6 a3, p. 56). To be sure "act" may here designate an
exercise of the proper function of reason and need not be equivalent to "action" in the sense
relevant to voluntarism, but here I leave open matters of detailed Thomistic interpretation.
15. This kind of view survives in ethical theory under the name "motivational internalism."
For a detailed appraisal and many references to relevant literature, see my "Moral Judgment
and Reasons for Action," Moral Knowledge and Ethical Character (New York: Oxford
University Press, 1997).
16. This version of the contrast is apparently due to Elizabeth Anscombe. See Intention
(Oxford: Blackwell, 1956).
17. Williams, "Deciding to Believe," Problems of the Self, is among many who treat be-
lief as aiming at truth. He says that "belief aims at truth" (p. 136) and adds that this is in part
to say that "to believe p is to believe p is true" (p. 137). I make a case against the latter claim
in chap. 3 of Moral Knowledge. For a useful discussion of voluntarism that both criticizes
and draws on Williams' treatment, see Dion Scott-Kakures, "On Belief and the Captivity of
the Will," Philosophy and Phenomenological Research 54 (1994): 77-103; and for a valu-
able critique of this paper, see Dana Radcliffe, "Scott-Kakures on Believing at Will," Phi-
losophy and Phenomenological Research 57 (1997): 145-51.
18. Donald Davidson, for example, has taken intending to be a kind of belief. See "In-
tending," Essays on Actions and Events (Oxford: Oxford University Press, 1980). Others have
supposed there are attitudes (sometimes called fesires) with the function of both belief and
desire.
110 Robert Audi

19. In Thinking and Doing (Dordrecht, The Netherlands: Reidel, 1975) Hector-Neri
Castafieda construed the objects of the practical attitudes as "practitions," which he contrasted
with propositions as objects of the theoretical attitudes. Although 1 do not adopt his terminol-
ogy (or his specific views on this issue), I have benefited from his far-reaching work on the
ontology of the intentional attitudes.
20. We sometimes say, in answering "What was your reason for doing that?," "I be-
lieved . . ." or (especially) "I wanted . . .", I call these attitudes reason states and think it
wise to distinguish them from reasons proper, which are their contents. A partial account
of this distinction is given in my "Acting for Reasons," Philosophical Review (1986). That
paper also deals with the special case of things we do for their own sake, for which "in
order to" is perhaps not entirely appropriate in giving an explanation. In any event, I sug-
gest that where a belief provides a reason for acting, it does so at least in part because there
is a suitably related want or a reason for having such a want, as where one's reason for
telephoning is given by citing a belief that one has too little time for e-mail.
21. One epistemologist who uses "reason" to encompass grounds is Marshall Swain, Rea-
sons and Knowledge (Ithaca, N.Y.: Cornell University Press, 1981). A sorting out of differ-
ent uses of "reason" is given in my "Reasons for Believing," Routledge Encyclopedia of Phi-
losophy (1998).
22. We cannot say "causing it"; beliefs about the future are not caused by future events.
But there is still a causal connection: from causes of the relevant events, such as my decision
to do something tomorrow, to the belief representing them.
23. The main implicit contrast is between causes of events and causal sustainers of non-
events, such as beliefs or desires. Causal sustainers need not be events. That causes of events
must be events has been widely defended. One defense and a number of references are given
in my Practical Reasoning (London: Routledge, 1989). For a recent defense of agent causa-
tion, see Randolph Clarke, "Toward a Credible Agent-Causal Account of Free Will," Nous
27 (1993). Supposing there is agent causation, however, it would still seem that the acts of
will caused by the agent must be events to affect the future in the relevant ways.
24. To be sure, if a belief is added to my cognitive inventory, the world is changed; but /
(or some event) would produce this change, not the belief. The belief cannot even change the
world in the direction of one's believing more truths (though it can, given other factors, guide
action).
25. Matthias Steup has emphasized such analogies. See, for example "A Defense of
Internalism," in Theory of Knowledge, 2d ed., Louis P. Pojman, ed. (Belmont, Calif.:
Wadsworth, 1997).
26. In, for example, chap. 3 of Moral Knowledge and Ethical Character (Oxford: Ox-
ford University Press, 1997).
27. In "Volition and Agency," in my Action, Intention, and Reason, I have explicated
and criticized various forms of volitionalism in detail.
28. This is does not apply to Montmarquet, Epistemic Virtue, at least if I am correct in
thinking that his weak version of voluntarism is at least largely consistent with the position
of this paper. See chap. 5 on doxastic voluntariness, especially p. 83.
29. Despite Descartes's sometimes appearing to hold a direct doxastic voluntarism, I do
not believe that internalization of any of his Rules for the Direction of the Mind is precluded
by the critique of voluntarism given in this paper.
30. I am indebted to Isaac Levi for urging me to accommodate distinctions of cognitive
degree and standards for inquiry as well as for appraisal of standing beliefs. I should add that
my approach is neutral with respect to another distinction he emphasizes: between treating
belief itself as a focus of appraisal and treating change of belief as a focus. For a partial state-
ment of his theory on these matters, see The Fixation of Belief and Its Undoing (Cambridge:
Cambridge University Press, 1991). The distinction between commitment and performance
(6-9) and among types of belief change (esp. pp. 64-70) are pertinent here.
31. For instructive discussion of difficulties with construing belief as under voluntary
control see Alston, "Deontological Conception," Epistemic Justification. For a partial reply
and contrasting view, see Matthias Steup, "A Defense of Internalism" and "The Deontic Con-
 Doxastic Voluntarism and the Ethics of Belief 111

ception of Epistemic Justification," Philosophical Studies 53 (1988): 65-84. By contrast with

my treatment of doxastic voluntarism, especially as a position in the philosophy of mind,
Alston's concentrates on empirical and apparently contingent difficulties with it.
32. Condition (b) is hedged because in certain trivial matters, or in certain cases where
one is scrutinizing one's beliefs and ready to revise one's view, one may not need to have
adequate grounds (as opposed to the absence of negative grounds). As to why the condition
does not give a sufficient as well as a necessary condition, the point is that it is not reasonable
to form all the beliefs one can that are adequately supported by one's grounds, say that there
are fewer than a million chairs in the room, fewer than two million, fewer than three million,
and so forth.
33. W. K. Clifford's view is set forth in his famous essay on the ethics of belief in his
Lectures and Essays (1879), reprinted in The Theory of Knowledge, Louis P. Pojman, ed.
(Belmont, Calif.: Wadsworth, 1993). For discussion of some specific principles one might
adopt in living up to a broadly Cliffordian ethic, see van Fraassen, "Belief and the Will,"
Journal of Philosophy, and his sequel, "Belief and the Problem of Ulysses and the Sirens,"
Philosophical Studies 77 (1995), 7-37.
34. I have discussed the ethics of belief, in connection with both self-deception and ratio-
nalization, in chap. 6 of Moral Knowledge and Ethical Character, and the preceding two
paragraphs draw on some of the points made there.
35. I omit consideration of knowledge here. Insofar as it is an external notion, the con-
cept of epistemic responsibility is less likely to provide a basis for understanding it than in
the case of justification; if knowing entails being justified, then the points suggested in the
text may be applied to knowledge insofar justification is an essential element in it.
36. This essay was originally written for this volume but appeared earlier in Facta Philo-
sophica 1.1 (1996), 87-109, and I thank the editors for permission to reprint it. Earlier ver-
sions were presented at Dartmouth College, the University of Helsinki, and the University of
New Mexico, and I thank the audiences for many helpful responses. For comments on one or
another draft I thank Elizabeth Fricker, Bernard Gert, Isaac Levi, and, especially, Matthias
Steup.
This page intentionally left blank
Part III

EPISTEMIC DEONTOLOGY AND

THE INTERNALITY OF JUSTIFICATION
This page intentionally left blank
 7

Internalism Exposed

ALVIN GOLDMAN

In recent decades, epistemology has witnessed the development and growth of

externalist theories of knowledge and justification.1 Critics of externalism have fo-
cused a bright spotlight on this approach and judged it unsuitable for realizing the
true and original goals of epistemology. Their own favored approach, internalism, is
defended as a preferable approach to the traditional concept of epistemic justifica-
tion.2 In this essay, I turn the spotlight toward internalism and its most prominent
rationale. Fundamental problems that lie at the core of internalism will be revealed
and the viability of its most popular rationale will be challenged. Although particu-
lar internalist theories such as (internalist) foundationalism and coherentism will
occasionally be discussed, those specific theories are not my primary concern. The
principal concern is the general architecture of internalism, and the attempt to justify
this architecture by appeal to a certain conception of what justification consists in.

I. Deontology, Access, and Internalism

I begin with a certain rationale for internalism that has widespread support. It can be
reconstructed in three steps:
1. The guidance-deontological conception of justification is posited.
2. A certain constraint on the determiners of justification is derived from the guidance-
deontological conception, that is, the constraint that all justification-determiners must
be accessible to, or knowable by, the epistemic agent.
3. The accessibility or knowability constraint is taken to imply that only internal con-
ditions qualify as legitimate determiners of justification. So justification must be a
purely internal affair.3

115
116 Alvin Goldman

What motivates or underlies this rationale for internalism? Historically, one cen-
tral aim of epistemology is to guide or direct our intellectual conduct, an aim ex-
pressed in Descartes's title, "Rules for the Direction of the Mind." Among contem-
porary writers, John Pollock expresses the idea this way:
I have taken the fundamental problem of epistemology to be that of deciding what to
believe. Epistemic justification, as I use the term, is concerned with this problem.
Considerations of epistemic justification guide us in determining what to believe. We
might call this the "belief-guiding" or "reason-guiding" sense of "justification." (Pol-
lock 1986, p. 10)

The guidance conception of justification is commonly paired with the deontological

conception of justification. Locke wrote of a person's "duty as a rational creature,"4
and the theme of epistemic duty or responsibility has been echoed by many contem-
porary epistemologists, including Laurence BonJour, Roderick Chisholm, Richard
Feldman, Carl Ginet, Hilary Kornblith, Paul Moser, and Matthias Steup.5 Chisholm
defines cousins of the concept of justification in terms of the relation "more reason-
able than," and he reexpresses the relation "p is more reasonable than q for S at f by
saying: "S is so situated at t that his intellectual requirement, his responsibility as an
intellectual being, is better fulfilled by p than by q."6 Similarly, Feldman says that
one's epistemic duty is to "believe what is supported or justified by one's evidence
and to avoid believing what is not supported by one's evidence" (Chisholm 1989,
254).
The guidance and deontological conceptions of justification are intimately related
because the deontological conception, at least when paired with the guidance con-
ception, considers it a person's epistemic duty to guide his doxastic attitudes by his
evidence or by whatever factors determine the justificational status of a proposition
at a given time. Epistemic deontologists commonly maintain that being justified in
believing a proposition p consists in being (intellectually) required or permitted to
believe p; and being unjustified in believing/) consists in not being permitted, or being
forbidden, to believe/?. When a person is unjustified in believing a proposition, it is
his duty not to believe it.
It is possible to separate the deontological conception from the guidance idea. In
ethical theory, a distinction has been drawn between accounts of moral duty that aim
to specify what makes actions right and accounts of moral duty that aim to provide
practical decision procedures for what to do.7 If an account simply aims at the first
desideratum, it need not aspire to be usable as a decision guide. Similarly, accounts
of epistemic duty need not necessarily be intended as decision guides. However, when
the deontological conception is used as a rationale for epistemic internalism of the
sort I am sketching, it does incorporate the guidance conception. Only if the guid-
ance conception is incorporated can the argument proceed along the intended lines
to the accessibility constraint and from there to internalism. This is why I shall hence-
forth speak of the "guidance-deontological" (GD) conception of justification.
I turn now to the second step of the argument for internalism. Following William
Alston, I shall use the term "justifiers" for facts or states of affairs that determine
the justificational status of a belief or the epistemic status a proposition has for an
epistemic agent.8 In other words, justifiers determine whether or not a proposition
 Internalism Exposed 117

is justified for an epistemic agent at a given time. It seems to follow naturally from
the GD conception of justification that a certain constraint must be placed on the sorts
of facts or states of affairs that qualify as justifiers. If a person is going to avoid vio-
lating his epistemic duty, he must know, or be able to find out, what his duty requires.
By "know," in this context, I only mean to have an accurate, or true, belief. I do not
mean to have a justified true belief (or whatever else is entailed by the richer concept
of knowledge). Admittedly, it might be possible to avoid violating one's duties by
chance, without knowing (having true beliefs about) what one's duties are. As a prac-
tical matter, however, it is not feasible to conform to duty on a regular and consistent
basis without knowing what items of conduct constitute those duties. Thus, if you are
going to choose your beliefs and abstentions from belief in accordance with your
justificational requirements, the facts that make you justified or unjustified in believ-
ing a certain proposition at a given time must be facts that you are capable of know-
ing, at that time, to hold or not to hold. There is an intimate connection, then, between
the GD conception of justification and the requirement that justifiers must be acces-
sible to, or knowable by, the agent at the time of belief. If you cannot accurately as-
certain your epistemic duty at a given time, how can you be expected to execute that
duty, and how can you reasonably be held responsible for executing that duty?9
The knowability constraint on justifiers that flows from the GD conception may
be formulated as follows:

KJ The only facts that qualify as justifiers of an agent's believing p at time t

are facts that the agent can readily know, at t, to obtain or not to obtain.

How can an agent readily know whether candidate justifiers obtain or do not obtain?
Presumably, the agent must have a way of determining, for any candidate class of
justifiers, whether or not they obtain. Such a way of knowing must be reliable, that
is, it must generate beliefs about the presence or absence of justifiers that are usually
(invariably?) correct. Otherwise, the agent will often be mistaken about what his
epistemic duty requires. The way of knowing must also be "powerful," in the sense
that when justifiers obtain, it is likely (certain?) that the agent will believe that they
obtain; at least he will believe this if he reflects on the matter or otherwise inquires
into it.10 As we shall soon see, internalists typically impose additional restrictions on
how justifiers may be known. But the minimal, generic version of KJ simply requires
justifiers to be the sorts of facts that agents have some way of knowing. In other words,
justification-conferring facts must be the sorts of facts whose presence or absence is
"accessible" to agents."
Given the KJ constraint on justifiers, it becomes fairly obvious why internalism
about justification is so attractive. Whereas external facts are facts that a cognitive
agent might not be in a position to know about, internal facts are presumably the sorts
of conditions that a cognitive agent can readily determine. So internal facts seem to
be the right sorts of candidates for justifiers. This consideration leads to the third step
of our rationale for internalism. Only internal facts qualify as justifiers because they
are the only ones that satisfy the KJ constraint—at least so internalists suppose.
One possible way to criticize this rationale for internalism is to challenge the GD
conception directly. This could be done, for example, by arguing that the GD con-
118 Alvin Goldman

ception of justification presupposes the dubious thesis of doxastic voluntarism, the

thesis that doxastic attitudes can be "guided" by deliberate choices or acts of will.
This criticism is developed by William Alston,12 and I have sympathy with many of
his points. However, the voluntarism argument against the GD conception is disputed
by Feldman and John Heil, among others.13 Feldman, for example, argues that epistemic
deontologism is not wedded to the assumption of doxastic voluntarism. Many obli-
gations remain in force, he points out, even when an agent lacks the ability to dis-
charge them. A person is still legally obligated to repay a debt even when his finan-
cial situation makes him unable to repay it. Perhaps epistemic obligations have
analogous properties.14 Since the complex topic of doxastic voluntarism would re-
quire article-length treatment in its own right, I set this issue aside and confine my
attention to other issues. Although I do not accept the GD conception of justifica-
tion, I take it as given for purposes of the present discussion and explore where it
leads. In any case, what is ultimately crucial for internalism is the accessibility re-
quirement that the GD conception hopes to rationalize. Even if the GD conception
fails to provide a good rationale, internalism would be viable if some other rationale
could be provided for a suitable accessibility requirement.

II. Direct Knowability and Strong Internalism

The initial KJ constraint was formulated in terms of knowability plain and simple,
but proponents of internalism often add the further qualification that determinants
of justification must be directly knowable by the cognitive agent. Ginet, for example,
writes as follows:
Every one of every set of facts about S's position that minimally suffices to make S, at
a given time, justified in being confident that p must be directly recognizable to S at
that time. (p. 34)

Similarly, Chisholm writes:

The concept of epistemic justification is ... internal and immediate in that one can
find out directly, by reflection, what one is justified in believing at any time.ls

Thus, Ginet and Chisholm do not endorse just the minimal KJ constraint as earlier
formulated, but a more-restrictive version, which might be written as follows:

KJdir The only facts that qualify as justifiers of an agent's believing p at time
t are facts that the agent can readily know directly, at t, to obtain or not
to obtain.

An initial problem arising from KJdir is this: What warrants the imposition of KJdir as
opposed to the looser constraint, KJ? KJ was derived from the GD conception on the
grounds that one cannot reasonably be expected to comply with epistemic duties
unless one knows what those duties are. How does such an argument warrant the
further conclusion that direct knowledge of justification must be available? Even
indirect knowledge (whatever that is) would enable an agent to comply with his
 Internalism Exposed 119

epistemic duties. So the second step of the argument for internalism cannot properly
be revised to feature KJdir in place of KJ. Proponents of KJdir might reply that direct
forms of knowledge are more powerful than indirect knowledge, but this reply is
unconvincing. The power requirement was already built into the original version of
KJ, and it is unclear how directness adds anything of significance on that score.
Whether KJdir can be derived from GD is a serious problem, because the argument
for internalism rests on something like the directness qualification; I shall say more
about this later. For now I set this point aside to explore where KJdir leads.
What modes of knowledge count as direct? At least one form of direct knowledge
is introspection. A reason for thinking that introspection is what Chisholm means by
direct knowledge is that he restricts all determiners of justification to conscious states:
A consequence of our "internalistic" theory of knowledge is that, if one is subject to
an epistemic requirement at any time, then this requirement is imposed by the con-
scious state in which one happens to find oneself at that time.16

Since he restricts justifiers to conscious states, it is plausible to assume that direct

knowledge, for Chisholm, means introspective knowledge, and knowledge by "re-
flection" coincides with knowledge by introspection.17 At least in the case of
Chisholm, then, KJdir might be replaced by KJim:

KJint The only facts that qualify as justifiers of an agent's believing p at time
t are facts that the agent can readily know by introspection, at t, to ob-
tain or not to obtain.

Now the only facts that an agent can know by introspection are facts concerning
what conscious states he is (or is not) currently in, so these are the only sorts of
facts that qualify as justifiers under KJint. This form of internalism may be called
strong internalism:

SI Only facts concerning what conscious states an agent is in at time t are

justifiers of the agent's beliefs at t.

Strong internalism, however, is an unacceptable approach to justification, for it has

serious, skepticism-breeding, consequences. This is demonstrated by the problem of
stored beliefs. At any given time, the vast majority of one's beliefs are stored in
memory rather than are occurrent or active. Beliefs about personal data (e.g., one's
social security number), world history, geography, or the institutional affiliations of
one's professional colleagues are almost all stored rather than occurrent at a given
moment. Furthermore, for almost any of these beliefs, one's conscious state at the
time includes nothing that justifies it. No perceptual experience, no conscious memory
event, and no premises consciously entertained at the selected moment will be
justificationally sufficient for such a belief. According to strong internalism, then,
none of these beliefs is justified at that moment. Strong internalism threatens a dras-
tic diminution in the stock of beliefs ordinarily deemed justified, and hence in the
stock of knowledge, assuming that justification is necessary for knowledge. This is
a major count against this type of theory.
120 Alvin Goldman

Richard Feldman anticipates this problem because his own account of having
evidence also implies that only consciously entertained factors have evidential force.18
Feldman tries to meet the threat by distinguishing between occurrent and disposi-
tional senses of epistemic terms. (Feldman actually discusses knowledge rather than
justification, but I shall address the issue in terms of justification because that is the
target of our investigation.) Feldman is not simply restating the familiar point that
"belief" has occurrent and dispositional senses. He is proposing that the term "justi-
fied" is ambiguous between an occurrent and a dispositional sense. Feldman appar-
ently claims that in the case of stored beliefs, people at most have dispositional jus-
tification, not occurrent justification.
There are two problems with this proposal. First, if having a disposition to generate
conscious evidential states qualifies as a justifier of a belief, why wouldn't this extend
from memorial to perceptual dispositions? Suppose a train passenger awakes from a
nap but has not yet opened his eyes. Is he justified in believing propositions about the
details of the neighboring landscape? Surely not. Yet he is disposed, merely by open-
ing his eyes, to generate conscious evidential states that would occurrently justify such
beliefs. So the dispositional approach is far too permissive to yield an acceptable sense
of "justified."19 Second, can an internalist, especially a strong internalist, live with the
idea that certain dispositions count as justifiers? Having or not having a disposition (of
the requisite type) is not the sort of fact or condition that can be known by introspec-
tion. Thus, the proposal to supplement the occurrent sense of "justified" with a dispo-
sitional sense of "justified" is simply the abandonment of strong internalism.

III. Indirect Know/ability and Weak Internalism

The obvious solution to the problem of stored beliefs is to relax the KJ constraint:
Allow justifiers to be merely indirectly knowable. This yields KJind:

KJ jnd The only facts that qualify as justifiers of an agent's believing/? at time
t are facts that the agent can readily know at t, either directly or indi-
rectly, to obtain or not to obtain.

The danger here is that indirect knowledge might let in too much from an internalist
perspective. How are externalist forms of knowledge to be excluded, for example,
perceptual knowledge? Clearly, internalism must propose specific forms of knowl-
edge that conform with its spirit. It is fairly clear how internalism should deal with
the problem of stored beliefs. Simply allow knowledge of justifiers to include memory
retrieval. Stored evidence beliefs can qualify as justifiers because the agent can know
that they obtain by the compound route of first retrieving them from memory and
then introspecting their conscious contents. This yields the following variant of the
KJ constraint:

KJjnt+ret The only facts that qualify as justifiers of an agent's believing p at

time t are facts that the agent can readily know, at t, fo obtain or not
to obtain, by introspection and/or memory retrieval.
 Internalism Exposed 121

This KJ constraint allows for a more-viable form of internalism than strong internal-
ism. We may call it weak internalism, and initially articulate it through the following
principle:

WI Only facts concerning what conscious and/or stored mental states an agent
is in at time t are justifiers of the agent's beliefs at t.

WI will certify the justification of many stored beliefs, because agents often have
other stored beliefs that evidentially support them. A person who believes that Wash-
ington, D.C. is the capital of the United States may have a stored belief to the effect
that a map of the United States he recently consulted showed Washington as the
capital. The latter stored belief is what justifies the former one. So weak internalism
is not plagued with the problem of stored justified beliefs. Weak internalism seems
to be a legitimate form of internalism because even stored beliefs qualify, intuitively,
as internal states.
Although weak internalism is better than strong internalism, it too faces severe prob-
lems. First is the problem of forgotten evidence.70 Many justified beliefs are ones for
which an agent once had adequate evidence that she subsequently forgot. At the time
of epistemic appraisal, she no longer possesses adequate evidence that is retrievable
from memory. Last year, Sally read about the health benefits of broccoli in a New York
Times science section story. She then justifiably formed a belief in broccoli's benefi-
cial effects. She still retains this belief but no longer recalls her original evidential source
(and has never encountered either corroborating or undermining sources). Nonethe-
less, her broccoli belief is still justified, and, if true, qualifies as a case of knowledge.
Presumably, this is because her past acquisition of the belief was epistemically proper.
But past acquisition is irrelevant by the lights of internalism (including weak internalism)
because only her current mental states are justifiers relevant to her current belief. All
past events are "external" and therefore irrelevant according to internalism.
It might be replied that Sally does currently possess evidence in support of her
broccoli belief. One of her background beliefs, we may suppose, is that most of what
she remembers was learned in an epistemically proper manner. So doesn't she, after
all, now have grounds for the target belief? Admittedly, she has some evidence, but
is this evidence sufficient for justification? Surely not. In a variant case, suppose that
Sally still has the same background belief—viz., that most of what she remembers
was learned in an epistemically proper manner—but she in fact acquired her broc-
coli belief from the National Inquirer rather than the New York Times. So her broc-
coli belief was never acquired, or corroborated, in an epistemically sound manner.
Then even with the indicated current background belief, Sally cannot be credited with
justifiably believing that broccoli is healthful. Her past acquisition is still relevant—
and decisive. At least it is relevant as long as we are considering the "epistemizing"
sense of justification, in which justification carries a true belief a good distance to-
ward knowledge. Sally's belief in the healthfulness of broccoli is not justified in that
sense, for surely she does not know that broccoli is healthful, given that the National
Inquirer was her sole source of information.
The category of forgotten evidence is a problem for weak internalism because,
like the problem of stored beliefs facing strong internalism, it threatens skeptical
122 Alvin Goldman

outcomes. A large sector of what is ordinarily counted as knowledge are beliefs for
which people have forgotten their original evidence.
In reply to the problem of forgotten evidence, Matthias Steup offers the follow-
ing solution.21 An additional requirement for memorial states to justify a belief that
p, says Steup, is that the agent have adequate evidence for believing the following
counterfactual: "If she had encountered p in a questionable source, she would not
have formed the belief that p." Steup's suggestion is that in the National Inquirer
variant, Sally fails to have adequate evidence for this counterfactual, and that is why
her broccoli belief is not justified. My response to this proposal is twofold. First, the
proposed requirement is too strong to impose on memoriaHy justified belief. It is quite
difficult to get adequate evidence for the indicated counterfactual. Second, the pro-
posed requirement seems too weak as well. Sally might have adequate evidence for
the counterfactual but still be unjustified in holding her broccoli belief. She might
have adequate evidence for the counterfactual without it being true; but if it is not
true and the rest of the story is as I told it, her broccoli belief is not justified. So Steup's
internalist-style solution does not work.
A second problem confronting weak internalism is what I call the problem of
concurrent retrieval. Principle WI says that only conscious and stored mental states
are justifiers, but it does not say that all sets or conjunctions of such states qualify as
justifiers.22 Presumably, which sets of such states qualify is a matter to be decided
by reference to KJint+rel. If a certain set of stored beliefs can all be concurrently re-
trieved at time t and concurrently introspected, then they would pass the test of KJim+ret
and could qualify as justifiers under the principle of indirect knowability. But if they
cannot all be concurrently retrieved and introspected at t, they would fail the test.
Now it is clear that the totality of an agent's stored credal corpus at a time cannot be
concurrently retrieved from memory. So that set of stored beliefs does not qualify as a
justifier for purposes of weak internalism. Unfortunately, this sort of belief-set is pre-
cisely what certain types of internalist theories require by way of a justifier. Consider
holistic coherentism, which says that a proposition p is justified for person S at time t
if and only if p coheres with S's entire corpus of beliefs at t (including, of course, the
stored beliefs). A cognitive agent could ascertain, at t, whether/? coheres with her en-
tire corpus only by concurrently retrieving all of her stored beliefs. But such concur-
rent retrieval is psychologically impossible.23 Thus, the critically relevant justificational
fact under holistic coherentism does not meet even the indirect knowability constraint,
much less the direct knowability constraint. Here is a clash, then, between a standard
internalist theory of justification and the knowability rationale under scrutiny. Either
that rationale is indefensible, or a familiar type of internalism must be abandoned at
the outset. Nor is the problem confined to coherentism. Internalist foundationalism might
also require concurrent retrieval of more basic (or low-level) beliefs than is psycho-
logically feasible to retrieve.

IV. Logical and Probabilistic Relations

As these last examples remind us, every traditional form of internalism involves some
appeal to logical relations, probabilistic relations, or their ilk. Foundationalism re-
 Internalism Exposed 123

quires that nonbasically justified beliefs stand in suitable logical or probabilistic re-
lations to basic beliefs; coherentism requires that one's system of beliefs be logically
consistent, probabilistically coherent, or the like. None of these logical or probabi-
listic relations is itself a mental state, either a conscious state or a stored state. So
these relations do not qualify as justifiers according to either SI or WI. The point
may be illustrated more concretely within a foundationalist perspective. Suppose that
Jones possesses a set of basic beliefs at t whose contents logically or probabilistically
support proposition p. This property of Jones's basic beliefs, the property of sup-
porting proposition p, is not a justifier under WI, for the property itself is neither a
conscious nor a stored mental state. Nor is the possession of this property by these
mental states another mental state. So WI has no way of authorizing or permitting
Jones to believe/>. Unless WI is liberalized, no nonbasic belief will be justified, which
would again threaten a serious form of skepticism.
Can this problem be remedied by simply adding the proviso that all properties of
conscious or stored mental states also qualify as justifiers?24 This proviso is unac-
ceptably permissive for internalism. One property of many conscious and stored
mental states is the property of being caused by a reliable process, yet surely internal-
ism cannot admit this archetypically externalist type of property into the class of
justifiers. How should the class of properties be restricted? An obvious suggestion is
to include only formal properties of mental states, that is, logical and mathematical
properties of their contents. But should all formal properties be admitted? This ap-
proach would fly in the face of the knowability or accessibility constraint, which is
the guiding theme of internalism. Only formal properties that are knowable by the
agent at the time of doxastic decision should be countenanced as legitimate justifiers
under internalism. Such properties, however, cannot be detected by introspection and/
or memory retrieval. So some knowing operations suitable for formal properties must
be added, yielding a liberalized version of the KJ constraint.
How should a liberalized KJ constraint be designed? The natural move is to add
some selected computational operations or algorithms, procedures that would enable
an agent to ascertain whether a targeted proposition p has appropriate logical or proba-
bilistic relations to the contents of other belief states he is in. Precisely which com-
putational operations are admissible? Again, problems arise. The first is the problem
of the doxastic decision interval.
The traditional idea behind internalism is that an agent is justified in believing p
at time t if the evidential beliefs (and perhaps other, nondoxastic states) possessed at
t have an appropriate logical or probabilistic relation to p. In short, justification is
conferred simultaneously with evidence possession. Feldman makes this explicit: "For
any person S and proposition p and time t, S epistemically ought to believe p at t if
and only if p is supported by the evidence S has at t."25 Once the knowability con-
straint is introduced, however, simultaneous justification looks problematic. If justi-
fication is contingent on the agent's ability to know what justifiers obtain, the agent
should not be permitted to believe a proposition p at t unless she can know by t whether
the relevant justifiers obtain. Since it necessarily takes some time to compute logical
or probabilistic relations, the simultaneity model of justification needs to be revised
so that an agent's mental states at t only justify her in believing p at t + e, for some
suitable e. The value of e cannot be too large, of course, lest the agent's mental states
124 Alvin Goldman

change so as to affect the justificational status of p. But emust be large enough to

allow the agent time to determine the relevant formal relations.
These two conditions—(1) avoid mental change but (2) allow enough time to
compute formal relations—may well be jointly unsatisfiable, which would pose a
severe problem for internalism. Mental states, including perceptual states that gen-
erate new evidence, change very rapidly, and they could easily change before re-
quired computations could be executed. On the other hand, although mental states
do change rapidly, the agent's belief system might not be epistemically required to
reflect or respond to each change until interval e has elapsed. Some doxastic deci-
sion interval, then, might be feasible.
Is there a short-enough decision interval during which justificationally pertinent for-
mal properties can be computed? Coherentism says that S is justified in believing propo-
sition p only iip coheres with the rest of S' s belief system held at the time. Assume that
coherence implies logical consistency. Then coherentism requires that the logical con-
sistency or inconsistency of any proposition p with S's belief system must qualify as a
justifier. But how quickly can consistency or inconsistency be ascertained by mental
computation? As Christopher Cherniak points out, determination of even tautological
consistency is a computationally complex task in the general case.26 Using the truth-
table method to check for the consistency of a belief system with 138 independent atomic
propositions, even an ideal computer working at "top speed" (checking each row of a
truth table in the time it takes a light ray to traverse the diameter of a proton) would take
20 billion years, the estimated time from the "big-bang" dawn of the universe to the
present. Presumably, 20 billion years is not an acceptable doxastic decision interval!
Any reasonable interval, then, is too constraining for garden-variety coherentism.
The knowability constraint again clashes with one of the stock brands of internalism.27
Dyed-in-the-wool internalists might be prepared to live with this result. "So much
the worse for traditional coherentism," they might say, "we can live with its demise."
But this does not get internalism entirely off the hook. There threaten to be many
logical and probabilistic facts that do not qualify as justifiers because they require
too long a doxastic interval to compute. Furthermore, it is unclear what is a prin-
cipled basis for deciding what is too long. This quandary confronting internalism has
apparently escaped its proponents' attention.
A second problem for logical and probabilistic justifiers is the availability prob-
lem. Suppose that a particular set of computational operations—call it "COMP"—is
provisionally selected for inclusion alongside introspection and memory retrieval.
COMP might include, for example, a restricted (and hence noneffective) use of the
truth-table method, restricted so as to keep its use within the chosen doxastic deci-
sion interval.28 This yields a new version of the KJ constraint:

KJjnt+ret+coMP The only facts that qualify as justifiers of an agent's believing

p at time t are facts that the agent can readily know within a
suitable doxastic decision interval via introspection, memory
retrieval, and/or COMP.

Now the KJ constraint is presumably intended to apply not only to the cleverest or
best trained epistemic agents but to all epistemic agents, including the most naive
 Internalism Exposed 125

and uneducated persons on the street. After all, the point of the knowability con-
straint is that justifiers should be facts within the purview of every epistemic agent.
Under the GD conception, compliance with epistemic duty or responsibility is not
intended to be the private preserve of the logical or mathematical elite. It is some-
thing that ought to be attained—and should therefore be attainable—by any human
agent. The truth-table method, however, does not seem to be in the intellectual
repertoire of naive agents, so it is illegitimate to include COMP operations within
a KJ constraint. Unlike introspection and memory retrieval, it is not available to
all cognitive agents.
It may be replied that computational operations of the contemplated sort would
be within the capacity of normal human agents. No superhuman computational powers
are required. Computing power, however, is not the issue. A relevant sequence of
operations must also be available in the agent's intellectual repertoire; that is, she
must know which operations are appropriate to obtain an answer to the relevant (for-
mal) question.29 Since truth-table methods and other such algorithms are probably
not in the repertoire of ordinary cognitive agents, they cannot properly be included
in a KJ constraint.
A third problem concerns the proper methodology that should be used in se-
lecting a KJ constraint that incorporates computational operations. As we see from
the first two problems, a KJ constraint that conforms to the spirit of the GD ratio-
nale must reflect the basic cognitive skills or repertoires of actual human beings.
What these basic repertoires consist in, however, cannot be determined a priori. It
can only be determined with the help of empirical science. This fact fundamen-
tally undermines the methodological posture of internalism, a subject to which I
shall return in section VII.
Until now I have assumed a universal accessibility constraint, one that holds for
all cognitive agents. But perhaps potential justifiers for one agent need not be poten-
tial justifiers for another. Justifiers might be allowed to vary from agent to agent,
depending on what is knowable by the particular agent. If two agents have different
logical or probabilistic skills, then some properties that do not qualify as justifiers
for one might yet qualify as justifiers for the other. Indeed, the constraint KJint+ret+COMP
might be read in precisely this agent-relativized way. The subscripts may be inter-
preted as indicating knowledge routes that are available to the agent in question, not
necessarily to all agents.
If KJ constraints are agent relativized as a function of differences in knowledge
skills, this means that two people in precisely the same evidential state (in terms of
perceptual situation, background beliefs, etc.) might have different epistemic entitle-
ments. But if the two agents are to comply with their respective epistemic duties,
each must know which knowledge skills she has. This simply parallels the second
step of the internalist's original three-step argument. If one's epistemic duties or
entitlements depend on one's knowledge skills (e.g., on one's computational skills),
then compliance with one's duties requires knowledge of which skills one possesses.
There are two problems with this approach. First, it is unlikely that many people—
especially ordinary people on the street—have this sort of knowledge, and this again
threatens large-scale skepticism. Second, what is now required to be known by the
agent is something about the truth-getting power of her cognitive skills—that is, the
126 Alvin Goldman

power of her skills in detecting justifiers. This seems to be precisely the sort of ex-
ternal property that internalists regard as anathema. How can they accept this solu-
tion while remaining faithful to the spirit of internalism?30

V. Epistemic Principles

When the KJ constraint speaks of justifiers, it is not clear exactly what these com-
prehend. Specifically, do justifiers include epistemic principles themselves? I be-
lieve that principles should be included, because epistemic principles are among
the items that determine whether an agent is justified in believing a proposition,
which is just how "justifiers" was defined. Furthermore, true epistemic principles
are items an agent must know if she is going to determine her epistemic duties
correctly. Knowledge of her current states of mind and their properties will not
instruct her about her epistemic duties and entitlements unless she also knows true
epistemic principles. How are epistemic principles to be known, according to internal-
ism? Chisholm says that central epistemic principles are normative supervenience
principles, which (when true) are necessarily true.31 Since they are necessary truths,
they can be known a priori, in particular, they can be known "by reflection."

The internalist assumes that, merely by reflecting upon his own conscious state, he
can formulate a set of epistemic principles that will enable him to find out, with re-
spect to any possible belief he has, whether he is justified in having that belief.32

This passage is ambiguous as to whether (correct) epistemic principles are acces-

sible on reflection just to epistemologists or are accessible to naive epistemic agents
as well. The latter, however, must be required by internalism, because justifiers are
supposed to be determinable by all epistemic agents.
Are ordinary or naive agents really capable of formulating and recognizing cor-
rect epistemic principles? This seems highly dubious. Even many career-long epis-
temologists have failed to articulate and appreciate correct epistemic principles. Since
different epistemologists offer disparate and mutually conflicting candidates for
epistemic principles, at most, a fraction of these epistemologists can be right. Per-
haps none of the principles thus far tendered by epistemologists is correct! In light
of this shaky and possibly dismal record by professional epistemologists, how can
we expect ordinary people, who are entirely ignorant of epistemology and its mul-
tiple pitfalls, to succeed at this task?33 Nor is it plausible that they should succeed at
this task purely "by reflection" on their conscious states, since among the matters
that epistemic principles must resolve is what computational skills are within the
competence of ordinary cognizers. I do not see how this can be answered a priori,
"by reflection."
A crippling problem emerges for internalism. If epistemic principles are not know-
able by all naive agents, no such principles can qualify as justifiers under the KJ
constraint. If no epistemic principles so qualify, no proposition can be justifiably
believed by any agent. Wholesale skepticism follows.
 Internalism Exposed 127

VI. The Core Dilemma for the Three-Step Argument

In this section, I raise doubts about whether there is any cogent inferential route
from the GD conception to internalism via an acceptable KJ constraint. Here is the
core dilemma. The minimal, unvarnished version of the KJ constraint does not
rationalize internalism. That simple constraint merely says that justifiers must be
readily knowable, and some readily knowable facts might be external rather than
internal. If all routes to knowledge of justifiers are allowed, then knowledge by
perception must be allowed. If knowledge by perception is allowed, then facts of
an external sort could qualify for the status of justifiers. Of course, no epistemolo-
gist claims that purely external facts should serve as justifiers. But partly external
facts are nominated by externalists for the rank of justifiers. Consider properties
of the form: being a reliable perceptual indicator of a certain environmental fact.
This sort of property is at least partly external because reliability involves truth,
and truth (on the usual assumption) is external. Now suppose that a certain audi-
tory perceptual state has the property of being a reliable indicator of the presence
of a mourning dove in one's environment. Might the possession of this reliable
indicatorship property qualify as a justifer on the grounds that it is indeed readily
knowable? If every route to knowledge is legitimate, I do not see how this possi-
bility can be excluded. After all, one could use past perceptions of mourning doves
and their songs to determine that the designated auditory state is a reliable indica-
tor of a mourning dove's presence. So if unrestricted knowledge is allowed, the
(partly) external fact in question might be perfectly knowable. Thus, the unvar-
nished version of the KJ constraint does not exclude external facts from the ranks
of the justifiers.
The simple version of the KJ constraint, then, does not support internalism. Tacit
recognition of this is what undoubtedly leads internalists to favor a "direct"
knowability constraint. Unfortunately, this extra rider is not rationalized by the GD
conception. The GD conception at best implies that cognitive agents must know what
justifiers are present or absent. No particular types of knowledge, or paths to knowl-
edge, are intimated. So the GD conception cannot rationalize a restrictive version of
the KJ constraint that unambiguously yields internalism.
Let me put the point another way. The GD conception implies that justifiers must
be readily knowable, but are internal facts always more readily knowable than
external facts? As discussed earlier, probabilistic relations presumably qualify as
internal, but they do not seem to be readily knowable by human beings. An entire
tradition of psychological research on "biases and heuristics" suggests that naive
agents commonly commit probabilistic fallacies such as the "conjunction fallacy"
and use formally incorrect judgmental heuristics, for example, the representative-
ness heuristic and the anchoring-and-adjustment heuristic.34 If this is right, people's
abilities at detecting probabilistic relationships are actually rather weak. People's
perceptual capacities to detect external facts seem, by contrast, far superior. The
unqualified version of the KJ constraint, therefore, holds little promise for restricting
all justifiers to internal conditions in preference to external conditions, as internalism
requires.35
128 Alvin Goldman

VII. The Methodology of Epistemology:

Empirical or a Priori?

Internalism standardly incorporates the doctrine that epistemology is a purely a priori

or armchair enterprise rather than one that needs help from empirical science. Chisholm
puts the point this way:
The episteraic principles that [the epistemologist] formulates are principles that one
may come upon and apply merely by sitting in one's armchair, so to speak, and with-
out calling for any outside assistance. In a word, one need only consider one's own
state of mind.36

Previous sections already raised doubts about the merits of apriorism in epistemol-
ogy, even in the context of the theoretical architecture presented here. In this final
section, I want to challenge the viability of apriorism in greater depth.
Assume that, despite my earlier reservations, an internalist restriction on justifiers
has somehow been derived, one that allows only conscious states and certain of their
nonexternal properties to serve as justifiers. How should the epistemologist identify
particular conscious states and properties as justifers for specific propositions (or types
of propositions)? In other words, how should specific epistemic principles be crafted?
Should the task be executed purely a priori, or can scientific psychology help?
For concreteness, consider justifiers for memory beliefs. Suppose an adult con-
sciously remembers seeing, as a teenager, a certain matinee idol. This ostensible
memory could have arisen from imagination, since he frequently fantasized about
this matinee idol and imagined seeing her in person. What clues are present in the
current memory impression by which he can tell whether or not the recollection is
veridical? This is precisely the kind of issue that internalist epistemic principles should
address. If there are no differences in features of memory states that stem from per-
ceptions of real occurrences versus features of states that stem from mere imagina-
tion, doesn't this raise a specter of skepticism over the domain of memory? If there
are no indications by which to distinguish veridical from nonveridical memory im-
pressions, can we be justified in trusting our memory impressions? Skepticism aside,
epistemologists should surely be interested in identifying the features of conscious
memory impressions by which people are made more or less justified (orprima facie
justified) in believing things about the past.
Epistemologists have said very little on this subject. Their discussions tend to be
exhausted by characterizations of memory impressions as "vivid" or "nonvivid."
There is, I suspect, a straightforward reason for the paucity of detail. It is extremely
difficult, using purely armchair methods, to dissect the microfeatures of memory
experiences so as to identify telltale differences between trustworthy and question-
able memories. On the other hand, empirical methods have produced some interest-
ing findings, which might properly be infused into epistemic principles in a way
entirely congenial to internalism. Important research in this area has been done by
Marcia Johnson and her colleagues.371 shall illustrate my points by brief reference
to their research.
Johnson calls the subject of some of her research "reality monitoring." She tries to
characterize the detectable differences between (conscious) memory traces derived from
 Internalism Exposed 129

veridical perception of events versus memory traces generated by mere imaginations

of events.38 Johnson and Raye (1981) propose four dimensions along which memory
cues will typically differ depending on whether their origin was perceptual or imagi-
native. Compared with memories that originate from imagination, memories originat-
ing from perception tend to have (1) more perceptual information (e.g., color and sound),
(2) more contextual information about time and place, and (3) more meaningful detail.
When a memory trace is rich along these three dimensions, this is evidence of its hav-
ing originated through perception. Memories originating from imagination or thought,
by contrast, tend to be rich on another dimension: They contain more information about
the cognitive operations involved in the original thinkings or imaginings (e.g., effortful
attention, image creation, or search). Perception is more automatic than imagination,
so a memory trace that originates from perception will tend to lack attributes concern-
ing effortful operations. Johnson and Raye therefore suggest that differences in aver-
age value along these types of dimensions can form the basis for deciding whether the
origin of a memory is perceptual or nonperceptual. A memory with a great deal of vi-
sual and spatial detail and without records of intentional constructive and organiza-
tional processes should be judged to have been perceptually derived.39
Epistemologists would be well advised to borrow these sorts of ideas and incor-
porate them into their epistemic principles. A person is (prima facie) justified in
believing in the real occurrence of an ostensibly recalled event if the memory trace
is strong on the first three dimensions and weak on the fourth dimension. Conversely,
an agent is unjustified in believing in the real occurrence of the recalled event if the
memory trace is strong on the fourth dimension but weak on the first three dimen-
sions. All of these dimensions, of course, concern features of conscious experience.
For this reason, internalist epistemologists should be happy to incorporate these kinds
of features into their epistemic principles.
Let me distinguish two categories of epistemologically significant facts about
memory experience that empirical psychology might provide. First, as we have seen,
it might identify types of representational materials that are generally available in
people's memory experiences. Second, it might indicate which of these representa-
tional materials are either reliable or counterreliable indicators of the veridicality of
the ostensibly recalled events. Is the reliability of a memory cue a legitimate issue
from an internalist perspective? It might be thought not, since reliability is usually
classed as an external property. However, epistemologists might use reliability con-
siderations to decide which memory characteristics should be featured in epistemic
principles. They need not insert reliability per se into the principles. There is nothing
in our present formulation of internalism, at any rate, that bars the latter approach.
Any KJ constraint provides only a necessary condition for being a justifier; it leaves
open the possibility that additional necessary conditions, such as reliable indication,
must also be met. Indeed, many internalists do use reliability as a (partial) basis for
their choice of justifiers. BonJour (1985, 7) says that the basic role of justification is
that of a means to truth, and he defends coherence as a justifier on the ground that a
coherent system of beliefs is likely to correspond to reality. This point need not be
settled definitively, however. There are already adequate grounds for claiming that
internalism cannot be optimally pursued without help from empirical psychology,
whether or not reliability is a relevant consideration.
130 Alvin Goldman

VIII. Conclusion

Let us review the parade of problems infecting internalism that we have witnessed,
though not in their order of presentation. (1) The argument from the GD conception
of justification to internalism does not work. Internalism can only be derived from a
suitably qualified version of the KJ constraint because the unqualified version threat-
ens to allow external facts to count as justifiers. No suitably qualified version of the
KJ constraint is derivable from the GD conception. (2) A variety of qualified KJ
constraints are possible, each leading to a different version of internalism. None of
these versions is intuitively acceptable. Strong internalism, which restricts justifiers
to conscious states, is stuck with the problem of stored beliefs. Weak internalism,
which allows stored as well as conscious beliefs to count as justifiers, faces the prob-
lem of forgotten evidence and the problem of concurrent retrieval. (3) The question
of how logical and probabilistic facts are to be included in the class of justifiers is
plagued by puzzles, especially the puzzle of the doxastic decision interval and the
issue of availability. (4) Epistemic principles must be among the class of justifiers,
but such principles fail internalism's knowability requirement. (5) The favored meth-
odology of internalism—the armchair method—cannot be sustained, even if we grant
the assumption that justifiers must be conscious states.
Internalism is rife with problems. Are they all traceable to the GD rationale? Could
internalism be salvaged by switching to a different rationale? A different rationale
might help, but most of the problems raised here arise from the knowability con-
straint. It is unclear exactly which knowability constraint should be associated with
internalism, and all of the available candidates generate problematic theories. So I
see no hope for internalism; it does not survive the glare of the spotlight.40

Notes
1. Prominent statements of externalism include D. M. Armstrong, Belief, Truth and Knowl-
edge (Cambridge, Mass.: Cambridge University Press, 1973); Fred Dretske, Knowledge and
the Flow of Information (Cambridge, Mass.: MIT Press, 1981); Alvin Goldman, Epistemol-
ogy and Cognition (Cambridge, Mass.: Harvard University Press, 1986); Robert Nozick,
Philosophical Explanations (Cambridge, Mass.: Harvard University Press, 1981); and Alvin
Plantinga, Warrant and Proper Function (New York: Oxford University Press, 1993).
2. Major statements of internalism include Laurence BonJour, The Structure of Empiri-
cal Knowledge (Cambridge, Mass.: Harvard University Press, 1985); Roderick Chisholm,
Theory of Knowledge, 3d ed. (Englewood Cliffs, N.J.: Prentice Hall, 1989); Richard Foley,
The Theory of Epistemic Rationality (Cambridge, Mass.: Harvard University Press, 1987);
Keith Lehrer, Theory of Knowledge (Boulder, Colo.: Westview, 1990); and John Pollock,
Contemporary Theories of Knowledge (Totowa, N.J.: Rowman and Littlefield, 1986). In
addition to relatively pure versions of externalism and internalism, there are also mixtures of
the two approaches, as found in William Alston, Epistemic Justification (Ithaca, N.Y.: Cornell
University Press, 1989); Robert Audi, The Structure of Justification (New York: Cambridge Uni-
versity Press, 1993); and Ernest Sosa, Knowledge in Perspective (New York: Cambridge Univer-
sity Press, 1991).
3. Alvin Plantinga also traces internalism to the deontological conception: "If we go
back to the source of the internalist tradition, . . . we can see that internalism arises out of
deontology; a deontological conception of warrant. . . leads directly to internalism." War-
rant: The Current Debate (New York: Oxford University Press, 1993), 24-25. William
Alston (Epistemic Justification, 236) proposes a slightly different rationale for internalism,
 Internalism Exposed 131

although his rationale also proceeds via the knowability constraint. Alston suggests that
the concept of justification derives from the interpersonal practice of criticizing one another's
beliefs and asking for their credentials. A person can appropriately respond to other people's
demands for credentials only if he knows what those credentials are. So it is quite under-
standable, says Alston, that justifiers must meet the requirement of being accessible to the
agent. Clearly, this is one way to derive the accessibility constraint without appeal to the
deontological conception. But Alston is the only one I know of who advances this ground
for the accessibility constraint. In any case, most of the problems I shall identify pertain to
the accessibility constraint itself, which Alston's rationale shares with the deontological
rationale.
4. An Essay Concerning Human Understanding, vol. II, A. C. Fraser, ed. (New York:
Dover, 1959), 413.
5. BonJour, The Structure of Empirical Knowledge', Chisholm, Theory of Knowledge',
Feldman, "Epistemic Obligations," in Philosophical Perspectives, vol. II, J. Tomberlin, ed.
(Atascadero, Calif.: Ridgeview, 1988); Ginet, Knowledge, Perception, and Memory (Dordrecht,
The Netherlands: Reidel, 1975); Moser, Empirical Justification (Dordrecht, The Netherlands:
Reidel, 1985); Steup, "The Deontic Conception of Epistemic Justification," Philosophical Studies
53 (1988): 65-84.
6. Chisholm, Theory of Knowledge, 2d ed. (Englewood Cliffs, N.J.: Prentice Hall, 1977),
14; emphasis added.
7. For example, R. Eugene Bales distinguishes between two possible aims of act-utilitarianism:
as a specifier of a right-making characteristic or as a decision-making procedure. See "Act-
Utilitarianism: Account of Right-Making Characteristics or Decision-Making Procedure,"
American Philosophical Quarterly 8 (1971): 257-65. He defends utilitarianism against certain
critics by saying that it does not have to perform the latter function.
8. See "Internalism and Externalism in Epistemology," reprinted in Alston, Epistemic
Justification (Ithaca, N.Y.: Cornell University Press, 1989), 189.
9. Some internalists explicitly reject externalism on the grounds that it cannot be used as
a decision guide. For example, Pollock says: "fl]t is in principle impossible for us to actually
employ externalist norms. I take this to be a conclusive refutation of belief externalism" (Con-
temporary Theories, 134). Pollock would not subscribe to the full argument for internalism I
am discussing, however, because it is committed to the "intellectualist model" of epistemol-
ogy, which he disparages.
10. For the distinction between reliability and power (phrased slightly differently), see
Goldman, Epistemology and Cognition, chap. 6.
11. Jack Lyons points out that to comply with one's epistemic duty it suffices to know
that one has (undefeated) j ustifiers for proposition p; one does not have to know which justifiers
these are. So the argument is not entitled to conclude that knowledge of particular justifiers is
required by epistemic duty. Practically speaking, however, it is difficult to see how a cogni-
tive agent could know that relevant justifiers exist without knowing which particular ones
exist. So I shall pass over this objection to the internalist line of argument.
12. See "The Deontological Conception of Justification," reprinted in Alston, Epistemic
Justification.
13. Feldman, "Epistemic Obligations"; Heil, "Doxastic Agency," Philosophical Studies
40 (1983): 355-64.
14. Feldman's response, however, undercuts the step from the GD conception of justifi-
cation to the knowability constraint. If epistemic duty does not require that the agent be able
to discharge this duty, there is no longer a rationale for the knowability constraint. A differ-
ent line of response to the voluntarism worry is taken by Lehrer, who suggests that epistemo-
logical analysis should focus not on belief but on acceptance, where acceptance is some sort
of action that is subject to the will. See Lehrer, "A Self-Profile," in Keith Lehrer, R. Bogdan,
ed. (Dordrecht, The Netherlands: Reidel, 1981).
15. Chisholm, Theory of Knowledge, 3d ed., 7; emphasis added and original emphasis
deleted.
16. Chisholm, Theory of Knowledge, 3d ed., 59-60.
132 Alvin Goldman

17. Other epistemologists who restrict justifiers to conscious states or discuss access in
terms of introspection include Moser, Empirical Justification, 174; Feldman, "Having Evi-
dence," in Philosophical Analysis, D. Austin, ed. (Dordrecht, The Netherlands: Kluwer, 1988);
and Robert Audi, "Causalist Internalism," American Philosophical Quarterly 26, no. 4 (1989):
309-20.
18. Feldman, "Having Evidence," 98-99.
19. Feldman might reply that there is an important distinction between memorial and
perceptual dispositions; but it isn't clear on what basis a principled distinction can be drawn.
20. This sort of problem is discussed by Gilbert Harman, Change in View (Cambridge,
Mass.: MIT Press, 1986); Thomas Senor, "Internalist Foundationalism and the Justification
of Memory Belief," Synthese 94 (1993): 453-76; and Robert Audi, "Memorial Justification,"
Philosophical Topics 23 (1995): 31-45.
21. Steup's proposal occurred in a commentary on an earlier version of this paper, pre-
sented at the Central Division meeting of the American Philosophical Association, April 25,
1997.
22. Obviously one would need to reject the principle that the knowability of fact A and
the knowability of fact B entail the knowability of the conjunctive fact, A & B.
23. The "doxastic presumption" invoked by BonJour (The Structure of Empirical Knowl-
edge, 101-106) seems to assume that this is possible, but this is simply an undefended
assumption. Pollock (Contemporary Theories of Knowledge, 136) also raises the problem iden-
tified here, though in slightly different terms.
24. More precisely, the contemplated proviso should say that the possession of any prop-
erty by a mental state (or set of mental states) qualifies as a justifier. This reading will be
understood wherever the text talks loosely of "properties."
25. Feldman, "Epistemic Obligations," 254.
26. "Computational Complexity and the Universal Acceptance of Logic," The Journal of
Philosophy, 81, no. 12 (1984): 739-58.
27. This computational difficulty for coherentism is identified by Hilary Kornblith, "The
Unattainability of Coherence," in The Current State of the Coherence Theory, J. Bender, ed.
(Dordrecht, The Netherlands: Kluwer, 1989).
28. Because of the contemplated restriction, there will be many questions about formal
facts to which COMP cannot deliver answers. Thus, formal facts that might otherwise qualify
as justifiers will not so qualify under the version of the KJ constraint that incorporates COMP.
29. Propositional (or "declarative") knowledge of the appropriate sequence of operations
is, perhaps, an unduly restrictive requirement. It would suffice for the agent to have "proce-
dural" skills of the right sort. But even such skills will be lacking in naive cognitive agents.
30. It might be argued that internalism's spirit leads to a similar requirement even for
universal versions of a KJ constraint, not just for agent-relativized versions. Perhaps so; but
so much the worse for the general form of internalism.
31. Chisholm, "The Status of Epistemic Principles," Nous 24 (1990): 209-15.
32. Chisholm, Theory of Knowledge, 3d ed., 76; emphasis omitted.
33. A similar worry is expressed by Alston in "Internalism and Externalism in Episte-
mology," 221-22.
34. See Amos Tversky and Daniel Kahneman, "Judgment under Uncertainty: Heuristics
and Biases," in Judgment under Uncertainty, D. Kahneman, P. Slovic, and A. Tversky, eds.
(New York: Cambridge University Press, 1982); and Amos Tversky and Daniel Kahneman,
"Extensional versus Intuitive Reasoning: The Conjunction Fallacy in Probability Judgment,"
Psychological Review 91 (1983): 293-315.
35. It is not really clear, moreover, why logical or probabilistic facts intuitively count as
"internal" facts. They certainly are not internal in the same sense in which mental states are
internal. This is an additional problem about the contours of internalism.
36. Chisholm, Theory of Knowledge, 3d ed., 76.
37. See Marcia Johnson and Carol Raye, "Reality Monitoring," Psychological Review
88 (1981): 67-85; and Marcia Johnson, Mary Foley, Aurora Suengas, and Carol Raye, "Phe-
 Internalism Exposed 133

nomenal Characteristics of Memories for Perceived and Imagined Autobiographical Events,"

Journal of Experimental Psychology: General 117 (1988): 371-76.
38. Memory errors are not confined, of course, to confusions of actual with imagined
events. There are also errors that arise from confusing, or blending, two actual events. But
this research of Johnson's focuses on the actual/nonactual (or perceived versus imagined)
problem.
39. They also recognize that people can compare a target memory with memories of con-
textually related events to assess the target's veridicality. This kind of "coherence" factor is
a stock-in-trade of epistemology, however, and hence not a good example of the distinctive
contributions psychology can make to this subject. I therefore pass over it.
40. An earlier version of this paper was presented at the Central Division meeting of
the American Philosophical Association, Pittsburgh, Penn. April 25, 1997. My commenta-
tor on that occasion was Matthias Steup, and I am much indebted to him for valuable cor-
respondence on this topic. I am also grateful to Tim Bayne and Holly Smith for very useful
suggestions.
 8

Epistemic Duty, Evidence,

and Internality

MATTHIAS STEUP

In his essay, "Internalism Exposed" (essay 7), Alvin Goldman has issued a tough
challenge to those who think of themselves as internalists.1 I think of myself as an
internalist, and thus I attempt in this essay to rebut what I take to be the most impor-
tant of Goldman's objections.2 In the first section, I explain what I take to be the proper
epistemological context in which the issue over the internality of justification arises.
In the second section, I state what I take to be the rationale that supports internalism.
In the remaining sections, I will discuss Goldman's objections.

Internalism, Externalism, and the Conditions of Knowledge

What turns true belief into knowledge? Let us say that what accomplishes this is
epistemization. According to externalists, epistemization is accomplished by one
single property.3 Some externalists refer to it as "justification," others as "warrant."4
Since what epistemizes would have to "degettier" true belief, it cannot be the inter-
nal property internalists have in mind when they talk about justification. For the sort
of thing that degettierizes true belief is certainly not internal to the subject in the way
internalists require justifiers to be internal.5 It is thus only to be expected that what
externalists call "justification" or "warrant" is not the same thing as that which
internalists call "justification."
Internalists do not think that there is one single property that turns true belief into
knowledge. Rather, they think that it is two distinct things that accomplish the episte-
mization of true belief: internal justification and degettierization. Both are necessary
conditions of knowledge. Since I favor a deontological and evidentialist understanding

134
 Epistemic Duty, Evidence, and Internality 135

of justification, the reason why I take justification to be a necessary condition of

knowledge is that, as it seems to me, you do not know that/? unless you are in pos-
session of evidence that makes it epistemically responsible of you to believe that p.
But, of course, I do not take belief that is true and justified in this way to be suffi-
cient for knowledge. Rather, justified true belief amounts to knowledge only if it meets
the fourth condition of knowledge, that is, if it is degettiered.
Internalists need not deny the epistemological significance of reliability. Since
reliability is an external matter, it is neither necessary nor sufficient for internal jus-
tification. But this is not to say that reliability is not necessary for knowledge. And in
fact, although I cannot argue the point here, I am inclined to think that it is.

Giving Internalists What They Want

The internalism/externalism debate is about the things that determine whether a be-
lief is justified. I will follow Goldman in referring to these things as "justifiers."6
Internalists insist that justifiers must be internal, and that, consequently, an external
thing such as the reliability of a cognitive process cannot be a justifier. Why, how-
ever, do internalists think that justifiers must be internal? According to Goldman,
the standard rationale in support of internalism goes as follows (see p. 115). First,
internalists take epistemic justification to be a deontological concept.7 I will refer to
this first step as "deontology."

Deontology
S is justified in believing that p at £ if and only if it is epistemically responsible
of (or permissible for) S to believe that p at t.

Second, internalists derive from deontology what we may call the accessibility con-
straint: If something is to play the role of a justifier, it must be readily accessible to
the subject. Third, from the accessibility constraint, they derive the conclusion that
justifiers are internal in such a way that reliability is precluded from entering the club
of justifiers.
Goldman argues that this rationale does not work, for internalists face the follow-
ing dilemma. What deontology rationalizes is the accessibility constraint. But this
constraint does not give internalists what they want: the kind of internality that is
strong enough to preclude an external factor, such as the property of being produced
by a reliable cognitive process, from playing the role of a justifier (the first horn). So
they need a more stringent constraint. However, epistemic deontology does not ra-
tionalize a more-stringent constraint (the second horn). Hence internalism is left
without a justifying rationale (p. 127).
I agree with Goldman's claim that, if the rationale for internalism moves from
deontology to internality via the accessibility constraint, it does not give internalists
what they want. The accessibility constraint does not preclude reliable belief pro-
duction from functioning as a justifier. However, I do think that deontology supplies
internalists with the desired result if the rationale for internalism proceeds from de-
ontology to internality via evidentialism, the view that only items in a person's evi-
136 Matthias Steup

dence can play the role of justifiers. 8 1 will get back to this point at the end of this
section. Next, I will address the following two questions. First, why is the accessibil-
ity constraint not strong enough to give internalists what they want? Second, how do
we get from evidentialism to internality and from deontology to evidentialism?
Given deontology as a starting point, we must think of justifiers as things that make
it epistemically responsible of a person to believe a proposition. If we do so, then, as
Goldman points out, justifiers are subject to the accessibility constraint. This is so,
roughly, because what is not accessible to me cannot be the sort of thing that makes
me responsible (or, depending on the situation, irresponsible) in believing something.
However, if we consider carefully what the accessibility constraint, if properly put,
tells us, it looks initially as though it does give internalists what they want.

The Accessibility Constraint

If S is justified in believing thatp at t, then S has at t cognitive access to the
justifier (justifiers) that justifies (justify) S's belief.

Consider basic, unrefined reliabilism:

Reliabilism
S is justified in believing that p if and only if S' s belief that p is produced by a
reliable cognitive process.

The accessibility constraint is inconsistent with reliabilism thus defined. Recall

BonJour's Norman, who is endowed with the reliable faculty of clairvoyance.9 This
is something Norman has no way of knowing. Suppose Norman believes that p at t,
and his belief is produced by his faculty of clairvoyance. According to reliabilism,
Norman's belief is justified because it is produced by a reliable cognitive process.
Its being produced by a reliable cognitive process is a justifier for it. However, if we
endorse the accessibility constraint, then we cannot consider Norman's clairvoyance,
despite the fact that it is reliable, a justifier for his belief, for its reliability is not, as
we have stipulated, at t cognitively accessible to him. It would appear, therefore, that
the accessibility constraint does give internalists what they want: a conception of
internality that precludes the property of being produced by a reliable process from
being a justifier.10
However, we must distinguish between two different questions here. One ques-
tion is: Is reliabilism correct? Another question is: Can a reliable cognitive process
play the role of a justifier? Suppose we answer the second question thus: It can play
the role of a justifier at t if the reliability of the process is accessible to me at t. You
can reject reliabilism as defined above and still answer the second question in the
way I just indicated. In fact, you can accept the accessibility constraint and still give
that answer. To see why, consider the following example. Suppose Norman has been
a reliable clairvoyant for a while. He has a positive track record, which is something
he could easily access through memory retrieval. Suppose further Norman believes
that/7 at t, and his belief is produced by his faculty of clairvoyance. In this case, the
accessibility constraint is met. So if we define internality in terms of the accessibility
 Epistemic Duty, Evidence, and Internality 137

constraint, a reliable cognitive process, provided its reliability is at the relevant time
accessible, is permitted into the club of justifiers.
Should internalists be bothered by this outcome? It seems to me that internalists
who are evidentialists should be. For what they should say about the case we just
considered is this: What justifies Norman's belief is not that his belief has the prop-
erty of being produced by a reliable cognitive process; rather, what justifies it is the
evidence Norman has for believing that his belief was produced by such a process.
So, since I identify internalism with evidentialism, I agree with Goldman. The
accessibility constraint does not give internalists what they want. However, if we re-
place the rationale for internalism that Goldman has identified with the alternative
one I have proposed, internalists do get a constraint that precludes the property of
being produced by a reliable cognitive process from being a justifier even if it is
accessible. For, according to evidentialism, if something is to function as a justifier—
that is, if it is a thing that can make it epistemically responsible of a person to believe
that/?—then it must be an item in a person's evidence. A person's evidence consists
of her evidential states: mental states that have their source in sense experience,
introspection, memory, and (broadly) logical intuition." Evidentialism, then, imposes
what we might call the evidential states constraint: Only evidential states can play
the role of justifiers. Reliability—accessible or not—is not an evidential state and
thus does not, if we impose the evidential states constraint, qualify as a justifier. So
if internalists define internality in terms of the evidential states constraint, they se-
cure the desired notion of internality.

From Deontology to Evidentialism

But can evidentialism (as I have construed it here) be derived from deontology? It
seems to me it can. The derivation is a complex one, and an adequate account and
defense of it would go beyond the confines of this essay. I will therefore merely outline
how it would go. It involves three arguments. The first is an ingredient of BonJour's
well-known counterexamples to reliabilism.12 If some item x is to make it episte-
mically responsible of a person S to believe that p, then x must be in S 's cognitive
possession. The second argument would have to establish that there is one and only
one way in which x can come into S's cognitive possession: S must be in an eviden-
tial state that constitutes evidence for x and thus enables x to justify S in believing
that/7. The job of the third argument would be to show that, if this condition is met,
then x is either an evidential state itself or a mere fifth wheel: an item that plays no
role as a justifier for S's belief that p since all the needed justificatory work is be
done by S's evidence for x.
It will be instructive to apply this line of reasoning to the view we considered above,
the view that a reliable cognitive process is a justifier for a belief at t, provided its
reliability is accessible at t. Suppose Norman believes that/?. Suppose further his belief
is produced by a reliable process whose reliability is accessible to Norman because
he has an R-meter: a device that, if its electrodes are affixed to Norman's skull, will
tell him whether his belief was reliably produced. However, Norman has not yet
consulted his R-meter and thus is not in any evidential state that supplies him with a
138 Matthias Steup

reason to ascribe reliability to the process in question. Is he, in the deontological sense,
justified in believing that pi It would seem he is not. He would first have to get an
R-meter reading into his cognitive possession. If he gets such a reading into his cog-
nitive possession, he will then be in a requisite evidential state. He will then have
acquired evidence for ascribing reliability to the process in question, and his belief
will be justified. But at that point the reliability of the process does not do any justi-
ficatory work for Norman's belief. What justifies Norman's belief is his evidence
for ascribing reliability to the process. Indeed, this evidence would justify his belief
even if the process were in fact not reliable.13
Let me sum up. My starting point is deontology. Deontology mandates (along the
lines indicated in the previous two paragraphs) that justifiers be evidential states and
thus leads to evidentialism. Evidentialism, however, is full-fledged internalism, for
it does not allow for reliability (accessible or not) to play the role of a justifier. This
conclusion does not undermine the epistemic significance of reliability, but rather
shifts it into the context of degettierization instead of justification.
Before proceeding, I will clarify one further point. Evidentialism as I conceive of
it is construed by giving it a twist that is taken from the externalist textbook. Accord-
ing to reliabilist externalism, reliable belief production is sufficient for justified
belief. No further conditions need be met. Likewise, according to evidentialism, hav-
ing (undefeated) evidence for p is sufficient for being justified in believing that p.
No further conditions must be met.
It might be objected that, actually, further conditions do have to be met. How can
a subject meet her duty without knowing what her duty is? And how can a subject
know what her duty is without knowing what determines her duty? My reply to this
objection is that we must distinguish—in both ethics and epistemology—between
meeting one's duty, and knowing oneself to meet one's duty. Knowing that one meets
one's duty might require knowing what one's duty is and what it is that determines
one's duty. However, for the bare fact of duty fulfillment, such knowledge is not
necessary.14 For example, somebody who tells the truth in a certain situation meets
her duty to tell the truth in that situation even if she does not know that she ought to
tell the truth in that situation and what sort of thing determines whether a person ought
to tell the truth. However, it is not without plausibility to argue that, if one is to know
that one ought to tell the truth in a given situation, one must know what sort of thing
determines whether one ought to tell the truth in a given situation. Likewise, some-
body whose belief that p is supported by undefeated evidence is justified in believ-
ing thatp even if she does not know that she is justified in believing thatp, and what
it is that determines a person's epistemic duty. However, if she is to know that in a
given situation she is justified in believing that/?, then, so we might plausibly claim,
she needs to know what sort of thing determines what a person's epistemic duty is.15
My point, then, is that we must distinguish between the conditions of first-order
and second-order justification. First-order justification, even if deontologically con-
strued, is easier to come by than second-order justification is. For a subject to be
justified in believing that p, she need not believe that she has undefeated evidence
for p. Nor must she form a belief about what justifiers justify her belief that p, let
alone a belief about what logical or probabilistic support relations obtain between
these justifiers and her belief that p. Nor need she have any concurrent beliefs in
 Epistemic Duty, Evidence, and Internality 139

epistemic principles. Meeting some of the conditions I just mentioned might be nec-
essary for the recognition of justification, for justifiably believing oneself to be jus-
tified in believing that/?. But there is no reason why deontologists need to consider
them necessary for first-order justification. As we will see, this point will play a large
role in the subsequent discussion.

Forgotten Evidence

Goldman begins his exposure of internalism by objecting to the sort of internalism

that results from restricting justifiers to conscious states. He argues that strong internal-
ism cannot handle what he calls stored beliefs: beliefs about one's birth date, social
security number, or telephone number. For almost any of such beliefs, according to
Goldman, "one's conscious state at the time includes nothing that justifies it" (p. 119).
But why should internalists restrict justifiers to conscious states? Goldman an-
ticipates that at least some internalists would not want to endorse such a severe
restriction and thus considers what he calls "weak internalism," which restricts
justifiers to conscious and/or stored mental states (see p. 121). Weak internalism is
roughly equivalent to evidentialism as I understand it. An evidentialist solution to
the problem of stored beliefs would go thus. Your telephone number is something
you remember. When you consider your own telephone number, you can experience
a certain phenomenological quality, a quality that is missing when you consider a
telephone number that isn't yours. Let us call what you experience when you con-
sider your own telephone number, as opposed to someone else's number, a memo-
rial seeming: You clearly seem to remember that the number is yours.16 This memo-
rial seeming qualifies as an evidential state; it is evidence you have for taking your
belief to be true. According to evidentialism (or weak internalism), then, stored
beliefs are justified by evidential states that have their source in memory.
Goldman argues, however, that weak internalism is still too restrictive. To make
his point stick—the point that some beliefs are justified by things other than one's
evidential states and thus are not internally justified—he construes a case for which
he claims the appeal to retrievable memorial states does not work. This is how he
describes the case.
Last year Sally read about the health benefits of broccoli in a New York Times science-
section story. She then justifiably formed a belief in broccoli's beneficial effects. She
still retains this belief but no longer recalls her original evidential source (and has never
encountered either corroborating or undermining sources). Nonetheless, her broccoli
belief is still justified, and, if true, qualifies as a case of knowledge, (p. 121)

In this case, Sally has forgotten where she picked up the information that broccoli is
healthy, and thus it appears as though there is nothing in her present evidence that
justifies her belief. And if that were true, this would certainly be bad news for the
kind of evidentialism that I am advocating.
As one would expect, the evidentialist will reply that there is in Sally's present
evidence an item that justifies her belief: a memorial seeming. She clearly seems to
remember that broccoli is healthy. Add to this the justified background belief that
140 Matthias Steup

what she clearly remembers is usually true, and we have what looks like a pretty good
justification for the belief in question. Goldman argues, however, that the kind of
justification the internalist can in this way secure is not the kind of justification we
need to find. What we need to find is the kind of justification that can supply Sally
with knowledge. What the internalist can come up with falls short of this. For the
purpose of illustration, let us consider the following example, which is a somewhat
embellished version of the modified Sally case Goldman describes (p. 121).
Sally is an epistemically responsible person, and thus, while she tends to accept what
she reads in the science section of the New York Times, she would not believe anything
printed in the National Inquirer, at least not without further corroboration. If she were
at all to lay eyes on this publication, it would be merely for the purpose of amusing
herself. However, it was indeed in the National Inquirer where she read that eating
broccoli is healthy. Normally she would not have formed this belief without further
corroboration, but on this particular occasion an unusual and atypical accident occurred:
she did form the belief that broccoli is healthy solely because she read it in the Na-
tional Inquirer. She retains this belief, but she does not subsequently remember from
where she got it.
As far as this variation of the case is concerned, Goldman seems prepared to agree
that Sally's belief enjoys internal justification. For the point of this case is not that
her belief is internally unjustified, but rather that the justification her belief enjoys is
not the kind of justification that is epistemologically significant: the kind of justifi-
cation that can turn true belief into knowledge.17 For clearly, Sally does not know
that broccoli is healthy given that she got this information from the National Inquirer.
Goldman's point, then, is that whatever internal justification Sally's belief might enjoy
is not epistemizing; it is not the sort of justification that can turn a true belief into
knowledge.
In reply to this argument, I claim indeed that Sally's belief is internally justified,
and that the justification it enjoys is epistemologically significant because it is nec-
essary for knowledge. But, and this is the main point, I reject the suggestion that saying
this commits me to the verdict that Sally's belief amounts to knowledge.18 It does
not although it is justified and (let us suppose) true. Sally's belief, then, is a justified
true belief that is not knowledge, which is to say that Sally's belief is the sort of be-
lief that figures in Gettier cases.
Consider a well-known Gettier case. Henry is in an area in which what to the
unsuspecting observer appear to be barns are merely barn facades.[9 But this is some-
thing Henry does not know. He happens to look at the only real barn in this region,
and so believes not only justifiably but also truly that there is a barn over there. Does
he know that there is a barn over there? Clearly he does not. The reason for denying
him knowledge is that the truth of his belief is, in a way that is not easy to pin down,
just an accident.
Compare the barn case with the second broccoli case. Sally's belief is, as we are
supposing, true and also justified because (i) she has a strong memorial seeming that
broccoli is healthy, and (ii) she has the justified background belief that what she clearly
seems to remember is usually true. Yet her belief does not amount to knowledge
because its truth is, in a way that is not easy to pin down, just an accident.
 Epistemic Duty, Evidence, and Internality 141

Recall that internalists, as I already mentioned in the first section, consider

internal justification to be a necessary condition of knowledge but not sufficient to
epistemize true belief. The sort of thing that turns true belief into knowledge is
internal justification plus degettierization. The latter, however, is an external mat-
ter and cannot be built into a conception of justification that is evidentialist and
deontological. My reply to Goldman's objection, then, is that the second broccoli
case does not refute the claim that internal justification is necessary for knowledge.
What it does refute is the claim that internal justification is sufficient for turning
true belief into knowledge. But this is a claim I do not think internalists should
make.

Logical and Probabilistic Relations

What Goldman refers to as "weak internalism" restricts justifiers to conscious and/

or stored mental states. With regard to logical and probabilistic relations, this con-
straint has the same effect as the evidential states constraint: It precludes such rela-
tions from playing the role of justifiers. This, according to Goldman, has an undesir-
able consequence. To explain what the problem is, Goldman considers internalist
foundationalism. Suppose Jones' s belief that p is justified by his belief that q, where
the latter is a basic belief. Let us further suppose that q functions as a justifier for p
by virtue of a logical relation R that obtains between p and q. Since R is not an evi-
dential state, it does not qualify as a justifier. Consequently, so Goldman argues, weak
internalism (WI) "has no way of authorizing or permitting Jones to believe/?. Unless
WI is liberalized, no nonbasic belief will be justified, which would again threaten a
serious form of skepticism" (p. 123).
Is the sort of internalism that results from imposing the evidential states constraint
threatened by skepticism in the way Goldman suggests? Let us examine this ques-
tion using entailment as an example. Suppose p entails q, Jones sees that this entail-
ment obtains, justifiably believes that p, and justifiably believes that q. The question
is whether the kind of internalism I have construed can issue authorization for Jones' s
belief that q. Goldman would say that it cannot because it does not count the entail-
ment relation as a justifier. Given that it does not, how can evidentialist internalism
authorize Jones's justified belief that ql
Here is how. There are two justifiers that justify Jones's belief that q: (i) his belief
that p, and (ii) his state of seeing that p entails q. According to evidentialism, these
two justifiers suffice to justify Jones in believing that q. No further justifier is neces-
sary. It is not the case, therefore, that evidentialist internalism is faced with embar-
rassment because it cannot explain what justifies us in believing propositions that
are entailed by other propositions.20
It might be objected that Jones would not be justified in believing that q if p did
not in fact entail q (for, as we may suppose, if/? did not entail q, it would not support
q in any other way). The entailment relation between p and q, then, is necessary for
the justification of Jones's belief that q. It is, therefore, a justifier for it. However, it
is erroneous to think that anything that is necessary for justification is a justifier. For
142 Matthias Steup

example, for Jones to be justified in believing that q, Jones must understand that q.
But surely Jones's understanding of q is not a justifier for his belief that q.21
Parallel reasoning applies to Goldman's example regarding foundationalist
internalism. The justifier for Jones's nonbasic belief is his basic belief. The latter
is a justifier for the former because a suitable support relation obtains between the
two beliefs. There is no need, however, to count the support relation itself as a
justifier. If Jones were to become justified in believing that his nonbasic belief is
justified, then it might be that he would have to grasp thatp supports q. But, in that
case, what generates second-order justification would be his grasp of the support
relation, not the support relation itself.
There is no need, then, for evidentialists to count logical and probabilistic rela-
tions as justifiers. As far as first-order justification is concerned, S's belief that p
suffices, provided a suitable support relation obtains between p and q, to justify S in
believing that q. And as far as second-order justification is concerned, it is not logi-
cal and probabilistic relations themselves, but rather justified belief in, or the grasp
of, such relations that play the role of justifiers.

The Epistemic Decision Interval

Suppose you sit on a bench in a park and notice a dog on the lawn before you. You
ask yourself: In believing that there is a dog over there, am I meeting my epistemic
duty? According to Goldman, if internalism is to enable ordinary subjects to meet
their epistemic duty, it must incorporate suitable procedures for identifying justifiers
and checking whether they support the target belief (p. 123f). I will follow Goldman's
terminology and let COMP be a candidate for such a procedure. What Goldman wor-
ries about is the problem of the epistemic interval. To ensure that you can answer
your question, the application of COMP must deliver a result in a timely fashion.
Suppose the dog is there for three minutes, but the application of COMP delivers a
result only after two hours. In the meantime, the dog went elsewhere, you ceased to
believe in its presence, and long gone is the opportunity to check on your epistemic
performance while you were having the belief the justification of which you meant
to ascertain.
What is the significance of this problem? It poses, according to Goldman, first of
all a problem for holistic coherentism, for checking on one's entire belief system might
take an awfully long time (see p. 124).22 But then, not all internalists are coherentists.
However, Goldman argues that those who are not also need to worry about the inter-
val problem for the following reason: "There threaten to be many logical and proba-
bilistic facts that do not qualify as justifiers because they require too long a doxastic
interval to compute. Furthermore, it is unclear what is a principled basis for deciding
what is too long" (p. 124).
The length of the epistemic decision interval and the availability of procedures to
recognize logical and probabilistic support relations strike me as interesting problems.
I do not see, however, that they pose a threat to internalism as I have construed it. I
think it is indeed a necessary condition of a beliefs justification that it stand in a suit-
able logical or probabilistic support relation to its evidence. However, such a relation
 Epistemic Duty, Evidence, and Internality 143

is not an evidential state itself and thus cannot play the role of a justifier. Hence a
subject's failure, or even inability, to ascertain such facts does not have the consequence
of leaving her without the justifiers she needs if her beliefs are to be justified.
Moreover, as I have already indicated, I do not think it is a necessary condition of
justified belief—even if construed deontologically—that the subject form a raetabelief
about what justifiers justify her belief. What is necessary is merely that there be suit-
able justifiers, that is, that the subject be in the requisite evidential states. Consequently,
as far as first-order justification is concerned, it does not matter whether we can recog-
nize justifiers in a timely fashion. This matters only as far as second-order justification
is concerned. If I cannot recognize what justifies me in believing that/? as long as I am
believing that p, then it is not possible for me to form the justified metabelief that I am
now justified in believing that p. I could then determine only later that when I believed
that p I was justified in believing that p. Now, I am inclined to accept that it may on
occasion be quite difficult to ascertain whether what we believe about a certain subject
matter is justified. On such occasions, it may be that we can determine only ex post
facto that our belief was justified (or unjustified). Why, however, suppose that the
possibility of such situations is a problem for internalism?

The Availability Problem

It would seem that Goldman saddles internalism with a view about justification that
I reject; that to have justified beliefs, subjects must form beliefs about where the
justification for their beliefs comes from. On the assumption that this view is indeed
an ingredient of internalism, a problem arises with regard to the cognitive processes
through which agents can recognize which justifiers justify their beliefs. Let us as-
sume the internalist suggest that COMP is the cognitive process through which
justifiers and justificatory relations can be recognized. If so, the internalist constraint
I have proposed reads thus: Justifiers must be evidential states of which the subject
can become aware through COMP. Such a constraint, Goldman writes, "is presum-
ably intended to apply not only to the cleverest or best-trained agents but to all
epistemic agents, including the most naive and uneducated persons on the street"
(p. 124f). I agree with Goldman that the constraint that defines internality should be
an equitable constraint; it should apply to all subjects equally. However, as Goldman
points out, equal opportunity to having justified beliefs is threatened by the qualifi-
cation that the cognitive accessibility of the justifiers must be guaranteed by COMP.
What about those agents who are not sophisticated enough to master COMP? It would
seem they will be unable to acquire knowledge of justifiers and thus will be left without
the resources to meet their epistemic duty.
The main problem with this argument is the premise on which it rests: that accord-
ing to internalism, for S to be justified in believing that p (to meet her epistemic duty
with regard to p), S must form a concurrent belief about which justifiers justify her in
believing that p—a belief that would have to be the result of using COMP or some
other procedure. Once again, my reply is that internalism should not be saddled with
any such condition. Certainly a deontological-evidentialist construal of internalism does
not require imposing it.
144 Matthias Steup

As mentioned before, I am not unsympathetic to the alternative premise that, to

recognize that one is justified in believing that/?, one must recognize which justifiers
justify one in believing that p. The problem Goldman has raised, then, concerns the
procedures through which agents can come to recognize second-order justification.
With regard to this problem, my reply is that I see no reason why internalists should
feel obliged to constrain the procedures through which agents can become aware of
the logical relations between the justifiers and justifiees in their belief systems. Why
place restrictions on these procedures? It is not necessary that a given justifier be
such that everybody (or n-number of other people) can recognize it. Rather, what is
necessary is merely that, if I am justified in taking myself to be justified in believing
that p, then there is some procedure available to me through which I can recognize
the justifiers that justify me in believing that p.
Goldman refers to this move as agent-relativization and objects to it on the
ground that it implies the following consequence: "Two people in precisely the same
evidential state might have different epistemic entitlements" (p. 125). From the
evidentialist point of view, this scenario cannot arise. If the two subjects are in type-
identical evidential states, they will not differ in their ability to recognize logical
and probabilistic relations. If they so differ, then they will not be in type-identical
evidential states. Furthermore, as far as their ability to recognize justifiers is con-
cerned, we must distinguish between first-order and second-order evidential states.
If they are in type-identical first-order evidential states, their epistemic entitlements
will be the same at the first level but might differ at the second. The second subject
might recognize justifiers that the first subject cannot recognize. If so, then even if
the two subjects are in type-identical first-order evidential states, the second sub-
ject will at the second level have entitlements that the first subject does not have.
For Goldman's objection to stick, however, it would have to be possible for two agents,
while being in type-identical evidential states at a given level, to have different
epistemic entitlements at that level. This, however, is not possible. At least, Goldman
has not established the possibility of it. I conclude, therefore, that if we wish to make
second-order justification available to all agents, then agent-relativization is the right
move.

Epistemic Principles

According to Goldman, one reason why internalism is burdened with skeptical con-
sequences is that we must include epistemic principles among the set of justifiers.
However, ordinary agents cannot know what the correct epistemic principles are. (Not
even professional epistemologists have succeeded in the task of identifying them.)
Let us call this claim the "bleak diagnosis." If the bleak diagnosis is true, then ordi-
nary agents cannot know which beliefs are licensed by the correct epistemic prin-
ciples, which means that they cannot have justified beliefs. Goldman concludes:
"Wholesale skepticism follows" (p. 126).
It seems to me that there are three problems with this argument. First, it imposes
a condition on first-order justification that is too stringent. Second, there are good
reasons to reject the premise that epistemic principles are justifiers. Third, the bleak
 Epistemic Duty, Evidence, and Internality 145

diagnosis is actually quite doubtful. Let me begin with the first problem. Goldman's
argument rests on the following implicit premise:

(P) If S is justified in believing that p at t, then S knows a relevant epistemic

principle that permits her to believe that p at t.

There is no reason why internalists, even if they endorse deontology, should accept
(P). As I mentioned before, to meet one's duty, one needs to know neither what one's
duty is nor what determines one's duty. (P) might be correct if its antecedent were
about second-order justification. But in that case, supposing the bleak diagnosis is
correct, we would merely get the conclusion that ordinary agents could not know
which of their beliefs are justified. This would certainly not be wholesale skepticism,
though it would in my opinion be an unfortunate outcome.
The second problem concerns the status of epistemic principles. Are they justifiers?
Goldman thinks they are, for two reasons. The first is that justifiers are the "deter-
miners" of justification, and epistemic principles are among the things that determine
justification (see p. 126). I find this reason unconvincing. Epistemic principles "deter-
mine" justification only in the sense that they tell us which conditions are sufficient—
or necessary—to justify beliefs. Clearly, however, that doesn't mean that these prin-
ciples themselves are among the things that justify beliefs. Consider an example of an
epistemic principle: If S has undefeated sensory evidence for p, then S is justified in
believing that p. This principle tells us that sensory evidence for p is a justifier and
succeeds in justifying the belief that p if it is undefeated. But surely this principle itself
does not function as a justifier for beliefs that are justified by undefeated sensory evi-
dence. And this point, it seems to me, applies to all epistemic principles: They tell us
what sorts of things justify beliefs, but they do not themselves play the role of justifiers.
Consider moral principles. Are moral principles the sort of thing that makes acts
right or wrong? I think not. Let us call the things that make acts right "right-making"
characteristics. The correct moral principles tell us what the right-making character-
istics are. But that doesn't mean that they themselves are right-making characteris-
tics. Similarly, we can conceive of justifiers as justification-conferring characteris-
tics. The correct epistemic principles tell us what these characteristics are. But they
themselves are not such characteristics.
Goldman's second reason is the following: "True epistemic principles are items an
agent must know if she is going to determine her epistemic duties correctly. Knowl-
edge of her current states of mind and their properties will not instruct her about her
epistemic duties .. . unless she also knows true epistemic principles" (p. 126). How-
ever, what follows from this premise is merely that knowledge of epistemic principles
is a justifier. It does not follow from it that epistemic principles themselves are justifiers.
Third, let us consider Goldman's bleak diagnosis. I am not convinced that it is
true. Compare the situation with regard to moral principles. There is vast disagree-
ment among professional ethicists on which moral principles are correct. Yet who
would want to conclude from this that ordinary agents cannot tell their moral duty or
that they do not know basic moral principles? It could, of course, be that epistemol-
ogy and ethics are different in this respect, but I would argue that they are not. It
seems to me that ordinary people can grasp, and typically do grasp, basic epistemic
146 Matthias Steup

principles such as that one should base one's beliefs on one's evidence, and that one
should not believe what one's evidence undermines.

Common Ground

Reading the literature on the internalism/externalism debate gives rise to the impres-
sion that internalists and externalists are engaged in a fundamental, irreconcilable
controversy. I'd like to conclude this paper by putting this controversy in a more
benevolent perspective. As already mentioned in the first section, although I favor
internalism about justification and Goldman rejects it, I agree with Goldman about
the epistemological relevance of reliability and the importance of a belief's causal
history. However, whereas he subsumes these issues under the heading of justifica-
tion, I prefer to think of them as belonging into the context of degettierization: of
capturing the fourth condition of knowledge.
I also agree with Goldman that internalist justification does not epistemize. Of
course, with regard to such justification, we also disagree. I take it to be necessary
for knowledge; Goldman (I suppose) does not. But how deep does this disagreement
go? Perhaps we are simply engaged in different projects. Goldman, it seems to me,
wants to analyze a concept of knowledge that extends to both humans and animals.
I do not. I think that there is a difference between the kind of knowledge we attribute
to the mouse who knows where the cheese is and the kind of knowledge we ordi-
narily attribute to humans. What I take to be the difference is this: Unlike animals,
humans can come to know that they know. This is not to say that humans never know
in the way animals know. When they acquire knowledge on the basis of reliable
cognitive processes the reliability of which they cannot detect, then for the sake of
consistency we must attribute to them the same kind of knowledge we attribute to
animals. But when humans acquire knowledge through processes that they can rec-
ognize as reliable (i.e., when they acquire knowledge that they can recognize as
knowledge), then they enjoy a higher kind of knowledge, a kind of knowledge that
animals cannot acquire.23 This kind of knowledge—the kind one has only if one is in
a position to know that one knows—is what I am concerned with. It seems to me,
however, that what is required for knowing that one knows that p at t is to have in
one's cognitive possession at / evidence for p. And, as I have argued, evidence that
is in one's cognitive possession consists of evidential states. What I claim, then, is
that internal justification is necessary, not for knowledge in the broad sense in which
externalists understand this concept, but only for knowledge in the narrow sense that
I just explained. Further discussion will show whether Goldman thinks that these
considerations narrow the gulf between internalism and externalism.24

Notes
1. See The Journal of Philosophy, 96, no. 6 (June 1999): 739-58, reprinted in this vol-
ume, pp. 115-133. (Page numbers in parentheses refer to this volume.)
2. I will discuss most of Goldman's objections, but within the confines of this essay, I
cannot address the arguments in section VII of his essay, which concern the methodology of
epistemology.
 Epistemic Duty, Evidence, and Internality 147

3. It could be objected that at least according to some versions of externalism, what

epistemizes true belief are two things: justification (or perhaps warrant), and degettierization.
But this move makes it hard to understand why justification (or warrant) should be necessary
in the first place. Moreover, if justification (or warrant) is taken to be external, why not build
degettierization right into it by adding a suitable clause? It seems to me, therefore, that exter-
nalism is best construed as a theory according to which what turns true belief into knowledge
is one single thing, and thus as a theory according to which knowledge has three necessary
conditions.
4. In Epistemology and Cognition (Cambridge: Harvard University Press, 1986), Goldman
refers to this property as "justifiedness." Plantinga refers to it as "warrant." See Alvin Plantinga,
Warrant: The Current Debate and Warrant and Proper Function (New York: Oxford Uni-
versity Press, 1993).
5. See my "In Defense of Internalism," in The Theory of Knowledge. Classical and Con-
temporary Readings, L. Pojman, ed. (Belmont, Calif.: Wadsworth, 1999).
6. I do so somewhat reluctantly. In addition to justifiers, there are things that diminish or
even destroy a belief s justification. Internalists take these to be internal as well. However, as
long as this point is kept in mind, use of the term "justifier" should be unproblematic.
7. I do not think that epistemology should offer guidance to its practitioners, nor, for
that matter, to anybody else. It is a theoretical discipline concerned with questions such as,
What is knowledge? and What is justified true belief? According to Goldman, however,
"only if the guidance conception is incorporated can the argument proceed along the in-
tended lines to the accessibility constraint, and from there to internalism." (See 273f.) Un-
fortunately, Goldman does not explain what he takes to be the argument in support of this
claim. Moreover, I think it is unclear to begin with what exactly the guidance-aspect adds
to epistemic deontology as it is usually understood. In any case, an appeal to a guidance
role of epistemology plays no role in the way I derive evidentialism—and thus internalism—
from deontology.
8. For a defense of evidentialism, see E. Conee and R. Feldman, "Evidentialism," Philo-
sophical Studies 48 (1985).
9. See Laurence BonJour, The Structure of Empirical Knowledge (Cambridge, Mass.:
Harvard University Press, 1985), 41.
10. The title of this section should not be misunderstood. Internalists do not start out with
the desire to prevent the property of being produced by a reliable cognitive process from being
a justifier and then doctor their view accordingly. Rather, they start out with an intuitive
understanding of justification, an understanding that, on the basis of many intuitive cases,
disqualifies reliable belief production from playing the role of a justifier. But then the ques-
tion arises of precisely what aspect of justification has this effect. The answer would appear
to be: its internality. Thus the litmus test for any account of internality is this: Does it yield
the intended result of implying that reliability does not qualify as a justifier? For effective
counterexamples to reliabilism that motivate an internalist understanding of justification, see
L. BonJour, The Structure of Empirical Knowledge; C. Ginet, "Contra Reliabilism," The Monist
68 (1985) 175-87, and K. Lehrer, Knowledge (Boulder, Colo.: Westview, 1990).
11. This account of evidence applies to ordinary human beings. Idiot savants and Alpha
Centaurians might possess additional cognitive faculties, and thus be able to base their be-
liefs on evidential states not available to ordinary humans.
12. See BonJour, The Structure of Empirical Knowledge, chap. 3.
13. Bear in mind that this argument rests on the starting point of thinking of justification
deontologically. It does not bear plausibility if we expect justification to epistemize.
14. However, it seems to me that the possibility of such knowledge is required for duty
fulfillment. Beings who are cognitively not equipped to understand, and come to know, that
they have duties—such as nonhuman animals—cannot have duties.
15.1 am saying that second-order justification might require these further conditions be-
cause I actually doubt that it does. According to Chisholm, being justified in believing that
one is justified in believing that p requires no more than considering the evidence one has in
support of p. If this is correct, then second-order justification is not so hard to come by. See
148 Matthias Steup

R. Chisholm, "Knowing That One Knows," in The Foundations of Knowing (Minneapolis:

University of Minnesota Press, 1982), 50-58.
16. For a recent defense of the view that we can think of justifiers as seemings, see Wil-
liam Tollhurst, "Seemings," American Philosophical Quarterly 35, no. 3 (1998), 293-302.
17. Actually, Goldman is somewhat cautious here. He identifies the relevant kind of jus-
tification as the kind of justification that "carries a true belief a good distance toward knowl-
edge." (281) But if this is taken literally, what is the relevance of pointing out that Sally doesn't
know that broccoli is healthy?
18. In "Internalism Exposed," Goldman discusses a different reply of mine to the prob-
lem of forgotten evidence. This reply, however, was directed against a different kind of case,
a case that, as I now think, is not the sort of case Goldman has in mind.
19. See A. Goldman, "Discrimination and Perceptual Knowledge," The Journal of Phi-
losophy 73 (1976), 771-91.
20. I'm indebted to Michael Bergman for an objection that prompted me to change this
argument from the way I made it in a previous version of this essay.
21. If this doesn't strike you as obvious, consider what is implied by taking understand-
ing of p to be a justifier for believing that p. What is implied is that you are justified (at least
to a certain degree) in believing every proposition you understand, supposing that you do not
have a defeater for it. This certainly strikes me as an unacceptable consequence.
22. It seems to me, though, coherentists can reply to this objection along the lines I am
suggesting. Why should a coherentist accept the premise that, to enjoy first-order justifica-
tion for believing that/?, S needs to identify the totality of justifiers that justify her belief that
/?? Of course, if according to coherentism a grasp of one's entire belief system were neces-
sary for the recognition of justification, coherentism would make second-order justification
rather hard to come by.
23. Ernest Sosa also employs a distinction between animal knowledge and reflective
knowledge, but his way of drawing the distinction differs from mine. See "Reflective Knowl-
edge in the Best Circles," The Journal of Philosophy 94. 8 (Aug. 1997), 410-30, reprinted in
this volume, pp. 187-203.
24. Ancestors of this paper were presented at the Central Division meeting of the Ameri-
can Philosophical Association, April 1997, in Pittsburgh (as a commentary on an earlier
version of Goldman's paper "Internalism Exposed)," and at the 1999 Epistemology confer-
ence in Bled, Slovenia. I am indebted to Alvin Goldman for helpful e-mail correspondence
on these issues and to the participants in the Bled conference for valuable discussion.
Part IV

JUSTIFICATION AND TRUTH

This page intentionally left blank
 9

Truth as the Epistemic Goal

MARIAN DAVID

Epistemologists of all persuasions tend to invoke the goal of obtaining truth and
avoiding error. This goal seems to be of special importance to epistemology. No other
goal is invoked as frequently as this one. No other goal is given as much weight or is
treated with as much respect as this one. Here I want to explore some aspects of this
theme: the theme of truth as an epistemic goal. In particular, I am interested in what
role invocation of the truth-goal plays in epistemology and in the prospects for the
idea that truth is the only epistemic goal.

I trust the reader will already have a fairly good sense of how popular the goal of
truth is among epistemologists. Still, I want to provide some passages that indicate
to what purpose the truth-goal is usually invoked. William Alston and Laurence
BonJour have been especially explicit on this:
[Epistemic justification] has to do with a specifically epistemic dimension of evaluation.
Beliefs can be evaluated in different ways. One may be more or less prudent, fortunate,
or faithful in holding a certain belief. Epistemic justification is different from all that.
Epistemic evaluation is undertaken from what we might call "the epistemic point of view."
That point of view is defined by the aim at maximizing truth and minimizing falsity in a
large body of beliefs. . . . For a belief to be justified is for it, somehow, to be awarded
high marks relative to that aim. . . . Any concept of epistemic justification is a concept
of some condition that is desirable or commendable from the standpoint of the aim at
maximizing truth and minimizing falsity . . . (Alston 1985, 83-84)

151
152 Marian David

What makes us cognitive beings at all is our capacity for belief, and the goal of our
distinctively cognitive endeavors is truth: We want our beliefs to correctly and accu-
rately depict the world.. . . The basic role of justification is that of a means to truth, a
more directly attainable mediating link between our subjective starting point and our
objective goal. . . . If epistemic justification were not conducive to truth in this way, if
finding epistemically justified beliefs did not substantially increase the likelihood of
finding true ones, then epistemic justification would be irrelevant to our main cogni-
tive goal and of dubious worth. . . . The distinguishing characteristic of epistemic jus-
tification is thus its essential or internal relation to the cognitive goal of truth. It fol-
lows that one's cognitive endeavors are justified only if and to the extent that they are
aimed at this goal, which means very roughly that one accepts all and only those beliefs
that one has good reason to think are true. (BonJour 1985, 7-8)
Descartes sums up his Fourth Meditation with the claim that attending to what he
understands clearly and distinctly will allow him "to avoid ever going wrong" and
"to arrive at the truth." Chisholm says, at one point, that believing what is justified
and not believing what is not justified is the most reasonable thing to do "if I want to
believe what is true and not to believe what is false." Paul Moser characterizes
epistemic justification as "essentially related to the so-called cognitive goal of truth,
insofar as an individual belief is epistemically justified only if it is appropriately
directed toward the goal of truth." Richard Foley identifies the goal "of now believ-
ing those propositions that are true and now not believing those propositions that are
false" as a "purely epistemic goal." Keith Lehrer, who holds that accepting some-
thing for the purpose of attaining truth and avoiding error is a requisite to knowl-
edge, maintains that "a concern for truth and nothing but the truth drives the engine
of justification." Talking about our desire for truth acquisition, Alvin Goldman says
that "true belief is a prime determinant of intellectual value, and in particular, a criti-
cal value for justifiedness." Ernest Sosa, who holds that knowledge requires true belief
arising from intellectual virtue, characterizes an intellectual virtue as "a quality bound
to help maximize one's surplus of truth over error"; he assumes "a Ideological con-
ception of intellectual virtue, the relevant end being a proper relation to the truth."
Finally, Alvin Plantinga holds that positive epistemic status (warrant) is conferred
"by one's cognitive faculties working properly, or working according to the design
plan insofar that segment of the design plan is aimed at producing true beliefs."1
These passages come from advocates of various approaches to epistemology—
foundationalism, coherentism, reliabilism, virtue epistemology, and proper-function
epistemology are all represented. No doubt there are significant difference in em-
phasis and detail. But our theme is clearly discernible in all of them. Truth is either
explicitly referred to as a goal or aim, or it is implicitly treated as such. Moreover, it
is noteworthy that most of our authors invoke the truth-goal in connection with the
epistemic concept that is central to their account of knowledge: justification, clear
and distinct understanding, intellectual virtue, and warrant.2

II

Let us characterize the truth-goal, somewhat loosely, as the goal of believing truths
and not believing falsehoods. Using '/?' as an objectual variable ranging over propo-
 Truth as the Epistemic Goal 153

sitions, we can abbreviate this, again somewhat loosely, as the goal of believing p if
and only ifp is true. Note that the goal has two parts, a positive part (believing truths)
and a negative part (not believing falsehoods). The label "truth-goal" is less than ideal
because it deemphasizes the negative part; nevertheless, I will use it for the sake of
convenience.3
The truth-goal has just been characterized in terms of belief. This is worth em-
phasizing because most ordinary formulations expressing a desire for truth are in fact
ambiguous. Consider "I want the truth and nothing but the truth" and "I want to obtain
truth and avoid error." Although their negative parts clearly refer to not having false
beliefs, their positive parts can be interpreted as saying, "I want to believe the truth"
or as saying "I want to know the truth"—the latter may even be the more natural
interpretation. Commonplaces like "the search for truth" and "the pursuit of truth"
are afflicted with the same ambiguity. In general, any undifferentiated "desire for
truth" could refer to a desire for mere true belief or to a desire for true belief that
is knowledge. So we should be careful to distinguish the truth-goal from a more-
inclusive knowledge-goal, which might be characterized like this: It is the goal of
knowing truths and not believing falsehoods. (Note that this goal differs from the
truth-goal only with respect to its positive part. A negative knowledge-goal would
be uninteresting; it would be the goal of not knowing falsehoods, which is trivially
satisfied.) When contemporary epistemologists refer to truth as a goal or aim, they
usually make it quite clear that they are referring to the truth-goal rather than to the
knowledge-goal. As far as I can tell, this unambiguous preference for the truth-goal
over the knowledge-goal is a relatively recent phenomenon. Earlier epistemologists,
including Descartes, tended to be systematically ambiguous on this matter.4
The contemporary emphasis on the truth-goal raises a simple question: What about
the goal of having knowledge? Isn't epistemology the theory of knowledge? Given
this, wouldn't it be reasonable to expect that the knowledge-goal be at least as promi-
nent in epistemology as is the goal of having true belief, maybe even more promi-
nent? Yet, the knowledge-goal is hardly ever brought up in epistemology nowadays.
Unlike invocations of the truth-goal, invocations of knowledge as a goal are pretty
much absent from contemporary epistemological theorizing. Isn't this a bit puzzling?
I think the answer, in a nutshell, is this: Invocation of the truth-goal serves primarily
a theoretical need, a need that arises from the overall structure of epistemology. The
knowledge-goal would not serve this need. As far as epistemology is concerned, the
knowledge-goal is theoretically impotent.
Knowledge, the epistemic concept par excellence, is usually defined in terms of
belief, truth, and some other epistemic concept, say, justification: S knows p iff p is
true, S believes p, and S is justified in believing p in a manner that meets a suitable
anti-Gettier condition.5 Belief and truth, although fundamental to epistemology, are
not themselves epistemic concepts. They are the nonepistemic ingredients in knowl-
edge. This means that epistemology is not responsible for them; that is, as far as
epistemology is concerned, belief and truth are given and can be invoked to account
for epistemic concepts. The distinctly epistemic ingredient in knowledge is justifi-
cation: the concept of S's being justified in believing p (or alternatively, the concept
of a belief's being justified for S). Epistemology is certainly responsible for this
concept. Indeed, once an account of knowledge is at hand, the task of epistemology
154 Marian David

pretty much reduces to the task of giving a theory of justification. Eventually, such
a theory should do two things: (1) It should give some account of the nature of jus-
tification, of what it is for a belief to be justified; and (2) it should offer principles of
justification, principles that specify the conditions under which beliefs of one sort or
another are justified. Such a theory may to some extent relate justification to other
epistemic concepts. However, it is usually held that, at some point, the theory of
justification has to "break out of the circle" of epistemic concepts and provide a
nonepistemic "anchor" for justification by connecting it in some significant manner
with nonepistemic concepts.
It is not hard to see how the truth-goal fits into this picture. It promises to provide
a connection between the concept of justification and the concept of true belief,
tying together the different ingredients of knowledge. Most of the passages quoted
earlier seem designed to do just that; they invoke the truth-goal as a way of connect-
ing justification (or some other epistemic concept that plays a similar role in the
author's preferred account of knowledge) to the nonepistemic concept of truth. Why
is truth typically cast as a goal when this connection is made? Alston provides the
reason. It is generally agreed that being justified is an evaluative concept of some
sort: To say that believing p is justified or unjustified is to evaluate believing p, in
some sense, as a good thing or as bad thing, as having some positive status or some
negative status. The suggestion is that this type of evaluation, epistemic evaluation,
is most naturally understood along broadly teleological lines, as evaluating beliefs
relative to the standard, or goal, of believing truth and avoiding error. This may not
be a huge step toward a theory of justification—the very schematic idea of a belief's
being a "good" thing "relative" to the truth-goal has to be filled in—still, it may be
an important step.
Although knowledge is certainly no less desirable than true belief, the knowledge-
goal is at a disadvantage here because it does not fit into this picture in any helpful
manner. Invoking the knowledge-goal would insert the concept of knowledge right
into the specification of the goal, which would then no longer provide an indepen-
dent anchor for understanding epistemic concepts. In particular, any attempt to
understand justification relative to the knowledge-goal would invert the explanatory
direction and would make the whole approach circular and entirely unilluminating.
After all, knowledge was supposed to be explained in terms of justification and not
the other way round. This does not mean that it is wrong in general to talk of knowl-
edge as a goal, nor does it mean that epistemologists do not desire to have knowl-
edge. However, it does mean that it is bad epistemology to invoke the knowledge-
goal as part of the theory of knowledge because it is quite useless for theoretical
purposes: The knowledge-goal has no theoretical role to play within the theory of
knowledge.
It is tempting to describe the resulting picture of epistemology by drawing an
analogy to ethics. Epistemology treats justified belief somewhat like ethics treats right
action: Holding a justified belief is like holding a belief in the right kind of way. Truth
is treated in analogy to the good—truth is, as it were, the good as far as epistemology
is concerned. Note that knowledge does not show up as a goal here at all. Knowl-
edge is just the state of having reached the truth-goal in the right kind of way. This
picture presents the truth-goal as the epistemic goal, meaning that, for the purposes
 Truth as the Epistemic Goal 155

of epistemological theorizing, the truth-goal is to be treated as an ultimate goal and

as the only ultimate goal. Justification is to be explained somehow by relating it to
the truth goal; hence, it is not an ultimate epistemic goal. Knowledge is not a goal at
all as far as epistemology is concerned; it is not an epistemic goal.6 This picture, the
picture of truth as the epistemic goal, reflects the intuition that epistemic concerns
are, at the bottom, all about truth. Whether it is a good picture of epistemology would
seem to depend in large part on whether the truth-goal can really play the part as-
signed to it.

Ill

Do we have the truth-goal? Do we desire to believe truths and not to believe false-
hoods? A worry that might arise here is that we desire knowledge rather than mere
true belief. In response, one can point out that the truth-goal is contained in the knowl-
edge-goal and that having the truth-goal in this sense is sufficient for the theoretical
purposes for which the truth-goal is needed in epistemology.7 Another worry might
be that we do not usually desire truth for its own sake; rather, we usually desire it
because we think that having true beliefs will increase our chances of satisfying our
other desires. This seems plausible enough. However, the view that truth is the
epistemic goal does not say that we desire truth for its own sake. It says that any further
goals for the sake of which we might desire truth must fall outside the domain of
epistemology. Such further goals will be practical goals or ethical goals—goals that
do not contribute anything of theoretical relevance to epistemology as such.
But there is a more serious worry. Consider again the goal-oriented approach to
justification, that is, the idea, roughly expressed in (G) below, that the concept of
justification is to be understood as involving an evaluation of beliefs relative to the
truth-goal. Is this approach committed to the desire-thesis (D)?

(G) S's being justified/unjustified in believing p somehow amounts to S's

believing p being a good/bad thing relative to the truth-goal;

(D) S cannot be justified/unjustified in believing anything at any time, un-

less S desires believing truths and not believing falsehoods at that time.8

Note first, if the relevant desire is construed as a conscious desire, (D) is obviously
absurd. Although most of us probably do have conscious desires for truth at some
times (much like we sometimes have conscious desires for ripe bananas) such de-
sires come and go. Yet, our beliefs are surely justified/unjustified even at times when
we are not conscious of any desire for truth. So the relevant desire has to be a stand-
ing desire, one that we have even when we are not aware of it, maybe a subconscious
desire. Taken this way, (D) is not obviously absurd. But it is still worrisome. The
claim that all of "us" (all those capable of having knowledge) have such a standing
desire for truth appears to be a somewhat daring empirical claim about human psy-
chology—a claim not well supported by empirical evidence. Relevant evidence may
become available some day. When scientists finally decode magnetic resonance
156 Marian David

images from peoples' brains, will they find that everyone has I WANT TRUTH; I DON'T
WANT FALSEHOOD in their "desire box"? Or is the desire for truth supposed to be an
"implicit" desire? What would be evidence for the claim that everyone has such an
implicit desire? Moreover, it seems there could in principle be a person who never
had a standing or even an implicit desire for truth (or who lost it for a time). Wouldn't
the beliefs of such a person (at that time) still be justified/unjustified? By my lights,
the answer to this is yes. I think we can at least agree to this: If the goal-oriented
approach to justification, (G), is committed to the desire-thesis, (D), then the picture
of truth as the epistemic goal is in a certain amount of trouble. Can the approach get
by without committing itself to this thesis?
One option here is to hold that we ought to have truth as a goal, that we ought to
desire having true beliefs, or maybe better, that we ought to aim at having true be-
liefs; and just as a person's actions can be morally right or wrong relative to a good
she ought to desire or aim at, even if she does not actually desire it, so a person's
believings can be epistemically good or bad relative to a goal she ought to have, even
if she does not actually have it.9 Of course, something will have to be said about the
sense in which we ought to have it as a goal to have true beliefs. This could be taken
as a moral "ought" or as a pragmatic "ought." It might be tempting to take it as a
rational "ought." But one has to stay away from the claim that we ought to desire
having true beliefs because it is rational, or reasonable, to believe that having true
beliefs will help secure our other goals. These concepts of rational belief and reason-
able belief appear to be epistemic concepts. But the truth-goal is supposed to help
anchor epistemic concepts in nonepistemic concepts, so epistemic concepts should
not be allowed to enter into the very claims that are supposed to do the anchoring. If
we did construe the "ought" in terms of reasonability, then (G) would boil down to:
S's being justified in believing/? somehow amounts to S's believing p being a good
thing relative to its being reasonable for S to believe that having true beliefs will help
secure S's desires. A similar problem arises when it is said that we have the goal of
having true beliefs qua rational beings or qua intellectual beings. Given the present
context, "rational" and "intellectual" must not be spelled out in epistemic terms. An
alternative option is to hold that having true beliefs and lacking false beliefs is simply
good, ethically good.10 Being justified can then be said to be a good thing relative to
the truth-goal in the sense that it is a good thing relative to the good of having true
beliefs, which again allows for the possibility that we have justified beliefs even though
we might not desire this good. A rather more low-key proposal is that having true be-
liefs is a goal for all of us in the subjunctive sense that we all would desire having true
beliefs were we to reflect carefully on such matters. This is again an empirical, or quasi-
empirical, claim, and one might wonder whether it gets our psychology right.
There are some influential versions of the goal-oriented approach that do take (G)
to be committed to the thesis that we have to have the truth-goal in order to be justi-
fied, but only in the very extended sense that our cognitive faculties or virtues must
be aimed at the truth. On Plantinga's (1988) view, for example, justification (he pre-
fers "warrant") is a good thing relative to the state of having true beliefs in the sense
that a belief's being justified is a matter of its being produced by faculties that are
aimed at the truth, that is, are designed to provide us with true beliefs, and are func-
tioning properly, that is, functioning in the way they were designed to function (in
 Truth as the Epistemic Goal 157

the environment for which they were designed). Since our faculties do not have de-
sires, the difficulties arising from (D) never come up here. However, the view has to
address the intuition that our beliefs could be justified (warranted), even if the rele-
vant faculties were not designed to aim at the truth. Moreover, if evolution designed
our belief-forming faculties, then it is unlikely that they were designed to provide us
with true beliefs. It is more likely that they were designed to provide our genes with
instruments for successful reproduction; and our having false beliefs may often serve
our genes' purposes just as well as our having true beliefs. If one believes that our
faculties were designed by God, one might still worry that they were designed to let
us cope with our environment rather than to provide us with access to the truth.11
Let us consider a different take on the question whether (G) is committed to the
desire-thesis (D). It is far from obvious that a person's beliefs cannot be epistemically
evaluated relative to a goal the person does not actually have. For epistemic evalua-
tion to make sense, it may well be sufficient that there be "enough" people who ac-
tually have the truth-goal, enough to sustain a general practice of using concepts like
justification for evaluative purposes. After all, there are various types of evaluations
that evaluate objects relative to a standard but don't require that the evaluated object
have the desire to satisfy the relevant standard, and, in case the object has an "owner"
of some sort, don't require that its owner desires the object to satisfy the relevant
standard. This is a plausible point. But where does it lead?
Here is one reaction: The point leads nowhere, for it is too concerned with evalua-
tion in the sense of an activity or practice. The primary issue under discussion is
whether, on the goal-oriented approach, S's beliefs canfoejustified or unjustified even
if S lacks any desire for truth—the issue is not whether there can then be a practice
of evaluating S's beliefs as being justified or unjustified. This can be brought out by
strengthening the original worry. It could have been that all humans lacked the de-
sire for truth. In that case, there would have been no practice of evaluating beliefs as
good or bad relative to the truth-goal, for such a practice requires that there be at
least some people who think that having truth is desirable. But wouldn't our beliefs
still be justified/unjustified? What is needed is an objective account of how S's
believings can be good or bad relative to the truth-goal, an account that does not make
being justified and being unjustified depend on there being a practice of evaluating
beliefs as good or bad. Each of the proposals mentioned earlier can be seen as pre-
senting such an objective account.
Here is a naturalistic reaction to the point about evaluative practices: It has more
going for it than meets the eye. Distinguish the (nonnatural) evaluative concept of jus-
tification from the natural (nonevaluative) property, F, that the concept refers to, or
picks out, or supervenes on. The concept of justification as an evaluative concept ex-
ists only because we, or enough of us, engage in the practice of evaluating beliefs in
epistemic terms. So (G) is indeed committed to (D), with the consequence that our belief s
wouldn't be justified, that is, wouldn't fall under the concept of justification, if no one
desired having true beliefs, because there would then be no such evaluative concept
for beliefs to fall under. But this consequence is not disturbing, for it does not mean
that we wouldn't have knowledge in that case. Knowledge is true belief having natural
property F. Knowledge also happens to be true belief falling under the concept of jus-
tification. This is because we, or many of us, happen to have the truth-goal; and since
158 Marian David

beliefs having F tend to be true, we happen to desire beliefs having F, which is why we
often refer to F with the goal-oriented evaluative concept of justification: The goal-
oriented approach, (G), has latched on to an "accidental" feature of knowledge. On a
view like this, a view that would seem to sit well with strongly naturalistic theories of
knowledge, principle (G) is far less central than the picture of truth as the epistemic
goal would like it to be; it governs our evaluative epistemic concepts (the concept of
justification and its relatives) but does not get at the nature of knowledge.

IV

Let us take a closer look at the truth-goal itself. It is the two-part goal of believing
what is true and not believing what is false. It is said sometimes that the goal has to
have both parts because taken separately they could be "trivialized": Believing
everything would trivially satisfy the positive part; believing nothing would trivially
satisfy the negative part. This strikes me as a curious consideration. Surely, believ-
ing everything is humanly impossible; it may even be impossible tout court, for it
involves believing explicitly contradictory propositions. Moreover, I find even be-
lieving three things very hard, when one of them is, say, the proposition that I have
no head. Although believing nothing is not quite as impossible as believing every-
thing, those of us who are actually able to have beliefs (those of us who are not brain-
dead, who are not fetuses, who are not trees) will not find it much easier. In any case,
it is not clear what the consideration is supposed to show. It could be meant to show
that we have, or are more likely to have, the two-part goal rather than just one of the
subgoals, say, the goal of believing what is true, because we realize (implicitly?) that
the latter goal is best pursued by a strategy that would make things "too easy" on us,
thereby devaluing the goal. Since it is so obvious that the "strategy" to believe
everything is extremely hard to implement, this reasoning seems rather farfetched; it
assumes that we are wildly confused about which things are easy for us to do. Maybe
the idea is that we do not have just the goal of believing what is true because we
(implicitly) realize that the strategy that would guarantee reaching that goal (the strat-
egy to believe everything) would, if we could implement it, lead us toward a state we
do not want to be in. However, this presupposes that we already have the goal of not
having false beliefs. It is hard to find much merit in the point about trivialization.
The goal of believing what is true and not believing what is false is rather indefi-
nite. An advocate of the idea that truth is the epistemic goal may want to be a bit
more precise about what the truth-goal looks like, especially with respect to the ques-
tion of how much truth and how little falsehood is being aimed at here. The most
straightforward proposal for making the goal more precise is to interpret the content
of the goal as a universal generalization:

(T) For all propositions p, Up is true then S believes p, and S believes p only
if p is true.

Competitors to (T) are typically advanced on the grounds that they are more plausi-
bly ascribed to us than a goal with (T) as content. Such plausibility considerations
 Truth as the Epistemic Goal 159

are, of course, relevant. However, in the present context, the first question to ask is
whether the competing goal can play the part assigned to the epistemic goal, that is,
whether it can serve the need of the goal-oriented approach to justification charac-
terized by principle (G) from the previous section. If not, plausibility considerations
become irrelevant.
One might object to (T) on the grounds that its positive part is easily seen to be
crazy because there are infinitely many truths. Let us restrict p in (T) to propositions
that can in principle be grasped by S—no harm can come from that, and it assuages
worries about infinity. One might still think that the positive part makes for too ambi-
tious a goal. Adapting a suggestion from Chisholm (1977, 14), let us try to restrict (T)
to those propositions that S considers. But the resulting goal would be too narrow to
serve in (G): Issues of justification reach beyond what we consider. We can have jus-
tification for propositions we do not believe and have never considered. More impor-
tant, we have acquired many of our beliefs willy-nilly, without ever considering the
propositions involved. Surely, our willy-nilly acquired beliefs can be justified or un-
justified. The restriction would unhinge the idea that being justified can be understood
as a good thing relative to the truth-goal. Maybe it will improve matters to go sub-
junctive—to formulate the goal as the goal to be such that, for every p, if one were to
consider p, one would believe p if and only if p is true. This is more promising, but it
may still be too narrow. You may be justified in holding beliefs which, on consider-
ation, you would reject on bad grounds, say, because considering things tends to bring
out the raving skeptic in you. It would then be a bad thing, relative to the subjunctivized
goal, for you to hold such beliefs, even though intuitively you are justified in holding
them. It is suggested sometimes that (T) is implausible because we simply are not
interested in all those truths; that it should be restricted to propositions that are at least
to some extent important to S.12 It is easy to see that the resulting goal will not do at
all for (G). The goal of having all and only important true beliefs is, again, too narrow
because even our most unimportant beliefs can be justified or unjustified. Alston (1985,
83-84) proposes the goal of maximizing truth and minimizing falsehood in a large
body of beliefs. This, too, may have a flaw. Take the admittedly unlikely case of some-
one who has a large body of true beliefs without having any false beliefs. Given Alston' s
goal, one may wonder how a person could posses justification for an additional truth
that she does not believe. After all, adding a further truth to her body of beliefs will
not improve the already perfect truth-ratio.
The first Chisholmian goal and the goal of believing important truths are clearly
too narrow to play the role the truth-goal is supposed to play on the goal-oriented
approach to justification. The other two candidates, Alston's goal and the subjunc-
tive Chisholmian goal, are much more suitable, but they are somewhat problematic.
Moreover, once (T) is restricted to propositions that S can grasp in principle, the
resulting truth-goal is not subject to the worry that, given our finite nature, it is abso-
lutely impossible for us to reach. No doubt, it is extremely unlikely that anyone of us
could ever reach the state described by (T). But then, the states described by the sug-
gested alternatives are also incredibly hard to reach.131 would say that the discussed
alternatives should not keep an advocate of the view that truth is the epistemic goal
from holding out for (T) as the content of the truth-goal, if he feels the need for a
precise characterization of the goal.
160 Marian David

Of course, there is still the issue whether the goal of believing all and only grasp-
able truths can plausibly be attributed to us. For one thing, the goal seems too pre-
cise. It seems more plausible to think that people have an indefinite goal, like having
true beliefs. On the other hand, those of us who have the indefinite goal may be said
to "implicitly" have the precise goal too. Maybe quite a few of us are implicitly com-
mitted to high ideals, to goals we realize, or could easily come to realize, are practi-
cally impossible to reach. There may also be something to the thought that we would
like to be omniscient, that we would like to know all that can be known and not make
any mistakes. A desire for omniscience could be said to carry with it the goal of
believing all and only (graspable) truths. Many of us may have some such goal, but
it is daring to claim that all of us have at least one such goal. Of course, as I pointed
out in the previous section, the goal-oriented approach to justification may get by
without the thesis that we actually desire truth. There is no need to rehearse the avail-
able options. One additional point deserves mention though. The important-truth-
goal could be called on for help along the following lines (cf. Foley 1993, 17): We
have the goal of believing all and only truths that are important to us. We realize
(implicitly, on reflection) that pretty much any truth could become at least a bit im-
portant to us and might be needed without there being time for engaging in research.
This leads us (would lead us on reflection) to adopt the goal of believing all and only
graspable truths—or if not quite that, then at least some more indefinite goal in the
neighborhood. This kind of consideration will also be of help to the view that, al-
though we do not have the goal of believing all and only graspable truths, we ought
to have it, pragmatically speaking.

According to (G), the truth-goal-oriented approach to justification, having a justi-

fied belief amounts, somehow, to a good thing with respect to the truth-goal. Justifi-
cation applies to beliefs one by one. The truth-goal, on the other hand, is a global
goal: It is the goal of having beliefs that are true and not having beliefs that are false.
How are we to understand the idea that being justified in holding a particular belief
p is a good thing relative to this global goal? The most natural answer to this ques-
tion is that having a justified belief promotes the truth-goal, that it is a means for
reaching the global goal of believing what is true and not believing what is false.14
Note that having a justified belief cannot be a causal means for reaching the truth-
goal. A person can be justified in believing p even if believing p will cause her to
hold a massive amount of false beliefs later on, even if all later beliefs to which her
present belief causally contributes will be false. Similarly, a person can be unjusti-
fied in believing p even if believing p will cause her to hold a massive amount of
true beliefs later on, even if all later beliefs to which her present belief causally con-
tributes will be true. Being justified in believing p has nothing at all to do with the
causal consequences of believing p. More generally, it seems that being justified in
believing p has nothing to do with what beliefs you are going to hold in the future.
The truth-goal cannot be a diachronic goal if it is to play the role assigned to it in the
goal-oriented approach to justification; it cannot be the goal of having beliefs that
 Truth as the Epistemic Goal 161

are true and not having beliefs that are false in the long run—if it were, the causal
consequences of our beliefs would be relevant to their epistemic status. Instead, it
must be a synchronic goal: It must be the goal of now having beliefs that are true and
now not having beliefs that are false.
One can see how having a true belief is a "means," in a broad sense, to "reaching"
the synchronic truth-goal; it is what Foley (1993, 19-20) calls a constitutive means.
Having a true belief is part of what is involved in now having beliefs that are true and
now not having beliefs that are false. But given the truth-goal-oriented approach to
justification with (1) justification understood as a means to the goal and (2) the as-
sumption that the synchronic truth-goal is the goal for justification to promote, it is
now hard to see how justification could be anything but a constitutive means to the
goal, which will make justification collapse into truth. The reason is, roughly, that
with a synchronic goal only constitutive means count, and a constituent of the goal
must always be a better constitutive means than a nonconstituent. (i) Assume you have
a true belief p that is (intuitively) unjustified. The goal-oriented approach must never-
theless count believing p as a good thing relative to the goal—certainly as better than
not believing p and as better than believing the negation of p, both of which conflict
with the goal, (ii) Assume you have a false belief p that is (intuitively) justified. The
goal-oriented approach must count believing p as a bad thing relative to the goal—
certainly as worse than believing the negation of/?, which would be a constitutive means
to the goal. The upshot is that the goal-oriented approach will count all true beliefs as
justified and will not allow justified beliefs that are false: Justification collapses into
truth. Let us refer to this argument as the reductio argument.15
Stephen Maitzen (1995) gives an argument to the same conclusion but he makes
a problematic transition from one set of issues to another. The truth-goal can be in-
voked at two different levels: first, to characterize the nature of justification in broadly
Ideological terms—this is the home of principle (G); second, to explain why we value
justification, why we care whether our beliefs are justified—the presently relevant
idea being that we care only because justification is a path to the truth.16 Maitzen
describes the view that truth is the epistemic goal primarily as a monistic view about
why we value justification. His argument then moves from why we value justifica-
tion, on the monistic view, to what justification is, on that view: "If the nominal aim
[the truth-goal] is the reason for having, or pursuing justification, then it ought to
follow that beliefs are justified insofar as they serve the nominal aim and unjustified
insofar as they do not" (Maitzen 1995, 870). But this transition is worrisome. I might
value one thing merely as a means to getting another; it does not follow that the na-
ture of the one reduces to the other: I might value my dog merely to keep away the
neighbors, but the nature of my dog does not reduce to keeping away the neighbors.
In my version, the argument does not make such a transition. It proceeds entirely in
terms of considerations pertaining to what justification is according to the goal-
oriented approach.
The reductio argument should be seen as a reductio of at least one of the premises
entering into it. The consequence that all true beliefs are justified is absurd. The con-
sequence that a belief cannot be justified unless it is true is very unpalatable—con-
temporary epistemology assumes that knowledge requires only fallible justification.
I take the point of the argument to be a diagnostic one: Thinking about how a theory
162 Marian David

of justification would block it should tell us something about the role it assigns to
the truth-goal. Let us look at some of the options in broad outline.
I. A theory of justification might block the reductio argument by disconnecting
the concept of justification from the concept of truth. This option completely aban-
dons the idea that truth is an epistemic goal in the substantive sense that we are con-
cerned with here. It holds that the truth-goal has no serious theoretical role to play in
the theory of knowledge because it has no role whatsoever to play in the theory of
justification—frequent invocations of the truth-goal notwithstanding. Of course, truth
still plays a role in the definition of knowledge; it is still an epistemic goal in the
meek sense in that reaching the truth is required for the possession of knowledge,
but that is all there is to it. Knowledge is merely a conjunctive state combining two
independent goods: truth and justification. There cannot be any account of what ties
these goods together because there is no explanatory connection between them: Pos-
sessing justification cannot be understood as being a good thing relative to the goal
of having true beliefs. The position is unsatisfying and hard to sustain for any length
of time. Having characterized justification without any reference to truth, we will
feel the urge to add: "We hope . . . that the marks of evidence will also be marks of
truth" (Chisholm 1957, 38), and soon we will find ourselves asking: "What is it about
evidence, or justification, that makes us cling to this hope?"
II. A theory of justification might want to block the reductio argument by setting
up a second goal for justification to promote. This option abandons the view that the
truth-goal is the epistemic goal but retains the idea that the goal plays an essential
role in the theory of justification. As on the original picture, having a justified true
belief is interpreted as the state of having reached the truth-goal in the right sort of
way. But believing in the right sort of way is now regarded as an autonomous part of
the overall epistemic goal whose other part is the truth-goal. Possessing justification
is understood as a good thing relative to two goals, neither one being merely a means
to the other.17 Such a double-goal view is vulnerable to the objection that it addresses
only half of the reductio argument. It does block the absurd result that all true beliefs
are justified, which was the conclusion of part (i) of the argument: holding a true
belief may now fail to be a good thing relative to the overall epistemic goal because
a belief that has reached the truth may be wanting with respect to the other subgoal.
But part (ii) of the argument still goes through because justification is still required
to promote the truth-goal: Holding a false belief must be a bad thing relative to the
overall goal because it conflicts with one of its subgoals. So the unpalatable conse-
quence that there is no fallible justification is still with us.
This is not the last word, however. So far, I have paid little attention to the point
that an adequate definition of knowledge requires not only that S have a justified
true belief, but also that S's justification meet a suitable anti-Gettier condition—a
condition that makes the definition immune to the sort of examples originally pro-
duced by Gettier (1963). Let us refer to justification that meets such a condition as
justification-1-: Justification+ is whatever turns true belief into knowledge. Now, it may
be that none of the anti-Gettier conditions that have been proposed is fully success-
ful. Still, the work that has been done in this area strongly suggests that any success-
ful condition will have the following feature: A belief cannot meet the condition unless
the belief is true.18 Assuming this to be correct, it follows trivially that justification+
 Truth as the Epistemic Goal 163

will not allow for fallibly justified* belief. One might then try to defend position (II)
in the following way: The "unpalatable consequence" is not so unpalatable after all;
on the contrary, part (ii) of the reductio argument merely shows that the concept of
justification under consideration is the concept of justification*, the concept that turns
true belief into knowledge, which had better be infallible. Fair enough. However,
the defense has an unpalatable consequence of its own. Remember that any epistemic
concept one might want to employ in the theory of knowledge was supposed to be
anchored in the truth-goal. Consider now the second goal that justification, or rather,
justification*, is supposed to promote. Surely, it must be specified in epistemic terms:
"Believing in the right sort of way" must amount to something along the lines of
believing in accordance with the evidence, believing reasonably, and so on. But as
soon as we connect, say, the concept of evidence to the truth-goal, and be it only
partially, we can run the reductio argument on this concept, with the unpalatable
consequence that beliefs that are in accordance with the evidence cannot be false.
What now? The response that this merely proves evidence to be evidence*, the con-
cept that turns true belief into knowledge, would seem to imply that the concept of
evidence, which was supposed to specify the subgoal to be served by justification*,
already coincides with the concept of justification*. (Alternatively, it will send us
off in search for the next epistemic concept, the one that specifies the second goal
that is supposed to be served by evidence*; and off we go.) The upshot is that a theory
of the second goal of justification* (say, the theory of evidence) cannot anchor evi-
dence in the truth-goal. It cannot say that believing in accordance with the evidence
is a good thing with respect to the goal of having true beliefs. The justification* de-
fense merely turns position (II) into a notational variant of position (I).
III. A theory of justification might want to reconceive the relation between justi-
fication and the truth-goal. The idea is to give an account of "being justified i s a good
thing relative to the truth-goal" that applies to individual beliefs but does not boil
down to the claim that being justified is a constitutive means to reaching the truth-
goal. The hope is to block the reductio argument, without discarding the view that
the truth-goal plays a crucial role in the theory of justification and without setting up
an independent epistemic goal.
Those who favor "internalist" approaches to the theory of justification may want
to follow Chisholm and construe the connection between justification and truth as a
"rational" connection:
If I want to believe what is true and not to believe what is false, then the most reason-
able thing for me to do is to believe what is justified and not to believe what is not
justified. (Chisholm 1982, 4)19

But in what sense of "reasonable" is this the reasonable thing to dol Certainly not in
the sense that (1) it is the best means to reaching the truth-goal, for this answer would
take us back to the reductio argument. The concept of reasonableness that Chisholm
uses here to connect justification with truth is a bit like the concept of reasonable
action and a bit like the concept of reasonable belief, but not quite like either one.
Let us try to spell out this concept in terms of the ordinary epistemic concept "it is
reasonable to believe that . . .". This yields a second answer to the question raised
above: If one has the truth-goal, then having a justified belief is the reasonable thing
164 Marian David

to do in the sense that (2) it is epistemically reasonable to believe that having a jus-
tified belief is the best means to reaching the truth-goal. Note first that this answer
has a curious feature. If the reductio argument is successful, then the answer is cer-
tainly correct. Sure enough, it is reasonable to believe that having a true belief is the
best means to reaching the truth-goal—it is a constitutive means. However, the an-
swer is supposed to block the argument by not construing justification merely as a
means to truth. Assuming the answer is successful in thus blocking the argument:
Why is it then still reasonable to believe that having a justified belief is the best
means to reaching the truth-goal? This question, I think, can have no answer. One
should also worry about the epistemic concept of reasonable belief that is used in
answer (2) to connect justification with the truth-goal. If this concept is not itself con-
nected with the truth-goal, then the present position abandons the view that truth is
the epistemic goal with respect to the "connecting" concept of reasonable belief. If,
on the other hand, one connects this concept with the truth-goal by replacing the word
"justified" with the word "reasonable" in Chisholm's principle (quoted above), the
result appears to be trivial. However, the deepest worry about (2) seems to me to be
this: (2) is supposed to tell us in what sense having a justified belief is a good thing
relative to the truth-goal; yet, I find it difficult to see how it manages to convey more
than that it is reasonable to believe that having a justified belief is a good thing rela-
tive to the truth-goal. An alternative rendering of Chisholm's principle in terms of
the epistemic concept of reasonable belief would go like this: (3) If I have the truth-
goal, then, for every/?, it is reasonable for me to believe/?, if I am justified in believ-
ing/?, and it is reasonable for me not to believe/), if I am not justified in believing/).
But the consequent of this conditional is trivially true—it is true even if it is my goal
to believe all and only falsehoods.
Maybe there simply is no answer to the question in what sense believing what
is justified and not believing what is unjustified is "the reasonable thing to do" if
one has the truth-goal. Say the concept that forges the rational connection between
epistemic concepts and the truth-goal cannot be circumscribed in terms of (other)
epistemic concepts. If this is the view, then one has to decide whether the connecting
rationality-concept is nevertheless supposed to be an epistemic concept of sorts. If it
is, then the resulting position would seem to abandon the idea that truth is the epistemic
goal, for the concept that forges the rational connection between the truth-goal and
all other epistemic concepts will not itself be anchored in the truth-goal (self-anchoring
principles will end up in trivialities to the effect that it is rational to have rational
beliefs—adding an invocation of the truth-goal to such a principle merely adds an
idle wheel). If, on the other hand, the connecting rationality-concept is regarded as a
nonepistemic concept, then one can indeed block the reductio argument without
abandoning the view that truth is the epistemic goal, but at the price that the "ratio-
nal" connection between justification and the truth-goal becomes rather mysterious.
Let us take a look at how an "externalist" approach to the theory of justification
will block the reductio argument. I will focus on a simple form of process reliabilism.
It is natural to expect that a reliabilist will be disinclined to abandon the idea that the
truth-goal plays a crucial role in the theory of justification, and that he will be equally
disinclined to set up another epistemic goal besides truth. Indeed, reliabilism blocks
the reductio argument with its account of epistemic goodness: Being justified in
 Truth as the Epistemic Goal 165

believing p is a good thing relative to the truth-goal because a justified belief is one
that is reliably produced, that is, produced by a belief-forming process, or mecha-
nism, that produces mostly true beliefs.20 Not surprisingly, this looks like a promis-
ing way of handling the truth-goal. It seems to avoid construing the epistemic good-
ness of an individual justified belief merely as a constitutive means to the truth-goal;
and it seems to take seriously the idea that truth is the epistemic goal. However, as
Maitzen (1995, 873-74) points out, it is not obvious that reliabilism really is faithful
to the truth-goal. Isn't it, rather, that the truth-goal is being replaced by the goal of
having reliably produced beliefs? After all, reliabilism counts a true belief that is not
reliably produced as one that is bad relative to the goal, and it counts a false belief
that is reliably produced as one that is good relative to the goal. It seems to follow
that "the goal" referred to here cannot be the goal of having true beliefs; it must be
the goal of having beliefs that are reliably produced. If so, then the reliabilist way of
blocking the reductio argument abandons the view that the truth-goal is the epistemic
goal. The conclusion of this line of reasoning is a bit surprising. One would have
thought that, if the truth-goal plays a serious role anywhere in epistemology, then it
does in reliabilism. I take it that the natural response to this charge, the charge that
reliabilism has traded the truth-goal for the goal of having reliably produced beliefs,
is this: No such trade has taken place because "having reliably produced beliefs" does
not refer to a goal other than the truth-goal; it is just another name for (part of) the
truth-goal. Let us see whether this response is successful.
Read "G" as an operator saying something like "it is my goal that," or "I want it
to be the case that." We should distinguish between the following

(1) G[(Yp) <Bp <-» T/0 ]

(2) (Yp) [G(Bp <-» Tp) ]

The first expresses the truth-goal—a goal you could have. The second ascribes to
you a vast number of goals, not all of which you can have. It makes a false claim,
saying that every proposition (you can grasp) is such that you have a de re desire
about it to the effect that you want to believe it if and only if it is true. Obviously,
you can desire the goal-state specified in (1) without desiring each one of the goal-
states specified by the de re instances of (2). Similarly, you may be committed to the
goal expressed by (1)—it does not follow from this that each of the de re instances of
(2) is such that you are committed to it. Reliabilism sees justification in this light. It
counts an individual true belief as unjustified, when its production history implies
that (1) will be ill served if the believer should go on forming beliefs like that; and it
counts an individual false belief as justified if its production history implies that (1)
will be well served if the believer should go on forming beliefs like that. So reliabilism
regards justification as serving the greater good while being prepared to forego serv-
ing individual instances of (2). Since reliabilist justification promotes (1), it promotes
the truth-goal. Since commitment to (1) does not carry with it commitment to each
instance of (2), it does not follow that reliabilist justification is not committed to the
truth-goal if it does not promote each instance of (2).
Now, what does it mean to have a reliably produced belief? Roughly, it means
that one's belief is produced by a belief-forming process M such that, in most cases,
166 Marian David

if a belief is produced by M, then the belief is true. Take a "family" of your beliefs to
be the beliefs produced by the same process. A reliably produced belief is a member
of a family most of whose members satisfy "(Bp —> T/>) & Bp" for any proposition
p they are concerned with. To say that an individual belief is reliably produced
amounts to saying that one half of goal (1), namely (Vp) (Ep -> T/?), is being collec-
tively looked after by the family to which the belief belongs, though not necessarily
by the belief itself. So reliabilism does not abandon the truth-goal for the goal of
having reliably produced beliefs; on the contrary, to say that a belief is reliably pro-
duced (justified) is to talk about half of the truth-goal and to say that it is being pro-
moted.21 However, this does not really work. The reliability of a process that is pres-
ently producing a particular belief depends (almost) exclusively on how successful
the process's past offspring has been and on how successful the process's future
offspring will be. So, on the above account, reliabilism has to maintain that the truth-
goal is the diachronic goal of having true beliefs in the long run, where the "long
run" is meant to include the past as well as the future. This is troubling. It is very
implausible to hold that whether I am now justified in believing p depends partly, or
wholly, on the truth values of my future beliefs. In particular, the diachronic goal
will allow that the truth values of beliefs that are the causal consequences of my present
beliefs bear significantly on whether my present beliefs are justified. This is quite
unacceptable. The diachronic truth-goal has to go.
The obvious goal to put in place of the diachronic truth-goal is a subjunctive truth-
goal: For every/?, if I were to believe/), then/? would be true, and if p were true, then
I would believe p. To make things fit together, reliability must also be spelled out in
subjunctive terms; this is an intuitively more plausible understanding of reliability
anyway. Take a subjunctive family of your beliefs to be the beliefs a given process
would produce, if it were to produce a large number of beliefs. A reliably produced
belief is a member of a subjunctive family most of whose members satisfy "(Bp —» Tp)
& Bp", for any p they would be concerned with, where you should now read "—»" as a
subjunctive arrow. To talk of an individual reliably produced belief amounts to saying
that the subjunctive truth-goal is being promoted collectively by the belief's subjunc-
tive family. So where does this version of reliabilism stand with respect to the charge
of truth-goal trading? It has indeed adopted a new goal, the subjunctive truth-goal. But
it has not abandoned the old goal, for the old goal is contained in the new one. No real
trade has taken place. Is the subjunctive truth-goal an ad hoc device? I do not think so.
It strikes me as a serious contender for the part of the epistemic goal—maybe a genu-
ine improvement over the old truth-goal. Insofar as we have a desire for having true
beliefs, that desire may well be for something stronger than happy-go-lucky truth.22

Notes
1. See Descartes (1641, 1984), AT VII 62; Chisholm (1982), 4; Moser (1985), 4; Foley
(1993), 19; Lehrer (1990), 112; Goldman (1986), 98; Sosa (1985,1991), 225; and Plantinga
(1988), 39. I should point out that Chisholm's attitude toward the goal of truth has varied
quite a bit; cf., e.g., Chisholm (1977), 14 and (1986), 90-91. Moser too puts things some-
what differently in another work; cf. Moser (1989), 42 and chap. 5.
2. I will not pay much attention to the minority view that believing truth and avoiding
error might be a duty rather than a goal—an "objective duty" to truth itself. Richard Feldman
 Truth as the Epistemic Goal 167

discusses it briefly and identifies one of its main flaws: if p is false but all our evidence indi-
cates that p is true, we do not have a duty to refrain from believing p; if anything, we have a
duty to believe p and vice versa; see Feldman (1988), 245-^7.
3. I assume that authors who do not make explicit mention of the negative part still take
it to be included in the goal. If one acknowledges propositions that lack truth value (e.g., the
proposition that Jones is bald, where Jones's skull exhibits a borderline case of baldness),
then one will have to distinguish two versions of the negative part: (a) the goal of not believ-
ing falsehoods, which is silent about propositions that lack truth value; and (b) the goal of not
believing untruths, which treats propositions that lack truth value just like false ones. Are
these of equal importance to epistemology or is one to be preferred? Here I will suppress this
issues by assuming/pretending that all propositions are either true or false.
4. When Descartes says he has learned "what to do to arrive at the truth" (1641, 1984),
AT VII 62, does he mean arrive at knowledge of the truth or at true belief? It is hard to tell.
William James's well-known commandment—"We must know the truth: and we must avoid
error" (1911), 17—turns out to be ambiguous too, for right away he reformulates it as: "Be-
lieve truth! Shun error!" Of course, here the ambiguity is between commandments or duties
rather than between goals.
5. Let us not worry for now about the intricacies of handling Gettier cases. And let us not
worry too much about the choice of justification in the third condition. Various alternative
epistemic concepts might be used: reason, evidence, rationality, warrant, intellectual virtue,
and others. To simplify things, let us take justification as a representative of this extended
family of epistemic concepts (other than knowledge and its equivalents).
6. Since simplicity, explanatory power, and other features that are deemed desirable in
scientific theories, do not show up as independent elements in the account of knowledge, they
are not even prima-facie candidates for being ultimate epistemic goals. They can be relevant
only insofar as they are ingredients of justification, only insofar as they help make believing
the theory a good thing relative to the truth-goal.
7. Could one desire knowledge and, not being aware that knowledge requires truth, fail
to desire truth? Would it then still be knowledge that one desires?
8. Compare the end of the passage from BonJour quoted in section I.
9. Locke: "He that believes, without having any Reason for believing . . . neither seeks
Truth as he ought, nor pays the Obedience due to his Maker . . ." (1700, 1975), IV.xvii.24.
Chisholm: "Every person is subject to a purely intellectual requirement—that of trying his
best to bring it about that, for every proposition h that he considers, he accepts h if and only
if h is true" (1977), 14. Chisholm may well have thought that the requirement of "trying one's
best" is derived from the fact that we ought to have the goal to bring about the state he de-
scribes, and that we are subject to this requirement even if we do not have that goal. This
view should be distinguished from another view that might underlie this well-known pas-
sage, namely that the relatively subjective duty of "trying one's best" is derived from an
objective duty to truth itself—i.e., from the duty to accept the propositions we consider if
they are true—as opposed to being derived from a duty to have truth as a goal. See note 2 on
objective duties to truth itself.
10. Russell: "As for the preference which most people . . . feel in favor of true proposi-
tions, this must be based, apparently, upon an ultimate ethical proposition: 'It is good to be-
lieve true propositions, and bad to believe false ones.' This proposition, it is to be hoped, is
true; but if not, there is no reason to think that we do ill in believing it" (1904, 1973), 76.
11. Cf. Sosa (1993); and Descartes (1641, 1984), AT VII 82-89.
12. Compare Alston (1985, 1989), 83; Plantinga (1993), 33; and Steup (1996), 72.
13. There is a sense in which Alston's goal is easier to reach than (T). A person who has
a large body of only true beliefs has reached Alston's goal but not (T). However, this is ex-
actly the case that seemed to create a difficulty for Alston's goal. Note that it will be very
hard to reach Alston's goal for a person who does have false beliefs, because adding a true
belief will always improve the ratio. Since it is very hard to avoid having false beliefs, this
makes Alston's goal not significantly easier to reach than (T).
14. Compare the passage from BonJour quoted in section I.
168 Marian David

15. More or less related arguments can be found in Sartwell (1992), 172-76 (although he
handles these issues rather oddly by my lights); DePaul (1993), chap. 2.4; and Maitzen (1995).
16. BonJour: "Why should we, as cognitive beings, care whether our beliefs are episte-
mically justified? Why is such justification something to be sought and valued? . . . The basic
role of justification is that of a means to truth . . . If epistemic justification were not condu-
cive to truth . . . [then it] would be irrelevant to our main cognitive goal and of dubious worth
.. . Epistemic justification is therefore in the final analysis only an instrumental value not an
intrinsic one." (1985, 7-8).
17. Cf. DePaul (1993), 77-78.
18. Cf., e.g., Lehrer (1990), chap. 7; Moser (1989), chap. 6; and Steup (1996), 12-18.
These works contain further references to the literature on the Gettier problem.
19. My italics. Later Chisholm rejects all his earlier attempts to forge a connection of any
sort between justification and the truth goal; see his (1986), 90-91.
20. See Goldman (1979). I use an extremely simplified version of reliabilism. Also, I as-
sume there is a solution to the "generality problem"—otherwise reliabilism itself will col-
lapse justification into truth; cf. Goldman (1979), 12-13; Plantinga (1988), 24-31; Steup
(1996), 165-67. A reliabilist account of justification will have to be supplemented with a "no
undermining evidence"-condition; cf. Goldman (1979), 18-20. If this concept of evidence is
not anchored in the truth-goal, reliabilism will end up as a version of position (II) rather than
(III).
21. Of course, the other part of (1) is also promoted. But the concept of a reliable process
has to do with something of this form "B/j —> Tp" rather than that "Tp —> Bp."
22. Thanks to Leopold Stubenberg and Mike DePaul for helpful discussions. A very early
precursor of this paper was given as a talk at the Bled 1999 Conference on Epistemology in
Bled, Slovenia. Thanks to participants Stewart Cohen, Mi Ellin, Terry Horgan, Peter Klein,
Michael Lynch, Eugene Mills, Nenad Miscevic, George Pappas, Matthias Steup, Ruth Weintraub,
and Michael Williams for helpful suggestions.

References
Alston, W. P. [1985] 1989. "Concepts of Epistemic Justification," The Monist 68. Reprinted
in Epistemic Justification: Essays in the Theory of Knowledge. Ithaca, N.Y.: Cornell
University Press.
BonJour, L. 1985. The Structure of Empirical Knowledge. Cambridge, Mass.: Harvard Uni-
versity Press.
Chisholm, R. M. 1957. Perceiving: A Philosophical Study. Ithaca, N.Y.: Cornell University
Press.
. 1977. Theory of Knowledge, 2d ed. Englewood Cliffs, N.J.: Prentice Hall.
. 1982. "A Version of Foundationalism." In The Foundations of Knowing. Brighton,
U.K.: The Harvester Press.
1986. "The Place of Epistemic Justification." Philosophical Topics 14: 85-92.
DePaul, M. 1993. Balance and Refinement. London: Routledge.
Descartes, R. [1641] 1984. Meditations on First Philosophy, in The Philosophical Writings
of Descartes, vol. 2, edited by J. Cottingham, R. Stoothoff, and D. Murdoch. Cambridge:
Cambridge University Press.
Feldman, R. 1988. "Epistemic Obligations." Philosophical Perspectives 2: 235-56.
Foley, R. 1993.Working Without a Net: A Study of Egocentric Epistemology. New York:
Oxford University Press.
Gettier, E. 1963. "Is Justified True Belief Knowledge?" Analysis 23: 121-23.
Goldman, A. 1979. "What is Justified Belief?" In Justification and Knowledge, edited by
George Pappas. Dordrecht, The Netherlands: Reidel.
. 1986. Epistemology and Cognition. Cambridge, Mass.: Harvard University Press.
James, W. 1911. The Will To Believe and Other Essays in Popular Philosophy. New York:
David McKay.
Lehrer, K. 1990. Theory of Knowledge. Boulder, Colo.: Westview.
 Truth as the Epistemic Goal 169

Locke, J. [1700] 1975. An Essay Concerning Human Understanding, edited by P. H. Nidditch.

Oxford, U.K.: Clarendon Press.
Maitzen, S. 1995. "Our Errant Epistemic Aim." Philosophy and Phenomenological Research
55: 869-76.
Moser, P. K. 1985. Empirical Justification. Dordrecht, The Netherlands: Reidel.
. 1989. Knowledge and Evidence. Cambridge: Cambridge University Press.
Plantinga, A. 1988. "Positive Epistemic Status and Proper Function," Philosophical Perspec-
tives 2: 1-50.
. 1993. Warrant: The Current Debate. New York: Oxford University Press.
Russell, B. [1904] 1973. "Meinong's Theory of Complexes and Assumptions", Mind 13;
reprinted in Essays in Analysis, edited by D. Lackey. New York: George Braziller, 21-
76.
Sartwell, C. 1992. Why Knowledge is Merely True Belief. The Journal of Philosophy 89:
167-80.
Sosa, E. [1985] 1991. "Knowledge and Intellectual Virtue", The Monist 68; reprinted in
Knowledge in Perspective: Selected Essays in Epistemology. Cambridge: Cambridge
University Press, 225-244.
. 1993. "Proper Functionalism and Virtue Epistemology." Nous 27: 51-65.
Steup, M. 1996. An Introduction to Contemporary Epistemology. Upper Saddle River, N.J.:
Prentice Hall.
 10

Value Monism in Epistemology

MICHAEL R. DEPAUL

Here is a little story. It may not be true, but it may still be informative.
Once upon a time, analytic epistemologists thought that there was one significant
concept of epistemic evaluation. The epistemologists had their own pet names for
this concept. Some called it "justification," others called it "rationality"; some used
the term "warrant"; others spoke of things being "evident," while still others formu-
lated their views in terms of a "right to be sure." There were still more names than
these being used, but the epistemologists figured they were all talking about the same
thing. To simplify, I'll stick with the term "justification."
The epistemologists differed about more than their favorite name for justification.
They held significantly different views about the nature of justification and the con-
ditions under which a belief is justified as well. Some thought that for a belief to be
justified it had to be based on a proper foundation, others that it had to cohere with
the rest of the believer's beliefs, while others took being justified to require having
the right sort of cause. Many held that justification was a matter of fulfilling epistemic
obligations. Still others held that justification's essence was reliability, while others
proclaimed that proper function was the key. And these were only the more widely
held positions. As you might imagine, the advocates of the various contending views
did their share of bickering over who was right.
"Why then," you might fairly wonder, "did the epistemologists think they were
arguing about the same thing?" Because they all agreed about knowledge, and each
assumed that what he or she was talking about was what made for the difference
between mere true belief and knowledge. We now know things turned out to be a

170
 Value Monism in Epistemology 171

little more complicated. Edmund Gettier cooked up some nice examples of justified
true beliefs that were not knowledge.1 But even after Gettier had produced his ex-
amples, the epistemologists still felt quite sure that justification played a crucial role
in knowledge, even if more was involved. So they went on developing their accounts,
and responding to criticisms and criticizing the accounts that others developed in their
turn. On and on they went, criticizing and responding and refining, and refining and
criticizing and responding.
Then something started to change. Perhaps it was because no one managed to
produce a nice easy definition of knowledge. But for whatever reason, the episte-
mologists stopped worrying quite so much about knowledge. They started to realize
that every one of them (or nearly every one—some poor souls maybe did come up
entirely empty-handed) had latched onto something interesting and important and
valuable, even if at most one of them had managed to identify what was involved in
knowing. They began to think, "Having a proper foundation and reliability and
cohering and functioning properly and even fulfilling one's epistemic obligations
are all valuable and important after all. None of these concepts is perfectly clear; they
all require more clarification and explication and analysis. So why fight. There is
plenty of work for all of us to do, each on his or her own favorite concept."
And so ended the terrible, contentious time when epistemologists assumed that
there was only one important concept of epistemic evaluation and endlessly fought
about the correct analysis of this one concept. A fine new day of pluralistic tolerance
dawned. There still were clouds, of course. A few who had a taste for it continued to
argue over who had discovered what was required to turn true belief into knowledge.
But for the most part, the epistemologists from each of the various camps happily
worked away at their own favorite concept and watched with interest as the others
worked away at theirs. Each camp was secure in the knowledge that it was working
at understanding something important and valuable. There was no longer any need
to feel threatened by the successes of those in other camps.
The end.

II

What about the real epistemological world? Do we now live in the fine pluralistic
times envisioned at the end of my little story? Not really—or better, it depends on
exactly what sort of pluralism is at issue. The problem is that pluralism regarding
concepts of evaluation is compatible with monism at a deeper level—the level of
one's theory of value. Consider a simple hedonistic version of utilitarianism: Plea-
sure is good, and it is the only good thing, while pain is bad. Pain and pleasure are
negative and positive values along the same dimension. The utility of an action is a
sum of the action's positive and negative consequences, that is, the pain and plea-
sure caused by the action. One is morally obliged to maximize utility, and any act
that produces less than maximal utility is wrong. It seems that hedonistic utilitarians
clearly accept a monistic theory of value, since they hold that pleasure is the one good
thing. The contrasting pluralistic theory of value would hold, as G. E. Moore held,2
that there are various independently good things, for example, knowledge, aesthetic
172 Michael R. DePaul

appreciation, and pleasure. However, as we all know, despite their acceptance of a

monistic theory of the good, hedonists can easily generate a plurality of concepts of
moral evaluation on the basis of their one good thing: An action is objectively wrong
just in case it produces less utility than some alternative action would have produced.
An action is subjectively wrong just in case it did not seem, from the agent's per-
spective, to be an action that would have produced at least as much utility as any
alternative action. An action is blameworthy if the utility of blaming the agent for
performing that action is higher than the utility of not blaming the agent. An action
is prudent if and only if it in fact yields at least as much utility, for the agent, as any
alternative. And so on.
It is not too hard to find epistemologists acknowledging that there are a number
of different concepts of epistemic evaluation.3 Although they may not be quite so
accepting of concepts other than their own favorites as my little story suggests, there
is clearly a recognition that a fair number of these concepts are important and wor-
thy of investigation. Nevertheless, I do not think that there is much question that the
vast majority of epistemologists accept a theory of epistemic value very similar to
the hedonistic theory just described. They take truth (or true belief) to be the only
intrinsic epistemic good and falsity (or false belief) to be the only thing that is intrin-
sically bad. I would not need to be much of a scholar to put together an impressive
string of quotations to back up my claim. So excuse me if I offer only the following
representative expression of the sort of epistemic value monism I am talking about.
It is taken from Larry Bon Jour.

What then is the differentia which distinguishes epistemic justification, the species of
justification appropriate to knowledge, from these other species of justification? The
answer is to be found, I submit, by reflecting on the implicit rationale of the concept of
knowledge itself. What after all is the point of such a concept, and what role is epistemic
justification supposed to play in it? Why should we, as cognitive beings, care whether
our beliefs are epistemically justified? Why is such justification something to be sought
and valued?
Once the question is posed in this way, the following answer seems obviously cor-
rect, at least in first approximation. What makes us cognitive beings at all is our capac-
ity for belief, and the goal of our distinctively cognitive endeavors is truth'. We want
our beliefs to correctly and accurately depict the world. . . . The basic role of justifica-
tion is that of a means to truth, a more directly attainable mediating link between our
subjective starting point and our objective goal. . . . If epistemic justification were not
conducive to truth in this way, if finding epistemically justified beliefs did not sub-
stantially increase the likelihood of finding true ones, then epistemic justification would
be irrelevant to our main cognitive goal and of dubious worth. It is only if we have
some reason for thinking that epistemic justification constitutes a path to truth that we
as cognitive beings have any motive for preferring epistemically justified beliefs to
epistemically unjustified ones. Epistemic justification is therefore in the final analysis
only an instrumental value, not an intrinsic one.4

Oops! I said I would provide only one representative quote from a host of possibili-
ties, and then it seems I can't manage to pick one that expresses the view I'm after.
BonJour here commits himself to the claim that truth is of epistemic value and the
claim that justification has value only in virtue of leading to the truth, but BonJour
 Value Monism in Epistemology 173

doesn't say that truth is the only thing of epistemic value. For all he says in this pas-
sage, he could perfectly well hold that there are many things in addition to truth that
are intrinsic epistemic goods. It looks like the passage from BonJour does not illus-
trate value monism, and supposing that I was at least right about its being represen-
tative of the view that is common among epistemologists, it seems to provide no reason
to think that there is any widely shared commitment to monism.
Let's distinguish epistemology proper and broad epistemology. Epistemology
proper is the easy one to identify. It is concerned with knowledge, and more specifi-
cally, the evaluative component of knowledge. So if justification is evaluative and it
is what distinguishes knowledge from mere true belief, then justification falls within
the scope of epistemology proper. I'm not sure it is possible to delimit broad episte-
mology precisely, but I can give an example of something that falls within broad
epistemology but outside epistemology proper. In a couple of books, Richard Foley
has examined what he calls epistemic rationality.5 "Foley rationality," as some others
have taken to calling his target concept, is a matter of believing in a way that would
stand up to one's own critical reflection or, as Foley sometimes explains, believing
in accord with one's own most deeply held epistemic standards. Foley does not claim
that epistemic rationality is either necessary or sufficient for knowledge, and he seems
to be right about this. Yet, when one classifies a belief as Foley rational, one is clearly
evaluating that belief, and, I think it is safe to say, evaluating it epistemically rather
than prudentially or morally or in some other way. So Foley rationality provides us
with one example of something that falls outside epistemology proper, but within
broad epistemology. It is not too hard to think of other things that fall only within
broad epistemology—indeed, that is where the various sorts of justification will end
up that turn out not to be involved in knowledge that I mentioned in my little story.
What is hard, as I've already admitted, is specifying what distinguishes broad
epistemic evaluation from other sorts of evaluations. Perhaps this indicates that the
things that fall under broad epistemology are linked only by a family resemblance.
Or perhaps they do not even share that much, being linked only by the fact that they
were once, mistakenly, thought to have been a component in knowledge. Neverthe-
less, I hope it is clear enough what I mean by broad epistemology.
Now personally, I suspect that epistemologists have a kind of unthinking tendency
to assume that truth is the only epistemic good even within broad epistemology. I
will not attempt to substantiate that suspicion, however, for I think that deep down
we do all recognize that truth is not the only thing of epistemic value. Here is an easy
demonstration. Take your favorite example of a well-established empirical theory, a
theory you believe that we know. Throw in all the evidence on the basis of which we
accept that theory. Depending on the theory you selected, all this will likely add up
to a substantial number of beliefs. Now, compare this set of beliefs with an equal
number of beliefs about relatively simple arithmetic sums and about assorted ele-
ments of one's current stream of consciousness. I suspect that most of us would want
to say that the first set of beliefs is better, epistemically better, than the second set.
But the two sets contain the same number of true beliefs. And so, to the extent that
we are inclined to say that these sets differ with respect to broad epistemic value, it
would seem that we are committed to saying that truth is not the only thing that has
broad epistemic value.6
174 Michael R. DePaul

But that's broad epistemology. When I selected the passage from BonJour, I had
epistemology proper in mind, and when it comes to epistemology proper, I think the
tendency to hold that truth is the only epistemic good becomes very much stronger.
BonJour could not have provided a much stronger statement of the view that as far
as justification is concerned, truth is the only epistemic good. He holds justification
to be of value solely as an instrument for the acquisition of truth. This brings him
nearly all the way to a version of value monism for epistemology. During the per-
haps mythic age that Gettier brought to an end, when it was universally assumed that
knowledge is justified true belief, he would have been all the way to value monism.
After Gettier, it was clear that something more was required for knowledge than jus-
tified true belief, so it is possible that something other than justification adds value
to true belief. But I think I can safely set that possibility aside, at least for the time
being, since justification has always been taken to be the significant evaluative ele-
ment within knowledge.7 And so I think the passage from BonJour does represent
the sort of value monism I wish to examine. And to the extent that this statement by
BonJour and the numerous similar statements that could easily be found in the writ-
ings of other epistemologists have long been accepted without question, I think it is
safe to go so far as to say that the assumption that truth is the only epistemic good
dominates epistemology proper.

Ill

So where does this leave us? I have made a couple of distinctions:

(a) monism versus pluralism with respect to concepts versus value
(b) epistemology proper versus broad epistemology
And I have made three claims:
(i) Contemporary analytic epistemology has come to accept pluralism with respect
to concepts,
(ii) At least as far as epistemology proper is concerned, epistemology is dominated
by value monism.
(iii) The sort of value monism that dominates epistemology proper takes true belief to
be the one and only epistemic good.
I realize that I have not gone to any great lengths to explain my two distinctions, but
then, they are fairly straightforward. And I confess I've done next to nothing to ex-
plain or defend my three claims, but that is because I don't believe they're at all
controversial. So far so good. Now let me add one more trivial claim:

Knowledge is epistemically better than mere true belief.

Now, I'm afraid, we have a problem, or, to be more precise, contemporary episte-
mologists have a problem. Personally, I've got no problem, but that's because I'm
not a monist with respect to epistemic value. What is the problem supposed to be? It
is that knowledge cannot be epistemically better than mere true belief IF true belief
is the only epistemic good. The point seems so simple and clear that I'm not sure
 Value Monism in Epistemology 175

how to go about arguing for it, yet it does not seem to have been recognized. Indeed,
far from being recognized, as the quote from BonJour illustrates, the general view
seems to be that the rationale of the concept of knowledge requires that justifica-
tion has value merely as an instrument to the truth. But certainly a central, and I
would think nonnegotiable, element of our concept of knowledge is that it is of
great epistemic value, and, more specifically, that it is more valuable than mere
true belief. But it simply cannot be if truth is the only epistemic good. When one has
a true belief, one has already attained the only thing of epistemic value—one's be-
lief is as epistemically good as beliefs can get. No matter what other characteristics
one's true belief might have, these characteristics cannot add any additional epistemic
value, for we are supposing that truth is the only epistemic good, and the belief in
question is a belief that is true! The belief cannot be true two times or three times
over; it is true and that is that.

IV

Before trying to go any further along this line of argument, there is something I should
try to clear up. It should already be obvious that my focus in this paper is on ques-
tions regarding epistemic good or value. Claims that something is good or valuable,
or that it is better or more valuable than something else, can be unclear. Particularly
when they are made in an unfamiliar context—and we really are not used to focus-
ing explicitly on questions regarding epistemic value—such claims can leave us
wondering about a number of questions, for example: "From what perspective is the
evaluation being made? For whom is this thing supposed to be good? Who values
this more than that? Are we talking about intrinsic goods or instrumental goods?"
How, then, do I intend my claims about what is epistemically good or valuable to be
understood?
One way in which claims that a thing is good or valuable can be intended is as
applying to some particular person, as when we say that something is good/or some-
one.8 When we say that something is good for a person, we ordinarily mean more
than merely that the person in question desires or wants or prefers or values the thing.
We recognize the possibility that the actual desires, wants, preferences, or values of
the person can be deficient in some way. A very simple way in which they might be
deficient is by not extending to things that are necessary for, or highly efficient means
to, the things the person actually desires or wants or prefers or values. Another simple
way in which the actual desires, wants, preferences, or values of a person might be
deficient is by being inconsistent. A person might, for example, want two things that
are, as a practical matter, incompatible with each other. If the person wanted one of
these things very much more than the other, we would be unlikely to say that the
thing the person wants much less is good for her; other things being equal, we would
probably say the thing she desires more is good for her. Yet another way in which
actual desires, wants, preferences, or values can be deficient is by having been formed
on the basis of an incomplete or mistaken understanding of the facts. A person might,
for example, want to eat a rare steak, not realizing that it is infected with mad cow
disease. Given that the actual desires, wants, preferences, or values of a person can
176 Michael R. DePaul

be deficient in at least these ways, perhaps when we assert that something is good
for a person we mean to assert that the person would desire, want, prefer, or value
the thing in some sort of improved or ideal circumstances, for example, if the person
were informed of the relevant facts or resolved conflicts.
Many philosophers consider evaluative claims understood along these lines to be
relatively secure or respectable, since the evaluation is explicitly grounded in psy-
chological states. Indeed, not a few philosophers seek to interpret all claims that
something is good or valuable as claims about what a person would desire, want,
prefer, or value in appropriately specified circumstances. I have doubts about whether
this approach can succeed even for claims that something is good for someone, but
I do not want to get involved in any controversy regarding the analysis of evaluative
claims here. For my purposes, it is sufficient to note that we surely sometimes claim
that things are good without having anyone in mind for whom the things are sup-
posed to be good. We surely sometimes assert that things are good regardless of
whether any person actually desires, wants, prefers, or values the things; any person
should desire, want, prefer, or value the things given what he or she does desire, want,
prefer, or value; or whether any person would desire, want, prefer, or value the things
under improved circumstances. Such evaluative claims are sometimes described as
being made "from the point of view of the universe as a whole."
Even though this sort of evaluation strikes many philosophers as mysterious and
problematic, I'm afraid that this is the way to understand the claims about epistemic
goods that concern me. I want to raise questions about what is epistemically good,
and what is epistemically better than what, from a first person point of view. I do not
want to be looking at some other person or group and asking what is good for them.
I want to ask what is epistemically good and what is epistemically better than what.
When we approach questions about what is good and what is better from this point
of view, it is most natural to understand them in the absolute, from the point of view
of the universe as a whole sort of way. When we consider for ourselves whether
something is good, or better than something else, we are obviously trying to decide
whether or how much to value the thing in question or whether we should, or how
much we should, value the thing. But I think it is pretty clear that we are not explic-
itly pondering whether we already actually desire, want, prefer, or value it. And since
we obviously sometimes try to decide whether things are intrinsically good, I don't
think we need to be considering whether the thing is necessary for or an efficient
means to some other thing we value. And I surely don't think we are considering
whether we would desire, want, prefer, or value it in some sort of ideal circumstances.
At least in the case of intrinsic goods, it seems to me that we look at things the other
way around: We try to figure out what is good and assume that our desires, wants,
preferences, or values will adjust themselves accordingly. At least this is how things
appear on the surface. Perhaps at some deeper level of semantic or metaphysical
analysis a different story must be told. I here take no stand on that issue.
Another issue I do not wish to take a stand on is whether the epistemic good or
goods are good for their own sakes or in virtue of leading to nonepistemic goods of
one sort or another. Having confined myself to epistemology proper, I have so far
assumed that truth is epistemically good, knowledge is epistemically good, and knowl-
edge is epistemically better than truth. It seems obvious that in many instances, true
 Value Monism in Epistemology 177

belief is good as an instrument to nonepistemic ends. For example, having a true belief
regarding my PIN is good as a means to my aim of attaining money from an ATM.
Is true belief also intrinsically good, or is it good only as a means to nonepistemic
goods? Interesting question, but it is not what concerns me here. My aim is to con-
sider the things we take to be epistemic goods, consider them in isolation from other
sorts of goods, and see what sort of sense we can make of them. I hope I can reach
some interesting conclusions regarding epistemic goods considered on their own.
It would certainly be worth going on to consider how epistemic goods fit in with
goods of other types. Perhaps it will turn out that epistemic goods are intrinsically
good and stand on a par with such other intrinsic goods as happiness. This is cer-
tainly the view that most philosophers have accepted, but perhaps it will turn out that
epistemic goods are good merely as means to other good things. Alternatively, it might
turn out that epistemic goods conflict with other things we value, so that we must
seek some sort of balance. It might even turn out that when we attempt to put to-
gether a broad view of the things we consider to be good, we will not be able to find
a place for epistemic goods at all. Perhaps they can only be good as means to non-
epistemic goods but in fact are not effective means to those goods. This last would
be a disturbing conclusion, much at odds with the bulk of philosophical reflection.
But it is a possibility we need to consider. The general question of how epistemic
goods fit into an overall picture of the good is, in my opinion, a much-neglected matter
to which epistemologists need to devote much more attention than they have. Unfor-
tunately, it is not a question I can here address. I will be well satisfied if I can make
a reasonable start on a sensible view of the epistemic good.

Let's return to the main issue. I have already indicated that I take it to be uncon-
troversial that knowledge is epistemically better than mere true belief. I also think it
is uncontroversial that true belief is epistemically good. But I maintain that if mo-
nism is true—if true belief is the only epistemic good—then it cannot be the case
that knowledge is better than mere true belief.
I do not pretend to know how knowledge should be analyzed. As the story goes,
philosophers always assumed that knowledge was simply justified true belief until
Edmund Gettier published his two counterexamples to this analysis. Few took the
moral of the Gettier and related subsequent counterexamples to be that justification
is not the significant element in the analysis of knowledge. Most assumed that knowl-
edge had to be justified true belief plus something else that rules out the sort of prob-
lem Gettier uncovered. For convenience, I'll call the something else "the Gettier
thing." So a widely held view, no doubt the dominant view, is that knowledge is true
belief plus justification plus the Gettier thing. If this view is right, then the most
obvious way for knowledge to be more valuable than mere true belief is by either
justification or the Gettier thing adding some additional value to the value of true
belief.
In a number of writings, Alvin Plantinga has proposed a significant alternative
account.9 He proposes to use the term "warrant" to indicate whatever it is that makes
178 Michael R. DePaul

the difference between mere true belief and knowledge. Plantinga prefers "warrant"
to the more usual "justification" because he thinks justification is heavily laden with
"deontological" associations—it suggests rights, duties and obligations—and accord-
ing to Plantinga, it is a serious mistake to try to account for knowledge in terms of
deontological evaluation, for example, in terms of responsible belief or belief that
violates no epistemic obligation. On Plantinga's way of dividing things up, we need
not mention the Gettier thing, at least not up front. We can simply say that the most
obvious way for knowledge to be more valuable than mere true belief is by warrant
adding some additional value to the value of true belief.
Despite the greater simplicity of Plantinga's account, I am going to work with the
account in terms of justification. My main reason for this is that "justification" is the
more familiar, traditional term.10 By opting for justification, I certainly do not intend
to presuppose a deontological understanding of the evaluative component of knowl-
edge. At this point, I wish to be noncommittal about the nature of justification, under-
standing it, as Plantinga initially understands warrant, by factoring it out of the equa-
tion for knowledge. The important difference between justification and warrant
therefore has to do with the Gettier thing. Warrant includes it and justification does
not. I think it is best, for my purposes, not to include the Gettier thing so as to call
attention, first, to the possibility that the Gettier thing is either in part, or significantly,
or even primarily responsible for the superior value of knowledge, and second, to
the assumption that this possibility is unrealized. I think this is a widely shared as-
sumption. Despite the large literature attempting to repair the traditional justified true
belief account of knowledge in response to the Gettier problem, there is a tendency
to downplay the significance of the Gettier thing and to view it as some kind of tech-
nical gimmick. When viewed in this way, it is very natural to assume that the Gettier
thing has no significant role to play in accounting for the value of knowledge. I will
go along with this assumption, but I will not argue for it, and I must admit that it
might be mistaken.11
Justification, then, is that which must be added to true belief and the Gettier thing
to get knowledge. And I'm for the moment assuming that the superior value of knowl-
edge is to be explained in terms of the value that justification adds to true belief. We
now need to ask about the value of justification. As we saw above, the dominant view
is that justification is instrumentally good; its value is to be understood in terms of
its connection to the truth. No matter what the ultimate nature of justification turns
out to be, if it is an instrumental good of this sort, it must be connected, in some sort
of fairly reliable way, to truth.12 Considered on its own, it isn't apparent that there is
anything wrong with trying to understand the good of justification in this way. There
are plenty of things that are good because of a reliable connection with some sort of
more basic good. But what happens when we turn our attention to knowledge? It
would seem that the idea is to understand the good of knowledge by conceiving of it
as a state where the good of truth has been attained, and it has not been attained in
just any old way, for example, by luck or by accident, but has been attained by way
of a reliable means to this good.
I do not think that there is anything preposterous about the notion of a thing that
is good in this general way. It is easy enough to think of examples of such goods.
There used to be a TV commercial for a financial institution in which a pompous
 Value Monism in Epistemology 179

older gentleman said, "We make money the old fashioned way: we earn it."13 Allow
me to twist this statement, just a little, so that it serves my purpose. It suggests a good—
wealth—that might have been attained in various ways. One might have worked for
it, that is, adopted a course of action that one could count on to produce the good
reliably, or one might have attained the good in some other way, for example, by
placing a winning bet on a horse race. The suggestion seems to be that the state of
having attained the good in the first way is better than the state of having attained the
good in one of the other ways. Here is another example. General A formulates a bril-
liant strategy for battle, carries through on it and wins. General B doesn't adopt any
coherent strategy at all, but wins because of several unlikely occurrences over which
he had no control. It would not be unreasonable to consider General A's victory su-
perior to General B's. Presumably, the brilliance of General A's strategy largely
consists in the fact that the strategy maximizes the chances for victory, so General
A's victory is a state where a good is attained, and it is attained in a way that reliably
leads to the good in question.
Perhaps we do not usually have particular terms to designate states of success-
fully attaining goods by reliable means. But my two examples suggest that states of
so attaining goods are familiar enough and that we not only value such states, but not
uncommonly value them more than states of having attained the same goods by other
means. So what is wrong with holding that the concept of knowledge is a concept of
such a state, and the value of knowledge is to be understood accordingly? Quite
possibly nothing. There is only a problem if we strictly adhere to the idea that
justification is good merely as a means to the truth. If we do, then one just cannot
maintain that a justified true belief is better, in terms of the good of believing the
truth, than a true belief plain and simple. To the extent that we really consider attain-
ing the truth by reliable means to be better than attaining the truth by other means,
we value forming beliefs reliably for its own sake, apart from any connection with
the truth.
I would want to say the same sort of thing about my other two examples. If all
you care about is money, it makes no sense whatsoever to value the attainment of X
number of dollars more or less depending on the means of attainment. If all you care
about is victory on the battlefield, it makes no sense to value victory won in one way
to victory won in some other way. It does not, of course, follow that it makes no
sense to value one attainment of X number of dollars more than another attainment.
I hope we would all value the honest attainment of a fortune over the dishonest at-
tainment of a similar fortune. But this just means we value honesty as well as for-
tune, that is, that we are not thinking about the case as monists. Similarly, it makes
perfect sense to value the victory won by a brilliant strategy more than a victory won
by dumb luck. But that is because in addition to taking victory to be good, we con-
sider excellence a good thing and incompetence a bad thing. Once again, it makes
sense to value attaining the goal in one way more than attaining it in another, but
only because more than one value is in play.
There is another, less happy, possibility. We might simply be confused about
knowledge. Clearly some good things are good only because they reliably lead to
other things we value. Think again about the investment firm earning its profits.
Clearly the people who made the ad want us to picture the firm working hard to re-
180 Michael R. DePaul

search the investments it makes. We are supposed to value well-researched invest-

ments because such investments tend to pay off. That's why we are supposed to entrust
our money to the firm. There is no puritanical valuing of sweat and toil for its own
sake at work. We value money and we value well-researched investments because
we rightly think that they are more likely to make us more money over the long haul.
But if that is the whole story about our values, then it really would not make any
sense to value money made the old-fashioned way more than money acquired in any
other way. If we do value it more, it is because we get into the habit of valuing good
means to an end and then assign more value to attaining the end via those means
than we assign to attaining the end in some other way. But such valuing is simply
confused. It is a result of our having lost sight of why we valued the means in the
first place. The case of knowledge might be similar. Justification is valuable purely
as a means to truth. We get into the habit of valuing it, forgetting that we value it
merely as a means to truth and recalling only that we value it. Then, when we en-
counter a justified true belief, we think it is more valuable than a mere true belief
because it has the good of justification as well as the good of truth. But this is clearly
a mistake—the good of justification is no different from the good of truth, so when
we add justification to a true belief we add nothing of value. In essence, we count the
value of truth twice over and so think knowledge is better than true belief.
I don't think we are mistaken or confused in valuing knowledge more than mere
true belief, but I do not think there is any quick proof that we are not. The decision
would have to be made on the basis of an overall comparison of the alternative ac-
counts we might give of the value of knowledge. I do believe that the account we
have just been considering, which attributes our inclination to regard knowledge as
better than mere true beliefs to confusion, is a live option. But it is not the only live
option. I have tried to show that we cannot coherently explain how knowledge is
better than mere true belief if we assume that truth is the only epistemic good. But if
we are pluralists about epistemic value and take justification to be good apart from
any connection it might have to the truth, then we can explain the superior value of
knowledge. So we have at least two live options. I happen to favor the pluralist op-
tion, but I will not argue for it here.

VI

I would now like to examine a less obvious way in which we might account for the
value of knowledge.14 In the last section, I considered several "additive" accounts of
the value of knowledge. The accounts assume that true belief is a good thing and
then consider whether or how justification might add some value to the value of true
belief to account for the value of knowledge. On such an approach, the value of
knowledge is seen as a sum of the values of its parts. But perhaps knowledge is an
organic unity. An organic unity is a thing with a value that does not equal the sum of
the values of its parts. The most familiar sort of organic unity is a thing with valuable
parts but with a value that is greater than the sum of the values of its parts. Plausible
candidates for such organic unities can be found among collectibles and art objects.
 Value Monism in Epistemology 181

Individual pieces of fine antique porcelain are valuable, but a complete set is more
valuable than the sum of the values of the individual pieces. Each panel of a early
Christian triptych is valuable, but the whole is more valuable than the sum of the
three parts.
It is arguable that there are other sorts of organic unities as well. Some have main-
tained, for example, that in certain circumstances something that is ordinarily valu-
able can decrease the value of a whole, as in a case where a person takes pleasure in
the suffering of another. Pleasure is a good thing, but in such a case it does not to any
degree mitigate the bad situation of a person suffering; it only makes the total situa-
tion worse. It would seem to be good to feel empathy for the suffering person, which
is to say that the addition of a kind of suffering, ordinarily a bad thing, serves to in-
crease the value of the total state of affairs. Otherwise, neutral states of affairs might
play similar roles in certain circumstances. If a person takes no pleasure in the suc-
cess of a close friend, this would seem to make the total situation worse. I have heard
military people maintain that it is appropriate for soldiers to be indifferent to the
violence they must do to their enemies in battle. It certainly would be a bad thing for
soldiers to take pleasure in such violence. And perhaps if they were pained by the
violence, this would interfere with them performing their duty. If so, then perhaps
adding the neutral state, taking neither pleasure nor pain in their actions, makes the
state of affairs better where soldiers must do violence in battle.
One might think the supposition that knowledge is an organic unity might enable
the monist to account for the superiority of knowledge over mere true belief. The
monist might claim that even though justification is good only as a means to truth
and hence can, strictly speaking, add no value to a true belief, the combination is an
organic unity that has a value in excess of the sum of the values of its parts. Alterna-
tively, the monist might claim that justification is not good as a means to truth, that
it is always neutral, but its inclusion in knowledge still renders that organic whole
more valuable than the sum of its parts, specifically, more valuable than mere true
belief. Perhaps, but I cannot help feeling that both possibilities are highly implau-
sible. I don't mean to say it is highly implausible that knowledge is an organic unity.
I'm willing to grant, at least tentatively, that it is. What I find implausible are the two
suggestions about how to turn this to the advantage of monism. The suggestion that
the combination of attaining a good (truth) and attaining it by way of something good
only as a means to that good (justification) has more value than the good itself does
not strike any intuitive cord with me. Quite the contrary, it strikes me as intuitively
suspect. I am, I think, even less sympathetic to the other suggestion, since it seems
obvious that justification is a good thing, not a neutral thing.
However, even if I am just missing something wonderful about these suggestions,
they cannot really provide comfort to the monist. For no matter how the details of
the story go, if true belief is a good thing, and knowledge is a good that cannot be
wholly understood in terms of the goods involved in its parts, then we just do not
have a version of monism. We have just admitted that there are at least two good
things: true belief and knowledge. One cannot maintain the claim that true belief is
the only good involved in epistemology proper while holding that knowledge is a
good that cannot be accounted for in terms of the good of true belief.
182 Michael R. DePaul

VII

It's time to wind down. I have taken it for granted that knowledge is a good thing
and that it is better than mere true belief. I tried to show that if a form of epistemic
value monism that takes true belief to be good is true, there is no way to account for
knowledge being epistemically superior to mere true belief. More particularly, I have
examined the very common notion that epistemic justification is good as a means to
the truth and tried to explain that if it is, and the Gettier thing does not make a sig-
nificant contribution, there is no satisfactory explanation for the value of knowledge.
Let me close by briefly mentioning where taking seriously the idea that knowl-
edge is an epistemic good superior to mere true belief might lead. I have made it
clear that I am sympathetic to value pluralism. I take true belief to be a good thing
and knowledge to be a good thing. I also think justification is good. But if justifica-
tion is to play a role in accounting for the good of knowledge, it must be good inde-
pendent of any connection between justification and the truth. How so? We need an
account of the good of justification, and I am well aware that I have not taken any
steps toward providing such an account. In trying to provide one, it would make sense
to consider various accounts of justification that have been offered to see whether
any suggest plausible accounts of the value of justification and then knowledge. I
can put this idea more aggressively: If knowledge is better than true belief, and jus-
tification plays the most significant role in accounting for the value of knowledge,
then no account of justification that fails to provide an understanding of the values
of justification and knowledge can be acceptable. Of course, it isn't as though we are
buried under apparently acceptable accounts of justification, wishing we could find
some way of eliminating some of them. But that does not mean it is a mistake to
eliminate accounts of justification that do not allow us to understand how justifica-
tion, and especially knowledge, are good. Even if it has been neglected, providing
such an understanding is one of epistemology's more important tasks.

Notes
I presented a version of this essay at the 1999 Bled Conference on Epistemology. I am in-
debted to the audience there for feedback and also to the Institute for Scholarship in the Lib-
eral Arts at the University of Notre Dame for funding my travel to the conference. I should
also acknowledge debts to Marian David and Linda Zagzebski. I have benefited from many
discussions with Marian of the issues addressed in this essay. Although I have not had occa-
sion to refer to her work at any specific point in this essay, I have benefited from hearing
Linda's criticism of reliabilist accounts of the value of knowledge in "From Reliabilism to
Virtue Epistemology," presented at the 1998 World Congress of Philosophy.

1. "Is Justified True Belief Knowledge?" Analysis 23 (1963): 121-23.

2. G. E. Moore, Ethics (Oxford: Clarendon, 1912).
3. See, for example, William Alston, "Epistemic Desiderata," Philosophy and Phenome-
nological Research 53 (1993).
4. Lawrence BonJour, The Structure of Empirical Knowledge (Cambridge, Mass.: Harvard
University Press, 1985): 7-8.
5. Richard Foley, The Theory of Epistemic Rationality (Cambridge, Mass.: Harvard Uni-
versity Press, 1987) and Working Without a Net (New York: Oxford University Press, 1993).
6. Notice that the set of beliefs containing the scientific theory is not epistemically better
because its elements are better justified than the elements of the other set. According to the
 Value Monism in Epistemology 183

usual view, beliefs about simple sums and about current elements of our conscious experi-
ences will be maximally well justified and quite possibly certain.
7. I will have a little more to say about this issue in section V.
8. Such a claim could be made for a group as well as for an individual.
9. See especially Platinga, Warrant: The Current Debate (New York: Oxford University
Press, 1993), and Platinga, Warrant and Proper Function (New York: Oxford University Press,
1993).
10. I also suspect that readers would tend to take "warrant" to indicate Plantinga's par-
ticular theory of warrant.
11. I am indebted to Peter Klein, who does not assume that the Gettier thing is insignifi-
cant, for calling this assumption to my attention. There are two other things to note about the
assumption: (1) If the assumption is that the Gettier thing does not add any value to justified
true belief, it is obviously mistaken. The very examples that Gettier gave to show that justi-
fied true belief is not sufficient for knowledge also serve to show that we do not value justi-
fied true belief as much as we value as knowledge. The interesting assumption, then, con-
cerns the relative contributions of justification and the Gettier thing to the value of knowledge.
It is, specifically, that justification does the lion's share of the work. (2) Even if the Gettier
thing is responsible for a significant amount of the value of knowledge, epistemic value plu-
ralism does not obviously follow. It may be that the value of the Gettier thing can be explained,
as many assume the value of justification can be explained, in terms of some connection with
truth. Such an approach is at least suggested by efforts to understand the Gettier thing in terms
of the person's justification not essentially involving any falsehoods.
12. How reliable the connection must be depends on the value of truth and what alterna-
tive instruments are available. Where something very significant is at stake and there is no
very effective means to the end, people value things that are, by any objective measure, very
poor means to the end. A good example is provided by the drugs first used to treat AIDS.
When better instruments are available, even a very reliable means can seem worthless. A person
who discovered a 90 percent effective cure for AIDS might well win an Noble Prize, but an
army procurement officer who sent soldiers into battle with small arms that fire only 90 per-
cent of the time would be court-marshaled.
13. I think "old fashioned" here must mean something like "moderately old fashioned,"
since the really old fashioned ways of making money are by inheriting it or by taking or extort-
ing it from those who are weaker.
14. Thanks to George Pappas for reminding me of this possibility.
This page intentionally left blank
Part V

EPISTEMIC VIRTUE AND CRITERIA

OF JUSTIFIED BELIEF
This page intentionally left blank
 11
Reflective Knowledge
in the Best Circles

ERNEST SOSA

Moore's Proof

Is the existence of external things just an article of faith? Certainly not, says G. E.
Moore, and offers us a proof, thus aiming to remove Kant's "scandal to philosophy."

Moore's Proof
Here is a hand (a real, flesh and bone hand). Therefore, there is at least one
external thing in existence.1

According to Moore, his argument meets three conditions for being a proof: first,
the premise is different from the conclusion; second, he knows the premise to be the
case; and, third, the conclusion follows deductively.2 Further conditions may be re-
quired, but he evidently thinks his proof would satisfy these as well.
As Moore is well aware, many philosophers will feel he has not given "any satis-
factory proof of the point in question."3 Some, he believes, will want the premise
itself proved. But he has not tried to prove it and does not believe it can be proved.
Proving that here is a hand requires proving that one is awake, and this cannot be
done.
Does Moore adequately answer the skeptic?4 Many have denied it for the reason
that he fails to rule out a crucial possibility: that our faculties are leading us astray,
for example, that we are dreaming.
Aware of this objection, Moore grants, in "Certainty," that to know he is standing
he must know he is awake.5 The point "cuts both ways," however, and he would prefer
to conclude that he does know he is awake since he does know he is standing.

187
188 Ernest Sosa

This has persuaded nearly no one. On the contrary, some have thought him com-
mitted to an argument, M (below), like the following:

Argument A
Al. This map is a good guide to this desert.
A2. According to the map an oasis lies ahead.
A3. Therefore, an oasis lies ahead.

Argument M
M1. My present experience is a veridical guide to reality (and I am not dreaming).
M2. My present experience is as if I have a hand before me.
M3. Therefore, here (before me) is a hand.

When challenged on premise Al, our desert dullard responds: "I must know Al since
the only way I could know A3 is through argument A, and I do know A3." Is this a
just comparison? Is Moore's response to the skeptic relevantly similar?6
If Moore depends on argument M for his knowledge of M3, his response seems
like the dullard's. The dullard is wrong to respond as he does. He must say how he
knows his premise without presupposing that he already knows the conclusion. And
Moore would seem comparably wrong in the analogous response to the skeptic. In
explaining how he knows Ml, he must not presuppose that he already knows M3.
Does Moore depend on argument M for his knowledge of M3? There is reason
to think that he does not, given his emphatic acknowledgment that he cannot prove
M3. After all, M would seem a proof of M3 just as good as Moore's own "proof of
an external world." Moore concedes, in effect, that r/he does not know that he is
not dreaming, then he does not know of the hand before him. But that is not neces-
sarily because he takes himself to know M3 only through M or any other such argu-
ment. And, in any case, even if he is relying on some such argument, which would
require making that concession, the defender of common sense has other options.
One might, after all, make that concession only because of the following "prin-
ciple of exclusion":

PE If one is to know that h, then one must exclude (rule out) every possibil-
ity that one knows to be incompatible with one's knowing that h.

As Moore grants explicitly, the possibility that he be just dreaming is incompatible

with his knowing (perceptually) that he has a hand before him. And this, in combi-
nation with PE, is quite sufficient to explain his concession above.
Suppose Moore is not depending on argument M for his knowledge of M3. Al-
though he recognizes his need to know he is not dreaming, suppose that is only be-
cause he accepts PE, our principle of exclusion. Then the sort of ridicule cast on the
dullard is misdirected against Moore. What is more, it is not even clear that Moore
must know how he knows he is not dreaming if he is to know M3. That is not
entailed by the application of the principle of exclusion. All that follows from the
 Reflective Knowledge in the Best Circles 189

application of that principle is that Moore must know that he is not dreaming, not
that he must know how he knows this.
In fact, however, the historical Moore did rely on something very much like argu-
ment M (more on this below). So is he not after all exposed to the damaging com-
parison with the desert dullard?
Not at all. There seems no good reason why, in responding to the skeptic, Moore
must show how he knows he is not dreaming. Of course, his response to the skeptic
would be enhanced if he could show that. But it now seems not properly subject to
ridicule even if he is not then in a position to show how he knows he is not dreaming.
The question he is addressing is whether he knows that he is not dreaming and, at
most, by extension, what grounds he might have for his answer to that question; in
answering which he does not, nor need he, also answer the question of how he knows
himself to be awake and not dreaming.
It might be replied that one cannot know that here is a hand if one's belief rests
on the unproved assumption that one is awake. According to Moore, however, things
that cannot be proved might still be known. Besides, even though he cannot prove
that he is awake, he has "conclusive evidence" for it. Unfortunately, he cannot state
his evidence, and the matter is left in this unsatisfactory state at the end of "Proof
of an External World." But Moore has more to say in another paper of the period,
"Four Forms of Scepticism."7 There he takes himself to know for sure about the
hand before him and takes this knowledge to be based on an inductive or analogi-
cal argument. We are told that introspective knowledge of one's own sensory
experience can be immediate, unlike perceptual knowledge of one's physical
surroundings. While agreeing with Russell that one cannot know immediately
that one sees a hand, Moore thinks, contra Russell, that he can know it for certain.
And he disagrees with Russell more specifically in allowing knowledge for cer-
tain about his hand through analogical or inductive reasoning from premises known
introspectively.
However, it is doubtful that any allowable form of inference—whether deduc-
tive, inductive, or analogical—will take us from the character of our experience to
the sort of knowledge of our surroundings that we ordinarily claim.
Familiar skeptical scenarios—dreaming, evil demon, brain in a vat, and so on—
show that our experience prompts but does not logically entail its corresponding
perceptual beliefs. Experience as if there is a fire before us does not entail that there
is a fire there, experience as if here is a hand does not entail that here is a hand, and
so forth. Perhaps what is required for one's beliefs and experiences to have certain
contents entails that these could not possibly be entirely false or misleading. Indeed,
some such conclusion follows from certain externalist and epistemic requirements
on one's justified attribution of familiar contents to one's own experiences or be-
liefs. But even if that much is right—which is still controversial—one's experience
or belief that here is a hand, or yonder a fire, might still be wildly off the mark. We
cannot deduce much of our supposed knowledge of the external from unaided prem-
ises about our experience.
As for inductive or analogical reasoning, only abductive reasoning (inference to
the best explanation) offers much promise, but it seems questionable as a solution to
190 Ernest Sosa

our problem.8 Suppose (1) that we restrict ourselves to data just about the qualitative
character of our own sensory experience, and (2) that we view belief in a common-
sensical external world as a theory postulated to explain the course of our experi-
ence. What exactly is the proposal? Is it proposed that when ordinarily we accept the
presence of a hand before us, we do know, and know on the basis of an abductive
inference? Or is it proposed, rather, that in such circumstances we have resources
that would enable us to know if only we used those resources to make effective
abductive arguments? The second, more modest, proposal is too modest, since it leaves
our ordinary perceptual beliefs in a position like that of a theorem accepted through
a guess or a blunder, one that we do have the resources to prove after much hard
thought, but one that we have not come close to proving at the time when we are just
guessing or blundering.
Even the modest proposal, moreover, seems unlikely to succeed. Could we form
a rich enough set of beliefs purely about the qualitative character of our sensory
experience, one rich enough to permit abductive inferences yielding our com-
monsense view of external reality? This seems doubtful when we consider (1) that
such pure data beliefs could not already presuppose the external reality to be in-
ferred and (2) that the postulated commonsense "theory" of external reality must
presumably meet constraints on abductive inference, for example, that the postu-
lated theory be empirically testable and also simpler and less ad hoc than alterna-
tives (e.g., Berkeley's). These requirements plausibly imply that our data must go
beyond detached observations and include some acceptable correlations. Yet these
correlations are unavailable if we restrict ourselves to beliefs about the character
of our experience.9 Most especially are they unavailable, and most especially is
the postulated inference implausible, when our database is restricted, as it is by
Moore, to introspectively known facts of one's own then present subjective expe-
rience and to directly recalled facts of one's own earlier experience. (If deprived
of the epistemic resources of testimony and of retentive memory—except insofar
as such resources can be validated by reason-cum-introspection, which is not very
far if at all—then there is precious little we can any longer see ourselves as know-
ing, thus deprived.)
Accordingly, the skeptic has a powerful case against Moore's claim that our knowl-
edge of the external is based on an inductive or analogical inference from such in-
formation about our experience. It is not realistic to suppose that we consciously make
such inferences in everyday life. It is less implausible to conceive of such inferences
as implicit and/or dispositional, but even this strains belief. Besides, even granted
that we make such inferences if only implicitly, do they yield simpler and less
ad hoc hypotheses than alternatives? That is far from clear; nor do such hypotheses
seem empirically testable and credible simply as explanations of the purely qualita-
tive character of our then present or directly recalled experience.
Having reached a dead end, let us have some second thoughts on Moore's view
of perceptual beliefs as inferential. Here he joined a venerable tradition along with
Russell himself. If perceptual knowledge is thus mediate and inferential, what knowl-
edge can qualify as immediate and foundational? Modern philosophy begins with
Descartes's canonical answer to this question.10
 Reflective Knowledge in the Best Circles 191

Descartes's Circles

Descartes had two circles, not only the big famous one involving God as guarantor
of our faculties, but also a smaller one found in the second paragraph of his third
meditation, where he reasons like this:
I am certain that I am a thinking being. Do I not therefore also know what is required
for my being certain about anything? In this first item of knowledge there is simply
a clear and distinct perception of what I am asserting; this would not be enough to
make me certain of the truth of the matter if it could ever turn out that something
which I perceived with such clarity and distinctness was false. So I now seem to be
able to lay it down as a general rule that whatever I perceive very clearly and dis-
tinctly is true.11

And yet when he looks away from particular clear and distinct items such as the propo-
sition that he thinks, Descartes grants that a powerful enough being could deceive
him even about what seems most manifest. Descartes grants that he could be astray
in his beliefs as to what he perceives or remembers, even in taking himself to intuit
something as quite clear and distinct. This doubt must be blocked if one is to attain
certainty by intuiting something as clear and distinct. Accordingly, Descartes launches
the theological reflections that lead eventually to his nondeceiving God.
Even without the further boost of certainty provided by the proof of a nondeceiving
God, however, Descartes takes himself to have attained some positive justification.
Early in the third meditation he takes himself to perceive clearly and distinctly that
he thinks, which he takes to be what gives him the certainty that he thinks. And he
reasons that this clear and distinct perception would not give him such certainty if it
were less than perfectly reliable and apparently concludes from this that his clear and
distinct perception is perfectly reliable. One could demand how he knows all these
things: how he can be sure that he does clearly and distinctly perceive that he thinks,
for one thing; how he can be sure that there is nothing else in his situation that could
provide the degree of certainty involved; and how he can be sure that the clarity and
distinctness of his perception could not possibly provide that degree of certainty unless
it were infallible. What could he say in response? Descartes might well have a uni-
form response to all such questions: In each case he might appeal once again to clear
and distinct perception, each of the things in question being something we are as-
sured of by our clearly and distinctly perceiving it.
About the cogito, I wish to highlight, not Descartes' s answers to such questions,
but the inference that he draws: So I now seem to be able to lay it down as a gen-
eral rule that whatever I perceive very clearly and distinctly is true. Just what is
Descartes's argument in support of this general rule? Would his reasoning take the
following form?
1. Datum: I know with a high degree of certainty that I think.
2. I clearly and distinctly perceive that I think, and that is the only, or anyhow the best
account of the source of my knowledge that I think.
3. So my clear and distinct perception that I think is what explains why or how it is
that I know I think.
192 Ernest Sosa

4. But my clear and distinct perception could not serve as a source of that knowledge
if it were not an infallibly reliable faculty.
5. So, finally, my clear and distinct perception must be an infallibly reliable faculty.

The move from (1) and (2) to (3) is an inference to an explanatory account that one
might accept for the coherence it gives to one's view of things in the domain involved.
Descartes does elsewhere appeal to coherence at important junctures.12 So he may
be doing so here as well, although questions do arise about how Descartes views
coherence. Does he accept the power of coherence to add justified certainty, and, in
particular, would he claim infallibility for (sufficiently comprehensive and binding)
coherence as he does for clear and distinct intuition?13 In any case, the comprehen-
sive coherence of his world view would be enhanced by an explanation of how clear
and distinct perception comes to be so highly reliable, even infallible. And this is
just what Descartes attempts, through his theological and other reasoning. Descartes
can see that reason might take him to a position that is sufficiently comprehensive
and interlocking—and thereby defensible against any foreseeable attack, no holds
barred, against any specific doubt actually pressed or in the offing, no matter how
slight. Unaided reason might take him to that position. Need he go any further? What
is more, might one reach a similar position while dispensing with the trappings of
Cartesian rationalism?

Circular Externalism

Compare now how Moore might have proceeded:

1. Datum: I know with a high degree of certainty that here is a hand.
2. I can see and feel that here is a hand, and that is the only, or anyhow the best ac-
count of the source of my knowledge that here is a hand.
3. So my perception that here is a hand is what explains why or how it is that I know
(with certainty) that here is a hand.
4. But my perception could not serve as a source of that degree of justified certainty if
it were not a reliable faculty.14
5. So, finally, my perception must be a reliable faculty.
Moore could, of course, go on to say more about the nature of the perception that
assures him about the hand. He might still say that such perception involves an im-
plicit inference from what is known immediately and introspectively, perhaps an
inductive or analogical inference of some sort. And that might make his view more
comprehensively coherent, but we have already seen reasons why postulating such
an inference is questionable. So we focus rather on a second alternative: Moore might
well take perceiving to involve no inference at all, not even implicit inference, but
only transfer of light, nerve impulses, and so on, in such a way that the character of
one's surroundings has a distinctive impact on oneself and occasions corresponding
and reliable beliefs. This might also amount eventually to a comprehensively coher-
ent view of one's knowledge of the external world. Its epistemologically significant
features would not distinguish it in any fundamental respect from the procedure fol-
lowed by Descartes.
 Reflective Knowledge in the Best Circles 193

The theme of accidentally true belief has loomed large in the epistemology of recent
decades. The Gettier problem, for example, is posed by a justified belief true for
reasons far removed from whatever causes it to be held and justified. Externalist
conceptions of prepositional knowledge focus on this theme, as do one offered by
Peter Unger (nonaccidentally true belief) and one offered by Alvin Goldman (belief
caused by the truth of its content). And Nozick's tracking account is also a concep-
tion of this sort: S knows that/? if and only if S believes correctly that/?, and also (in
the circumstances): both it would have been true that/? only if S had believed it, and
if it had not been true that p, then S would not have believed it.
Why are these conceptions of knowledge of special interest to us here? Because
each offers a way to explain how one can know that p without reasoning from prior
knowledge. The key idea exploited is this: You can know something noninferentially
as long as it is no accident or coincidence that you are right.
Both the tracking and the causal accounts defensibly require a special nonaccidental
connection between the belief and the fact believed. Nevertheless, in each case, other
levels of accidentality remain. Suppose I fancy myself a connossieur of tomato ripe-
ness but suffer from a rare form of color blindness that precludes my discerning nearly
any shade of red except that displayed by this particular tomato. Therefore, my judg-
ments of tomato ripeness are in general apt to be right with no better than even chance.
But when it's the particular (and rare) shade of red now displayed, then I am nearly
infallible. Oblivious to my affliction, however, I issue judgments of tomato ripeness
with abandon over a wide spectrum of shades of red. Assuming that, unknown to
me, the variety of tomato involved always ripens with this shade of red, then my
belief that this tomato is ripe is in step with the truth, and arguably satisfies the re-
quirements of Unger, Goldman, and Nozick. But, again, it is nevertheless in some
relevant sense or respect only an accident that I am right in my belief.15 We need a
clearer and more comprehensive view of the respects in which one's belief must be
nonaccidentally true if it is to constitute knowledge.
Unaided, the tracking or causal requirements proposed suffer from a sort of tun-
nel vision. They permit too narrow a focus on the particular target belief and its causal
or counterfactual relation to the truth of its content. Just widening our focus will not
do, however, if we widen it only far enough to include the process that yields the
belief involved. We need an even broader view.

Virtue Epistemology

[When]. . . thought is concerned with study, not with action or production, its good or
bad state consists [simply] in being true or false. For truth is the function of whatever
thinks. . . , 16

Hence the function of each of the understanding parts is truth; and so the virtue of each
part will be the state that makes that part grasp the truth most of all.17

Virtue epistemology is distinguished by its emphasis on the subject as seat of justi-

fication. To qualify as knowledge, a belief must be "apt," epistemically so, in a strong
sense that goes beyond its being just a belief that coheres well within the subject's
194 Ernest Sosa

perspective. The "tracking account" (Nozick) sees here little more than a claim about
that belief's counterfactual relation to the truth of what is believed. "Reliable indica-
tor" accounts require rather that the belief itself or the reasons for it have properties
nomically sufficient for its truth (Armstrong, Swain). "Reliable process" accounts
focus instead on the cognitive process, beneath the skin, that yields the belief, and
on the truth ratio in the products of that process, actual and counterfactual (Goldman).
It is rather the subject and her cognitive virtues or aptitudes that hold primary inter-
est for virtue epistemology. Consider the athletic virtues of a tennis champion. When
we say that a shot is not just a winning shot but a skillful one, we imply a comment on
the player as shotmaker. Suppose a tyro wields a racquet on a court and, unaware of the
approaching ball, issues what amounts by luck to a stylish and effective backhand stroke.
Such a shot might be an unreturnable winning shot, but it would manifest no real skill.
Why are we unwilling to admire a performance as "skillful" if it manifests only a
fleeting or even an instantaneous state of the agent's? Skills, abilities, competences,
aptitudes, prowess—these come and go, true enough, but they do not flit by instan-
taneously. Why not? Why do we tend to define these concepts so as to require such
stability? We might have defined similar concepts without requiring stability. Why
do we define these concepts as we do? Why have we adopted these and not others?
Should one not expect that, other things being equal, the more clearly useful a con-
cept is to us, the more likely it is that we shall retain it? People need to know who are
dependable members of their group—this is a kind of thing we need to monitor in a
great variety of contexts, with a great variety of objectives. Cooperative success
depends on the group's ability to monitor people's aptitudes and ineptitudes. So it is
no surprise that the sorts of aptitudes (skills, competences, virtues) that we recog-
nize and admire are those that linger stably.18
To praise a performance as skillful or an action as right, or a judgment as wise or
apt, accordingly, is to assess not only the action or the judgment, but also the reflected
aptitude or character or intelligence. This is a distinctive view with versions both in
epistemology and in ethics. It is distinctive in that the Tightness of an action (or a
choice) and the aptness (or positive epistemic status) of a belief would involve not
just whether the performance is optimific (if an action) or true (if a belief); nor just
whether a good-enough procedure was followed, perhaps accidentally, in arriving
at that choice, or whether a good-enough cognitive process chanced to lead to the
belief; nor even just whether a rule somehow in effect demands that choice or that
belief in those circumstances. Our virtue epistemology and virtue ethics focus rather
on the agent and cognizer. When the agent's actions are said to be right and the
cognizer's beliefs to be knowledge, we speak implicitly of the virtues, practical or
intellectual, seated in that subject, which (1) give rise to that action or belief, adding
to the subject's worth as agent or cognizer, and which (2) make him reliable and trust-
worthy over an interesting spread of possible choices (or beliefs) and circumstances.

Virtue and Coherence

Can we explain what distinguishes a system of beliefs (and experiences) that, inter-
nally regarded, is intellectually virtuous and admirable? Presumably our explana-
 Reflective Knowledge in the Best Circles 195

tion would involve the system's explanatory coherence, its overall simplicity and lack
of ad hoc epicycles, and so on. Philosophical mythology contains creatures who excel
in all such respects, though their beliefs fall short of being knowledge even when
true. Take the brain in a vat, for example, or the victim of the evil demon. An adult
recently envatted, or victimized by the demon, can be indistinguishable from the best
of us in respect of the comprehensiveness and coherence of their beliefs and experi-
ences. Even when right about environing objects and events, such a victim's beliefs
are far from being knowledge.
Might such comprehensive internal coherence exhaust all cognitive or intellec-
tual virtue, at least when it comprises not only beliefs but also experiences? To as-
sent here would be overbold even for a rationalist. Yet coherence is, of course, val-
ued not only by philosophers but more generally by the reflective. One also wants
faculties and virtues beyond reflective, coherence-seeking reason: perception, for
example, and memory. Equally, internal coherence goes beyond such faculties, and
requires reason, which counts for a lot in its own right. But why should that be so, if
comprehensive coherence is no guarantee of truth, if the internal coherence enjoyed
by the envatted yields little if any truth?
Compare first reason with memory. Input beliefs are required for retentive memory
and inferential reason, which then yield beliefs as outputs. Retentive memory yields
again the input belief itself, while inferential reason yields a new belief. Even the
most excellent transmission faculty will not guarantee the truth of its output, which
will depend not only on the transmission but also on the inputs. But our transmission
faculties are valuable even so, if only because they combine with other faculties to
increase vastly the total yield of true beliefs.
How does internal coherence, of little significant epistemic value in itself, be-
come more valuable when combined with external aptness? Coherence-seeking
inferential reason, like retentive memory, is of epistemic value when combined with
externally apt faculties of perception because when so combined, it (like retentive
memory) gives us a more comprehensive grasp of the truth than we would have in
its absence.
Good perception is in part constituted by certain transitions from experiences to
corresponding beliefs, as is the transition from the visual experience characteristic
of a tomato seen in good light to belief in the tomato. Other such transitions help
constitute good introspection, as when one's headache prompts awareness of it as a
headache. Finally, if the comprehensive coherence of one's system of beliefs is at
least in part responsible for its constitution and persistence, it thereby manifests a
virtuous faculty of reason. Such comprehensive coherence is not just mechanical,
but must reflect appropriate sensitivity to factors like "adhocness," simplicity, and
explanatory power. And it must include not only belief/belief connections, but also
experience/belief connections constitutive of good perception, and conscious-state/
belief connections constitutive of good introspection. This broader conception of the
coherence of one's mind involves not only the logical, probabilistic, and explana-
tory relations among one's first-order beliefs, but also coherence between these be-
liefs and one's sensory and other experiences, as well as comprehensive coherence
between first-order experiences, beliefs, and other mental states, on one side, and on
the other beliefs about first-order states.
196 Ernest Sosa

We may well ask about certain aspects of broad coherence—for example, the
experience/belief transitions as well as the enumerative and abductive inferences
involved—why these should be viewed as adding to the subject's intellectual worth
or merit. "Because it is truth conducive," or at least in good measure for that reason,
we are told, "because it increases the likelihood that the subject will have true beliefs
and avoid false ones." But that is obviously false of victims in skeptical scenarios,
who nevertheless are internally coherent and even epistemically justified.
Although that seems undeniable, we can perhaps understand it comfortably if we
distinguish two sorts of epistemic justification: (1) S is "same-world justified" in
believing P in world W iff S believes P in W in virtue of a faculty that in W is truth
conducive; and (2) S is "actual-world justified" in believing P in world W iff S be-
lieves P in W in virtue of a faculty that in our actual world is truth conducive.
Such relativizing and con textual i zing is familiar enough in ordinary thought and
speech. Here it enables us to combine the following theses: (a) Our broad coherence
is necessary for the kind of reflective knowledge traditionally desired, and (b) such
broadly coherent knowledge is desirable because in our actual world it helps us ap-
proach the truth and avoid error. This is not to deny that there is a kind of "animal
knowledge" untouched by broad coherence. It is rather only to affirm that beyond
"animal knowledge" there is a better knowledge. This reflective knowledge does
require broad coherence, including one's ability to place one's first-level knowledge
in epistemic perspective. But why aspire to any such thing? What is so desirable,
epistemically, about broad coherence is, in part, integrated understanding that it can
help us to attain, and for which it is requisite. Partly because it is truth conducive,
even if in a demon world broad coherence fails this test and is not truth conducive.
Even so, we can still regard broad coherence as intellectually valuable and admirable
so long as we do not regard our world as such a world.
We are now, it seems to me, in just the position of arch-internalist Descartes.
Consider the following passage:

The fact that an atheist can be "clearly aware that the three angles of a triangle are equal
to two right angles" is something I do not dispute. But I maintain that this awareness
of his [cognitionem] is not true knowledge [scientia], since no act of awareness that
can be rendered doubtful seems fit to be called knowledge [scientia]. Now since we
are supposing that this individual is an atheist, he cannot be certain that he is not being
deceived on matters which seem to him to be very evident (as I fully explained). And
although this doubt may not occur to him, it can still crop up if someone else raises the
point or if he looks into the matter himself. So he will never be free of this doubt until
he acknowledges that God exists.19

Descartes considers reasons to doubt not only one's faculties of perception, memory,
and introspection, but even one's faculty of intuitive reason, by which one might know
that 3 + 2 = 5, that if one thinks one exists, and the like. And he defends against such
doubts by coherence-inducing theological reasoning that yields an epistemic perspec-
tive on himself and his world, in terms of which he can feel confident about the re-
liability of his faculties, including the very faculties employed in arriving, via a priori
theological reasoning, at that perspective on himself and his world, the perspective
that enables him to see his world as epistemically propitious.20
 Reflective Knowledge in the Best Circles 197

In structure, virtue perspectivism is thus Cartesian, though in content it is not.

Radical rationalism admits only (rational) intuition and deduction (along with
memory) as its faculties of choice (or anyhow of top choice) and wishes to validate
all knowledge in terms of these faculties. Thus the Cartesian grand project. Virtue
perspectivism admits also perception and introspection, along with intuition and
deduction, as well as inductive and abductive reasoning. Gladly using all such facul-
ties, through testimony it accepts also the aid of one's epistemic community. Fortu-
nately, the overview thus attained inspires confidence in the means used.
Rejected as viciously circular by Descartes's critics, and by many today, our pro-
cedure does present a troubling aspect of circularity. However, a closer look may
show this to be only an illusion.21

Epistemic Circularity: What's the Problem?

"I think, therefore I am," says Descartes, adding: "Here at last is something I really
know. But what is it about this knowledge that makes it knowledge? As far as I can
see, it is knowledge because it is a clear and distinct intuition. But it would not be
real knowledge unless such intuition were reliable. So I can already lay it down as a
general rule that clear and distinct intuition is reliable."
"Here is a hand," says G. E. Moore, adding: "Here is something I really know.
But what gives me this knowledge? As far as I can see, it is knowledge in virtue of
being a deliverance of perceptual experience. But it would not be knowledge if I were
dreaming. So I can already conclude that I am not dreaming."
Descartes goes on to buttress the reliability of his rational intuition by developing
a theology through vigorous use of that very rational intuition. And Moore can simi-
larly appeal to what he knows about his reliable senses on the basis largely of those
very senses.
But isn't any such reasoning circular? Yes, circular it does seem to be, "episte-
mically circular," let us say. But is it viciously circular? Skeptics through the ages have
attacked it as such. Sextus Empiricus already used the tropes of Agrippa to develop
the so-called diallelus, or "problem of the criterion." And many have followed his
lead in a long tradition. Today skepticism cum relativism has spread beyond episte-
mology and ethics, beyond philosophy, and even beyond the academy, and its cham-
pions often wield circularity as a weapon. But, again: Is such circularity vicious? To
say that it is vicious, in the present context, is to say that it is somehow bad, intellec-
tually bad, that it puts us in a situation that is somehow intellectually unsatisfactory.
When we ask how the circularity is vicious, therefore, what we want to know is just
how it puts us in an unsatisfactory state: When we reason in the way alleged to be
viciously circular, wherein lies the defect in our reasoning or in the resulting state?
Largely through the use of rational intuition, Descartes supports the view that
rational intuition is reliable and that through its exercise he knows that he thinks and
exists.
Largely through the use of perception, Moore could support the view that percep-
tion is reliable, that he is not misled by a dream, and that through the exercise of
perception he knows of the hand before him.
198 Ernest Sosa

If a crystal ball claims itself to be reliable, then, largely through the crystal ball, a
crystal ball gazer could support the view that such gazing is reliable, that it is rarely
misleading, and that through the crystal ball he can foretell the future.
Epistemic circularity is vicious, it might be said, because it would make the gazer
as well justified as Descartes or Moore. Since there is no way to support adequately
the view that intuition is reliable, or that perception is reliable, without employing
those very faculties; and since the same goes for memory, deduction, abduction, and
testimony; therefore, there is no way to arrive at an acceptable theory of our knowl-
edge and its general sources.
Perhaps that shows only how defective is the attempt to develop such a general
theory of one's knowledge and it sources. There is an easy way to avoid the intellec-
tual discomfort of having to use a faculty in answering the question whether that
faculty is reliable: namely, not to ask the question. Call this the avoidance strategy.
Of course, we will hardly lack company if we avoid philosophy because we find
it frustrating. But the avoidance strategy that I wish to consider is not just a rejection
of what seems too difficult for one's own intelligence. The implication of the avoid-
ance strategy is not that there is something lacking in one's intelligence but that there
is something wrong with the questions avoided.
Much might indeed be wrong with our very general, philosophical questions. Many
find them too abstract, too impractical, too useless, and so on. But these are not the
concerns of our avoidance strategist. He is after all a philosopher. His concern is
neither that the questions are just too hard for his intelligence nor is it their abstract-
ness, impracticality, or uselessness. He would hardly have gone into philosophy with
such concerns, nor are they his concerns now. Difficulty, abstractness, impractical-
ity, and uselessness are not in his view disqualifying drawbacks.
Why then should one as philosopher avoid questions of epistemology, such as
those about the reliability of one's faculties? These questions become pressing with
the realization that only if they reliably yield truth can our faculties yield knowledge.
This is not just a commitment peculiar to contemporary reliabilism. Indeed, it is found
already in Descartes, who, as we have seen, also stresses that intuition (and clear and
distinct perception) yields knowledge only if reliable.
Consider again our principle of exclusion:

PE If one is to know that p then one must exclude (rule out) every possibility
that one knows to be incompatible with one's knowing that p.22
(By "excluding" here I mean "knowing not to be the case.")

On the basis of PE, we can see that, to know that/?, one must know that the faculties
employed in arriving at one's belief thatp are reliable faculties. After all, just con-
sider the possibility that one's operative faculties be unreliable. That is surely a pos-
sibility generally known to be incompatible with attaining knowledge through them.
Unreliable mechanisms of belief acquisition will not yield knowledge.
If the principle of exclusion is right, therefore, one cannot possibly know that
p unless one knows that the faculties involved are reliable. But this is just the
sort of knowledge that we seem able to attain only through epistemically circular
reasoning.
 Reflective Knowledge in the Best Circles 199

One might, of course, question the principle of exclusion.23 One might hold that
to know that p, one's pertinent faculties need only be reliable; one need not know
them to be reliable.
One might, for example, appeal to a conception of knowledge as mere tracking.
One might grant Rorty that causation should not be confused with justification, while
joining Nozick in taking tracking as the essence of knowledge. "To know is just to
mirror (or to track) nature. Justification is quite another matter. Justification of some
sort may well require the principle of exclusion. Thus it may be that in order to be
justified in believing that p, one must exclude every possibility one knows to be in-
compatible with one's knowing that/?. But such justification is not required for simple
knowledge."
That response seems essentially right. What is more, even Descartes would agree.
For Descartes, you will recall, our knowledge that our faculties are reliable, even our
faculty of reason, depends on our knowledge of God's epistemic good will. Yet
Descartes grants explicitly that the atheist mathematician can know some mathemat-
ics—as we have seen.
The knowledge of an atheist is said to be cognitio, however, a second-class
accomplishment by comparison with scientia. Scientia, by contrast, does require rele-
vant knowledge of one's reliability. Only thus can one repel doubts about the pos-
sible unreliability of one's faculties. Only thus can one exclude a possibility evidently
incompatible with one's knowing that/?: namely, the possibility that only unreliable
faculties yield one's belief.
By analogy we can more generally distinguish animal knowledge, which requires
only that one track nature on the one hand, and on the other reflective knowledge,
which requires also awareness of how one knows, in a way that precludes the
unreliability of one's faculties. Unlike Descartes's cognitio and scientia, our more
general animal and reflective knowledge do not require infallible reliability, but
only a high level of reliability.24
The avoidance strategy now has not only the cost of suppressing philosophical
curiosity about knowledge. We can now see how it also precludes first-level reflec-
tive knowledge, and, of course, scientia.
Given these costs, what again counts in favor of avoidance? So far we have been
told that we must avoid epistemic circularity because it entails arriving at a generally
positive view of one's faculties only by use of those very faculties. But why should
that be frustrating when it is the inevitable consequence of its generality? So far the
answer is only that the superstitious crystal-gazer could reason analogously and with
equal justification in defense of his own perspective. How damaging is this?
Suppose we grant the gazer epistemic justification and internal coherence equal
to our own. Still, internal coherence is clearly insufficient. Isn't that obvious in view
of paranoia, hypochondria, and similar psychoses? Logical brilliance permits logi-
cal coherence but does not even ensure sanity, much less general epistemic aptitude.
There are faculties other than reason whose apt functioning is also crucial to the
subject's epistemic welfare.
In light of that result, why not distinguish between the gazers and the perceivers
in that, although both reason properly and attain thereby coherence and justifica-
tion, only the perceivers are epistemically apt and attain knowledge?
200 Ernest Sosa

On this view, the crystal ball-gazers differ from the perceivers in that gazing is
not reliable whereas perceiving is. So the theory of knowledge of the perceivers is
right, that of the gazers wrong. Moreover, the perceivers can know their theory to be
right when they know it in large part through perception, since their theory is right
and perception can thus serve as a source of knowledge. The gazers are by hypothe-
sis in a very different position. Gazing, being unreliable, cannot serve as a source of
knowledge. So the perceivers have a good source or basis for their knowledge, but
the gazers, lacking any such source or basis, lack knowledge.
Still one might insist that the perceivers should not be so smug. They should still
feel acute discomfort and intellectual frustration. This I find a very widely shared
view, in epistemology and, mutatis mutandis, far beyond. According to Barry Stroud,
the perceivers can at best reach a position where they can affirm the conditional propo-
sition that if their perception is reliable, then they know.25 And he has recently reem-
phasized what is essentially the same thesis as follows:

Sosa's "externalist" could say at most: "If the theory I hold is true, I do know or have
good reason to believe that I know or have good reason to believe it, and I do under-
stand how I know the things I do." I think . . . we can see a way in which the satisfac-
tion the theorist seeks in understanding his knowledge still eludes him. Given that all
of his knowledge of the world is in question, he will still find himself able to say only
"I might understand my knowledge, I might not. Whether I do or not all depends on
how things in fact are in the world I think I've got knowledge of."26

However, it is not easy to understand this position. If our perceivers believe (a) that
their perception, if reliable, yields them knowledge, and (b) that their perception is
reliable, then why are they restricted to affirming only the conditional, a, and not its
antecedent, bl Why must they wonder whether they understand their relevant knowl-
edge? Indeed, to the extent that they are really convinced of both a and b, it would
seem that, far from being logically constrained to wondering whether they know,
they are, on the contrary, logically constrained from so wondering. After all, first, if
you are really certain that p, then you cannot well consider whether you know it
without thinking that you do. Moreover, second, isn't it incoherent to be convinced
that p and yet wonder whether pi
In sum, I see no sufficient reason to settle either for irresoluble frustration or for
the avoidance strategy. The main argument we have seen for that depends on the
claim that if we allow the circular defense offered for externalist epistemology, then
the gazers turn out no less epistemically justified than the perceivers. In a sense that
is true: But then in a sense they are equally internally justified, equally coherent.
Nevertheless, they are not equally apt in all epistemically relevant respects. Percep-
tion is, of course, reliable whereas gazing is not. Therefore, the perceivers are right
and apt both in their particular perceptual beliefs, at least generally, and in their theory
of knowledge—for it all rests in large measure on their reliable perception. By con-
trast, the gazers are wrong and inapt both in their particular gaze-derived beliefs and
in their theory of knowledge—for it all rests on their unreliable gazing. Moreover, I
see no reason why the perceivers must be restricted to affirming only the conditional
that if perception is reliable, then they know. T see no reason why they cannot also
 Reflective Knowledge in the Best Circles 201

affirm the antecedent, why they cannot believe, both rationally and aptly, that per-
ception is reliable and does enable them to know.

Circles Beyond Belief

Why require the appeal to comprehensive enough coherence for justification, an

appeal that I have attributed, tentatively, to Descartes, as part of what justifies his
recourse to theology in accounting for true knowledge (scientia)! Why not say that
what justifies is that one's beliefs be caused by the gods? And if the question arises,
why not add that this belief itself is justified because it is itself caused by the gods?
We could, of course, proceed in this simplified way without worrying about coher-
ence or about the source of these beliefs beyond attributing them to divine agency.
But that is not the way we are built, most of us: We just do not acquire such beliefs
the way we do acquire beliefs willy-nilly when we open our eyes in good light. But
what if we were built that way? Would we then be justified in having such beliefs,
and in explaining our justification for having them, by their origin in divine agency?
Would we then be justified to the degree and in the way in which Descartes is justi-
fied or in the way in which our imagined Moore would be justified through his ap-
peal to a more ordinary reliabilism than that of Descartes? Internally regarded, the
structure of beliefs would share prominent features in all three cases. Of course, from
our Moorean, commonsense position, we can object both to Cartesianism and to the
invocation of the gods. These views are internally coherent, but we might still reject
them as wrong. And we might be able to explain what is wrong with them, from our
point of view, especially if our point of view rules out their leading ideas. But they
can, for their part, return the favor. Besides, we can anyhow imagine someone bril-
liant but insane, who weaves a system of immense interlocking complexity, but one
wholly detached from reality as we know it commonsensically. Such a madman could
object to our commonsense beliefs in a way that would seem relevantly analogous to
the way in which we would object to his mad beliefs.
What all of that shows, it seems to me, is nothing more than that knowledge does
not live by coherence and truth alone. Knowledge requires truth and coherence, true
enough, but it often requires more: for example, that one be adequately related, caus-
ally or counterfactually, to the objects of one's knowledge, which is not necessarily
ensured by the mere truth-cwm-coherence of one's beliefs, no matter how compre-
hensive the coherence. Madmen can be richly, brilliantly coherent; not just imagi-
nary madmen, but real ones, some of them locked up in asylums. Knowledge requires
not only internal justification or coherence or rationality, but also external warrant
or aptness. We must be both in good internal order and in appropriate relation to the
external world.27

Notes
1. A simplified version of Moore's Proof.
2. G. E. Moore, "Proof of an External World," Philosophical Papers (New York: Collier
Books, 1962), 144-45. Originally in the Proceedings of the British Academy 25 (1939).
202 Ernest Sosa

3. Ibid., p. 147.
4. Henceforth "knowledge" here will be short for "plain knowledge," leaving aside Car-
tesian superknowledge.
5. Moore, "Certainty," Philosophical Papers.
6. Compare Barry Stroud, The Significance of Philosophical Scepticism (Oxford: Oxford
University Press, 1984), especially chap. 1 and 3.
7. Moore, "Certainty," Philosophical Papers, 193-223.
8. For Russell, the "common-sense hypothesis" of independent physical objects is "sim-
pler" than the supposition that life is but a dream (as he explains in chap. 2 of The Problems
of Philosophy, Oxford: Oxford University Press, 1912). For Quine, the "hypothesis of ordi-
nary physical objects" is "posited" or "projected" from the data provided by sensory stimu-
lations. "Subtracting his cues from his world view, we get man's net contribution as the dif-
ference" (Word and Object, (Cambridge: MIT Press, 1960), 5.) That Quine's position is deeply
problematic is shown by Stroud (The Significance of Philosophical Scepticism, chap. 6).
9. This is argued by Wilfrid Sellars in "Phenomenalism," Science, Perception, and Real-
ity (London: Routledge and Kegan Paul, 1963).
10. The shift to discussion of Descartes may seem abrupt; however, what we find about
the nature of immediate knowledge in that discussion has important implications for a posi-
tion that Moore failed to explore. Skeptics who are willing to grant Descartes his immediate
knowledge through introspection or rational intuition would need to explain exactly why
perception could never yield such knowledge. (And what of memory?) The discussion of
Descartes to follow is meant to highlight this issue.
11. Descartes, The Philosophical Writings of Descartes, J. Cottingham, R. Stoothoff, and
D. Murdoch, eds. (Cambridge: Cambridge University Press, 1985), vol. 2, 24.
12. In his Principles of Philosophy (part 4, art. 205) for example, he notes that if we can
interpret a long stretch of otherwise undecipherable writing by supposing that it is written in
"one-off natural language," where the alphabet has all been switched forward by one letter,
and so on, then this is good reason for that interpretation. There he also argues for his scien-
tific account of reality in terms of certain principles by claiming that ". . . it would hardly
have been possible for so many items to fall into a coherent pattern if the original principles
had been false" (Descartes, Philosophical Writings).
13. My attribution to Descartes is tentative because of the enormous bibliography on the
"Cartesian Circle." In deference to that important tradition of scholarship, I do no more than
suggest that there is logical space for an interpretation of Descartes that is perhaps more com-
plex than many already tried, but that seems coherent and interesting. (I am myself convinced
that this is Descartes's actual position, and I defend this more fully in "How to Resolve the
Pyrrhonian Problematic: A Lesson from Descartes," in Philosophical Studies (forthcoming).
In "Mythology of the Given," in The History of Philosophy Quarterly (forthcoming), I argue
for the relevance of this Cartesian strategy to issues of empirical foundations that have di-
vided philosophers since the Vienna Circle, and have pitted, for example, Sellars, Rorty, and
Davidson on one side, against C. I. Lewis, Hempel, and Chisholm on the other. My warm
thanks to Lex Newman and James Van Cleve for helpful discussion of, and further references
relevant to, this way of viewing Descartes.)
14. Here one would reduce Descartes's requirement of infallible certainty.
15. For an early statement of this sort of problem, urged against Nozickian tracking, see
Colin McGinn, "The Concept of Knowledge," Midwest Studies in Philosophy 9 (1984): 529-
54.
16. Aristotle, Nichomachean Ethics, trans. T. Irwin, 1139a27-30. (Indianapolis: Hackett,
1985).
17. Ibid., 1139bll-13.
18. The social utility of concepts is invoked occasionally to defend a proposed account
of knowledge in my Knowledge in Perspective (Cambridge: Cambridge University Press,
1991); for example, 27 and 275. The form of argument involved is used insightfully in E. J.
Craig's Knowledge and the State of Nature (Oxford: Clarendon, 1990), where it takes center
stage.
 Reflective Knowledge in the Best Circles 203

19. This passage is from the Second Set of Replies as it appears in Descartes, The Philo-
sophical Writings of Descartes, vol. 2, 101. However, I must add that where this translation
says that an atheist can be "clearly aware," Descartes's Latin is "clare cognoscere."
20. Although unremarked by Descartes, the role of dreams in his perception skepticism
is analogous to a role assignable to paradoxes and aporias in a parallel skepticism vis-a-vis
rational intuition.
21. See my "Philosophical Skepticism and Epistemic Circularity," Proceedings of the
Aristotelian Society Supplementary Volume 64 (1994): 263-90; and compare Barry Stroud's
response, "Scepticism, 'Externalism,' and the Goal of Epistemology," ibid.
22. Chapter 1 of Stroud's The Significance of Philosophical Scepticism is an illuminat-
ing discussion of this principle and its importance for understanding philosophical skepti-
cism. In "How to Resolve the Pyrrhonian Problematic," I suggest how to derive it from other
principles with independent plausibility.
23. For one thing, as it stands it leads, apparently, to a vicious regress. But that is an illu-
sion. After all, what PE requires one to rule out is, not every possibility incompatible with
one's knowing that;?, but rather every possibility known to be thus incompatible. Since, for
one thing, knowledge requires belief, the regress is hence not infinite, nor does it seem vicious.
24. This distinction figures in my Knowledge in Perspective (Cambridge: Cambridge
University Press, 1991); see, for example, 240 and 282.
25. See "Understanding Human Knowledge in General,"in Knowledge and Scepticism,
M. Clay and K. Lehrer, eds. (Boulder, Colo.: Westview, 1989), 47.
26. B. Stroud, "Scepticism, 'Externalism', and the Goal of Epistemology," 303-4.
27. I am pleased to acknowledge helpful discussion, first here at Brown, in a faculty dis-
cussion group and in my graduate seminar; at the University of California, Santa Barbara; at
the Oxford Philosophical Society; at Cambridge University; and at the University of Granada.
 12

Commonsensism in Ethics
and Epistemology

NOAH LEMOS

Some philosophers think it appropriate to appeal to commonsense beliefs in the course

of philosophical reflection. To others, this appeal is anathema and unphilosophical.
In this essay, I want to describe what I take to be certain features of the "commonsense"
tradition in philosophy, especially in connection with epistemology and ethics. I shall
also consider some main objections to the use of commonsense epistemic and ethi-
cal beliefs. Some of these objections presuppose a certain view about what is required
for justified belief in, or knowledge of, particular epistemic or ethical propositions
or claims. Other objections raise different concerns. I shall argue that none of these
objections provide compelling reasons for rejecting the commonsense tradition in
either epistemology or ethics.

I want to describe briefly what I take to be certain features of the commonsense tra-
dition in philosophy. First, in answering certain philosophical questions, common-
sensism holds that it is appropriate or reasonable to take as data certain ordinary, yet
widely and deeply held, beliefs. These beliefs may have as their objects propositions
that are widely accepted, such as that there are other people, or they might involve
attributing to oneself properties that almost everyone attributes to himself, such as
having hands and having a body. Among the beliefs that commonsensism takes as
data are beliefs about the world around us, for example, that there are other people
who think and feel, that the earth has existed for many years, and that I have hands
and a body.

204
 Commonsensism in Ethics and Epistemology 205

But it is also worth noting that philosophers in the commonsense tradition take as
data various epistemic beliefs, that is, beliefs about what is known or reasonable or
justified. For example, in his essay "A Defence of Common Sense," G. E. Moore
begins by noting various nonepistemic propositions that, in his opinion, he knows.
These include that there exists a living human body that is his body, that it was born
in the past and has existed continuously ever since. But Moore also claims to know
that many other people know similar things about themselves.1 Moore thus claims to
know something about what other people know. Similarly, Chisholm, in the second
edition of Theory of Knowledge, tells us that we may take as a working hypothesis
that we do know most, if not all, of those things that we ordinarily think we know.2
Presumably, this would include various epistemic facts such as that most people know
their own names and that most people know they have hands.
A commonsense approach to moral philosophy would also take as data various
deeply and widely held ethical or moral beliefs. These beliefs would include moral
claims or propositions that almost everyone accepts or moral properties or features
that almost everyone attributes to himself. Among these might be the beliefs that we
ought to keep our promises, that gratitude is better than ingratitude, that I ought not
to be tortured to death, that I ought not to torture to death my parents.
Second, the commonsense philosopher takes these beliefs as data without having
any proof for them. He holds that it is appropriate to take these beliefs as data even
though he has not proved or given any argument for them. Whatever feature makes
it appropriate or reasonable for such beliefs to serve as data does not depend on one's
having proved or argued that these beliefs are true. Furthermore, he might take these
beliefs as data without knowing how he knows them and without knowing what is
his evidence for these beliefs. Again, Moore writes, concerning his commonsense
beliefs,

I do not know them directly, that is to say, I only know them because, in the past, I
have known to be true other propositions which were evidence for them . . . And yet I
certainly do not know what the evidence was. Yet this certainly seems to me to be no
good reason for doubting that I do know it. We are all, I think, in this strange position
that we do know many things, with regard to which we know further that we must have
had evidence for them, and yet we do not know how we know them, i.e., we do not
know what the evidence was.3

Third, the commonsense tradition is not committed to the view that our common-
sense beliefs are indefeasible or immune to revision, but it does assign to our com-
monsense beliefs a great deal of "weight" or importance. I know of no clear or simple
way of explaining how much weight the philosophers in this tradition assign to these
beliefs. Perhaps we could say the commonsense philosopher finds a heavy presump-
tion in favor of those commonsense beliefs he takes as data and holds that the grounds
for abandoning them must be clear and compelling. Of course, the commonsense
philosopher accepts this much. But it isn't clear that accepting this claim would dis-
tinguish him from the follower of Zeno who abandons many commonsense beliefs
on hearing the master's arguments. It is not a sufficient condition for being a "com-
monsensist" that one take one's commonsense beliefs as data or assume as a provi-
sional starting point that we know pretty much what we think we know. Perhaps we
206 Noah Lemos

may say simply that when our commonsense beliefs conflict with one another or with
a philosophical theory, the commonsense philosopher seeks to resolve the conflict
in the way that does the least violence to the body of his commonsense beliefs. Thus,
the commonsense philosopher seeks to be conservative in his revisions of his com-
monsense beliefs. Confronted with Zeno' s arguments, he might, therefore, find it more
reasonable to continue to hold his vast body of commonsense beliefs and believe
that Zeno's arguments are unsound than to accept the soundness of Zeno's arguments
and radically change his commonsense beliefs.
In describing the commonsense tradition, it is tempting to think of it as a kind of
"method" whose followers begin with some criterion for identifying commonsense
beliefs and then reject various principles inconsistent with the beliefs picked out by
the criterion. Undoubtedly certain ways of describing a commonsense approach fos-
ter this view. Thus, if we say that a commonsense approach holds that our epistemic
principles must be compatible with our commonsense beliefs, then one might natu-
rally expect that a proponent of such an approach had some criterion of identifying
commonsense beliefs. Such a temptation, though natural, should be resisted. Some
philosophers who take a commonsense approach, such as Chisholm and Moore, do
not begin or finish with a criterion or definition of a commonsense belief. Of course,
they reject various views because they are not compatible with beliefs that we may
call commonsense beliefs because they are ordinary and commonplace and deeply
and widely held. But clearly Chisholm and Moore do not begin with a criterion for
something's being a commonsense belief, and it is not essential for their approach
that they do so.
A commonsense approach to the theory of knowledge or right action has not been
universally accepted. Indeed, there are illustrious philosophers who have apparently
rejected the view that we may take various commonsense beliefs as data for the theory
of evidence or right action. Descartes seems to have rejected such an approach in the
Meditations, preferring to take as data only that which is certain and indubitable. Kant,
for example, scornfully rejects appeals to commonsense in the introduction to his
Prolegomena to any Future Metaphysics:
It is indeed a great gift of God to possess right or (as they now call it) plain common
sense. But this common sense must be shown in action by well-considered and reason-
able thoughts and words, not by appealing to it as an oracle when no rational justifica-
tion for one's position can be advanced. To appeal to common sense when insight and
science fail, and no sooner—this is one of the subtile discoveries of modern times by
means of which the most superficial ranter can safely enter the lists with the most thor-
ough thinker and hold his own. But as long as one particle of insight remains, no one
would think of having recourse to this subterfuge. Seen clearly, it is but an appeal to
the opinion of the multitude, of whose applause the philosopher is ashamed, while the
popular charlatan glories and boasts in it.4
More recently, R. M. Hare has written:
The appeal to moral intuitions will never do as a basis for a moral system. It is cer-
tainly possible, as some thinkers even of our own times have done, to collect all the
moral opinions of which they and their contemporaries feel most sure, find some rela-
tively simple method or apparatus which can be represented, with a bit of give and
take, and making plausible assumptions about the circumstances of life, as generating
 Commonsensism in Ethics and Epistemology 207

these opinions; and then pronounce that that is the moral system which having reflected,
we must acknowledge to be the correct one. But they have absolutely no authority for
this claim beyond the original convictions, for which no ground or argument was given.
The "equilibrium" that was reached is one between forces which might have been gen-
erated by prejudice, and no amount of reflection can make that a solid basis for moral-
ity. It would be possible for two mutually inconsistent systems to be defended in this
way; all that this would show is that their advocates had grown up in different moral
environments.5

Hare rejects the appeal to moral intuitions and proposes to argue for various moral
principles on the basis of logical principles, including principles about the logic of
moral terms and various empirical, nonmoral premises. He proposes to justify vari-
ous moral principles by appealing to nonmoral premises.
Suppose, then, a defender of a commonsense approach to the theory of evidence
or right action were to ask himself why should he take various commonsense beliefs
as data? What might he say?
In the second edition of Theory of Knowledge, Chisholm says that in investigat-
ing the theory of evidence from a philosophical or "Socratic" point of view, we make
three general presuppositions. The first of these is "that there is something that we
know and we adopt the working hypothesis that what we know is pretty much that
which, on reflection, we think we know. This may seem the wrong place to start. But
where else could we start?"6 Thus, one might ask, as Chisholm asks, where else could
we begin? One might hold that there is simply no other way to proceed except by
taking these "commonsense" beliefs as data. But surely there are alternatives such as
those adopted by Descartes and Hare that do not take such beliefs as data.7
Alternatively, one might hold that it is rational for us to take our commonsense
beliefs in an external world and other people as data because we simply cannot give
them up, or at least we cannot give them up for long. Such a line of argument is sug-
gested in the following passage from Thomas Reid's Inquiry:
The belief of a material world is older, and of more authority, than any principles of
philosophy. It declines the tribunal of reason, and laughs at all the artillery of the logi-
cian. It retains it sovereign authority in spite of all the edicts of philosophy, and reason
itself must stoop to its orders. Even those philosophers who have disowned the author-
ity of our notions of an external material world, confess that they find themselves under
a necessity of submitting to their power.8

Perhaps the idea is that if our principles of evidence or right are to be part of a coher-
ent body of beliefs, then, given our incapacity to give up our commonsense beliefs,
our principles must be compatible with our commonsense beliefs. The demand for
coherence along with our inescapable commonsense doxastic commitments requires
us to accept only those principles that cohere with those commitments. But this line
of argument is not persuasive. First, suppose that one has some belief or set of be-
liefs that one cannot give up, could it not be rational for one to hold that the belief is
false or unworthy of one's rational acceptance even while admitting that one cannot,
like a bad habit, get rid of it? Could not some beliefs be like bad habits or prejudiced
preferences that one rationally regards as unworthy of having even while one has
them? Even if I can't get rid of my belief that p, could I not rationally hold that p is
208 Noah Lemos

unworthy of my acceptance? And why should I accept principles of evidence or right

action because they cohere with beliefs that I rationally regard as unworthy of my
acceptance? Second, perhaps there are some commonsense beliefs that we cannot
give up, but it is not clear that this is true of everything the commonsense philoso-
pher would take as data. Thus, even if one cannot give up some beliefs, such as that
there is an external world or there are other people, is it really impossible to give up
the epistemic belief that one knows or is justified in believing these things? I do not
think the commonsense philosopher needs to claim that all those beliefs he takes as
data have this character of unrevisability.
What, then, makes it reasonable for the proponent of a commonsense approach to
take as data the things that he does? The best answer, I suggest, is that the common-
sense beliefs that he takes as data are either instances of knowledge or justified be-
lief. It is reasonable for him to take them as data because they have some positive
epistemic status for him, because he either knows them or is justified in believing
them. Indeed, I think we may take Reid to be making this point in the passage quoted
above. Though Reid says the belief in a material world is one we cannot give up and
older than conflicting principles, he also points out that the belief has "authority."
The belief has a certain normative status; it is worthy of our acceptance and more
worthy of our acceptance than any philosophical principles that imply there is no
material world. We should not confuse the power of a belief with its authority, and it
is the latter that makes it reasonable for us to use them as data.
The defender of a commonsense approach to the theory of evidence may thus hold
that it is reasonable for him to take various commonsense beliefs, beliefs he ordi-
narily thinks he knows, as data because those beliefs are justified for him or instances
of knowledge. Thus, he may hold that he is justified in believing various epistemic
propositions such as "I know I have two hands" or "It is beyond reasonable doubt
that I was alive five minutes ago." Similarly, a defender of a commonsense approach
to the theory of right action may hold that it is reasonable for him to take various
commonsense beliefs as data because those beliefs are justified for him or instances
of knowledge. Thus, he might hold that he is justified in believing various moral
propositions such as "I ought to keep my promises" or "It would be wrong to torture
my parents to death." In each case, he may take various evaluative propositions
(whether epistemically or morally evaluative) as data because those propositions have
a positive epistemic status for him, because they are instances of knowledge, or he is
justified in believing them.

II

The view that it is reasonable to take various commonsense beliefs as data because
they have a positive epistemic status faces a variety of objections. Some of these
objections stem from a certain conception of what is required for a belief to be jus-
tified. Let us consider the following argument:

1. In order to use one's particular epistemic or moral belief as data one must be justi-
fied in that belief.
 Commonsensism in Ethics and Epistemology 209

2. In order to be justified in believing any proposition, p, one must show that one's
belief that p is true or one must at least justifiably believe true premises from which
p follows.
3. Therefore, in order to use one's particular epistemic or moral belief that p as a
datum one must show thatp is true or at least justifiably believe true premises from
which p follows.
Perhaps it is the conception of justification embodied in (2) that leads Kant and Hare
to complain about appeals to commonsense or intuition "when no rational justifica-
tion for one's position can be advanced" or "for which no ground or argument was
given." Given the conception of justification embodied in (2) and the plausible view
that the beliefs one takes as data must themselves be justified, it would be quite un-
satisfactory for the commonsense philosopher to appeal to particular epistemic or
moral beliefs without showing that they are true or that they are justified in believing
the premises from which they followed.
But surely (2) is false. It leads to an implausible infinite regress. To be justified in
believing any proposition, one would need to justifiably believe premises from which
it followed and one would need, in turn, to justifiably believe premises from which
those premises followed and so on. As a general conception of justification, (2) is
false if we are justified in believing anything at all.
Moreover, why should we think "rational justification" and "argument" are nec-
essary for our commonsense epistemic and moral beliefs to have such positive sta-
tus? There are approaches to knowledge and justification that do not take such things
to be necessary. Among these are various forms of foundationalism, coherentism,
and externalism. For the foundationalist, whether some beliefs are justified or knowl-
edge depends on their being takings of what is given in experience. For the coherentist,
whether a belief is known or justified depends on its belonging to a (sufficiently and
appropriately comprehensive) coherent body of beliefs. For the externalist, whether
a belief is known or justified depends on one's belief being formed by a reliable pro-
cess or one's "tracking the truth" of the proposition that is the content of the belief or
one's belief's being the product of a truth-conducive intellectual virtue. Neither
foundationalism, coherentism, nor externalism require that in order for a belief to be
justified or known, one must give or have a rational argument for the belief in question.
Consider again the commonsense epistemic beliefs that Moore takes himself to
know. Could not these beliefs have the status of knowledge for Moore at least in part
because they are things whose truth he tracks or because his beliefs issue from some
truth-conducively reliable process or intellectual virtue of belief acquisition? Fur-
thermore, given his view that the sorts of things he takes himself to know are also
known by almost everyone, his beliefs about what he knows are coherent with his
views about the cognitive powers of ordinary people. Moore, at least, is not in the
position of someone who has some reliable cognitive faculty and yet believes or has
good reason to believe that no one has that faculty. Moore's commonsense beliefs
on this view would have the status of knowledge in part because of his reliability
about the beliefs in question and the fact that those beliefs fit coherently into his pic-
ture about the cognitive powers of ordinary people.
Still, one might think that while (2) as a general thesis about justification is false,
it is more plausible when restricted to a particular domain of beliefs, viz. those par-
210 Noah Lemos

ticular epistemic and moral beliefs that commonsense philosophers often take as data.
Thus one might accept instead of (2),

(2') In order to be justified in believing any particular epistemic or moral propo-

sition,/?, one must show that one's belief that/? is true or one must at least jus-
tifiably believe premises from which p follows.

But why should this restricted claim be true? That some philosophers appear to hold
such a view seems clear. Consider the following passage from Hare's The Language
of Morals'.
[I]f we knew all of the descriptive properties which a particular strawberry had . . . and
if we knew also the meaning of the word 'good', then what else should we require to
know, in order to be able to tell whether a strawberry was a good one? Once the ques-
tion is put in this way, the answer should be apparent. We should require to know,
what are the criteria in virtue of which a strawberry is to be called a good one, or what
are the characteristics that make a strawberry a good one, or what is the standard of
goodness for strawberries. We should require to be given the major premiss.9
In Hare's view, knowing that a particular strawberry is good requires or depends on
knowing a criterion or standard for the goodness of strawberries. Though Hare re-
fers to our knowledge of good strawberries, we may take him to be making a general
claim about our knowledge of the goodness of other sorts of particulars. William
Alston suggests a similar view about what is required for us to be justified in believ-
ing that a particular belief is justified. He writes,
In taking a belief to be justified, we are evaluating it in a certain way. And, like any
evaluative property, epistemic justification is an evaluative property, the application
of which is based on more fundamental properties . . . Hence, in order for me to be
justified in believing that S's belief that/? is justified, I must be justified in certain other
beliefs, viz., that S's belief that/? possesses a certain property Q, and that Q renders its
possessor justified. (Another way of formulating this last belief is: a belief that there is
a valid epistemic principle to the effect that any belief that is Q is justified.)10
Alston's remarks suggest that it is the supervenient character of evaluative properties
that makes justified attributions of them dependent on our being justified in believing
other propositions and general principles. Thus, even if one rejects (2), one might ac-
cept (2') because of the supervenient character of epistemic and moral properties.
Reflection on the passages by Hare and Alston suggests the following general
principle:

(P) One is justified in believing that x has F (where F is some evaluative prop-
erty) only if one is justified in believing both (a) x has some nonevaluative
property Q, and (b) whatever has Q has F.

Given P, justified particular attributions of an evaluative property must meet two

conditions. First, one must be justified in believing a certain kind of general evalua-
tive principle and if Q is the property on which the evaluative feature supervenes,
one must be justified in believing that the thing in question has Q. Thus, given this
 Commonsensism in Ethics and Epistemology 211

view, particular attributions of justification to beliefs or Tightness to actions is de-

pendent on one's being justified in believing certain kinds of general epistemic or
moral principles, and one must be justified in believing that the particular belief or
action has the properties on which its Tightness or justification supervenes.
If P is true, then anyone who seeks to discover general evaluative criteria by tak-
ing his particular evaluative judgments as (revisable) data faces a dilemma. Consider
the following argument about the appeal to particular moral beliefs in order to dis-
cover criteria of rightness. Suppose that S is justified in believing that x is right only
if for some Q, S is justified in believing both (a) x has Q, and (b) whatever has Q is
right. Either S is not justified in believing (b) or S is justified in believing (b). If S is
not justified in believing (b), then S's particular belief that x is right is not justified
and cannot serve as a (good) reason for accepting or rejecting any criterion of right-
ness. On the other hand, if S is justified in believing (b), then one is already justified
in believing the criteria, and there is no point in using the belief that x is right in order
to discover the criteria. Therefore, either the particular belief that x is right cannot
serve as a (good) reason for accepting a criterion of rightness, or there is no point in
using the belief that x is right in order to discover the criteria. Clearly, a similar argu-
ment could be advanced concerning the appeal to particular epistemic beliefs in order
to discover epistemic criteria.
But is P true? Does it provide us with a good reason for accepting (2')? I think the
answer to both is "no." First, we should not confuse claims about exemplification with
claims about application. In assuming that evaluative properties are supervenient, we
assume that they are exemplified in virtue of the exemplification of nonevaluative prop-
erties. But this is very different from Alston's claim that the application of an evalua-
tive property is based on more fundamental properties. That things have evaluative
properties in virtue of their having nonevaluative properties does not imply that the
application or attribution of evaluative properties must be based on more fundamental
properties or on one's accepting a general criterion. Thus, P does not seem to follow
from the mere fact that evaluative properties are supervenient.
Second, it seems plausible that there are justified attributions of supervenient prop-
erties that do not depend on justified belief in general principles. Many philosophers
hold, for example, that mental properties supervene on more basic nonmental prop-
erties. Thus, the mental properties of being in pain or being conscious supervene on
more basic nonmental properties. But do our attributions to ourselves of such prop-
erties depend on our being justified in believing other propositions such as (1) I have
F, and (2) whatever has F is in pain? Surely, I can know that I am in pain or con-
scious without knowing on what more basic nonmental properties pain or conscious-
ness supervenes or knowing that I have those properties.
Third, it seems plausible that there are justified attributions of supervenient evalua-
tive properties that do not depend on justified belief in general principles. Consider,
for example, someone driving in the mountains coming suddenly upon a spectacular
mountain vista. He thinks, "That's beautiful!" It is plausible, I believe, that one can
be justified in such a belief where that belief does not depend for its justification on
some general principle about beautiful-making properties. We often form justified
beliefs about the beauty of things when we are struck by their beauty. In such cases,
our beliefs seem based on our experience of the thing and not on our beliefs that it
212 Noah Lemos

has some property, F, and that whatever has F is beautiful. In such a case, it may be
that our nondoxastic experience, our seeing or hearing the thing's beauty-making
properties, plays a role in our justifying our belief that P mistakenly assigns to be-
liefs, beliefs I think we often simply do not have or form.
I conclude that accepting the supervenient character of justification, knowledge,
or moral rightness does not commit us to accepting P or (21). It is not obvious that the
justified attribution of these properties to particular beliefs or actions depends on our
being justified in believing some general criterion or principle of the sort Hare or
Alston require.
I shall conclude this section by making two additional points. First, if we reject
principle P, we need not claim that our knowledge that a particular has an evaluative
property is direct or basic in the sense that it does not epistemically depend on our
knowing other things. Moore, as we have seen, denied that his commonsense knowl-
edge was direct. If we reject P, we need not deny, for example, that, in at least some
cases, knowing or justifiably believing that some thing, X, has a particular evalua-
tive feature, F, depends on our knowing or being justified in believing that X has
some feature Q. Thus, we may hold that my knowing that some particular act is wrong
depends on my knowing that it is an instance of ingratitude and promise-breaking.
Similarly, my knowing that a particular belief is justified may depend on my know-
ing that it is visual belief about a medium-sized physical object held at arm's length
in daylight.
But, second, we need not assume that this further knowledge is knowledge of those
properties on which the evaluative feature supervenes. In other words, we need not
assume that if my knowing or justifiably believing that X has some evaluative fea-
ture, F, epistemically depends on my knowing or justifiably believing that X has Q,
then X's having F supervenes on X's having Q or that it is in virtue of having Q that
Xhas F. Consider the connection between someone's facial expression and his men-
tal states. It may be that my knowing that Sheila is sad epistemically depends on my
knowing something about her facial expression, but we need not assume that her being
sad supervenes on her having that facial expression or that she is sad in virtue of
having that facial expression. The relation between her facial expression and mental
state is epistemic or evidential rather than one of supervenience or ontological
dependence. Similarly, we may hold that X's being an instance of ingratitude and
promise-breaking is epistemically or evidentially related to X's being wrong with-
out holding that X is wrong in virtue of its being an instance of ingratitude or promise-
breaking. If this is right, then the utilitarian may concede that one's justification for
believing that some act is wrong epistemically depends on one's justifiably believ-
ing it is an instance of ingratitude and promise-breaking while denying that the wrong-
ness of the act supervenes on these features. Similarly, the coherentist may hold that
his belief that B is justified depends on his belief that B is the product of a reliable
faculty of perception while denying that justification supervenes on a belief's being
the product of a reliable faculty of perception. Thus, even if a moral philosopher or
epistemologist can point to various features of an act or belief as reasons for think-
ing that the act has some evaluative character, he need not hold that it is those fea-
tures in virtue of which the act has that evaluative character.
 Commonsensism in Ethics and Epistemology 213

III

Let us return to Kant's and Hare's complaint about appeals to commonsense beliefs
or moral intuitions "when no rational justification for one's position can be advanced"
or "for which no ground or argument was given." Let us consider two prominent
objections to the use of moral "intuitions." The first objection is raised by Richard
Brandt:
Various facts about the genesis of our moral beliefs militate against the mere appeal to
intuitions in ethics. Our normative beliefs are strongly affected by the particular cul-
tural tradition which nurtured us, and would be different if we had been in a learning
situation with different parents, teachers, or peers. Moreover, the moral convictions of
some people derive to use words from Peter Singer, "from discarded religious systems,
from warped views of sex and bodily functions, or from customs necessary for the
survival of the group in social and economic circumstances that now lie in the distant
past." What we should aim to do is step outside our own tradition somehow, see it from
the outside, and evaluate it separating what is only a vestige of a possibly once useful
moral tradition from what is justifiable as present. The method of intuitions in prin-
ciple prohibits our doing this. It is only an internal test of coherence, what may be no
more than a reshuffling of moral prejudices.11
The second objection is raised by Hare:
The intuitions that give rise to the conflict are the product of our upbringings and past
experiences of decision-making. They are not self-justifying; we can always ask whether
the upbringing was the best we could have, or whether the past decisions were the right
ones, or even if so, whether the principles then formed should be applied to a new situa-
tion, or, if they cannot all be applied, which should be applied. To use intuition itself to
answer such questions is a viciously circular procedure; if the dispositions formed by
our upbringing are called into question, we cannot appeal to them to settle the question.12
There are three points that should be made about these passages. First, both Brandt
and Hare object to the method of wide reflective equilibrium employed by Rawls
and others to discover or defend moral criteria. The method of wide reflective equi-
librium takes as (revisable) data one's considered moral judgments about particular
cases and general principles. One thing that distinguishes the method of "wide" from
"narrow" reflective equilibrium in moral philosophy is that the former seeks coher-
ence not simply between one's moral beliefs, but a broader coherence that includes
one's beliefs about other areas that seem relevant such as beliefs about psychology
and economics. Second, Brandt rejects such a method since it "may be no more than
a reshuffling of moral prejudices" and Hare complains, in a passage quoted earlier,
that the equilibrium reached "is one between forces which might have been gener-
ated by prejudice." Third, for both Brandt and Hare, the remedy or alternative is to
"get outside" our own moral beliefs by appealing to nonmoral beliefs to defend a
moral criterion. In their view, we must seek an independent argument for a moral
criterion, independent in the sense that it does not presuppose the truth of any of our
moral beliefs. Thus, Brandt appeals to beliefs about "rational" action and desire and
Hare to beliefs about various logical principles, including principles about the logic
of moral terms, and various empirical, nonmoral premises.
214 Noah Lemos

Let us grant that Brandt and Hare are right that even the most prejudiced and vi-
cious moral beliefs might be woven into a coherent web. Of course, that is not a unique
feature of our moral beliefs. Wildly defective perceptual beliefs and fanatical epistemic
beliefs can with enough ingenuity or stubbornness be spun into a coherent fabric.
Let us grant that mere coherence alone does not distinguish reasonable or justified
moral beliefs from mere vicious prejudice.
But such a point can be made by either the foundationalist or the externalist since
neither sees mere coherence as sufficient for justification or takes accidentally true
beliefs to be knowledge simply because they cohere with the rest of a subject's be-
liefs. Neither is committed to the view that all coherent bodies of beliefs are episte-
mically on a par. The foundationalist may hold that the justification of one's moral
beliefs also depends on the character of one's experience including, perhaps, one's
moral and emotional experiences. The externalist may hold that the justification of
one's moral beliefs also depends on whether those moral beliefs have been produced
by a sufficiently truth-conducive intellectual virtue.
Suppose we grant that mere coherence alone does not distinguish between rea-
sonable or justified belief and mere vicious prejudice; then why would such a con-
cession imply that we must "get outside" our moral beliefs, that we cannot use our
moral beliefs as data to reach knowledge or justified belief about general criteria of
right action? Perhaps the answer may be put in terms of the following argument:
1. In order to distinguish between those coherent bodies of moral belief that are true
and those that are merely coherent yet false, we must distinguish between those moral
beliefs that are reliably formed and those that aren't.
2. In order to distinguish between reliable and unreliable ways of forming moral be-
liefs, we can either (a) use our moral beliefs as data or (b) "go outside" our moral
beliefs in order to discover a criterion of moral Tightness and then see which dispo-
sitions lead one to form moral beliefs that accord with that independently estab-
lished criterion.
3. (a) is viciously circular.
4. Therefore, we must follow (b).
If this is the line of reasoning for rejecting moral beliefs as data, it does not seem
compelling. First, consider premise (1). If "distinguishing" between true and false
bodies of beliefs is understood as knowing which are true and which are false, then
it is not obvious that either a foundationalist or an externalist would accept the first
premise. Each might agree that S's knowing that p requires that S's belief that/? must
be reliably formed, but why, in addition, must it also require that S know that his
belief that;? is reliably formed? Neither foundationalist nor externalist accounts of
knowledge are committed to such a requirement.
Second, even if we thought that S's knowing that/? required that S know that his
belief that p was reliably formed, how would going outside our moral beliefs put us
in a better position? Can we not also wonder about the reliability of the dispositions
to produce those nonmoral beliefs on which we would base our criterion? Consider
those nonmoral beliefs, for example, beliefs about what it is rational to do or beliefs
about the logic of moral terms and various empirical premises, to which we might
appeal to independently establish our moral criterion. Surely, if we are to know our
criterion of right action on the basis of these nonmoral beliefs, then they must also
 Commonsensism in Ethics and Epistemology 215

be known. But if knowing that/) requires that one know one's belief is reliably formed
and if this in turn requires that we "go outside" such beliefs for a noncircular argu-
ment, then it seems that any knowledge will be impossible. For S to know that p, S
must also know that his belief that p is reliably formed and to know that he must
know it on the basis of knowing q and r. But to know q, he must know that his belief
that q is reliably formed and to know that he must know it on the basis of s and t, and
so on.
But perhaps it will be argued that we do not need to establish the reliability of our
dispositions to form these nonmoral beliefs. One might claim that various facts about
the genesis of our moral beliefs require the need for external validation and similar
worries do not apply to these nonmoral beliefs. But what facts are these supposed to
be? Surely, our beliefs about what it is rational to do, the logic of moral terms, and
our empirical beliefs are influenced by our culture and upbringing, by our teachers
and peers. And though there is serious disagreement about some of our moral be-
liefs, is there really serious disagreement about them all? What about the belief that
we ought not to be tortured to death or that we ought not to torture to death our par-
ents? And would not the judgment about what is a "serious" disagreement requiring
a move to an independent justification be itself a normative matter over which people
might disagree? Surely, if it is reasonable to admit that some of our moral and episte-
mic beliefs might be mistaken, would it not be as reasonable to admit that some of
our beliefs about what rationality requires, or our empirical beliefs, or our beliefs in
putative necessary truths, including the logic of moral terms, might also be mistaken?13
In short, it is not clear that there is any reason to think that our commonsense moral
beliefs have some special character that requires an external validation that would
not equally apply to these nonmoral beliefs to which we might appeal.
Third, even if we reject the view that S's knowing that/? requires S's knowing
that his belief that p is reliably formed, we might still wonder how is S to know that
his belief is reliably formed. Is the only satisfactory alternative to "go outside",
appealing to other beliefs that do not presuppose the truth of one's moral beliefs?
This is a question that might interest us even if we reject the view that to know p, one
must know that one is reliable with respect to p. For suppose philosophical reflec-
tion convinces us that S knows p only if S's belief that p is reliably formed; can we
insist, as Moore does, that we now know various commonsense propositions, whether
ordinary commonplaces such as that there are hands or commonsense epistemic and
moral propositions, without knowing that our beliefs are reliably formed? In other
words, doesn't our acceptance of both (1) S knows p only if S's belief thatp is reli-
ably formed and (2) we know some commonsense propositions commit us to (3) our
beliefs in those propositions are reliably formed?
How then might we know that our belief in some commonsense moral proposi-
tion, say, that it would be wrong to torture our parents to death (p), is reliably formed?
Well, could we not infer it from (1) and the proposition that we do know/?? Of course,
such an inference would be rejected as "viciously circular" by Hare and Brandt. It
does not "go outside" our moral beliefs but rather presupposes the truth of p since
one could not know p unless p is true. Could not a moral madman argue analogously
that his belief that not-/) is reliably formed since it follows from (1) and his knowing
not-p? Could not the crystal ball gazer who, on peering into the crystal ball, forms
216 Noah Lemos

the belief that q argue similarly that he infers that his belief that q is reliably formed
from (1) and he knows that ql The conclusions of the moral madman or crystal ball
gazer might cohere as well with their beliefs as our belief about our reliability with
respect to p coheres with ours.
But even if this way of arguing for the reliability of our belief in p is circular, is it
viciously circular? Is the madman's or crystal ball gazer's belief in his reliability
epistemically on a par with our belief thatp is reliably formed? Ernest Sosa argues,
persuasively I think, that it is not: "What privileges our positions, if anything does,
cannot be that it is self-supportive, as we have seen; but nor can it possibly be that it
is ours. Our position would be privileged rather by deriving from cognitive virtues,
from the likes of perception and cogent thought, and not from derangement or super-
stition or their ilk."14 What would privilege our position would not be its mere
coherence but its deriving from truth-conducive cognitive virtues concerning what
we know and what is right. Thus, if Sosa is right, then premise (3) is false.
Finally, suppose we think that some externalist reliability condition must be met
for a belief to be knowledge. Or suppose we believe, I think less plausibly, that such
a condition is necessary for a belief to be justified. Some philosophers object to such
a condition because it would take the question of the status of our epistemic beliefs,
our beliefs to the effect that some belief is justified or known, out of the hands of
philosophers and put it into the hands of cognitive scientists. Consider the following
remark by Richard Fumerton:

If externalist metaepistemologies are correct, then normative epistemology is an inap-

propriate subject matter for philosophy. Philosophers as they are presently trained have
no special philosophical expertise enabling them to reach conclusions about which
beliefs are or are not justified. Since the classical issues of skepticism fall under nor-
mative epistemology, it follows that if externalism is correct, then philosophers should
stop addressing the questions raised by the skeptic. The complex causal conditions that
determine the presence or absence of justification for a belief are the subject matter of
empirical investigations that would take the philosopher out of the easy chair and into
the laboratory.15

Fumerton is concerned here with the notion of justification, but I think his remarks
apply as well to other epistemically evaluative notions, such as knowledge, for which
we might think an externalist reliability condition must be met. Fumerton writes as
though the question whether a belief is justified must be answered either by a phi-
losopher via his philosophical expertise or the expertise of the laboratory scientist,
and if some externalist reliability condition must be met then it must answered by
the latter. But why think that knowledge of epistemic facts is a matter of special
expertise at all? Ordinary people can know many things and know that they know
them. Who, but a philosopher, would have thought justified belief in epistemic facts
was the sole purview of the philosopher? Moreover, if justification is a property that
attaches to beliefs, including epistemic beliefs, in virtue of complex causal condi-
tions, why does it follow that the question of whether a belief has that property must
be answered in the laboratory? Headaches and toothaches are the product of com-
plex causal processes, but surely one need not go to a laboratory to learn whether
 Commonsensism in Ethics and Epistemology 217

one has them. It is far from clear that an externalist requirement on knowledge or
justification is committed to the view that knowledge of epistemic facts requires the
expertise of the laboratory scientist.

Notes
1. G. E. Moore, "A Defence of Common Sense," Philosophical Papers (London: George
Allen and Unwin, 1959), 34.
2. Roderick Chisholm, Theory of Knowledge, 2d ed. (Englewood Cliffs, N.J.: Prentice Hall,
1977), 16.
3. Moore, 44.
4. Immanuel Kant, Prolegomena to any Future Metaphysics, trans, with an introduction
by Lewis White Beck (Indianapolis, Ind.: Bobbs-Merrill, 1950), 7.
5. R. M. Hare, Moral Thinking (Oxford: Clarendon, 1981), 12.
6. Roderick Chisholm, Theory of Knowledge, 16.
7. I do not mean to suggest that this is Chisholm's only response.
8. Thomas Reid, Thomas Reid's Inquiry and Essays, Ronald E. Beanblossom and Keith
Lehrer, eds., with an introduction by Ronald E. Beanblossom (Indianapolis, Ind.: Hackett,
1983), 54. Reid goes on to say: "Methinks, therefore, it were better to make a virtue of neces-
sity; and, since we cannot get rid of the vulgar notion and belief of an external world, to rec-
oncile our reason to it as well as we can; for, if Reason should stomach and fret ever so much
at this yoke, she cannot throw it off; if she will not be the servant of Common Sense, she
must be her slave."
9. R. M. Hare, The Language of Morals (Oxford: Oxford University Press, 1952), 111.
10. William Alston, "Two Types of Foundationalism," The Journal of Philosophy 73
(1976), 170. Cf. also "Some Remarks on Chisholm's Epistemology," Nous 14 (1980), 579.
11. Richard Brandt, A Theory of the Good and the Right (Oxford: Clarendon, 1979), 21-22.
12. Hare, Moral Thinking, 40. A similar objection has been raised by Alan Gewirth. See
Gewirth's Reason and Morality (Chicago: The University of Chicago Press, 1978), 18-19.
Whatever the merits of such an inductive justification with regard to the principles of logical
inference, it suffers from a serious difficulty when applied to morality—a difficulty already
encountered above in my discussion of the possible self-evidence of a moral principle. The
inductive justification assumes that we can differentiate the morally right from the morally
wrong in person's particular moral judgments and can hence infer what is the morally right
general principle by generalizing from the morally right particular judgments. This would
work well enough if there is no serious dispute over the particular judgments that are taken as
the sources. But what if there is such a dispute? What if the judgments are challenged by a
Callicles, an Aristotle, or a Nietzsche? In this case, to appeal to the principles to settle the
dispute would beg the question, since the principles rest exclusively on the particular judg-
ments the opponents are disputing. A principle that shows that alternatives like those just
mentioned are morally wrong must do so not simply because it is itself a generalization from
the opposed alternatives—for this is question-begging—but because it has an independent
rational justification of its own.
13. Hare asks us to consider the following two propositions:
1. I now prefer with strength S that if I were in that situation x should happen rather
than not.
2. If I were in that situation, I would prefer with strength S that x should happen
rather than not.
Hare claims that it is a conceptual truth that to know (2), (1) must be true (Hare, Moral Think-
ing, 95-96). But is it really reasonable to think that I am less likely to be mistaken about this
"conceptual truth" than the moral claim that I ought not to torture my parents to death? Is it
really reasonable for me to believe that I am less reliable about that moral claim than about
218 Noah Lemos

this "conceptual truth"? For a criticism of Hare's view, see Tom Carson, "Hare's Defense of
Utilitarianism," Philosophical Studies 50 (1986), 97-115.
14. Ernest Sosa, "Philosophical Skepticism and Epistemic Circularity," Empirical Knowl-
edge, 2d ed., Paul K. Moser, ed. (Lanham, Md.: Rowman and Littlefield, 1996), 325.
15. Richard Fumerton, Metaepistemology and Skepticism (Lanham, Md.: Rowman and
Littlefield, 1995), 171.
Part VI

BEYOND DEONTOLOGY
This page intentionally left blank
 13

Knowing People

VRINDA DALMIYA

To get off "the wheel" and get on with the positive task of epistemological theo-
rizing, we must start with certain uncontroversial samples of knowing—or so goes
the particularist credo. These instances are then analyzed to yield universally ap-
plicable epistemic principles and become the lens through which all knowledge is
understood. Now, nothing could be less problematic than claiming to know (under
"normal" circumstances, of course) that there is a computer on my desk in front of
me. Equally unproblematic is my claim to know (given a "normal" mother-and-
eight-year-old relationship) that my daughter is deeply upset at not being invited
to a sleep-over party next door. Which of these "uncontroversial" knowledge claims
should we pick to serve as the starting point of epistemological theorizing? And
does it even matter?
Resonating with a largely positivistic legacy, conventional epistemological wis-
dom opts for knowledge of middle-sized physical objects as exemplars of knowing.
Lorraine Code1 speaks to the political bias underlying this traditional choice and
argues that if instances of knowing selves rather than knowing objects had been the
exemplars of knowing, then the contours of epistemological theory would have been
very different from what they are now. I try to make good this claim here. Code herself
develops the idea into a critique of the subjective-objective dichotomy used to theo-
rize knowledge and shows how infusion of "subjectivity" make questions of cogni-
tive agency and character central in an epistemology. Partially building on this, I
first spell out why and how knowing selves involves cognitive agency (in a way that
knowing things does not) and then argue that peculiarities of such agency recasts the
debate between internalism and externalism and ultimately reconfigures the notion
of epistemic responsibility. These changes suggested by taking the knowledge of

221
222 Vrinda Dalmiya

selves seriously converge, as we shall see, with a "responsibilist" version of virtue

epistemology.2
Now, this entire argument presupposes that knowing selves is significantly dif-
ferent from knowing middle-sized physical objects—and I turn to this assumption in
the next section.

I. Knowledge of Selves

It would be ridiculous to claim that knowing people has found no place in traditional
epistemology. The engagement with the problem of "other minds" is, if not anything
else, at least testimony to the fact that knowledge of subjectivity—in the first, sec-
ond, and third persons—cannot be and has not been ignored. That it is considered to
be a special "problem" is put down to broadly Cartesian metaphysical assumptions
of what the self is. However, whether Cartesian or not, selves are a very different
sort of entity than "things," and they remain epistemically intractable as long as their
knowledge has to be assimilated within the general parameters of knowing things.
Let us quickly rehearse the steps taken to affect this (mistaken) assimilation.

Analogical Inference
Just as we know physical objects through our external senses, we are said to know
ourselves through an "internal sense." I know that there is a computer on my desk by
looking outward, and similarly, I know that I am anxious about meeting the deadline
for this paper by "looking within" (introspection). Furthermore, anxieties and fears
or haste and sleepiness can lead me to mistake the computer on my desk for a visit-
ing Martian just as much as they can result in mistaken claims about myself. In ei-
ther case, concentration and a bracketing out of emotions and other "irrational" fac-
tors ("pure objectivity") is necessary to prevent errors. We have here a quasi-visual
model of self-knowledge. The step from knowing myself to knowing others (from
knowing that I am anxious to knowing that you are anxious) is more complex. Since
you and I are not identical, I obviously need to "look outside" for such knowledge.
But no matter how concentratedly I look, I can only come to know your outward or
bodily conditions. (Whether or not there is a separate "mind" behind them, what I
encounter through external perception is clearly not the whole of you). According to
conventional wisdom, knowledge of physical objects not present to or hidden from
view is derived through inference. Thus, some form of analogical inference becomes
the "solution" to the problem of other minds.

Simulation Theory
Now the difficulties of this approach are too well known even to its practitioners to
be rehearsed here. But what is the alternative? Can we explain knowledge of selves
in a way that is not modeled on knowledge of things? Simulation theory in the works
of Alvin Goldman,3 Robert Gordon4 and some developmental and cognitive psycholo-
gists has been used to articulate the ontology of mental states as well as to explain
 Knowing People 223

mastery of mental concepts. But concentrating on the epistemological dimension of

simulation suggests a unique manner in which selves are known.
Simulation opposes itself to the dominant picture according to which we deploy
a "theory"—laws relating mental states to one another and to external stimuli and
behavior (the knowledge of which can be "tacit") to ascribe mental states to others.
It resurrects instead an old idea of "empathy" or "sympathy," which Adam Smith
had explained in the following way:
As we have no immediate experience of what other men feel, we can form no idea of
the manner in which they are affected, but by conceiving what we ourselves should
feel in the like situation. Though our brother is upon the rack, as long as we ourselves
are at our ease, our senses . . . never did, and never can, carry us beyond our own per-
son, and it is by imagination only that we can form any conception of what are his
sensations. . . . By the imagination we place ourselves in his situation, we conceive
ourselves enduring all the same torments, we enter as it were into his body, and be-
come in some measure the same person with him, and thence form some idea of his
sensations. . . .5

Contemporary simulation theorists change the example of "our brother upon the rack"
to our brothers missing their planes in the much discussed case of Mr. Crane and
Mr. Tees (originally used by Kahneman and Tversky, 1982). Arriving at the airport
thirty minutes late for their respective flights, Mr. Crane is told that his flight left on
time while Mr. Tees is informed that his flight had been delayed and had left just
five minutes ago. In this situation, we can know that Mr. Tees is more upset than
Mr. Crane because we put ourselves first in Mr. Crane's and then in Mr. Tees's
circumstances and figure out that we would clearly have been more upset in the lat-
ter case. Since we would have been more upset, we conclude that Mr. Tees is the
more upset person in the situation. This claim is not based on the application of any
general law like "people who are late by five minutes are more upset than those who
miss their targets by a more substantial margin." Rather, we come to know by using
ourselves as an analog, as it were, and "putting ourselves in Mr. Tees's shoes." The
actual steps of such role-playing can be brought out by considering yet another ex-
ample used by Goldman6: our attempt to understand a victim of sexual harassment
like Anita Hill. The process is comprised of the following steps:
1. Construction of Anita Hill's "situation," that is, a retrieval of facts about her, such
as that she is a smart, ambitious woman who moved with Clarence Thomas to the
EEOC.
2. Adoption of this constructed situation as "pretend inputs" to be fed into my cogni-
tive mechanism, that is, my imaginatively putting myself in Anita Hill's shoes and
adopting her situation as "mine."
3. Playing out the consequences of such pretense, that is, seeing what results are gen-
erated by my cognitive mechanism once these pretend inputs have been fed in. Pre-
tending to be in Hill's situation, I may come up with "guilt" and "a desire to keep
quiet."
4. Running these states "off line" (as outputs of pretend inputs, these states are after
all "pretend" outputs). I do not actually start feeling guilty or desire to keep quiet
but ascribe them to Hill and conclude that Anita Hill desired to keep quiet and felt
guilty even as she followed Clarence Thomas to the EEOC.
224 Vi-inda Dalmiya

Two important clarifications are needed to pin down the process of simulation fur-
ther: How is simulation not a more complicated form of analogical inference? In other
words, are not "off line" ascriptions of mental states a form of arguing from myself
to another? And, as an epistemological route to subjectivity, can simulation be a way
of knowing myself! Or must it start with an introspective account of my own mental
states and consequently be grounded in a quasi-perceptual model of self-knowledge
after all?
To address the latter issue first: Simulation is involved in hypothetical and counter-
factual thinking. I role-play and imagine occupying not only others' situations but
actual or possible situations other than what I am in now. And this is often a crucial
step in figuring out what I really want. Thus, the decision whether to accept a lower
paying but less-stressful job in a smaller town is negotiated by my imaginatively
"enacting" this counterfactual and seeing "how it feels." This helps me clarify my
own priorities and reveals myself in a more nuanced manner than reflection involv-
ing deductions from generalizations like "I am put off by faceless interactions and
the nonsustainable life-style in a big city." Even though I may sincerely believe the
latter to be true, I may not actually want to move to a smaller place—a conclusion
that I can reach only by "living through" the situation either imaginatively or actu-
ally. Objective "reasoning" can in some instances serve to hide that fact from my-
self. Simulation thus is a source of self-knowledge.
That simulation is not analogical reasoning depends on the precise form of the
pretense in step (2) above. Here I, Vrinda Dalmiya, do not pretend to occupy Anita
Hill's situation, but I pretend to be as Anita Hill would be in her situation. I pretend
to be in Anita Hill's shoes as Anita Hill and not as myself. We have here "not a trans-
fer but a transformation."7 In the imaginative play of step (2), the referent of "I" shifts
from me to the person being simulated and, consequently, the "off line" ascription
of mental states generated by my cognitive apparatus is not meant to be an analogi-
cal move from me to the other.
However, the difficulties in this strategy of knowing selves are many. The one
that I want to raise here pertains to the issue of reliability. Do we have reason to claim
that belief/desire ascriptions based on simulation are anywhere near accurate? In fact,
given the way simulation works, there is more reason to think of it being a fanciful
process than not. And predilections of simulation being unreliable make knowledge-
claims based on it not even prima facie justified and render the process otiose from
an epistemological point of view. The problems begin right at step (1) above that
requires us to construct the "situation" of the target-subject under consideration. Such
construction includes not just a description of empirical facts but what they mean to
the target-subject and how she perceives them (the "worlds" of Lugones8). This con-
dition is hard to satisfy. The examples of successful simulation (Gordon's knowing
which book his sister would like, and Adam Smith's knowing what it is like for a
person being tortured to suffer) are all very simple cases of knowing people very
similar to us in social location or of people in "situations" (like torture) where these
locations are not that important (and even this may be an oversimplification in some
cases). Given a highly stratified society and the ensuing complexities of crossing
boundaries of class, race, ethnicity, and sexual preferences to the "worlds" of genu-
ine others render the formal constraints (Gordon's "making adjustments for relevant
 Knowing People 225

differences" and Goldman's "to isolate or quarantine any beliefs or desires of his
own that are not shared by the target agent") seem simple-minded as strategies for
simulating across differences. As a way of knowing people, simulation is danger-
ously close to "remaking" the other according to our own lights rather than "encoun-
tering" her.

The "Method of Care"

The option of skepticism can restrict knowledge of other minds to a small group of
like-situated people or to those in relatively unproblematic and unidimensional situa-
tions. However, before opting for such epistemic insularity, I would like to suggest
a third epistemic route to selves that I call the "method of care." Since "care" is a
much used and much maligned concept in recent times, I clarify the terminology used
to explicate what is meant by the term here.9 In very broad outline, the "method of
care" incorporates the insight that simulation has to work along with other processes
to lead to justified claims about other people. Embedding simulation in a broader
context that provides built-in veracity checks and mechanisms to fine-tune and modify
raw simulationist data is what I call "care."
The birth of simulation can be traced to an a priori ethical moment of making a
person in all his or her particularity important. Call this "caring about." Such acknowl-
edgment that the other makes a difference in our lives leads to three further steps. It
motivates us to attempt the imaginative projection at the heart of simulation. Call
this shift to the other's point of view a "caring for." The deeper the initial ac-
knowledgment of worth, the stronger is our involvement with the other. And this
may go beyond just an "attitudinal" shift (in simulation) to a "volitional" shift—where
the projects of the other (learned through a simulationist "caring for" her) are adopted
as our own, and we work for their flourishing. This is "taking care." Further, if a
person p genuinely matters, then it is also important that we think ofp as important.
Thus, the initial ethical moment of "caring about p" leads to a reflexivity—to "car-
ing about caring (aboutp)." "Caring about caring" would, in turn, imply an "involve-
ment" with the nature of the caring relation—an attempt to understand it and make it
(the caring) flourish—just as much as "caring about p" led to "caring for" and "tak-
ing care" of p. This self-reflexivity turns what may have begun as an unreflective
natural ability to care into a self-consciously entrenched trait. But caring has an
emotional trajectory as well. What is important if I care, is not just the other person
but also that she recognize that I consider her important. In a full-blown caring, I
want not merely that the other be content, but that part of her contentment be caused
by my involvement with her. In other words, I desire what may be called "care-
reception"—an acknowledgment (by the person cared-for) of my efforts of "caring
for" and "taking care." This five-step process involving "caring about," "caring for,"
"taking care," "care reception," and "caring about caring" constitutes a complex
interrelational system. This I claim can be the basis for justified claims about other
minds.
It is not hard to see why such a "method of care" has the richness to turn mere
simulation into something more reliable. The key lies in that "caring about" a person
not only makes me want to express what I learn about her through simulation ("car-
226 Vrinda Dalmiya

ing for") in a "taking care" but also makes me sensitive to how that care is being
"received" by the person. "Care-reception" provides the mechanism whereby the
simulationist data, the basis of my "taking care," can be tested. The subject here is
free to resist or to express dissent, and I am motivated to take these responses seri-
ously and modify my initial assumptions about what she desires because I "care about
caring" and want the relationship to continue.
Let us consider a concrete example to make the point clearer. Suppose Ravi is
important for me and I genuinely "care about" him. Using simulation, I ascribe to
him the desire to get a full-time teaching position in a research university ("care for"),
and based on this I start sending him all the announcements of academic vacancies
that I come across ("taking care"). One day, Ravi makes a snide remark about my
efforts ("care reception"). I am hurt but realize that I had overestimated Ravi's
enthusiasm for academics. Now I could very well have turned a deaf ear to Ravi's
response because I do believe that his interests would be best served by his entering
academics. But the nature of my involvement and rny desire to be related to Ravi
("caring about care") make me "needy" of his approval (of my efforts). So I am
impelled to take his responses seriously and change my beliefs about his wants and
desires. But, in the process, I may also come to realize that my own bias for aca-
demics comes with a prejudice against the value of manual labor. Alternatively, sup-
pose through simulation, I am forced to ascribe to Ravi some racist beliefs that make
me unwilling to "take care" of (what I think are) his projects. Would my caring abort
here? Not immediately. The ethical moment ("caring about") that made Ravi impor-
tant for me, also makes it important that I care for him. And so, in virtue of my "car-
ing about caring (for Ravi)," I would be led to engage with him in an attempt to keep
the relation going—I would try and present my point of view and get him to explain
his. But at the same time, I would also be reflecting on my own need that makes him
important in the first place and trying to figure out whether I want my investment in
him to continue, whether it is worthwhile modifying my own projects to accommo-
date his. Because of this dynamic a number of different possibilities emerge: (1)1
may realize that I was mistaken, that Ravi is not a racist after all; after our conversa-
tions (2) Ravi himself may agree to soften his position; or (heaven forbid) (3) he may
even manage to change my views about the legitimacy of his position. It is also pos-
sible that while reflecting on the nature of my care for him, on why I thought him to
be "worthy" to begin with, (4) I realize that I do not want to engage with him any-
more after all, and call a halt to my caring.10
What the "method of care" underscores, then, is that knowing people is a genu-
inely interactive process with the following implications: (1) First, interrelationality
provides a check to bare simulations because results of projective empathy need to
be acknowledged and "received" by the person being simulated. Of course, this does
not mean that we can never go wrong nor that the subject being simulated always
provides infallible checks. Reliability is not infallibility. (2) Second, the process of
knowing can actually change both the knower and the person known. Knowing and
the known become dynamically related, and we give up the perceptual model's bias
that knowing leaves the known totally unchanged and gets it as it would be even if it
were unknown. (3) Third, knowing others is at the same time a process of knowing
myself. The reflexivity of care forces a contemplation of the roots of caring and the
 Knowing People 227

priorities of the knower. What I care for and why shows who I am. The known and
the knower are involved in almost a joint project of mutual self-understanding.
(4) Finally, such reflexivity makes the question why I know as important as, and in-
tegrally interwoven with, what I know. Knowing is dynamically related to engaging
with the reasons and motivations behind the knowing itself.
These points, particularly (3) and (4), underscore how the subjectivity of the
knower is involved in the knowing of selves in a manner in which it does not come
into play in the knowing of things. Knowing another self is at the same time a pro-
cess of coming to know one's own self and is an engagement with our values and
motivations that initiate knowing in the first place.

II. From Knowledge of Self to the "Knowing-Self"

Since the knowing of selves spelled out above is significantly different from know-
ing things, can it serve as a model for the latter and for all knowing? Just as the
positivists distorted knowledge of selves by trying to make it look like the knowl-
edge of middle-sized physical objects, we could well distort knowledge of things
by modeling it on a knowledge of selves. In fact, this attempt would require our
empathetically "pretending to be" rocks and cells and "scientists to begin talking
to their rocks and cells or to claim that the process does not work because the rocks
fail to respond."11 Knowledge of selves has the potential to revolutionize episte-
mology precisely because it is so different; but because it is so different, it cannot
meaningfully advance our understanding of knowledge of things. Maybe it is a
nagging paradox of this sort that leads Code to dilute her original insight by the
qualifier, "I am not suggesting that knowing other people should become the new
epistemological paradigm, but I am asserting that it has equally strong claims to
paradigm status, despite worries that purists might have about allowing for more
than one kind of paradigm."12
The issue here hinges on how we want paradigms to work. Obviously, there cannot
be a point-to-point correspondence between knowing people and knowing things. But
I think the power of Code's original idea is that knowing people situates knowledge in
a completely different discursive space by creating a new network of epistemic, onto-
logical, and ethical concepts—and that is why its paradigmatic status can help us re-
think epistemic evaluation and responsibility for all kinds of knowledge.
Let us step back and consider when claims about knowing people are considered
epistemically "worthy." Like any other knowledge-claim, there is a premium set on
"getting it right," and so consideration of its truth is part of the assessment. But this
is only a small part of the cognitive game now. Selves change, as we have seen, in
the process of being known and knowing; and so every "conclusion" reached about
a self comes stamped as being interim. Consequently, it is important to put "time-
slice" appraisals of selves—even when these appraisals happen to be true—in a wider
context of a questioning that cloaks every claim about a person in an acknowledged
fallibility and fluidity. Being essentially an interactive affair, such knowledge aims
not at a simple awareness of truth but an awareness grounded in seeing connections
that can convince. For example, the truth of a psychologist's diagnosis about my
228 Vrinda Dalmiya

troubled eight-year-old is, of course, important but equally relevant is how and the extent
to which the psychologist can "establish" her theory in her dialog with (say) a resistant
me and my daughter. In confronting our resistance, she must be willing to reflect on
the grounds of her own convictions, and her attempt to know my daughter, therefore,
may well dislodge a perception she had of herself. My knowledge-claims (about my
daughter) are to be taken seriously here because they are products of my "attunement"
to her: they are symbols of a cognitive effort that is epistemically worthy quite inde-
pendently of the truths that they happen to lead to. I and my daughter (given that we
have certain dispositions and make certain efforts) have epistemic value because we
have made that effort and we, along with the psychologist, form an epistemic commu-
nity of nonsymmetric, nontransitive relations. Thus I and my daughter relate in a way
that the psychologist and my daughter do not. The psychologist and 1 relate in a way
that my daughter and the psychologist do not. But I may be related cognitively and
caringly to my daughter only insofar as I am so related to myself. It is interactions within
such a community that leads one of us (the psychologist here) to the truth, and without
such community there would be no path to truths about selves. But note that the com-
munity itself is constituted by persons who are willing and able to make certain kinds
of effort. Knowing selves thus underscores a crucial shift from a belief-assessing to an
agent-assessing perspective that makes being an (epistemically) caring person as im-
portant for the cognitive enterprise as the truth of particular beliefs. The "method of
care," after all, is ultimately grounded in a character that engenders trust.
Taking knowledge of selves as paradigms of knowing thus broadens the under-
standing of epistemic worth. Since the knowing self is of primary significance in
"virtue responsibilism,"13 changing cognitive exemplars might well serve as an argu-
ment for a kind of virtue epistemology. But what of the notion of epistemic respon-
sibility now?

III. From the Knowing-Self to

the (Epistemically) Responsible Self

Our assessment of Clifford's shipowner as a sloppy thinker implies not only the
unacceptability of the intellectual product (the belief) he comes up with, but indi-
cates that he is to be held accountable for it. But in what sense does praise or blame
accrue to the knowing, the epistemically deficient, or the negligent self?
On the traditional picture, accountability is traced to the knower abiding (or not
abiding) by epistemic obligations and duties definitive of good thinking. An epistemic
agent who cannot be blamed is one who has been "responsible" and to have been so
is to have believed what is "permissible" and not in violation of epistemic "duties."
Epistemic responsibility is thus tied to deontology. Of course, we can be blamed only
for what could have been done otherwise and is within our control in some sense.
Responsibility for our beliefs thus implies a doxastic voluntarism. This in turn, sug-
gests an internalism emphasizing cognitive access to duties in accordance with which
the knower is expected to structure her doxastic system. Epistemologists have mostly
been engaged in articulating such epistemic duties as "principles of justification." A
responsible knower is expected to be at least implicitly aware of and to use such a
 Knowing People 229

system of deontic principles. A belief h responsible (justified) when it is formed (or

maintained) through the application of these epistemic principles. A knower is
responsible or not according as she has a store of such justified beliefs. Epistemic
responsibility therefore, is constituted by the three motifs of deontology, voluntarism,
and internalism14 as applying primarily to beliefs.
When the "method of care" is made paradigmatic, the above picture changes dra-
matically. The success of knowing others is not, as we have seen, dependent on our
following certain predetermined rules. The "duty" to be responsive to others cannot
translate into a set of principles. Responsibility in the context of care requires sensi-
tivity to contextual detail that necessitates modification, and can require even aban-
doning rules and principles that we may have begun with. Certainly we can be blamed
even here for not having done enough—but this blame cannot be traced to our not
having fulfilled a predetermined range of obligations. Certainly there is scope for
improvement even here, but our efforts to become better at caring are at best guided
by "exemplars" of caring exchanges rather than by formulas.
Such slackening of deontology when extended to all knowing signals the primacy
of doxastic effort and motivational structure—the character of the knowing self—
over epistemic norms. "Responsibility" when affixed directly to character (a "respon-
sible person") is not exhausted by conformity to duties. In fact, to the extent that virtues
like creativity and intellectual courage are constitutive of a worthy cognitive charac-
ter, rule-following cannot be definitive of epistemic responsibility. Of course, it is
important that a virtuous knower acquire and follow certain rules, but the point is
that her epistemic worth is not determined exclusively by them. The attempt to be-
come epistemically praiseworthy is now not so much a matter of not violating spe-
cific duties and discharging obligations but cultivating and reinforcing attitudes
deemed admirable in the community.
Along with such a backgrounding of principle-based deontology, the process of
knowing others reconfigures internalism. Since the method of care and by extension
an agent-evaluative epistemology, allows for ascription of blame and provides scope
for improvement, there must be an element of choice and control in constituting
epistemic agency. However, control here is not generated by an access to principles
that can then be used to restructure character. How and what must we access so that
it becomes meaningful to ascribe blame in situations of care and in epistemic con-
texts where the virtuous knower is foregrounded? What is the shape that internalism
takes once we forsake deontology? I shall approach changes in the notion of epistemic
responsibility through answers to this question.
Of course, internalism has assumed different forms even within traditional dis-
course. But in the broadest of terms it can be conceived as a theory that puts specific
constraints on justifiers (what contribute to epistemic worth) along two axes: con-
straints pertaining to (1) content (what justifiers are) and (2) accessibility (how they
are grasped). From the point of view of content, internalism says that justification-
conferring states are "internal" to the subject—they are states like belief, perceptions,
recollective and introspective awareness. Even Greco's "norm internalism"'5 requires
that the antecedents of the norms that make beliefs justified be constituted by one or
more of these states. As a theory about access, on the other hand, internalism under-
scores that justification is "directly recognizable on reflection."16 The idea here is
230 Vrinda Dalmiya

that any information that is not available to the subject at a certain time is not neces-
sary to determine whether the belief is justified for the subject at that time. Thus what
the knower has reflective access to at any particular time is sufficient to judge the
epistemic status of the belief at that time. (The "content" and the "access" conditions
may be linked17 because what the knower can reflectively access at any time are "in-
ternal states" like her doxastic, perceptual, memorial, and introspective awarenesses).
Now, postulation of the knowing-self 'as the ground of epistemic worth obviously
affects both "content" and "access" internalism. Since character of the knower is now
the primary justification-conferring factor, the issue from the point of view of con-
tori-internalism hinges on whether character is an "internal" state of the subject on
par with beliefs and other kinds of awarenesses spoken of in the literature. Here it
can be argued that even though not a state "of" the subject like beliefs and percep-
tions, it is hard to think of character as being "external" to the subject in any substan-
tive sense. As the normative organization of discrete states of awareness, character
is "internal" to the subject in being constitutive of the subject—it constitutes the very
point of view that an agent stands for. Zagzebski's arguments18 also come in handy
here. Character is constituted by intellectual virtues and vices. These, in turn, are
defined by characteristic motivations to achieve certain ends accompanied by suc-
cess (generally) in achieving those ends. The concept of an intellectually virtuous
character is "internalist" to the extent that it involves motivations regulating actions
and attitudes in the quest for truth. On the other hand, it is "externalist" because such
motivations have as their end truth, and for the most part, they succeed in achieving
that end. In this way, given the "dual-aspect" analysis of intellectual virtue, an agent-
based epistemology reconciles the divergent streams of internalism and externalism.
Locating epistemic worth in the knower thus remains internalist but it is clearly an
internalism with a difference.
Equally interesting changes are brought about in access-internalism once we shift
to knowers as the ground for justification. The issue here hinges on whether we can
access or recognize who we are (our characters) directly "through reflection alone."
Now, the burden of our discussion in the previous sections has been to make a case
for the contrary. Reflective looks within often tend to hide rather than reveal our true
natures. Self-knowledge is dependent on others in more senses than one. (1) We need
to look at what we actually make important in our lives (instead of what we think is
important) to grasp who we really are. So it is what we do to and with others, rather
than what we can introspectively access, that reveals the nature of our character. For
example, I may think that my aging aunt is important for me. But if I keep shelving
involvement with her, that is, if I constantly find excuses to avoid "caring for" or
"taking care" of her, then clearly she does not matter to me however much I might
believe that she does. (2) Moreover, selfunderstanding emerges through taking into
account how others perceive us. It is in the dynamic interaction of responding to these
perceptions that we get a clearer picture of ourselves and our own motivations (re-
member the case of Ravi and the psychologist trying to convince me about my daugh-
ter). Consequently, (3) knowledge of the self emerges through an active and serious
interest in looking "outward," in trying to understand others and the world around
us. The more we genuinely try to know others, the more we are forced into being
in touch with ourselves. But the more we focus directly and exclusively on self-
 Knowing People 231

knowledge, the more we are prone to error and self-deception. Thus given a "method
of care," a pure form of access-internalism is not in order either.
However, "having a character" or "having a characteristic motivation" is differ-
ent from "having evidence" in that the latter allows for gaps between (a) grasping
justification-conferring rules and (b) grasping that our beliefs follow from these rules.
Greco19 makes a distinction between a belief being merely "in accordance" with an
epistemic principle and its actually being "in conformance" to it. It is the latter—
when beliefs actually follow from our acknowledged epistemic principles and are not
just accidentally in accord with them—that makes them epistemically worthwhile.
But it is unlikely, he points out, that ordinary knowers ever have sophisticated meta-
beliefs about such "conformance." Inability to recognize that our beliefs are "in con-
formance" with whatever may be the justification-conferring rules is, according to
Greco, an essential externalist element in theorizing justification according to what he
calls "perspectival internalism." Now, when character is the justification-conferring
state, grasping that our beliefs in question follow from a certain character (the ana-
log of (b) above) is not needed as a distinct step over and above knowing that we do
possess a certain character or motivation (the analog of (a) above). Recognizing the
characteristic motivation behind a belief is to grasp that the belief follows from it
because that is what having a motivation means. Of course, it is quite possible to be
unclear about our motivations. A belief could conceivably be the outcome of many
alternative motives, some of them even conflicting, and we may be unsure as to what
actually led us to the beliefs we hold (a desire for truth? peer pressure? a will to re-
volt against parents?). But once (and if) we are able to sort out our motives, that is,
once we have (a), we do not need an extra cognitive step to determine whether our
beliefs actually followed from that motive or not. A motive would not be one if our
beliefs did not follow from it, while a norm could still be a norm even if we failed to
conform to it. Thus, in a character-based epistemology, "externality" is not intro-
duced because of the distinction between "accordance" and "conformance." How-
ever, as we have argued, there is an element of externality in the very identification
of our motives (or step (a) above).
The changes brought about by a character-based epistemology are actually deeper.
Note that our discussion above works within the broad framework of evaluating
beliefs—where it differs is in introducing character or the knowing-self (instead of
specific duties) as the justification-conferring ground for beliefs. According to a full-
fledged agent-based theory of knowledge, however, knowers are the primary locus
of epistemic worth, and their evaluation is not derivative on the prior appraisal of the
beliefs they have. In fact, beliefs are justified to the extent that they follow from an
epistemically worthy character, but characters are judged to be so quite independently
of the beliefs that they might lead to. Now, when the principal target of evaluation
shifts from beliefs to knowers in this way, is it still possible to entertain an internalism?
Can we, in other words, craft the two special internalist constraints of "content" and
"access" on conditions affecting the epistemic worth of charactert
From the point of view of content, appraisal of a knower as intellectually praise-
worthy or not depends on (1) the extent to which she approximates the character of
"exemplary" knowers and (2) the extent to which this exemplary motivational struc-
ture is entrenched in her. Incidentally, neither of these two conditions is strictly
232 Vrinda Dalmiya

deontological. An exemplary knower is "good" not simply because she stays within
the boundaries of the "permissible" or performs a fixed set of duties. She has the flexi-
bility to fine-tune her responses to the object of knowledge and be guided by it. And
entrenchment of any kind of motivational structure into a habit is again not specifiable
in terms of a set of rules. This loosening of deontology leads in turn to a loosening of
internalism, which is what I want to underscore here. While ideal/exemplary charac-
ter (the touchstone of a "good" knower) is "external" to the subject of knowledge, the
effort to emulate these exemplars is obviously an "internal" affair. You could "con-
tract" a certain ideal character unbeknownst to you by simply keeping good cognitive
company. But you could not try to cultivate such a character without knowing that you
are so trying: Trying needs to be transparent to the person trying.
From the point of view of access, we once again find an intrusion of externalist
elements. The identification of a person as "exemplary" is not a product of reflec-
tion alone. It is clearly a social assessment requiring engagement with what the com-
munity thinks best rather than a product of the subject's isolated deliberations. Note
here that I cannot be a caring person simply because / think (quite sincerely) that I
am one—at least the person that I care for must suggest or imply it (through a posi-
tive "care-reception") before I can lay claim to that status. Apart from the fact that
there can be no self-styled exemplars, externality is heightened by the fact that con-
siderations of reliability and truth-effectiveness figure crucially in communal delib-
erations to determine exemplars. The entrenchment condition of becoming a good
knower also points to a slackening of internalist autovalidation. A kind of "self-
forgetful ness" occurs when traits and motivations become habitual. Part of the point
of entrenchment is to make motivations tinsel/conscious and recede from the thresh-
old of active reflection. Not only is it possible for a virtue to exist unrecognized, but
it may even be a constitutive mark of it. Doesn't an acute self-consciousness that I
am caring detract from the value of the moves if it accompanies every caring ges-
ture? Could Mother Teresa's acts of charity be accompanied by a self-righteous self-
awareness on her part that she is a charitable person without changing the quality of
her goodness?20 This, of course, need not entail that once entrenched, motives can
never be reflectively retrieved but that they can be so only to the extent that motives
themselves are recognizable by reflection.
The upshot of this for our purposes is that when the knowing-self moves to center
stage, epistemic evaluation, whether it is of beliefs or of character, cannot function
within the constraints of a strict internalism. The relaxation of internalist criteria occurs
on two fronts. First, considerations of reliability and success in achieving truth
become relevant, and second, a social dimension is introduced to rupture the isola-
tionism of purely "internal" looks "within." But what are the contours of epistemic
responsibility and accountability that ultimately emerge from all of this?
Put briefly: Epistemic responsibility now is not a function of either not violating
epistemic obligations (deontology) nor of factors purely transparent to the knower
(internalism). Yet the sense of accountability is sufficiently strong to ground epistemic
blame and praise and hence, must involve a kind of agency and control that makes
such ascription (of praise and blame) meaningful. This is the agency involved in
constituting and being true to a kind of character. What we need to access to make
and remake ourselves in a specific way are facts about ourselves—our motives and
 Knowing People 233

desires, the reliability/truth of the strategies we adopt, and what others think both
about ourselves and about reliable truth-generating strategies. All these, therefore,
factor into what makes us epistemically responsible. We cannot be intellectually
responsible and be systematically wrong just as much as we cannot be epistemically
responsible either by simply having a coherent reflective "perspective" that is ut-
terly idiosyncratic or by being generally right and conforming to communal stan-
dards but without a clue about our own motivations.
The importance of such self-awareness, once again, brings in the importance of
care for the construction of epistemically responsible selves. The method of care, we
saw, is not just a way of knowing others but a way of knowing and remaking our-
selves. It begins in an ethical moment of making someone valuable and loops back
to a self-consciousness about our reasons for ascribing such value. Ascription of
importance to some things or to some people when coupled with our finite ability to
take interest (can we make everything important? Doesn't making something impor-
tant imply a making of some other things wmmportant?) creates a salience that
deprioritizes some other things and people. The reflexivity of care entails that we be
accountable for these ethical choices, that we be ready to bear the blame or claim the
credit for our ethico-political perspective on the world. To conclude then: The me-
chanics of knowing people puts the normative burden on us as knowing people. But
the knowing-self is at the same time the ethico-political self who constantly negoti-
ates her cognitive agency within a community of other knowing people.

Notes
1. Lorraine Code, What Can She Know? (Ithaca, N.Y.: Cornell University Press, 1991);
"Taking Subjectivity into Account" in Feminist Epistemologies, Linda Alcoff and Elizabeth
Potter, eds. (New York: Routledge, 1993), 15-48.
2. See Guy Axtell, "Recent Work on Virtue Epistemology," American Philosophical
Quarterly 34, 1-26 for the distinction between virtue reliabilism and virtue responsibilism.
The convergence with virtue responsibilism is to be expected here because Code herself is
among its first proponents. See her Epistemic Responsibility (Hanover, N.H.: University Press
of New England, 1987).
3. Alvin Goldman, "Empathy, Mind and Morals" in Mental Simulation, Martin Davies
and Tony Stone, eds. (Oxford: Blackwell, 1995), 185-208.
4. Robert Gordon, "Simulation Without Introspection or Inference From Me to You" in
Mental Simulation, Martin Davies and Tony Stone, eds. (Oxford: Blackwell, 1995), 53-67.
5. Adam Smith, Theory of Sentiments, D. D. Raphael and A. L. Macfie, eds. (Indianapo-
lis, Ind.: Liberty Fund, 1984), p. 9.
6. See Goldman, "Empathy, Minds and Morals."
7. Gordon, "Simulation Without Introspection or Inference from Me to You," 275-90.
8. Maria C. Lugones, "Playfulness, 'World'-Travelling and Loving Perception," in
Women, Knowledge and Reality: Explorations in Feminist Philosophy, Ann Garry and Marilyn
Pearsall, eds. (Boston: Unwin Hyman, 1989), 275-90.
9. My analysis of care draws on the work of Nel Noddings, Caring: A Feminine Approach
to Ethics and Education (Berkeley, Calif.: University of California Press, 1984); Harry Frank-
furt, "The Importance of what We Care About," Synthese 53 (1982), 257-72; Annette Baier,
"The Importance of What We Care About: A Reply to Frankfurt," Synthese 53 (1982), 273-
90; loan Tronto, Moral Boundaries: A Political Argument for an Ethic of Care (New York:
Routledge, 1993). I have developed the notion of care in more detail in my "Why Should a
Knower Care?" (unpublished).
10. This possibility is often overlooked in feminist criticisms of care.
234 Vrinda Dalmiya

11. Code, What Can She Know?, 164.

12. Code, What Can She Know?, 31.
13. As Axtell points out in "Recent Work on Virtue Epistemology," there are many ver-
sions of virtue responsibilism. I have in mind particularly Zagzebski's agent-based responsi-
bilism. See her Virtues of the Mind (Cambridge: Cambridge University Press, 1996).
14. See Alvin Plantinga, Warrant: The Current Debate (New York: Oxford University
Press, 1993).
15. John Greco, "Internalism and Epistemically Responsible Belief," Synthese 85 (1990),
245-77.
16. Matthias Steup, "A Defense of Internalism," in Theory of Knowledge, 2d ed., Louis
Pojman, ed. (Belmont, Calif.: Wadsworth, 1999), 337-84.
17. Ernest Sosa, "Skepticism and the Internal/External Divide," in The Blackwell Guide
to Epistemology, John Greco and Ernest Sosa, eds. (Cambridge, Mass.: Blackwell, 1999),
145-57.
18. Zagzebski, Virtues of the Mind.
19. Greco, "Internalism and Epistemically Responsible Belief."
20. A point along these lines about the nature of virtue was made by Arindam Chakrabarti
in his response titled "Unknown Knowledge" to Nicholas Smith's paper, "Cognitive Self-
Evaluation, Animal Knowledge and the KK Regress" at the 47th Annual Conference of the
Far Western Philosophy of Education Society held in Honolulu, Jan. 1999.
 14

Recovering Understanding

LINDA ZAGZEBSKI

There have been radical changes in epistemology in recent years, although not al-
ways in the same direction. The naturalistic epistemologists treat it as a branch of
psychology, whereas I treat it as a branch of ethics. Still others gleefully pronounce
the field dead on the grounds that the central epistemological questions are based on
defective assumptions. In the first part of this essay I want to look at what we can
learn from the death-of-epistemology camp because I think that the lesson of their
argument is that epistemology ought to be reoriented in a direction I believe is desir-
able for many other reasons. In particular, I want to focus on the neglect of the
epistemic value of understanding. Recovering understanding requires an approach
like the virtue approach I have endorsed elsewhere. It also may alter the way we re-
spond to skepticism.
The death-of-epistemology theorists claim that epistemology is primarily con-
cerned with demonstrating that knowledge is possible, and that project arises only if
we take the threat of radical skepticism seriously. But the peril of skepticism is not
perennial. It is only because of certain historical contingencies that it has assumed so
much importance in the last three hundred years. And the danger arises only within
the context of a certain set of theoretical ideas—ideas that can and must be challenged.
Beyond this, the death-of-epistemology theorists do not agree on the particular
set of ideas that are at fault. Richard Rorty (1979) blames skepticism on the "repre-
sentational" conception of belief and the correspondence theory of truth, while
Michael Williams (1991) blames foundationalism. But they agree that the demise of
the presuppositions behind the skeptical challenge undermines its danger and makes

235
236 Linda Zagzebski

the attempt to answer it pointless. Since answering skepticism is reputedly the major
job of epistemology, it makes epistemology pointless as well, and to the extent to
which the rest of philosophy is built on epistemology, philosophy itself is allegedly
in danger of collapse.
I have no interest in pointing out the many defects in this line of reasoning. In-
stead, I want to start with what I think is a worthwhile contribution to philosophy
from this argument, and that is the way it reminds us how the questions that we con-
sider central to epistemology change over time and are not presented for our reflec-
tion singly, but in clusters. An important way in which questions bunch together is
around the issue of skepticism. There have been significant periods of philosophical
history in which skepticism was thought to be a serious threat, and other periods in
which it was not. In those eras in which it was, philosophers gave preeminence to the
epistemic value of certainty and focused on the nature of justified belief. In those
eras in which it was not, the questions that assumed most importance were quite dif-
ferent. Philosophers gave most of their attention to the value of understanding and
focused on the nature of explanation rather than on the nature of justification. We
may be ending a period in which radical skepticism is taken in full seriousness, but
whether or not that is the case, we should admit that the questions of most signifi-
cance to epistemology in the askeptical periods have been neglected. It is time we
cease the obsession with justification and recover the investigation of understand-
ing. At least, that is what I intend to propose.
My position is that (1) the death-of-epistemology theorists are right that the dan-
ger of skepticism has been perceived quite differently in different periods of philo-
sophical history, and one reasonable conclusion to draw from that (but not the only
one) is that skepticism does not arise from the pure a priori use of presuppositionless
reason, but only within a context of certain substantive philosophical positions.1
But (2) the death-of-epistemology theorists are mistaken if they think the serious-
ness of skepticism is a peculiarly modern phenomenon. There are, in fact, revealing
similarities between the modern skeptical period and the post-Aristotelian skeptical
period, both of which are significantly different from the askeptical periods that make
up most of the rest of philosophical history. But I am not interested in the theoretical
presuppositions of skepticism for the purposes of this paper. My point is just that
many epistemologists are ready to put skeptical worries aside, although not all for
the same reasons. This is not to say that most of them accept the claims of the death-
of-epistemology theorists. In fact, they clearly do not. But many now agree that epis-
temology is stultified if the issue of skepticism is allowed to dominate epistemologi-
cal inquiry. It is perfectly legitimate to pursue epistemological issues that leave the
skeptical challenge to one side.
But what are the questions that should then capture our attention? A good way to
answer that, I propose, is to look at the questions that dominated epistemology dur-
ing askeptical periods. Since I believe this opens up an important range of neglected
issues involving the epistemic value of understanding, I conclude that (3) the death-
of-epistemology theorists are quite wrong that there is nothing left for epistemology
to do. In what follows I will look at the work of some historians of philosophy for
interpretations of the epistemological enterprise in previous eras. This will lead to
 Recovering Understanding 237

some suggestions for the future of epistemology. But, unfortunately, it is not enough
to announce that skepticism should be treated independently and then continue to
use the approaches that are most common in Anglophone epistemology. Most of the
dominant theories make it almost impossible to undertake an investigation of under-
standing. I will then argue that virtue epistemology has the greatest promise of giv-
ing us a new and useful way to approach the neglected value of understanding.

II

Understanding and certainty are two epistemic values each of which has enjoyed pride
of place for long periods in the history of epistemology, but rarely, if ever, at the
same time. In Hellenistic philosophy and much post-Cartesian philosophy, certainty
was given more attention than understanding, while in Plato and Aristotle, in the long
medieval period, and even in some of the major modern philosophers such as Spinoza,
it was the reverse. Usually whichever one of the two concepts dominated was the
one connected with the concept of knowledge, so Plato comes very close to identify-
ing knowledge with understanding, while Descartes comes very close to identifying
knowledge with certainty.2 Historians have attributed this difference in focus to dif-
ferences in the way skepticism has been handled. The rise of skepticism is accompa-
nied by the concern for certainty, and that brings with it a batch of questions, most of
which focus on propositional belief and the process of justifying belief, since justi-
fication is what is needed to defend the right to be sure. In contrast, the askeptical
periods have been mostly concerned with understanding, and the questions accom-
panying it show little concern for justification, often not even much interest in propo-
sitional belief, but instead, an interest in the process of explanation, since explana-
tion is what is needed to defend a claim to understand. The understanding/explanation
orientation is much less atomistic and more social than the certainty/justification ori-
entation. This is noteworthy in the present climate of opinion, since these features of
the understanding approach are the very things that would meet some recent criti-
cisms of contemporary work in epistemology.3
So the rise of skepticism led to a move from a concern with understanding and
explanation to a concern with certainty and justification in ancient epistemology.
As Stephen Everson points out in the introduction to his book on ancient episte-
mology (1990, p. 7), "It has become generally accepted . . . that what marks off
post-Aristotelian epistemology from what went before is a novel concern with justi-
fication . . . and that this concern was elicited by the onset of scepticism." In contrast,
the central epistemological aim of Plato, as Julius Moravcsikhas argued (1979,1992),
was the delineation of what it means to understand something (1979, p. 53). Propo-
sitional knowledge was of secondary interest (p. 55) and derivative from it:
The only propositional knowledge that will be of interest will be that which is derived
from the kind of theoretical understanding that Plato envisages. Mere knowledge of
truths is of no interest to Plato; propositional knowledge figures in the dialogues only
insofar as this may be, in some contexts, evidence for understanding, and needed for
practical activity, (p. 60)
238 Linda Zagzebski

Gail Fine (1990) has a similar interpretation:

On the account [of Plato] I have proposed, one knows more to the extent that one can
explain more; knowledge requires, not a vision, and not some special sort of certainty
or infallibility, but sufficiently rich, mutually supporting, explanatory accounts. Knowl-
edge, for Plato, does not proceed piecemeal; to know, one must master a whole field,
by interrelating and explaining its diverse elements, (p. 114)

Gail Fine translates the word episteme in Plato as "knowledge," but in a sense
that includes understanding, whereas Moravscik translates it simply as "understand-
ing."4 Either way, understanding is a central epistemic value in Plato. But not justi-
fication. Occasionally, a contemporary philosopher such as Chisholm (1966) will
claim to find the origin of the idea of epistemic justification in Plato's Theaetetus
(201c-210b), where Socrates proposes and then rejects the suggestion that knowl-
edge is true belief plus a logos. But both Burnyeat (1980, pp. 134ff) and Fine (1990,
p. 107) have disputed the idea that Plato's notion of a logos can be identified with
justification, maintaining that a logos is an account or an explanation, and as already
mentioned, Everson says that the historians agree that the concept of justification
was a new idea in Hellenistic philosophy, associated with the rise of skepticism.
In examining Plato's early theory of knowledge, Paul Woodruff argues that Plato
distinguished expert from nonexpert knowledge,5 where he indicates expert knowl-
edge by techne and its cognates, and, in many contexts, also by episteme and sophia.
To be expert at a techne is therefore associated both with knowledge/understand-
ing and with wisdom. The expert, Woodruff says, is a person on whom others may
rely in difficult or highly technical matters (p. 68). What makes the expert reliable
is that he knows the essential nature of his product (p. 76). Whether he has that
knowledge is revealed in his ability to give a Socratic definition of the good pro-
duced by his techne (p. 74). Woodruff then argues that it turns out that the same
basic knowledge is essential to every techne—knowledge of the good. It follows
that the person who is the most reliable source of knowledge for other people is
someone who has mastered a skill and in doing so has a basic knowledge of the
good shared by all experts.
Aristotle's notion of episteme is also one that is far removed from a concern with
justification arising out of the threat of skepticism, and for that reason, I have heard
some scholars remark that they have trouble finding his epistemology at all. But that
reveals more about the contemporary philosopher than about Aristotle. Consider what
C.C.W. Taylor says at the beginning of a paper on Aristotle's epistemology:
While Aristotle was certainly aware of sceptical challenges to claims to knowledge,
whether in general or in specific areas, the justification of knowledge claims in response
to such challenges, which has been central to most epistemology since Descartes, is at
best peripheral to Aristotle's concerns. On the whole, he does not seek to argue that
knowledge is possible, but, assuming its possibility, he seeks to understand how it is
realised in different fields of mental activity and how the states in which it is realised
relate to other cognitive states of the agent. In particular, the central problem of post-
Cartesian epistemology, that of showing how our experience may reasonably be held
to be experience of an objective world, is hardly a problem for Aristotle. The problem
for the post-Cartesian philosopher is how, once having retreated in the face of Carte-
 Recovering Understanding 239

sian doubt to the stronghold of private experience, he or she can advance sufficiently
far beyond that experience to recover the objective world. Aristotle, never having made
the retreat, does not have the problem of the advance . . . (1990, pp. 116-117)
And Burnyeat:
All through the Hellenistic period, both positive philosophy and the negative attacks
of scepticism take their starting point to be the problem of perceptual certainty. Aristotle
does not. But not because he is unacquainted with sceptical arguments for conclusions
which would undermine his enterprise, nor because he does not think (some of) them
worth extended discussion. He is simply very firm that he is not going to let them struc-
ture his inquiries or dictate his choice of starting-points. (1980, p. 138)
Like Plato, Aristotle was particularly interested in understanding, either as a form
of knowledge or as a special cognitive state.6 The dispute about the connection be-
tween understanding and knowledge in Aristotle is reflected in the fact that episteme
in Aristotle is variously translated "science," "scientific knowledge," or "scientific
understanding," but sometimes nous is translated as "understanding" instead. I will
not investigate Aristotle's rather complicated notion of understanding, as it has been
discussed in the literature. My point is merely that Aristotle's epistemological inter-
ests were quite different from ours, concerned with a complex of ideas associated
with understanding rather than with justification, and that those concerns can be
explained in part by his lack of interest in skepticism.
So the practice of epistemology in the ancient world was quite different when
menaced by skepticism than when no such peril was perceived. The rise of skepticism
in Hellenistic philosophy is both the source of and the explanation for the shift
from the Platonic/Aristotelian concern for understanding to the post-Aristotelian
concern for certainty, and that shift was associated with the emerging dominance
of the concept of justification. It would be interesting to investigate how philoso-
phers went about the return to the askeptical epistemic values of understanding and
wisdom in the early medieval period, since our situation is similar. We also live at
a time in which the focus of epistemological investigation is the individual belief
state. For decades, epistemologists have concentrated on justification, and scarcely
a word is said about understanding. There is not even an entry for "understanding"
in the new Routledge Encyclopedia of Philosophy nor in the recent Blackwell
Companion to Epistemology, and rarely is there an entry for "understanding" in
the indices of scholarly books in epistemology.7 On the other hand, justification is
given even more attention than knowledge, notwithstanding the fact that the term
"justification" may sometimes be explicitly rejected, as in Plantinga's recent work
on warrant. In Plantinga's case, he rejects it because of the ambiguities he finds in
the concept of justification. Notice that this means that justification can direct the
inquiry, even when the theorist refuses to use the word "justification."8 But if I
am right that the concept of justification is associated both historically and
conceptually with the perceived danger of skepticism, and if I am also right that
epistemologists are now prepared not to let skeptical worries direct their inquiry,
it follows that it is advisable to give up the preoccupation with justification and
perhaps even with the prepositional belief and to turn instead to an investigation
of understanding.
240 Linda Zagzebski

III

So far we have only hints about the nature of understanding from its use in ancient
epistemology, particularly by Plato. But we need much more than hints if we are to
have reason to attempt to recover it. If understanding is a notion peculiar to Plato's
philosophy, it is unlikely that very many contemporary philosophers believe it ex-
ists at all. It is always problematic to use a single term to apply to the object of study
of philosophers from completely different historical periods and with completely
different background theories. This occurs whether we are talking about knowledge,
freedom, causation, the human person, justice, the good, or many others. So when
we discuss Plato's theory of knowledge, can we be sure that it is a theory of the same
thing as Chisholm's theory? Are they giving two different accounts of the same thing
or accounts of two different things? I am not convinced that this question has a de-
terminate answer, but it can be finessed to some extent if we find a common ground
between us and Plato. What I will try to do in this section is to find the common ground
in the concept of understanding.
There is another problem that makes it hard to find our target. A consequence of
the neglect of understanding as an object of philosophical investigation is that the
word "understanding" is used loosely, with a wide variety of meanings, and usually
without any notice of ambiguity. When Alston (1993) and Plantinga (1990) detected
ambiguities in the concept of justification, they both treated that as a serious prob-
lem and concluded that the concept must be disambiguated before there could be
further epistemological progress. In contrast, even when people notice that "under-
standing" is ambiguous, they do not consider that a real problem, since they assume
that all they have to do is to stipulate the meaning they want before proceeding. The
fact that so little contemporary literature depends on any given sense of understand-
ing seems to make the choice inconsequential. This situation makes it very difficult
to even identify the value that needs recovering. One of the consequences of schol-
arly neglect is fragmentation of meaning. And the more that meaning is fragmented,
the more reason there seems to be to neglect it since there no longer seems to be any
"it" that is being ignored. This vicious circle of neglect has probably occurred be-
fore in philosophical history with other concepts.9
Let us see, then, if we can find some commonality between us and Plato's notion
of episteme. From Moravscik's and Woodruff's work, we see that understanding
(episteme) in Plato has something to do with techne—practical human arts or skills.
These include such complex activities as medicine, hunting, and shipbuilding, two
of which are no longer widely practiced, but they also include more-specific practi-
cal skills, such as cooking or even pastry-making. Understanding is a cognitive state
that arises from techne, and since techne includes certain practical activities that are
by no means wholly cognitive, it follows that understanding in Plato is a state that
arises from practices that are not purely cognitive. The person who has mastered a
techne understands the nature of the product of the techne and is able to explain it.
She also knows the good in a sense that gives her a common understanding with the
practitioners of other technai.
This seems to be the case even when the techne in question is an academic field.
Fine says in the passage quoted above that to have episteme one must have mastered
 Recovering Understanding 241

an entire field. One does not have episteme of an astronomical fact10 without interre-
lating and explaining its relation to diverse elements within the field of astronomy.
And one can do that only by mastering the techne of being an astronomer. In other
words, one does not understand a part of a field without the ability to explain its place
within a much larger theoretical framework, and one acquires the ability to do that
by mastering a skill. Similarly, one does not have episteme of some feature of
human psychology without the ability to explain how that feature fits into the larger
framework of human psychology, and that requires having mastered the techne of
the psychologist. Assuming that epistemology is also a techne, it follows that one
does not have episteme of some object of epistemology, such as having episteme of
what knowledge is, without the ability to explain how knowledge fits the other ob-
jects of study in the field, and one cannot do that without having mastered the skills
of the epistemologist.
So far, two ideas about understanding have emerged from the Platonic notion of
episteme that are live options in the contemporary milieu. One is that understanding is
a state gained by learning an art or skill, a techne. One gains understanding by know-
ing how to do something well, and this makes one a reliable person to consult in mat-
ters pertaining to the skill in question. This way of looking at understanding makes it
unlikely that understanding is achieved by a single mode of reasoning, but it involves
more-complex processes, including, perhaps, processes that are noncognitive. This leads
to the second idea, which is that understanding is not directed toward a discrete object,
but involves seeing the relation of parts to other parts and perhaps even the relation of
part to a whole. It follows that the object of understanding is not a discrete proposition.
One's mental representation of what one understands is likely to include such things as
maps, graphs, diagrams, and three-dimensional models in addition to, or even in place
of, the acceptance of a series of propositions.
The formal structure of some states of this kind have been examined by John
Etchemendy and John Barwise.11 They have investigated the complex reasoning
involved in problem solving that they call heterogeneous reasoning. This form of
reasoning would be used in such situations as planning the layout of a group of
offices that meet certain requirements, designing a cellular telephone, a computer,
or an airplane, or figuring out a food distribution system that maximizes utility. In
the cases Barwise and Etchemendy consider, there is no unique solution to the prob-
lem, but the goal is to find any solution that meets the requirements, and these could
include cost, efficiency, safety, style, aesthetic quality, as well as probabilistic con-
siderations. The reasoning involved is not only complex, but often collaborative. The
process uses diagrams, graphics, and other representational forms in addition to
sentential reasoning.
People who make discoveries or solve problems often do so by a process of visu-
alization, sometimes even in dreams. One of the most famous examples of this was
the discovery of the structure of the benzene molecule by Auguste Kekule. While
working on the problem, he dreamed of a snake biting its own tail. Upon awakening
he realized that the benzene molecule had a ring structure. There are many other
examples of this in science. The American mathematical geneticist, Sewall Wright,
used the metaphor of a landscape with hills and valleys as a way of explaining adap-
tive genetic mechanisms. Selection drives populations up the slopes, he argued. Sci-
242 Linda Zagzebski

entists who think that populations seek the closest lowest level have suggested turn-
ing the image upside down. Either way, people find that the image gives them an
understanding they would not have had otherwise and it has become well known.
We now have identified three features of understanding: It is acquired through
mastering a techne; its object is not a discrete proposition but involves the grasp of
part/whole relations; and it involves representing some portion of the world non-
propositionally. Let us look more closely at this last feature.
I take it for granted that reality has a structure or structures and that the compre-
hension of these structures is an important epistemic goal. I also take it that it is un-
likely that propositional structure exhausts the structure of reality. This point raises
deep metaphysical questions about the relation between mind and reality. Whether
or not propositions exist independent of the mind, and whether or not something
sentential in form is the bearer of truth value, we can all probably agree that there is
a form of knowledge that is mediated through a sentential structure, leaving aside
the question of whether the mind imposes such a structure or whether it is there—
independent of the mind. But I think it is consistent with theoretical neutrality to
maintain that reality has a propositional structure as long as we are not committed
to the position that propositional structure is the only structure it has. In fact, I want
to claim now that it is very unlikely that propositional structure exhausts the struc-
ture of reality, and it is even unlikely that propositional structure exhausts the struc-
ture of what is intelligible to us.
I propose that understanding is the state of comprehension of nonpropositional
structures of reality. In this sense of understanding, we can understand such things
as an automobile engine, a piece of music, a work of art, the character of a human
person, the layout of a city, a causal nexus, a Ideological structure, or reality itself—
this last being the object of the science of metaphysics. There is no reason to think
that any of these things has an exclusively propositional structure, if indeed it has a
propositional structure at all. I am not denying the possibility that all of reality can
be represented propositionally, but I am denying that the proposition is the only form
in which reality can be made intelligible to the human mind.
The structures of music include harmonic structures and rhythmic structures ex-
tended in time as well as formal patterns, such as the sonata form and structures that
blend ingredients simultaneously to produce a sound with a certain color, such as
the simultaneous blending of many distinct instruments in the creation of a single
note in a musical composition.
The structure of works of art is also quite obviously nonpropositional, and for this
reason it is very difficult for interpreters of art to translate their understanding of that
structure into a propositional form. Understanding literary figures is probably easier
to achieve than the understanding of paintings or sculpture, since it is probable that
the aesthetic sense of many persons is not very well developed, whereas understand-
ing a literary character is an extension of understanding oneself and one's friends,
something that is presumably more common. Some, but not all, literature has a nar-
rative structure. The pieces are propositional, but the structure of the whole is not.
Narrative structure is relatively easy to understand because it is the same as the struc-
ture of our own lives.
 Recovering Understanding 243

The technai of art, music, and literature can produce a state that has epistemic
value. The arts enable us to penetrate reality more deeply than we could without them.
Understanding works of art and literature is probably one step removed from under-
standing basic features of reality, since the arts are in part an attempt to understand
reality. That is, to understand a work of art is to understand something that is itself
the product of an attempt to understand something else, although this is, of course,
arguable. But it is much less arguable that the state we get from the successful cre-
ation or appreciation of works of art is understanding, and it has much in common
with the episteme discussed by Plato. I have said that I think it is likely that academic
fields are technai also, although that is more controversial. But it is uncontroversial
that one can have understanding of an academic field and only somewhat more con-
troversial that understanding some part of an academic field may require understand-
ing how that part relates to other parts of it.
Philosophy aims to understand the whole of reality, not simply that portion of it
or aspect of it that is successfully represented by propositions and their constituent
concepts. An enormous advantage of language is that we have rules codifying the
logical relations among propositions, and these are very useful in comprehending a
certain kind of structure. What we do not have is a set of rules codifying the relations
among pieces of the structures that make up an artwork or a piece of music or the
motivational structure of a person, although metaphysicians have sometimes proposed
rules of a sort for the deepest structures of reality, as in Plato's theory of Forms. Of
course, there is little chance that we can reach agreement on any of these matters,
and that is no doubt one of the reasons that so few people attempt to figure things out
in these areas. It is probably also a reason for the lack of prestige of the attempt.
There is one form of understanding that epistemologists have not given up, and
that is the understanding we get from a theory. Philosophical theories these days are
considerably less grandiose than Plato's, of course, but despite their modest preten-
sions, they often attempt to do what Plato did in the respect we have been consider-
ing: to represent the objects of epistemological inquiry nonpropositionally. For ex-
ample, the two main models of justified doxastic structures for some time were the
coherentist raft and the foundationalist pyramid.12 Even though many epistemolo-
gists have given up these models, they look for alternatives that serve the purpose of
the raft and the pyramid—to represent the structure of an entire system of beliefs
when it is epistemically good. So even when one way of attempting to give us under-
standing of the whole of a good belief system is unsuccessful, nobody rejects the
point of the attempt.

IV

What is the relationship between understanding and knowledge? As we have seen,

understanding in Plato is closely connected with knowledge, since episteme is a state
in which the two are not clearly distinguishable, and the same word is variously trans-
lated as "knowledge" or "understanding." But since knowledge these days almost
always means propositional knowledge, and since I have proposed that understand-
244 Linda Zagzebski

ing has a nonpropositional object, understanding differs from knowledge as it is

normally understood. I argue elsewhere (1996) that we should not assume that propo-
sitional knowledge is the only kind of knowledge. Nonetheless, I think that knowl-
edge and understanding differ. We can have both understanding and knowledge about
the same part of reality. Understanding deepens our cognitive grasp of that which is
already known. So a person can know the individual propositions that make up some
body of knowledge without understanding them. Understanding involves seeing how
the parts of that body of knowledge fit together, where the fitting together is not it-
self propositional in form.
Moravscik uses the example of understanding a mathematical theorem to make
the same point. A proof of a mathematical theorem is a sequence of propositions, but
understanding the proof involves seeing the relations between the propositions, and
that is not itself the knowledge of a proposition (1979, p. 55).13 It is possible that
understanding a theorem is just a different kind of knowledge, but it differs from
knowledge in the usual sense and it often cannot be attained until after knowledge in
the usual sense is reached. In some cases, then, understanding is a stronger state than
knowledge.
On the other hand, understanding does not always build on a base of knowledge.
It may be achieved in more than one way about the same portion of reality. More
than one alternative theory may give understanding of the same subject matter. This
makes sense if we think of a theory as a representation of reality, where alternative
representations can be better or worse, more or less accurate. But more than one may
be equally good, equally accurate.14 This form of understanding does not presup-
pose knowledge or even true belief, and if we assume that two competing represen-
tations of the same part of reality cannot both constitute knowledge, it cannot be a
form of knowledge.15
Another reason for thinking that understanding is not a form of knowledge is that
in some cases, truth can actually be an impediment to understanding, as Catherine
Elgin (1996) has pointed out. Her example is propositional. "Objects in a vacuum
fall toward the Earth at a rate of 32 ft/sec2" is not strictly true, since it ignores the
gravitational attraction of everything except the earth, yet it gives more understand-
ing for most purposes than the vastly more complicated truth (p. 123). The same
applies to Boyle's Law (p. 124). The strictly true (that is, accurate) law can be grasped,
but only at high cognitive cost.
These considerations suggest that understanding is achieved partly by simplify-
ing what is understood, highlighting certain features and ignoring others. This pro-
cess compensates for our cognitive limitations. Understanding aims at comprehen-
siveness, not exactness, and we usually need to sacrifice one for the sake of the other.
As van Fraassen has noted in a different context, a more informative theory is less
likely to be true, but I would say it is more likely to produce understanding.16 In each
of these instances, understanding aims in a different direction than knowledge, and
yet it does not necessarily have a lesser status.
At the beginning of this essay, I observed that many epistemologists are willing
to put the skeptical threat aside, but that is not because they all agree that knowledge
can be taken for granted. Some do think that, of course, but many others merely think
that the need to answer skepticism should not direct all epistemological projects, but
 Recovering Understanding 245

that skepticism can be put aside while other questions are addressed. In proposing
that the value of understanding ought to be recovered, I do not mean to imply that
we can take for granted that the object of understanding is also something known or
even that it is true. We can leave aside questions of knowledge and threats to knowl-
edge in addressing understanding, but as we will see in the next section, a form of
skepticism arises anyway.
Truth is a thin epistemic goal, and knowledge is derivative from it, since believ-
ing the truth is a component of knowing. Understanding is a thicker goal, and its
connection with truth is often indirect. When we want an expert about a problem, we
consult a person who has understanding of the subject matter, since such a person is
likely to be a reliable problem solver. A reliable problem solver ordinarily will also
be a reliable source of prepositional information, but her reliability is not limited to
being a reliable truth-bearer. The problem solver may use something like the com-
plex process of heterogeneous reasoning investigated by Etchemendy and Barwise,
but whether she does or not, what enables her to figure out the solution to the prob-
lem is her understanding of a complex chunk of the world, not simply her knowl-
edge or true beliefs about that portion of the world.
This suggests another feature of understanding that links it with a techne: Under-
standing makes its bearer reliable in carrying out the goals of the techne, some of
which are not epistemic goals. It enables him to produce a flakey pastry, repair an
automobile, design a bridge that will not collapse, or figure out why the vintner failed
this year. This means that understanding is a property of persons. It is not carried by
propositions or states of belief. This consequence is important because so many con-
temporary theories in epistemology focus exclusively on the proposition or the state
of believing a proposition. We will get to that in section VI.
If understanding is a goal worth pursuing, there ought to be criteria for success in
reaching it. We do have such criteria, but they are not as clear as we would want. We
expect students to demonstrate their understanding of a text or a conceptual point by
reconstructing it in their own words. Understanding in some areas may be displayed
by drawing a diagram. We attempt to give understanding to others by constructing
models of what we want them to understand. If they can construct their own models,
we take that as a sign of their understanding. In Plato, understanding is demonstrated
by successfully passing the elenchus test. A person demonstrates episteme of the
product of the relevant techne by giving a definition of its essential nature that with-
stands attack. Problem solving of the sort examined by Etchemendy and Barwise is
an attempt to understand the relations among a given set of requirements. Finding a
solution is a matter of figuring out how to make them all compossible. So more than
one solution is permitted. Any solution that meets the specifications is successful.

Once we bring up success, we implicitly bring up failure, and that means there is a
danger of skepticism. Skepticism about understanding is as real as skepticism about
truth. Skepticism threatens whenever there is a goal that is such that we cannot tell
for sure whether we have attained it even after we think we have done so. It is a threat
246 Linda Zagzebski

to our motivation because motivation to reach X requires both the belief that reach-
ing X is possible and some way of telling how well we have done after we have made
the attempt. So skepticism about our ability to get knowledge or true beliefs is not
unique. It arises for understanding because success in gaining understanding may be
illusory. It may seem to me that I clearly understand something even when I do not.
It might even appear to me that I have demonstrated my understanding by passing
the elenchus test or constructing a model or producing a solution to a problem even
when I have not. But unlike truth skepticism, understanding skepticism has never
had a significant hold on the philosophical imagination. That is because the test for
success is largely within the practice, the techne, itself. Reliably carrying out the goals
of a techne can be verified within the techne. One's understanding of an art work
can be proven by successfully giving that understanding to others by teaching it in
an art history class, and success in teaching is defined within the practice of teach-
ing. Success in problem solving is proven by the workability of the solution produced.
Again, success is defined within the confines of the techne that gives rise to the prob-
lem to be solved.
In each of these cases, of course, the test for success is not infallible. It is still
possible that failure has occurred. But skepticism about infallibility is not the most
serious form of skepticism. As we have seen, justification is associated with the re-
sponse to skepticism because it is the state we want to defend our right to be certain
we believe the truth, but the justification test was never an infallible test for certainty.
In recent epistemology, foundationalism and coherentism have been treated as alter-
native ways of answering the challenge of truth skepticism by describing the condi-
tions for having a justified belief structure. But there are many ways fallibility creeps
into both kinds of structure, and certainty is never guaranteed. There always has been
considerable slippage between justification and certainty. So the fact that there is also
slippage between explanation and understanding is not a special problem. Skepti-
cism at the level of infallibility arises within any epistemological framework. The
most threatening form of skepticism occurs when it takes away the motive to even
try to reach an epistemic goal—at the level of motivation. At this level, understand-
ing skepticism is less serious than truth skepticism. That is because, as I have ar-
gued, success in understanding is demonstrated within a practice.
Understanding has internalist conditions for success, whereas knowledge does not.
Even when knowledge is defined as true justified belief and justification is construed
internalistically, the truth condition for knowledge makes it fundamentally a con-
cept whose application cannot be demonstrated from the inside.17 Understanding, in
contrast, not only has internally accessible criteria, but it is a state that is constituted
by a type of conscious transparency. It may be possible to know without knowing
that one knows, but it is impossible to understand without understanding that one
understands. To repeat, this does not eliminate every form of skepticism. Skepticism
can appear at the second-order level. Nonetheless, there is less reason to doubt my
understanding than there is to doubt other internally accessible states, such as the
justifiability of my beliefs even on strong internalist theories of justification. That is
because for most of my beliefs, the belief is justified only if a long string of other
beliefs are justified, and even if I have cognitive access to the entire string of beliefs,
I cannot cognitively access them simultaneously and probably cannot consciously
 Recovering Understanding 247

go through them one by one. In contrast, understanding is a state in which I am di-

rectly aware of the object of my understanding, and conscious transparency is a cri-
terion for understanding. Those beleagured by skeptical doubts therefore can be more
confident of the trustworthiness of putative understanding states than virtually any
other epistemic state. Ironically, then, even though understanding dominated episte-
mology in askeptical periods, there is good reason for the skeptic to recover it as well.

VI

The recent interest in virtue epistemology makes the recovery of understanding more
likely, but it is difficult to reclaim it within the other theories that have the most in-
fluence. Epistemology is dominated by the information model of knowledge. This is
even true of many theories that are commonly called forms of virtue epistemology,
such as process and faculty reliabilism. Almost all contemporary theories agree that
to know is to believe a true proposition plus something else. A true proposition is
treated as a piece of information that can be passed from one person to another. One
of the critical issues in this model is how the original informant obtains the informa-
tion. Knowledge is akin to a physical object that is discovered in some basic way by
one person and is then passed from person to person through the epistemic commu-
nity. There is little attention in the contemporary literature to this part of the theory,
since almost all of the competing theories agree with that part of it. The disagree-
ment arises over the "plus something else" component of the definition. In evidentialist
theories, a true belief is an instance of knowledge, just in case it is based on the ap-
propriate evidence.'8 Defeasibility theories maintain that to be knowledge, a true belief
must also be justified in certain counterfactual circumstances—if certain other pieces
of information (defeaters) were also known.19 Process and faculty reliabilism and
the proper function theory maintain that it is not a true belief's justificatory relations
that make it knowledge, but the epistemic process or faculty by which it was acquired.
That process or faculty must be a reliable truth-producer, or alternatively, a properly
functioning faculty.20 The earlier causal theory maintains that to be knowledge, a
true belief must be properly caused by the fact the belief is about.21
All of these theories focus on prepositional objects, and none of them can be reinter-
preted as theories about valuable epistemic states that have a nonpropositional object.
That is, in each case, the "something else" that is a component of knowledge in addition
to true belief is something about the proposition believed or the state of believing a
proposition. The theory does not make sense when applied to anything other than a prepo-
sitional object, and it has nothing to say about epistemic states that do not have prepo-
sitional objects. Consequently, they have nothing to say about understanding.
Epistemologists presumably think that their own theories have positive epistemic
value, yet these theories generally do not meet the criteria for the epistemic concepts
they address—justification, warrant, or knowledge. In other words, most theories of
justification are such that the theory itself does not pass its own criteria for justifica-
tion, and similarly for warrant and knowledge. This is not an objection to these theo-
ries because most of them are couched in terms that make it clear the theory is not
intended to be self-referential. The theory aims at giving an account of justification
248 Linda Zagzebski

or warrant for beliefs of a certain kind, often empirically based beliefs, and the ac-
count itself is not a belief in the category it addresses. Even so, epistemologists must
think that their own theories are better than competing ones, which means that they
must think, at least implicitly, that there are valuable epistemic states other than the
ones they address in their theories. I propose that understanding is one of those goals
and that most epistemologists tacitly aim at achieving understanding through the
theories they advance. Since understanding is an epistemic state they implicitly value,
it would be a good idea to investigate it.
Virtue theories make the "something else" in the definition of knowledge a prop-
erty of the knower.22 Since theories of this kind identify a property of the knower as
the value-conferring property of epistemic states, they not only have the advantage
of not being committed to the view that justification, warrant, and knowledge have
propositional objects, but they have the additional advantage of making a recovery
of the investigation of understanding much easier to do. Valuable epistemic proper-
ties of agents produce valuable epistemic states of agents. The states produced need
not be limited to justification or knowledge. Some of them may not have proposi-
tional objects. They may be states of understanding.
In Virtues of the Mind, I propose two definitions of knowledge that differ only in
the way the object of knowledge is identified. In both definitions, knowledge is a state
that is the result of acts of intellectual virtue, where I define "act of intellectual virtue"
in a way that is intended to capture the idea of an intellectual act that is good in every
respect.23 The definition of an act of intellectual virtue is unimportant for my purposes
here. My point is that I give two definitions, and they differ only in the way the object
of knowledge is specified. That is because I do not address the issue of the object of
knowledge in that book except in passing. The first definition is neutral on the issue of
whether knowledge has a propositional object. To maintain neutrality I propose that
knowledge is cognitive contact with reality plus something else. I take it that everyone
can agree on that. The issue is what that entails and what else has to be added to it. I
have proposed that the something else is that it is the result of an act of intellectual
virtue, but, as I say, that is not important for the topic of this essay. The second defini-
tion follows the contemporary convention of defining knowledge as true belief plus
something else. That definition presupposes that the object of knowledge is a proposi-
tion. It is important to notice that this definition of knowledge can remain neutral about
the object. In contrast, most other definitions do not make sense unless the object is a
proposition. Since understanding has a nonpropositional object, it is easier to connect
it with knowledge in the kind of virtue theory I propose than in those that define knowl-
edge by looking at properties of propositional beliefs or the processes that produce them.
I am not going to propose a definition of understanding in this paper that is as spe-
cific as my definition of knowledge, but I think that a definition can be generated by
looking at those intellectual virtues that aim not at truth, but at understanding. The vir-
tues I have previously examined were in the former category. An investigation into the
latter would be very interesting. I suspect that understanding arises from special and
unanalyzed, even unrecognized, virtues. It follows from what I have said that we edu-
cate people in these virtues when we teach them a techne, but what we do, exactly, is
very hard to pinpoint. I think there is a form of teaching that can produce these virtues
so they are not simply natural talents—but even their names are elusive.
 Recovering Understanding 249

There is a difference between the kind of understanding a person has who acquires
it from a teacher and the kind that a person has who has figured it out for herself.
Good teachers learn how to give their students understanding of difficult subject matter
by the use of diagrams, vivid examples, and explanations of the way the new subject
matter connects to things the students already understand. Understanding can be
taught, like knowledge; and like knowledge, there is probably a qualitative differ-
ence between the state one gets from another and the state one gets on one's own.
Recovering understanding as an object of epistemological investigation should in-
clude an investigation of these differences.
What about wisdom? Is there any hope of recovering that also? I have claimed
that the neglect of understanding is associated with the fixation on certainty and jus-
tification and the accompanying focus on the prepositional belief, a complex of in-
terests associated with pessimism about the possibility of knowledge. But unlike the
neglect of understanding, the neglect of wisdom is probably the result of another kind
of pessimism as well—pessimism about the concept of the good life, the life a wise
person understands. This pessimism is primarily an ethical one, and leaving aside
radical skepticism will not be sufficient to make wisdom the object of inquiry. The
task of recovering wisdom, then, will be more difficult than the task of recovering
understanding, but I believe the latter is a necessary condition for the former.

VII

The epistemologist asks what knowledge is and how it can be acquired. The skeptic
suspects that knowledge as the epistemologist defines it cannot be acquired at all.
The two have been locked in a battle that has lasted many centuries. Paul Woodruff
remarks in response to this situation: "Which came first, the sceptic or the episte-
mologist? The answer is, 'Neither: Plato came first'" (p. 61). The philosophical in-
quiry into understanding came before the philosophical inquiry into prepositional
knowledge and doubts about the latter. Epistemology should not only make the in-
vestigation of understanding one of its aims, but if epistemology is itself a techne, it
should aim to understand how understanding relates to knowledge and the other
objects of epistemological study. And if Plato is also right that the ultimate object of
a techne is some good and all the forms of good are related, we can deepen our under-
standing of understanding and other epistemic goods by inquiry into the good itself.
With luck, that might give us a glimpse of what wisdom is.

Notes
1. The reason I say this is not the only possible response is that it is still possible that there
is progress in philosophy, and the lack of concern for skepticism in some previous eras might
just be a mistake destroyed once and for all by Descartes. I believe this is Richard Foley's view.
2. Roger Florka has argued that even though Descartes was interested in certainty, his
concerns were more metaphysical than epistemological. Florka (1997) addresses Descartes's
view of reasoning according to which the reasoning faculty attempts to follow the metaphysical
structure of the universe. He claims that what Descartes wanted to achieve in gaining cer-
tainty was close to what I mean by understanding.
250 Linda Zagzebski

3. See, especially, Kvanvig (1992), 181-2, and Code (1987) for the complaint that con-
temporary epistemology is too atomistic and insufficiently social.
4. See also Everson (1990), 4-5.
5. The distinction between expert and nonexpert knowledge is important for the consis-
tency of Socrates's position since, according to Woodruff (1990), it is only expert knowl-
edge that Socrates says he lacks.
6. Taylor (1990), 116.
7. See the Routledge Encyclopedia of Philosophy, edited by Edward Craig and Luciano
Floridi, Oxford: Routledge 1998, and the Companion to Epistemology, edited by Jonathan
Dancy and Ernest Sosa, Oxford: Blackwell 1992. Even when the term "understanding" is
used in the title of a work on epistemology, that often means nothing when it comes to an
account of the nature of understanding. I am pleased to find that one recent book on epis-
temology does not neglect understanding and is even concerned to reclaim it. See Elgin
(1996).
8. Note that Plantinga's use of the word "warrant" is a replacement for one of the mean-
ings for "justification" that he says he has identified (1990). Large sections of this essay ap-
pear in (1993a).
9. I suspect that this happened for a time with the concept of virtue and the concepts of
some of the individual virtues.
10. I am using the word "fact" as a placeholder, since we have not yet identified the ob-
ject of episleme.
11. Etchemendy and Barwise have a number of papers on this. For a good overview and
bibliography, see Hammer (1995).
12. See Sosa (1980) for these analogies.
13. This is also the moral of Lewis Carroll's famous paper, "What the Tortoise Said to
Achilles," (1895). The tortoise keeps trying to turn a logical inference of the form of modus
ponens into another premise. The effort leads to an infinite regress.
14. This way of looking at it forces us to make a decision about an issue I left up in the air
earlier. There I wanted to be neutral on the question of whether understanding gives us a grasp
of nonpropositional structures that are actually there in reality, or whether it gives us a way
of representing reality nonpropositionally, where there is no commitment to the idea that the
structure of reality and the structure of the theory are isomorphic. A theory seems to give us
understanding in the latter sense.
15. This point raises deep metaphysical questions that I cannot address here. For those
who disagree with me, I can only say that nothing of importance in this paper hinges on the
point. If it turns out that all forms of understanding are forms of knowledge, that actually
strengthens my main point, which is that understanding has been neglected and ought to be
investigated.
16. van Fraassen (1991), 3.
17. A priori knowledge is an exception, of course, but most of our most interesting knowl-
edge is not a priori.
18. See, for example, Conee and Feldman (1985).
19. For the strong defeasibility theory, see Lehrer (1990) and Klein (1971), 471-82.
20. For reliabilist theories, see Goldman (1986) and Sosa (1991). Plantinga gives his theory
of warrant as proper function in (1993b).
21. See Goldman (1967).
22. In addition to the virtue theory I endorse, John Greco's agent reliabilism is a form of
virtue epistemology in which properties of the knower are the focus of the theory. Sosa's
faculty reliabilism is arguably a form of agent reliabilism also.
23. Zagzebski (1996), 264-73.

References
Alston, William. 1993. "Epistemic Desiderata." Philosophy and Phenomenological Research
53 (3): 527-51.
 Recovering Understanding 251

Burnyeat, Miles. 1980. "Aristotle on Understanding Knowledge." In Aristotle on Science:

The Posterior Analytics, edited by E. Berti. Padua, Italy and N.Y.: Editrice Antenoire.
Carroll, Lewis. 1895. "What the Tortoise Said to Achilles." Mind 4.
Chisholm, Roderick. 1966. Theory of Knowledge. Englewood Cliffs, N.J.: Prentice Hall.
Code, Lorraine. 1987. Epistemic Responsibility. Hanover, N.H.: University Press of New
England for Brown University Press.
Conee, Earl, and Richard Feldman. 1985. "Evidentialism." Philosophical. Studies (48): 15-
34.
Elgin, Catherine. 1996. Considered Judgment. Princeton, N.J.: Princeton University Press.
Everson, Stephen, ed. 1990. Epistemology. Vol. 1 of Companions to Ancient Thought. Cam-
bridge: Cambridge University Press.
Fine, Gail. 1990. "Knowledge and Belief" in Republic v-vii. In Epistemology. Vol. 1 of Com-
panions to Ancient Thought, edited by Stephen Everson.
Florka, Roger. 1997. Descartes' Metaphysical Reasoning. Los Angeles: University of Cali-
fornia Press.
Fraassen, Bas van. 1991. Quantum Mechanics: An Empiricist View. Oxford: Clarendon.
Franklin, R. L. 1981. "Knowledge, Belief and Understanding." The Philosophical Quarterly
31 (124): 193-208.
Goldman, Alvin I. 1967. "A Causal Theory of Knowing." Journal of Philosophy 64 (12):
357-72.
. 1986. Epistemology and Cognition. Cambridge: Harvard University Press.
Greco, John. Forthcoming. "Agent Reliabilism." In Philosophical Perspectives, edited by
James Tomberlin.
Guerriere, Daniel. 1975. "The Aristotelian Conception ofEpisteme." The Thomist 39 (2): 341-
48.
Hammer, Eric M. 1995. Logic and Visual Information. CLSI.
Hankinson, R. J. 1995. The Sceptics (The Arguments of the Philosophers series). London:
Routledge.
Klein, Peter. 1971. "A Proposed Definition of Prepositional Knowledge." Journal of Phi-
losophy 67 (16): 471-82.
Kvanvig, Jonathan. 1992. The Intellectual Virtues and the Life of the Mind. Lanham, Md.:
Rowman and Littlefield.
Lehrer, Keith. 1990. Theory of Knowledge. Boulder, Colo: Westview.
Moravscik, Julius. 1979. "Understanding and Knowledge in Plato's Philosophy." Neue Hefte
fur Philosophie (15/16): 53-69.
. 1992. Plato and Platonism. Oxford, U.K.: Blackwell.
Plantinga, Alvin. 1990. "Justification in the Twentieth Century." Philosophy and Phenom-
enological Research, 50 suppl (Fall): 45-71.
. 1993a. Warrant: The Current Debate. New York: Oxford University Press.
. 1993b. Warrant and Proper Function. New York: Oxford University Press.
Rorty, Richard. 1979. Philosophy and the Mirror of Nature. Princeton, N.J.: Princeton Uni-
versity Press.
Sosa, Ernest. 1980. "The Raft and the Pyramid." In Midwest Studies in Philosophy, Vol. 5 of
Studies in Epistemology.
. 1991. Knowledge in Perspective. Cambridge: Cambridge University Press.
Sosa, Ernest, and Jonathan Dancy. 1992. Companion >o Epistemology. Oxford, U.K.: Blackwell.
Taylor, C. C. W. 1990. "Aristotle's Epistemology." In Everson 1990: 116-42.
Williams, Michael. 1991. Unnatural Doubts. Cambridge: Blackwell.
Woodruff, Paul. 1990. "Plato's Early Theory of Knowledge." In Everson 1990: 60-84.
Zagzebski, Linda. 1996. Virtues of the Mind: An Inquiry into the Nature of Virtue and the
Ethical Foundations of Knowledge. Cambridge: Cambridge University Press.
This page intentionally left blank
 Index
circularity, 187-201, 214-17
and the avoidance strategy, 198
common sense philosophy, 204—18
as advocated by Chisholm and Moore, 205
and the appeal to intuitions, 213-17
as criticized by Hare, 206-7, 210
as criticized by Kant, 206
and the crystal ball problem, 215-16
and its features, 204—6
and indefeasibility, 205
and supervenience, 210-12
death-of-epistemology, 235-37
deciding to believe, 63-76
and the aim of truth, 72
and Alston's objection to its possibility, 74
and belief as a dispositional state, 67-70
and Cohen's account of belief, 69
and counting on its being the case that p,
65-66, 71
and epistemic irrationality, 73
and interested vs. disinterested reasons
for believing, 70-71
as a phenomenon we are familiar with,
70
as a rare phenomenon, 70
253
and staking something on its being the
case that p, 65
and Williams's objections to its
possibility, 71-74
deontology, epistemic, 12-16, 77-92, 104-
8, 115-18, 135-39, 228-33
and contractual obligations, 86
and Locke's evidentialism, 00
and the objection from doxastic
involuntarism, 78-79
and paradigm vs. responsibility
obligations, 87
and role oughts, 87-89
Descartes, 191-92, 196-97
doxastic voluntarism vs. involuntarism, 5,
12-16, 41-43, 77-92, 93-111, 118,
228
Alston's view on, 5, 78, 80-81
and the analogy/disanalogy between
belief and action, 12-16, 93, 103
and behavioral vs. genetic voluntarism,
94
and compatibilism, 12-16
and the conceptual possibility vs.
impossibility of voluntary belief, 71—
74, 79-80
254 Index
doxastic voluntarism vs. involuntarism
(continued)
and the contingent possibility vs. impossi-
bility of voluntary belief, 80-86
and control over actions, 12-16
and deciding to believe, 63-76
and determinism, 12-16
and the difference between coming to
believe vs. causing oneself to believe,
101
and epistemic evaluation, 77-92
and intentions as essential to voluntary
control, 85
and libertarianistn, 14-16
and perceptual beliefs, 13
Plantinga's view on, 78
and practical reasons, 100-105
and H. H. Price, 15
and role oughts, 87-89
and states vs. events of acceptance, 96
Steup's view on, 12-16, 41-43, 83-86
and taking a stance, 96
and the way belief and desire differ with
regard to fit to the world, 98
and withholding assent, 97
duty, epistemic, 12-16, 36-43, 86-90
as different from moral duty, 41
and Feldman's distinction between
subjective and objective justification,
38-39
and Goldman's distinction between
strong and weak justification, 37-39
and Kant's grocer example, 41
and Lehrer's example of Mr. Raco, 40-
41
and objective justification, 36
and Plantinga's church bell example, 43-
44, 46
and Plantinga's red London bus example,
44-46
and the possibility of knowledge of
animals and young children, 46
and the problem of doxastic
involuntarism, 41-43
and its relevance to knowledge, 43
duty, moral, 34-36
and Kant's grocer example, 35
and subjective vs. objective moral duty,
35
ethics of belief, the, 3, 21-33, 104-8
and Chisholm, 22, 23, 29-30
and W. K. Clifford, 26-29, 106
and cognitive inadequacy, 24
and the correlation thesis, 21, 23-24
and William James, 27-29
and C. I. Lewis, 30
and morally culpable ignorance, 25
and the overlap thesis, 21, 25-26
and the special-case thesis, 21-23
epistemization, 134, 146
epistemology, 128-129
and commonsensism, 204-18
and its death, 235-37
and evaluation, 22
as proper or broad, 173
and value monism, 170-203
evidentialism, 135-39
Foley rationality, 173
internalism, epistemic 115-33, 228-33,
246
and animal vs. human knowledge, 146
and the accessibility constraint, 135
and the availability problem, 124, 143-
44
and Chisholm's constraint on justifiers,
119
and the core dilemma, 127, 135-38
and the deontological conception of
justification, 115-18, 135-39
and epistemic principles, 126, 144-46
and the evidential states constraint,
137
and evidentialism, 134-39
and the Gettier problem, 134, 140-41
and the knowability constrain on
justifiers, 117
and logical and probabilistic relations,
122-26, 141-42
and the methodology of epistemology,
128-29
and the problem of the doxastic decision
interval, 123, 142-43
and the problem of forgotten evidence,
121, 139-41
and the problem of stored beliefs, 119,
139

as strong, 119
as weak, 121
Gettier problem, the, 171, 177
justification, epistemic
as a causal means to reach the truth-goal,
160-61
as conceived of by BonJour, 172
as a constitutive means to reach the
truth-goal, 161
first-order and second order, 138-39
as a means to truth, 172, 178-79
justifiers, 116, 135
knowledge, 43-47, 134-35, 146, 153-55,
174, 177, 187-203, 221-34, 237-41
as better than mere true belief, 174
and the distinction between animal and
reflective knowledge, 199
as an organic unity, 180-81
as related to justification, 177-82
as requiring epistemic duty fulfillment, 43
as requiring external aptness, 201
knowledge of people, 221-34
and analogical inference, 222
as compared with knowledge of objects,
221
and epistemic deontology, 228-33
and the knowing-self, 227-28
and the method of care, 225-27, 229
and simulation theory, 222-25
Moore's proof of the external world, 187-
91, 197
and the problem of the desert dullard,
188
moral wrongness, 171-72
normativity of epistemic justification, the,
49-60
and attacks on externalism, 49
and cognitivism vs. non-cognitivism, 52,
53
and goals, 54-56
and moral judgments, 51-53
and praise and criticism, 56-58
and rules, 53-54
and value terms, 50-51
Index 255
reflective knowledge, 187-203
and circular externalism, 192-93
and the distinction between cognitio and
scientia, 199
and epistemic circularity, 197-201
and the principle of exclusion, 188, 198
reliabilism, 136, 164-66
responsibility, epistemic. See deontology,
epistemic, and duty, epistemic
skepticism, 187-90, 236-37, 244-45, 249
truth as the epistemic goal, 72, 151-69
and the definition of knowledge, 153-54
and the desire to have true beliefs, 155-
58, 160
as diachronic, 166
and the externalist approach, 164-66
and the goal of not believing what's
false, 158
and human psychology, 155-156
and the internalist approach, 163-64
and justification as a causal means of
reaching it, 160-61
and justification as a constitutive means
of reaching it, 61
as opposed to the knowledge-goal, 153-
54
and the option of abandoning it, 162
and the option of a double-goal view,
162
and the reductio of justification to truth,
161
and its relation to epistemic justification,
160-66
and reliabilism, 164-66
and restricting the truth-goal to
propositions one considers, 159
and restricting the truth-goal to important
propositions, 159
as subjunctive, 166
as synchronic, 161
understanding, 235-51
and Aristotle's epistemology, 238-39
as an epistemic goal, 245
and expert knowledge as techne, 238
and internalist conditions, 246
and Plato's epistemology, 237-38, 240
256 Index
understanding (continued)
and its relationship to knowledge, 243
and visualization as a means of discovery
and problem solving, 241
as the state of comprehension of
nonpropositional structures of reality,
242
utilitarianism, 171
value monism vs. pluralism in
epistemology, 170-83
and the analysis of knowledge, 177-79
and the conception of justification as an
instrumental value, 172, 178
virtue epistemology, 193-97, 247-49
voluntary belief. See doxastic voluntarism
and involuntarism