Landau Metaethics 10

OXFORD STUDIES IN METAETHICS
Oxford Studies
in Metaethics
Volume 10
Edited by
RUSS SHAFER-L ANDAU
1
1
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© the several contributors 2015
The moral rights of the authors‌have been asserted
First Edition published in 2015
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2015936890
ISBN 978–0–19–873869–5 (hbk.)
ISBN 978–0–19–873870–1 (pbk.)
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Contents
List of Contributors vi
Introduction vii
1 Cognitivism about Moral Judgement 1

Alison Hills
2 Acting for the Right Reasons, Abilities, and Obligation 26
Errol Lord
3 The Irrelevance of Moral Uncertainty 53
Elizabeth Harman
4 Justification and Explanation in Mathematics and Morality 80
Justin Clarke-Doane
5 Deliberative Indispensability and Epistemic Justification 104
Tristram McPherson and David Plunkett
6 Rationality and Moral Authority 134
David Copp
7 Disagreement, Correctness, and the Evidence for Metaethical
Absolutism 160
Gunnar Björnsson
8 Grounding the Autonomy of Ethics 188
Barry Maguire
9 Irreducibly Normative Properties 216
Chris Heathwood
10 How to Be a Moral Platonist 245
Knut Olav Skarsaune
11 Explaining the Quasi-Real 273
Jamie Dreier
Index 299
List of Contributors
Gunnar Björnsson is Professor of Philosophy at the Department of

Historical, Philosophical and Religious Studies, Umeå University, and
Research Fellow in Philosophy at the University of Gothenburg
Justin Clarke-Doane is Honorary Research Associate at Monash
University, Honorary Research Fellow at the University of Birmingham,
and Assistant Professor of Philosophy, Columbia University
David Copp is Distinguished Professor of Philosophy, University of
California, Davis
Jamie Dreier is Professor of Philosophy, Brown University
Elizabeth Harman is Associate Professor of Philosophy and Human
Values, Princeton University
Chris Heathwood is Associate Professor of Philosophy, University of
Colorado, Boulder
Alison Hills is Tutorial Fellow and University Lecturer, University
of Oxford
Errol Lord is Assistant Professor of Philosophy, University of Pennsylvania
Barry Maguire is Research Assistant Professor of Philosophy and Core
Faculty member in Politics, Philosophy, and Economics, University of
North Carolina-Chapel Hill
Tristram McPherson is Associate Professor of Philosophy, Ohio State
University
David Plunkett is Assistant Professor of Philosophy, Dartmouth College
Knut Olav Skarsaune is Postdoctoral Research Fellow at the University
of Oslo
Introduction
Russ Shafer-Landau
Oxford Studies in Metaethics marks its tenth anniversary with this volume.
The entries in this edition reveal just how broad a field metaethics has
become, and how deep is its roster of talent.
We begin with a chapter by Alison Hills that develops her innovative
view about the kinds of mental states that play a role in moral judgment.
Traditional cognitivist views regard moral judgments as a type of belief.
Hills is sympathetic with cognitivism, but introduces us to a puzzle that
requires refining our conception of the sort of belief that plays a role in
moral judgment. The puzzle emerges once we accept the claim that beliefs
ought to be responsive to evidence, accept also that testimony is a basic kind
of evidence, and yet reject the idea that our moral beliefs are justified if they
are formed solely on the basis of testimony. We can resolve this tension by
invoking a special class of beliefs—those that Hills has termed uliefs. Unlike
other forms of belief that aim at knowledge, uliefs aim at understanding.
Forming a moral belief solely on the basis of testimony cannot give an agent
understanding. Moral judgments, Hills believes, are aimed at expressing
and aiding an agent’s understanding; they are uliefs, rather than beliefs.
Much of her provocative essay is devoted to developing her conception of
uliefs and their role in moral psychology.
Errol Lord’s chapter on acting for the right reasons and moral obligation
was the winner of the inaugural Marc Sanders Prize in Metaethics. The prize
is awarded to the best unpublished paper in the area by a scholar within
fifteen years of having received his or her Ph.D. Lord’s paper is focused on
objectivism about obligation—the thesis that obligations are determined by
all of the normatively relevant facts. Critics of objectivism, whom Lord calls
perspectivalists, hold that only facts within one’s perspective can determine
what we are obligated to do. Lord here argues for the perspectivalist view,
on the grounds that our obligations depend on the normative reasons we
possess. The central argument for this view is anchored in the thought that
our obligations have to be action-guiding in a certain sense—we have to be
able to act for the reasons that obligate us. Lord argues that we have this
ability only if we possess those reasons. Because some normative facts may
be inaccessible to agents, or they may fail for other reasons to possess such
viii Introduction
facts, these facts cannot, on Lord’s view, contribute to the determination

of which obligations apply to us. If this line of reasoning is correct, then
perspectivalism is true.
Elizabeth Harman next offers her reflections on a topic that has received
a lot of attention only recently—that of deciding how to act when one
is uncertain about the normative status of one’s options. She invites us
to imagine a case in which you believe that you are morally required to φ,
while also holding two further judgments: (i) that this requirement to φ is
not very important, and (ii) that it might be deeply morally wrong to φ.
A natural thought is that in such a case you should not φ, because φing
would be too morally risky. She rejects this thought by arguing that if φing
is in fact morally required, then you should φ, and this is so even taking
into account your moral uncertainty. The primary basis for her view is the
claim that if this natural thought were correct, then being caught in the grip
of a false moral view would be exculpatory; it isn’t; therefore, the natural
thought is mistaken, and an agent’s moral obligations are not conditioned
on the uncertainty she has about the normative status of her options.
In his chapter, “Justification and Explanation in Mathematics and
Morality,” Justin Clarke-Doane takes as his starting point a well-known and
highly influential remark of Gilbert Harman: “In explaining the observa-
tions that support a physical theory, scientists typically appeal to math-
ematical principles. On the other hand, one never seems to need to appeal
in this way to moral principles.”1 Clarke-Doane argues that both ethicists
and philosophers of mathematics have misunderstood the importance of
the contrast that Harman is pointing to. According to Clarke-Doane, these
philosophers have conflated what he calls the justificatory challenge for real-
ism about an area D—the challenge to justify our D-beliefs—with the reli-
ability challenge for D-realism—the challenge to explain the reliability of
our D-beliefs. Clarke-Doane allows that Harman’s contrast is relevant to
the first of these projects, but is skeptical that it is relevant to the second.
He argues that once we do the needed untangling, two very important
implications become clear. First, genealogical debunking arguments that
target realism are fallacious. Second, indispensability considerations cannot
answer the Benacerraf–Field challenge for mathematical realism—at best,
such considerations can justify our mathematical beliefs (or explain how
they are justified). They cannot explain how those beliefs could be reliable.
When it comes to moral realism, and metanormative realism more
broadly, Tristram McPherson and David Plunkett doubt that this best-case
scenario can be realized. Their selection carefully investigates an important
1
G. Harman, The Nature of Morality (Oxford: Oxford University Press, 1977), p. 10.
Introduction ix
effort by David Enoch to account for the justification of our moral beliefs
on the assumption of moral realism’s truth. Enoch argues that we are justi-
fied in believing in something if we must presuppose its existence in order
to vindicate the point and value of our deliberations. In other words, delib-
erative indispensability can justify ontological commitment. Enoch argues
that we are able to deliberate only if we take for granted the existence of
normative reasons, construed as a robust realist would do. McPherson and
Plunkett argue that Enoch’s proposal fails because it conflicts with a central
fact about epistemic justification: that the norms of epistemic justification
have the content that they do in part because of some positive connection
between those norms and the truth of the beliefs that these norms govern.
They then argue that the most salient alternatives to Enoch’s attempt to
defend the idea that deliberative indispensability confers epistemic justifica-
tion fail for parallel reasons. If they are correct, then deliberative indispen-
sability does not provide epistemic justification.
We turn next to the ages-old problem of the relationship between ration-
ality and moral authority. In his selection, David Copp asks us to consider
what he calls the Rationality Doctrine—the view that the normativity of
morality depends metaphysically on the existence of a link of an important
kind between morality and rationality. Copp focuses on one view about the
kind of link that is required, which he calls the Basic Linkage Thesis: neces-
sarily, an agent is morally required to φ in circumstances C only if there is
a requirement of rationality that she φ or at least be motivated to φ in cir-
cumstances C. Copp argues that this linkage thesis, which he takes to figure
in the best formulation of the Rationality Doctrine, is mistaken. If this is
correct, then either morality is not normative, or the Rationality Doctrine
is false, and the normativity of morality does not depend on the rationality
of adhering to moral demands. Copp opts for the latter view.
In our next selection, Gunnar Björnsson devotes his attention to a doc-
trine that he calls metaethical absolutism—the view that moral concepts
have non-relative satisfaction conditions that are constant across judges
and their particular beliefs, attitudes, and cultures. The prospect of genuine
moral disagreement seems to presuppose absolutism. Further, absolutism
appears to be supported by the fact that many features of paradigmatically
absolutist non-moral discourse are shared by moral discourse. Björnsson
raises doubts about such arguments, proposing independently motivated
general accounts of attributions of agreement, disagreement, correctness,
and incorrectness. These accounts provide the basis for his efforts to explain
the phenomena above in a way that is friendly to non-absolutist interpreta-
tions of moral discourse.
There is a widespread belief that morality is somehow autonomous.
Hume gave the most influential expression of this idea when claiming
x Introduction
that one cannot deduce an ought from an is. While the literature that has
descended from Hume’s remark has focused very largely on logical charac-
terizations of autonomy theses, Barry Maguire argues that this emphasis is
misplaced. According to Maguire, the important autonomy thesis is rather
a metaphysical one that maintains that ethical facts are not fully grounded
just in non-ethical facts. After explaining and defending this thesis, he also
argues for the converse thesis that all facts partly grounded in ethical facts
are themselves ethical facts. He then argues that this pair of theses can help
with debates about the plausibility of nihilism and the classification of revi-
sionary metaethical theses.
Chris Heathwood’s chapter nicely picks up where Maguire leaves off.
Heathwood focuses on the metaethical non-naturalists’ claim that norma-
tive or evaluative properties cannot be reduced to, or otherwise explained
in terms of, natural properties. This tenet has caused non-naturalists some
trouble—among other things, they have had difficulty explaining what
these irreducibly normative properties are supposed to be, usually offering
a negative characterization in terms of what they are not. Heathwood offers
a partial, positive characterization of irreducible normativity in naturalistic
terms: roughly, to attribute a normative property to something is necessar-
ily to commend or condemn (where such commendation or condemna-
tion can be understood naturalistically) that thing, due to the nature of the
property attributed. After elaborating on and defending this characteriza-
tion, Heathwood proceeds to argue that his hypothesis does some impor-
tant explanatory work: it provides for an account of the “queerness” of
normative properties, one superior to other accounts; it explains why meta-
ethical reductionism is bound to fail, in a way friendly to non-naturalism
(as opposed to non-cognitivism); and it can help deflect arguments against
non-naturalism from the “essential practicality” of normativity.
In the first of a pair of essays devoted to moral supervenience, Knut Olav
Skarsaune takes up the challenge to non-naturalist realism of explaining
both why normative properties supervene on descriptive properties, and
why this pattern is analytic. The explanation proceeds by positing a subtle
polysemy in normative predicates such as “good.” Such predicates express
slightly different senses when they are applied to particulars (e.g., Florence
Nightingale) and to kinds (such as altruism). According to Skarsaune, the
former sense (which he labels goodpar) can be defined in terms of the latter
(goodkin) as follows: x is goodpar = df. there is a kind K such that x is a token
of K, and K is goodkin. Skarsaune believes that if x and y are descriptively
exactly similar, then they are tokens of exactly the same kinds; x is a token
of a goodkin kind if and only if y is. Therefore, by definition, x is goodpar if
and only if y is. Thus the definition of “goodpar” directly entails the truth of
the moral supervenience claim.
Introduction xi
Simon Blackburn’s seminal 1971 paper on moral supervenience2 put the

topic front and center; ever since, considerations of supervenience have
been thought to plague realists (especially non-naturalists) and, to a lesser
extent, offer support to non-cognitivists. In the chapter that concludes this
volume, Jamie Dreier revisits this history and raises some doubts about the
extent to which moral supervenience should both cause realists to worry
and non-cognitivists to rejoice. After helpfully working to distinguish
quasi-realism from robust realism (thereby making progress on solving a
problem that Dreier himself introduced to the literature in an influential
paper a decade ago),3 Dreier argues that quasi-realists have failed to provide
any plausible explanation for the phenomenon of moral supervenience. He
proceeds to sketch some possible ways forward on behalf of quasi-realists,
and concludes with an illuminating synopsis of the state of the dialectic on
this topic.
The essays collected here are very substantially revised versions of presen-
tations offered at the 10th Annual Wisconsin Metaethics Workshop, held
in Madison in the fall of 2013. The workshop is supported each year by a
grant from the Dean of the College of Letters and Sciences at the University
of Wisconsin-Madison, and I am grateful to Deans Gary Sandefur and Karl
Scholz, as well as Associate Dean Sue Zaeske, for their confidence in the
merits of the workshop and their willingness to continue to fund it. I’d
like to thank the members of the Workshop’s program committee—Terence
Cuneo, David Enoch, Jimmy Lenman, Sarah McGrath, Sharon Street, and
Mark van Roojen—for their efforts in helping to select these entries from
over 100 that were submitted for consideration. As usual, my terrific edi-
tor, Peter Momtchiloff, managed to recruit two outstanding reviewers who
provided dozens of pages of very careful, supremely sharp and constructive
criticisms to our authors. I wish I could thank these reviewers here, but
they prefer anonymity to their well-deserved public recognition. And as
you’ll see from the materials within, the OUP production team continues
to maintain its impeccable standards.
2
S. Blackburn, “Moral Realism,” in John Casey (ed.), Morality and Moral
Reasoning: Five Essays in Ethics (London: Methuen, 1971), 101–24.
3
J. Dreier, “Meta-Ethics and the Problem of Creeping Minimalism,” Philosophical
Perspectives 18 (2004): 23–44.
1
Cognitivism about Moral Judgement
Alison Hills
When someone tells me that what I am proposing to do is immoral,

I do not react by asking for his credentials but for his reasons.
(Coady 1992: 71)
1.1 COGNITIVISM AND NON-COGNITIVISM

ABOUT MORAL JUDGEMENT
What is it to make a moral judgement?1 There are two standard views, cog-
nitivist and non-cognitivist, plus hybrid options according to which moral
judgements have cognitivist and non-cognitivist components.2 In this con-
text, cognitivism is typically defined as follows:
Cognitivism: moral judgements are beliefs.
By contrast, there are two standard definitions of non-cognitivism. The first
is negative, characterizing non-cognitivism as simply the denial of cognitivism:
Non-cognitivism 1: moral judgements are not beliefs.
1
Questions concerning the nature of moral judgement can relate to the proper inter-
pretation of moral sentences (a semantic thesis) or the nature of the state of mind which
is typically expressed by moral sentences (a psychological thesis). I will be discussing
the latter in this chapter (see Van Roojen (2013) for further discussion), though the
debates are related. Many authors mentioned in this chapter primarily discuss the former
(though their views have implications for the latter question too).
2
See Van Roojen (2013). Non-cognitivists include Blackburn (1998) and Gibbard
(2003). Cognitivists include Brink (1989). Hybrid theorists include Copp (2001) and
Ridge (2006, 2007); for discussion and assessment, see Schroeder (2009).
2 Alison Hills
The second makes a positive claim:

Non-cognitivism 2: moral judgements are ‘desire-like’ mental states
(e.g., desires, pro-attitudes, commitments, emotions, intentions, etc.).
This essay is not a defence of cognitivism (or of non-cognitivism). Rather,
the aim is to get clearer about what it means for a moral judgement to be
a belief.
I begin by setting out, in section 1.2, a tension between three claims: cog-
nitivism, an account of belief, and an account of moral judgement. I think
all three claims are plausible, but I will not be defending them here. Rather,
my interest is in whether they can be reconciled. In section 1.3, I distin-
guish between broad and narrow belief. In sections 1.4 and 1.5 I give an
example of a mental state that is a broad belief but not a narrow belief, a
moral ‘ulief ’. Section 1.6 addresses some objections to the idea of a moral
ulief. In the final section I set out new definitions of cognitivism and
non-cognitivism and draw out some further implications of the argument.
1.2 BELIEF, MORAL JUDGEMENT, AND

TRUSTING TESTIMONY
Cognitivism is an appealing theory of moral judgement. Like beliefs,

moral judgements appear to describe the world. The semantic expres-
sion of moral judgements, like beliefs, is in declarative sentences that are
truth-evaluable.
Cognitivism faces difficulties, naturally, when moral judgements do not
appear to behave like beliefs. One such is very well known. It is widely
believed that moral judgements are ‘practical’: when you make a moral
judgement, ceteris paribus, you are motivated to some degree to act on that
judgement. But beliefs have no such connection to motivation.
This has a good claim to be the most discussed argument in metaethics.
I am not convinced by it, but I will not be adding to the already vast litera-
ture. Instead, I am going to discuss a different problem for cognitivism, a
tension between it and two further claims, one about the nature of beliefs,
and one about moral judgement.
The claim about beliefs is: beliefs should be responsive to evidence. This
might be a descriptive claim—that beliefs typically and characteristically do
respond to evidence—but I will mostly discuss the normative claim—that
they should.
Of course, beliefs are not always formed as they should be; we are care-
less, or self-deceived, or victims of wishful thinking. But if it’s brought to
your attention that you have formed a belief that is not properly supported
Cognitivism about Moral Judgement 3
by evidence, you should suspend judgement or form a new belief that fits
better with your evidence.
One of our most important sources of evidence about the world is tes-
timony: under the right circumstances, testimony that p is good evidence
that p.
For example, suppose you are having a discussion about tax, and some-
one says ‘an increase in sales tax of 1% will raise X millions for the govern-
ment’. You might check her credentials: is she really knowledgeable about
tax? Has she thought about the matter carefully? And if her credentials
are fine, you should trust what she says, because by doing so you can gain
knowledge.
This is an example of what I call ‘pure’ testimony, that is, testimony that
p without any indication of the reasons why p might be true.
Of course, things are not always this straightforward. First, you should
not trust testimony when the speaker is unreliable, trying to deceive, or has
simply made a mistake. Here, testimony is not good evidence, so beliefs
should not be responsive to it.
Secondly, testimony is only one piece of evidence.
Suppose that a speaker says that p. But you have lots of evidence that
not-p. Her evidence is outweighed.
Suppose she says that p, but you have lots of other, very strong evidence
for p. Her evidence is swamped. Though you do form the belief that p, the
belief is responsive to other, stronger evidence and is responsive only a little,
if at all, to her testimony.
But even in these cases, your beliefs are responsive to pure testimony
insofar as you treat it as evidence in its own right that is assessed along with
the other evidence that you have regarding p, and which plays a role in
determining whether or not you believe that p.
In short: beliefs should be (and are) characteristically responsive to testi-
mony that p, provided that it is regarded as good evidence that p.
The problem for cognitivism arises when this claim about belief is com-
bined with a claim about moral judgements, namely that moral judgements
should not be responsive to pure moral testimony.
Suppose that in your tax conversation, someone says: ‘an increase in sales
tax is morally wrong’. Should you check her moral credentials, make sure
that she was knowledgeable about tax, and then trust her moral judgement?
Some people say: yes you should. Doing so will get you moral knowledge.3
3
There are of course religious and moral traditions that do endorse relying on moral
testimony. I have argued at length elsewhere that these traditions are missing something
of moral significance by accepting moral testimony (Hills 2009).
4 Alison Hills
If this is right, moral judgements are just like beliefs, in this regard at least,
and there is no problem for cognitivism.
But other people say no.4 Consider the quote from Coady with which I
began which says that you should not trust pure moral testimony, instead
you should ask for the reasons. Why is an increase in sales tax wrong? You
are looking for an explanation that you yourself can follow and accept or
reject. That is, you are not ignoring this testimony, but you are not putting
your trust in it either. You are assessing the case for the moral conclusion
yourself.5
When we first make moral judgements as children, we are strongly influ-
enced by those around us. But many people regard it as an ideal that, at
a certain point, people make up their own minds about morality with-
out relying on the authority of others. What kind of ideal is this? Moral?
Presumably. Epistemic? That isn’t so clear, and I will return to this question
later. It is striking that moral philosophers who are in deep disagreement
about ethics—including Kant and Bernard Williams—all say that trusting
moral testimony is not the best way of making moral judgements.6 Aristotle
4
See Jones (1999), Driver (2006), Hopkins (2007), Hills (2009), and McGrath
(2011).
5
Is this how we actually do make moral judgements? As far as I am aware, no psychol-
ogist has investigated directly our responses to moral testimony. The very well-known
theories of Greene and Haidt, for instance, do not discuss pure moral testimony directly
(e.g., in Greene (2007), Haidt (2001), and Haidt and Bjorkland (2008)). Kohlberg does
say that some moral judgements may be made ‘autonomously’ suggesting that they are
typically not based on pure testimony (Kohlberg (1973)). However, there is some sug-
gestive evidence in Hussar and Harris (2010). This is a study of children under the age
of 10 who are vegetarians though their families eat meat. When asked why they were
vegetarian, these children tended to cite animal welfare as the reason (rather than the
taste of meat or health considerations) and would condemn themselves for eating meat
as they would other moral transgressions. This suggests that their vegetarianism is based
on a moral judgement that eating meat is morally wrong, and that they have ignored any
moral testimony from their family that meat eating is morally acceptable (it is consist-
ent with the study that they have based their view on moral testimony from non-family
members, but also consistent with it that they have not and chosen to use their own
moral judgement instead). Though this study is not conclusive, it does suggest that some-
times children may choose to make up their own minds about moral questions rather
than trust moral testimony, even from a source like their parents whom they normally
trust. Nevertheless, in this chapter I do not make assumptions about how moral judge-
ments are actually formed, rather I am interested in how many people (including many
moral philosophers) have thought that they should be formed.
6
Williams (1995: 205). Kant’s views about moral testimony are of course closely con-
nected to his views about autonomy, expressed in all his ethical texts, for instance in the
Groundwork at 4.425–7 and especially 4.431. He specifically discusses teaching ethics
in the Metaphysics of Morals and rejects a lecturing model in which the pupil trusts her
teacher in favour of a ‘catechism’ in which the teacher asks the pupil questions in order
to draw out her understanding (6.478–84).
too suggests that a virtuous person will have good judgement (phronesis)
and should use it to decide what to do.7 I assume (and will later defend
the assumption) that it is at least not epistemically irrational to form moral
judgements in this way.
It is important to be clear about what this means, because of course,
testimony can still be important to these people when they are thinking
about morality. Their judgements should be responsive to testimony about
relevant non-moral matters.8 And these judgements should be responsive
to moral reasons and moral argument. But ceteris paribus, they should not
be responsive to pure moral testimony, that is, to a speaker simply asserting
that X is morally wrong.9
Why shouldn’t moral judgements be responsive to pure moral testimony?
One possibility is that there is no good evidence for any moral judge-
ments. Perhaps there are no moral truths; perhaps there are, but we have
no good access to them. Assessing this claim properly is beyond the scope
of this chapter. But in any case, this cannot explain why moral judgements
should be insensitive to pure moral testimony, but sensitive to other kinds
of evidence (including non-moral facts, moral arguments, thought experi-
ments, and so on).
A second possibility is that though there is some good evidence for our
moral judgements, moral testimony is not among it. Other people’s moral
judgements are unreliable, so we should not place our trust in them.
This argument might be framed in terms of responsibility for our beliefs.
When forming beliefs about very important matters—as moral questions
frequently are—the responsible thing to do is to make up your own mind,
rather than to put your trust in others.
Of course, it would not be responsible to trust unreliable testimony, and
much, perhaps most, moral testimony is unreliable. It is difficult to find
someone whose views you can trust in ethics. There are wide divergences in
opinion and sometimes people’s own interests can be at stake, so that their
judgements may be biased. This is certainly an important factor in explain-
ing why we should not trust moral testimony.
But there is no reason to think that moral testimony is always unreliable.
It is implausible that other people’s moral judgements are, quite generally,
much less reliable than one’s own. And then why not trust moral testimony
that is, and that you take to be, reliable? Surely this would actually be the
7
For instance, NE 1144a6–8, NE 1113a32–3.
8
I discuss this further in Hills (2009).
9
I discuss whether there is ever good reason to treat the opinion of others as evidence
for moral judgements in much more detail in Hills (2010), especially c hapter 10.
6 Alison Hills
most responsible way of forming moral beliefs: more responsible than using

your own judgement, when that too can go wrong.10
A third possibility is that our moral judgements actually should be sen-
sitive to evidence from pure moral testimony, but that evidence is either
outweighed or swamped by other evidence that we have. It may seem so
obviously true that it is wrong to torture innocent children that testimony
either for or against the practice can safely be ignored.
But again, it is not plausible that other evidence I have always outweighs
or swamps moral testimony. Not all moral questions have obvious answers,
and other people may well be better at making moral judgements than I am.
More importantly, when I make a moral judgement, I do not always
think: ‘this person is unreliable; I can and should ignore what she says because
I am more likely to get the answer right that way’. Instead I think: ‘this is the
sort of issue about which I should make up my own mind. I will listen to
her arguments but in the end I need to come to my own conclusion rather
than put my trust in her’.
In other words, if I can make up my own mind properly about a moral
question, I should: in these circumstances, trusting pure moral testimony is
‘the wrong kind of reason’ for forming a moral judgement, quite aside from
any concerns about reliability.
Now we can see the tension for cognitivism. Anyone who accepts this
claim about moral judgement thinks that pure testimony can be the right
kind of reason for a belief, and at the same time ‘the wrong kind of reason’
for forming a moral judgement.
An easy way out of the problem for a cognitivist would be to deny the
claim about belief (it should be responsive to evidence) or the claim about
moral judgement and moral testimony. But what if, like me, you are per-
suaded by both of them? Can you still be a cognitivist about moral judge-
ment? How can the tension between those two claims and cognitivism be
resolved?
One possible move for a cognitivist is to accept a simple form of moral
subjectivism. If moral truths were dependent on an individual’s attitudes,
there would be little point in trusting someone else’s moral judgements
because what was true for them on the basis of their attitudes might not
be true for you. Your own attitudes might be different from theirs, and you
10
Note that this is true even for an internalist about knowledge or justification, who
requires the justification for a belief to be accessible to the subject. If you have good
evidence that someone is trustworthy (e.g., she has a good track record), then even an
internalist accepts that you are justified in accepting her pure testimony, and by doing so,
you can gain knowledge.
might be the best placed to work out what they are. So it might be best to
make moral judgements using only your own judgement.
Simple subjectivism would be a way of reconciling cognitivism with
the two further claims. Unfortunately this very simple form of subjectiv-
ism leads to an extreme kind of moral relativism: moral truths are rela-
tive to each person. This is not very appealing. On the other hand, a more
sophisticated kind of subjectivism that does not lead straight to relativism
cannot explain why you should not trust moral testimony. So it does not
seem that cognitivists can solve their problem satisfactorily by appealing to
subjectivism.
There have been some other attempts to explain why we might be reluc-
tant to trust moral testimony.
(1) The importance of autonomy.
Autonomous judgements are self-given or self-made. If it were important
that moral judgements were self-given, this could explain why you should
use your own judgement rather than trusting moral testimony, because put-
ting your trust in someone else’s judgement is not to make a judgement
yourself.
(2) Motivation.
Moral judgements tend to motivate people to act. They are essentially
practical. It may be that taking a moral judgement from someone else, on
trust, tends not to motivate. Hence, there is reason to make your own moral
judgements instead.
I am not convinced that either of these explanations will ultimately be
successful (McGrath and Hopkins argue convincingly that they will not).11
The second is problematic for cognitivism because it depends on an aspect of
moral judgements that is desire-like rather than belief-like. The first requires
further explanation because it is not clear why some beliefs (but not others)
should be formed ‘autonomously’, that is, without trusting testimony.
In the remainder of the chapter, I turn to a different kind of solution.
I begin with a distinction between two types of belief.
11
With regard to autonomy, it is notoriously unclear what autonomy requires, and
whether it is essential to morality. One possibility is that you can only be bound by
judgements that you make for yourself, and you can only have moral obligations if you
are bound by them. This would imply that at least one class of moral judgements (those
about obligations) need to be self-given. But the argument is not convincing. Why can-
not recognizing an obligation on the basis of someone else’s testimony bind you too?
With regard to motivation, it is certainly not obvious that we could not be motivated by
a ‘second hand’ moral judgement. And if motivation did come with the judgement, there
seems to be no reason not to take it on trust.
8 Alison Hills
1.3 BROAD AND NARROW BELIEF
It has been acknowledged for some time that the term ‘desire’ can be used in
a broad or a narrow sense. In the broad sense, ‘desire’ means a non-cognitive
‘desire-like’ attitude, sometimes called a pro-attitude. A desire in the narrow
sense is one of those attitudes, but there are others. For instance, here is
Davidson characterizing ‘pro-attitude’ as including:
Desires, wantings, urges, promptings, and a great variety of moral views, aesthetic prin-
ciples, economic prejudices, social conventions, and public and private goals and values.
(Davidson 1980: 4)12
And here is Bernard Williams characterizing a ‘subjective motivational
set’ or S:
I have discussed S primarily in terms of desires, and this term can be used, formally,
for all elements in S. But this terminology may make one forgot that S can contain
such things as dispositions of evaluation, patterns of emotional reaction, personal
loyalties, and various projects, as they may be abstractly called, embodying commit-
ments of the agent.
(Williams 1981: 105)
Desires in the narrow sense are a subset of desires in the broad sense.13
A very similar distinction can be made between a broad and narrow sense of
belief (what I will call broad belief and narrow belief or n-belief ).14
12
Davidson appears to contrast believing with other attitudes: ‘knowing, perceiving,
noticing, remembering’ (Davidson 1980: 3). But since it is natural to think that know-
ing, perceiving, etc. all involve belief, it is natural to assume that the parenthesis distin-
guishes different kinds of belief or knowledge (e.g., from different sources) rather than
different cognitive states that are to be distinguished from belief and knowledge.
13
In response to this variety, many different non-cognitivist views have been set out
and defended. To take just a sample, it has been claimed that a moral judgement that x is
wrong is a combination of attitudes of disapproval of x and disapproval of those who fail
to share this disapproval (Blackburn 1998); a positive attitude in favour of blaming for
the action in question (Schroeder 2008); and a planning state concerning feelings of guilt
and resentment (Gibbard 2003). These theories all concern desires in the broad sense but
not desires in the narrow sense.
14
The recognition of other kinds of cognitive state (e.g., in Velleman 2000: 244–82) has
not much influenced characterizations of cognitivism. There has been some discussion about
the possibility of two kinds of moral belief, a ‘minimal’ moral belief which even expressivists
can take moral statements to express, and a more robust kind of moral belief, which they
cannot (Blackburn 1998; Sinclair 2006; see also Horgan and Timmons 2006). Minimal
belief may ‘aim’ at truth in a minimal sense, or (in the case of Horgan and Timmons) may
not be a descriptive state at all. Nevertheless, all of these discussions assume that the cognitive
mental states in question are beliefs, and so do not address the objection to cognitivism in
which I am interested (in addition they face further issues, as to whether minimal beliefs or
non-descriptive beliefs can capture all the important features of moral judgements).
A broad belief is a representational state, a propositional attitude. It has a

‘mind to world’ direction of fit. It has a characteristic functional role, being
typically formed in response to evidence, and taking part in reasoning and
in action. It may also have a characteristic phenomenology:
This what-it-is-likeness typically involves: (1) psychologically “coming down” on
some issue, in a way that (2) classifies (sometimes spontaneously) some “object”
of focus as falling under some category, where one’s classificatory coming down is
experienced (3) as involuntary, (4) as a cognitive response to some sort of consid-
eration that is experienced (perhaps peripherally in consciousness) as being a suf-
ficient reason for categorizing as one does, and (5) as a judgement that is apt for
assertion and hence is naturally expressible in public language by a sentence in the
declarative mood.
(Horgan and Timmons 2006: 263)
Broad beliefs can be divided into sub-kinds. One sub-kind is a narrow
belief. Narrow beliefs are propositional attitudes that aim at truth in a dis-
tinctive way: they aim at knowledge.15 These metaphors of ‘aiming’ are
notoriously obscure. What I mean here is: a narrow belief that p should be
responsive to evidence regarding p in such a way that in favourable circum-
stances the result will be knowledge that p.16 I am not of course suggesting
that we consciously try to respond to evidence in this way, nor that we
deliberately aim at knowledge. Most of us do not. And I am not suggesting
that beliefs are always perfectly responsive to evidence. Sometimes we fail to
recognize evidence, sometimes we do not form beliefs properly in response
to it. Moreover, there may be other influences on belief (including desire).
But nevertheless narrow beliefs should be formed in response to evidence,
I suggest, in such a way that in favourable circumstances the result will be
knowledge and when they are not, we typically regard this as a mistake
which we try to correct.
In favourable circumstances, trusting pure testimony from a reliable
source will give you knowledge. It follows that narrow beliefs that aim at
knowledge should be (and characteristically are) responsive to this kind of
evidence.
15
There is a large literature attempting to explicate the sense, if any, in which beliefs
aim at the truth. See for example Velleman (2000), Wedgwood (2002), and Owens
(2003). Owens suggests that beliefs may aim at knowledge rather than truth, as does
Williamson (2000) and Engel (2004).
16
I aim to be as neutral as possible in characterizing knowledge, belief and evidence
here. For instance, this thesis has no implications for the question of whether knowledge
or belief is a ‘conceptually prior’ notion. I am also neutral on the question of what evi-
dence is (except that I claim that testimony that p can be evidence that p) and I am not
endorsing evidentialism (i.e., I do not assume that your belief is justified iff it fits your
evidence).
10 Alison Hills
Moral judgements that should not respond to pure moral testimony do

not have all the characteristics of narrow beliefs. So anyone who accepts that
should reject a version of cognitivism of the following form:
Cognitivism (narrow): moral judgements are narrow beliefs.
But recall that n-beliefs are a sub-kind of broad beliefs. There may be other
sub-kinds: types of broad belief that are not narrow beliefs.
In the remainder of this chapter, I introduce a mental state which I call
a moral ‘ulief’. Moral uliefs are propositional attitudes which represent the
world. They are most naturally expressed in declarative sentences which are
truth-evaluable. They have the ‘direction of fit’ that they ‘aim to fit the world’.
They are therefore cognitive states (broad beliefs). But they are not narrow
beliefs: they do not aim at knowledge. The best way to think of the differ-
ence between narrow beliefs and uliefs is that as moral beliefs stand to moral
knowledge, so moral uliefs stand to what I call exercised moral understanding.
But this clearly needs explanation, and I will begin with the idea of moral
understanding.
1.4 MORAL UNDERSTANDING
By moral understanding, I mean understanding why p where p is some

moral proposition (such as, killing is morally wrong, giving money to char-
ity is the thing to do). I have introduced and discussed my conception of
moral understanding elsewhere, so I shall simply sketch briefly its main
features in this section.17
Suppose that you understand why giving money to charity is good. Then
you judge that it is good, and you grasp why. What is it to grasp why giving
money to charity is good? It involves having a set of abilities. For instance,
you can give an explanation of why it is good: because doing so helps peo-
ple who are in great need. You can give explanations and draw conclusions
about similar cases (e.g., when there are other ways of helping people, when
the charity in question is not very efficient, and so on).
In some ways, moral understanding is quite similar to moral knowledge.
As I conceive of it, it is factive (you cannot understand why p, if p is false or
if you are wrong in judging that q is why p). And it is not transparent (you
may think you understand why p when you do not, and vice versa). But
the distinctive feature of moral understanding is this grasp or set of abilities,
which goes beyond what is required to know that some action is right, and
17
This account of moral understanding is based (with a few minor revisions and clari-
fications) on that in Hills (2009, 2010) where it is discussed in much more depth.
even what is needed to know why it is right.18 If you understand why p (and
q is why p), then you judge that p and that q is why p and in the right sort
of circumstances you can successfully:
(i) follow an explanation of why p given by someone else
(ii) explain why p in your own words
(iii) draw the conclusion that p (or that probably p) from the information that q
(iv) draw the conclusion that p′ (or that probably p′) from the information
that q′ (where p′ and q′ are similar to but not identical to p and q)
(v) given the information that p, give the right explanation, q
(vi) given the information that p′, give the right explanation, q′
To understand why p, you have to have the abilities (i)–(vi) to at least
some extent.19
It is clearly possible to have knowledge that p and even knowledge why
p without understanding why p, because you lack the ‘grasp’ distinctive of
understanding. I think that it is also possible to understand why p without
knowing why p (or knowing that p). For there may be defeaters present for
knowledge, that are not defeaters for understanding.
For instance, your judgement that p could depend on your judgement
that q, and you might be lucky in correctly judging that q. For instance,
suppose that you read in an otherwise very inaccurate textbook that Pol Pot
was responsible for the deaths of millions. You correctly judge that he was
evil. You could not know that he was evil (since your judgement was made
on an unreliable basis) but you could understand why he was.20
18
There is a widespread though not universal view that understanding why p is the
same as knowing why p (Kitcher (2002), Woodward (2003: 179), Lipton (2004: 30),
and Grimm (2006) agree, Zagzebski (2001), Kvanvig (2003), and Pritchard (2005) do
not). My view of understanding is not exactly the same as any of these (see Hills (2009,
2010) for more discussion).
19
A few more brief remarks about explanation. First, explanations can be more or less
full. To understand why m, you do not need to have the fullest possible explanation, an
explanation ‘all the way down’. How many of the morally relevant features you need to
be sensitive to, how full an explanation you need to be able to give may depend on your
circumstances. Explanations of moral truths that involve moral claims (e.g., it is wrong to
raise sales tax because doing so unfairly burdens the worse off) are perfectly legitimate. It is
possible that ultimately all moral claims may be explicable in non-moral terms (because they
ultimately are grounded on natural features of the situation) but I do not suppose that in
order to understand why m you must be able to give an explanation of it in non-moral terms.
20
In addition, if you judged that p on the basis of q in the face of widespread disa-
greement, you might understand why p without knowing that p. For instance, suppose
that many people whose judgement you had previously thought perfectly good, said
that not-p, this testimony would be a defeater for knowledge—it would defeat whatever
grounds you had for knowing that p. So even if you continue to believe that p, you could
not know that p (or know why p). But provided that your judgement that p (because q)
was true, you could have understanding, even if that judgement was made in the face of
misleading evidence. I discuss this sort of case at great length in Hills (2010).
12 Alison Hills
Understanding why p and knowledge why p are thus two separate states.
One way of conceiving of the difference uses a Platonic metaphor. Knowledge
requires a true judgement, and the judgement has to be ‘tied down’. It is contro-
versial exactly what this tie is. Perhaps the true judgement must ‘track the truth’
or be ‘safe’ or not be ‘lucky’. Understanding why p also requires a true judge-
ment, and also requires that the judgement be tied down. This ‘tie’ consists in
your having a set of abilities to make judgements and give explanations in this
and in similar cases. Exercising understanding requires an even firmer tie: that
you have made your judgement by exercising your abilities. Notice that the
kind of anchor required for understanding is quite different from that needed
for knowledge. You can have either kind of ‘tie’ without the other: knowledge
without understanding or understanding without knowledge.
How can you acquire moral understanding? You need to develop the key
abilities, to explain why actions are right and to make judgements about
what is right based on the right reasons. Trusting when you are told that this
or that action is right will usually not give you the grasp of the reasons why
that is essential to understanding.
Why acquire moral understanding in the first place? Because it is important
to use the abilities characteristic of moral understanding to decide what to do.
To see why, we need to think about moral virtue. Moral virtue is widely
agreed to be a kind of responsiveness to moral reasons, usually explained as
responsiveness through action and through feelings, emotions, and desires.
But it is also possible to respond appropriately or inappropriately to moral
reasons through your cognitive attitudes.
What is appropriate responsiveness to moral reasons? I suggest it is: form-
ing a moral judgement (e.g., this action is morally good) through your sen-
sitivity to the reasons why it is true (e.g., because this action helps people
who are in great need). In other words, it is a matter of making moral judge-
ments by exercising your moral understanding.21
To have virtue, you need to decide what to do by exercising your moral
understanding (i.e., by using some of the abilities essential to moral under-
standing).22 In Aristotelian terms, you need phronesis. This has serious
implications for moral testimony.
21
It is sometimes said that a virtuous person may ‘just see’ what it is right to do. And
when you ‘just see’ that an action is right, you do not deliberate about whether it is right
beforehand (and you may not even be aware exactly why it is right: McDowell (1979: 332)).
But even if that is right, it just shows that you do not need to exercise all of the abilities essen-
tial to moral understanding any time that you decide what to do. You still need to make your
judgement for the right reasons, i.e., to have and to use abilities (iii) and (iv), and you might
well need the others as well, to explain and to justify your action to others, or to give advice.
22
This is why moral uliefs are said to aim at ‘exercised’ moral understanding, rather
than merely that they aim at moral understanding. As I mentioned earlier, moral
Suppose that you have a friend who is very seriously ill and she asks you
to help her kill herself. You are not sure what to do. You ask your parish
priest, for whom you rightly have the highest regard. If you put your trust
in what he says, you will know what to do.
Suppose that you do. Now both you and he have moral knowledge. But
only he has responded directly to moral reasons: if he has given you pure
moral testimony, you are not aware of those reasons at all. You have just
done what he told you to do. This is not an ideal of moral virtue. You may
be well motivated and well meaning but you fall short of full moral virtue,
even if you do the right thing knowing that it is right.
Similarly, for your action to be fully morally worthy, you must do the
right thing for the right reasons, that is, for the reasons that make that
action right. Suppose that you trust your priest when he says that it is
acceptable for you to help your friend and act on that basis. Your action is
based on his pure moral testimony. It is not based on the reasons that make
it right—it could not be—because you have no idea what they are and you
are not responsive to them in any way. Again, you may be well motivated
and do the right action, but that is not enough for your action to be fully
morally worthy.23
Exercising your moral understanding to make moral judgements is an
ideal. Other things being equal, you should try to live up to this ideal. But
other things are not always equal. First, trusting testimony has a crucial
role in moral education. To acquire moral understanding, you need to start
with some (preferably true) moral claims, and you may have to take these
on trust.
Secondly, there may be some reason why you cannot gain or cannot use
your moral understanding. Perhaps you lack certain kinds of experience and
these are essential to good judgement. Perhaps you are just not very good at
weighing up values. Hopefully you will be able to get good advice that will
help you to reach a good judgement by explaining why some action is right
and others wrong. But perhaps you are not in a position to get or to appreciate
advice. It may be better for you to trust moral testimony instead. Again: this
understanding consists in making a judgement (that p because q (or that q is why p))
and in having certain abilities. You might have these abilities, but nevertheless make
your moral judgements on other grounds: testimony for example. But in order to be
virtuous, it is important not just to have these abilities, but to use (some of ) them
to make your moral judgements. And if you exercise moral understanding in making
your moral judgements—or even if you try to—you are forming moral uliefs, not
beliefs.
23
What if it was not a case of pure moral testimony and the priest explained his rea-
sons? This is a more complicated example that I do not have space to consider here, but
see Hills (2009) for further discussion.
14 Alison Hills
is a sign that you are not fully morally virtuous.24 Ideally, you would make up
your own mind on the basis of reasons that you yourself appreciate.
1.5 MORAL ULIEFS
Suppose that an increase in sales tax has been proposed and you are won-
dering whether this policy is just. You recognize that raising this tax will
force the worse off to pay a higher rate of tax than those who are better off.
You judge that it is unjust on this basis. To put it schematically, you are
wondering whether p, whilst you recognize that q. Exercising your moral
understanding, you judge that p and that q is why p.
Let us focus on your judgement that p, that a rise in sales tax is unjust.
What kind of mental state have you now formed?
The state is plainly a propositional attitude. It can play a role in action.
For instance, you might oppose the policy of an increase in sales tax, because
you judge it to be unjust. And you might use the judgement in further
reasoning (for example, since further tax income is needed, but a sales tax
increase is unjust, a different way of raising tax must be found). The state is
responsive to certain kinds of evidence regarding p, for example, arguments,
counterexamples, thought-experiments and so on.
At this point, we can describe your judgement as a cognitive state, a belief
in the broad sense. But is it a belief (narrow sense)? It is characteristic of
narrow beliefs to aim at knowledge. But we have seen that knowledge is not
the only epistemic aim. Instead, you might make moral judgements as the
virtuous person does, by exercising her moral understanding. I call a state
aimed at this goal, a moral ‘ulief ’.
You need not be consciously aware of trying to exercise moral under-
standing to be forming uliefs. Nevertheless uliefs should be formed in
response to evidence, in such a way that in favourable circumstances the
result will be exercised moral understanding.
If you exercise your moral understanding, you respond to certain kinds
of evidence only—evidence regarding the reasons why a rise in sales tax is
unjust. But testimony of the form: ‘a rise in sales tax is unjust’ is not a reason
24
See Jones (1999) for a good example. This is similar to the familiar problem of
when it is right and when it is not for a non-virtuous person to try to act as the virtuous
person does. The best-known examples concern non-virtuous non-cognitive attitudes.
For instance, you might be more prone to anger or jealousy than a virtuous person, and it
would be right for you to avoid certain situations, which the virtuous person has no need
to avoid. Similarly, sometimes it might be right for you not to make moral judgements
in the way a virtuous person does, because doing so would be likely to end in failure.
why that proposition is true. So if you are forming uliefs, you will typically
not put your trust in testimony of that form, however reliable or expert you
regard the speaker.
In other words, if your moral judgement that a rise in sales tax is unjust is
sensitive to evidence including moral testimony, it is a (narrow) belief. If it
is sensitive to other kinds of evidence, but not moral testimony (or similar),
it is a moral ulief.
There are a host of well-known problems for any view that moral judge-
ments are not beliefs. These include requirements to explain how such
judgements are truth-apt and subject to logical relations and how they can
be a proper focus for disagreement. These are all extremely demanding chal-
lenges for non-cognitivsts to meet. Are they equally difficult for a ‘moral
ulief ’ account of moral judgement?
Not at all. Moral uliefs are a kind of broad belief, therefore they are fac-
tive states that ‘aim at the truth’. Moral uliefs, like narrow beliefs, will typi-
cally be expressed by declarative sentences with descriptive content that can
be assessed as true and false. For instance:
Sales tax increases are unjust.
Sales tax increases are unjust because they disproportionately affect
the worst off.
Since truth is as important for uliefs as beliefs, uliefs can and do play
the same role in reasoning as beliefs. You can derive further uliefs or beliefs
from combinations of your current uliefs and beliefs in familiar ways. For
example, you might think though the following argument:
1. Rises in tax which disproportionately affect the worse off are unjust.
2. A rise in sales tax is a rise in tax which disproportionately affects the
worse off.
Therefore
3. A rise in sales tax is unjust.
A piece of mental reasoning like this may be constituted by narrow
beliefs—that is, you may n-believe the propositions expressed by (1) and
(2) and draw the appropriate conclusion, forming a narrow belief that a
rise in sales tax is unjust. But here is another possibility. You may n-believe
that this is a tax rise that disproportionately affects the worse off, but have a
moral ulief that rises in tax which disproportionately affect the worse off are
unjust (e.g., because you grasp that a just system of taxes is sensitive to need
and the ability to pay). And you may as a consequence draw the appropriate
conclusion, forming a moral ulief (that this rise in tax is unjust) rather than
an n-belief. Moral uliefs can play a role in reasoning and in valid arguments
16 Alison Hills
because the ‘logic’ of moral uliefs is sufficiently similar to the ‘logic’ of nar-
row beliefs. And finally, if you form the moral ulief that killing the innocent
is always wrong, whilst I form the moral ulief that it is not, it is perfectly
clear that we disagree with each other. Our judgements cannot both be
true, and anyone who ‘ulieved’ both propositions would have inconsistent
uliefs. Just as there is something prima facie wrong with having inconsistent
n-beliefs, there is something prima facie wrong with having inconsistent
uliefs (or indeed in having an inconsistent set of uliefs and n-beliefs).
The role of belief, particularly moral beliefs, in action is of course a matter
of dispute. On some views, beliefs cause action only in combination with a
desire. According to other theories, beliefs can sometimes cause action on
their own, without the need of an additional desire. I will not take a stand
on this controversy. I assume that whatever role moral n-beliefs play in caus-
ing action can also be played by moral uliefs.
I have introduced the idea of moral uliefs by describing moral under-
standing, its use in virtue and morally worthy action. In the next section,
I want to consider some objections that might be raised to this account of
moral uliefs as belief-like mental states.
1.6 OBJECTIONS
1.6.1 Is a Ulief a Kind of Cognitive State At All?

I will take it that a cognitive state is a broad belief. In section 3, broad
beliefs were defined in terms of a set of properties. It is a propositional
attitude. It has a ‘mind to world’ direction of fit. It has a set of characteristic
functions—being typically formed in response to evidence, taking part in
reasoning and in action. It may have a characteristic phenomenology:
This what-it-is-likeness typically involves: (1) psychologically “coming down” on
some issue, in a way that (2) classifies (sometimes spontaneously) some “object”
of focus as falling under some category, where one’s classificatory coming down is
experienced (3) as involuntary, (4) as a cognitive response to some sort of consid-
eration that is experienced (perhaps peripherally in consciousness) as being a suf-
ficient reason for categorizing as one does, and (5) as a judgement that is apt for
assertion and hence is naturally expressible in public language by a sentence in the
declarative mood.
(Horgan and Timmons 2006: 263)
I think it is clear that moral uliefs will have all these features. Uliefs aim
to fit the world, though in a slightly different way from n-beliefs: I have
characterized as n-beliefs aiming at knowledge whilst moral uliefs aim at
exercised moral understanding. Moral uliefs have the functional role of a
cognitive state: they can play the same role in argument as n-beliefs and can
be used in the same mental inferences (aside from any specifically relating
to testimony or deference to experts or similar). They do not have exactly
the same functional role with respect to action as narrow beliefs, because
they play a crucial role in fully morally worthy action and in full moral vir-
tue. But nevertheless, their causal contribution to action is in other respects
similar.
Finally, moral uliefs have the phenomenology of a cognitive state. A moral
ulief is formed on the basis of your grasp (or your attempted grasp at least)
of the reasons why it is true. You cannot just choose to judge whether a rise
in sales tax is unjust or not, once you recognize (what seem to you to be) the
reasons why it is unjust. This is no doubt experienced as a ‘coming down’
in classifying the rise in sales tax (‘the object’) as falling into a particular
category (unjust policies). It is naturally expressed in an assertion (‘A rise in
sales tax is unjust’) by a declarative sentence.
I leave it open whether there are cognitive states other than narrow beliefs
and uliefs, though I think it likely that there are.25 But in any case, there is
no doubt about it: uliefs are a type of cognitive state.
1.6.2 Why Distinguish Uliefs and N-Beliefs? Why Not

Call Them All Beliefs?
No one has until now suggested that there is such a thing as a ulief, a cogni-
tive state that is not a (narrow) belief. Surely, if there were any such states,
we would know about them? But though many people who make moral
judgements describe them as beliefs (‘I believe that a rise in sales tax is mor-
ally wrong’), absolutely no one has ever before talked about her uliefs.
In addition, isn’t proliferating the number of different cognitive states
multiplying entities beyond necessity? Why not just say that there is one
fundamental type of cognitive state: a belief? It is extravagant to posit two
different states to do basically the same thing.
With regard to the first objection: it is legitimate to use the term ‘belief ’
as a general term for all cognitive mental states (I have done so here, though
for clarity I have called them broad beliefs). It is certainly true that no one
talks about uliefs. Does this matter? Only if we are some sort of authority
on our own mental processes, that these are transparent to us. But obviously
they are not. A failure to distinguish between two cognitive states is hardly
unexpected.
25
Perhaps an ‘alief ’ as described in Gendler (2008) might be a cognitive state (though
it also has non-cognitive and affective aspects too), and more familiarly, suppositions and
hypotheses might qualify as well (see Velleman (2000)).
18 Alison Hills
With regard to the second objection, it is true that the criteria for indi-
viduating types of mental state are not clear. Nevertheless, I have argued that
there are some good reasons for distinguishing between different ‘belief-like’
cognitive states, whether these are considered as different types of mental
state or different sub-kinds of broad belief. Some of these reasons are theo-
retical. I have argued that knowledge and (exercised) moral understanding
are different states, and that you can have one without the other. Narrow
belief ‘aims’ at knowledge, hence it is responsive to all evidence. Moral ulief
‘aims’ at exercised moral understanding, hence is responsive only to certain
kinds of evidence.
Secondly, there are practical reasons for distinguishing n-beliefs and
uliefs. Moral virtue is typically a matter of forming and acting on moral
uliefs, rather than n-beliefs.26
These are good reasons for concluding that n-beliefs and uliefs are two
different mental states. Nevertheless, there are costs to doing so. The major
difficulty consists in explaining the relationship between the two: when you
have a ulief that p do you (Sometimes? Always? Never?) have a narrow belief
that p as well? Can you have a ulief that p and a belief that not-p at the same
time? What are the inferential relations between the two? A full account of
this relationship is beyond the scope of this chapter, but in the following
section, I will make a start, hoping to say enough to vindicate drawing the
distinction between uliefs and n-beliefs.
1.6.3 What Is the Relationship between Moral Uliefs

and Moral N-Beliefs?
According to some people, all moral judgements are narrow beliefs: all
should be responsive to pure moral testimony. Even those who deny this,
and think that not all moral judgements should be so responsive, neverthe-
less typically think that at least some of them should: some moral judge-
ments are narrow beliefs.
Suppose that you begin with a narrow moral belief, formed on the basis
of testimony from your family or friends. But you begin to reflect, using
your moral understanding, on why it is true—and why it might not be.
Suppose that you change your mind. You n-believed that abortion was
always wrong and now you form the moral judgement that it is not. Your
n-belief has been replaced by a ulief (if it had not, you would have believed
(in a broad sense) a contradiction). But what if you reaffirm your prior
view? Is your n-belief supplemented or replaced with a ulief?
26
Both arguments are made at greater length in Hills (2009, 2010).
I suspect that typically it is replaced, that is, you suspend your n-belief
and replace it with a ulief that p. To form a ulief at all, you need to base it
on the exercise of your moral understanding, and I think that that is dif-
ficult to do (though perhaps not impossible) when you still have an n-belief
that p. For you already have the answer to your question: is p true? Instead
you need to regard the question as reopened, by suspending judgement on
whether p.
Of course at this stage you might ulieve that p on the basis of your moral
understanding and n-believe that p too. But there are a couple of reasons
why the n-belief might remain suspended. Once you have formed a ulief,
you have reached a settled opinion that p, you have no need of an n-belief
that p as well. And secondly, it makes sense for us to replace n-beliefs with
uliefs rather than to supplement them, for if the argument sketched here is
right, it is morally important that it is the ulief and not the narrow belief
that plays a role in further action, and if there is no n-belief, there is no
chance that it will play the role instead.
This is not to say that it is impossible to have both an n-belief and a ulief
that p at the same time. Indeed, I also think it possible to n-believe that p
and ulieve that not-p at the same time. Perhaps your moral views are not
completely consistent, maybe you are in the process of a moral ‘conversion’.
Consider the following example.
1.6.3.1 The Milgram Experiments

In the famous Milgram experiments, participants were asked to ‘shock’ a
victim to 450 volts (starting with a shock of 15 volts and increasing at 15
volt intervals) despite his screams of pain and demands to be released. The
shocks (which were not real) were apparently applied in the course of an
experiment purporting to be about learning. If the subjects raised ques-
tions about the experiment, the experimenter, who was present in the room,
ordered them to continue (saying, for instance ‘please continue’, ‘the experi-
ment requires that you continue’, ‘you must go on’). Before the experiment,
most people say that they would refuse to participate, but in fact 65 per cent
of the subjects continued to shock the victim until they reached 450 volts.
What is going on in the minds of participants in the experiment? This
is a highly controversial matter, about which there are different opinions.
Here is one of them: the subjects treated the experimenter as a moral expert:
He is the authority—not (or not just) in the sense of the person in charge but in
the sense of the person who knows what is to be done in the experimental situation.
The subjects, after all, are strangers to the world of the laboratory or, better, visitors
to this world. They are looking for clues about how to act, and the experimenter
provides them … The experimenter shows the subjects a vision of the moral order.
20 Alison Hills
He “tells” them that the right thing to do here is to shock the victim. He does this
not by arguing explicitly that it is morally correct to continue shocking but simply
by ordering them in the most matter-of-fact way to continue. In so doing, he shows
them that he (a seemingly reasonable, smart, competent fellow) takes it to be mor-
ally appropriate to do so.
(Sabini and Silver 2005: 550–1)27
According to Sabini and Silver, the subjects n-believe that it is appropri-
ate (perhaps even required) to continue to shock their victim. This narrow
belief is based on their taking the experimenter to think so too, on the basis
of his behaviour, and on their taking him to be an expert on the appropriate
way to behave in his laboratory. Of course they have no idea why the experi-
ment might be acceptable. And moreover they do not wholeheartedly judge
that it is right, even during the course of the experiment. They hesitate,
protest, feel under stress, and in a few cases refuse to participate further.
One possibility is that there is a conflict between their cognitive states
(their belief that the experiment is acceptable) and their non-cognitive atti-
tudes (their desire not to harm). Another possibility, however, is that there
is a conflict in their cognitive states. They n-believe that the experiment is
morally acceptable, but they also ulieve that it is not. This ulief is based on
their hearing the screams and cries from the subject in the next room: they
appear to be causing considerable suffering and possibly injury, for no obvi-
ous gain. They naturally form the ulief that continuing to participate is
morally wrong. This cognitive conflict causes their confused response: they
carry on with the experiment thanks to their n-belief but their ulief causes
them to hesitate; in a few people, the ulief wins out and they refuse to go on.
Prior to the experiment, almost everyone claims that they would not
participate. Why do they say this? Obviously, they think that in the course
of the experiment, they would form a ulief that participating is wrong, and
that they would and should act on that ulief. But, as Sabini and Silver put
it, in the setting of the experiment, most participants lose their ‘moral rud-
ders’ or their ‘moral compass’, that is, they fail to act on the basis of their
own moral understanding. There are a number of interfering factors in the
experiment, some motivational (a desire to obey authority, a wish to avoid
the embarrassment of making a scene) and some cognitive (a belief that the
experiment must be morally acceptable, because the experimenter appears
to think that it is). The two also combine: a strong desire not to disrupt the
experiment and disobey the authority figure makes the n-belief that partici-
pating is morally acceptable much more attractive than it would otherwise
27
Other interpretations are advanced by Harman (2000) and Kamtekar (2004),
amongst others.
be. It is hard—much harder than we predict—to form a clear ulief that

participating in the experiment is wrong, and it is harder than we predict
to act on that ulief.28
According to this interpretation, the Milgram participants have a clash of
cognitive attitudes. Even if you agree that that is possible, you might insist
that these clashes are straightforward conflicts of beliefs. But I think that
there are clear advantages in regarding them as involving both uliefs and
n-beliefs, which are also reasons for making the distinction among cognitive
attitudes in the first place.
Interpreting these situations as ulief/narrow belief conflicts emphasizes
the different grounds of the cognitive states in each case—whether it is
responsive to the authority of other people’s judgements or not. This distinc-
tion is worth making, linked as it is to the differences between knowledge
and moral understanding. But I have also argued that it is morally impor-
tant. A virtuous person would be disposed to use their own judgement and
would respond to moral reasons appropriately in this sort of situation.29
1.7 COGNITIVISM REDEFINED
Cognitivists about moral judgement have typically claimed that moral

judgements are beliefs and the view has often been defined in those terms.
I have argued that this is a confusing definition. There are cognitive states
(broad beliefs) that are not narrow beliefs, so it would be better to define
cognitivism about moral judgement as the following view:
Cognitivism: moral judgments are cognitive states (beliefs in the
broad sense, i.e., they may be narrow beliefs or moral uliefs or …).
28
Of course there is a complicating factor in the Milgram experiment, namely that
the shocks are not real and so no one at all is suffering. But anyone who was aware of
this would not be an appropriate subject for the experiment. And a virtuous person in
a real-life situation of the type the experiment is trying to model would not agree to
give significant shocks to the victim. Glover (1999) recounts similar real-life examples,
including the My Lai massacre, in which many ordinary US soldiers obeyed an order to
kill unarmed civilians and only a few refused. According to Glover (1999: 333) one of
the whistleblowers on the massacre, Ronald Ridenhour, had been the subject of a repeat
of the Milgram study at Princeton, and refused to give any shocks at all.
29
It does not follow from this that a morally virtuous person never defers to anyone
else’s moral judgement. There may be specific circumstances in which she has good rea-
son to think that her own judgement is impaired and she should trust someone else, but
in these circumstances she is not acting ‘characteristically’. Note also that even if a mor-
ally virtuous person would not defer to someone else in a particular situation, it does not
follow that the rest of us should not either. We should not always try to act as the virtuous
22 Alison Hills
And correspondingly, non-cognitivism should be defined either as the

negative view:
Non-cognitivism 1: moral judgements are not cognitive states (broad
beliefs).
Or, if you prefer a ‘positive’ characterization:
Non-cognitivism 2: moral judgements are ‘desire-like’ mental states
(e.g., pro-attitudes, commitments, etc.).
Since both n-beliefs and uliefs are cognitive states, a cognitivist who thinks
ideal moral judgements are not made on the basis of pure moral testimony
could claim that whilst some moral judgements are narrow beliefs, oth-
ers are moral uliefs (a cognitivist who disagreed could maintain that moral
judgements are all narrow beliefs). Hybrid theorists, who think that moral
judgements are a mixture of cognitive and non-cognitive states could state
their theory in terms of uliefs as well as narrow beliefs.
The main purpose of this chapter is to clarify one option open to cognitiv-
ists in the face of a problem about pure moral testimony. Of course, this does
not establish that cognitivism is correct. In particular, nothing I have said
here resolves the ‘practicality’ problem for cognitivism, for moral judgements
appear to have a connection to motivation which cognitive states do not. I am
not convinced by this argument, but I will not make the case against it here.30
Instead, I want to mention very briefly two wider implications of the
argument. First, though I have discussed only moral understanding here,
I think that understanding differs from knowledge in other fields too. For
instance, aesthetic understanding, unlike aesthetic knowledge, requires a
grasp of why a work of art is aesthetically valuable; philosophical under-
standing, unlike philosophical knowledge, requires a grasp of why a phil-
osophical thesis is true. We typically form judgements on these subject
matters by making up our own mind, rather than trusting testimony, that
is, we form uliefs, not narrow beliefs. Perhaps it is possible to form a ulief
on any subject matter on which you can form a narrow belief: I suspect that
it is. I hope to say more about this in future work.
person does—our own shortcomings can make such attempts disastrous. Making moral
judgements as a virtuous person does is always a moral ideal, and we should do so where
possible, but if our judgement is very bad (or the situation very unfavourable) we should
trust testimony, rather than trying to exercise moral understanding.
30
There is a well-known argument that no state can have both the direction of fit of
a belief and that of a desire (Smith 1987: 54f.). I am not convinced that this argument
succeeds, for reasons similar to those given by Setiya, in the course of his argument
that intentions are mental states with both belief-like and desire-like components (Setiya
2007: 48–51). Moreover, this chapter has shown that mental states can be ‘belief-like’ in
Secondly, I raised a question earlier about whether the ‘ideal’ of form-

ing moral judgements by making up your own mind, rather than trusting
moral testimony was a moral or an epistemic idea. I have argued so far
that it is a moral ideal, because a fully virtuous agent would form moral
judgements that way. But I want to now suggest that it is an epistemic
ideal too.
The study of the relationship between the world and our representations
of it has traditionally been called ‘epistemology’, and recently this has been
understood as being—by definition—the study of knowledge. But I sug-
gest that that is too narrow. Knowledge and justified (narrow) belief is only
a subset of that broader subject matter, which includes understanding and
uliefs as well. And we have very good reason to think that the epistemic
rationality and epistemic justification for understanding, particularly with
regard to testimony and related matters (notably, disagreements), will be
rather different.
Suppose that you ignore a piece of moral testimony from some-
one whom you correctly regard as reliable. Could this be epistemically
rational? If epistemic rationality is a matter of gaining relevant knowl-
edge, presumably not, because (ceteris paribus) you could gain knowledge
by doing so. It is true that you might have good moral reasons to ignore
testimony—it is what a virtuous person would do, after all—but these are
what are sometimes called the ‘wrong kind of reasons’. They are practical
reasons, not epistemic ones.
I suggest that this argument is not quite right. ‘Epistemic’ rationality and
justification for moral uliefs should relate to exercised moral understand-
ing, since that is the ‘aim’ of those states. So it can be perfectly rational to
form moral judgements without trusting the testimony and judgements of
others, because making up your own mind is the only way to exercise your
moral understanding. ‘Epistemology’ for moral uliefs with regard to moral
testimony and related matters (including moral disagreement) is going to
look rather different from the standard epistemology for (narrow) belief.
The study of the relationship between our cognitive states and the world,
epistemology in the broadest sense, is a bigger subject than we realized, of
which philosophers have so far studied only one part, and not necessarily
the most important part either.31
some respects but not in others. Perhaps there could be a mental state that is belief-like in
some respects—sufficiently belief-like to qualify as a cognitive state—but also desire-like
in some respects. But I cannot pursue this interesting possibility any further here.
31
Thanks to audiences at Oxford, BSET, SPAWN Stirling, and the Madison
Metaethics conference where this essay was first presented and especially Rachel Cohon
and Debbie Roberts.
24 Alison Hills
References
Aristotle. 2002. Nicomachean Ethics, trans. S. Broadie and C. Rowe. Oxford: Oxford
University Press.
Blackburn, S. 1998. Ruling Passions. Oxford: Clarendon Press.
Brink, D. O. 1989. Moral Realism and the Foundations of Ethics. New York: Cambridge
University Press.
Coady, C. A. J. 1992. Testimony. Oxford: Oxford University Press.
Copp, D. 2001. ‘Realist-Expressivism: A Neglected Option for Moral Realism’,
Social Philosophy and Policy 18: 1–43.
Davidson, D. 1980. ‘Actions, Reasons and Causes’, in his Essays on Actions and
Events. Oxford: Oxford University Press, 3–19.
Driver, J. 2006. ‘Autonomy and the Asymmetry Problem for Moral Expertise’,
Philosophical Studies 128: 619–44.
Engel, P. 2004. ‘Truth and the Aim of Belief ’, in D. Gillies (ed.), Laws and Models
in Science. London: King’s College Publications, 77–97.
Gendler, T. 2008. ‘Alief and Belief ’, Journal of Philosophy 105: 634–63.
Gibbard, A. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Glover, J. 1999. Humanity: A Moral History of the Twentieth Century. London: Cape.
Greene, J. D. 2007. ‘The Secret Joke of Kant’s Soul’, in W. Sinnott-Armstrong
(ed.), Moral Psychology, vol. 3: The Neuroscience of Morality: Emotion, Disease, and
Development. Cambridge, MA: MIT Press, 35–79.
Grimm, S. 2006. ‘Is Understanding a Species of Knowledge?’ British Journal for the
Philosophy of Science 57: 515–35.
Haidt, J. 2001. ‘The Emotional Dog and its Rational Tail: A Social Intuitionist
Approach to Moral Judgment’, Psychological Review 108: 814–34.
Haidt, J. and Bjorklund, F. 2008. ‘Social Intuitionists Answer Six Questions about
Moral Psychology’, in W. Sinnott-Armstrong (ed.), Moral Psychology, vol. 2: The
Cognitive Science of Morality: Intuition and Diversity. Cambridge, MA: MIT Press,
181–217.
Harman, G. 2000. ‘Moral Philosophy Meets Social Psychology: Virtue Ethics and
the Fundamental Attribution Error’, in his Explaining Value and Other Essays in
Moral Philosophy. Oxford: Oxford University Press, 165–78.
Hills, A. E. 2009. ‘Moral Testimony and Moral Epistemology’, Ethics 120: 94–127.
Hills, A. E. 2010. The Beloved Self: Morality and the Challenge from Egoism.
Oxford: Oxford University Press.
Hopkins, R. 2007. ‘What is Wrong with Moral Testimony?’, Philosophy and
Phenomenological Research 74: 611–34.
Horgan, T. and Timmons, M. 2006. ‘Cognitivist Expressivism’, in Horgan and
Timmons (eds.), Metaethics after Moore. Oxford: Oxford University Press, 255–98.
Hussar, K. M. and Harris, P. L. 2010. ‘Children Who Choose Not to Eat
Meat: A Study of Early Moral Decision-Making’, Social Development 19: 627–41.
Jones, K. 1999. ‘Second-Hand Moral Knowledge’, Journal of Philosophy 96: 55–78.
Kamtekar, R. 2004. ‘Situationism and Virtue Ethics on the Content of Our
Character’, Ethics 114: 458–91.
Kant, I. 1991. Metaphysic of Morals, trans. M. Gregor. Cambridge: Cambridge

University Press.
Kant, I. 1998. Groundwork of a Metaphysic of Morals, trans. M. Gregor.
Cambridge: Cambridge University Press.
Kitcher, P. 2002. ‘Scientific Knowledge’, in P. Moser (ed.), The Oxford Handbook of
Epistemology. Oxford: Oxford University Press, 385–407.
Kohlberg, L. 1973. ‘The Claim to Moral Adequacy of a Highest Stage of Moral
Judgment’, Journal of Philosophy 70: 630–46.
Kvanvig, J. 2003. The Value of Knowledge and the Pursuit of Understanding.
Lipton, P. 2004. Inference to the Best Explanation, 2nd edition. London: Routledge.
McDowell, J. 1979. ‘Virtue and Reason’, The Monist 62: 331–50.
McGrath, S. 2011. ‘Skepticism about Moral Expertise as a Puzzle for Moral Realism’,
Journal of Philosophy 108: 111–37.
Owens, D. 2003. ‘Does Belief Have an Aim?’, Philosophical Studies 115: 283–305.
Pritchard, D. 2005. ‘Knowledge, Understanding and Epistemic Value’, in A. O’Hear
(ed.), Epistemology. Cambridge: Cambridge University Press, 19–43.
Ridge, M. 2006. ‘Ecumenical Expressivism: Finessing Frege’, Ethics 116: 302–36.
Ridge, M. 2007. ‘Ecumenical Expressivism: The Best of Both Worlds?’, in R. Shafer-
Landau (ed.), Oxford Studies in Metaethics, vol. 2. Oxford: Oxford University
Press, 51–76.
Sabini, J. and Silver, M. 2005. ‘Lack of Character? Situationism Critiqued’, Ethics
115: 535–62.
Schroeder, M. 2008. Being For: Evaluating the Semantic Program of Expressivism.
Schroeder, M. 2009. ‘Hybrid Expressivism: Virtues and Vices’, Ethics 119: 257–309.
Setiya, K. 2007. Reasons without Rationalism. Princeton, NJ: Princeton University Press.
Sinclair, N. 2006. ‘The Moral Belief Problem’, Ratio 19: 249–60.
Smith, M. 1987. ‘The Humean Theory of Motivation’, Mind 96: 36–61.
Van Roojen, M. 2013. ‘Moral Cognitivism vs. Non-cognitivism’, Stanford Encyclopedia
of Philosophy <http://plato.stanford.edu/entries/moral-cognitivism/>.
Velleman, J. D. 2000. ‘On the Aim of Belief ’, in his The Possibility of Practical
Reason. Oxford: Oxford University Press, 244–82.
Wedgwood, R. 2002. ‘The Aim of Belief ’, Philosophical Perspectives 16: 267–97.
Williams, B. 1981. ‘Internal and External Reasons’, in his Moral Luck.
Cambridge: Cambridge University Press, 101–13.
Williams, B. 1995. Making Sense of Humanity. Cambridge: Cambridge University Press.
Williamson, T. 2000. Knowledge and Its Limits. Oxford: Oxford University Press.
Woodward, J. B. 2003. Making Things Happen: A Theory of Causal Explanation.
Zagzebski, L. 2001. ‘Recovering Understanding’, in M. Steup (ed.), Knowledge,
Truth and Duty: Essays on Epistemic Justification, Responsibility and Duty.
Oxford: Oxford University Press, 235–51.
2
Acting for the Right Reasons,
Abilities, and Obligation
Errol Lord
2.1 INTRODUCTION
Some moral theorists—we’ll call them objectivists about obligation—hold

that what an agent ought to do is a function of all of the normatively rel-
evant facts.1 Other moral theorists—we’ll call them perspectivalists about
obligation—hold that what an agent ought to do is rather a function of some
perspectival facts.2 Perspectival facts are facts within the agent’s perspective.
What counts as the agent’s perspective is a matter of debate amongst per-
spectivalists, as is the debate about which facts matter within one’s perspec-
tive. We’ll get to this in due time.
The debate between objectivists and perspectivalists is important because
the two views can come apart dramatically. Consider Sick Mother:
Sick Mother
Jack’s mother is in the hospital. She needs an operation in order to
survive past this week. Her insurance won’t pay. Jack, being a fledgling
art dealer, doesn’t have the money. It looks like his mother is going
to die. She would be extremely comforted by Jack’s presence in her
final days. She lives in California; Jack lives in New York. Jack needs
to decide whether to go see her. As it happens, a pawn shop owner in
Queens has just unknowingly (and legitimately) bought a rare Picasso.
1
See Moore (1912), Thomson (1986), and Graham (2010).
2
I am using ‘obligation’ such that A is obligated to φ just in case A ought to φ. There is
a popular usage amongst moral philosophers where obligations are always things that we
owe to other agents. I am not using the word in this way. I mostly use the word ‘obliga-
tion’ to make the prose more elegant. I also think that my use of obligation is a natural
use in English.
Acting for the Right Reasons, Abilities, and Obligation 27
He’s selling it at a fraction of the price it’s worth. If Jack were to buy
it, he would be able to use it as collateral for a loan that would pay
for his mother’s surgery. Unfortunately, he has no idea that this pawn
shop exists.
Ought Jack go to California or ought he go to Queens? Objectivists hold
that all the normatively relevant facts matter when it comes to Jack’s obliga-
tions. Moreover, it’s plausible that all the normatively relevant facts point
towards going to Queens. He can save his mother that way, and that takes
priority one. Perspectivalists, on the other hand, think that only perspecti-
val facts matter when it comes to Jack’s obligations. Moreover, it’s clear that
the perspecitival facts don’t support going to Queens; they support going
to California. After all, Jack’s beliefs, knowledge, and evidence support the
thought that it is best to go comfort his mother in her last days. Given
his perspective, going to Queens is at best a fool’s errand at the cost of his
mother dying a lonely death.
This chapter is dedicated to arguing for a perspectival view. On this view,
the perspectival facts that determine obligation are possessed normative rea-
sons. Possessed normative reasons are the normative reasons that are within
one’s epistemic ken. In sections 2.2 and 2.3 I will provide an argument for
this view. The key thought behind the argument is that the facts that obli-
gate must be potentially action-guiding in a certain sense—the facts that
obligate must at least potentially be the reasons for which we act. This is
because when we are obligated to perform some act φ, we must at least have
the ability to φ for the right reasons. The rub will be that we can have the
ability to act for the right reasons only if we possess those reasons. This is a
huge step forward in a full defense of my view. It also follows that objectiv-
ism is false.
Providing the positive argument for my view is not my only goal. I also
aim at defusing what I take to be the most compelling objection to perspec-
tivalism. This objection—which goes back to at least Moore (1912) and
Ross (2002) and is prominently developed in Thomson (1986) and Graham
(2010)—holds that only objectivism can explain the fact that in delibera-
tion we aim to do what is supported by all the facts. In short, deliberation
aims at what’s best.3 Data often proffered in support of the thought that
only objectivism can explain this fact is that it seems like onlookers with
more information can have true thoughts about what one ought to do that
come apart from what one ought to do given one’s perspective.
I think my view is compatible with the claim that deliberation aims at
what’s best. The key thought is that deliberation can aim at what’s best even
3
By ‘what’s best’ I just mean the thing to do given all the facts.
28 Errol Lord
though our obligations are constrained by our abilities. This is very plausible
when it comes to our physiological abilities. Deliberation can aim at what’s
best even if we are never obligated to do what we are physiologically unable
to do. I will argue in section 2.4 that we should think of my view in a
similar way. My view just enforces a different ability condition. This doesn’t
threaten the claim that the aim of deliberation is to do what’s best. Thus,
considering the nature of ability constraints more broadly shows that the
perspectival view I defend survives the objection.
2.2 DELIBERATION, OBLIGATION,

SUBJECTIVE, OBJECTIVE
There are two important preliminaries. First, we need to get clearer about
the type of obligation that is at issue. We are interested in what I’ll call
deliberative obligations. These obligations are so-called because of their con-
nection to the central deliberative question—viz. what ought I do? For each
time the central deliberative question applies—every time there is some-
thing to be done—there is a correct answer about what is to be done. Of
course, this is not to say that there will always be a single act that is the act
to be done. Oftentimes many actions are permissible.
The correct answer to the central deliberative question will be the act
that you are deliberatively obligated to perform. The question central to this
chapter is what determines one’s deliberative obligations.
This leads to the second preliminary. It will be helpful to adopt an ideol-
ogy in order to explore the topic in more concrete terms. I will adopt the
ideology of normative reasons. Normative reasons are facts that recommend
certain reactions. We can frame the debate by appealing to normative rea-
sons. On this framing, objectivists hold that deliberative obligations are a
function of all of the normative reasons. Whatever is best supported by all
of the reasons is what one deliberatively ought to do. Perspectivalists hold
that only the reasons within one’s perspective can determine what you’re
deliberatively obligated to do. Whatever is best supported by the reasons
within your perspective is what you ought to do.
On my perspectival view, your perspective is made up of the normative
reasons you possess. The normative reasons you possess are the normative
reasons that are in your epistemic ken. So for example, while it is true that
there is a reason for me go to the store if we’re out of milk, I don’t possess
that reason to go to the store unless I’m aware of the fact that we’re out of
milk. If the last of the milk is currently being consumed while I’m at the
office, then even though there is a reason for me to go to the store, I do not
possess that reason to go to the store.
It is, as one might expect, controversial what constitutes one’s epistemic
ken and thus it is controversial which epistemic relation is constitutive of
the possession relation. I will be mostly neutral on this here. As we’ll see, my
argument in section 2.3.2 requires that the epistemic relation be a positive
one. That is to say, in order to possess a reason, one’s belief must have some
epistemic pedigree. Examples of the kind of pedigree required are knowl-
edge and justification. Most views of possession embrace this requirement
(and for good reason).4 For simplicity I will assume that the possession
relation is knowledge.5 This is because all of the going views in the literature
hold that knowing some reason r is sufficient for possessing r.
2.2.1 A Dialectical Primer
The dispute between objectivists and perspectivalists has been dominated
by two types of cases, which I’ll call simple ignorance cases and sophisticated
ignorance cases. In order to appreciate the dispute between the two camps, it
is helpful to think about the cases.
Let’s start with the simple cases. Sick Mother is a simple case. In Sick
Mother, Jack is ignorant of the pawn shop and the Picasso. Given his per-
spective, he ought to go to California. Given all the facts, he ought to go
to Queens.
In all of the simple cases the characters are ignorant of some normatively
relevant facts. Given what the characters know, some act φ-ing is obligatory.
Given all the facts, another act is obligatory. I think it is fair to say that most
people’s initial intuitions about the simple cases support perspectivalism.
However, things are more complicated than they initially seem because the
objectivist has a compelling response.
The response has two parts. The first is that in the simple cases it is always
reasonable to believe that the action that is obligatory in light of one’s per-
spective is also permitted by the balance of all the reasons. This is true in
Sick Mother. It is reasonable for Jack to think that the balance of all the
reasons permits him to go to California. In fact, it’s reasonable for him to
believe that they require him to do so.
4
See Williamson (2000), Neta (2008), Gibbons (2013), Lord (2010), and Sylvan and
Sosa (forthcoming). In Lord (2013: ch. 3) I argue against views of possession that deny
one needs to stand in a positive epistemic relation by appealing to the same kinds of cases
I appeal to in section 2.3.2. See also n. 23.
5
My considered view is that being in a position to know is the relevant relation. See
Lord (2013: ch. 3).
30 Errol Lord
The second part consists in the claim that we should divorce the deontic
from the hypological. That is, we shouldn’t hold that there are any necessary
connections between claims about what ought to be done and claims about
what we’re praiseworthy or blameworthy for doing. In particular, claims
the objectivist, we shouldn’t think that doing wrong is sufficient for being
blameworthy. Importantly for our purposes, there is blameless wrongdoing
when one falsely but reasonably believes that the balance of all the reasons
supports φ-ing and one φ-s. Moreover, we can nicely explain why this is.
The φ-ing is blameless because it was reasonable to believe that φ-ing was
supported by the balance of all the reasons. But φ-ing was wrong because
this belief is false.
This response is dialectically compelling. This is because it is anchored in
the very plausible claim that there can be blameless wrongdoing. Moreover,
given objectivism, the simple cases are paradigm cases of blameless wrong-
doing. At the very least, this response should dampen the strength of one’s
intuitions about the simple cases.
Fortunately for perspectivalists, there are the sophisticated cases. The
standard objectivist response to the simple cases is not available when it
comes to the sophisticated cases. The most famous sophisticated case is
Mine Shaft.6
Mine Shaft
A group of 10 miners are trapped in a mine. They are either trapped in
shaft A or in shaft B. It is not easily knowable which shaft they are in.
Flood waters are approaching the shafts. Billy has the choice to sand-
bag shaft A, sandbag shaft B, or not sandbag either. She knows that if
she sandbags A and the miners are in A, all the miners will survive. She
knows the same is true of B. She also knows that if she sandbags either
shaft and the miners are in the other shaft, they will all die. Finally, she
knows that if she does nothing, then 9 of the 10 will survive.
It is very plausible that Billy ought to do nothing. She ought to guarantee
that 9 miners will survive. It is simply too risky to sandbag either shaft. At
best she will save one life and at worse she will cause ten deaths.
The most important feature of the sophisticated cases is that one is not
in a position to reasonably believe that the balance of reasons supports the
act that is best supported by the facts in one’s perspective. In Mine Shaft,
Billy knows that doing nothing is not the act that will bring about the best
outcome. Nevertheless, it seems like she should do nothing.
6
The case was made famous by Parfit (2011). It originated in Regan (1980). See
Jackson (1991) and Ross (2006) for similar cases.
This blocks the objectivist response to the simple cases from applying to
the sophisticated cases. This is because it was crucial to that response that
the characters reasonably believe that the balance of all the reasons supports
the act that the facts in their perspective support. The characters in sophis-
ticated cases can’t reasonably believe this. Thus, we can’t explain why they
are blameless in doing the second best option by appealing to a reasonable
but false belief. The best explanation of why they are blameless for doing the
second best option, it seems, is that they ought to.
Before moving on, it is important to say something about a common
reaction to the above dialectic. The first reaction of many is that we can
explain all that needs to be explained by appealing to the distinction
between objective and subjective obligation. Objective obligations are a
function of all the normatively relevant facts. Subjective obligations are in
some way perspectival.
The common thought is that in both the simple and sophisticated cases
there is something we objectively ought to do and something we subjec-
tively ought to do. Our intuitions in favor of perspectivalism are really
tracking subjective obligations and our intuitions in favor of objectivism
are really tracking objective obligations. However, neither type of obligation
takes precedence. They are just associated with different things of interest
to normative theory.7
It is important to stress that the appeal to the objective/subjective distinc-
tion I am interested in right now is deflationary when it comes to what I’ve
called the ought of deliberation. On the view under consideration, there is
no conceptual room for the ought of deliberation. There are just the subjec-
tive obligations and the objective obligations. Those who hold this view
think that the debate between objectivists and perspectivalists is built on
sand. There is just no interesting question to ask about the ought of delib-
eration. Let’s call this view the deflationary view.
The deflationary view has a nice conciliatory tone to it, but I think that
it is hard to maintain. I will mention two difficulties. First, this kind of
conceptual deflationism does not seem plausible upon reflection. To see
this, think of Mine Shaft. Billy knows that doing nothing is the second
best option. If she is conceptually sophisticated enough, then she is in a
position to know that she objectively ought to sandbag one of the shafts.
Moreover, if she is conceptually sophisticated enough, she is in a position
to know that she subjectively ought to do nothing. If the deflationist is
right, these exhaust the deontic facts. But it is implausible that Billy has
deliberated about all that can be deliberated about. Billy can ask a further
7
Cf. Schroeder (2009).
32 Errol Lord
question: Which obligation ought I satisfy? The ‘ought’ in this question is

just the ought of deliberation. This is good reason to think the deflationist
is wrong.8
The second problem is clear once we appreciate what the deflationist has
to say about Billy when she asks which requirement she ought to satisfy.
The deflationist has to say that this question is confused. There are only the
subjective and objective obligations. But that means that in this case Billy
is subject to a deontic dilemma. This is because her objective and subjective
obligations come apart and those exhaust the deontic facts. Intuitively, Billy
is not subject to a dilemma. She ought to do nothing!
It gets even worse for the deflationist. For if her view is right, then there
will be widespread deontic dilemmas. This is because there are many cases
where our subjective obligations come apart from our objective obligations.
All of the ignorance cases are like this. In those cases, defenders of the defla-
tionary view must hold that there are simply two incompatible things one
ought to do.
Defenders of the deflationary view will quickly point out that they think
the two forms of obligation are incommensurable in a certain sense. This
is supposed to help with the dilemma objection, and it does to a certain
extent. It would be worse if you were obligated in the same sense to perform
incompatible actions. However, I don’t think it fully dampens the force of
the objection. It is implausible that Billy is subject to any kind of deontic
dilemma. There is one thing she ought to do in the sense tied to her delib-
eration, and that is to do nothing.
Of course, we can deny deflationism without denying the importance
of objective obligation or subjective obligation. Those notions might have
important roles to play. Moreover, it is not to deny the relevance of objec-
tive or subjective obligations to the debate over deliberative obligations.
After all, objectivists hold that the deliberative obligations metaphysically
reduce to the objective obligations. Moreover, a serious contender holds
that the deliberative obligations metaphysically reduce to the subjective
obligations.9 If the deliberative obligations reduced to the objective or sub-
jective obligations, then in a sense—a metaphysical one—there would only
be the objective and subjective obligations. But in another sense—a concep-
tual one—there are still three notions.
Putting the deflationary view to one side, the rest of the chapter will be
dedicated to defending my perspectival view.
8
Cf. Jackson (1991), Kolodny and MacFarlane (2010), Graham (2010).
9
Perhaps my view is a view like this. In order to find out, we’d have to investigate
the essential properties of subjective obligations. I am not interested in doing this here.
2.3 ABILITIES, ACTING FOR THE RIGHT REASONS,

AND OBLIGATION
It is close to uncontroversial that our abilities can constrain our obligations.

Perhaps the least controversial ability condition is the Physiological Ability
Condition:
Physiological Ability Condition: If A ought to φ, then A has the
physiological ability to φ.
The Physiological Ability Condition is supported by strong intuitions.
Consider Dunk for Money:
Dunk for Money
Mark Cuban decides to have a raffle. The person whose name is cho-
sen gets a shot at a $10 million prize. In order to win the prize, one
has to do a 360° dunk on an NBA regulation sized hoop. Just for fun,
Sam enters the raffle. She is the lucky winner. Unfortunately for Sam,
she is only 4′ 11″ tall. Because of this, she lacks the ability to dunk on
an NBA regulation sized hoop.
It is very plausible that it is not the case that Sam ought to perform the
360° dunk. A plausible hypothesis about why this is true is that she lacks
the physiological ability to dunk. Sam’s physiological abilities seem to be
constraining her obligations. If she were able to dunk, it would be the case
that she ought to. This is because dunking would be best.
The Physiological Ability Condition is fairly weak. It is compatible with
objectivism. I will argue for an ability condition that is not compatible with
objectivism. This is the Right Reasons Ability Condition:
Right Reasons Ability Condition: If A ought to φ, then A has the
ability to φ for the right reasons.
The right reasons are the normative reasons that determine one’s obligation.
The intuitive idea behind the Right Reasons Ability Condition is that the
normative reasons that determine our obligations can be action guiding.
They can be the reasons for which we act.
It will be important going forward to have an intuitive grip on the notion
of acting for the right reasons. First, consider a case.10
Jenny needs to get to work. Unfortunately there was a big snow recently.
Jenny’s car is thus covered in snow. She deliberates about alternative ways
of getting to work, but decides—rightly—that she needs to drive. She thus
10
Note that I do not take these features to immediately support the argument below.
34 Errol Lord
digs her car out. The reason for which she digs her car out is that she needs
to get to work. That seems like a good reason to dig her car out.
I think we can tease out of this case some general truths about acting
for the right reasons. The first thing to say is that it is very plausible that
we can provide a certain kind of explanation of why Jenny digs her car out
by appealing to the right reasons. Jenny digs her car out because she has
to get to work. This explanation isn’t merely causal. It is also normative
in a particular way. It explains why the action Jenny performs is justified.
Jenny is justified because she needs to get to work. Let’s call explanations of
this kind justificatory explanations. It’s plausible, then, that the Explanatory
Condition is true:
Explanatory Condition: If A φs for a normative reason r, r provides
a justificatory explanation of why A φs.
What is it that makes it the case that we can provide a justificatory explana-
tion of why Jenny did what she did? A plausible answer to this question is
that we can provide the justificatory explanation because Jenny is sensitive
to the right reasons. She is sensitive to the support relation between the fact
for which she acts and the act she performs. It’s plausible to suppose that
she wouldn’t dig her car out if that fact didn’t provide her with normative
reason to dig her car out.11 She in some way tracks the relevant normative
considerations.12 This seems like a very important part of acting for the
right reasons.13 This supports the Sensitivity Condition:
Sensitivity Condition: If A φs for a normative reason r, A’s φ-ing is
sensitive to the fact that r is a normative reason to φ.
With these conditions in hand, back to the objectivist. The rub for the
objectivist is that I think that in order to meet the Right Reasons Ability
Condition for some reason r, one must possess r. In order to possess r, r has
11
This is not intended to be an analysis of sensitivity. I don’t think any counterfactual
analysis is adequate. I think that we analyze this sensitivity in dispositional terms (and
I don’t analyze dispositions counterfactually). The sensitivity involved is the disposition
not to perform the action if the reasons were defeated. See Lord (2013: ch. 4) and Lord
and Sylvan (n.d.) for more.
12
I should note that I don’t think this means that one generally needs to be a good per-
son to act for the right reasons. The relevant abilities might be very local and quite fragile
and thus it might be easy for them to not manifest in similar situations. This position is
possible given my rejection of a counterfactual analysis of abilities.
13
If you’re not convinced of this yet, keep reading. This will be supported further by
the discussion of creditworthiness below (see also Lord and Sylvan (n.d.)). I admit now
that there might be a thinner notion of ‘acting for the right reasons’ that doesn’t require
this type of sensitivity. But I maintain that the notion moral philosophers have been
interested in (the notion tied to credit) requires sensitivity.
to be within one’s perspective. Thus, objectivism is false. Here’s the argu-

ment in a more perspicuous form:
(1) If A ought to φ, then A has the ability to φ for the right reasons (Right
Reasons Ability Condition).
(2) If A has the ability to φ for the right reasons, then A possesses the right
reasons.
(C) If A ought to φ, then A possesses the right reasons.
If this argument is sound, then objectivism is false and some perspectival
view is true. The rest of this section is dedicated to defending (1) and (2).
2.3.1 In Defense of the Right Reasons Ability Condition

The intuitive idea behind the Right Reasons Ability Condition is that the
reasons that determine our obligations must be able to guide our actions.
The paradigm way in which reasons guide our actions is by being the rea-
sons for which we perform actions. I think there are at least two arguments
that flesh out this intuition.
The first argument turns on a principle tying acting for the right reasons
to acting non-accidentally in a certain sense. It is very plausible that when
you act for the right reasons, you non-accidentally perform the act you
ought to perform. We can see this by considering pairs of cases like Good
Husband and Bad Husband:
Good Husband
Brandon’s wife Jen’s birthday is tomorrow. She badly wants a new hat.
He buys her a new hat. Moreover, the reason for which he buys her
the hat is that she wants it.
Bad Husband
Brandon’s wife Jen’s birthday is tomorrow. She badly wants a new hat.
He buys her a new hat. However, the reason for which he buys her
the hat is that he dislikes her hair and hopes that she will cover it up
with the hat.
In both Good Husband and Bad Husband Brandon ought to buy Jen a new
hat.14 Moreover, in both cases the reason that obligates Brandon is provided
by the fact that Jen wants a new hat. The only relevant difference between
the cases is that it is only in Good Husband that Brandon’s action is per-
formed for the right reasons.
14
I’ll assume Brandon knows all of the relevant facts.
36 Errol Lord
This is plausible, in part, because in Bad Husband it is an accident that

the act Brandon performs for bad reasons happens to be the action he ought
to perform.15 It is accidentally true in Bad Husband that Brandon buys Jen
a new hat and Brandon ought to have bought her a new hat. Given that
he is not sensitive to the relevant normative reason, he could have easily
been moved to do something that he ought not. This is not true in Good
Husband. Given that Brandon is sensitive to the relevant normative con-
sideration in Good Husband, it is plausible to think that it is no accident
that the action he performs is the action he ought to have performed. These
thoughts support No Accident:
No Accident: If A φs for the right reasons, then it is not an accident
that the action A performs is the action A ought to have performed.
No Accident is not the whole truth. It being no accident that Brandon
performs the action he ought to have performed is merely a necessary
condition on acting for the right reasons. This is because it seems possi-
ble for Brandon’s action to be modally robust in the relevant sense even
though he doesn’t act for the right reasons. For example, it might be that,
in Bad Husband, there is a demon who will guarantee that, no matter what
Brandon ought to buy for Jen, he will be selfishly motivated to buy that
for Jen. In that case, it won’t be a modal accident that the action Brandon
performs is the action he ought to have performed. Intuitively, though, he
still doesn’t act for the right reasons.16
What’s missing in the demon case is that the fact that Brandon performs
the act he ought to perform non-accidentally is not explained in the right
way. Specifically, that fact is not explained by the fact that Brandon is sensi-
tive to the normative feature of the fact for which he acts. This supports
Sensitive No Accident:17
Sensitive No Accident: A φs for the right reasons just in case A φs
non-accidentally because A is sensitive to the right reasons.
Now for the problem for denying the Right Reasons Ability Condition.
Given Sensitive No Accident, if the Right Reasons Ability Condition is
false, then there will be cases where one ought to φ even though one will
be unable to φ non-accidentally in the right way. This is implausible. If it
were true, then there would be cases where you ought to φ even though the
reasons that make this true cannot get any legitimate grip on you—i.e., they
15
For similar remarks, see Arpaly (2003: ch. 3) and Markovits (2010).
16
It is not an accident that these thoughts mirror familiar thoughts about how knowl-
edge is non-accidental in a certain sense. See also Arpaly (2003: ch. 3).
17
This is not meant as an analysis of acting for the right reasons.
cannot move you in a non-accidental way. In these cases you will have to get
lucky in order to do what you ought.
The second argument builds on the first. It turns on a principle tying
credit to acting for the right reasons.18 It is plausible that token actions are
creditworthy just in case they are performed for the right reasons. We can
see this by reflection again on Good Husband and Bad Husband.
Brandon only acts for the right reasons in Good Husband. Given Sensitive
No Accident, it follows that only in Good Husband does Brandon do what
he ought in a non-accidental way. It is because of this that it is plausible
to think that Brandon’s token act is creditworthy only in Good Husband.
After all, in Bad Husband it is an accident that Brandon performs the action
that he is obligated to perform. It is just a coincidence that what Brandon
is motivated to do happens to align with what he is obligated to do. This is
not so in Good Husband. Brandon performs the action he does precisely
because there are decisive reasons to. This kind of non-accidentality seems
required for creditworthiness.
This supports Credit:
Credit: A’s φ-ing is creditworthy just in case A φs for reasons that
make φ-ing permissible.
In cases where one is obligated to φ, it follows from Credit that a token
φ-ing is creditworthy just in case it is performed for the reasons that obli-
gate one to φ.
If Credit is true and Right Reasons Ability Condition is false, then there
will be cases where one ought to φ even though one is unable to φ in a
way that would be creditworthy. This is intuitively unsatisfying. It is not
plausible that one can be obligated to φ even though one couldn’t φ in a
normatively kosher way. It is implausible that there are cases where the best
one can do is get completely lucky.
Denying the Right Reasons Ability Condition thus has at least two major
costs. First, one has to deny that one is always able to perform the action
one ought to perform in a non-accidental way. Second, one has to deny
that one is always able to do what one is obligated to do in a way that is
creditworthy.
It is tempting for the objectivist to reply to these arguments by again
appealing to the fact that the deontic comes apart from the hypological.
18
Perhaps the debate where the notion of acting for the right reasons crops up the
most is the debate about moral worth (see, e.g., Arpaly (2003) and Markovits (2010)).
I think that moral worth is too narrow a notion to cover what I mean to cover. Plenty of
actions are creditworthy that are not morally worthy—e.g., actions that are required for
prudential reasons.
38 Errol Lord
Credit is a hypological notion. Thus, since the deontic doesn’t necessarily

align with the hypological, we shouldn’t expect that it’s always possible to
do what’s obligated in a creditworthy way.
In reply, note that this is a much more radical severing of the deon-
tic from the hypological than the one we already granted. Recall that it
is quite plausible that blameworthiness doesn’t necessarily co-travel with
wrongdoing. It’s possible to be blameless even though you’ve done some-
thing wrong. By denying the Right Reasons Ability Condition, one makes a
much more radical claim. Rather than saying that in some cases it’s possible
for the deontic and hypological to come apart, one is saying that in some
cases it’s impossible for the deontic and hypological to stay together. No one
denies that in cases where it’s possible to be blameless even though you do
something wrong, it’s also possible to be blameworthy even though you’ve
done something wrong. To deny the Right Reasons Ability Condition is to
hold that in some cases it’s impossible to do the right thing in a way that is
creditworthy. Thus, I don’t think that what I’m willing to grant about the
deontic and hypological opens the door very wide for this type of response.
Denying the Right Reasons Ability Condition is costlier than that.
2.3.2 In Defense of (2)
I suspect that objectivists are so far unperturbed. This is because they feel
no need to deny the Right Reasons Ability Condition. They can accept it as
long as they hold a liberal view of what it takes to have the ability to act for
the right reasons. In this subsection I will argue that in order to act for the
right reasons, one must possess those reasons.
The easiest way to see why possession is necessary is by considering pairs
of cases. Delusional Andy and Surprised Andy is one such pair.19
Delusional Andy
Andy knows that his wife has always been an extremely loyal person.
He also knows that he has no reason to think that she is cheating on
him. Despite this knowledge, he does believe that she is cheating on
him. He thus files for divorce. In fact, his wife is cheating on him.
Surprised Andy
Andy knows that his wife has always been an extremely loyal per-
son. However, much to his surprise, he learns that she is cheating
on him—her best friend tells him, he finds some love letters, and he
catches his wife with her lover. He thus files for divorce.
19
Cf. Hyman (2006), Hornsby (2008), Gibbons (2013).
In both cases, Andy reasons from a belief that his wife is cheating on
him to the act of filing for divorce. Indeed, we can suppose that delusional
Andy’s deliberation is phenomenologically indistinguishable from surprised
Andy’s. Moreover, in both cases Andy’s belief is true. Finally, the fact that
his wife is cheating on him is a weighty reason to perform that action.
However, it’s very plausible that only in Surprised Andy does Andy file for
divorce because his wife is cheating on him. That is, it’s very plausible that
only Surprised Andy acts for the right reason.
We have several ways of testing this intuition. First, we have Sensitive
No Accident and Credit. If Credit is true and Andy acts for the right reason
in Delusional Andy, then Andy’s token act of filing for divorce should be
creditworthy. But it intuitively isn’t. Andy is delusional in Delusional Andy!
Despite the fact that he performs the best action, he does not deserve credit
for it. This is because he is just lucky that the act he actually performed
turned out to be the best one.
Second, we have the Sensitivity Condition. If Andy acts for the right
reason in Delusional Andy, then he is sensitive to the support relation
between the fact that his wife is cheating on him and the action he per-
forms. Intuitively Andy is not sensitive to that fact in the right way. He
has no legitimate contact with that fact. He is just lucky that his irrational
belief happens to be true. Because of this, it is hard to see how he is sensi-
tive to that fact in the right way.
Third, we have the Explanation Condition. If Andy acts for the right
reason in Delusional Andy, then those reasons explain why he’s justified for
filing for divorce. This doesn’t seem right, either. His token act is not even
justified. Thus, the fact that his wife is cheating on him can’t explain why
he’s justified. Since he isn’t sensitive to that fact, it doesn’t seem like it can
explain his actions.
We should come to the opposite conclusions about Surprised Andy.
His action in that case does seem creditworthy, does seem sensitive
to the relevant facts, and does seem to be explained by the fact that
his wife is cheating on him. The only relevant difference between the
two Andys is that in Surprised Andy the relevant fact is within Andy’s
epistemic ken. Thus, it seems like in order to act for the right reasons,
those reasons have to be within your ken. If this is true, then it is very
plausible that in order to act for the right reasons, you have to possess
those reasons.
This, of course, does not yet show that in order to be able to act for the
right reasons, you have to possess those reasons. I think that this is a plau-
sible step to take. After all, Delusional Andy sure seems to be exercising all
the abilities he has when it comes to the action at hand. He isn’t, as it were,
holding anything back. So if his actions are not done for the right reasons,
40 Errol Lord
I think it is plausible that he lacks the ability to act for the right reasons.20
Moreover, given the fact that Surprised Andy does act for the right rea-
sons, it is plausible to conclude that what Delusional Andy is missing is
possession.
At this point we should consider an important objection to this defense of
(2). The anchor of the objection is the obvious fact that even in Delusional
Andy, the consideration that Andy’s wife is cheating on him plays an impor-
tant role in Andy’s deliberation and subsequent action. There is a sense in
which the reason for which Andy files for divorce is that his wife is cheating
on him. That is the thought that ultimately motivates Delusional Andy to
file for divorce. Why not think that having the ability to do this is sufficient
for meeting the Right Reasons Ability Condition? Isn’t having the ability to
be motivated by the relevant considerations all that is required?
I agree that there is a sense in which the reason for which Andy files for
divorce is that his wife is cheating on him. We can make his actions intelli-
gible by citing that consideration. We can, that is, understand why he acted
the way he did rather than in some other way by appealing to the content
of his belief that his wife is cheating on him. We can do this even though he
is delusional. To give it a name, the consideration that his wife is cheating
on him is his rationale for filing for divorce.
As it happens, most theories of acting for reasons are theories of intel-
ligibility. These theories seek to understand what the two Andys have in
common. Moreover, they all hold that delusional Andy’s belief that his wife
is cheating on him plays an important part in explaining why he acted as
he did.21 So it is not unmotivated to think that there is a sense in which the
reason for which Delusional Andy acts is that his wife is cheating on him.
The question now is whether having as one’s rationale a consideration
that happens to be a normative reason is sufficient for acting for the right
reasons. If it is, then Delusional Andy does act for the right reason and
hence does have the ability to act for the right reason. If this is right, then
the objectivist can deny (2) while retaining (1).
Not surprisingly, I don’t think that having as one’s rationale a considera-
tion that happens to be the right reason is sufficient for acting for the right
20
We can make this more precise once we have a precise view about what his abilities
consist in. My preferred view is that they consist in some dispositions. Given that view,
my point is that it is plausible he lacks the dispositions that constitute the ability to act
for the right reasons because the relevant psychological manifestation conditions are met,
there aren’t any obvious finks or masks, and yet the relevant disposition is not manifested.
21
The role it plays is different in different theories. I’ll be assuming something like
Dancy’s (2000) view, which holds that the content of the belief is the motivating reason.
For pushback on the thought that we can assimilate the two Andys, see Gibbons (2013),
Hornsby (2008), Lord (2013).
reasons. The problem is that intelligibility comes cheap. It is because of this

that having a normative reason as one’s rationale is not sufficient for acting
for that normative reason.
The following variation on Mine Shaft makes this vivid.
Random Picking
Everything that is true in Mine Shaft is true in this case. Moreover,
Billy is disposed to sandbag A if she believes the miners are in A and
disposed to sandbag B if the miners are in B. She has also recently
taken a pill that will arbitrarily cause her to either believe the miners
are in A or the miners are in B. The pill kicks in and she believes the
miners are in A. She acts on this belief and sandbags A. As it happens,
the miners are in A.
Clearly Billy doesn’t act for the right reason in this case. However, her
act can be made intelligible by citing the consideration that the miners are
in shaft A. Moreover, she treats this consideration in the right way. Indeed,
she isn’t that much different from Delusional Andy. They both reason—in
the right way—from an irrational belief to an action. It seems clear upon
reflection that they lack the ability to act for the right reasons.
The important feature of cases like Delusional Andy and Random Picking is
that the characters’ beliefs lack a certain epistemic pedigree. Thus far we haven’t
confirmed that this is the same pedigree involved with possession. We have
ruled out that merely believing some proposition and acting on it in the right
way is sufficient for acting for the right reasons.22 What seems to be required is
that agents stand in some positive epistemic relation to the reason.
To argue for any specific view about which positive epistemic relation is
involved would be to break my neutrality about what the possession rela-
tion involves. I won’t do that. I am content to leave it here: It is very plau-
sible that possession requires a positive epistemic pedigree and it is very
plausible that having the ability to act for the right reason requires a positive
epistemic pedigree. It would be quite odd if it wasn’t the same epistemic
pedigree. I think this is enough to infer that in order to meet the Right
Reasons Ability Condition, one must possess the right reasons.23 In other
words, (2) is true.
From (1) and (2) we can infer that if you ought to φ, then you possess the
right reasons to φ. Since possession requires that the reason be within your
perspective, it follows that objectivism is false.
22
I think this result can be parlayed into an argument against certain views of posses-
sion. See Lord (2013: ch. 3).
23
I should stress that I don’t think the story ends here. I go much further in Lord
(2013: ch. 3). In order for the story to continue I have to take sides on the nature of pos-
session, which I do not want to do here.
42 Errol Lord
This argument doesn’t immediately establish the view that the reasons
you possess determine what you’re obligated to do. It just establishes that
possession is necessary for a reason to obligate. It doesn’t establish that pos-
sessing a set of reasons that conclusively support φ-ing is sufficient for those
reasons to obligate you to φ. While it doesn’t establish this, it is telling. The
most natural view to take once you’re on board up to this point is that the
reasons you possess determine your obligations. I will rest content with
establishing the necessary condition and hence showing that objectivism
is false.
2.4 NEW INFORMATION PROBLEMS
My argument for perspectivalism notwithstanding, there are some powerful

arguments in the literature for objectivism. In the last half of the chapter
I will provide new replies to these arguments.
2.4.1 Two New Information Problems
2.4.1.1 The Past Obligations Problem

Suppose that right before the moment of truth, Billy figures out a way to
determine where the miners are. She comes to find out that the miners are
in shaft A. What should she think she was obligated to do before she found
out this information? It is quite intuitive that she should think that her
obligations didn’t change. Rather, she discovered what her obligations were
by discovering where the miners were.24 As Ross (2002: 32) puts it, ‘Many
people would be inclined to say that the right act for me is … that which on
all the evidence available to me I should think to be my duty. But suppose
that from the state of partial knowledge in which I think act A to be my
duty, I could pass to a state of perfect knowledge in which I saw act B to be
my duty, should I not say “act B was the right act for me to do”?’
This thought is in tension with perspectivalism. This is because the per-
spectivalist is committed to thinking that before she got the new infor-
mation, Billy’s perspective best supported doing nothing. Moreover, when
she gained the new information, her perspective best supported blocking
A. Thus, it seems that Billy must be mistaken if she thinks that her obliga-
tion at the earlier time was the same as her obligation at the later time.
24
This is often taken as data, but not always. Some have argued that these types
of hindsight judgments are incorrect. See Bjornsson and Finlay (2010) and especially
Dowell (2013).
It seems that we can confirm that my perspectival view makes these pre-
dictions. Before finding out where the miners are, the reasons she possesses
conclusively support doing nothing. After she finds out, this is no longer
true. After she finds out, the reasons she possesses decisively support block-
ing shaft A. So if Billy thinks she’s discovered what her obligation was all
along, she is mistaken, according to my view.
2.4.1.2 The Advice Problem

When we seek advice about what to do, we don’t seek advice about what
our perspectives best support. We want to know what’s best, not what’s best
given our perspective. To see this, suppose that Billy’s method of figuring
out where the miners are is to ask you what she ought to do. She has found
out, suppose, that you know where the miners are. It would be a mistake
for you to tell her she ought to do nothing. You should tell her to block
shaft A. As Thomson (1986: 179) bitingly puts a similar idea, ‘On the rare
occasions someone conceives of asking my advice on a moral matter, I do
not take my field work to be limited to a study of what he believes is the
case: I take it incumbent upon me to find out what is the case.’ Similarly,
we might think, an advisor’s field work (whether it’s a moral matter or not)
is not limited to what the evidence or knowledge of the advisee suggests is
best; it’s what is best.
Again, this seems to be in tension with perspectivalism. If perspectivalism
is correct, then Billy ought to do nothing. So if she asks you what she ought
to do, you should tell her she ought to do nothing. Since you should in fact
tell her she ought to block A, it is very plausible that she ought to block
A. Thus, perspectivalism must be false.
Once again, it seems as if my view makes the bad predictions. Before
Billy gets the new information, the reasons she possesses conclusively sup-
port doing nothing. So this seems to be the answer to give when she asks
what she ought to do. This is not what you should say. You should tell her
she ought to block shaft A.
2.4.2 A Diagnosis
I grant that these arguments have great appeal. What I want to know right
now is why they have such appeal. For it is quite puzzling, to me at least,
that one’s intuitions about sophisticated ignorance cases can be tossed and
turned so easily. It is very plausible, even upon reflection, to think that Billy
ought to do nothing when she is ignorant. However, it also seems plausible
that if she is relieved of her ignorance, her judgments about what she was
obligated to do while ignorant should match her judgments about what
44 Errol Lord
she ought to do with more information. It is especially plausible that when

giving advice one should not always focus on the epistemic situation of the
agent one is giving advice to, even when one is giving advice to a character
in a sophisticated ignorance case. What gives?
What we want first is an explanation for why it is our hindsight judg-
ments and advice giving practices behave this way. I think they do because
deliberation aims at what’s best or what’s supported by all the reasons. This
is why Billy thinks that her past obligation is the same as her current
one—because her current one is getting at what she was aiming at all along.
Moreover, it is plausible that advice is parasitic on our deliberative aims.
That is, correct advice is guided by the aims of deliberation. This is why we
seek to inform the advisee what’s best or what’s supported by the balance of
all the reasons when we give advice.
The million dollar question is whether my perspectival view is compat-
ible with the claim that deliberation aims at what’s best. I think that it is.
Indeed, I think that it is also compatible with thinking that Billy’s hindsight
judgment and your advice about what Billy ought to do are true.
I think there are two burdens here. The first is to explain what is going
on with Billy’s hindsight judgment and with the advisor’s assertion. Are
they true? Is this compatible with perspectivalism? The second burden is to
show that perspectivalism is compatible with the intuition that motivates
our judgments—viz., that deliberation aims at what’s best. I’m going to take
these in reverse order. I will first argue that my view is compatible with the
claim that deliberation aims at what’s best. I will then explain why I think
the hindsight judgment and the advice are true and how this relates to
perspectivalism.
2.4.3 Abilities, Obligation, and ‘Ought’

My solution to the new information problems has two parts. The first part
is about the metaphysics of obligation. The second part is about the semantics
of ‘ought’ in English. I think that the two come apart in predictable ways
given the aim of deliberation. When it comes to the metaphysics, my view
is compatible with thinking that the aim of deliberation is to do what’s
best. Our ‘ought’ thought and talk tends to track the aim of deliberation,
so to speak, and in contexts where one party has more information, this
will lead us to make judgments about our obligations that come apart from
our deliberative obligations. This is why the truth-values of the English
sentences relevant to the evaluation of the hindsight judgment and advice
come out true. Let’s start with the metaphysics.
2.4.3.1 New Information as a Problem for the Metaphysics

The first point to make is that it is not clear that the new information prob-
lems speak decisively in favor of objectivism. This is because there are cases
that provide the same lesson even though the new information provided
doesn’t put us in a position to know what is best. Consider a version of
Mine Shaft where Billy starts out being more ignorant than in the original
case (new details in bold).
More Ignorant Mine Shaft
A group of 10 miners are trapped in a mine. They are either trapped
in shaft A or in shaft B. It is not known which shaft they are in. Flood
waters are approaching the shafts. Billy has the choice to sandbag
shaft A, sandbag shaft B, or not sandbag either. She knows that if she
sandbags A and the miners are in A, all the miners will survive. She
has strong but misleading evidence, however, that if she sandbags
A and they aren’t in A, nothing bad will happen and vice versa for
getting it wrong about B. Finally, she knows that if she does nothing,
then 9 of the 10 will survive.
Given what Billy knows, doing nothing is definitely not what she ought
to do. She ought to block A or B. Now imagine that you know that if she
blocks A and the miners are in B, then all the miners will die and you know
that if she blocks B and the miners are in A, all will die. If she asks you what
the thing to do is, you should tell her to block neither shaft, even though
what she ought to do given her perspective is block A or B.
We can say similar things about Billy’s thoughts about past obligations.
Suppose you tell her the new information and it becomes true that from her
perspective she ought to block neither shaft. It would be natural for Billy to
think that she’s discovered what she ought to have done all along. She will
thus judge that her earlier thought about what she ought to do was false.25
The rub is that in this case both you and Billy know that blocking neither
shaft is not best. So it seems that the new information problems don’t show
that the nature of deliberation and advice entail or even support objectiv-
ism. It seems as if the lessons can be learned by focusing on different sophis-
ticated ignorance cases. Back to this in a moment.
The second point to make is that it is very plausible that ability con-
ditions generally are compatible with the thought that deliberation aims
at what’s best. This is obvious when it comes to the Physiological Ability
25
Many who have a strong intuition in the original hindsight case don’t have as strong
an intuition in this case. This is some reason to doubt the veridicality of our intuition
about the original case. Again, I will grant the data for the sake of argument.
46 Errol Lord
Condition. Deliberation can aim at what’s best even though our obligations
are constrained by our physiological abilities. Deliberation can aim at what’s
best even though we aren’t always obligated to bring about the best state of
affairs because sometimes we don’t have the physiological ability to bring
about the best state of affairs.
I think the same is true of the Right Reasons Ability Condition.
Deliberation can aim at what’s best even though our obligations are con-
strained by some of our agential abilities. That is, deliberation can aim at
what’s best even though we aren’t always obligated to bring about the best
state of affairs because sometimes we don’t have the agential abilities needed
to bring about the best state of affairs in a way deserving of credit.
So far we’ve seen that it is intelligible to think that my perspectival view
is compatible with thinking deliberation aims at what’s best, but we haven’t
been told explicitly why we should think this is true. I think that cases like
More Ignorant Mine Shaft provide some strong evidence that we implicitly
recognize the relevant constraints. I see no reason to think that in More
Ignorant Mine Shaft Billy doesn’t seek what’s best in her deliberation. Nor
is there any reason to think that you, her advisor, are eschewing the aim of
having Billy do what’s best. However, you recognize that pursuit of that aim
is constrained by the information within your perspective.
It’s helpful here to compare practical deliberation with epistemic delib-
eration and its aims. Plausibly, epistemic deliberation—deliberation about
what to believe—aims at the truth.26 Given this, you’d expect there to be a
new information argument for the conclusion that one is always delibera-
tively obligated to believe the truth. At the very least, epistemic advisors try
to advise their advisees to believe truths. And in cases where the advisee’s
information suggests ¬p and the advisor’s better information suggests p, the
good advisor should tell the agent they ought to believe p. Does this show
that we’re always deliberatively obligated to believe the truth?
No, this argument is bad. It is incredibly plausible that we are some-
times deliberatively obligated to refrain from believing the truth. There are
two relevant cases. In the first, we are deliberatively obligated to believe
something that is false. Sometimes the evidence available is misleading and
strongly supports believing p even though ¬p. In these cases, it’s plausible
that we are deliberatively obligated to believe p.
Even if you think that we are never deliberatively obligated to believe
a falsehood, it is still overwhelmingly plausible that we are sometimes
not deliberatively obligated to believe the truth. This is because we are
26
Some (e.g., Williamson (2000)) hold that epistemic deliberation aims at knowl-
edge. This won’t matter for my point.
sometimes deliberatively obligated to withhold belief. To take the easiest

case, when we lack evidence for both p and ¬p, we deliberatively ought to
withhold belief about p. Consider an example. There is a grassy field in one
corner of Central Park. The number of blades of grass in this field is either
odd or even. Consider the question of whether the number is odd or even.
If you’re like me, you have no evidence either way. Because of this, it is very
plausible to think that we ought to withhold belief on this question. If the
new information argument for objectivism worked, it seems like what we in
fact ought to do is either believe it is odd or believe it is even.27
It is worth emphasizing the implausibility of this result. If epistemic
deliberation aims at the truth and the new information argument is sound,
then we are forced to think that we are always obligated to believe the truth.
One result of this is that we are always obligated not to withhold since with-
holding is incompatible with believing. This is extremely implausible.
Thus, it doesn’t seem like obligations to withhold threaten the thought
that epistemic deliberation aims at the truth. What we’re after is the
truth even though sometimes the only permitted option is to withhold.
Intuitively, this is because we are only allowed to rely on certain information
in deliberation and, alas, sometimes that information doesn’t adequately
support either p or ¬p. Thus, it seems very plausible that our epistemic obli-
gations are constrained by our perspectives. This does not seem to threaten
the thought that epistemic deliberation aims at the truth.
The most important lesson for my purposes here is that in the epistemic
case it is very plausible that we aren’t always obligated to believe the truth
even though epistemic deliberation aims at the truth. There seem to be
constraints on our epistemic obligations. This is the structure I am suggest-
ing practical deliberation and obligation have. The epistemic case provides
a nice model of how I think the practical works. Appreciating how natural
the structure is in the epistemic case is the main lesson I want to glean from
the epistemic.
Thus, I don’t think that the new information problems present much
of a problem for the metaphysics of my perspectival view of obligation.
This is because my view can account for the claim that motivates the new
information problems, which is the claim that deliberation aims at what’s
best. Deliberation aims at what’s best even though our obligations are con-
strained in various ways by our abilities.
27
Notice that cases where we ought to withhold are just like sophisticated ignorance
cases. They are cases where we are in a position to know that the option that we ought
to take is second best. Given how plausible it is that this is the right answer in the epis-
temic case, we should be more confident that the sophisticated ignorance cases are indeed
counterexamples to objectivism.
48 Errol Lord
2.4.3.2 New Information and the Semantics of ‘Ought’

While the metaphysical problem is, I take it, the biggest problem posed by
new information, it isn’t the only problem. This is because so far we haven’t
accounted for the data in the new information cases. Namely, we haven’t
accounted for the fact (if it is a fact) that Billy’s judgment about her past
obligation—her judgment that her past obligation is the same as her obli-
gation after gaining more information—is true. We also haven’t explained
how it is that Billy’s advisor says something true when she tells her that she
ought to block shaft A.
It certainly seems like these bullets must be bitten. I think this seeming
is misleading. In order to see why we need to understand a bit about how
‘ought’ works in English.
On the canonical view of ‘ought’ in linguistics, ‘ought’ operates as a
quantifier over possible worlds.28 But it doesn’t (always) quantify over
all possible worlds. Rather, it quantifies over a restricted set of possible
worlds. Which set it quantifies over is determined by context. One way
in which context often restricts the domain is by restricting the amount
of information that can be taken as true. The most natural way this hap-
pens is by limiting the domain to the worlds compatible with some sali-
ent body of information. Sometimes this is just the knowledge of the
speaker and sometimes it’s the knowledge that a group of contextually
salient speakers has.29 When the domain gets restricted in this way, we
can have thoughts and talk about what one ought to do given some lim-
ited body of information.30 We often take advantage of this nice feature
of the word ‘ought.’
Given the role that context plays in the semantics of ‘ought,’ ‘ought’
thought and talk is flexible. We can think and talk about what ought to be
28
It is the canonical view mostly because of the work of Angelika Kratzer. See Kratzer
(2012). Recently there has been much debate about the role these relativizations play in the
semantics. Contextualists like Kratzer think that the relativization plays a role in determin-
ing the content of the propositions expressed, whereas truth-relativists like Kolodny and
MacFarlane (2010) hold that the content is contextually invariant but that the truth-value
is relativized to contexts of assessment in another way. This debate is orthogonal to our
discussion here. I will assume contextualism given that it is the canonical account.
29
There is a second way that context can play a role in the semantics. Namely, by
fixing which standards will be germane for the evaluation of the options. We can just
ignore this here and assume that the standards chosen are the standards that evaluate the
deliberative ‘ought.’
30
Sometimes this relativization to information will be explicitly contained in what’s
said (e.g., ‘Given what Billy knows, she ought to φ’). Most often, though, we just make
bare ‘ought’ claims and context determines the relativization.
done given X for a very large amount of Xs. This means that Billy can think
about her past obligations in light of her new information—she can think
what she ought to have done previously in light of what she knows now.
Moreover, it means that advisors can think about the obligations of advisees
from the perspective of their information—Billy’s advisor can think of what
Billy ought to do given the advisor’s information. I think these are the con-
tents of Billy’s thought and of the advisor’s thought (and assertion). Surely
those contents are true.
Doesn’t that show that perspectivalism is false? In a word: No. As we’ve
already seen, there are lots of true ‘ought’ claims in this case. It’s true rela-
tive to some bodies of information that Billy ought to do nothing, it’s true
relative to some bodies of information that she ought to block shaft A, and
it’s true relative to other bodies of information that she ought to block A or
block B. Those truths don’t necessarily establish anything about what she
deliberatively ought to do.
‘Fair enough,’ one might respond, ‘but this leaves out the important fact
that Billy and the advisor are having those true thoughts in a deliberative
context and, moreover, the content of Billy and the advisor’s thoughts seems
to be the answer to the central deliberative question. This provides very
strong evidence that Billy and the advisor really are getting at the delibera-
tive ought.’
This is a powerful response. However, I think it can be resisted. Those
who want to resist it have at least two burdens. First, they have to explain
why it is that Billy and the advisor’s thoughts don’t track Billy’s deliberative
obligations in these particular cases. This isn’t enough to be fully satisfying.
For once we have this explanation, we’ll want to know if it overgeneral-
izes. That is, we’ll want to know whether the explanation, if correct, shows
that we never or rarely track deliberative obligations. This would be bad.
It’s a very serious problem with a theory of deliberative obligations if it’s
committed to holding that our ‘ought’ thought and talk very rarely tracks
our deliberative obligations. So this commitment should be resisted. If it
is, then—and this is the second burden—one needs to explain why it is
that in these cases we don’t track our deliberative obligations but in most
cases we do.
I think both burdens can be met. Let’s start with the first: Why is it that
Billy’s and the advisor’s thoughts don’t track Billy’s deliberative obligations?
It is because Billy and her advisor are more concerned with what’s best
rather than what Billy’s deliberative obligations are in a more ignorant state.
This is not surprising given that the aim of deliberation is doing what’s
best. This is what we’re trying to get at in deliberation. Given that, it is no
surprise that the ‘ought’ judgments we are disposed to make will always be
50 Errol Lord
relativized to the best information available. What our deliberative obliga-

tions are given worse information is of no interest to us given our aims.31
This explanation doesn’t overgeneralize and thus the second burden can
also be met. If my metaphysical story above is correct, then deliberation
aims at what’s best even though our obligations are constrained by some of
our abilities. One of these constraints is tied to how much information one
has. Given this, what’s best now in light of the information currently had is
of great interest to a deliberator (and the advisors of the deliberator). True
thoughts about this will track one’s deliberative obligations. So the expla-
nation of why Billy’s thought about her past obligation and the advisor’s
thought don’t track Billy’s deliberative obligations doesn’t overgeneralize.
Since our obligations are constrained by the information we have and the
information we have will be all deliberation can go on, most of our true
‘ought’ thoughts will track our deliberative obligations. It is only when we
get differences in how much information is possessed between agents (or
time-slices of agents) that we get the two coming apart.
So, my response to the new information problems is twofold. First,
I think that new information does not put pressure on my view of the
metaphysics of obligation. This is because the motivating idea behind the
arguments—that deliberation aims at what’s best—is compatible with my
view. Deliberation can aim at what’s best even if obligation is constrained
by our abilities.
The second part of my answer has to do with the semantics of ‘ought.’
Given the flexibility of ‘ought,’ we can have all kinds of true thoughts about
what we ought to do. This means that our hindsight judgments can be
about what we ought to have done given the information we have now and
our advisors’ thoughts can be about what we are obligated to do given their
information. I think that these are the contents of our thoughts in hind-
sight cases and the content of our advisors’ thoughts. Those contents are
true. Moreover, there is a plausible story to be told about why our thinking
about what ought to have been done in hindsight and our advisors’ think-
ing about what we ought to do can come apart from thinking about our
deliberative obligations.
2.5 CONCLUSION
This chapter had two main ambitions. The first was to provide an argument
for perspectivalism. The anchor of that argument was that in order for a
Cf. Bjornsson and Finlay (2010).

31
reason to obligate, it has to be possible for that to be the reason for which
we act. I argued that a reason can be potentially action guiding in this way
only if we possess that reason. Thus, perspectivalism is true and objectivism
is false.
The second ambition was to respond to what I take to be the strong-
est argument against perspectivalism. I argued that the motivating thought
behind that argument is compatible with my perspectival view. Moreover,
I provided explanations of the key data that are both compatible with and
friendly to my view. 32
References
Arpaly, N. 2003. Unprincipled Virtue: An Inquiry into Moral Agency. Oxford: Oxford
University Press.
Bjornsson, G. and Finlay, S. 2010. ‘Metaethical Contextualism Defended,’ Ethics
121: 7–36.
Dancy, J. 2000. Practical Reality. Oxford: Oxford University Press.
Dowell, J. 2013. ‘Flexible Contextualism about Deontic Modals: A Puzzle about
Information-Sensitivity,’ Inquiry 56: 149–78.
Gibbons, J. 2013. The Norm of Belief. New York: Oxford University Press.
Graham, P. A. 2010. ‘In Defense of Objectivism about Moral Obligation,’ Ethics
121: 88–115.
Hornsby, J. 2008. ‘A Disjunctive Conception of Acting for Reasons,’ in A. Haddock
and F. Macpherson (eds.), Disjunctivism: Perception, Action, Knowledge.
Hyman, J. 2006. ‘Knowledge and Evidence,’ Mind 115: 891–916.
Jackson, F. 1991. ‘Decision-Theoretic Consequentialism and the Nearest-Dearest
Objection,’ Ethics 101: 461–82.
Kolodny, N. and MacFarlane, J. 2010. ‘Ifs and Oughts,’ Journal of Philosophy
107: 115–43.
Kratzer, A. 2012. Modals and Conditionals: New and Revised Perspectives. Oxford:
Oxford University Press.
Lord, E. 2010. ‘Having Reasons and the Factoring Account,’ Philosophical Studies
149: 283–96.
Lord, E. 2013. ‘The Importance of Being Rational,’ PhD thesis, Princeton University.
Lord, E. and Sylvan, K. n.d. ‘Prime Time (for the Basing Relation),’ manuscript.
Markovits, J. 2010. ‘Acting for the Right Reasons,’ Philosophical Review 119: 201–42.
Moore, G. E. 1912. Ethics. Oxford: Oxford University Press.
32
Thanks to Kurt Sylvan, Andrew Sepielli, Michael Smith, Tom Kelly, Gideon Rosen,
Nat Tabris, Daniel Wodak, Eden Lin, Robert Audi, David Enoch, Derek Baker, Steve
Sverdlik, Liz Harman, Joe Rachiele, Michael Titelbaum, Dustin Locke, and an anony-
mous referee.
52 Errol Lord
Neta, R. 2008. ‘What Evidence Do You Have?’ British Journal for the Philosophy of
Science 59: 89–119.
Parfit, D. 2011. On What Matters. Oxford: Oxford University Press.
Regan, D. 1980. Utilitarianism and Cooperation. Oxford: Oxford University Press.
Ross, J. 2006. ‘Acceptance and Practical Reason,’ PhD thesis, Rutgers University.
Ross, W. 2002 [1930]. The Right and the Good. Oxford: Oxford University Press.
Schroeder, M. 2009. ‘Means-End Coherence, Stringency, and Subjective Reasons,’
Philosophical Studies 143: 223–48.
Sylvan, K. and Sosa, E. Forthcoming. ‘The Place of Reasons in Epistemology,’ in
D. Star (ed.), The Oxford Handbook of Reasons and Normativity. Oxford: Oxford
University Press.
Thomson, J. J. 1986. ‘Imposing Risks,’ in her Rights, Restitution, and Risk: Essays in
Moral Theory. Cambridge, MA: Harvard University Press, 173–91.
Williamson, T. 2000. Knowledge and its Limits. Oxford: Oxford University Press
3
The Irrelevance of Moral Uncertainty
Elizabeth Harman
3.1 MORAL UNCERTAINTY
Consider this case:
B believes that φing is morally required, while failing to φ would be
morally wrong. B thinks that failing to φ wouldn’t be deeply morally
wrong; it would only be minorly morally wrong. B is only 90% sure
that φing is morally required. B has 10% confidence that φing is actu-
ally morally wrong, and indeed is deeply morally wrong.
What should B do? It might seem that B should φ; after all, B believes that
φing is morally required. But consider this line of argument:
B ought to refrain from φing. Taking a 10% chance is taking a sub-
stantial chance of doing a deeply morally wrong thing. One should
be very averse to risking doing a deeply wrong thing. It is better to
do what is very likely a minorly wrong thing to do than to risk doing
what is a deeply wrong thing to do. Suppose that B does φ, φing is
in fact morally wrong, and a victim of B’s φing were to later try to
hold B responsible for φing. Could B defend him- or herself by saying
“but I believed that φing was morally required”? No! B knew that B
was risking doing a deeply morally wrong thing, only for the sake of
what B believed was a minor moral requirement. For this reason, B
is blameworthy for φing, and it is appropriate to hold B responsible
for φing.
This way of thinking about cases like B’s is attractive, and there is an inter-
esting philosophical literature that takes this way of thinking to be correct
and then seeks to answer further questions that arise. I will call a proponent
of this line of argument an Uncertaintist, and the view that this line of
thought is correct Uncertaintism. According to Uncertaintism, an agent’s
54 Elizabeth Harman
moral uncertainty (and specific moral credences) are crucially relevant to

how the agent should act.1
Uncertaintism begins by considering and rejecting the following view: an
agent should be guided by any moral claims she believes. Cases like B’s
bring out that sometimes an agent takes a big moral risk by being guided by
what she believes. If her beliefs hold that very little of moral significance is
at stake, but she has a small credence in the claim that a great deal is mor-
ally at stake, then she may be taking a big moral risk in simply ignoring her
small credence. B’s φing would be like a homeowner’s failing to buy fire
insurance: one believes that one’s home won’t burn down, but the small
credence one gives to a fire makes it reasonable to buy fire insurance even
though there is simply a net loss if no fire occurs.
Uncertaintists then consider the following view: one should maximize
the expected moral value of one’s actions. This is not a consequentialist
view. Rather, the view is that some actions are morally worse than oth-
ers, according to various moral claims. One should maximize the expected
moral value of one’s action. On this view, B has 90% credence that φing is
morally good, but only minorly morally good, 90% credence that failing to
φ is minorly morally bad, 10% credence that φing is very morally bad, and
10% credence that failing to φ is very morally good. The overall expected
moral value of his action is maximized by failing to φ.
The view that one should maximize expected moral value faces a difficult
puzzle, which can be illustrated by the following case. Suppose that a per-
son has some credence in Utilitarianism and some credence in Kantianism,
and she is trying to decide whether to push one person in front of a train
to stop it from hitting five people. Utilitarianism holds that pushing is
morally required, while Kantianism holds that it is morally wrong. Does
Utilitarianism hold that letting five die is just as morally bad as the Kantian
holds that killing the one is morally bad? The answer to this question mat-
ters if agents should maximize expected moral value; but it is very hard to
answer it. More generally, it is hard to know how to compare the moral
values that different moral claims or principles assign to an agent’s options.2
This puzzle about how to compare moral value between conflicting
moral views is interesting, but I will argue that this literature is based on a
1
Uncertaintist thinking appears in Ross (2006) and Sepielli (2008, 2013). (Though
these authors do not appeal to considerations of blameworthiness to support their
claims.) Related thinking appears in Lockhart (2000), Guerrero (2007), and Moller
(2011); all three of these authors claim that an agent’s moral credences are relevant to
how she should act. My argument against Uncertaintism can be adapted to target these
three views, as I will explain in notes 5 and 6.
2
Lockhart (2000), Ross (2006), and Sepielli (2008, and n.d.) have offered solutions
or partial solutions to this puzzle.
The Irrelevance of Moral Uncertainty 55
mistake: the mistake of thinking that Uncertaintism is true. Once we see

that Uncertaintism is false, we will see that we do not really face this puzzle.
To see why Uncertaintism is false, let’s look more closely at the initial
line of argument I laid out concerning B, which the Uncertaintist endorses.
I will make three observations about the Uncertaintist’s view.
First: In saying that B should refrain from φing, the Uncertaintist is mak-
ing a moral claim about B, not a claim about how B might best pursue
B’s goals. (The description of B does not mention whether B cares to do
what is moral; though the description may imply that he does.) Suppose
we consider Bill, a version of B who doesn’t care about morality at all. We
might run through that same line of argument about Bill, finding it just
as compelling: Bill should be morally cautious, even though we know Bill
won’t be morally cautious.
Second: In saying that B should refrain from φing, the Uncertaintist is
making a moral claim that is subjective rather than objective. To see the dis-
tinction, consider the following case. Anne’s husband is dying and Anne
gives him what she has every reason to believe is the cure to his illness. In
fact it is poison. The following normative claims both seem truth of Anne:
Anne does something she shouldn’t do. (She poisons him.)
Anne does something she should do. (She gives her husband what she
has every reason to believe is his cure.)
The first normative statement is true, and it is an objective moral truth.
The second normative statement is true, and it is a subjective moral truth.
Loosely speaking, objective moral truths are made true by features of the
world independently of anyone’s (perhaps mistaken) beliefs about what the
world is like, while subjective moral truths are made true by someone’s
mental states (which may be mistaken), often but not always the agent’s
mental states.3
When a subjective moral claim “A ought to refrain from φing” is true
and is made true by A’s beliefs and credences (rather than by an advisor’s
or an observer’s beliefs and credences), one can express the same truth by
saying “Given A’s whole mental state, A ought to refrain from φing.” That is
what the Uncertaintists are saying. They are saying: given the entirety of B’s
beliefs and credences, B ought to refrain from φing.
3
There is a rich literature on subjective and objective normative statements, and on
subjective claims made by advisors and remote observers. Subjective statements may also
have true readings relative to an agent’s evidence rather than her credences (see section 3.7).
See Dowell (2013), Jackson (1991), Kolodny and MacFarlane (n.d.), MacFarlane (2014),
Smith (n.d.), and others.
56 Elizabeth Harman
My third observation is that the Uncertaintist is committed to the fol-

lowing claim: if someone is caught in the grip of a false moral view (if she
is sure of that view), then she ought to act as her false moral view requires.
This is simply the limiting, uninteresting case of one’s moral credences
determining how one should act. According to Uncertaintism, if one is sure
that one is morally required to φ, then one should φ.
I will now argue that Uncertaintism is false. Uncertaintism is the view that
in light of an agent’s whole mental state—given all of her credences—she
should do the morally cautious thing in cases like B’s. Furthermore, some-
one caught in the grip of a false moral view should do what that moral view
holds is morally required. These are subjective moral claims, made true by
the agent’s whole mental state. But the following principle holds regarding
such subjective moral claims:
An agent is blameworthy for her behavior only if she acted as she
subjectively should not have acted.
Consider Anne, who acts wrongly by poisoning her husband. She is not
blameworthy because it is not the case that she subjectively should not have
acted that way. Subjectively, she should have given him what she took to
be the cure.
Of course, Anne might be blameworthy for causing the poisoning if she
is blameworthy for having the credences she has. If Anne didn’t look at
the bottle before giving it to her husband, for example, she may be blame-
worthy. But Anne would not be blameworthy for behaving as she did; she
wouldn’t be blameworthy for poisoning him. Rather, she would be blame-
worthy for causing the poisoning to occur by failing to look at the bottle.4
Uncertaintism makes some subjective moral claims. They have impli-
cations for blameworthiness. They imply, for example, that in B’s case, if
φing is indeed morally required, as B believes, but B does not φ, then B
is blameless for failing to φ, because B did as B subjectively ought to have
done. According to Uncertaintism, B’s failure to φ would be blameless for
the same reason that Anne’s poisoning of her husband is blameless: they did
the right thing, given their beliefs and credences.
4
My discussion here illustrates a terminological choice I have made in this chapter.
I distinguish blameworthiness for behavior, which I construe narrowly, from blameworthi-
ness for causing that behavior. Other authors (including myself in other papers) count
both kinds of blameworthiness as blameworthiness for behavior, sometimes distinguish-
ing them as “original blameworthiness” and “derivative blameworthiness,” respectively.
Nothing hangs on which kind of terminology one uses, but it is important to bear in
mind throughout the chapter that when I say a view implies that a person is blameless for
her behavior, this leaves open that she may be blameworthy for having caused herself to
engage in that behavior (she may be derivatively blameworthy for the behavior).
The Uncertaintist is committed to the view that those caught in the grip
of false moral views, who do morally wrong things while sure that those
things are morally required, are blameless for their behavior. According to
Uncertaintism, these agents acted as they subjectively should have acted.
Uncertaintism is committed to the view that being caught in the grip of a
false moral view exculpates. That view is false, I claim, and so Uncertaintism
is false.
My Main Argument:
1. Uncertaintism implies that being caught in the grip of a false moral view
is exculpatory.
2. It is not true that being caught in the grip of a false moral view is
exculpatory.
Therefore:
3. Uncertaintism is false.
This argument has a controversial premise: premise 2. I won’t seek to estab-
lish that premise in this chapter, but in section 3.3, I will say some things
to defend it and to make it seem plausible. (In section 3.6, I will discuss
whether premise 1 can be denied.)5
In section 3.2, I will discuss the implications of my main argument by
offering and explaining an alternative to Uncertaintism. In section 3.3,
I will support my main argument by explaining and defending premise
2. In sections 3.4–3.7, I will discuss objections to my arguments. In section
3.4, I also offer another argument against Uncertaintism.
5
My main argument can be adapted to target the views of Lockhart, Guerrero, and
Moller. Lockhart claims that an agent should minimize her chances of acting wrongly.
Guerrero and Moller claim that, at least for cases of killing, an agent should avoid doing
something she believes may well be morally wrong, if she believes her alternative is defi-
nitely morally permissible. As I read these authors, all three of their views imply that
an agent caught in the grip of a false moral view should act as her moral view dictates;
thus, all are committed to the view that being caught in the grip of a false moral view is
exculpatory. (Though these authors do not necessarily embrace this commitment or agree
with me that they are so committed.)
Weatherson (2014) argues against views along the lines of Lockhart’s, Moller’s, and
Guerrero’s by arguing that such views are implausible in cases of prudential uncertainty
and then arguing that moral uncertainty is more analogous to prudential uncertainty
than to non-moral uncertainty; he argues that such views inappropriately fetishize either
prudence or morality, relying on Smith (1994).
58 Elizabeth Harman
3.2 THE ALTERNATIVE TO UNCERTAINTISM
In this section, I will discuss the implications of my argument.

Because Uncertaintism is false, the puzzle we discussed above, about how
to compare moral value between conflicting moral views, is not important.
It may be interesting as a puzzle; but nothing normatively important hangs
on solving it.
If Uncertaintism is false, what is true in its place? Here is my proposal:
Actualism: A person’s moral beliefs and moral credences are usually
irrelevant to how she (subjectively) should act. How a person (sub-
jectively) should act usually depends solely on her non-moral beliefs
and credences; her moral beliefs and credences are relevant only inso-
far as they provide warrant for beliefs and credences about what her
non-moral situation may be.6
Why is “Actualism” a good name for this view? Because according to
Actualism, what a person (subjectively) should do depends crucially on
what’s actually the true moral theory, and not on what people believe is
the true moral theory. The contrast is between what is believed and what is
actual. “Actual” here doesn’t have anything to do with contingency, as our
moral requirements (at a sufficient level of generality) are necessary.7
What does Actualism say about B? As B’s story stands, it is too underd-
escribed to settle what B should do. Different ways of filling out B’s story
(or a similar story) will lead to very different results as to how B should act.
Consider this way of filling out B’s story:
Barbara is a police officer who has bad marksmanship and has promised
her superior that she’ll practice today. She is at a shooting range, decid-
ing whether to shoot at the target in front of her. She is 90% sure that
this is a normal shooting range environment, in which no one would
be in any danger if she shoots. But she has a 10% credence that there is
a cleaning person moving around behind the targets, who would very
likely be shot if Barbara were to shoot at the target in front of her. If the
shooting range were empty, Barbara would be morally required to prac-
tice. If the shooting range had a person behind the targets, it would be
6
Actualism implies both that Uncertaintism is false and that the related claims of
Lockhart, Guerrero, and Moller are false. (See note 5.) In section 3.7, I offer a revised
statement of Actualism which explicitly denies two possible revisions to Uncertaintism;
I endorse both the initial and the revised statements of Actualism.
7
“Actualism” is used as a name for various philosophical views, including even a view
in ethics. I’m not using it in any of these already existing ways. I’m using it in a new,
stipulative way.
morally wrong to shoot. Barbara has a 90% credence that she’s in a sce-
nario in which it is morally required that she shoot (to practice, and to
keep her promise) and she has 10% credence that she’s in a scenario in
which it is morally required to refrain from shooting. (These are the cre-
dences she has regarding the objective moral truth about her situation.)
In this case, it is obvious that Barbara should refrain from shooting, for the
reasons given in the initial line of argument we considered. Barbara should
not take the moral risk of killing someone just to be sure to practice and
keep her promise. But what’s important, according to Actualism, is not that
Barbara believes (as any ordinary person does) that shooting an innocent
cleaning woman in this context would be morally wrong, but simply that
she knows that the cleaning woman might be there. What makes it the
case, according to Actualism, that Barbara should refrain from shooting is
simply that she’d be taking a risk of killing someone, not that she’d be doing
something that she knows is taking a risk of doing something objectively morally
wrong—though that is also true in this case.
Now consider this story which is very much like B’s story:
Bob’s daughter Sue has been asking him to teach her to drive and he
has finally promised to do so. He already taught her twin brother to
drive. Bob has a 90% credence that he is morally required to teach
Sue to drive. But Bob has recently been listening to some conservative
speeches about the morally appropriate place of women in society.
According to the conservative speaker he’s been listening to, women
should not drive and no one should teach a woman to drive; in fact
that is a grave moral wrong. The conservative speaker does not chal-
lenge any non-moral facts Bob already believes; the challenge is sim-
ply to Bob’s normative beliefs. Bob thinks the conservative speaker is
probably wrong; he’s 90% sure of that. But Bob finds the conservative
picture being offered somewhat compelling, so that he is 10% sure it
is the correct picture. (He’s quite sure that a compromise position is
false; so he’s simply torn between the liberal ideals he grew up with
and the more conservative picture he’s learning about.) Bob has a 90%
credence that failing to teach Sue would be wrong, but not very seri-
ously wrong; he has a 10% credence that teaching her to drive would
be deeply morally wrong. (These are the credences he has regarding
the objective moral truth about his situation.)
Uncertaintism holds that, if the conservative picture holds that teaching a
woman to drive is wrong enough, Bob should not teach Sue to drive. He
would be taking a moral risk that he should not take.
But that is false. Bob should teach Sue to drive. Breaking his promise
would be treating her badly, just because she is female, in a way that a father
should not mistreat his daughter. (Let’s not be distracted by the fact that it
60 Elizabeth Harman
wouldn’t be a minor wrong to Sue, so it’s not an elaboration of the initial

case of B (but rather is a similar story). That’s true. But just so long as the
conservative picture holds that teaching a woman to drive is wrong enough,
then Uncertaintism will imply that Bob should not teach Sue to drive.)
Bob knows his situation. He knows that he would be treating his daugh-
ter differently from his son, simply because she is a woman. He knows that
women are not inherently less good drivers, or less intelligent than men.
(Remember, the conservative picture does not challenge the non-moral
beliefs that Bob already had as a liberal.)
If Bob refuses to teach Sue to drive, breaking his promise, her resenting
him for it would be reasonable and appropriate. Even if she knows that he
refrains because he does not want to take this moral risk, her resentment is still
appropriate. Bob would be blameworthy for refusing to teach Sue to drive.
These are the claims that Actualism makes about the case. Whether they
seem plausible partly depends on what one thinks about whether being
caught in the grip of a false moral view is exculpatory. If it is, then being
“partly caught” in the grip of a false view, as Bob is, would also seem to
be exculpatory. If it is not, then being “partly caught” is not exculpatory.
Again, the dispute between Actualism and Uncertaintism hangs at least in
part on a dispute over whether being caught in the grip of a false moral view
is exculpatory. I’ll say more about that in section 3.3.
Here is another way of filling out B’s case:
Bernard is a slaveholder in the early 1800s in the United States. Like
every slaveholder in his community, once a week he beats his slaves if
they have failed to do their work as well as he believes they can. He
believes that he is morally required to continue this practice, because
(as he has been taught) it is a crucial ingredient in sustaining his com-
munity’s way of life, which he believes he is morally required to do.
But lately Bernard has heard some compelling abolitionist speeches.
He thinks abolitionism is misguided; but he’s not sure. Bernard is
90% sure that he is morally required to beat his slaves if they disobey,
though it is not seriously morally wrong to refrain. He is 10% sure
that it is deeply morally wrong to beat his slaves. (These are his cre-
dences in the objective moral truth about his situation.)
What should Bernard do? As Uncertaintism says, Bernard should refrain
from beating his slaves. That’s true. I believe that cases like Bernard’s provide
crucial and central motivation for Uncertaintism. It does seem that Bernard
should refrain from beating his slaves. Actualism holds that Bernard should
refrain from beating his slaves, but not because of his moral credences. He
should refrain because he knows it is beating his slaves, he knows it is beat-
ing people who are kept captive and who work for him without compensa-
tion and without freedom. One should not beat a person in this situation.
(One should almost never beat another person, of course. Nor should one
keep slaves.) Bernard should refrain because he knows what he is doing,
Actualism holds, and because what he is doing is actually wrong.
Uncertaintism may be motivated in part by the thought that Bernard is
blameworthy if he beats his slaves. But Actualism agrees that he is blame-
worthy and that he should not beat his slaves; the disagreement is over what
explains these truths.
Here is another way of filling out B’s case:
Betsy is presented with the option of pressing a red button, without
knowing what the button does. Then an advisor speaks. The advi-
sor says, “I’m 90% sure that pushing the button is morally required,
though it wouldn’t be a grave moral wrong to fail to push it. I’m 10%
sure that pushing the button is gravely morally wrong.” That is all the
advisor says. Betsy knows this advisor well. Betsy in fact has sound moral
views, and knows that the advisor does too.
Betsy should be cautious and refrain from pressing the button, for just the
reasons offered in the initial line of thought we considered. Pushing the but-
ton is too risky. Betsy has reason to believe that the advisor is uncertain what
pushing the button does, but that the advisor has a 10% credence that push-
ing the button does one of a number of things that the advisor and Betsy
both think are gravely wrong—and these things are wrong, because they have
sound moral views—while the advisor likely has 90% credence that pushing
the button does one of a number of things that the advisor and Betsy both
think are morally required, but not seriously morally wrong to do—and those
things really are morally required but not seriously wrong to do. So, Betsy
should have 10% credence that pushing the button is: killing an innocent
person, seriously harming an innocent person, or etc. And she should have
90% credence that pushing the button is: breaking a minor promise, dividing
a benefit unfairly, or etc. When a person’s credences are distributed that way,
her case is very much like Barbara’s case, and she should be morally cautious.
What’s crucial to our understanding of Betsy’s case is that her moral
uncertainty gives rise to non-moral uncertainty. Because she believes the
advisor, she comes to be in a state of non-moral uncertainty that makes it
reasonable to be cautious and refrain from pushing the button.
But consider this variant, involving an agent who holds a true moral view
about helping hurricane victims but a false moral view about gay marriage:
It is the final moments of a U.S. state’s legislative session. Unless a bill is
delivered to the statehouse by midnight, the bill cannot be signed by the
governor. In the statehouse, Ben is presented with the option of pressing
a red button. He knows that the button delivers a piece of legislation to
the governor, but he does not know what the legislation is. He knows that
62 Elizabeth Harman
two pieces of legislation were before the legislature, one providing aid to
towns recently hit by a hurricane and one allowing gay marriage in the
state. The governor has pledged to sign each bill if it is delivered to him
in time. An advisor says, “I’m 90% sure that pushing the button is mor-
ally required, though it wouldn’t be a grave moral wrong to fail to push
it. I’m 10% sure that pushing the button is gravely morally wrong.” That
is all the advisor says. Ben knows that he and the advisor agree about the
morality of both pieces of legislation: they agree that it is good to aid the
hurricane victims; and they agree that it is seriously morally wrong for the
state to allow gay marriage or for anyone to do anything to aid the state
in allowing gay marriage. Ben is an ordinary person with no duties in the
legislation, but due to an odd computer set-up, he alone is able to push
this button in time to get the legislation, whatever it is, to the governor. If
the governor does not get the legislation today, the voting will be re-done
in one month by a newly constituted group elected in a recent election,
who are expected to decide differently on both bills. In fact, both bills are
good bills that should be enacted, though the hurricane bill is not terribly
important because federal aid will also be provided.
In this case, Ben has a 90% credence that pushing the button is morally
required though failing to push would not be seriously wrong, and a 10%
credence that pushing is gravely morally wrong. But he also has credences
regarding the non-moral upshot of each choice: he is 90% confident that
pushing would provide the hurricane relief, and 90% confident that failing to
push would withhold that relief; he is 10% confident that pushing the button
would result in the legalization of gay marriage and 10% confident that failing
to push would prevent that legalization. Uncertaintism holds that Ben should
refrain from pushing the button because it would be taking too big a moral
risk. Actualism holds that Ben should push the button, because it is morally
required to push the button in either scenario he is considering.8
3.3 DO FALSE MORAL VIEWS EXCULPATE?
In this section, I will defend premise 2 of my main argument.

It is controversial whether being caught in the grip of a false moral view
exculpates. I will lay out the relevant issues and briefly argue for and defend
8
The following could seem like a counterexample to Actualism.
The oracle tells Arthur that pushing the green button is morally required. Arthur has lots
of bad moral views, so Arthur forms the belief that pushing the button is X1, X2, X3, …
or Xn, where these are all in fact morally wrong to do.
Actualism holds that Arthur should be guided by his non-moral views, so Arthur should
refrain from pushing the button according to Actualism. But, the objector says, pushing the
button does a morally good thing, and the Oracle told Arthur that. Surely he should push.
the view that being caught in the grip of a false moral view does not exculpate.
(A person is “caught in the grip of a false moral view” just in case she is certain
the false moral view is correct.)
Some people have argued for this claim:
Moral Ignorance Exculpates: A person who acts wrongly is blameworthy
for so acting only if she believes she is acting wrongly.
The claim that Moral Ignorance Exculpates is clearly false. Even if we
take seriously that a person’s moral beliefs and credences can sometimes
render her blameless for moral wrongdoing, it is implausible that mere
ignorance—that is, failure to believe the relevant moral truth—is sufficient
for blamelessness. Cases such as those that motivate Uncertaintism show
Moral Ignorance Exculpates to be false. The mere fact that someone does not
believe her action is wrong does not rule out that she has some non-trivial
credence that it is wrong, and so her moral beliefs and credences may not
vindicate her action at all. One cannot defend one’s wrongful action later
by saying, “I didn’t know it was wrong, though I had a 30% credence that it
was wrong.” That is not a good defense, if one knew that one’s other option
was morally permissible.9,10
The more difficult question is whether being caught in the grip of a false
moral view renders one blameless for wrongful actions:
False Moral Views Exculpate: If a person behaves in a morally wrong
way, while certain of a false moral view according to which that
behavior is morally required, then she is not blameworthy for her
behavior.11
Actualism implies that Arthur should not push the button. In fact, if Actualism is true,
it seems that Arthur is blameworthy for pushing the button. This might seem bizarre,
though in section 3.3 I offer a view on which false moral belief is typically blameworthy;
that Arthur is blameworthy for pushing the button may seem less strange if he is also
blameworthy for his moral beliefs.
If Arthur does not push the button, then Actualism implies he’s not blameworthy;
whereas he may seem to be blameworthy. But at least the Actualist can say, in that case,
that Arthur is blameworthy for his false beliefs, though not for failing to push.
9
The point that moral ignorance does not exculpate, because one might be igno-
rant while being uncertain as to whether one’s action is morally wrong, is made by
Guerrero (2007).
10
Note that, on my view, this is definitely not a good defense because moral belief and
credence is not exculpatory (except when it warrants non-moral credences that would be
exculpatory). But my point is that even if one thinks that moral belief and credence can be
exculpatory, one should not think that mere ignorance can be exculpatory.
11
In my (2011), I discuss an argument that moral ignorance exculpates; I argue that
the real issue is not whether moral ignorance exculpates but rather whether false moral
belief exculpates, and I argue that it does not. Rosen (2003, 2004), Wolf (1987), and
Zimmerman (1997) offer arguments that moral ignorance or false moral belief exculpates.
64 Elizabeth Harman
My main argument in this chapter holds that Uncertaintism is committed

to the claim False Moral Views Exculpate; and I claim that False Moral
Views Exculpate is false.
How might we figure out whether false moral views exculpate? Some
writers have seemed to suggest that we can learn that false moral views
exculpate from the fact that false non-moral views exculpate.12 Consider
Anne, who poisons her husband, certain that she is curing him. Anne is not
blameworthy for behaving in this way. What is the explanation of Anne’s
blamelessness? It does seem that Anne’s ignorance renders her blameless. But
how can we characterize the ignorance that renders her blameless? There are
two competing principles that would explain Anne’s blamelessness. (Note
that we are looking for principles that explain why Anne is not blamewor-
thy for behaving as she did; she might still be blameworthy for having caused
her behavior if she is blameworthy for having come to be ignorant.) Here is
the first principle:
A person is blameworthy for behaving in a particular way only if her
behavior has certain features that make it morally wrong and she
believes that her behavior has those features.
This principle implies that Anne is not blameworthy for her behavior.
Anne’s behavior was a poisoning of her husband, that makes it mor-
ally wrong, but she did not believe her behavior had that feature. Anne
believed her behavior was an attempt to save her husband’s life, but that
feature does not make her behavior morally wrong. There is no feature of
her behavior that both makes it morally wrong and is such that she was
aware of it.
The second principle that would explain Anne’s blamelessness for her
behavior is this:
A person is blameworthy for behaving in a particular way only if her
behavior has certain features that make it morally wrong and she
believes both that her behavior has those features and that they make
it morally wrong.
The second principle implies that false moral views exculpate. So, if cases of
exculpatory non-moral false views like Anne’s are explained by this second
principle, then these cases can directly support the claim that false moral views
exculpate.
But cases like Anne’s are not explained by the second principle. These
cases give us no reason to believe the second principle. Anne’s blame-
lessness follows simply from her failure to know that what she is doing
is a poisoning. (Importantly, she does not even know that it might be a
12
Rosen (2004) seems to suggest this, though this may not be intended.
poisoning.) Anne’s blamelessness is fully explained by the first principle,

which is simply a weaker claim than the second principle. We do not
need to make the stronger claim in the second principle to explain Anne’s
blamelessness.
Thus, while it might have seemed that the fact that false non-moral views
exculpate can support the claim that false moral views exculpate, there is no
support from the former to the latter.13
To know whether false moral views exculpate, we must confront cases
involving false moral views head on, and ask whether they involve blame-
worthiness. When we do, we see that many cases of wrongful behavior by
agents caught in the grip of false moral views are paradigm cases of blame-
worthiness.14 Consider these two cases:
Max works for a Mafia “family” and believes he has a moral obligation
of loyalty to the family that requires him to kill innocents when it is
necessary to protect the financial interests of the family. This is his
genuine moral conviction, of which he is deeply convinced. If Max
failed to “take care of his own” he would think of himself as disloyal
and he would be ashamed.
Gail is a gang member who believes that she has a moral obligation
to kill a member of a neighboring gang as revenge after a member of
her own gang is killed, although her victim was not responsible for the
killing. This is her genuine moral conviction, of which she is deeply
convinced. If Gail failed to “take care of her own” she would think of
herself as disloyal and she would be ashamed.
I claim that Max and Gail are paradigm cases of agents blameworthy for
their wrongful actions. They know that they are killing innocent people;
this is sufficient for the agents to be blameworthy.
A proponent of the claim that false moral views exculpate might respond
as follows:
While these two agents are not blameworthy for killing the inno-
cent people, they are blameworthy for causing these killings to occur
because they must be blameworthy for having come to have these
beliefs. Agents are blameworthy for their moral views if and only if
13
My point here does not depend on how the second principle is worded. We might
instead consider the view that a person is blameworthy for acting in a particular way
only if she knew she was acting wrongly. This principle is more concise than the second
principle I state; but my objection to the second principle is not that it is too compli-
cated, or that it adds a needless further condition, but simply that it is unmotivated by
consideration of cases of non-moral ignorance exculpating. (Sliwa (n.d.) holds that moral
knowledge is necessary for praiseworthiness; but I don’t think she would endorse the
strong claim that moral knowledge is necessary for blameworthiness.)
14
See my (2011).
66 Elizabeth Harman
they have violated procedural moral obligations regarding learning the

moral truth (and these violations were themselves blameworthy).15
These include the obligation to think a reasonable amount about
morality in general, the obligation to take seriously moral arguments
one hears, etc., but do not include any substantive obligations to
believe certain things. These two agents cannot possibly have ful-
filled their procedural moral obligations regarding learning the
moral truth.
I have two responses. First, it underplays the extent and gravity of these
agents’ blameworthiness to treat them as merely blameworthy for having
come to have certain beliefs, and having thereby caused some deaths, while
not seeing them as blameworthy for their killings, which occurred while
they knew full well that they were killing innocent people. Second, it is
a grave mistake to think that people cannot become convinced of deeply
false moral views, such as these, without violating this kind of procedural
obligation. Consider versions of Max and Gail who have thought an ordi-
nary amount about morality and have taken seriously the moral arguments
that have been presented to them. They are aware that many people think
their moral views are false, but they believe they understand where others
have gone wrong: others have been “suckered” into a “wimpy” morality,
when what is really important is taking care of one’s own. These versions of
Max and Gail are certainly possible; they would be blameworthy for their
wrongful behavior.
On my opponent’s view, the question of whether false moral belief
exculpates turns out to be intimately connected to the question: is ethics
hard? Ethics is indeed quite hard, and this is why we see so many cases of
false moral belief even among those who have fulfilled their procedural
moral obligations regarding learning the moral truth. Some people do
not think hard enough about morality in general, or they dismiss moral
arguments that they ought to take seriously. But many people think hard
about morality, take the arguments they hear seriously, and still get it
wrong.16 A failure to appreciate how hard ethics is can make the claim that
false moral belief exculpates seem less radical than it is.17 But that claim
is very radical.18
15
Rosen (2004) offers this view of blameworthiness for moral beliefs.
16
I make this point in my (2011).
17
FitzPatrick (2008), in discussing whether false moral belief exculpates, assumes that
ethics is only hard in special cases.
18
My paper “Ethics is Hard. What Follows?” (n.d.) expands on some of the argu-
ments in this section.
I have argued that false moral views do not exculpate. In my view, peo-
ple who act wrongfully are blameworthy not in virtue of what their moral
beliefs and credences are, but in virtue of what their non-moral beliefs
and credences are, and how these influence their choices. Someone who
knows she is killing an innocent person, and does so anyway, does not
care adequately to avoid killing the innocent.19 A view of blameworthiness
that can undergird the claim that false moral views do not exculpate
is this:
A person is blameworthy for her wrongful behavior just in case it resulted
from her failure to care de re about what is morally important—that is,
from her failure to care adequately about the non-moral features of the
world that in fact matter morally.20
A person cares de dicto about morality if she wants to be moral. A person
cares de re about morality if she wants to keep her promises, to help the needy,
etc., and if keeping one’s promises, helping the needy, etc. are in fact morally
important.
A proponent of the view that false moral views exculpate holds the following:
(*) Someone who behaves morally wrongly while caught in the grip
of a false moral view (according to which what she is doing is morally
required) is not blameworthy for this behavior; she is blameworthy for
causing this behavior only if and only because she is blameworthy for
her false moral view.
I deny claim (*) but I grant that there is something intuitively compelling
about it. I grant that the following claim is true:
(**) Someone who behaves morally wrongly while caught in the grip
of a false moral view is blameworthy for this behavior only if she is
blameworthy for her false moral view.
I grant that there is something odd about holding someone blameworthy
for her morally wrong behavior while acknowledging that she is blameless
for a false moral view according to which that behavior is morally required.
I hold that people who do morally wrong things while caught in the grip of
false moral views are blameworthy for their actions and are also blamewor-
thy for their beliefs. But they are not blameworthy for their actions merely
because they are blameworthy for their beliefs; and they are not blame-
worthy merely for having caused themselves to behave in this way. Rather,
19
Unless this is a case in which it is morally permissible to kill an innocent person.
20
Here I am adopting a view along the lines of Arpaly (2003) and Markovits (2010).
68 Elizabeth Harman
they are blameworthy for both their actions and their beliefs for related
reasons—because both their actions and their beliefs involve their failing to
care adequately about what matters morally:
Believing that one’s wrong action is morally required involves car-
ing inadequately about the features of one’s action that make it mor-
ally wrong, because believing that an action is morally wrong on the
basis of the features that make it wrong is a way of caring about those
features.21
False moral belief is blameworthy. Actions done on the basis of false moral
belief are often blameworthy. On my view, they are blameworthy for similar
reasons.
In this section, I have explained and defended premise 2 of my main
argument, my claim that false moral views do not exculpate. I’ve argued
that consideration of the way that false non-moral views exculpate in no
way supports the claim that false moral views exculpate. I claim that if we
look directly at some cases of wrongdoing due to false moral views, we see
that these are paradigm cases of blameworthy behavior. And I’ve offered
a view of blameworthiness on which false moral views do not exculpate,
which accommodates the thought that if the behavior is blameworthy, then
the false moral belief is blameworthy as well.22,23
So far in this chapter, I have offered my main argument: If Uncertaintism
is true, then false moral views exculpate. But false moral views do not excul-
pate. So, Uncertaintism is false. And I have articulated and explained an
alternative to Uncertaintism: Actualism. In the remaining sections of the
chapter, I defend this argument in the face of some objections. In section
3.4, I discuss the objection that Uncertaintism and Actualism do not really
disagree; and I also offer a further argument against Uncertaintism, based on
an analogy with epistemology. In section 3.5, I discuss an objection that the
analogy with epistemology can be used to support Uncertaintism. In section
3.6, I discuss an objection to the first premise of my main argument, which
holds that if Uncertaintism is true, then false moral views exculpate. In sec-
tion 3.7, I discuss whether either of two revisions of Uncertaintism can resist
my arguments.
21
As I understand it, this part of my view goes beyond anything in Arpaly (2003) and
Markovits (2010).
22
I respond to some objections to this view in my (2011) and in my “Ethics is Hard.
What Follows?”
23
Michael Smith has argued that caring about morality, and acting in a certain way
because it is morally required rather than because of the reasons that make it morally
required, is being a moral fetishist, and is thereby objectionable. (See Smith (1994).) This
is a different thought than the Actualist’s thought that some concern for morality lacks
moral value because it is in fact concern for things that do not matter morally.
3.4 IS THERE REALLY DISAGREEMENT BETWEEN

THE UNCERTAINTIST AND THE ACTUALIST?
Let’s focus on Bob, who must decide whether to teach his daughter Sue to
drive. I have said that the Uncertaintist and the Actualist disagree about
Bob. The Uncertaintist says:
(1) Bob should not teach Sue to drive
because teaching Sue to drive would be taking a serious moral risk.
The Actualist says:
(2) Bob should teach Sue to drive
because refusing to teach Sue to drive would be limiting her options because
she is a woman.
The dialectic here is complicated. The Uncertaintist will grant that (2) is
true, or has a true reading. To see this, remember the case of Anne, who
poisons her husband thinking she is curing him. The following claim can be
truly made of Anne:
(3) Anne should not give the drink to her husband
because it is poison. We can also truly say:
(4) Anne should give the drink to her husband
because one should do what one believes will save one’s husband’s life.
Claim (3) is true of Anne, ignoring her beliefs and credences about her situa-
tion, we might say; it is made true by what her situation really is. Claim (4) is
true of Anne, given her beliefs and credences about her situation, we might say.
Similarly, the Uncertaintist can claim that there are three ways for claims
about Bob to be true. First:
(2) Bob should teach Sue to drive.
This is true objectively. Ignoring Bob’s beliefs and evidence about his sit-
uation, given what his situation really is, he should teach his daughter
to drive.
But also:
(2) Bob should teach Sue to drive.
This is true for one kind of subjectivity. Ignoring Bob’s moral beliefs and
credences, just focusing on his non-moral beliefs and credences, Bob should
teach Sue to drive. (The same claim is true on two different readings.)
Finally,
(1) Bob should not teach Sue to drive.
70 Elizabeth Harman
According to the Uncertaintist, this claim is true taking into account all of
Bob’s beliefs and credences about his situation (including his moral beliefs and
credences). This claim is true for a second kind of subjectivity.24,25
The Uncertaintist may then claim that the Actualist and the Uncertaintist
do not disagree. Rather, both agree that claim (2) is true. The Uncertaintist
simply raises a third question that the Actualist does not appear to be inter-
ested in: suppose we do not ignore a person’s moral beliefs and credences,
but take into account her whole mental state. What should she do, on the
basis of this whole mental state?
I believe that some Uncertaintists would see the dialectic this way.26 But
they are wrong. Actualism is a proposed answer to the very same question
the Uncertaintists are interested in, namely: how should a person act, tak-
ing into account her beliefs and credences (including her moral beliefs and
credences), given that one sometimes must act while experiencing moral
uncertainty?
I will now draw a lesson using an analogy with epistemology.
24
Here is an example in which three different moral claims might be true, one in each
of these three ways, according to the Uncertaintist:
John’s sixteen-year-old daughter wants to take the morning-after pill. John has two pills
in front of him, A and B. John believes that A is an aspirin and B is the morning-after pill.
In fact, it is the reverse: A is the morning-after pill and B is an aspirin. John is sure that
taking the morning-after pill is wrong and that it is wrong to give it to one’s daughter.
Consider:
(i) John should give his daughter pill A.
(ii) John should give his daughter pill B.
(iii) John should give his daughter neither pill.
(i) is true as an objective moral claim. John should give his daughter what is actually
the morning-after pill: pill A. (ii) is true as a subjective moral claim relative to John’s
non-moral beliefs but ignoring his moral beliefs: a person should give his daughter what
he takes to be the morning-after pill if she wants to take the morning-after pill. (iii) is true
relative to John’s entire mental state, according to the Uncertaintist.
25
Now that we have distinguished these ways in which moral claims may be true, we
can clarify the best interpretation of certain general claims that Uncertaintism makes,
such as: “Someone who is 90% sure that φing is minorly morally wrong, but 10% sure
that failing to φ is deeply morally wrong, should refrain from φing.” Here the agent’s
two mentioned credences are best understood as credences in subjective moral claims that
are relative to the agent’s non-moral credences, ignoring the agent’s moral credences; the final
“should” claim, which the Uncertaintist makes, is best understood as a subjective claim
relative to the agent’s entire mental state.
26
Seppieli (2008) distinguishes the “Non-Normative Belief-Relative ‘Should’ ” from
the “most belief-relative ‘should’ of all—relative to the agent’s beliefs about both the
normative and the non-normative” which seem to correspond to the two kinds of sub-
jectivity I mention. He clarifies that his claims are in terms of the latter (as I say the
Uncertaintist’s claims are).
Uncertaintism is a view about how a person should behave; consider

the analogue of the Uncertaintist regarding how a person should believe.
The Epistemic Uncertaintist holds that how a person should believe is
determined by what she believes about how she should believe. If a person
is sure that she should form her beliefs in accord with a certain rule, then
indeed she should do so. If a person is unsure between two different ways
of forming beliefs, then her beliefs should be formed in a compromise
way between those two methods, etc. To see why this view fails, consider
this case:
Mary believes that a particular way of reasoning is a good way of
reasoning. In fact it’s not. It involves coming to beliefs on the basis
of claims that don’t really support those beliefs. This way of reason-
ing is not a way of becoming justified in the newly formed beliefs.
But Mary does believe it is a good way of reasoning. Now Mary
considers following a particular line of reasoning. She correctly sees
that it is an instance of that way of reasoning, the one she believes
to be a good way of reasoning. She correctly sees that this line of
reasoning would lead her to believe P. In fact Mary’s evidence and
the reasoning available to her (ignoring her false belief that it is a
good way of reasoning) do not support a belief in P. Should Mary
believe P?
Mary should not believe P. The mere fact that she believes a bad way of
reasoning to be a good way of reasoning does not make that way of reason-
ing a good way for her to reason or a way that she now should reason. We can
say this while taking into account that she believes that it is a good way of
reasoning. The true epistemological view, whatever it is, will hold that Mary
should not believe P. The claim that Mary should not believe P is an ana-
logue of Actualism’s claim that people caught in the grip of false moral views
nevertheless should not do the particular deeply morally wrong things they
believe to be morally required; and it is an analogue of Actualism’s claim
that Bob should teach Sue to drive, even taking into account his whole
epistemic state.
Mary’s case shows that the following general claim is false, where X ranges
over either behaving in a certain way or believing in a certain way:
If an agent is convinced of principles that imply she should X, and she
sees that these principles imply she should X, then all things consid-
ered, in light of her whole mental state, she should X.
While this principle would indeed imply that Uncertaintism is right about
cases involving certainty, consideration of Mary’s case shows that this prin-
ciple is false. Seeing how it is false in the epistemological case helps to see
how it may be false in the moral case, I claim.
72 Elizabeth Harman
Mary’s case also provides a counterexample to the following more general

claim (again letting X range over either behaving in a certain way or believ-
ing in a certain way):
An agent’s beliefs and credences about how she should X alone deter-
mine how she should X.
This general claim would imply that Uncertaintism (or something like it) is
true, and that Actualism is false; but this general claim is false.
For all I have said so far in this section, it might seem that we could
state the disagreement between Actualism and Uncertaintism as follows.
Uncertainists believe that claims like (1) and (2) both have true readings.
Actualists believe, rather, that claims like (1) are false on any reading. But
the dialectic is not that simple. The Actualist need not deny that there is any
true reading of (1), but simply that the Uncertaintist’s intended reading of
(1) is false. The Actualist can grant that there is a true reading of (1), while
denying both that it is what the Uncertaintist says and that it is interesting.
How can the Actualist see (1) as having a true reading? The true reading
occurs in a speech like this:
“If Bob were to decide what to do by being guided by his moral beliefs
and credences, what should he do? He should refrain from teaching
Sue to drive.”
This isn’t a crazy way to talk. But it doesn’t tell us anything interesting about
Bob, because he shouldn’t be guided in that way. This speech is an analogue
of the following.
“What should Mary believe, if she were to be guided by her beliefs
about how she should believe? She should believe P.”
The claim that Mary should believe P may be true in this context. But
it’s not an interesting truth. It is no more interesting that the last claim in
this speech:
“Nora has a lot of evidence about whether the earth is more than 6,000
years old. She knows that Tom says the earth is only 6,000 years old.
She knows that there is a scientific consensus that the earth is more than
6,000 years old. But let’s ignore her information about the scientists.
Focus on what she knows about what Tom says. If Nora’s belief is to be
guided by her knowledge of what Tom says, what should she believe?
She should believe that the earth is not more than 6,000 years old.”
The final claim in this speech is true. But it is not an interesting truth.
Actualism holds that there are contexts in which (1) expresses a truth,
but it is as uninteresting as the claim that Nora should believe the earth
is not more than 6,000 years old. Most importantly, Actualism holds that
Uncertaintists put forward claims like (1) in contexts in which those claims
are not true. Uncertaintists claim that a person who faces moral uncertainty,
such as Bob, is doing the best he can in light of his full epistemic situation,
including both his moral and non-moral beliefs and credences, when he
chooses not to teach Sue to drive. That claim is false.27
3.5 OBJECTION: THE ANALOGY WITH

EPISTEMOLOGY CAN BE USED TO ESTABLISH
UNCERTAINTISM
At this point, someone might object as follows on behalf of Uncertaintism.

Consider the following case.
George follows a particular line of reasoning to a conclusion. It leaves
him with a high credence in P. In fact, this is good reasoning. But
George has recently been told, by someone he respects, that this kind
of reasoning is bad reasoning. George was not convinced, but he has
a small credence that this person was correct; his credence is 10%.
The objector’s claim: George should be less confident in P than if he
did not have this small doubt about this form of reasoning: his cre-
dence in P should be lower than it should be in the alternative.
If the objector’s claim about George is right, then in the epistemic case, a per-
son’s beliefs and credences about how she should believe are relevant to what
she should believe. This might seem to support Epistemic Uncertaintism.
I will respond to this worry in two ways.
First, I will make a dialectical point. I brought up the epistemic anal-
ogy to make the following point: Mary’s case is a counterexample to a
general principle that would support both Uncertaintism and Epistemic
Uncertaintism.
Mary’s case shows that one’s beliefs about whether one should X do not
alone determine whether one should X. If the objector’s claim about George
is correct, then George’s case shows that, at least sometimes, one’s beliefs
about whether one should X are relevant to whether one should X.
27
It’s a bit unclear how to read Sepielli (2008) on this question. On the one hand, he
does say that his claims are in terms of the “most belief-relative ‘should’ of all—relative to
the agent’s beliefs about both the normative and the non-normative.” That’s how I read
him. On the other hand, he comments that he is making claims about “local rationality”
rather than “global rationality,” explaining what an agent should do “relative to” her cre-
dences in claims of the form “action A is better than action B,” etc. I take this to stipulate
that an agent might have other moral beliefs that Sepielli’s view doesn’t take into account;
I do not take him to be saying that his claims ignore the agent’s non-moral beliefs.
74 Elizabeth Harman
The Actualist acknowledges that sometimes one’s beliefs about how one
should behave are relevant to how one should behave—in particular, in
cases in which one’s beliefs about how one should behave are themselves the
warrant for further beliefs about what one’s non-moral situation is.
The lesson of Mary’s case and George’s case is that there is a substantive
epistemological question “how should this agent believe?” that is answered by
consideration of the agent’s whole mental state. The answer cannot be read off
of the agent’s beliefs about how she should believe. Similarly, there is a sub-
stantive normative question “how should this agent behave?”—which is in part
a moral question—that is answered by consideration of the agent’s whole men-
tal state. The answer cannot be read off of the agent’s beliefs and credences
about how she should behave. But epistemology and morality are different,
and the way that these substantive questions get answered may be different.
Now for my second response to the objection. It is not clear to me that
the objector’s claim about George must be true.28
Nevertheless, I do find the objector’s claim about George plausible and I am
not going to deny it. An Actualist does not need to deny the claim about George.
3.6 OBJECTION: THE FIRST PREMISE OF MY MAIN

ARGUMENT IS FALSE
In section 3.4, I distinguished several types of subjective moral claims. This

provides a way for the Uncertaintist to object to the first premise of my
main argument. My main argument is:
1. Uncertaintism implies that being caught in the grip of a false moral view
is exculpatory.
2. It is not true that being caught in the grip of a false moral view is exculpatory.
Therefore:
3. Uncertaintism is false.
28
The Objector makes his claim; my suggestion is that the Objector’s claim may be
false; a third option is that the there is more than one sense of “epistemically justified”
or “rational” and that the claim is true on one reading and false on another. Miriam
Schoenfield (n.d.) develops a view according to which there are two senses of “rational.”
One credence in P may be rational for George, in light of his evidence. Another credence
for P may be rational for George, in light of his doubt about his reasoning process.
Sepielli (2013) similarly offers a view on which there are various kinds of local rationality
(rationality relative to some of an agent’s evidence and other credences), but there is no
such thing as a credence being globally rational, or rational in light of the agent’s entire
epistemic situation.
My argument for premise 1 can be unpacked as follows:

(i) Uncertaintism claims that agents should act as their moral credences
dictate. (This is a subjective “should.”)
(ii) A person is blameworthy for some behavior only if she should not have
behaved in that way. (This is a subjective “should.”)
(iii) Persons caught in the grip of false moral views, who behave in ways
those views morally require, should behave as they do, according to
Uncertaintism.
Therefore:
(iv) Persons who behave morally wrongly while caught in the grip of false
moral views (that require their behavior) are not blameworthy for their
behavior, if Uncertaintism is true.
An objector might hold that this argument equivocates between two
different kinds of subjective claims: subjective claims that are true relative
to an agent’s moral beliefs but ignoring her non-moral beliefs and credences;
and subjective claims that are true relative to an agent’s entire mental state,
including all her beliefs and credences. Claim (ii) uses a subjective “should”
that is relative to an agent’s entire mental state. But claim (i), the objector
holds, uses a subjective “should” that is relative only to the agent’s moral
beliefs and credences, ignoring her non-moral beliefs and credences. The
objector holds that my argument for premise 1 equivocates.
My response to this objection is that if this objection is correct, then
Uncertaintism is an uninteresting claim. It is no more interesting than the
claim that Nora should believe the earth is not more than 6,000 years old,
discussed above.29
3.7 TWO REVISIONS OF UNCERTAINTISM
In this section, I will consider an objection that offers two revisions to

Uncertaintism that may seem to avoid some of the worries I have raised and
that provide some of what Uncertaintism offers. They are:
Revision #1 of Uncertaintism: An agent should be guided not by her
actual moral beliefs and credences but only by her epistemically justified
beliefs and credences.
29
While my reading of Sepielli (2008) is that he does not make what I call the uninterest-
ing version of his claim, my reading of Sepielli (2013) suggests he would prefer to embrace
what I call the uninteresting version of his claim, while disputing that it is uninteresting.
76 Elizabeth Harman
Revision #2 of Uncertaintism: An agent should be guided not by her

actual moral beliefs and credences but by the moral beliefs and credences
that she would be justified in holding.
The objector claims that one or both of these views does better than
Uncertaintism at addressing the worries I have raised.
The following question is relevant:
Is it possible for agents to be epistemically justified in being certain of
false moral views according to which deeply morally wrong actions are
morally required—including all such false moral views discussed in this
chapter?
In other work, I have entertained an answer of “no” to this question.30 I assume
that those interested in rescuing Uncertaintism by offering Revision #1 or #2
believe that the answer to this question is “yes”; let’s assume that is so. It seems
this can only hold if testimony is a way to become epistemologically justified in
believing false moral views, so let’s assume that it is.31
We should note that these revisions of Uncertaintism abandon some of
what might have motivated Uncertaintism. The idea that a person cannot do
better than use her own beliefs32 is given up as motivation for Uncertaintism.
One Uncertaintist has suggested that we should separate the question of what
credences a person should have from the question of how she should act in
light of the credences she actually has and has said that Uncertaintism answers
the latter question only.33 (But he has since changed his mind.34) This is also
abandoned by the revisions.
Can my main argument be revised to address these new views? It can.
30
In my (2011).
31
Yet another view is that some moral truths are such that it is not possible to be
justified in getting them wrong, while other moral truths are such that it is possible to be
justified in getting them wrong. This view may involve a misguided assumption that we
ourselves occupy a privileged position, only wrong or unsure about moral truths that are
deeply hard to know, while others get wrong truths that are easy to know. In fact, others
in the future may themselves be able to see easily how we are going wrong. Even moral
truths that are deeply obvious to some may be hard to see by others trying earnestly,
and even moral truths that seem hard for us are obvious to others. Because there is no
privileged perspective from which to separate the easy from the hard, there are no facts
that some moral questions are easy and some are hard, and thus there is no privileged
class of moral truths such that those—and only those—are the ones one can be justified
in getting wrong.
32
Sepielli (2008: 8) writes: “we cannot base our actions on the correct normative
standards … we cannot guide ourselves by the way the world is, but only by our repre-
sentations of the world.”
33
Sepielli (2008).
34
Sepielli (2013).
My Main Argument, Revised to Target Revisions #1 and #2 of Uncertaintism:

1′. If Revision #1 or Revision #2 of Uncertaintism is true, then being epis-
temically justified in being certain of a false moral view is exculpatory.
2′. It is false that being epistemically justified in being certain of a false
moral view is exculpatory.
Therefore:
3′. Neither Revision #1 nor Revision #2 of Uncertaintism is true.
I endorse premise 2′ of this revised argument for the same reasons that
I endorse premise 2 of the original argument. First, consideration of the
way that justified non-moral certainty exculpates gives us no reason to think
that justified moral certainty exculpates. Second, the very same kinds of
cases I discussed, the Mafia family member and the gang member who kill
innocent people, provide counterexamples to the claim that justified moral
certainty exculpates. If we take testimony seriously as a source of justification
for false moral belief, then these agents may be epistemically justified while
they are nevertheless paradigm cases of blameworthy agents. Finally, we can
grant that it would be weird for their actions to be blameworthy while their
beliefs are blameless, claiming that their beliefs are indeed blameworthy.
One might object that epistemically justified beliefs cannot be morally
blameworthy. But why not?
One (controversial) example of an epistemically justified but morally
blameworthy belief is a sister’s belief in her brother’s guilt of a serious crime
given just enough evidence to make that belief epistemically justified; she
should give him the benefit of the doubt and withhold belief. We can say
this even while acknowledging that more than enough evidence to make the
belief epistemically justified would make the belief blameless; the moral
duty to give one’s sibling the benefit of the doubt just requires being some-
what reluctant to believe badly of him.35
One might think that epistemically justified beliefs cannot be morally
blameworthy because a person who is epistemically justified could not have
believed differently, and a person cannot be blameworthy for something if
she could not have done otherwise. But it is not true that all epistemi-
cally justified beliefs are such that one could not have believed differently.
I might take your word for something, and come to be epistemically justi-
fied in believing it, even though I was capable of thinking it through for
myself, and had I done so, I would have realized it was false.36
35
Keller (2004) and Stroud (2006) argue that we owe our loved ones the benefit of
the doubt, and that we may owe it to them to refrain from holding beliefs that would be
epistemically justified. Lackey (n.d.) disagrees.
36
See my (2011).
78 Elizabeth Harman
In response to Revisions #1 and #2 of Uncertaintism, here is a revised

statement of my view. (I endorse both the original statement and the revised
statement.)
Actualism (Revised): A person’s moral beliefs and moral credences
(whether justified or not) are usually irrelevant to how she (subjectively)
should act (as are the moral beliefs and moral credences she would be
justified in holding). How a person (subjectively) should act usually
depends solely on her non-moral beliefs and credences; her moral
beliefs and credences (or those she should hold) are relevant only inso-
far as they provide warrant for beliefs and credences about what her
non-moral situation may be.
3.8 CONCLUSION
I have argued that Uncertaintism is false. If Uncertaintism is true, then

false moral views exculpate. But false moral views do not exculpate. So,
Uncertaintism is false.
There are two main ways of rejecting my argument. First, one might
hold that false moral views do exculpate. If the Uncertaintist goes that
route, she takes on a significant commitment that Uncertaintists have
so far not acknowledged (as far as I know). Second, one might hold that
Uncertaintism is not committed to the view that moral false views excul-
pate. If the Uncertaintist goes that route, it turns out that her claims are not
interesting. The Uncertaintist is not telling us how an agent should act, in
light of her whole mental state (including both moral and non-moral beliefs
and credences); rather, she is simply telling us how agents should act, ignor-
ing their non-moral beliefs and credences.37
References
Arpaly, N. 2003. Unprincipled Virtue. Oxford: Oxford University Press.
Dowell, J. 2013. “Flexible Contextualism about Deontic Modals: A Puzzle about
Information-Sensitivity,” Inquiry 56: 149–78.
FitzPatrick, W. 2008. “Moral Responsibility and Normative Ignorance: Answering
a New Skeptical Challenge,” Ethics 118: 589–613.
37
I owe much thanks to Tyler Doggett, Peter Graham, Alex Guerrero, Sarah McGrath,
Dan Moller, Gideon Rosen, Miriam Schoenfield, Andrew Sepielli, Paulina Sliwa, Brian
Weatherson, and the participants at the 2013 Wisconsin Metaethics Workshop for com-
ments on drafts of this chapter.
Guerrero, A. 2007. “Don’t Know, Don’t Kill: Moral Ignorance, Culpability, and

Caution,” Philosophical Studies 136: 59–97.
Harman, E. 2011. “Does Moral Ignorance Exculpate?” Ratio 24: 443–68.
Harman, E. n.d. “Ethics is Hard. What Follows?” manuscript.
Jackson, F. 1991. “Decision-Theoretic Consequentialism and the Nearest and
Dearest Objection,” Ethics 101: 461–82.
Keller, S. 2004. “Friendship and Belief,” Philosophical Papers 33: 329–51.
Kolodny, N. and MacFarlane, J. n.d. “Ought: Between Subjective and Objective,”
manuscript.
Lackey, J. n.d. “Why There is No Epistemic Partiality in Friendship,” manuscript.
Lockhart, T. 2000. Moral Uncertainty and Its Consequences. Oxford: Oxford
University Press.
MacFarlane, J. 2014. Assessment Sensitivity: Relative Truth and its Applications.
Markovits, J. 2010. “Acting for the Right Reasons,” Philosophical Review
119: 201–42.
Moller, D. 2011. “Abortion and Moral Risk,” Philosophy 86: 425–43.
Rosen, G. 2003. “Culpability and Ignorance,” Proceedings of the Aristotelian Society
103: 61–84.
Rosen, G. 2004. “Skepticism about Moral Responsibility,” Philosophical Perspectives
18: 295–313.
Ross, J. 2006. “Rejecting Ethical Deflationism,” Ethics 116: 742–68.
Schoenfield, M. n.d. “Bridging Rationality and Accuracy,” manuscript
Sepielli, A. n.d. “Normative Uncertainty and Intertheoretic Comparisons of Value,”
manuscript.
Sepielli, A. 2008. “What to Do When You Don’t Know What to Do,” in R.
Shafer-Landau (ed.), Oxford Studies in Metaethics. Oxford: Oxford University
Press, 5–28.
Sepielli, A. 2013. “What to Do When You Don’t Know What to Do When You
Don’t Know What to Do …” Noûs 48: 521–44.
Sliwa, P. n.d. “Moral Worth and Moral Knowledge,” manuscript.
Smith, H. n.d. “Making Morality Work,” manuscript.
Smith, M. 1994. The Moral Problem. Oxford: Blackwell.
Stroud, S. 2006. “Epistemic Partiality in Friendship,” Ethics 116: 498–524.
Weatherson, B. 2014. “Running Risks Morally,” Philosophical Studies 167: 141–63.
Wolf, S. 1987. “Sanity and the Metaphysics of Responsibility,” in F. Schoeman (ed.),
Responsibility, Character, and the Emotions. Cambridge: Cambridge University
Press, 46–62.
Zimmerman, M. 1997. “Moral Responsibility and Ignorance,” Ethics 107: 410–26.
4
Justification and Explanation
in Mathematics and Morality
Justin Clarke-Doane
In an influential book, Gilbert Harman writes:

In explaining the observations that support a physical theory, scientists typically
appeal to mathematical principles. On the other hand, one never seems to need to
appeal in this way to moral principles.
(Harman 1977: 10)
What is the epistemological relevance of this contrast, if genuine? In this
chapter, I argue that ethicists and philosophers of mathematics have misun-
derstood it. They have confused what I will call the justificatory challenge for
realism about an area, D—the challenge to justify our D-beliefs—with the
reliability challenge for D-realism—the challenge to explain the reliability of
our D-beliefs. Harman’s contrast is relevant to the first, but not, evidently,
to the second. One upshot of the discussion is that genealogical debunk-
ing arguments are fallacious. Another is that indispensability considerations
cannot answer the Benacerraf–Field challenge for mathematical realism.
4.1 THE JUSTIFICATORY CHALLENGE
Let the justificatory challenge for realism about an area, D, be the challenge
to justify our D-beliefs (realistically construed).1 By “justify,” I mean argue
1
I will not consistently add the qualification “realistically construed” in what follows.
But this is always intended. (Obviously, no argument supports or threatens our beliefs
under any construal.) Realism about an area, D, is roughly the view that D-sentences
should be interpreted literally, and that some atomic or existentially quantified ones
are true relevantly counterfactually, constitutively, and causally independent of any-
one’s believing them to be. For a detailed explication of “D-realism,” see Clarke-Doane
(2012a: section 1).
Justification and Explanation in Mathematics and Morality 81
for, or defend. (I do not assume that being able to justify our belief that p
is either necessary or sufficient for being justified in believing that p.) Then
one answer to the justificatory challenge for mathematical realism is that
the contents of our mathematical beliefs figure into the best explanation of
our observations. By “observation,” I mean any “immediate judgment made
in response to the situation without any conscious reasoning” (Harman
1977: 208), where a judgment is a mental event, not a propositional con-
tent. Note that both mathematical and moral beliefs, understood in the
occurrent sense, may qualify as observations. W. V. O. Quine writes:
Objects at the atomic level and beyond are posited to make the laws of macroscopic
objects, and ultimately the laws of experience, simpler … Moreover, the abstract
entities which are the substance of mathematics … are another posit in the same
spirit. Epistemologically these are myths on the same footing with physical objects …
neither better nor worse except for differences in the degree to which they expedite
our dealings with sense experiences.
(Quine 1951: 42)2
One advantage to Quine’s answer to the justificatory challenge for math-
ematical realism is that it seems to afford an empirical justification of our
mathematical beliefs (realistically conceived). By an “empirical justifica-
tion,” I mean an argument or defense that should convince an empiri-
cal scientific realist. In particular, one can argue that, merely by being an
empirical scientific realist, one is committed, on pain of incoherence, to
being a mathematical realist. As Hilary Putnam writes:
[Q]‌uantification over mathematical entities is indispensable for science … therefore
we should accept such quantification; but this commits us to … the existence of
the mathematical entities [that satisfy our theories]. This type of argument stems,
of course, from Quine, who has for years stressed both the indispensability of quan-
tification over mathematical entities and the intellectual dishonesty of denying the
existence of what one daily presupposes.
(Putnam 1971: 347)
The drawback of Quine’s answer to the justificatory challenge is that it
seems to afford no way to justify all of our mathematical beliefs. It seems
hopeless to argue that the contents of our higher set-theoretic beliefs figure
into the best explanation of our observations, for example. Quine him-
self was compelled to pronounce such truths “mathematical recreation and
without ontological rights” (Quine 1986: 400).
2
I am not sure whether Quine means the same thing by “sense experience” as Harman
means by “observation,” though their usage of these terms is clearly related. But I will
be concerned with the Quinean position where “sense experience” means what Harman
means by “observation.”
82 Justin Clarke-Doane
But however compelling Quine’s answer to the justificatory challenge for

mathematical realism may be, notoriously, a similar answer does not seem
possible in the moral case.3 Gilbert Harman writes:
Observation plays a part in science it does not appear to play in ethics, because
scientific principles can be justified … by their role in explaining observations …
(Harman 1977: 10)
In particular:
In explaining the observations that support a physical theory, scientists typically
appeal to mathematical principles. On the other hand, one never seems to need to
appeal in this way to moral principles.
(Harman 1977: 9–10)
Let the Indispensability Thesis be the claim that the contents of our typical
mathematical beliefs figure into the best explanation of every observation (in
Harman’s sense), and let Harman’s Objection be the claim that the contents of
our moral beliefs do not figure into the best explanation of any of our obser-
vations.4 It will not matter for my purposes whether either claim is actually
true. But what would be the epistemological upshot if they both were true?
One upshot is obvious. If the Indispensability Thesis and Harman’s
Objection were both true, and if Quinean empiricism were also true, then
our mathematical beliefs would be justified, but our moral beliefs would
not be. By “Quinean empiricism,” I mean the view that our belief that p is
justified if and only if p figures into the best explanation of our observations.
A more interesting epistemological upshot has been touched upon. If
the Indispensability Thesis and Harman’s Objection were both true, then
we would seem to be able to empirically justify our typical mathematical
beliefs (realistically construed), but could not so justify any of our moral
3
Of course, some ethicists have challenged this appearance—just as some philosophers
of mathematics have challenged the appearance that the contents of our typical mathe-
matical beliefs (realistically construed) figure into the best explanation of our observations.
See, for example, Sturgeon (1985), Brink (1989), and Boyd (2003a, 2003b), in the moral
case, and Chihara (1990), Field (1980, 1989), and Hellman (1989) in the mathematical.
4
Pedantically: the Indispensability Thesis says that for any typical (i.e., not higher set-
theoretic) mathematical truth, p, and for any observation, O, the best explanation of O
implies p, and Harman’s Objection says that there is no moral truth, q, and observation, O,
such that the best explanation of O implies q. Why does anyone believe the Indispensability
Thesis, so strongly formulated? Because typical mathematics seems to play a role in our
empirical scientific theories that is like the role played in them by their logic. It is a back-
ground assumption, so that every explanation “will contain the axioms of number theory
and analysis” (Steiner 1973: 61). (Harman’s Objection is sometimes understood as the claim
that the contents of our moral beliefs do not figure into the best explanation of intuitively
“observable phenomena” more generally—including, e.g., the movements of planets. The
argument which follows would work equally under this reading of “Harman’s Objection.”)
beliefs. Again, it is hard to see how to be an empirical scientific realist while

rejecting mathematical realism, but there is no apparent obstacle to being
an empirical scientific realist while rejecting moral realism.5 This point is
merely dialectical, however. Does anything else follow?
4.2 GENEALOGICAL DEBUNKING ARGUMENTS
An increasingly common suggestion is that (knowledge of ) Harman’s

Objection—or a consequence of it—undermines our moral beliefs (realisti-
cally construed), and the Indispensability Thesis blocks an analogous argu-
ment in the mathematical case.6 Richard Joyce writes:
Nativism [the thesis that moral beliefs are evolutionarily innate] offers us a gene-
alogical explanation of moral judgments that nowhere … presupposes that these
beliefs are true … My contention … is that moral nativism … might well … render
[moral beliefs] unjustified … In particular, any epistemological benefit-of-the-doubt
that might have been extended to moral beliefs … will be neutralized by the avail-
ability of an empirically confirmed moral genealogy that nowhere … presupposes
their truth.
(Joyce 2008: 216)
Harman’s Objection implies that there is a “moral genealogy that
nowhere … presupposes” the truth of our moral beliefs. Hence, if “the
availability” of such a genealogy will “render moral beliefs unjustified,”
then a fortiori so will Harman’s Objection.7
Joyce’s reasoning is of interest. If sound, then the upshot of Harman’s
Objection is not merely the dialectical one that we cannot empirically jus-
tify our moral beliefs (realistically construed), as we apparently can our
mathematical ones. It is that we ought (epistemically) to give them up.
But how could Harman’s Objection undermine our moral beliefs? Again,
if Quinean empiricism were true, then, rather than undermining our moral
5
But see Clarke-Doane (2014: section 2). Note that it does not follow that we can-
not justify our moral beliefs in any way. Traditional realists about both mathematics and
morality have tended to justify their (epistemically basic) beliefs by appeal to their intui-
tive evidentness. See Gödel (1947) and Ross (1939) for classic arguments.
6
I will not consistently add the qualification “knowledge of,” but this is always
intended. Note that (knowledge of ) information, E, undermines our belief that p only if
p was antecedently justified.
7
Joyce’s argument strictly speaks of the truth of our moral beliefs, while Harman’s
Objection speaks of their contents. But if the contents of our moral beliefs fail to figure
into the best explanation of any of our observations, then so does their truth, and vice
versa, by (uncontroversial instances of ) the T-schema. I will, therefore, mostly ignore the
distinction between the contents of our beliefs and their truth in what follows.
beliefs, Harman’s Objection would show that they were never justified to
begin with. But even if this were what Joyce intended to show, such an argu-
ment would have no traction with Joyce’s primary targets—“non-naturalist”
moral realists—who explicitly accept Harman’s Objection and so, of course,
reject Quinean empiricism.8
The only promising answer to the question of how Harman’s Objection
could undermine our moral beliefs of which I am aware is suggested by the
work of Sharon Street.9 She writes:
[T]‌he realist must hold that an astonishing [inexplicable] coincidence took place—
claiming that as a matter of sheer luck, evolutionary pressures affected our evaluative
attitudes in such a way that they just happened to land on … the true normative
views … [T]o explain why human beings tend to make the normative judgments
that we do, we do not need to suppose that these judgments are true.
(Street 2008: 208–9, emphasis in original)
Let the reliability challenge for realism about an area, D, be the chal-
lenge to explain the reliability of our D-beliefs (realistically construed).
Then I understand genealogical debunking arguments as follows. Harman’s
Objection (or a consequence of it) undermines our moral beliefs (realisti-
cally construed) by showing that the reliability challenge for moral realism
is unanswerable. The assumption, familiar from Hartry Field, is that “our
belief in a theory [is] undermined if … it would [appear to] be a huge coin-
cidence if what we believed about its subject matter were correct” (Field
2005: 77). What debunkers must add is that in order to relevantly explain
the reliability of our D-beliefs (realistically construed), it is at least necessary to
show that their contents figure into their best explanation.10
The Indispensability Thesis blocks a genealogical debunking argument
against mathematical realism (where our elementary mathematical beliefs
are Harmanian observations). Joyce writes:
There is some evidence that natural selection has provided humans with an inbuilt
faculty for simple arithmetic … [D]‌oes the fact that we have such a genealogical
explanation of our simple mathematical beliefs serve to demonstrate that we are
8
See Joyce (2006: ch. 6). In his (forthcoming), p. 17, Joyce seems not to appreciate this.
9
I will discuss two answers which are suggested by Joyce’s work in section 4.3.
10
I borrow the term “reliability challenge” from Schechter (2010), though I under-
stand it in accord with Field (1989: 26). Unlike debunkers and Field (see section 4.7),
Schechter does not believe that the apparent impossibility of explaining the reliability of
our D-beliefs undermines them (realistically construed). He merely thinks that it “counts
against” them. I believe that the interest of the reliability challenge for D-realism is greatly
diminished if it is not supposed to undermine our D-beliefs (realistically construed). But
I will not be concerned with Schechter’s view.
unjustified in holding these beliefs? Surely not, for we have no grasp of how [math-
ematical] belief[s] might have been selected for … independent of [their] truth.
(Joyce 2006: 182)11
If it is assumed that such an argument not only blocks one way of show-
ing that the reliability challenge for mathematical realism is unanswerable,
but also suffices to answer that challenge, then it follows that showing that
the contents of our D-beliefs figure into their best explanation is both neces-
sary and sufficient for relevantly explaining their reliability. How plausible is
this conclusion? In what follows, I consider the necessity and the sufficiency
claims in reverse order.
4.3 THE BENACERRAF–FIELD CHALLENGE
The sufficiency claim is immediately suspect. The reliability challenge for

mathematical realism is widely thought to seem unanswerable even given the
Indispensability Thesis. Paul Benacerraf writes:
[O]‌n a realist (i.e., standard) account of mathematical truth our explanation of how
we know the basic postulates must be suitably connected with how we interpret
the referential apparatus of the theory … [But] what is missing is precisely … an
account of the link between our cognitive faculties and the objects known … We
accept as knowledge only those observations which we can appropriately relate to
our cognitive faculties.
(Benacerraf 1973: 674, emphasis in original)
Benacerraf ’s concern stems from the view that if X knows that p, then
there obtains a causal relation between X and the subject matter of p.12 This is
11
Clarke-Doane (2012a: section 3) argues, contra Joyce, that we can explain the use-
fulness of arithmetic beliefs in terms of corresponding first-order logical truths alone (and
similarly in the case of geometry). But for the purposes of this chapter, I set this argument
aside, and assume that Joyce is correct. (In his (forthcoming), Joyce responds that he,
unlike Street, is “free to maintain that an explanation of the usefulness of the ancestral
belief that 2 + 3 = 5 in terms of first-order logic is nevertheless one that presupposes that
our ancestors’ arithmetic beliefs were true” (p. 8, fn. 7, emphasis in original). But I do
not know what this means. Surely Joyce does not claim that arithmetic truths are first-
order logical truths. In his (forthcoming), Joyce writes that he does “not have space here
to argue against Clarke-Doane’s complicated argument, so [he]’ll take the simple way
out and point out that in [his] work the … mathematics example is just that—a passing
illustration—and if it fails then [he]’ll still confidently stand by the general claim [that his
argument does not overgeneralize].” But in his (2006), Joyce writes: “the dialectic within
which I am working here assumes that if an argument that moral beliefs are unjustified
… would by the same logic show that our belief that 1 + 1 = 2 is unjustified … this would
count as a reductio ad absurdum” (Joyce 2007: 182, fn. 5). For more on overgeneraliza-
tion, see fn. 29.)
12
See Benacerraf (1973: 671–3).
different from the view that if X knows—let alone justifiably believes—that

p, then p figures into the best explanation of our “observation” that p (or
figures into the best explanation of any of our observations, in Harman’s
sense).13 Benacerraf does not discuss the latter constraint, which is trivially
satisfied in the mathematical case, by the Indispensability Thesis. As Mark
Steiner writes:
[S]‌uppose that we believe … the axioms of analysis or of number theory … [S]ome-
thing is causally responsible for our belief, and that there exists a theory—actual or
possible, known or unknown—which can satisfactorily explain our belief in causal
style. This theory, like all others, will contain the axioms of number theory and analysis.
(Steiner 1973: 61, emphasis in original)14
Of course, causal constraints on knowledge and justification are now
widely rejected. But the reliability challenge for mathematical realism still
appears to be unanswerable. The problem is that it can appear impossible
to explain the reliability of our mathematical beliefs, whether in terms of
a causal relation between their subject matter and us or in any other way.
Field writes:
Benacerraf ’s own formulation … relies on a causal theory of knowledge … [But t]he
key point … is that … it seems hard to give any account of our beliefs about …
mathematical objects that doesn’t make the correctness of the beliefs a huge
coincidence.
(Field 2005: 77)15
13
See Clarke-Doane (2014) and (forthcoming a). In fact, there is no plausible ana-
logue of Benacerraf ’s concern in the moral case (Clarke-Doane 2014: section 3). There
does obtain a causal relation between us and the subject matter (the values of the names
and bound variables) of our moral beliefs. Their subject matter is the likes of people,
actions, and events. One might respond that there does not obtain a causal relation
between us and the subject matter of sentences like “Generosity is a virtue.” But whether
we ought to believe any sentence of the form “F-ness is G” on a face-value construal is
just the problem of universals. If we ought, then there fails to obtain a causal relation
between us and the subject matter of all manner of our beliefs—e.g., that red is a color,
that force is the product of mass and acceleration, and that Beethoven’s Fifth Symphony
is in the key of C minor. If we ought not, then such examples are irrelevant. Either way,
unlike our mathematical beliefs, our moral beliefs generate no new Benacerraf problem
(a similar point applies if one takes apparent talk of “reasons” as ontologically commit-
ting). This point is widely missed. See, for example, Bengson (forthcoming) and Joyce
(forthcoming: 9).
14
A similar point can be made with respect to logic. Every logical truth is a con-
sequence of every explanation at all. Hence, for any logical truth that we believe, p, p
figures into the best explanation of our “observation” that p. But surely this truism does
not suffice to answer the reliability challenge for logical realism.
15
Note that Field does not challenge the Indispensability Thesis in this context.
To be sure, some advocates of the Indispensability Thesis claim that the

Indispensability Thesis answers the reliability challenge for mathemati-
cal realism. They claim that since mathematical knowledge is like theoretical
empirical scientific knowledge, arrived at by way of an inference to the best
explanation of our observations, then it is no more mysterious than the latter.
On this view, the likes of numbers, sets, and tensors are not epistemically dif-
ferent from electrons. Mark Colyvan writes:
[L]‌et’s take a … charitable reading of the Field version of the [Benacerraf ] challenge,
according to which the challenge is to explain the reliability of our systems of beliefs …
Once the challenge is put this way, we see that Quine has already answered it: we justify
our system of beliefs by testing it against bodies of empirical evidence.
(Colyvan 2007: 111, emphasis in original)16
But this confuses the justificatory challenge and the reliability challenge.
Showing that the contents of our mathematical beliefs figure into the best
explanation of our observations—mathematical or otherwise—may help to
empirically justify those beliefs. But it does not suffice to explain their reliability.
It leaves it totally mysterious how those beliefs manage to reliably align with
the mathematical truths (realistically construed).17 In general, showing that the
contents of our D-beliefs figure into their best explanation is not sufficient for
answering the reliability challenge for D-realism. Why, then, would debunkers
assume that it is necessary?
4.4 SENSITIVITY
There are two reasons latent in the literature on debunking arguments. The first
is that an explanation of the reliability of our D-beliefs would show that had the
D-truths been different, our D-beliefs would have been correspondingly different.18
16
See also Hart (1996).
17
See also the quote from Field in section 4.7, as well as his (2005) where he
writes: “[A]‌lthough [those who argue that Benacerraf ’s problem doesn’t arise for the
empiricist] say that empirical evidence bears on mathematical claims, they have not
offered (and could not easily offer) even a clear sketch of how the experiences that alleg-
edly might overturn our mathematics are reliable symptoms of the facts about mathe-
matical objects” (Field 2005: 71). Colyvan claims not to fall prey to the above confusion,
but I do not understand how he supposes himself to avoid it. Hart does not seem to
acknowledge a distinction between justifying our beliefs and explaining their reliability.
18
Or, more weakly, it would block the worry that had the D-truths been different, our
D-beliefs would have failed to be (perhaps on the grounds that such counterfactuals are
unintelligible). See Field (2005) and Clarke-Doane (2012a). The differences between the
two demands will not matter for my purposes.
For convenience, I will say that our D-beliefs are sensitive if this counterfactual is
true.19 The assumption is that if the contents of our D-beliefs fail to figure into
their best explanation, then they fail to be sensitive. Joyce writes:
Suppose that the actual world contains real categorical requirements—the kind that
would be necessary to render moral discourse true. In such a world humans will be
disposed to make moral judgments … for natural selection will make it so. Now
imagine instead that the actual world contains no such requirements at all—noth-
ing to make moral discourse true. In such a world, humans will still be disposed to
make these judgments … for natural selection will make it so … [D]‌oes the truth of
moral judgments … play a role in their usefulness? … I believe the answer is “No.”
(Joyce 2001: 163, emphasis in original)20
The problem with such arguments is well-known.21 Suppose that our
explanatorily basic moral beliefs—our conditional beliefs which purport
to state the conditions under which a moral property is instantiated—are
(actually) true, and that the explanatorily basic moral truths would be nec-
essary if true at all. Then our explanatorily basic moral beliefs are vacuously
sensitive on a standard semantics. Our non-basic moral beliefs seem to be
sensitive even if the explanatorily basic moral truths would be metaphysi-
cally contingent. Had Bush’s invasion of Iraq not been wrong, it would
have been different in non-moral respects, and our moral beliefs would have
varied correspondingly (since, even if the explanatorily basic moral truths
would not be metaphysically necessary, the closest worlds in which the ante-
cedent is true are presumably worlds in which those truths are the same).22
Note that neither conclusion depends on the claim that the contents (or
truth) of our moral beliefs figure into their best explanation.23
19
Note that, while related, the present notion of sensitivity is different from that of
Nozick (1981: ch. 3). Both notions must plausibly be relativized to a method of belief
formation, though how they ought to be is irrelevant for my purposes.
20
Similarly, Michael Ruse writes:
You would believe what you do about right and wrong, irrespective of whether or not a
“true” right and wrong existed! The Darwinian claims that his/her theory gives an entire
analysis of our moral sentiments. Nothing more is needed. Given two worlds, identical
except that one has an objective morality and the other does not, the humans therein
would think and act in exactly the same ways.
(Ruse 1986: 254)
And Walter Sinnott-Armstrong writes:
The evolutionary explanations [of our moral beliefs] work even if there are no moral
facts at all.
(Sinnott-Armstrong 2006: 46)
21
See Sturgeon (1985, 1986) for the basic insight.
22
Thanks to Alex Silk for helpful discussion of this point.
23
Joyce no longer wishes to rely on such counterfactuals (see his (forthcoming)).
Unfortunately, he does not seem to replace this reliance with an alternative answer to the
Of course, the above argument assumes that our explanatorily basic

moral beliefs are (actually) true. But this assumption is unobjectionable in
the context of the reliability challenge, as debunkers themselves point out.24
Consider the perceptual case. What we can arguably offer in this case is an
evolutionary explanation of how we came to have sensitive mechanisms for
perceptual belief, and a neurophysical explanation of how those mecha-
nisms work such that they are sensitive.25 But these explanations blatantly
assume the (actual) truth of our explanatorily basic perceptual beliefs. If the
reliability challenge for D-realism requested an explanation of the reliability
of our D-beliefs which failed to assume the (actual) truth of our explanatorily
question at issue—or, correlatively, an alternative answer to the question of how Harman’s

Objection could undermine our moral beliefs. In his most recent work, he says that
moral nativism shows that our moral beliefs are the product of a “non-truth-tracking”
process which is “independent of their truth.” But by this he just means that the contents
of our moral beliefs do not figure into the best explanation of our having them (forth-
coming: 8). In other words, Joyce simply repeats the relevant consequence of Harman’s
Objection, and does nothing to explain how it could “render moral beliefs unjustified.”
Similarly, in his (2006), he makes an argument by analogy. He claims that learning the
truth of moral nativism is like learning that we have taken a “Napoleon lost Waterloo”
pill (2006: 179). But, as Joyce apparently now recognizes, this case is not analogous. By
all accounts, whether Napoleon lost at Waterloo is metaphysically contingent. Hence,
there is no problem arguing that, had Napoleon won there, we still would have believed
that he had lost.
24
See Balaguer (1995), Field (1989: 26) and the quotation from this page in section
4.7, Gibbard (2003: section 13), Schechter (2010), and Street (n.d.) for discussion of
this issue. (Field, Gibbard, and Street add that, while any explanation of reliability will
be “circular” in the present sense, adequate explanations must meet a non-triviality con-
straint. My own view is that once the challenge to explain the reliability of our abductive
methodology is properly distinguished from the challenge to explain the reliability of our
epistemically basic mathematical or moral beliefs, this suggestion cannot be made out.
Only Field (1996: 371) ventures a statement of the needed constraint, but ends up just
restating the sensitivity constraint at issue in this section. For my purposes here, however,
all that matters is that the explanation of the sensitivity of our uncontroversial beliefs will
be “circular.”)
25
See Schechter (2010) for something like this distinction of explanatory tasks.
According to Schechter, such a distinction is important because the question of how we
came to have a reliable mechanism for D-belief may remain open even under the assump-
tion that it is unintelligible to imagine the D-truths being different (even if the question
of how that mechanism works such that it is reliable may not). But I believe that this is
mistaken. Schechter is explicit that the question of how we came to have a reliable mech-
anism for D-belief is different from the question of how we came to have the mechanism
for D-belief that we actually came to have (since the latter question is clearly answerable in
principle). However, in order to decide whether we were, say, selected to have a reliable
mechanism for D-belief, as opposed to being selected to have a mechanism for D-beliefs
with property, F, that is in fact reliable, we would seem to need to have to decide what
mechanism it would have benefited our ancestors to have had had the D-truths been differ-
ent. For discussion, see Clarke-Doane (2012a, forthcoming a), and Field (2005).
basic D-beliefs, then the apparent impossibility of answering it could not be

thought to undermine those beliefs.
In order for the argument for the sensitivity of our explanatorily basic
moral beliefs to work, it must also be assumed that the explanatorily basic
moral truths would be metaphysically necessary (again, the argument for
the sensitivity of our non-basic moral beliefs seems to work even without
this assumption). But debunkers have typically allowed that they would
be, and it is clear why. The belief that the explanatorily basic moral truths
would be metaphysically necessary is commonly thought to have a similar
status as the belief that they are (actually) true. It is a (defeasibly) justified
belief that must be undermined. If debunkers merely claimed to undermine
our moral beliefs under the assumption that the belief that the explanatorily
basic moral truths would be metaphysically necessary is not itself (defeasibly)
justified, then the interest of their argument would be greatly diminished.
In fact, debunkers tend to allege that whether the explanatorily basic
moral truths would be metaphysically necessary is irrelevant. Even if there
are no metaphysically possible worlds in which those truths are different,
debunkers claim that the counterfactual, “had the explanatorily basic moral
truths been different, our moral beliefs would have been correspondingly
different,” is not vacuously true. It must be evaluated with respect to “con-
ceptually possible” worlds—where these are, roughly, ways the world could
be, for all that we can intelligibly imagine. The explanatorily basic moral
truths could have been different as a matter of “conceptual possibility.”
Speaking of normative truths generally, Street writes:
As a purely conceptual matter … normative truths might be anything … Noting
this sense in which the normative truth might be anything, and noting the role of
evolutionary forces in shaping the content of our basic evaluative tendencies, we
may wonder whether … it somehow promoted reproductive success to grasp the
independent normative truth, and so creatures with an ability to do so were selected
for. Unfortunately for the realist … to explain why human beings tend to make the
normative judgments that we do, we do not need to suppose that these judgments
are true.
(Street 2008: 208)
The problem with this allegation is that we seem to be equally unable to
show that our beliefs which are not supposed to be in doubt in this context
are sensitive with respect to “conceptually possible worlds.” Consider our
beliefs about ordinary objects, such as rocks. Street is explicit that her argu-
ment does not threaten realism about ordinary objects such as these. She
writes:
[Evolutionary debunking arguments do not] go through against realism about non-
evaluative facts such as facts about fires, predators, cliffs, and so on … In order to
explain why it proved advantageous to form judgments about the presence of fires,
predators, and cliffs, one will need to posit in one’s best explanation that there were
indeed fires, predators, and cliffs, which it proved quite useful to be aware of, given
that one could be burned by them, eaten by them, or could plummet over them.
(Street 2006: 160, fn. 35)
But even if the contents of our ordinary object beliefs figure into their
best explanation, we do not seem to be able to show that our explanato-
rily basic ones are sensitive with respect to “conceptually possible worlds.”26
While it may be metaphysically necessary that the conditions under which
the property of being a rock is instantiated are what they are, it seems that
they could have been different “as a purely conceptual matter.” That they are
is just what some ontologists allege. These ontologists allege that particles
arranged rockwise fail to compose a rock.27 But had—for all that we can
intelligibly imagine—this been the case, our rock beliefs would have been
the same.28
The mathematical case makes the point vividly. By the Indispensability
Thesis, the contents of our typical mathematical beliefs figure into the
best explanation of every observation. But it is commonly held that, even
if the Indispensability Thesis is true, virtually none of our mathematical
beliefs is sensitive with respect to “conceptually possible” worlds.29 In fact,
26
We seem to be able to show that our non-basic moral beliefs are sensitive with
respect to conceptually possible worlds for the same reason that we seem to be able to
show that they are sensitive with respect to metaphysically possible ones. The closest
worlds in which the non-basic moral truths are different are presumably still worlds in
which the explanatorily basic moral truths are the same.
27
See Van Inwagen (1990) and Merricks (2001). See also Clarke-Doane (2014: sec-
tion 3), and Berker (2014: section 8).
28
This shows that the widely assumed view that we were “selected to have true ordi-
nary object beliefs,” but not true moral beliefs, is obscure at best. For virtually any area,
D, had the explanatorily basic D-truths been different, it would have benefited our
ancestors to have correspondingly different D-beliefs—given the (actual) truth of our
D-beliefs and that this counterfactual is evaluated with respect to metaphysically possible
worlds. For virtually any area, D, the explanatorily basic D-truths would be metaphysi-
cally necessary, so the counterfactual is (vacuously) true. In particular, if D is morality,
then the counterfactual is true. However, for virtually any area, D, the counterfactual
is false if it is evaluated with respect to “conceptually possible” worlds. For virtually no
D-truths would be conceptually necessary in the sense in question. In particular, the
counterfactual is false if D is ordinary object discourse, as explained above. (Again, had
the non-basic D-truths been different, it would have benefited our ancestors to have
correspondingly different D-beliefs—given the (actual) truth of our D-beliefs—whether
this counterfactual is evaluated with respect to metaphysically or conceptually possible
worlds. In particular, this is true when D is morality, as explained above.)
29
See Azzouni (1994: 56), Balaguer (1999: 113), Ellis (1990: 113), Horgan
(1987: 281), and the quotation from Field below.
problematically, Field sometimes suggests that this is the Benacerraf problem.

He writes:
The Benacerraf problem … seems to arise from the thought that we would have had
exactly the same mathematical … beliefs even if the mathematical … truths were
different … and this undermines those beliefs … [T]‌here is a reasonably clear con-
tent (at least prima facie) to [this counterfactual]; that’s what gives the Benacerraf
problem its initial bite in the mathematical case.
(Field 2005: 81)
Merely arguing that the contents of our mathematical beliefs figure into
their best explanation does nothing evidently to block this counterfactual.
It “gives no sense to the idea that if the … facts had been different then our
… beliefs would have been different too” (Field 1996: 371).30
To sum up: our moral beliefs appear to be sensitive with respect to meta-
physically possible worlds, even if their contents do not figure into their
best explanation—and reflection on the ordinary object and mathematical
cases suggests that our explanatorily basic ones would fail to be sensitive
with respect to “conceptually possible” worlds even if their contents did so
figure. Hence, even if showing that our D-beliefs are sensitive is necessary
for “explaining their reliability,” this is no reason to assume that showing
that our D-beliefs figure into the best explanation is necessary for explain-
ing their reliability. What of the second reason to assume this?
4.5 SAFETY
The second reason is harder to pin down, but is suggested by Charles

Darwin himself. He writes:
In the same manner as various animals have some sense of beauty, though they
admire widely different objects, so they might have a sense of right and wrong,
though led by it to follow widely different lines of conduct. If, for instance … men
were reared under precisely the same conditions as hive-bees, there can hardly be
a doubt that our unmarried females would, like the worker-bees, think it a sacred
duty to kill their brothers, and mothers would strive to kill their fertile daughters,
and no one would think of interfering. Nevertheless the bee, or any other social
animal, would in our supposed case gain … some feeling of right and wrong, or a
conscience.
(Darwin 1871: 70)
30
There is an argument from the Indispensability Thesis to the sensitivity of our
mathematical beliefs with respect to “conceptually possible” worlds. But it is highly sus-
pect. See Field (1989: 18–20) and Clarke-Doane (2012a).
More recently, Michael Ruse writes:

Had evolution taken us down another path, we might well think moral that which
we now find horrific, and conversely. This is not a conclusion acceptable to the
[moral realist].
(Ruse 1986: 254)
It is tempting to read Darwin and Ruse as merely noting that had our explan-
atorily basic moral beliefs been different, they would have been false.31 But this
counterfactual is a trivial consequence of moral realism. It says that the moral
truths do not counterfactually depend on our moral beliefs. No explanation of
the reliability of our moral beliefs (realistically construed) should counter this.
A more charitable reading of Darwin and Ruse adds that we could have
easily had different explanatorily basic moral beliefs. The conclusion would
then be that we could have easily had false moral beliefs. If we say that our
D-beliefs are safe just in case we could not easily have had false ones, then it
follows that our moral beliefs are not safe.32 The second reason may then be
understood as the claim that an explanation of the reliability of our moral
beliefs would show that they are safe.
But showing that the contents of our D-beliefs figure into their best
explanation is not necessary for showing that they are safe as genealogical
debunking arguments illustrate. Street’s point is naturally taken to be that we
were (all but) bound to have at least the “core” moral beliefs that we do have,
whether or not they were true. To take a simple example, it does not seem
that we could have easily believed that killing our offspring is good, since
creatures who did would seem to have been less effective at passing on their
genes than creatures who believed what we do. Street writes, “among our
most deeply and widely held judgments, we observe many … with exactly
the sort of content one would expect if the content of our evaluative judg-
ments had been heavily influenced by selective pressures” (Street 2006: 116).
But if we could not have easily had different “core” moral beliefs, then, given
that our core moral beliefs are (actually) true and that the explanatorily basic
moral truths would be necessary, at least our core moral beliefs could not
have easily been false. Of course, showing that our “core” moral beliefs are
safe is a far cry from showing that our moral beliefs generally are.33 But what
31
It must be our explanatorily basic beliefs that are in question since, again, had our
non-basic beliefs been different, they would not plausibly have been false (given their
actual truth).
32
Again, a method of belief formation must plausibly be held fixed. But how exactly
it ought to be is irrelevant for my purposes, so I ignore this complication. (Note that the
present formulation of safety avoids the consequence that beliefs in necessary truths are
automatically safe. See Pritchard (2008) for a related treatment.)
33
There is an argument that if our “core” moral beliefs are safe, then our moral beliefs
generally are as safe as our beliefs from any area are. See Clarke-Doane (2012b: section 3).
matters is whether showing that our moral beliefs are safe requires showing
that their contents figure into their best explanation. It clearly does not.34
In fact, showing that the contents of our moral beliefs figure into their best
explanation does not even seem to suffice to establish their safety. The appeal
of pluralist accounts of mathematics, such as Mark Balaguer’s “Full-Blooded
Platonism,” according to which every (consistent) mathematical theory is
equally true (under a face-value Tarskian truth definition), is arguably that our
mathematical beliefs would not be safe if (non-pluralist) standard mathemati-
cal realism were true.35 Even if we could not have easily believed that 1 + 1 = 3,
or that addition is not commutative, arithmetic trivialities wildly underdeter-
mine the content of our mathematical theories. The Indispensability Thesis
does nothing evidently to show that our mathematical beliefs are suitably con-
strained. Field writes:
[Pluralists solve] the [Benacerraf ] problem by articulating views on which though
mathematical objects are mind independent, any view we had had of them would have
been correct … [T]‌hese views allow for … knowledge in mathematics, and unlike
more standard [realist] views, they seem to give an intelligible explanation of it.
(Field 2005: 78)36
A similar point applies to ordinary objects. A common motivation for
“mereological permissivism” is that our ordinary object beliefs would not be
safe if standard mereological realism were true. Korman writes:
Proponents of sufficiently permissive conceptions … can admit that we could easily
have come to have slightly or radically different conventions, and that we would then
have judged there to be various kinds of extraordinary objects. But our beliefs are none-
theless safe: whichever conventions we had ended up with, our judgments about the
existence of the relevant objects would still have been correct. The extraordinary objects
are all already out there waiting to be noticed; all that our conventions do is determine
which ones we do notice.
(Korman 2011: section 4.2, emphasis in original)37
34
Note that I am not saying that our moral beliefs are safe. I am saying that showing
that they are need not involve showing that their contents figure into their best explanation.
35
See Clarke-Doane (forthcoming a: section 3). The relevant notion of consistency is
a primitive modal one.
36
This is awkward, since, in the same section Field claims that the Benacerraf problem
is the problem of showing that our mathematical beliefs are sensitive. But mathemati-
cal pluralism does nothing to help with this. Had—“as a purely conceptual matter”—
Balaguer’s pluriverse failed to exist, or had it been different, our mathematical beliefs
would have been the same. Indeed, Balaguer concedes that “[i]‌f there were never any
such things as [mathematical] objects, the physical world [and, hence, our mathematical
beliefs] would be exactly as it is right now” (Balaguer 1999: 113).
37
Similarly, an earlier draft of Korman (2014)—which, coincidentally, is premised on
the thesis that I am challenging here—contained the following quotations from Heller,
Sider, and Hawthorne on pp. 1–2:
[I]‌f we conceptually divide up the world into objects one way rather than another
because doing so will serve our purposes better [which we do], then there is little chance
Again, merely arguing that the contents of our explanatorily basic ordi-
nary object beliefs figure into their best explanation does nothing evidently
to show that they could not have easily been different.
To sum up: debunking arguments themselves show that our moral
beliefs may be safe, even if their contents fail to figure into their best
explanation—and reflection on the ordinary object and mathematical
cases suggests that they may fail to be safe even if their contents do so fig-
ure. Hence, even if showing that our D-beliefs are safe is necessary for rel-
evantly explaining their reliability, this is no reason to assume that showing
that our D-beliefs figure in the best explanation is necessary for explaining
their reliability. Is there any other reason to assume this?
4.6 MODAL SECURITY
There is still the intuition that if the contents of our moral beliefs failed to
figure into their best explanation, then those beliefs would “have nothing
to do with” the moral truths (realistically construed).38 But stripped of its
connection with sensitivity or safety, what could this mean? It could just
amount to a restatement of Harman’s Objection (or a consequence of it).39
But, in that case, the intuition cannot be used to explain how Harman’s
that the resulting ontology will be the true ontology … In principle, we could by sheer
coincidence arrive at the true ontology by the use of conventions … I will discount the
possibility of such a coincidence.
(Heller 1990: 44)
On [a conservative) view, the entities that exist correspond exactly with the categories for
continuants in our conceptual scheme: trees, aggregates, statues, lumps, persons, bodies,
and so on. How convenient! It would be nothing short of a miracle if reality just hap-
pened to match our conceptual scheme in this way.
(Sider 2001: 156–7)
Barring a kind of anti-realism that none of us should tolerate, wouldn’t it be remarkable
if the lines of reality matched the lines that we have words for? The simplest exercises of
sociological imagination ought to convince us that the assumption of such a harmony is
altogether untoward, since such exercises convince us that it is something of a biological
and/or cultural accident that we draw the lines that we do.
(Hawthorne 2006: 109)
Note that, unlike mathematical pluralism, mereological permissivism only “vindicates”
positive beliefs of the relevant sort. It does not, for instance, vindicate the belief that there
are not any “incars”—cars that are, necessarily, located in garages.
38
Debunking arguments are often formulated in such a way as to leave this intuition
unanalyzed. See Street (2006).
39
The claim that there is no “explanatory connection” between our moral beliefs and
their contents (or truth) is an example of this suggestion. Again, this seems to be Joyce’s
in his (forthcoming). See n. 23.
Objection undermines our moral beliefs. Alternatively, it could amount to

the claim that there is no causal relation between the subject matter of our
moral beliefs and us. But, in that case, it just replays Benacerraf ’s origi-
nal argument, and cannot be thought to undermine our moral beliefs.40
Finally, it could amount to the claim that the probability that our moral
beliefs are true is low.41 But either the probability at issue is epistemic or it
is objective. If the probability is epistemic, then the suggestion is blatantly
question-begging. It effectively amounts to the conclusion that debunking
arguments are supposed to establish—that our moral beliefs are not justi-
fied. Suppose, then, that the probability is objective. Then for any (explana-
torily basic) moral truth, p, presumably Pr(p) = 1, given that such truths
would be metaphysically necessary.42 Moreover, as genealogical debunking
arguments themselves illustrate, it may be that Pr(we believe that p) ≈ 1,
because the probability of our having the (explanatorily basic) moral beliefs
that we do is high. But, then, Pr(p & we believe that p) ≈ 1, by the prob-
ability calculus. Since, Pr(our belief that p is true) = Pr(p & we believe that
p), it appears that Pr(our belief that p is true) ≈ 1, contrary to hypothesis.43
I suggest that debunkers’ key assumption—that in order to relevantly
explain the reliability of our D-beliefs it is at least necessary to show that
their contents figure into their best explanation—is not just unsupported: it
is dubiously coherent. I have argued that—assuming that our moral beliefs
are actually true—we may be able to show that they are both sensitive and
safe (in whatever sense we can show that our uncontroversial beliefs are),
even given that their contents fail to figure into their best explanation. If
debunkers’ key assumption were correct, then we may be unable to “explain
the reliability” of our moral beliefs despite being able to show that they
were (all but) bound to be true. Of course, there are surely senses of the
quoted phrase in which this is true (there are myriad senses of “explain the
reliability”). But the relevant sense of this phrase is such that the apparent
impossibility of explaining the reliability of our D-beliefs undermines them
(realistically construed). If debunkers’ key assumption were correct, then
40
Again, this suggestion appears to be confused anyway. See n. 13.
41
See Street (n.d.) for an apparent example of this suggestion.
42
What if we only assign objective probability 1 to contents which are necessary in
an even stronger sense—e.g., “conceptually necessary”? Then, again, the contents of our
uncontroversial beliefs—e.g., our explanatorily basic ordinary object beliefs—would
seem to have equal claim to being objectively improbable. See section 4.4. (Thanks to
David James Barnett for discussion.)
43
Perhaps the quoted phrase means that the truth of our belief that p is not “grounded
in” or “constituted by” the fact that p (see Bengson (forthcoming) for something like this
proposal)? Such hyperintensional ideology does not seem to me to be more perspicuous
than the quoted phrase itself. But, even if it were, this proposal could not serve debunk-
ers’ dialectical purposes, as will become clear below.
we may be obligated to give up beliefs which we can show were (all but)
bound to be true.
How could that happen? That it could not is the key idea behind the
following.
Modal Security: Information, E, cannot undermine our D-beliefs
without giving us some reason to believe that our D-beliefs are not
both safe and sensitive.44
Modal Security states a necessary condition on E. It does not say that if
E gives us some reason to believe that our D-beliefs are not both safe and
sensitive, then E undermines those beliefs. It says that if E does not even do
this, then E cannot be thought to undermine them.
This principle is weak. It is plausible that E cannot undermine our
D-beliefs without giving us some reason to believe that our D-beliefs are
not safe—i.e., without giving us some reason to believe that, even if our
D-beliefs are actually true (as we assume, for the sake of argument, in the
moral case), we might have easily had false ones. Even if this is incorrect,
it is hard to see how E could undermine our D-beliefs without giving us
some reason to believe either that our D-beliefs are not safe or that they are
not sensitive.45 If E fails to give us some reason to believe either of these
things, then E fails to threaten our judgment that our D-beliefs were (all
but) bound to be true. How could information obligate us to give up our
beliefs of a kind while failing to threaten our judgment they were (all but)
bound to be true?
It might be thought that Modal Security has absurd implications.
Suppose that we are astrological realists. If we are granted the (defeasible)
justification and (actual) truth of our astrological beliefs, and we can argue
both that the explanatorily basic astrological truths would be necessary if
true at all and that we could not have easily had different explanatorily basic
astrological beliefs, then, by Modal Security, we can relevantly explain their
reliability. That is, we can explain their reliability in any sense which is such
that the apparent impossibility of explaining their reliability undermines
them. Of course, it is doubtful that we could argue these things. But would
it be a reductio of Modal Security if we could? Arguably, it would be a
reductio of the idea that we should be granted the (defeasible) justification and
44
This principle is introduced in Clarke-Doane (forthcoming a). Again, sensitivity
and safety must plausibly be relativized to methods of belief formation. (Thanks to David
James Barnett for extensive discussion of this and related principles.)
45
“Rebutting” as well as “undercutting” underminers satisfies Modal Security. If E is
evidence for (the contents of ) alternative D-beliefs, then E is evidence that our D-beliefs
are false and so, a fortiori, not both safe and sensitive. (Thanks to Sinan Dogramaci for
helpful discussion of these issues.)
(actual) truth of our astrological beliefs. If we are not granted these things, then
there may be no hope of a dialectically effective argument against our position.
But absent an answer to the question above, such an argument does not, in
general, seem possible. If Modal Security is true, then debunking arguments
overreach.46
4.7 JUSTIFICATION AND EXPLANATION
If Modal Security is true, then it is false that we must show that the contents
of our moral beliefs figure into their best explanation in order to relevantly
explain their reliability. It suffices to show that our moral beliefs are both safe
and sensitive, and, as we have seen, in order to show this, we need not show
that their contents figure into their best explanation.47
But even absent an argument that debunkers’ key assumption is false,
I hope to have shown that the positive arguments that it is true are poor. Why,
then, have so many philosophers supposed otherwise? I suggest that, like some
advocates of the Indispensability Thesis, they have confused the justificatory
and reliability challenges for realism about an area. Field writes:
[W]‌e can formulate [Benacerraf’s] challenge so as to make indispensability considera-
tions of questionable relevance to answering it. The way to understand Benacerraf’s
challenge … is not as a challenge to … justify our mathematical beliefs, but as a chal-
lenge to … explain the reliability of these beliefs. We start out by assuming the existence
of mathematical entities that obey the standard mathematical theories; we grant also
that there may be positive reasons for believing in those entities. These positive reasons
might … be that the postulation of these entities appears to be indispensable … But
Benacerraf’s challenge … is to … explain how our beliefs about these remote entities
can so well reflect the facts about them … [I]f it appears in principle impossible to explain
this, then that tends to undermine the belief in mathematical entities, despite whatever
reason we might have for believing in them.
(Field 1989: 26, emphasis in original)
I take Field to suggest that the challenge to empirically justify our math-
ematical beliefs should not be confused with the challenge to explain their reli-
ability. Some explanations of the reliability of our mathematical beliefs—such
as Balaguer’s—fail to empirically justify them, since they fail to imply that their
contents figure into the best explanation of any of our observations (though,
again, Balaguer’s view is consistent with this).48 Some empirical justifications
46
See also Clarke-Doane (forthcoming a: section 6, and forthcoming b: section 4).
47
Note that it may be highly non-trivial to show this in either the mathematical or the
moral cases, since it may be highly non-trivial to show that our beliefs are safe.
48
Balaguer is explicit that his purpose is not to justify our mathematical beliefs, but
rather to explain their reliability (under the assumption that they are true). See his (1995).
of our mathematical beliefs, such as Colyvan’s, fail to explain their reliability,

since—contra Colyvan—they fail to indicate how those beliefs manage to reli-
ably align with the mathematical truths.
Debunkers appear to have confused these challenges. Consider Joyce’s
fuller account of why genealogical debunking arguments do not threaten
our mathematical beliefs. Joyce writes:
There is some evidence that natural selection has provided humans with an inbuilt
faculty for simple arithmetic … [L]‌et’s interpret this as implying that our belief that
1 + 1 = 2 is innate. This … is an eternal and necessary truth, and thus by “hard-
wiring” such a belief into our brains natural selection takes no risks—it is not as if
the environment could suddenly change such that 1 + 1 would equal 3. So does the
fact that we have such a genealogical explanation of our simple mathematical beliefs
serve to demonstrate that we are unjustified in holding these beliefs? Surely not, for
we have no grasp of how this belief might have been selected for … independent of
its truth … The truth of “1 + 1 = 2” is a background assumption to any reasonable
hypothesis of how this belief might have come to be innate.
(Joyce 2006: 182)
Joyce claims that “the truth of “1 + 1 = 2” is a background assumption
to any reasonable hypothesis of how [the belief that 1 + 1 = 2] might have
come to be innate.” The suggestion that the contents of our mathematical
beliefs figure into their best explanation helps to answer the justificatory
challenge for mathematical realism. It shows, as per the Indispensability
Thesis, that the contents of those beliefs figure into the best explanation of
our observations (in Harman’s sense). But we may be able to show that the
contents of our beliefs of a kind, D, figure into their best explanation despite
having no idea how those beliefs manage to reliably align with the D-truths.
Indeed, in the mathematical case, this seems to be the situation. Hence, the
Indispensability Thesis fails to answer the reliability challenge for math-
ematical realism.
In the other direction, Joyce suggests that mathematical truths are “eter-
nal and necessary.” The suggestion that the mathematical truths are neces-
sary helps to answer the reliability challenge for mathematical realism. It
shows, assuming the (actual) truth of our mathematical beliefs, that they
are sensitive (the claim that the mathematical truths are eternal seems to be
redundant, since, if they are necessary, then presumably they are eternal).
Joyce also suggests that our “core” mathematical beliefs are innate, implying
that we could not have easily had different ones. This suggestion also helps
to answer the reliability challenge. It shows, assuming the necessity of their
contents, that our “core” mathematical beliefs are safe. But we may be able to
show that our D-beliefs are both sensitive and safe despite our inability to show
that their contents figure into their best explanation (or, indeed, despite our
inability to show that their contents figure into the best explanation of any
of our observations). Hence, Harman’s Objection may not, and, if Modal
Security is true, cannot show that the reliability challenge for moral realism
is unanswerable.
4.8 CONCLUSIONS
I have argued that in order to relevantly explain the reliability of our

beliefs of a kind, D, it appears to be neither sufficient nor necessary to
show that their contents figure into their best explanation. It follows that
the Indispensability Thesis cannot answer the Benacerraf–Field challenge
for mathematical realism, and, if Modal Security is true, that Harman’s
Objection cannot undermine our moral beliefs by showing that the reli-
ability challenge for moral realism is unanswerable (or, indeed, in any other
way). The alternative view apparently arises from a confusion of the justifi-
catory and reliability challenges for D-realism—if it does not arise from the
false assumption that showing that the contents of our D-beliefs figure into
their best explanation helps to show that they are sensitive or safe.
I have not argued that the genealogy of our D-beliefs is irrelevant to
the reliability challenge for D-realism. For all that I have argued, a cul-
tural history of our moral or mathematical beliefs could undermine them
by showing that we might have easily had different (explanatorily basic)
ones—thereby giving us some reason to believe that they are not safe, and
so, a fortiori, not both safe and sensitive.49 Or perhaps an evolutionary his-
tory of those beliefs could help to answer the reliability challenge for moral
realism by blocking this possibility. The point has been that whether a gene-
alogy of our D-beliefs aggravates or alleviates the reliability challenge for
D-realism is independent of whether it assumes their contents. In order to
“explain the reliability” of our D-beliefs it does not suffice to argue that their
genealogy assumes their contents, and in order to “debunk” them it does
not suffice to argue the opposite. 50
49
The empirical fact of moral and mathematical disagreement might have the same effect,
in which case genealogical speculation would turn out to be unnecessary in this context. For
more on disagreements in mathematics, see Clarke-Doane (2014). For an example of a kind
of debunking argument which makes no use of Harman’s Objection, see Kitcher (2006).
50
Thanks to David James Barnett, Sinan Dogramaci, Hartry Field, Toby Handfield,
Colin Marshall, Josh May, Jennifer McDonald, Alex Silk, Jussi Suikkanen, two anony-
mous referees for Oxford Studies in Metaethics, and to audiences at the University of
Birmingham, Columbia University, the University of Cambridge, the University of
Oxford, the 2013 Wisconsin Metaethics Workshop, and the 2014 Set Theory of Semantic
Theories of Truth & Metaphysical Basis of Logic (STSTT/MBL) series at the Northern
Institute of Philosophy, for helpful feedback.
References
Azzouni, J. 1994. Metaphysical Myths, Mathematical Practice: The Ontology and
Epistemology of the Exact Sciences. Cambridge: Cambridge University Press.
Balaguer, M. 1995. “A Platonist Epistemology,” Synthese 103: 303–25.
Balaguer, M. 1999. “Review of Resnick,” Philosophia Mathematica 7: 108–26.
Benacerraf, P. 1973. “Mathematical Truth,” Journal of Philosophy 70: 661–79.
Bengson, J. Forthcoming. “Grasping the Third Realm,” Oxford Studies in
Epistemology.
Berker, S. 2014. “Does Evolutionary Psychology Show That Normativity Is Mind-
Dependent?,” in J. D’Arms and D. Jacobson (eds.), Moral Psychology and Human
Agency: Philosophical Essays on the Science of Ethics. Oxford: Oxford University
Press, 215–52.
Boyd, R. 2003a. “Finite Beings, Finite Goods: The Semantics, Metaphysics and
Ethics of Naturalist Consequentialism, Part I,” Philosophy and Phenomenological
Research 66: 505–53.
Boyd, R. 2003b. “Finite Beings, Finite Goods: The Semantics, Metaphysics and
Ethics of Naturalist Consequentialism, Part II,” Philosophy and Phenomenological
Brink, D. 1989. Moral Realism and the Foundations of Ethics. New York: Cambridge
University Press.
Chihara, C. 1990. Constructability and Mathematical Existence. Oxford: Oxford
University Press.
Clarke-Doane, J. 2012a. “Morality and Mathematics: The Evolutionary Challenge,”
Ethics 122, 313–40.
Clarke-Doane, J. 2012b. “Response to Braddock, Mogensen, and
Sinnott-Armstrong—Part II,” Ethics at PEA Soup. Available at: <http://peasoup.
typepad.com/peasoup/2012/03/ethicsdiscussions-at-pea-soup-justin-clarke-doa
nes-morality-and-mathematics-the-evolutionary-challe1.html>.
Clarke-Doane, J. 2014. “Moral Epistemology: The Mathematics Analogy,” Noûs
38: 238–55.
Clarke-Doane, J. Forthcoming a. “What is the Benacerraf Problem?,” in F. Pataut
(ed.), New Perspectives on the Philosophy of Paul Benacerraf: Truth, Objects, Infinity.
Clarke-Doane, J. Forthcoming b. “Genealogy and Reliability,” in N. Sinclair and
U. Leibowitz (eds.), Ethics and Explanation.
Colyvan, M. 2007. “Mathematical Recreation versus Mathematical Knowledge,”
in M. Leng, A. Paseau, and M. Potter (eds.), Mathematical Knowledge.
Darwin, C. 1871. The Descent of Man. New York: Appleton.
Ellis, B. 1990. Truth and Objectivity. Oxford: Basil Blackwell.
Field, H. 1980. Science without Numbers. Princeton: Princeton University Press.
Field, H. 1989. Realism, Mathematics, and Modality. Oxford: Blackwell.
Field, H. 1996. “The A Prioricity of Logic,” Proceedings of the Aristotelian Society
96: 359–79.
Field, H. 2005. “Recent Debates about the A Priori,” in T. S. Gendler and J.

Hawthorne (eds.), Oxford Studies in Epistemology, Volume 1. Oxford: Oxford
University Press, 69–88.
Gödel, K. 1947. “What is Cantor’s Continuum Problem?” Revised and reprinted in
P. Benacerraf and H. Putnam (eds.), Philosophy of Mathematics. Englewood Cliffs,
NJ: Prentice-Hall, 1964.
Harman, G. 1977. The Nature of Morality: An Introduction to Ethics. New York: Oxford
University Press.
Hart, W. 1996. “Introduction,” in W. Hart (ed.), The Philosophy of Mathematics.
Hawthorne, J. 2006. Metaphysical Essays. Oxford: Oxford University Press.
Heller, M. 1990. The Ontology of Physical Objects: Four-Dimensional Hunks of
Matter. New York: Cambridge University Press.
Hellman, G. 1989. Mathematics Without Numbers. Oxford: Oxford University Press.
Horgan, T. 1987. “Discussion: Science Nominalized Properly,” Philosophy of Science
54: 281–2.
Joyce, R. 2001. The Myth of Morality. Cambridge: Cambridge University Press.
Joyce, R. 2006. The Evolution of Morality. Cambridge, MA: MIT Press.
Joyce, R. 2008. “Precis of Evolution of Morality and Reply to Critics,” Philosophy
and Phenomenogical Research 77: 213–67.
Joyce, R. Forthcoming. “Evolution, Truth-Tracking, and Moral Skepticism,” in
B. Reichardt (ed.), Problems of Goodness: New Essays on Metaethics. Draft available
at: <http://personal.victoria.ac.nz/richard_joyce/onlinepapers.html>.
Kitcher, P. 2006. “Biology and Ethics,” in D. Copp, (ed.), The Oxford Handbook of
Ethical Theory. Oxford: Oxford University Press, 163–85.
Korman, D. 2011. “Ordinary Objects,” The Stanford Encyclopedia of Philosophy
(Winter Edition), ed. Edward N. Zalta. Available at: <http://plato.stanford.edu/
entries/ordinary-objects/>.
Korman, D. 2014. “Debunking Perceptual Beliefs about Ordinary Objects,”
Philospoher’s Imprint 14: 1–21.
Merricks, T. 2001. Objects and Persons. Oxford: Clarendon Press.
Nozick, R. 1981. Philosophical Explanations. Oxford: Oxford University Press.
Pritchard, D. 2008. “Safety-Based Epistemology: Whither Now?,” Journal of
Philosophical Research 34: 33–45.
Putnam, H. 1971. The Philosophy of Logic. New York: Harper & Row.
Quine, W. V. O. 1951. “Two Dogmas of Empiricism,” Philosophical Review
60: 20–43.
Quine, W. V. O. 1986. “Reply to Charles Parsons,” in L. E. Hahn and P. A. Schilpp
(eds.), The Philosophy of W. V. Quine. La Salle, IL: Open Court, 396–403.
Ross, W. D. 1939. Foundations of Ethics. Oxford: Oxford University Press.
Ruse, M. 1986. Taking Darwin Seriously. Oxford: Blackwell.
Schecther, J. 2010. “The Reliability Challenge and the Epistemology of Logic,”
Philosophical Perspectives 24: 437–64.
Sider, T. 2001. Four-Dimensionalism. Oxford: Oxford University Press.
Sinnott-Armstrong, W. 2006. Moral Skepticisms. Oxford: Oxford University Press.
Steiner, M. 1973. “Platonism and the Causal Theory of Knowledge,” Journal of

Philosophy 70: 57–66.
Street, S. 2006. “A Darwinian Dilemma for Realist Theories of Value,” Philosophical
Studies 127: 109–66.
Street, S. 2008. “Reply to Copp: Naturalism, Normativity, and the Varieties of
Realism Worth Worrying About,” Philosophical Issues 18, 207–228.
Street, S. n.d. “Objectivity and Truth: You’d Better Rethink It.” Available at: <https://
files.nyu.edu/ss194/public/sharonstreet/Writing.html>.
Sturgeon, N. 1985. “Moral Explanations,” in D. Copp and D. Zimmerman (eds.),
Morality, Reason, and Truth: New Essays on the Foundations of Ethics. Totowa,
NJ: Rowman and Allanheld, 49–78.
Sturgeon, N. 1986. “Harman on Moral Explanations of Natural Facts,” The Southern
Journal of Philosophy (Supplement) 24: 69–78.
Van Inwagen, P. 1990. Material Beings. Ithaca, NY: Cornell University Press.
5
Deliberative Indispensability and
Epistemic Justification
Tristram McPherson and David Plunkett
5.1 INTRODUCTION
According to one influential view in metaethics (e.g. Harman 1977; Boyd

1997), we are justified in believing in ethical facts just in case they (or their
reduction base) feature in our best explanations of scientifically respectable
phenomena. This naturalistic criterion, however, can seem to miss a crucial
point. Many of us care about the existence of ethical facts not because of
scientific-explanatory roles that they may play, but rather because we seem-
ingly need such facts to make adequate sense of our practical lives. This sug-
gests a Tempting Idea: that the (putative) indispensability of belief in ethical
facts for our practical projects—including, for example, the project of deliber-
ating about what to do—can justify our belief in such facts. Some version of
this idea has attracted a range of philosophers, including Christine Korsgaard
(1996), Ronald Dworkin (2011), and T. M. Scanlon (2014).
Any philosopher hoping to develop the Tempting Idea needs to answer two
questions. First: which beliefs are relevantly indispensable? And second: what is
the significance of this indispensability? David Enoch has recently spelled out a
powerful and novel version of the Tempting Idea, which he develops primarily
in “An Outline of an Argument for Robust Metanormative Realism” (2007)
and chapter 3 of his Taking Morality Seriously (2011b). Enoch’s account offers
ambitious answers to each of these questions.
In response to the first question, Enoch argues for the indispensability
of belief in what he calls Robust Realism about ethical facts.1 According to
1
Three clarificatory notes. First, strictly speaking, Enoch takes certain belief-forming
methods to be indispensable. In calling belief in Robust Realism “indispensable” here, we
Deliberative Indispensability and Epistemic Justification 105
Robust Realism, ethical facts are ungrounded, irreducibly normative, and

psychology-independent. Further, Enoch claims that Robust Realism is
incompatible with a quietist or quasi-realist interpretation of these meta-
physical claims. Enoch argues that such belief is indispensable to the project
of practical deliberation: the project that we engage in when we seek to
choose what to do (2011b: 70–3). This idea is intensely controversial and
well worth examining. However, in this chapter we set it aside, in order to
focus on Enoch’s answer to the second question.
In response to the second question, Enoch argues that deliberative indis-
pensability is significant because such indispensability can epistemically jus-
tify belief. More precisely, Enoch defends:
Indispensabilism If a belief-forming method is indispensable to the
project of practical deliberation, then that method is a source of basic
epistemic justification.
For someone hoping to develop the Tempting Idea into a response to the
naturalistic epistemic criterion mentioned at the beginning of this chap-
ter, Indispensabilism has three striking virtues. First, it claims to under-
write epistemic justification (the Tempting Idea itself is silent on the type
of justification provided). If defensible, it thus constitutes a direct rebuttal
to the naturalistic criterion. Absent this claim, a defense of the Tempting
Idea threatens to suggest that our capacity for practical deliberation dooms
us—either causally or rationally—to epistemic irrationality. Second,
Indispensabilism promises to ameliorate a standard worry about metaethi-
cal views that violate the naturalistic criterion: that they are forced to posit
a special capacity to directly perceive non-natural ethical facts. A proponent
of Indispensabilism can argue that our justification for belief in such facts is
explained by the deliberative indispensability of such belief, rather than by
a mythical perceptual capacity. A third virtue of Indispensabilism is its neu-
trality concerning the nature and grounds of ethical facts. One might try to
defend the Tempting Idea in part by arguing that facts about indispensabil-
ity explain the fundamental ethical facts. By contrast with this approach, if
signal that, according to Enoch, it follows from the proper deployment of an indispen-
sable method. We explain this part of Enoch’s reasoning in more detail below (in section
5.2). Second, we treat facts as the standard metaphysical relata throughout, while Enoch
typically talks of truths. We take this change to be unobjectionable given Enoch’s com-
mitments, a point that he himself emphasizes (2011b: 5). Third, we talk of ethical facts,
where Enoch tends to talk of normative facts. We mean “ethical” here broadly, to refer to
the normative and evaluative facts that govern our practical lives. We insist on this change
in wording because epistemic facts are also normative, and because the contrast between
ethical and epistemic normativity is central to our project here.
106 Tristram McPherson and David Plunkett
Indispensabilism could be defended, it might be adapted to permit propo-

nents of a variety of metaethical views to vindicate the Tempting Idea.
For these reasons, we take Indispensabilism to be an important thesis.
However, in this chapter, we argue that Indispensabilism should be rejected.
The core reason is this: Indispensabilism conflicts with part of what is dis-
tinctive of epistemic justification.
The distinctiveness of epistemic justification can be suggested by the fol-
lowing thought-experiment. Suppose that Hallie believes that when she
sings “Don’t Stop Believin’ ” in the shower, she sounds exactly like Journey’s
Steve Perry. This belief is strikingly irrational: merely attending carefully
to the sound of her own voice would suffice to disabuse her of this belief,
and her trustworthy friends have let her know how silly her belief is. Now
suppose that an evil demon lets Hallie know that if she ceases to hold this
belief, the demon will brutally torture every sentient being that exists. This
fact gives Hallie very strong practical reasons to retain her belief, but evi-
dently does nothing to epistemically justify her belief. The case of Hallie
dramatizes the familiar point that ethical and epistemic normativity appear
to be very different things, by showing that one can have overwhelming eth-
ical justification for a belief, while lacking any epistemic justification for it.2
This stark contrast helps to frame our thesis. At best, the fact that some-
thing is deliberatively indispensable can perhaps provide ethical reasons for
belief. However, it is not the right sort of thing to underwrite epistemic
justification. Thus, Indispensabilism must be rejected.
Our chapter proceeds as follows. We start by laying out our exemplary
stalking horse: Enoch’s case for Indispensabilism (section 5.2). We then
argue that Enoch’s case fails because it elides a distinctive feature of genu-
inely epistemic justification, a feature that we dub Truth-Directedness (sec-
tion 5.3). Briefly, according to Truth-Directedness, the norms of epistemic
justification have the content they do because of some positive connection
to the truth of the beliefs these norms govern. (We give a more precise
characterization of this thesis in section 5.3.) Because Enoch’s account fails
to respect a crucial part of what is distinctive about epistemic justification,
it thereby fails as a defense of Indispensabilism (which, recall, is a thesis
about epistemic justification). Enoch’s specific proposal, however, is just one
possible attempt to defend Indispensabilism. We go on to argue that our
2
There might be other ways of drawing the distinction between Hallie’s different rea-
sons that would work for our argumentative purposes in this chapter. Our point is that
however one draws this distinction, the case of Hallie brings out an intuitive and impor-
tant contrast between two different sets of norms that an agent can be subject to. We are
characterizing one set of these norms as the norms of genuine epistemic justification, and
we submit that it is deeply intuitive that only one set of these norms can plausibly be
understood this way.
objection generalizes to a range of salient alternatives to Enoch’s way of

defending Indispensabilism. These alternatives are based on leading general
approaches in contemporary epistemology. We argue that looking at these
alternative defenses of Indispensabilism reveals a general tension between
Truth-Directedness and Indispensabilism (and not just Enoch’s particular
defense of it). We thus conclude that deliberative indispensability does not
epistemically justify belief (section 5.4). Although our main argument in
this chapter is directed at Indispensabilism, part of our goal here is to get
clearer on what it would take to defend the initial Tempting Idea, and to
assess the prospects for doing so. We thus examine three ways of attempting
to salvage the Tempting Idea, in the face of our argument. We briefly argue
that each of these attempts faces significant costs. In light of this, we suggest
that pessimism is warranted about the Tempting Idea itself (section 5.5).
We conclude the paper with brief reflections on the broader metaethical
significance of our arguments (section 5.6).
5.2 ENOCH’S DEFENSE OF INDISPENSABILISM
In this section and the next, we aim to illustrate the promise and perils of
Indispensabilism by exploring its powerful recent defense by David Enoch.
In this section, we first explain the role of that defense in Enoch’s case for
his Robust Realism about ethical facts. This illuminates part of the potential
metaethical significance of Indispensabilism. We then lay out the details of
Enoch’s case for Indispensabilism, which puts us in a position to evaluate it
in the following section.
5.2.1 From Deliberative Indispensability to Robust Realism

To understand any indispensability argument, one must understand the
notion of indispensability being deployed. We thus begin by unpacking
two technical notions that Enoch introduces: an intrinsically indispensable
project, and something’s being instrumentally indispensable to such a project.
We then introduce the outline of Enoch’s ambitious deliberative indispen-
sability argument for Robust Realism about ethical facts.
On Enoch’s account, a project is intrinsically indispensable if it is “ration-
ally non-optional,” such that disengaging from it “is not a rationally accept-
able option” (2011b: 70).3 This entails that one is rationally criticizeable if
3
While some philosophers (e.g., Broome 1999; Kolodny 2005) use “rationality” to
talk about distinctively structural normativity, Enoch uses “rationality” here as a way to
talk about substantive normative facts.
one does disengage from such a project. It is worth emphasizing that this
account is explicitly normative. A project thus does not count as intrinsi-
cally indispensable in Enoch’s sense simply because one is stuck with engag-
ing in it (in the non-normative sense of “stuck with”).4
There can be necessary conditions for pursuing such a project. According
to Enoch, for something to be instrumentally indispensable to a project is for
its elimination to undermine or attenuate the reasons that we had for engag-
ing in that project in the first place. Crucially, Enoch wants to distinguish
such instrumentally indispensable features from mere “enabling conditions”
for a project. For example, take the relationship between getting adequate
sleep and the project of engaging in scientific inquiry. Enoch claims that
while getting enough sleep might be an enabling condition for this project,
it is not instrumentally indispensable to this project, in the sense of “instru-
mentally indispensable” that he has in mind (2011b: 68).
With these clarifications in hand, we can now present Enoch’s overall
indispensability argument in schematic form:5
1. If (implicitly or explicitly) treating a belief-forming method as a source
of basic epistemic justification is instrumentally indispensable to an
intrinsically indispensable project, then that method is a source of basic
epistemic justification (2011b: 60–4).
2. The project of practical deliberation is intrinsically indispensable (2011b:
70–3).
3. Treating our commitments in practical deliberation as a source of basic
epistemic justification is instrumentally indispensable to the deliberative
project (cf. 2011b: 67–9).
4. Therefore, our commitments in practical deliberation are a source of
basic epistemic justification (from premises 1–3).
4
On this point, Enoch (2011b) differs from his (2007) presentation of his indispen-
sability argument. This change makes Enoch’s metaethical views more consistent: as he
himself notes, the earlier version of his argument faces his own “schmagency” challenge
to attempts to explain authoritative normativity in terms of necessary facts about agency
(Enoch 2006). The change also allows Enoch to avoid intuitive worries that afflict his
earlier argument. For example, we can imagine possible creatures who are doomed to
engage in worthless projects—perhaps because they were designed to be doomed in this
way. It is especially hard to see why serving such a project could epistemically justify
otherwise unsupported beliefs. For these reasons, we take Enoch’s explicitly normative
gloss on intrinsic indispensability in his (2011b) to be a significant improvement on the
original (2007) version.
5
Enoch provides a simpler schematic summary of his argument (2011b: 83).
However, because that reconstruction elides detail in his argument that is crucial to our
discussion here, we have provided our own, slightly more complex summary here.
5. In practical deliberation, we are committed to belief in the existence of

ethical facts, as they are conceived of by Robust Realism (2011b: 71–9).
6. Therefore, (because sources of basic epistemic justification provide
defeasible epistemic justification) we have defeasible epistemic justifica-
tion for believing in the existence of Robustly Real ethical facts (from
premises 4–5).
Premises (1) and (2) of this argument entail (a variant of ) Indispensabilism.
The remainder of the argument shows that, together with the further claims
about practical deliberation in premises (3) and (5), Indispensabilism can
support an ambitious metaethical view: namely, Robust Realism about
ethical facts. Note that at various points in this chapter, we will abbreviate
the sort of case just sketched for (6)— and theses like it—by saying that,
according to Enoch, deliberative indispensability provides basic justification
for believing that P.
This argument illustrates the potential metaethical significance of
Indispensabilism, and also illustrates why it is such a powerful way of
developing the Tempting Idea that we introduced at the start of this
chapter. Enoch appears to offer a clear and principled account of how
epistemically justified belief in non-natural ethical facts is possible, and,
moreover, to do so in a way that also provides a kind of positive argu-
ment for Robust Realism in metaethics. It is a positive argument for
Robust Realism (a metaphysical thesis) for the following straightforward
reason: it is telling you that you have epistemic reason to believe this
metaphysical thesis.
We have significant worries about premises (2), (3), and (5) of the argu-
ment.6 However, the metaethical bite of Indispensabilism extends beyond
Enoch’s own defense of this thesis. This is because the basic Indispensabilist
idea could potentially be combined with a variety of auxiliary commitments
(in lieu of premises (2–5)), to epistemically justify various commitments in
or about ethics. The epistemic heart of Enoch’s argument that achieves these
results is premise (1). In the next subsection, we thus explore Enoch’s case
for this premise in detail.
6
Here are two examples. First, premise (3) is challenged by the existence of credible
anti-intuitionist approaches to moral epistemology. Second, with premises (2) and (5),
Enoch faces a version of a dilemma he himself has pressed against the constitutivist: the
more you build into a conception of practical deliberation, the less plausible it is that
doing that is rationally non-optional (2011b: 71–2). We find it especially doubtful that
belief in the existence of ethical facts, as conceived of by Robust Realism is delibera-
tively indispensable. For related challenges, see Husi (2013: §4), Lenman (2014), and
Björnsson and Olinder (forthcoming).
5.2.2 Enoch’s Strategy for Vindicating Indispensabilism

Enoch’s defense of Indispensabilism presupposes a specific kind of foun-
dationalist picture of the structure of epistemic justification. On this pic-
ture, certain belief-forming methods are epistemically derivative while others
are epistemically basic. Consider an example of an epistemically derivative
method: someone might be justified in using the results of a DNA test as
evidence of paternity, but only because she has prior evidence of the reli-
ability of the test. By contrast, consider belief-forming methods such as
reliance on sense perception and memory, inference to the best explanation,
and inference rules like modus ponens. Enoch claims that these methods are
epistemically basic: using these methods can produce defeasibly epistemi-
cally justified belief, even when we lack independent epistemic justification
for using them (Enoch 2011b: 58; cf. Enoch and Schechter 2008). For
uniformity, we will call these methods sources of basic epistemic justification.
Enoch argues that philosophers who endorse the foundationalist pic-
ture face the burden of explaining the facts in virtue of which only some
belief-forming methods are basic justifiers (Enoch 2011b: 59ff.; Enoch
and Schechter 2008: 547). What, we might demand, explains the contrast
between reliance on memory and reliance on DNA testing? Note that this
is not a request for an epistemic justification for treating certain sources
as basic. Rather, the question is: what explains why these sources have the
status of being epistemically basic?
Enoch calls the sort of explanation he is after here a “vindication.” The
details of Enoch’s own glosses on this term are not totally clear. We will
understand a vindication as an explanation of the distinctive epistemic sta-
tus of the methods that are basic sources of justification, where this explana-
tion fits with (or, ideally, supports) the intuitive normative significance of
these sources. This is in contrast to an explanation that debunks that pur-
ported significance or reforms it away (cf. 2011b: 60). One might be able
to provide a vindication in this sense by using one of a variety of different
types of philosophical explanation.7 We think that Enoch’s own approach
to offering a vindication is best understood as a grounding account: a meta-
physical account that explains the facts in virtue of which certain sources
provide basic epistemic justification.8 We join Enoch in taking the demand
7
Consider two familiar alternatives. First, one might propose a vindication of the
sources of basic epistemic justification by providing an ontological reduction of the source
of basic epistemic justification relation. Second, a proponent of ambitious conceptual anal-
yses (à la Jackson 1998) might propose a vindication via an analysis of the concept basic
source of justification that illuminated its extension.
8
For a helpful overview of grounding in contemporary metaphysics, see Trogdon
(2013).
for a vindication of the sources of basic epistemic justification to be force-

ful (modulo the controversial assumption of the truth of foundationalism).
And we are happy to grant for the sake of argument that such a vindication
should take the form of a grounding account. Our concern in this section is
with the specific grounding account that Enoch proposes. Enoch originally
developed the core of this account in joint work with Joshua Schechter
(Schechter and Enoch 2006: §6; Enoch and Schechter 2008).
Because Enoch calls his vindicating account a “pragmatic” one, we will
appropriate this handy (if slightly misleading) label to refer to his account.
This account can be stated as follows:
Pragmatic One complete ground for the fact that something is a
source of basic epistemic justification is the fact that treating it as
such a basic source is instrumentally indispensable to an intrinsically
indispensable project.9
Pragmatic trivially entails premise (1) of Enoch’s argument. And it is thus
the crucial step in his case for Indispensabilism (we are granting him the
other element of that case: the assumption that practical deliberation is an
intrinsically indispensable project). En route to assessing Pragmatic’s plau-
sibility as a vindicating account for the sources of basic epistemic justifica-
tion, we sketch three virtues of this thesis.
Pragmatic’s first virtue is that it appears to offer a credible explanation of
the substantive normativity of the basic epistemic justification facts. What
do we mean by this? Contrast epistemic norms with the norms of chess,
or fashion. Indifference to the “epistemic evaluation game” seems like an
objective flaw, in a way that indifference to chess or the norms of fash-
ion does not: the epistemic norms appear to have normative substance in
a way these other norms do not.10 As we have seen, Enoch’s conception
of indispensability appeals to ethical facts, which are themselves substan-
tively normative. It thus entails that basic epistemic justification facts will
be grounded partly in substantively normative facts. Because grounding can
arguably transmit normativity, being grounded in substantively normative
facts seems like a promising way to explain the substantive normativity of
the epistemic justification facts.
9
Enoch’s clearest official statement of his thesis (2011b: 63) provides a mere suffi-
ciency condition for being a source of basic epistemic justification. Enoch clearly intends
the thesis to be explanatory, and our formulation reflects that fact. It should be noted that
many important motivations for this thesis (including both motivations that we discuss
below) would be more compelling if Pragmatic were strengthened to purport to explain
the complete grounds of all basic sources of justification.
10
See McPherson (2011: section 4) for a brief exploration of this contrast, in terms of
“formal” vs. “robust” normativity.
Second, Enoch suggests that Pragmatic is plausible in part because it is

capable of explaining the epistemic status of a plausible range of the sources
of basic epistemic justification. Enoch takes inference to the best explana-
tion (IBE) as his leading illustration. On Enoch’s account, IBE is a source
of basic epistemic justification because (i) the project of understanding and
explaining the world around us is rationally non-optional, and (ii) deploy-
ing IBE is instrumentally indispensable for creatures like us pursuing this
project (2011b: 60–1).
Enoch’s focus on IBE has a further, dialectical payoff. The naturalistic cri-
terion for justifying commitment to ethical facts that we introduced at the
beginning of this chapter appeals crucially to IBE. And this criterion seri-
ously threatens Enoch’s Robust Realism, since, on his view, the fundamental
ethical facts are irreducible and do not explain anything non-normative. It
also threatens all arguments from deliberative indispensability, since delib-
erative indispensability does not entail explanatory indispensability.
If we suppose that Enoch is right that Pragmatic provides the most plau-
sible vindication for the epistemic status of IBE, however, the threat posed
by the naturalistic criterion is neutralized. This is because (as Enoch argues)
Pragmatic can vindicate other sources of basic epistemic justification
besides IBE. Further, deliberative indispensability arguments are no longer
threatened, because deliberative indispensability is claimed to explain the
epistemic status of IBE itself.
Enoch’s case for Pragmatic is part of a clear and carefully developed argu-
ment for Indispensabilism. The virtues just canvassed so far also suggest
that this argument for Indispensabilism is promising. However, in the next
section, we argue that despite its promising features, Enoch’s defense of
Pragmatic is ultimately unsuccessful.
5.3 EPISTEMIC JUSTIFICATION AND TRUTH
This section sets out our case against Pragmatic—Enoch’s vindicating

account of the sources of basic epistemic justification. We begin by articu-
lating and defending a partial characterization of what is distinctive of epis-
temic justification: that it is directed at the truth (section 5.3.1). We then
introduce three intuitive counterexamples to Pragmatic, and argue that the
force of these counterexamples is well explained by the truth-directedness
of epistemic justification (section 5.3.2). We argue that our theory of the
truth-directed nature of epistemic justification, coupled with these counter-
examples, gives us strong reason to reject Pragmatic (section 5.3.3).
A vindicating account of a phenomenon is supposed to explain and

uphold our intuitive commitments with respect to that phenomenon, rather
than debunking or substantially reforming them. A vindicating account of
the sources of basic epistemic justification should thus accomplish at least
the three following goals. First, it should explain (or at least be compatible
with) the most plausible theses about which sources of epistemic justifica-
tion are basic. Second, it should explain (or at least be compatible with) the
apparent normative substantiveness of the norms of epistemic justification.
Third, it should explain (or at least be compatible with) our sense of what is
distinctive of the norms of epistemic justification.
In the previous subsection, we saw that Enoch makes a prima facie case
that Pragmatic meets the first desideratum, by arguing that Pragmatic can
explain the status of belief-forming methods like IBE. We also saw that
there is a good case to be made that it meets the second desideratum. This
is because Pragmatic grounds facts about the sources of basic epistemic jus-
tification partly in ethical facts.
The third desideratum demands that a vindicating account of the sources
of basic epistemic justification explain (or at least be compatible with) what
is distinctive of epistemic justification. There is little in Enoch’s work that
explicitly addresses the third desideratum. Further, recall the case of Hallie
that we set out in section 5.1. As this case shows, epistemic and ethical justi-
fication look like radically different creatures. This licenses initial suspicion
that an account like Pragmatic, which seeks to ground the sources of basic
epistemic justification partly in ethical norms, will struggle to satisfy this
desideratum. We argue that this initial suspicion is warranted: Pragmatic
should be rejected precisely because it is incompatible with a central dis-
tinctive feature of epistemic justification. This is a feature that we call
Truth-Directedness.
5.3.1 Truth-Directedness
Recall the case of Hallie and the demon. The demon will torture every sen-
tient being if Hallie ceases to believe that she sounds exactly like Journey’s
Steve Perry when she sings “Don’t Stop Believin’ ” in the shower. As we
emphasized, this fact fails to provide epistemic justification for Hallie’s con-
tinuing to believe that her singing voice sounds like Steve Perry’s. A compel-
ling explanation of this failure is that this fact about the consequences of
Hallie’s belief is wholly unconnected to the truth of the proposition that her
singing voice sounds like Steve Perry’s. If this diagnosis is right, it suggests
that any adequate explanation of the sources of basic epistemic justification
will need to appeal in a central way to some link between those sources and
true belief.
We endorse a specific, although highly schematic, account of this link:

Truth-Directedness The sources of basic epistemic justification have
the content that they do (in part) because of some positive connection
between those sources and the truth of the beliefs that they govern.
Note that, as the “(in part)” locution suggests, Truth-Directedness provides
only a necessary condition: it is compatible with there being further con-
ditions that a source of basic epistemic justification needs to satisfy. We
now clarify three elements of this thesis: the appeal to truth, to explanation
(“because”), and to positive connection.
First, in adverting to truth, we do not intend to commit ourselves to a
specific account of truth. To see this, note that one could restate our thesis
(more clumsily) in terms of a positive connection between (a) belief that
P and (b) P. Many philosophers with a range of views about truth—both
substantive and minimalist—should find this thesis attractive.
Second, the thesis asserts that the connection between the truth and the
sources of basic epistemic justification must be explanatory. The thesis is
thus a constraint on theories that purport to explain why something is a
source of basic epistemic justification. According to Truth-Directedness, all
such theories must advert to some positive connection to truth as a crite-
rion. Consider an example where this connection fails: a crude epistemic
divine command theory. This theory states that a belief-forming mecha-
nism’s being a source of basic epistemic justification is grounded in God’s
commanding you to treat it as basic. This theory violates the explanatory
requirement of Truth-Directedness. Note that even if God in fact ensured
that the sources he commanded you to rely on are reliable, the link to truth
thereby secured is not part of this theory’s account of what explains why
something is a source of basic epistemic justification. We think that the fact
that this theory violates Truth-Directedness is one sufficient explanation of
why this theory fails as an account of epistemic justification.
Crucially, Truth-Directedness is a demand on the theory that explains
epistemic basicness, not on the basic sources themselves. Truth-Directedness
thus does not require that a source of basic epistemic justification must
“have truth as a goal.” For example, conceptual competence with the con-
cept and does not obviously involve having truth as a goal, any more than
competence with the concept tonk does. This is compatible with a theory
of the sources of basic epistemic justification saying (for example) that our
competence with the former, and not the latter, is a source of basic epis-
temic justification, in part because the first concept is truth-preserving while
the second is not.11
11
This simple proposal is intended only as an illustration. Perhaps, as Schechter and
Third, Truth-Directedness requires some positive connection between the

sources of basic epistemic justification and truth. We intend this thesis in
a very ecumenical spirit. Thus, some epistemologists might understand the
positive connection in modal terms: for example in terms of reliability or
safety. Others might flesh it out in terms of the constitutive goals of epis-
temic agents, or in terms of conditions for responsible pursuit of the truth.
Still others might appeal to conditions for the possibility of the pursuit of
truth, or on pursuing the truth efficiently.
To underscore the ecumenical spirit of our thesis, consider three more
examples of controversial theses that Truth-Directedness is compatible with,
but distinct from. First, some philosophers are attracted to the idea that
belief “aims at the truth” in some interesting sense.12 If combined with the
thought that the epistemic norms are the constitutive norms of belief, this
sort of approach could be developed into an appropriately truth-directed
grounding account of the sources of basic epistemic justification. Second,
Selim Berker (2013) argues against a view that he calls “epistemic teleology,”
which combines Truth-Directedness with the assumption that “directed-
ness” should be construed as a promotion relation, and that the goods to be
promoted can be aggregated. The thesis defended here is compatible with
either accepting or rejecting these further assumptions.
Third, pragmatic encroachment about knowledge is, roughly, the view that
whether a given body of evidence suffices to put a subject in a position to
know that P can depend upon the practical stakes involved in the agent’s rely-
ing upon P in their practical deliberation. Critics of pragmatic encroachment
about knowledge sometimes appeal to claims similar to Truth-Directedness.
However, the Truth-Directedness of epistemic justification is compat-
ible with the most influential defenses of pragmatic encroachment about
knowledge (e.g. Hawthorne 2004; Stanley 2005; Fantl and McGrath 2009).
Distinguish two issues: (1) when a token source of evidence provides (at
least some) justification for believing that P vs. (2) the threshold of justifica-
tion required for knowledge, or for adequately epistemically justified belief.
Standard views of pragmatic encroachment address the second issue, but we
intend Truth-Directedness to be a thesis about the first.13
Enoch (2006: 705) worry, direct appeal to truth-preservation in this example is objec-

tionably reliabilist. That would not count against Truth-Directedness, because there are
multiple ways of making good on Truth-Directedness that are not reliabilist. See section
5.4 for discussion.
12
For some of the different takes on how to best understand the idea that “belief aims
at the truth” see Velleman (2000), Wedgwood (2002), Shah (2003), and Gibbard (2008).
13
For example, all of the philosophers cited above appeal to (roughly) the claim
that if one knows that P, one can rationally rely on P in one’s practical reasoning.
Mark Schroeder (2012) offers an additional rationale for pragmatic encroachment,
As the discussion above makes clear, Truth-Directedness is a highly sche-

matic theory in the foundations of epistemology. This indeterminacy might
seem objectionable, but it is not. Rather, it is dialectically crucial. Disputes
between the sorts of approaches mentioned above are central to contem-
porary epistemology. Truth-Directedness is intended to be powerfully ecu-
menical, in virtue of being compatible with all of these approaches. Indeed,
we take commitment to something like Truth-Directedness to be close to
common ground in many parts of epistemology. Thus, when epistemolo-
gists seek to explain the foundations of epistemic justification, they often
predictably advert to similar theses. Here is one representative example,
from Paul Moser:14
Epistemic justification is 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.
(Moser 1985: 4)
This is not to say that Truth-Directedness is uncontroversial (see, promi-
nently, Wright 2004 on entitlement). However, the presumption of some-
thing like Truth-Directedness is so entrenched that it can be hard for even
a radical to put it into question. For example, Stephen Stich reports that,
when philosophers first confronted his heterodox argument that truth
should not be the aim of our epistemic practices, many of them sim-
ply “assume[d]‌I must be joking, or propounding silly skeptical puzzles”
(1990: 101).15
There is a powerful explanation of the strength and breadth of endorse-
ment of theses similar to Truth-Directedness. As we have noted, epistemic
justification appears both to be substantively normative and (as the example
of Hallie shows) very different from ethical justification. Truth-Directedness
promises to explain both of these features. On the one hand, the link to
truth is distinctive: for example, there does not appear to be a parallel
arguing that even if only evidence constitutes reason to believe, pragmatic considera-
tions can enter into an account of knowledge by providing reasons to suspend judg-
ment. Neither of these rationales can easily be adapted to an account of the sources
of basic epistemic justification. Note, however, that if one combined pragmatic
encroachment with the view that the evidence that constitutes a subject’s justification
for belief just is that subject’s knowledge, pragmatic encroachment on knowledge
would also infect prima facie justification. Such a combination is not compatible with
Truth-Directedness.
14
See Berker (2013: §3) for dozens of endorsements by epistemologists of similar
(often stronger) theses about the relationship between epistemic justification and truth.
15
We take Stich’s radical challenge to be significant. We set it aside here, however,
because addressing it would require an entirely distinct sort of argument.
constraint on our ethical norms.16 On the other hand, the substantiveness

of epistemic normativity can potentially be explained in part by appeal to
the intrinsic or instrumental significance of true belief.17
Truth-Directedness also helps to address an important challenge that
Enoch poses to those philosophers (such as ourselves) who want to reject
his account of epistemic justification, but who also want to preserve the
role of explanatory indispensability in epistemic justification. The chal-
lenge is to identify a principled distinction between those kinds of indis-
pensability that can justify belief, and those that do not (2011b: 67).
Truth-Directedness provides the tools needed to accomplish this task.
Truth-Directedness is a constraint on candidate vindicating theories,
and not a vindicating theory itself. It is thus compatible with many differ-
ent ways of seeking to vindicate the sources of basic epistemic justification
(for example). However, we know what attempts to vindicate inference to
the best explanation (IBE) within a truth-directed framework would look
like. This is because the overwhelming majority of the literature on the
vindication of IBE presupposes something like Truth-Directedness. For
example, Peter Lipton (2004: ch. 11) and Igor Douven (2011: §3) assume
without comment that a “justification” of IBE (i.e., roughly, a vindica-
tion) will show that IBE is “truth-tropic” or reliable. And the discussion
above suggests alternative accounts that appeal to epistemic responsibility
or virtue, or to transcendental conditions on the pursuit of truth. Fully
developing any such theory would be no small task. But that is not our
goal. Our aim here is to explain how, with Truth-Directedness in hand,
one can have confidence that explanatory indispensability could be vin-
dicated as a source of basic epistemic justification, even if deliberative
indispensability is not.
In this subsection, we have introduced Truth-Directedness, and made
what we take to be a strong initial case for its plausibility. To sum up: it is
intuitively plausible; it is ecumenical concerning many central controversies
in epistemology; and it promises to explain the distinctive normativity of
epistemic justification. This is the first half of our case for this thesis. The
next section completes that case, by sketching three counterexamples to
Pragmatic, and arguing that Truth-Directedness can well explain why these
cases are powerful objections to Pragmatic.
16
Compare Alston (2005): Alston abandons the idea that there is a single category of
epistemic justification. Still, he is able to characterize various features as epistemic desid-
erata in large part because all of them are in some way or another truth-directed.
17
One example: the central “problem of normativity” in contemporary
knowledge-centric epistemology is arguably to explain why knowledge is relevantly bet-
ter than mere true belief, given that truth is the fundamental normative currency of
epistemology. Cf., e.g., Sosa (2007: Lecture 4).
5.3.2 Three Cases against Pragmatic

Enoch’s accounts of intrinsic and instrumental indispensability are intended
to be capacity-relative (note Enoch’s judicious use of “to us” language at
2011b: 70–1). This means that, on Enoch’s account, the sources of epis-
temic justification that are basic for us may not be basic for a creature
with quite different capacities. For example, sense perception would not
be instrumentally indispensable to a god-like being who had direct intui-
tive epistemic access to the complete nature of reality. And so, on Enoch’s
account, it would not count as a source of basic epistemic justification for
such a being. We think that this assumption of capacity-relativity is essential
to the plausibility of Enoch’s account: why should instrumental indispen-
sability for some other sort of creature determine what we have reason to
believe? However, this feature of the account also makes it vulnerable to our
first two counterexamples.
5.3.2.1 Case One: Sparky and Sally

Suppose that ingenious artificial intelligence researchers have designed
an AI (“Sparky”) capable of having full-fledged beliefs about the world.
Sparky is epistemically similar to an ordinary person (“Sally”) in the follow-
ing respects: Sparky’s capacities (and limitations) with respect to memory
and to reasoning processes that conclude in belief are identical to Sally’s.
Further, across Sally’s whole life, Sparky has been rigged up to Sally so that
Sparky receives exactly the sensory evidence that Sally does. We stipulate
just one crucial difference between Sally and Sparky. Perhaps Sparky does
some things that count as actions, but unlike Sally, Sparky is simply incapa-
ble of deliberating about what to do.18
Plausibly, at any given time in Sally’s adult life, Sally and Sparky share
(very nearly) the same total evidence. After all, they have been exposed to
very nearly the same sensory evidence. Their memories are qualitatively
nearly identical. Their senses are similarly connected to the very same
18
One might question whether Sparky is genuinely possible. For example, on a stand-
ard functionalist account of psychology, a belief is a state that, inter alia, interacts with
desires in certain ways. Functionalism would thus take a dim view of Sparky imagined
as a “pure thinker” with beliefs but no desires. However, we insist only that Sparky lack
the capacity for practical deliberation (as Enoch conceives of that capacity; 2011b: 70–3).
This is compatible with Sparky possessing desires, because functionalists are paradigmati-
cally happy to ascribe beliefs and desires to animals that lack sophisticated deliberative
capacities of the sort Enoch’s argument appeals to. This point, combined with the intui-
tive conceivability of a creature like Sparky, constitutes a strong case for Sparky’s genuine
possibility.
environment. And they have identical capacities and dispositions to form

beliefs on the basis of these inputs.
Suppose further that, in light of their parallel capacities and circum-
stances, Sally and Sparky engage in ethical reasoning (that is: reasoning about
which ethical claims are true) in exactly the same ways at exactly the same
times: when Sally trusts some testimony, so does Sparky; when Sally finds a
thought-experiment or principle intuitively compelling, so does Sparky, etc.
So Sally and Sparky in fact accept all the same ethical propositions, on the
same bases. For Sparky, of course, this reasoning has purely theoretical sig-
nificance. In contrast, Sally’s ethical reasoning is often a part of her practical
deliberation: she often acts on the basis of her ethical conclusions.
This description makes it highly plausible that at any given time, Sally
and Sparky are almost always justified to the same extent in believing the
very same propositions.19 There is one plausible exception: Sally will have
plenty of introspective and memory evidence as of deliberating, which
Sparky will lack, so she (unlike Sparky) will have many beliefs that are justi-
fied partly on this basis. However, it is hard to imagine this affecting how
justified each of them is in accepting an ethical or metaethical claim. On
Enoch’s account, however, Sally has a rationally required project (practical
deliberation) that Sparky lacks. So, Enoch’s account suggests that Sally has
some defeasible epistemic justification for believing that there are ethical
facts that Sparky lacks. This, we submit, is very odd.
The force of the oddity can be illustrated by comparing our case to
the more familiar “new evil demon” case introduced by Keith Lehrer and
Stewart Cohen (1983). This case compares two agents who are “perspecti-
val” duplicates: they have identical beliefs, apparent memories, and con-
front identical perceptual appearances. The difference between these agents
is that the perceptual appearances which in the one agent arise from reli-
able sense perception are in the other the result of demonic illusion. There
is strong intuitive pull to think that agents like these are also intuitively
justificational duplicates, and that this constitutes a serious problem for
reliabilist accounts of justification. While the case is powerful, justification
“externalists” can potentially bite the bullet here, in part by appealing to
their central arguments that causal or modal connection to the world are
justification-conferring. Our case is more dialectically powerful than the
new evil demon case, because it holds parallel all of the features—both
“internal” and “external”— that contemporary epistemologists typically
find relevant to epistemic justification. Indeed, the case is intended to
19
Or, if epistemic permissivism is true, the same range of attitudes is permissible for
each of them to take towards a given proposition. On epistemic permissivism, see White
(2005).
isolate only the distinctive alleged source of epistemic justification entailed

by Enoch’s account.
However, this is exactly what makes Enoch’s account so implausible. Sally
is supposed to have justification for believing that there are ethical facts that
Sparky lacks. But the only difference between Sparky and Sally is that Sally
engages in a valuable activity that Sparky cannot engage in, and that activ-
ity would not be valuable if there were no ethical facts. It is very hard to see
how this difference could make Sally more justified than Sparky. Thus, we
claim, the case of Sparky and Sally is a counterexample to Enoch’s theory.
5.3.2.2 Case Two: Declan

Next consider Declan, an ordinary agent who has never thought about
metaethics. He has no beliefs about whether there are ethical facts, and if he
thought about it, he would simply suspend judgment on the matter.20 Enoch’s
account predicts that Declan—like us—has indispensability-grounded jus-
tification for believing that there are ethical facts. Now imagine that an evil
demon temporarily eliminates Declan’s capacity to deliberate (without his
noticing), and does nothing else. On a straightforward reading of Enoch,
Declan thereby has less justification for believing that there are ethical facts.
But it is very odd to think that such a demon can alter Declan’s epistemic
situation simply by switching on and off this capacity. By contrast, there is
nothing odd about a demon altering what Declan is justified in believing by
altering his access to uncontroversial sources of basic epistemic justification.
For example, if the demon eliminates Declan’s ability to remember that P,
this can undercut Declan’s justification for believing that P.
5.3.2.3 Case Three: Marjorie

Marjorie has strong empirical evidence that her practical deliberation is not
causally efficacious. This evidence suggests that she is in a science fiction
dystopia, where, whenever she deliberates and decides what to do, certain
diabolical scientists intervene, and prevent her intention from guiding her
action. In fact, however, Marjorie is in a different science fiction dystopia,
where the scientists interfere with her perceptual faculties, memories, and
sense of self-control, but leave the connection between her decisions and
actions untouched. Now consider the following de se belief: the results of my
practical deliberation have some chance of being causally efficacious. It may be
20
Does this imply that Declan is irrational, given Enoch’s account? No. For delibera-
tive indispensability provides only defeasible justification, and Declan could be in the
presence of relevant defeaters.
possible to deliberate absent this belief.21 However, this belief is plausibly

instrumentally indispensable to practical deliberation in Enoch’s sense: the
central reasons that one has to deliberate would surely be radically under-
mined by the inefficacy of one’s deliberation. This means that, on Enoch’s
account, Marjorie has indispensability-based basic justification for believing
that her practical deliberations are causally efficacious.
This case exploits the fact that if deliberative indispensability provides
basic justification for believing that there are ethical facts, it will also pro-
vide basic justification for believing ordinary contingent propositions. This
is embarrassing for Enoch because beliefs like the one mentioned—which
concern the distribution of contingent causal connections—seem paradig-
matically amenable only to justification on the basis of empirical evidence.22
Deliberative indispensability, however, appears to be an a priori mode of
justification. On Enoch’s account, Marjorie’s empirical evidence has to com-
pete with and outweigh her alleged deliberatively-based a priori evidence that
she can affect the world via her practical deliberations. This is implausible.23
5.3.3 Why Enoch’s Vindicating Account Fails

This puts us in a position to spell out our central objection to
Enoch: Truth-Directedness is a constraint on vindicating accounts
of the sources of basic epistemic justification, and Pragmatic violates
Truth-Directedness. This is because the fact that the belief that P is indis-
pensable to our deliberative projects bears no positive relationship to the
truth of P. Indeed, Enoch never claims that it does. Rather, he appeals to
a different normative significance for this belief: that it is indispensable to
a rationally non-optional project. But, as we have emphasized in section
5.3.1, it is not enough for a vindicating account of the sources of basic epis-
temic justification to be normatively significant; such an account must also
capture what is distinctive of epistemic justification. This, we have argued,
requires compatibility with Truth-Directedness.
21
Bratman (1987: 37–8) has offered counterexamples against the idea that ϕ-ing
intentionally requires belief that one can ϕ. For example, someone recovering from paral-
ysis might intentionally flex her hand behind her back, despite not knowing whether she
is doing so, or indeed whether she is able to do so. If one is compelled by some cases,
one should also allow that such an agent could deliberate about whether to flex her hand
behind her back.
22
We thank David Enoch for helping us to clarify the force of this case.
23
It should be noted that, according to some views, there are cases of a priori justifica-
tion of contingent claims (such as the claim I am here now). We are not convinced there
are such cases. But note that even if there are such cases, the best explanations of their
plausibility do not apply here.
The fact that Pragmatic violates Truth-Directedness explains the force

of the three counterexamples just offered. Consider them in turn. The dif-
ference between Sparky and Sally is that Sally has an indispensable project
that Sparky lacks. That may give Sally all sorts of ethical reasons. But it is
instructive to compare Sally to Hallie: the demon’s threat is unconnected
to the truth of Hallie’s belief in her golden voice, and that explains why the
demon’s threat does not epistemically justify. The instrumental indispensa-
bility of Sally’s belief in ethical facts appears unrelated to the truth of Sally’s
belief in just the same way. In the second case: turning on and off Declan’s
capacity to deliberate about what to do does not affect what he is justified
in believing, because this capacity does not add any connection (causal,
modal, perspectivally relative, or otherwise) to the facts that was lacking in
its absence. The distinctive feature of the third case is that Marjorie’s rel-
evant beliefs concern contingent facts about causal connections. Here, the
implausibility of deliberative indispensability making a positive difference
is well explained by our difficulty in understanding how there could be a
relevant positive connection between (a) the deliberative indispensability of
beliefs concerning such facts and (b) their truth. We take these highly plau-
sible theoretical explanations of the force of the intuitive cases to augment
the force of the cases themselves.
Further, in virtue of violating Truth-Directedness, Pragmatic offers infe-
rior explanations of why some candidate sources of evidence are not basic
justifiers.24 For example, why doesn’t reading tea leaves provide basic jus-
tification for beliefs about your fate? Here are two (schematic) candidate
explanations: (1) the beliefs that result from standard methods of tea-leaf
reading bear no positive relation to facts about your fate; (2) reliance on
tea-leaf reading is not instrumentally indispensable to a rationally indispen-
sable project. Both (1) and (2) seem true. (Although one can easily imagine
a benighted soul whose fragile grasp on the shreds of meaning in his life
required trust in the tea leaves). But (1) is intuitively a much better answer
to our explanatory question than (2).
For another example, consider a case of active controversy in foundation-
alist circles: whether testimony is a source of basic epistemic justification.
Pragmatic suggests that this controversy could be resolved in large part by
determining whether reliance on testimony is instrumentally indispensable
to an intrinsically indispensable project. It plausibly is: living well with oth-
ers is plausibly an intrinsically indispensable project for creatures like us,
and defeasibly trusting their testimony seems instrumentally indispensable
to that project. But this seems like the wrong sort of consideration to settle
24
We are indebted to Brad Cokelet for this point.
the outstanding debate about whether or not testimony is a source of basic

Finally, note that it may be possible to come up with piecemeal replies
to some of our counterexamples. However, we think this would be a losing
strategy for the defender of Pragmatic. With Truth-Directedness in hand,
we have a recipe for generating counterexamples to Pragmatic: construct
cases where Pragmatic endorses as epistemically basic methods that lack
plausible connection to the truth. Such counterexamples will be legion.25
One might object on Enoch’s behalf here that practical deliberation is
truth-directed. In particular, one might object that often, when we engage
in practical deliberation, we deploy our rational capacities to answer the fol-
lowing (purportedly) factual question: what ought I to do? One might then
take a page from another of Enoch’s arguments (2011b: ch. 7), and suggest
that we have reason to believe that our capacity for practical deliberation
is in fact acceptably reliable, because evolution has fitted us to track some-
thing close to the ethical facts.
While this objection is initially seductive, it ultimately misunderstands
either Enoch’s view or our case against it. Truth-directedness does not
merely require that a vindicating account point to a belief-forming method
that in fact bears some positive connection to the truth (our example of the
divine command theory of epistemic justification in section 5.3.1 satisfied
that criterion). Rather, on our theory, a legitimate vindicating account must
appeal to this positive connection to explain why something is a source of
basic epistemic justification. As we have emphasized, Pragmatic simply does
not do that. Because our argument here is directed against this explana-
tory theory, the question of whether the deliberative methods that theory
endorses also happen to be in some sense positively connected to the truth
is irrelevant.
We conclude that Pragmatic should be rejected. We have argued that it
faces serious intuitive counterexamples. Further, this vulnerability to coun-
terexample is well explained by the fact that Pragmatic violates a powerful
constraint on vindicating accounts of the sources of basic epistemic justifi-
cation, namely Truth-Directedness.
The failure of Pragmatic does not by itself show that no defensible vindi-
cating account can be given that would underwrite Indispensabilism. Nor
does it establish our more ambitious thesis: that Indispensabilism must
be rejected. However, our argument against Pragmatic does provide the
25
For another example that fits this recipe, see Cuneo’s case for the instrumental
indispensability of having a positive self-image (2012: 1064).
materials needed in order to extend our argument to both of these stronger

conclusions. We now turn to that task.
5.4 ALTERNATIVE VINDICATING ACCOUNTS

AND INDISPENSABILISM
As we saw in section 5.2, there are good reasons to seek a vindication of

the sources of basic epistemic justification. To repeat, a vindication of those
basic sources is, roughly, a non-epistemic explanation of why the sources are
what they are, which upholds (rather than debunks or reforms away) our
intuitive conception of the nature and importance of those sources. In the
previous section, we argued that Pragmatic—Enoch’s proposed vindicating
account—should be rejected. In this section, we argue that there are strong
reasons to think that our case will generalize to other attempts to vindi-
cate Indispensabilism. The core reason is that deliberative indispensability
appears unconnected to the truth. This means that those seeking to vindi-
cate Indispensabilism face a dilemma. On the one hand, you can develop an
account that aims to vindicate deliberative indispensability as a basic source
of justification. But those accounts will—like Pragmatic—be inconsistent
with Truth-Directedness, and thus, we claim, fail to provide an adequate
account of epistemic justification. On the other hand, you can aim to pro-
vide a vindicating account of the sources of basic epistemic justification that
is compatible with Truth-Directedness. Such accounts have some chance of
vindicating norms of epistemic justification. But such accounts will not vin-
dicate deliberative indispensability as a basic source. If sound, the dilemma
provides good reason to think that Indispensabilism is false.
To begin, consider an example of the second horn of the dilemma.
Alvin Goldman (2008) has recently suggested a view that treats reliabilism
as (in our terms) a vindicating account of the sources of basic epistemic
justification. For example, this account suggests that sense perception is
epistemically basic because, in a relevant range of cases, the processes that
link the world to perceptual states and thence to correlated belief are on
the whole marvelously reliable. Such a reliabilist vindicating account of
the foundations of epistemic justification is evidently compatible with
Truth-Directedness. Indeed, reliabilism is one natural framework for spell-
ing out a substantive justification–truth link. But this vindicating account
is also hostile to Indispensabilism.26 This example will generalize: similar
26
One could of course here try to present an ambitious argument that deliberative
indispensability is a reliable belief-forming method. We think that the prospects for such
an argument are poor, and thus leave this possibility to the side for now, in order to
streamline our main discussion.
considerations apply to possible vindicating accounts that emphasize alter-

native modal properties such as sensitivity or safety, rather than process
reliability (cf. Nozick 1981; Sosa 1999; Roush 2007).
The force of the tension may appear less clear as we move away from
modal approaches to satisfying Truth-Directedness. However, consider
one of the most influential alternatives to such conceptions, which we will
call “responsibilism” about epistemic norms.27 On this way of thinking
about epistemology, norms for epistemic justification are, or are grounded
in, (some of the) norms that govern what is required to be a responsible
epistemic agent (cf. Chisholm 1977; BonJour 1985; Greco 1990). On one
variant, for a belief to be justified is just for it to have been formed by
an epistemically responsible process. For example, on this sort of account, if
I believe that P in virtue of carefully acquiring and assessing the evidence,
which I reasonably find to strongly support P, I would be justified in believ-
ing P, even if this process is in fact unreliable.28 Conversely, if I form my
belief that P irresponsibly—e.g., on a whimsical hunch—I would not be
justified, even if my whimsical hunches are in fact very reliable.29
The central question for this sort of approach is how to characterize an
epistemically responsible process. It is no accident that many responsibilists in
epistemology (including those cited above) advert to truth in explaining what
epistemic responsibility amounts to. Think again about Hallie: her continu-
ing to believe that she sounds exactly like Steve Perry when she sings “Don’t
Stop Believin’ ” would be highly ethically responsible, but a clear violation of
her “epistemic duties,” if such things exist. And the lack of (first-personally
mediated) connection to the truth appears to be an excellent explanation
of the latter fact. With this contrast in hand, it becomes very unclear how
Indispensabilism could be vindicated as compatible with our fundamental
epistemic duties.
It is worth emphasizing that not all responsibilists do advert to a connec-
tion to truth in explaining epistemic responsibility. For example, on one
natural reading of Enoch, he is a responsibilist. On this reading, Pragmatic
reflects an underlying conception of epistemic responsibility as requiring
apt pursuit of our intrinsically indispensable projects. On this reading, our
core argument in this chapter can be seen as illuminating the difficulties
27
Another important group of approaches seeks to understand central epistemic cat-
egories like justification in terms of epistemically virtuous and vicious character traits
(e.g., Montmarquet (1993); Zagzebski (1996); Sosa (2007); Greco (2010)). On these
approaches, epistemic virtue tends itself to be understood either in modal or responsibil-
ist terms (or both), leading to a very similar dialectic as in the text. Because of this, we do
not discuss these approaches further here.
28
Compare again the “new evil demon”-style cases.
29
Compare BonJour’s classic case of Norman the clairvoyant (1985: 41ff.).
facing attempts to characterize epistemic responsibility in a way that is

inconsistent with Truth-Directedness.
These brief remarks demonstrate the challenges facing attempts to defend
Indispensabilism by providing a vindication of the sources of basic epis-
temic justification. The pattern suggested by these cases makes it plausible
that any vindicating account for the sources of basic epistemic justification
will either fail to respect the Truth-Directedness constraint (and thus, we
claim, fail to secure a link to genuine epistemic justification), or fail to sup-
port Indispensabilism.
Enoch’s thesis that our intuitive commitments in practical deliberation
are a basic source of evidence is highly controversial among foundational-
ists (contrast sources like sense perception, memory, and use of IBE). So if
this source is going to get on to the list of basic sources, it needs to do so
on theoretical grounds (Enoch’s defense of Pragmatic is one such theoretical
attempt). However, if we are right, Truth-Directedness is one of our cen-
tral theoretical commitments concerning epistemic justification. This sug-
gests that any theoretical attempt to vindicate Indispensabilism will need
to involve quite radical reform of our understanding of epistemology. For
example, one could reject Truth-Directedness as part of a broader meta-
philosophical pragmatism. We think that our case for Truth-Directedness
tells against such views. However, the dialectic here is notoriously murky,
with very little that can be assumed as non-question-begging common
ground. We are thus satisfied to say: Indispensabilism must be rejected,
modulo meta-philosophical pragmatism or other similarly radical views.
5.5 ABANDONING INDISPENSABILISM
In this section, we briefly discuss three distinctive replies to our argument.

These replies each target a central assumption of our argument: that delib-
erative indispensability arguments must focus on practical deliberation; that
practical deliberation does not ground the ethical facts; and that indispensa-
bility arguments must deliver epistemic normativity. Exploring these replies
helpfully illuminates the range of assumptions within which our argument
operates. These replies are also instructive because each reply abandons
Indispensabilism in some way, in order to defend a related thesis.
First, one might hope that the metaethical payoff of Enoch’s project could
be preserved, and our objection avoided, by the following variant on his argu-
ment. Begin by claiming that indispensability to the epistemic (or doxastic,
or explanatory) project confers basic epistemic justification. Assume that
the epistemic project is consistent with Truth-Directedness. (Whether it is
depends on how one characterizes this project.) Then adapt Enoch’s case for
the deliberative indispensability of belief in Robustly Real ethical facts, into

a case for the indispensability of belief in Robustly Real epistemic facts for
doxastic deliberation. This would permit a Truth-Directedness-respecting
case for Robust Realism about the epistemic facts. The Robust Realist about
ethical facts can now appeal to the epistemic as a companion in guilt: once
we posit some normative facts (conceived of along the lines of Robust
Realism), it might seem a small cost to take on commitment to additional
normative facts, such as the ethical facts.30
Note that this argument abandons Indispensabilism. It grants
Truth-Directedness, and seeks to find a variant of Enoch’s argument that
is compatible with the central points that we have made in this chapter.
This argument is thus substantially less theoretically exciting that Enoch’s.
However, if it succeeded in defending Robust Realism about ethical facts, it
would deserve substantial independent attention.
We doubt it succeeds, however, for at least two reasons. First, note that
deliberation about what to believe is, famously, transparent: we focus on
whether P, not whether we ought to believe that P. In light of this, it is not
clear that belief in epistemic facts is indispensable to such deliberation, as
the adapted argument requires. If the oracle tells you there are no epistemic
facts, the project of settling what to believe still evidently has a central and
valuable point. Second, if the companion in guilt is established, it might
assuage worries about belief in Robustly Real ethical facts that stem from
the alleged metaphysical peculiarity of those facts. However, it is not clear
that it helps with the central challenge of this chapter. To see this, consider
that the proposal says nothing about whether the justification of belief in
ethical facts is consistent with Truth-Directedness. Here the proponent of
this strategy faces a dilemma. If she claims that the justification of belief in
ethical facts is inconsistent with Truth-Directedness, then our main argu-
ment in this chapter targets her position in the same way that it targets
Enoch’s. And if she claims that the justification of belief in ethical facts is
consistent with Truth-Directedness, then her view is radically different from
Enoch’s, such that the central challenges that Enoch’s account were meant
to address will need to be addressed entirely anew.
A second important strategy for replying to our argument appeals to an
ambitious form of constructivism about ethical facts. Suppose, for example,
that ethical facts are determined by a certain privileged (and indispensable)
process of practical deliberation. This could be used to deliver an account of
our epistemic justification for belief in ethical facts that is entirely consist-
ent with Truth-Directedness. The basic idea here is that the very process of
30
See Cuneo (2007) for an extended case that many of the core challenges posed to
positing ethical facts carry over to positing epistemic ones.
practical deliberation that commits us to the existence of ethical facts also

functions to ground the existence of those facts. If this were the explanation
for why this commitment constituted a source of basic epistemic justifica-
tion, it satisfies Truth-Directedness.
This is a familiar attempt to make good on the Tempting Idea we men-
tioned in section 5.1. The challenges facing this view are equally familiar.
Enoch himself has forcefully posed some of these challenges in his (2006 and
2011a). We cannot hope to assess this strategy here, but we do want to empha-
size that this kind of constructivism about ethical facts does not underwrite
Indispensabilism. This kind of constructivism supports a domain-specific
claim about the metaphysics of ethics, rather than a global epistemic the-
sis about the significance of deliberative indispensability (which is what
Indispensabilism is). Perhaps this suggested constructivist view—which is
only about ethical facts in particular—is all that we should have wanted from
the Tempting Idea to start with. If so, this brings out the poor fit between
Indispensabilism and the intuitive idea it was supposed to underwrite.31
Finally, a third reply to our argument is to give up on the idea that
deliberative indispensability gives us distinctively epistemic justification.
Developed in this way, the Tempting Idea would be wholly compatible
with the naturalistic criterion that we initially posed as its foil. For, on that
criterion, we are epistemically justified in believing in ethical facts just in
case they (or their reduction base) feature in our best explanations of scien-
tifically respectable phenomena. Despite this compatibility, the third reply
also comes at a certain cost. It would be interesting, for example, if we were
ethically committed to the existence of Robustly Real ethical facts. But if
this provides us with no evidence for their existence, the upshot could be a
deep and potentially intractable conflict between the demands of practical
and epistemic reason. This is an interesting possibility, but not a particularly
attractive one.
Summing up, we think that each response that we have considered here
faces significant challenges, and we are not hopeful about any of them. Thus,
31
Notice that some philosophers drawn to the Tempting Idea have appealed to a
similar idea with respect to a range of types of facts, such as facts about the nature of
action, the self, or desire (e.g. Korsgaard (2009); Schapiro (2009)). Suppose that certain
beliefs about some such facts were practically indispensable. Indispensabilism would then
underwrite an inference from that indispensability to epistemic justification. Attempting
to replace Indispensabilism with a “constructivist” rationale, however, requires finding
an independent metaphysical argument that shows that the relevant facts are grounded
in facts about some practical project (e.g., the project of practical deliberation). Such
grounding claims will not be equally plausible for all types of facts. The crucial point
about this alternative to Indispensabilism is that the defensibility of indispensability
arguments for epistemically justified belief in a class of facts will stand or fall with the
plausibility of a “constructivist” metaphysics for that class of facts.
although our main argument in this chapter is directed at Indispensabilism,

we take our argument to have another upshot as well: it helps clarify the
range of ways that one might seek to make good on the Tempting Idea that
we started with, and underscores the significant philosophical burdens that
one takes on in trying to do so.
5.6 CONCLUSION
This chapter has evaluated the prospects of the following thesis:

Indispensabilism If a belief-forming method is indispensable to the
project of practical deliberation, then that method is a source of basic
One striking reason to be drawn to this thesis is that our commitment to
the existence of ethical facts can seem to arise from their apparent delibera-
tive indispensability, rather than from any role that they play in explaining
scientifically respectable phenomena. A modest methodological conserva-
tism enjoins us to try to uphold this etiology as justification-conferring.
This is a scrupulously reasonable project. However, we have argued that it
likely cannot be successfully accomplished.
The heart of our argument has been that in order to vindicate the epistemic
credentials of Indispensabilism, one would need to show that it is compat-
ible with what is distinctive of epistemic justification. We explored Enoch’s
sophisticated recent attempt to vindicate Indispensabilism, and argued that
it falls well short. Our argument rested heavily on a characterization of
(part of what) makes epistemic justification distinctive. According to this
thesis, which we dubbed Truth-Directedness, the explanation of what makes
something a source of basic epistemic justification must advert to a con-
nection between that source and the truth of the proposition justified. This
thesis attempts to bring out a fundamental difference between the norms of
epistemic justification and those of practical justification, suggested at the
start of the chapter by our example of Hallie and the demon. We argued
that Enoch’s defense of Indispensabilism violated Truth-Directedness. We
then argued that Enoch’s failure to successfully defend Indispensabilism is
no accident: we have reason to be confident that similar challenges will
plague other attempts to defend Indispensabilism. Finally, we discussed
ways of attempting to appeal to deliberative indispensability in metaethics
that abandon Indispensabilism.
Both Indispensabilism and Truth-Directedness share a substantive foun-
dationalist assumption about epistemic structure. This is a significant limi-
tation of our argument as developed here, because one can easily imagine
theses similar to Indispensabilism that eschew the foundationalist assump-

tion. We conjecture that variants of Truth-Directedness could be used to
raise very similar problems for these theses. However, we do not have the
space to defend this claim here.
We would like to close by briefly underscoring some of the broader
metaethical import our argument. First, consider the significance of this
conclusion for the assessment of Robust Realism in metaethics. Enoch’s
deliberative indispensability argument is crucial to his case for Robust
Realism. So, if this argument fails, one of the most interesting and innova-
tive defenses of Robust Realism simply stops well short of the finish line.
Beyond its central role in his positive case, the argument from deliberative
indispensability plays two other major roles in Enoch’s defense of Robust
Realism. First, he uses the argument to defend Robust Realism against a
common naturalist objection to non-naturalist metaethical views: given
that (according to the relevant kind of non-naturalist) these ethical facts (or
their reduction base) are not part of our best explanation of any scientifi-
cally respectable phenomena, we lack epistemic justification for believing in
them.32 Second, many philosophers have thought that non-naturalists will
be forced to posit a special perceptual faculty for detecting ethical facts. The
deliberative indispensability argument is supposed to allow Enoch to escape
the need to do so. Without his argument from deliberative indispensabil-
ity to turn to, Enoch’s innovative responses to these challenges to Robust
Realism are undercut. This leaves the Robust Realist about ethical facts still
searching for solutions to these challenges.
Our assessment of Indispensabilism has metaethical signifi-
cance that extends beyond the evaluation of Robust Realism. Many
philosophers—including philosophers who advocate metaethical positions
quite different from Robust Realism—have been attracted to some ver-
sion of the Tempting Idea that we introduced at the start of this chapter.
To repeat, this was the idea that the (putative) indispensability of belief in
ethical facts for our practical projects—including, for example, the project
of deliberating about what to do—can somehow justify our belief in such
facts. If our argument in this chapter is right, then one of the clearest ways
of making good on this idea, namely, Indispensabilism, should be rejected.
Moreover, as we discussed in section 5.5, there are serious worries with
some of the other salient ways that one might seek to make good on the
32
It is important to note that Enoch’s strategy is intended as an alternative to two
common replies to this objection. The first is to insist that ethical facts in fact do play a
role in the explanation of scientifically respectable phenomena. The second is that ethi-
cal facts themselves count as respectable explananda (even if not necessarily scientifically
respectable). Enoch (2011b: 53) expresses skepticism about both of these replies.
Tempting Idea, e.g., by giving up on the idea that the sort of justification
involved is genuinely epistemic justification, or by pursuing an ambitious
constructivism about ethical facts.
To sum up, then, we have argued for three conclusions concerning the
epistemic and metaethical significance of practical indispensability. First, we
should reject Enoch’s indispensabilist case for Robust Realism about ethi-
cal facts. Second, we should reject Indispensabilism in general: deliberative
indispensability does not epistemically justify belief. Third, there are good
reasons to be skeptical of the metaethical significance of deliberative indis-
pensability, as well as practical indispensability more broadly.33
References
Alston, W. 2005. Beyond Justification: Dimensions of Epistemic Evaluation.
Ithaca: Cornell University Press.
Berker, S. 2013. “Epistemic Teleology and the Separateness of Propositions,”
Philosophical Review 122: 337–93.
Björnsson, G. and Olinder, R. F. Forthcoming. “Enoch’s Defense of Robust
Meta-Ethical Realism,” Journal of Moral Philosophy.
BonJour, L. 1985. The Structure of Empirical Knowledge. Cambridge, MA: Harvard
University Press.
Boyd, R. 1997. “How to be a Moral Realist,” in S. Darwall, A. Gibbard, and P.
Railton (eds.), Moral Discourse and Practice. New York: Oxford University Press,
105–36.
Bratman, M. 1987. Intention, Plans, and Practical Reason. Cambridge, MA: Harvard
University Press.
Broome, J. 1999. “Normative Requirements,” Ratio 12: 398–419.
Chisholm, R. M. 1977. Theory of Knowledge: Englewood Cliffs, NJ: Prentice-Hall.
Cuneo, T. 2007. The Normative Web: An Argument for Moral Realism. Oxford: Oxford
University Press.
Cuneo, T. 2012. “Taking Morality Seriously: A Defense of Robust Realism, by David
Enoch,” Mind 121: 1059–64.
33
Thanks to David Enoch and Joshua Schechter for extensive and invaluable discus-
sion. We are also grateful for comments from Selim Berker, David Braddon-Mitchell,
Sarah Buss, Matthew Chrisman, Brad Cokelet, Terence Cuneo, Stephen Darwall, Billy
Dunaway, Kenny Easwaran, Allan Gibbard, Nadeem Hussain, Matt Kotzen, John Ku,
Dustin Locke, Kate Manne, Howard Nye, Peter Railton, Sharon Street, Mike Titelbaum,
Silvan Wittwer, two anonymous referees for Oxford Studies in Metaethics, participants in
Sarah McGrath’s Spring 2014 Metaethics graduate seminar at Princeton, participants at
the 2013 Wisconsin Metaethics Workshop, and participants at David Plunkett’s pres-
entation at the Author Meets Critics session for Taking Morality Seriously at the 2012
Eastern APA.
Douven, I. 2011. “Abduction,” in Stanford Encyclopedia of Philosophy, ed. E. N. Zalta.

<http://plato.stanford.edu/archives/spr2011/entries/abduction>.
Dworkin, R. 2011. Justice for Hedgehogs. Cambridge, MA: Harvard University Press.
Enoch, D. 2006. “Agency, Shmagency: Why Normativity Won’t Come from What
Is Constitutive of Action,” Philosophical Review 115: 169–98.
Enoch, D. 2007. “An Outline of an Argument for Robust Metanormative Realism,”
in R. Shafer-Landau (ed.), Oxford Studies in Metaethics, Volume 2. Oxford: Oxford
Enoch, D. 2011a. “Can There Be a Global, Interesting, Coherent Constructivism
about Practical Reason?” Philosophical Explorations 12: 319–39.
Enoch, D. 2011b. Taking Morality Seriously: A Defense of Robust Realism.
Enoch, D. and Schechter, J. 2008. “How Are Basic Belief-Forming Methods
Justified?” Philosophy and Phenomenological Research 76: 547–79.
Fantl, J. and McGrath, M. 2009. Knowledge in an Uncertain World. Oxford: Oxford
University Press.
Gibbard, A. 2008. “Rational Credence and the Value of Truth,” in T. Gendler and
J. Hawthorne (eds.), Oxford Studies in Epistemology, Volume 2. Oxford: Oxford
Goldman, A. 2008. “Immediate Justification and Process Reliabilism,” in Q. Smith
(ed.), Epistemology: New Essays. Oxford: Oxford University Press, 63–82.
Greco, J. 1990. “Internalism and Epistemically Responsible Belief,” Synthese
85: 245–77.
Greco, J. 2010. Achieving Knowledge: A Virtue-Theoretic Account of Epistemic
Normativity. Cambridge: Cambridge University Press.
Harman, G. 1977. The Nature of Morality. Oxford: Oxford University Press.
Hawthorne, J. 2004. Knowledge and Lotteries. Oxford: Oxford University Press.
Husi, S. 2013. “Why Reasons Skepticism is Not Self-Defeating,” European Journal
of Philosophy 21: 424–49.
Jackson, F. 1998. From Metaphysics to Ethics. Oxford: Clarendon Press.
Kolodny, N. 2005. “Why Be Rational?” Mind 114: 509–63.
Korsgaard, C. M. 1996. The Sources of Normativity. New York: Cambridge University
Press.
Korsgaard, C. M. 2009. Self-Constitution: Agency, Identity, and Integrity.
Lehrer, K. and Cohen, S. 1983. “Justification, Truth, and Coherence,” Synthese
55: 191–207.
Lenman, J. 2014. “Deliberation, Schmeliberation: Enoch’s Indispensability
Argument,” Philosophical Studies 168: 835–42.
Lipton, P. 2004. Inference to the Best Explanation, 2nd edn. London: Routledge.
McPherson, T. 2011. “Against Quietist Normative Realism,” Philosophical Studies
154: 223–40.
Montmarquet, J. 1993. Epistemic Virtue and Doxastic Responsibility. Lanham,
MD: Rowman & Littlefield.
Moser, P. 1985. Empirical Justification. London: Springer.
Nozick, R. 1981. Philosophical Explanations. Cambridge, MA: Harvard

University Press.
Roush, S. 2007. Tracking Truth: Knowledge, Evidence, and Science. Oxford: Oxford
University Press.
Scanlon, T. M. 2014. Being Realistic about Reasons. Oxford: Oxford University Press.
Schapiro, T. 2009. “The Nature of Inclination,” Ethics 119: 229–56.
Schechter, J. and Enoch, D. 2006. “Meaning and Justification: The Case of Modus
Ponens,” Noûs 40: 687–715.
Schroeder, M. 2012. “Stakes, Withholding, and Pragmatic Encroachment on
Knowledge,” Philosophical Studies 160: 265–85.
Shah, N. 2003. “How Truth Governs Belief,” Philosophical Review 112: 447–82.
Sosa, E. 1999. “How to Defeat Opposition to Moore,” Philosophical Perspectives
13: 137–49.
Sosa, E. 2007. A Virtue Epistemology: Apt Belief and Reflective Knowledge, Volume I.
Stanley, J. 2005. Knowledge and Practical Interests. Oxford: Clarendon Press.
Stich, S. 1990. The Fragmentation of Reason. Cambridge, MA: MIT Press.
Trogdon, K. 2013. “An Introduction to Grounding,” in M. Hoeltje, B. Schnieder,
and A. Steinberg (eds.), Varieties of Dependence. Munich: Philosophia Verlag,
97–122.
Velleman, J. D. 2000. “On the Aim of Belief,” in his The Possibility of Practical
Reason. Oxford: Oxford University Press, 244–82.
Wedgwood, R. 2002. “The Aim of Belief,” Philosophical Perspectives 16: 267–97.
White, R. 2005. “Epistemic Permissiveness,” Philosophical Perspectives 19: 445–59.
Wright, C. 2004. “Warrant for Nothing (and Foundations for Free)?” Proceedings of
the Aristotelian Society (Supplemental Volume) 78: 167–212.
Zagzebski, L. 1996. Virtues of the Mind: An Inquiry Into the Nature of Virtue and
the Ethical Foundations of Knowledge. Cambridge: Cambridge University Press.
6
Rationality and Moral Authority
David Copp
It is often taken to be a platitude that morality is authoritative or norma-

tive, but it is not entirely clear what this comes to. The terms “normative”
and “authoritative” are, after all, terms of art. This essay is about a view that
seeks to explain the normativity of morality on the basis of a putative link
of an important kind between morality and rationality. The view I have
in mind takes the normativity of morality to depend metaphysically on the
existence of a link of an important kind between morality and rationality.
This dependency thesis is the topic of my chapter. I call it the Rationality
Doctrine or RD.
The intuitive plausibility of the RD is difficult to deny. If a fully rational
person could ignore the requirements of morality, even if she were aware of
what morality demands, then, it might seem, morality lacks authority over
rational persons. And if it lacks authority over rational persons, it surely
is not normative in any interesting sense. Morality in this case would be
undermined in a fundamental way. The details matter, of course, and there
is disagreement about the precise nature of the link between morality and
rationality that is at issue. But the fundamental idea is that the authority of
morality depends on the existence of a link of an appropriate kind between
morality and rationality.
The RD is not universally accepted of course. It may be, for example,
that non-cognitivists would reject it,1 and many realists would also reject
it.2 I reject it. Yet the RD has an historical pedigree. Versions of it have
been endorsed by philosophers who otherwise disagree fundamentally. In
the Groundwork, Kant aimed to show that morality is “something real,” that
it is not “a vain delusion and a chimerical concept,” or “a mere phantom
1
They might think that, as I interpret the RD in what follows, it rests on a false pre-
supposition. The issue raises subtleties that I want to set aside.
2
Constructivists of certain kinds endorse it.
Rationality and Moral Authority 135
of human imagination.”3 To show that morality is “something real,” he

thought he needed to show that the fundamental principle of morality “is
an imperative, i.e., that the will of every rational being is necessarily bound
[by it]” (Ak 440, 445). He held, it seems, that morality has normative sig-
nificance only if it is binding on all rational agents such that compliance
with morality is a requirement of reason. Working in the Kantian tradition,
Christine Korsgaard takes up the view that moral principles are principles
of practical reason (2009: 47). But it is not only Kantians who endorse the
RD. David Gauthier uses a version of it in motivating his neo-Hobbesian
theory. Gauthier asks, “What theory of morals … can ever serve any useful
purpose unless it can show that all the duties it recommends are also truly
endorsed in each individual’s reason?” He goes on to say that if “a moral
appeal” would ever “contradict reason,” then “the moral enterprise, as tradi-
tionally conceived, is impossible” (1986: 1–2). The RD has a skeptical edge.
In arguing for his error theory, J. L. Mackie seems to assume that whether
moral requirements would be authoritative depends on whether there is a
requirement of practical reason to act as one is morally required to act. He
contends that there is no such requirement of practical reason (see Smith
1994: 65; Mackie 1977: 29–30).
People have complained that my own account of morality is unsuccessful
because it fails to answer the “Why be Moral?” question by establishing the
right kind of link between morality and rationality.4 If my account were
true, the objection runs, we would not necessarily be rationally required to
act as we are morally required to act, and so morality would not be authori-
tative or normative. This objection relies on a version of the RD.
In section 6.1, I introduce the issue, formulate the RD, and briefly
consider and criticize arguments in its favor. In section 6.2, I discuss the
concepts of normativity and rationality, concepts that are central to the
formulation of the RD. I consider and reject understandings of these con-
cepts on which the RD is trivially or obviously true. I want to understand
the concepts in such a way that the RD is an interesting substantive thesis.
The RD proposes that morality’s being normative depends on the truth of a
Linkage Thesis that postulates a suitable link between morality and rational-
ity. There are different views about how such a thesis should be formulated.
In section 6.3, I narrow attention to one formulation, which I call the Basic
Linkage Thesis or BLT. In section 6.4, I consider some strategies for support-
ing the BLT and present two objections. I think there is reason to doubt the
truth of the BLT, which means that, if we hold that morality is normative,
3
Kant (1981), Ak 445, 402, 407, respectively, in the pagination of the standard
Prussian Academy edition, as given in the Ellington translation.
4
For my answer to the “Why be Moral?” question, see Copp (2010b).
136 David Copp
we need to question the RD. In section 6.5, I return to the RD and pre-
sent three arguments intended to show at least that one cannot reasonably
invoke the RD in order to adjudicate among moral theories. The RD can-
not play a role in narrowing the field of competing theories. Section 6.6 is a
brief concluding discussion. I do think that morality is normative, but I am
inclined to reject the RD.
6.1 THE RATIONALITY DOCTRINE: FORMULATION

AND ARGUMENTS
The RD holds that the normativity of morality depends on there being a

link of an appropriate kind between morality and rationality. The kind of
link in question, and the kind of dependence that is putatively involved,
can be described in different ways (see Rosen 2010). We might say that,
according to the RD, if morality is normative, it is normative in virtue of a
relation between morality and rationality. Or we might say that, according
to the RD, the normativity of morality must be grounded in the normativity
of rationality, such that its authority is explained by the existence of a rela-
tion between morality and rationality. Or we might say that, according to
the RD, the normativity of morality would consist in there being a suitable
relation between morality and rationality. The RD postulates a metaphysical
dependence of the normativity of morality, if indeed morality is normative,
on the normativity of rationality. I understand it as follows:
Rationality Doctrine (RD): Morality is normative in the philosophi-
cally most important sense only if it has a normative authority that is
grounded in the existence of a suitable relation between morality and
rationality, as postulated by the Linkage Thesis.
On this understanding, the RD proposes a metaphysical grounding thesis.
This is not the place to investigate the idea of metaphysical grounding.
Yet let me make two points about the idea. First, to postulate a grounding
relation between morality and rationality is not to propose merely a neces-
sary condition for morality’s being normative. It is plausible that morality
is normative only if those who are subject to moral requirements also meet
the minimal conditions of rational competence. This would be agreed on
all sides, even by those who reject the RD. It is not a plausible candidate for
what grounds the normativity of morality. Second, to postulate a grounding
relation between morality and rationality is not to postulate merely neces-
sary and sufficient conditions for morality’s being normative. The ground-
ing relation is asymmetrical. If the facts of neurophysiology ground the
psychological facts, neurophysiology has a metaphysical, explanatory prior-

ity. The RD accords metaphysical, explanatory priority to rationality.
The RD could be interpreted in a completely different, non-metaphysical
way. One might interpret it as a methodological or dialectical claim about
what would be needed to vindicate morality in the face of challenges such as
the why-be-moral challenge. The RD might seem plausible on this under-
standing. For if the normativity of morality is challenged, it might seem
that the only viable response would be to show that it can be derived in
some way from the normativity of something the normativity of which
is not in question. And rational requirements may be the only ones with
authority that is not in doubt. On the methodological reading, then, the
RD claims that vindicating the authority of morality in the face of skeptical
doubts depends on whether it can be shown that morality is appropriately
linked to practical rationality. I have no need to deny this. In what follows,
I shall be interpreting the RD as proposing a metaphysical grounding thesis.
To be clear, the RD claims that the existence of a grounding relation
between morality and rationality is a necessary condition of morality’s being
normative and that the normativity of morality depends on the existence of
such a relation. To show that this is so, one would need to show that there
is no viable alternative account. Yet there certainly are alternative proposals.
Primitivism holds that the normativity of morality is metaphysically
ungrounded or primitive. It holds that there is no philosophically interest-
ing explanation of the normativity of morality. Alternative grounding views
are non-rationalistic. They hold that although the normativity of morality
has a metaphysical grounding, its grounding does not consist in a rela-
tion between morality and rationality. Some alternative grounding views
are reductionist. Reductionist views claim, roughly, that the normativity of
morality can be explained or grounded without referring to or quantifying
over anything normative. Normative naturalists contend specifically that
we can explain what it is for a moral fact or property to be normative with-
out referring to, or quantifying over, any “non-natural” facts or properties
(see Copp 2012). My “pluralist-teleology” is in this sense a reductionist
view (see Copp 2009). Some alternative grounding views are not reduc-
tive, however. Such views claim that the normativity of morality can be
explained or grounded, but the accounts they provide refer to or quantify
over something that is taken to be normative. Views of this kind purport to
reduce the normativity of morality to the normativity of something else, so
they are partially reductive, as I will say. The RD is partially reductive since it
purports to reduce the normativity of morality to the normativity of ration-
ality. A non-rationalistic partially reductive view might contend that there is
a kind of ur-normativity that grounds both the normativity of morality and
the normativity of rationality. So there are both reductionist and partially
138 David Copp
reductive alternative grounding views and there is also primitivism. These

views agree that the normativity of morality is not grounded in the norma-
tivity of practical reason.
I began the chapter with (what might have seemed to be) an argu-
ment for the RD. The idea was that whether morality is authoritative in
any philosophically interesting way depends on whether its authority can
be grounded in an authority it has over rational persons. The key premise
was that whether morality has authority over rational persons depends on
whether a person could ignore the requirements of morality while remain-
ing fully rational. The argument is question-begging, however, for anyone
who doubts the RD would doubt the key premise that whether morality has
authority over rational persons depends on whether a person could ignore
moral requirements while remaining fully rational. For one might think
that a person who ignored moral requirements would be failing in a norma-
tively significant way even if she were fully rational. One might think she
would be failing morally.
The following argument might seem better. Begin with the thought
that the existence of a rational requirement to be morally motivated could
explain or ground the normativity of morality. One might indeed think that
the normativity of morality simply consists in the fact that a person who is
subject to a moral requirement (or who judges that she is) must be moti-
vated accordingly unless she is practically irrational (see Smith 1994: 61–2;
Dreier 2015). We can perhaps capture the idea in the following way: First, if
morality is normative, then one must be failing in a normatively significant
way if one does not act morally, or if one is not at least morally motivated.
What kind of failure? The only suitable candidate is that it would be a fail-
ure of rationality. It is trivial that a failure to be morally motivated would
be a moral failing. It would be of importance, however, if a moral failing
were also a failure of rationality. To be rational is to have the fundamental
virtue of agency. So if there is not a rational requirement to act morally, or
at least to be morally motivated, then it would seem that one is not failing
in a significant way if one is not morally motivated. But if there is a rational
requirement to act morally, or at least to be morally motivated, then one is
indeed failing in a significant way if one is not at least morally motivated.
Second, then, a failure to be morally motivated is a significant failure only if
it is a failure of rationality. The conclusion is that morality’s being normative
depends on the existence of a rational requirement to act morally or at least
to be morally motivated. This is what the normativity of morality comes
to. In a nutshell, the argument is that morality’s being normative depends
on whether failing to be moral is a significant failing of some kind, which
would have to be a failing of rationality.
The main problem is the second premise, according to which a lack of

moral motivation is a significant failing only if there is a requirement of
practical rationality to be morally motivated. There may be a strong tempta-
tion to accept this premise. As Richard Joyce says (2001: 104):
[Morality] presents itself as something with ubiquitous and inescapable authority.
The philosopher turns to practical rationality as a way of understanding this author-
ity, for where else is there to turn? Practical rationality alone seems to have the kind
of immunity from legitimate questioning that is attributed to morality.
This premise is nevertheless question-begging, for the point at issue is pre-
cisely whether a failure to be morally motivated would be significant if it
were not also a failure of rationality. It is of course trivial that a failure to
be motivated by a moral requirement is a moral failure, but this does not
mean that it would be a trivial failure. It is trivial that a failure to survive an
automobile accident is a failure to live after the accident. But this does not
mean that failing to live after an accident is a trivial matter. So this triviality
point is no reason to think that a failure to be morally motivated is signifi-
cant only if it is a failure of rationality.
There is a kind of skepticism about morality according to which a failure
to be morally motivated is not in itself normatively significant. There is also
the stronger view that failures of rationality are the only significant failures.
On these views, for instance, a failure to do right by one’s children is not
itself a significant failure. An effort to answer a skeptical view of this kind
may be at the root of the argument we are considering for the RD. But we
need an argument to support this skeptical view and to think that the RD
is the best way to answer it.
6.2 NORMATIVITY AND RATIONALITY: HOW TO

AVOID TRIVIALIZING THE ISSUE
To some ways of thinking, the RD will seem obviously true given the nature
of normativity or, on another way of thinking, given the nature of rational-
ity. I will consider these views in turn. The issues here are difficult because
both the concept of normativity and the concept of rationality are disputed
technical concepts in philosophical theorizing. Indeed, it is not entirely
clear that philosophers who use the terms “normativity” and “rationality”
have the same concepts in mind. As Michael Smith has said, “the term
‘rationality’ is almost entirely a philosopher’s term of art” (1997: 91). He
might have said the same of the term “normativity.” I therefore need to
discuss both of these terms and the associated concepts. I need to do this
140 David Copp
in order to clarify the RD and in order, as well, to address the views I men-
tioned, according to which the RD is an obvious truth.
Let me begin with the concept of normativity. To some philosophers,
the RD may seem to be a trivial conceptual or analytic truth, since it may
seem trivial that any “normative” requirement is a requirement of “practical
rationality” and vice versa. On this understanding, it may seem a simple
conceptual truth that whether morality is normative depends on whether
moral requirements are normative, which depends in turn on whether
moral requirements are requirements of practical rationality. Hence, on this
understanding, the RD is trivial. Anyone who shares this understanding of
the relation between normativity and practical rationality should agree that
the RD is true.5
This is not the concept of normativity that I will be working with in this
chapter. Of course, nothing substantive turns on how we choose to use
words, so I have no objection to philosophers who use the term “normativ-
ity” in this way, such that it is trivially analytic that a “normative” require-
ment is a requirement of practical reason. I could let them have the word
“normative” and instead conduct my discussion using the term “authorita-
tive” or I could write about “bindingness.” I will not do this, however. I will
continue to use the term “normative.” But, to avoid misunderstanding, it
is important to understand that, as I am using the terms “normativity” and
“rationality,” the RD is a substantive claim rather than a trivial conceptual
or analytic truth. I believe that Kant, Korsgaard, Gauthier, and Mackie also
take the RD to be a substantive claim. The substantive issue is whether
morality’s being authoritative or binding depends on the existence of a suit-
able relation between morality and rationality.
There are other ways to understand normativity that also trivialize the RD,
although this may be more difficult to see. The dialectical situation is murky
here because different concepts of normativity might be in play. Some phi-
losophers favor a kind of “reasons-fundamentalism” that seeks to reduce nor-
mativity to reasons; a consideration is normative just in case (and because)
it is suitably related to reasons.6 Others favor an “ought-fundamentalism,”
according to which normativity reduces to claims about what ought to be
done or believed or the like. There is also a view that aims to reduce norma-
tivity to motivation or desire. Derek Parfit distinguishes four conceptions of
normativity, but since he is a reasons-fundamentalist, he contends that one
5
I here set aside the fact that, on this understanding, the normativity of morality does
not depend metaphysically on whether moral requirements are requirements of practical
rationality. The dependency is conceptual or perhaps even merely terminological.
6
The term comes from Scanlon (2014).
of the four, the conception of normativity in the “reason-implying sense,” is

the philosophically most important (2011: II, 267–9).7
Consider, then, reasons-fundamentalism and ought-fundamentalism.
A complication is that there are kinds of reason that presumably are not
normative in the most interesting and important sense, including rea-
sons of etiquette, and there are corresponding “oughts,” including facts
about what we ought to do as a matter of etiquette. To account for this,
reasons-fundamentalism needs to be understood to claim to reduce norma-
tivity to reasons that are normative, and ought-fundamentalism to reduce
normativity to “oughts” that are normative.8 This is where the trivializing
move may be made, for there is a temptation to turn to the idea of ration-
ality in order to explain the idea of a normative reason or a normative
“ought.” A reasons-fundamentalist might claim, for example, that a norma-
tive reason is a reason of a kind that any fully rational person who was aware
of it would take appropriately into account in deliberation, just in virtue of
being rational. Let me call reasons of this kind—reasons it would be irra-
tional to ignore— “practical reasons.” And an ought-fundamentalist might
claim that a normative “ought” is one that any fully rational person who
was aware of it would take appropriately into account in deliberation, just
in virtue of being rational. Call these “practical oughts.” On either of these
approaches, however, the RD has a true but trivial reading.9 For, according
to the reasons-fundamentalist, on this way of proceeding, the RD can be
understood to claim that whether morality is a source of practical reasons
depends on whether the reasons that it provides are practical reasons. And
according to the ought-fundamentalist, on the corresponding way of pro-
ceeding, the RD can be understood to claim that whether morality is a
source of practical oughts depends on whether moral requirements are or
entail practical oughts.
7
I have made my own proposal in Copp (2009).
8
We cannot avoid this move by invoking the idea of a categorical reason or a cat-
egorical ought. For on one understanding, reasons of etiquette and the requirements of
etiquette are categorical, since their status as reasons or requirements does not depend on
factors that can vary from person to person, such as what is desired or valued. For relevant
discussion, see Foot (1978: 161).
9
I am here interpreting these forms of fundamentalism as proposing analyses of the
concept of normativity in terms of the concept of a reason or the concept of an “ought,”
respectively. They could instead be viewed as substantive forms of partial reductionism.
So understood they would claim that normativity is grounded in reasons or “oughts,”
respectively. On both interpretations, they could be described as contending that rea-
sons or “oughts,” respectively, are the most fundamental normative consideration such
that all other normative considerations reduce to them. I thank Hille Paakkunainen for
this point.
142 David Copp
I am not claiming of course that reasons-fundamentalism or

ought-fundamentalism must make the trivializing move. The problem is
that they need to say something about the idea of a normative reason or a
normative “ought.” They might take the idea to be primitive and unana-
lyzable. But if they make the trivializing move in order to explain what
they have in mind, they leave the substantive issue unaddressed. There is
not a substantive issue whether morality’s being a source of practical rea-
sons depends on there being a suitable relation between moral reasons and
practical reasons or whether morality’s being a source of practical oughts
depends on there being a suitable relation between moral oughts and practi-
cal oughts. But there is a substantive issue whether morality’s being authori-
tative or binding depends on the existence of a suitable relation between
morality and rationality. This is the issue I want to address.
Having made this point, however, I am left with the problem of identi-
fying the concept of normativity that I will be using. I believe I share this
concept with advocates of the different views about normativity that I have
mentioned. And I believe that they take themselves to share the same concept
with each other, for they take themselves to disagree about the nature of nor-
mativity. For instance, reasons-fundamentalists and ought-fundamentalists
take themselves to be offering different accounts of what is fundamental
to normativity in the philosophically most important sense. We therefore
need some conception of what they disagree about. What we are looking for
here is not an analysis of the concept of normativity. One common view,
after all, is that the concept is unanalyzable. Rather, what we need is a way
to isolate, refer to, or identify the concept as an analysandum that is ready
for analysis. I can do no better, I am afraid, than to use the familiar vague
terms, “bindingness” and “normative oomph.” Normative requirements are
binding on us; they have normative oomph.
I can add that whether a requirement has normative oomph, as I under-
stand this idea, turns on whether a failure to act in accord with it would be
a normatively significant failure of a kind that is independently worrisome.
If a requirement has normative oomph, a failure to act in accord with it
would be worrisome or significant for some further reason, a reason that
goes beyond the triviality that such a failure is of course a failure to act in
accord with that very requirement. As I will say, it would be independently
worrisome. This idea also is not terribly helpful, but it will suffice for my
purposes. Nothing in the arguments that follow turns on my way of direct-
ing attention to normativity.
I now turn to the concept of rationality. To some philosophers, the RD
may seem to be obviously true, even if not trivial—in this case because
of the way they understand the concept of rationality. To see this, we can
begin by distinguishing between two concepts. There is, first, the idea of
rationality understood as a matter of agential competency. An agent is

rational in this, the competence sense, just in case, roughly, she has the capac-
ity to deliberate, to weigh reasons, and to act from reasons. The basic idea of
competence-rationality is not controversial. There is also the idea of ration-
ality in what I will call the performance sense. Agents are rational in this sense
just in case they meet relevant standards of performance in carrying out
the activities that agents who are competence-rational have the capacity to
carry out. Fully rational agents have the capacity to carry out the activities
in question and they also meet the further performance requirements. An
agent can of course be rational in the competence sense even if she occasion-
ally fails to meet the performance requirements. Fully rational agents can be
described as meeting requirements of both kinds, both the requirements of
competency and the performance requirements.
On one way of understanding the relation between these two concepts
of rationality, the RD may seem to be obviously true. For, given that
competence-rationality consists in having the capacity to carry out certain
activities, it may seem that performance-rationality must consist in per-
forming these activities well.10 And it may seem, moreover, that morality
would not be normative or important in any interesting sense if one could
do well in acting for reasons without acting morally.11 Hence, one might
think, whether morality is normative or important in any interesting sense
depends on whether a fully rational agent, who is doing well in acting for
reasons, must do what is morally required. That is, whether morality is nor-
mative or important in any interesting sense depends on whether moral
requirements are also requirements of full rationality.
The problem with this suggestion is that not all ways of failing to deliber-
ate well, or of failing to do well in acting for reasons, are ways of failing to
be rational. It would not be plausible, for instance, to view a chess player
who is very slow at reasoning her way to her next move as failing to be fully
rational. Similarly, even if she acts in a way that is impolite while preparing
for an important game, it would not be plausible to take her impoliteness
as a failure to be fully rational. She may not be reasoning and acting well
in all respects, but she may be fully rational. She may be doing well in all
relevant respects. Accordingly, we should identify performance-rationality
with meeting the standards of rational performance, not simply with acting
well in all respects. If we do this, we can see the RD as making an interesting
substantive claim. It claims that whether morality is normative depends on
whether a person who fails to act morally must thereby be failing to meet
10
An anonymous referee and Hille Paakkunainen both suggested this.
11
I thank an anonymous referee for this suggestion.
144 David Copp
the standards of rational performance—even given that she is obviously not

acting well in all respects.
It is of course highly controversial what is required by the standards of
rational performance. Rational egoists think, in general terms, that the only
rational requirement is a requirement to be efficient in serving one’s own
interests. Kantians think that one is rationally required to act in accord
with the Kantian categorical imperative. I assume that these philosophers
disagree, and that what they disagree about is what rationality requires. The
rational egoist might agree with the Kantian that there is a requirement to
act in accord with the Kantian categorical imperative but claim that this is
a requirement of morality rather than of rationality. Similarly, the rational
egoist agrees with the ethical egoist that there is a requirement to be efficient
in serving one’s own interests, but she thinks it is a rational requirement
and might deny that it is a moral requirement. To understand such disa-
greements we need to understand the concept of a requirement of rational
performance.
One might suggest that we can use the notion of a reason to analyze the
concept. On one view, for example, to be rational is to respond appropri-
ately to reasons.12 Unfortunately, however, at least in my view, there are
different kinds of reasons, including moral reasons. I think it is a conceptual
truth that a moral requirement to φ entails the existence of a moral reason
to φ. But it is a matter of some importance whether a moral reason is a
practical reason, a reason that any fully rational person who was aware of
it would take appropriately into account in deliberation, just in virtue of
being rational. The proposal to analyze rationality in terms of reasons is pre-
sumably intended to analyze rationality in terms of practical reasons. And
since a morally good person would respond in a morally appropriate way
to practical reasons, we should add that being rational is to respond in the
rationally appropriate way to practical reasons. Given these qualifications,
however, the notion of a reason is doing no work for us. In distinguishing
practical reasons from reasons of other kinds, all the work is being done by
the idea of rationality.
What we are looking for, importantly, is not an analysis of the concept
of a requirement of rational performance, but rather, a way to isolate or
identify the concept as an analysandum, which we would need before we
could go on to provide an analysis. Russ Shafer-Landau points out, help-
fully, that to call someone irrational is to charge her with an important
kind of normative failure (2003: 168). Derek Parfit suggests that, in the
“ordinary, non-technical sense,” when we call an action “rational” we are
12
Shafer-Landau (2003: 168). Compare Parfit (2011: I, 56).
expressing “a kind of praise or approval that we can also express with words
like ‘sensible’ [and] ‘reasonable’ ” (2011: I, 56).13 I will follow these sugges-
tions. I will say that an action is required as a matter of rationality if and
only if it is something that must be done in order for a person to exercise
her agential capacities in a way that is sensible or reasonable or that avoids
irrationality in an ordinary sense. This is not terribly helpful or illuminat-
ing, but it will be adequate for my purposes in this chapter. I will construe
the RD as saying roughly that whether morality is normative depends on
whether one must act morally or at least be morally motivated in order to
exercise one’s agential capacities in a way that is sensible or reasonable or
that avoids irrationality in the ordinary sense.
The plausibility of the RD depends on its being the case that require-
ments of rationality are normative. It would be much less plausible that the
normativity of morality depends on the existence of a suitable link between
morality and rationality if rational requirements were not normative.14
I shall simplify my task in this chapter by simply assuming for the sake of
argument that rational requirements are normative in the philosophically
most important sense.
This assumption is widely shared but it is controversial. On the one hand,
Joyce contends, for example, that practical rationality has a “kind of immu-
nity from legitimate questioning” (2001: 102, 104). He thinks it would be
incoherent to question the normativity of practical reason (2001: 50; but
see Copp 2010a). On the other hand, John Broome can find no grounds
for thinking that rational requirements are normative.15 It seems plausible
to me, however, that if the requirements of rationality are requirements one
must comply with in order to act sensibly or reasonably or to avoid irration-
ality in an ordinary sense, then they are normative, for it seems plausible
that a failure to act in accord with such requirements would be a significant
and worrisome failure. The important point, in any case, is that defenders of
the RD do best to claim that the requirements of rationality are normative.
13
He adds “ ‘intelligent’, and ‘smart’.” Rationality is not, however, a matter of intel-
ligence, so I will set aside these words. I thank Adam Morton for pressing this point.
14
That is, the RD is most plausible as a partially reductionist view. To be sure, it could
instead be proposed as a reductionist theory. In that case, it would claim that although
rational requirements are not normative, still, whether morality is normative depends on
whether morality stands in a grounding relation to rational requirements. I cannot see
what would motivate this proposal. What is it about rational requirements that suits
them to ground the normativity of morality if they are not themselves binding or author-
itative or normative?
15
See Broome (2007: 177–8). Broome is a reasons-fundamentalist. What he means
is that he can find no grounds for thinking that there are always practical reasons to
comply with rational requirements. I thank John Brunero for suggesting that I discuss
Broome’s views.
146 David Copp
To simplify the debate in this chapter, I simply assume that this is correct.
This assumption leaves it open what, if anything, grounds the normativity
of requirements of rationality.
It is important for my purposes that I have introduced the concepts of
normativity and of rationality in such a way that there is an interesting
substantive issue whether the RD is true. As I have explained the concepts,
there is a gap between the concept of a normative requirement and the
concept of a rational requirement. To be sure, I have assumed that rational
requirements are normative, but I have not assumed that all normative
requirements are rational requirements. Rational requirements are norma-
tive requirements with a certain “flavor.” On my proposal, rational require-
ments must be complied with in order to act sensibly or reasonably or to
avoid irrationality in an ordinary sense. But the crucial point is simply that
they have some flavor, that there is the gap I mentioned. Nothing in the
arguments that follow depends on my account of what this flavor is.
6.3 THE BASIC LINKAGE THESIS
The RD holds that morality is normative only if its authority is grounded in

the truth of a Linkage Thesis. So far I have been ignoring the issue of how
exactly the Linkage Thesis should be formulated. But there are different ver-
sions of the RD corresponding to different versions of the Linkage Thesis.
To be plausible, a candidate for the Linkage Thesis must cite a relation
between morality and rationality that plausibly grounds the normativity of
morality. I propose the following:
The Basic Linkage Thesis (BLT): Necessarily, an agent is morally required
to φ in circumstances C only if there is a requirement of rationality
that she φ or at least be motivated to φ in circumstances C.16
I shall treat the BLT as the preferred formulation of the Linkage Thesis. The
BLT answers the why-be-moral question by saying that there is a rational
requirement to be morally motivated. It implies that a failure to be morally
motivated is a failure of practical rationality, other things being equal.
The BLT does not imply that a failure to be morally motivated is neces-
sarily a rational failure. If an agent is ignorant of a moral requirement, there
may be no irrationality in failing to be motivated accordingly. Moreover,
there might be other rational requirements that override the requirement
to be morally motivated. The BLT implies only that a failure to be morally
16
Nothing in what follows depends on whether we take the putative necessity here to
be conceptual or metaphysical.
motivated is a failure of practical rationality when “other things are equal.”

It leaves it open that a failure to be morally motivated might not be irra-
tional when all things are considered.
The BLT is compatible with various more specific positions. It is compat-
ible with the view that moral requirements are requirements of rationality.
On this view, a moral requirement is a rational requirement so that a failure
to comply with it would not be sensible and reasonable or it would be irra-
tional in an ordinary sense, other things being equal. Moral requirements
belong to the set of all rational requirements. The BLT is also compatible
with the weaker view that a moral requirement entails that there is a rational
requirement to act accordingly. Finally, it is compatible with the still weaker
thesis that a moral requirement entails that there is a rational requirement
to be motivated to act accordingly.
One might think it is more plausible that a moral requirement entails the
existence of a practical reason than that it entails the existence of a rational
requirement, especially a requirement to be motivated in some way.17 One
might therefore propose a linkage thesis to the effect that moral require-
ments entail the existence of practical reasons.18 Recall, however, that a prac-
tical reason is a reason that any fully rational person who was aware of it
would take into account in deliberation. Other things being equal, it would
be a rational failure not to take into account a reason of this kind. In effect,
this means that, on the current proposal, a moral requirement entails the
existence of a reason such that there is a rational requirement to take such
reasons into account. We can therefore formulate the following thesis:
The Reasons Linkage Thesis (RLT): Necessarily, if an agent is morally
required to φ in circumstances C then there is a consideration in favor
of doing φ in C such that there is a requirement of rationality that the
agent take this consideration into account in deliberation about how
to act in C.
We can see by inspection that the RLT is more complex than the BLT. It is
stronger than the BLT in what it implies about deliberation. Furthermore,
like the BLT, it says that the existence of a moral requirement entails the
existence of a rational requirement.
For these reasons, in what follows, I shall take the BLT to be the pre-
ferred formulation of the Linkage Thesis. There are linkage theses that are
stronger than the BLT, but because they are stronger, it is better to focus on
17
I thank Andrew Alwood for this suggestion.
18
Smith defends a view of this kind (1994: 62). The BLT leaves us free to take no
position on whether a moral requirement entails the existence of a practical reason.
Broome (2007) denies that rational requirements entail the existence of practical reasons.
148 David Copp
the BLT. The BLT is strong enough to answer the why-be-moral question.
It is simpler and relevantly weaker than the RLT. Nevertheless, nothing in
the arguments to follow depends on my choice to use the BLT instead of
the RLT. The arguments would go through (mutatis mutandis) either way.
When the BLT is plugged into the RD, the RD reflects a rationalistic
conception of normativity that I have elsewhere called “authoritative nor-
mativity” (Copp 2007a). It seeks to explain the authority of morality as
grounded in and dependent on the normativity of rationality. It says more
specifically that the normativity of morality is grounded in the link between
morality and rationality that is postulated by the BLT.
6.4 FOR AND AGAINST THE BLT
The central issue in this chapter is of course whether the RD is true and
whether it can reasonably be used to narrow the field of contending moral
theories. But there is reason to pause here to consider the defensibility of the
BLT. As I have formulated it, the RD says that whether morality is norma-
tive depends on whether the BLT is true. I believe, however, that the BLT
is false. If it is indeed false, then if morality is normative, it follows that the
RD is false. For this reason it is important to pause here to consider objec-
tions to the BLT and to consider arguments in its favor. I think there is
reason at least to doubt that the BLT is true.
The literature contains several arguments for the BLT, a few of which
I have addressed in other places (Copp 2007a, 2010a, 2010b). I cannot
hope to discuss the arguments here. Instead, I will mention two ideas that
might motivate the BLT and I will look briefly at an important strategy that
might be used to support it.
One obvious motivation for the BLT is the belief that the RD is true.
For if the RD is true, then the normativity of morality depends on whether
the BLT is true. If the RD is true, the BLT had better be true. We can
construct on this basis an argument for the BLT that takes as its premises,
first, the proposition that morality is normative, and second, the proposi-
tion that the normativity of morality depends on whether there is a rational
requirement to be morally motivated. The second premise is the RD. The
argument proceeds from the RD and the claim that morality is normative
to the BLT. The trouble, of course, is that the argument is question-begging
in this context. In a context in which there is controversy among different
competing accounts of the normativity of morality, it is question-begging
to assume the RD is true.
A second important motivation for the BLT is that it seems to offer a

theoretically attractive unifying perspective on the relation between practi-
cal reason and morality. According to the BLT, a person is not fully rational
unless she is motived to do what she is morally required to do. Moral moti-
vation is at least partly constitutive of full rationality (Smith 2010: 134–8).
If, then, we think that morality and practical reason cannot plausibly be
independent normative systems, we may conclude that the BLT explains
the link between morality and practical reason. The problem is that we need
some independent reason to think that the demands of morality and the
demands of practical rationality are unified in such a theoretically clean
way.19 We need a reason to think that a unifying perspective of this kind
on the relation between rationality and morality is plausible and illuminat-
ing rather than distorting. Natural law theory in philosophy of law offers
a similarly unifying account of the relation between law and morality, but
legal positivists find it to be distorting rather than plausible.
The best strategy for supporting the BLT, it seems to me, would be to
construct an argument from a general account of the nature of rationality.
Such an argument would begin with an account of the nature of rationality
and then argue on this basis that a moral requirement entails the existence
of a rational requirement, a requirement at least to be morally motivated.
Arguments of this kind are found in Kant (1981), and in recent times,
in Gauthier (1986), Smith (1994, 2010), and Korsgaard (1996, 2009).
Obviously an argument of this kind would not be of interest in the present
context if it rested on assuming the RD is true. But this is not the place to
attempt a systematic evaluation of these arguments.
There are two points I want to make about such arguments. First, even
if one of these arguments is successful, it does not follow of course that
the RD is true. For even if the BLT is true, it may be that the normativ-
ity of morality does not depend on its truth. The normativity of morality
may depend on there being a requirement to be morally motivated that has
normative oomph without depending on the further point, claimed by the
BLT, that this requirement is a rational requirement. Second, the success
of an argument in supporting the BLT would make it seem unimportant
whether the RD is true. We are interested in the RD mainly because it pro-
poses a constraint that must be met by any theory that purports to account
for the normativity of morality. It says that any such theory must entail or
at least be compatible with the truth of the BLT. But if the BLT is true, it
follows from this fact alone that an adequate moral theory must entail or at
least be compatible with the BLT, and it won’t matter whether the reason for
19
Copp (2009) argues to the contrary and in favor of normative pluralism.
150 David Copp
this is that the normativity of morality depends on the truth of the BLT, as
the RD claims, or whether it is simply that the BLT is true. The truth of the
BLT would imply that an adequate non-skeptical moral theory must entail
or be compatible with the BLT even if the RD were false. So the truth of
the BLT would remove the main motivation for worrying about the RD.
I am unconvinced by arguments in favor of the BLT partly because
I think there are good reasons to doubt it is true. In the rest of this section,
I present two objections.
My first objection to the BLT is that it implies that what we are mor-
ally required to do depends on issues about the nature of rationality, issues
that might be arcane and that, in any case, seem beside the point. I find
it counter-intuitive, for example, to suppose that whether we are morally
required not to torture people depends on whether it is necessarily a require-
ment of rationality that we be averse to torturing people. The requirement
not to torture rests on facts about what torture does to its victims not on
whether causing such effects on victims is rationally permitted. It rests on
the horror of being subjected to merciless pain and the terror of being faced
with the prospect of imminently losing one’s life, not on abstract issues
about rationality. Nor, intuitively, does it rest on issues about the nature of
practical reasons.
There is a response to this worry. If the BLT is true, it follows that we are
rationally required to be morally motivated. One might take this to mean
that the BLT constrains the content of moral requirements such that they
must fit some independent account of the nature of rationality. My objec-
tion to the BLT was that it is counter-intuitive that what we are morally
required to do depends in this way on the nature of rationality. Defenders
of the BLT might, however, take a different view, as Korsgaard and Smith
have noted (Korsgaard 2009: 47; Smith 2010: 136–7). For suppose we have
an independent account of what morality requires. One might then view
the BLT as constraining the content of rational requirements such that they
must fit this account of morality. For example, if torture is morally pro-
hibited, it follows from the BLT that there is a rational requirement to be
motivated not to torture. On this approach, Smith points out, ordinary
moral reflection can “provide us with insight into the nature of rationality”
(2010: 136). “Moral theorizing and theorizing about the nature of rational-
ity are one and the same” (137).
This response does not avoid the charge of counter-intuitiveness. For
notice that the BLT entails only that if there are any moral requirements
then there is a rational requirement to be morally motivated. Hence, if
the BLT is true, the nature of rationality might preclude the existence
of any moral requirements. The BLT entails that whether there are any
moral requirements at all turns on issues about the nature of rationality.
Whether we are morally required not to torture would depend on whether

practical rationality is such that there is a rational requirement to be moti-
vated not to torture. So again, on the BLT, what we are morally required to
do, if anything, depends on issues about the nature of rationality. This is
counter-intuitive.
My second objection to the BLT is that, in many cases, a lack of moral
motivation seems not to be at all irrational. Consider Alice’s Pride. Suppose
that Alice is morally required all things considered to apologize to Bob for
something she said. She has been engaged in a political struggle with Bob
over the future direction of their department, and she mis-characterized
something Bob said. She has her pride, however, and a strong sense
of self-respect, and she would feel humiliated if she were to apologize.
Moreover, she sees that if she were to apologize, her political credibility in
the department would be weakened. As a result, she is not at all motivated
to apologize. Her sense of pride and her willpower have extinguished any
motivation she otherwise would have had to apologize. In this case, I sub-
mit, Alice is not failing to be sensible. Intuitively her lack of motivation to
apologize is not irrational in any ordinary sense, or unreasonable or sense-
less. One might think that she is unreasonable to be too proud to apologize
and to let her sense of self-respect prevent her from apologizing. Yet we
need not imagine an extreme or outlandish form of pride in order to make
sense of her lack of motivation to apologize. Alice may be responding to her
deepest intellectual goals for her department and this may fundamentally
be why she lacks any motivation to apologize to Bob. Intuitively, there need
be no irrationality on her part.
A defender of the BLT might challenge the significance of the intuition
that Alice is not irrational to lack any motivation to apologize. For if the
BLT is true, then if Alice is morally required to apologize, it follows imme-
diately that she is rationally required to be motivated to apologize. But this
response concedes that the BLT has counter-intuitive implications and one
might hope to avoid this concession. A defender of the BLT might instead
deny that Alice is morally required to apologize. For if the BLT is true,
then if Alice is not irrational to lack any motivation to apologize, it follows
that she is not morally required to apologize. We might have thought that
whether Alice is morally required to apologize turns on such things as the
impact on Bob of Alice’s mis-characterization of what he said, the nature of
her relationship with Bob, and so on. But on this response to the example,
a defender of the BLT would hold instead that whether Alice is morally
required to apologize to Bob turns on issues about rationality. This strikes
me as counter-intuitive. This was my first objection to the BLT.
A defender of the BLT might now contend that the example is not coher-
ent as I set it out. For if Alice is morally obligated to apologize, then since
152 David Copp
she is not being irrational in the example as I set it out, she must be moti-
vated at least to some degree to apologize. Hence, if Alice is not irrational
in the example, she also does not lack moral motivation. On this response,
whether Alice is morally obligated to apologize turns on an issue about
moral motivation. For if we assume that the BLT is true and that Alice is
not being irrational, then whether Alice is morally obligated to apologize
depends on whether she is motivated at least to some degree to apologize.
But as I set up the example, I stipulated that Alice is not at all motivated
to apologize. This stipulation forces the defender of the BLT to insist that
either Alice is not morally required to apologize or that she is irrational.
Both of these options are counter-intuitive. Intuitively, although Alice
exhibits a moral fault in lacking any motivation to apologize, she is being
sensible and reasonable and she is not irrational in any ordinary sense.
I conclude, then, that the BLT is counter-intuitive on two grounds. First,
it implies that what we are morally required to do (if anything) depends on
the requirements of practical rationality, which means that it depends on
arcane issues in the theory of practical reason. Second, as illustrated by the
example of Alice’s Pride, a lack of moral motivation need not intuitively be
irrational or senseless or unreasonable. So a failure to be morally motivated
does not invariably seem to be a violation of any requirement of practical
rationality. Although of course these objections are not decisive, they lead
me to doubt the BLT.
I am convinced that a defense of the BLT is hopeless, and I am also con-
vinced that morality is normative. For this reason, I believe that the RD is
false. Of course, to those who find the RD attractive, it will seem important
to develop arguments for the BLT. For, given the RD, a defense of the BLT
or of some similar linkage thesis will seem to be required in order to defend
the normativity of morality. My goal, then, is to undermine the RD. In
the rest of this chapter, however, I pursue the limited goal of arguing that,
in a context in which there is controversy about the plausibility of various
accounts of the normativity of morality, one cannot reasonably invoke the
RD in order to adjudicate among the contenders.
6.5 ARGUMENTS AGAINST THE RATIONALITY

DOCTRINE
The RD holds that whether morality has the philosophically important

kind of normative authority depends on whether it has a normative author-
ity that is grounded in a rational requirement to be morally motivated,
as postulated by the BLT. The main reason to be interested in the RD, as
I said before, is that it proposes a constraint that allegedly must be met by

any theory that purports to account for the normativity of morality. The
RD is of interest mainly in light of controversy regarding the plausibility of
various non-skeptical moral theories. I will evaluate it on the assumption
that there is such controversy. No doubt there are moral theories that entail
the truth of the RD. If we had a sound argument for such a theory, then,
of course, any true moral theory would have to be compatible with the RD
because, ex hypothesi, the true theory would entail the RD. My goal here,
however, is to argue that, in a context in which there is controversy about
the plausibility of a variety of non-skeptical theories, one cannot reasonably
invoke the RD in order to adjudicate among the contenders. The RD can-
not do any important work in such adjudication. In this section, I present
three arguments in support of this claim. The basic point, however, is that,
in a context in which there is controversy about the plausibility of a variety
of non-skeptical theories, the RD itself is controversial.
The arguments depend on interpreting the RD as concerned with the
metaphysical grounding of morality and they depend on rejecting inter-
pretations of the RD of the kind I discussed in section 6.2, which trivialize
the RD or make it seem an obvious truth. The objections presuppose an
interpretation of the RD on which it is a substantive thesis. They therefore
presuppose that the concept of normativity is not defined in some way in
terms of rationality, for in this case the RD would be uninteresting. I dis-
cussed this point in section 6.2. But the arguments do not turn on my
specific suggestions about the nature of rationality or the nature of norma-
tivity. They do not depend on my suggestion that rational requirements are
those that must be complied with in order to act sensibly or reasonably or
to avoid irrationality in an ordinary sense. Nor do they depend on my sug-
gestion that whether a requirement has normative oomph turns on whether
a failure to act in accord with it would be a normatively significant failure
of a kind that is independently worrisome.
The first argument is the argument “from other options.” Suppose that
there is controversy about the plausibility of a variety of non-skeptical the-
ories about the normativity of morality. There are theories according to
which the normativity of moral requirements is ungrounded and primitive.
There are theories according to which the normativity of moral require-
ments is grounded in some way other than by relating them to requirements
of practical reason. Unless these options can be ruled out, it will remain
doubtful that the normativity of morality depends on the existence of a
rational requirement to act morally or to be morally motivated. (Recall here
the importance of avoiding the trivializing move of equating the norma-
tivity of a requirement with its being a rational requirement.) This means
that, unless these options can be ruled out, the RD is doubtful. That is, it
154 David Copp
is doubtful in a context of controversy that the lack of a rational require-

ment to be morally motivated would entail that morality is not normative
in the sense that is of philosophical interest. Since defending the RD in a
context of controversy depends on ruling out the primitivist and alternative
grounding views, it would be question-begging in such a context to invoke
the RD to rule out these views.
The second argument is the “argument from parity.” The same question
can be raised about putative rational requirements as the why-be-moral
challenge raises about putative moral requirements. The question is whether
the putative requirements are normative (see Broome 2007). What would
this question come to? I have been working with the following idea. If a
requirement is normative, then a failure to act in accord with it must be
a normatively significant kind of failure, a failure of a kind that is inde-
pendently worrisome. The basic idea, however, is that we can challenge the
normativity of any putative requirement, whether it be a putative moral
requirement or a putative rational requirement. This is the premise of the
parity objection.
This premise comes into play as follows. Any showing that there is a
rational requirement to be morally motivated would depend on a sub-
stantive account of what rationality requires. Once we are offered such an
account, we can question whether the requirements that it postulates would
be normative. Consider, for example, the proposal that the sole requirement
of practical rationality is to maximize the degree to which one’s preferences
are satisfied. The problem is that it is doubtful that the putative requirement
to maximize the degree to which one’s preferences are satisfied has norma-
tive oomph. In the case of any proposal as to what rationality requires, there
is room to question the normativity of the putative requirement. But to
question this would be to challenge whether the requirements proposed by
this account could plausibly have metaphysical priority in explaining the
normativity of morality.
This brings me to the parity argument. I have been assuming that advo-
cates of the RD would hold that rational requirements are normative.
Rational requirements could not plausibly have metaphysical priority in
grounding the normativity of morality, as claimed by the RD, unless they
were normative. But given the premise that putative moral requirements
and putative rational requirements are equally open to challenge, putative
rational requirements cannot be assumed to be normative in a context in
which the normativity of moral requirements is at issue. To account for
the normativity of rational requirements, there appear to be three main
options. One option is primitivism, according to which their normativ-
ity is ungrounded. A second option is reductionism, according to which
their normativity can be explained or grounded without referring to or
quantifying over anything that is taken to be normative.20 A third option

may be a partial-reductionism that invokes the notion of an ur-normativity.
The plausibility of the RD rests on some account of the normativity of
rational requirements that takes one of these approaches. And if such an
account is possible, a similarly structured account of the grounding of the
normativity of morality might also be possible, an account that does not
rest on relating moral requirements to the requirements of practical reason.
It would be question-begging to invoke the RD to rule out theories that
take these approaches in the moral case unless some reason can be given
why approaches that may be viable in accounting for the normativity of
rational requirements would not be viable, mutatis mutandis, in accounting
for the normativity of moral requirements. But a further reason of this kind
would already have ruled out these approaches in the moral case. So again,
in a context of controversy, the RD cannot do any work in ruling out the
primitivist and alternative grounding views.
Third, there is what I will call the “basic argument.” The RD claims
that whether morality is normative depends on whether there is a rational
requirement to be morally motivated. Yet it seems clear that if there is a
requirement to be morally motivated that has normative oomph, this
would be sufficient for morality to be normative. It seems not to be neces-
sary that such a requirement have the further characteristics in virtue of
which it would be a rational requirement. A requirement to be morally
motivated might be normative even if it does not have the specific and dis-
tinctive nature possessed by rational requirements. (On the account I have
proposed, a rational requirement is such that one must comply with it in
order to act sensibly or reasonably and to avoid irrationality in an ordinary
sense.) It seems, then, that whether morality has normative oomph does not
depend on whether the requirement to be morally motivated is a rational
requirement. In a context of controversy, it would be question-begging to
invoke the RD to exclude this position because one could not accept the
RD unless one had already excluded the position.
One might object that one of the issues in the present context is whether
all normative requirements are requirements of rationality. If all normative
requirements are requirements of rationality, then if the requirement to be
morally motivated is normative, it follows that it is a rational requirement.
In this case there would be no further issue whether it is a rational require-
ment.21 In a context of controversy, however, and where the plausibility of
the RD is at issue, it would be question-begging to invoke the premise that
20
I proposed such an account in Copp (2007b).
21
An anonymous referee and Hille Paakkunainen both seemed to offer this objection.
156 David Copp
the requirements of rationality are the only normative requirements. For

this premise entails the RD; it entails that the normativity of moral require-
ments depends on whether moral requirements are requirements of ration-
ality. And the plausibility of this view, as well as primitivist and alternative
grounding views, is precisely what is at issue in a context of controversy.
One competing position holds that the normativity of the requirement
to be morally motivated would be sufficient for morality to be normative
even if it were not a rational requirement. It would be question-begging to
invoke the RD to exclude this position.
We can think of the problem in the following way. The RD claims that
the normativity of morality depends on whether there is a rational require-
ment to be morally motivated. But it surely depends more fundamentally
on whether the requirement to be morally motivated is normative. And it
seems plausible that whether the requirement to be morally motivated is
normative depends on the normativity of the relevant “first-order” moral
requirements—that is, it depends on whether the moral requirements that
a morally motivated person would thereby be motivated to comply with
are normative. If these requirements were not normative, then it seems
to me, the requirement to be motivated to comply with them would not
be normative. It therefore seems that the normativity of the requirement
to be morally motivated would be grounded, if at all, in the normativ-
ity of these first-order moral requirements. A partisan of the RD would
claim, of course, that the normativity of these moral requirements depends
on whether they are or entail rational requirements to act or at least be
motived accordingly. Partisans of primitivist and alternative grounding
views would claim otherwise. But in a context of controversy among such
views, it would be question-begging to invoke the RD to narrow down
the options. The RD cannot do any work in adjudicating among the
competing views.
In this section, I have argued that, in a context in which there is contro-
versy about the plausibility of a variety of theories about the normativity
of morality, one cannot reasonably invoke the RD in order to adjudicate
among the contenders. I have provided three arguments for this conclusion.
Of course, the underlying issue is what grounds the normativity of moral-
ity, and this issue remains unsettled by anything I have said. Nevertheless,
I have been attempting to show only that one cannot reasonably invoke
the RD to rule out primitivist and alternative grounding theories. This is
because the plausibility of the RD depends on having ruled out such theo-
ries. If a view of either of these kinds is correct, then the normativity of
moral requirements does not depend on the existence of a rational require-
ment to be morally motivated.
6.6 THE NORMATIVITY OF MORALITY
The RD takes the normativity of morality to depend on the nature of

rationality. It reflects the idea that practical rationality is the foundation
of normativity. This is why it holds that morality is normative only if its
normativity is grounded in a suitable relation between morality and practi-
cal reason. A second view holds that the normativity of morality is primi-
tive and ungrounded. A third view is reductionist. It holds roughly that
the normativity of morality can be explained or grounded without refer-
ring to or quantifying over anything normative. A fourth view is partially
reductive. It claims that the normativity of morality can be explained or
grounded, but not without referring to or quantifying over something
that is taken to be normative, such as, perhaps, a kind of ur-normativity.
The alternative grounding views agree that morality has a grounding that
is non-rationalistic; they agree that although the normativity of morality
has a metaphysical grounding, its grounding does not consist in a rela-
tion between morality and rationality. Like primitivism, these views reject
the RD.
Nothing in what I have said shows that the RD is false.22 Yet it has per-
haps emerged why I think it is doubtful. The substantive issue at hand is
whether the fact that an action is morally required entails that a failure to
do this action would be a failing of a normatively significant kind. Nothing
substantive seems to turn on whether this failure would be a rational failure,
a failure to act sensibly or reasonably or a failure to avoid irrationality in
an ordinary sense. From this perspective, the RD seems unmotivated and
implausible. Furthermore, as I argued, I think there is reason to doubt the
Basic Linkage Thesis. If the BLT is false, then if the RD is true, morality is
not normative. Since I believe that the BLT is false and also that morality
is normative, I must deny the RD. If we think that morality is normative,
then unless we accept the BLT, we must deny the RD.
I have not aimed, however, to show that the RD is false. Instead
I have attempted to argue that it cannot reasonably be used to exclude
non-rationalistic accounts of the normativity of morality. Of course, if
we had a sound argument to support the RD against non-rationalistic
accounts, we would not need to invoke it to exclude these other views.
The argument to support the RD would already have excluded them. But
in the absence of such an argument, and given that there is controversy
22
I thank an anonymous referee and Hille Paakkunainen for pressing me to clarify
this point.
158 David Copp
about what if anything grounds the normativity of morality, it would be

question-begging to invoke the RD in an argument intended to exclude the
other views. For one would accept the RD only if one had already excluded
non-rationalistic accounts. Hence the RD can do no work in adjudicating
among the competing views. In debates about the normativity of morality,
the RD is not a useful tool.23
References
Broome, J. 2007. “Is Rationality Normative?” Disputatio 2: 161–78.
Copp, D. 2007a. “Moral Naturalism and Three Grades of Normativity,” in Copp,
Morality in a Natural World. Cambridge: Cambridge University Press, 249–83.
Copp, D. 2007b. “The Normativity of Self-Grounded Reason,” in Copp, Morality
in a Natural World, 309–53.
Copp, D. 2009. “Toward a Pluralist and Teleological Theory of Normativity,”
Philosophical Issues 19: 21–37.
Copp, D. 2010a. “Normativity, Deliberation, and Queerness,” in R. Joyce and
S. Kirchin (eds.), A World Without Values: Essays on John Mackie’s Error Theory.
Berlin: Springer, 141–65.
Copp, D. 2010b. “The Wrong Answer to an Improper Question?” in S. Black and
E. Tiffany (eds.), Reasons to be Moral Revisited, Canadian Journal of Philosophy,
Supplementary Volume 33: 97–130.
Copp, D. 2012. “Normativity and Reasons: Five Arguments from Parfit
Against Normative Naturalism,” in S. Nuccetelli and G. Seay (eds.), Ethical
Naturalism: Current Debates. Cambridge: Cambridge University Press, 24–57.
Dreier, J. 2015. “Can Reasons Fundamentalism Answer the Normative Question?”
in G. Björnsson, C. Strandberg, R. F. Olinder, J. Eriksson, and F. Björklund
(eds.), Motivational Internalism. New York: Oxford University Press.
Foot, P. 1978. “Morality as a System of Hypothetical Imperatives,” in Foot, Virtues
and Vices. Berkeley: University of California Press, 157–74.
Gauthier, D. 1986. Morals by Agreement. Oxford: Oxford University Press.
23
I presented versions of this essay to the Central Division of the American
Philosophical Association in Chicago in February 2012 and to the Departments of
Philosophy at the University of British Columbia in October 2012 and at Washington
University in Saint Louis in March 2013. I also presented a version of the essay to the
BAFFLE discussion group, at the School of Law, University of California, Berkeley, in
April 2013 and to the Madison Metaethics Workshop in September 2013. I am very
grateful to those who contributed to the discussions on these occasions for their help-
ful comments. I am especially grateful to Andrew Alwood, Matt Bedke, Eric Brown,
John Brunero, Meir Dan-Cohen, Christian Coons, Dale Dorsey, Elizabeth Harman,
Jonathan Ichikawa, Carrie Jenkins, Nico Kolodny, Charlie Kurth, Adam Morton, Hille
Paakkunainen, Eric Rakowski, Paul Russell, Roy Sorensen, Kevin Toh, R. J. Wallace, Eric
Wiland, and two anonymous referees.
Kant, I. 1981 [1785]. Grounding for the Metaphysics of Morals, trans. J. W. Ellington.
Indianapolis: Hackett.
Korsgaard, C. 1996. The Sources of Normativity. Cambridge: Cambridge
University Press.
Korsgaard, C. 2009. Self-Constitution: Agency, Identity, and Integrity. Oxford: Oxford
University Press.
Mackie, J. L. 1977. Morality: Inventing Right and Wrong. Harmondsworth: Penguin.
Parfit, D. 2011. On What Matters. Volumes I and II. Oxford: Oxford University Press.
Rosen, G. 2010. “Metaphysical Dependence: Grounding and Reduction,” in
B. Hale and A. Hoffman (eds.), Modality: Metaphysics, Logic, and Epistemology.
Scanlon, T. M. 2014. Being Realistic About Reasons. Oxford: Oxford University
Press.
Shafer-Landau, R. 2003. Moral Realism: A Defence. Oxford: Oxford University Press.
Smith, M. 1997. “In Defense of The Moral Problem,” Ethics 108: 84–119.
Smith, M. 2010. “Beyond the Error Theory,” in R. Joyce and S. Kirchin (eds.),
A World Without Values: Essays on John Mackie’s Error Theory. Berlin: Springer,
119–39.
7
Disagreement, Correctness, and
the Evidence for Metaethical Absolutism
Gunnar Björnsson
Metaethical absolutism, or just absolutism, is the view that moral concepts

have non-relative satisfaction conditions that are constant across judges
and their particular beliefs, attitudes, and cultural embedding. Absolutism
is compatible with a variety of more widely discussed views about moral
semantics—descriptivist and non-descriptivist, naturalist and non-natural-
ist, realist and constructivist—but it arguably has more practically impor-
tant consequences than either of these. If absolutism is true, there is an
important sense in which parties of moral disputes are concerned to get the
same things right, such that their disputes can be settled by the facts. If it
is not true, as various forms of relativism and non-cognitivism imply, such
coordination of concerns will be limited.1
The most influential support for absolutism comes from two related
premises. According to the first, moral thinking and moral discourse display
a number of features that are characteristically found in paradigmatically
absolutist domains, and only partly in uncontroversially non-absolutist
domains. Among these features, we find ways of thinking about moral
disagreement, and ways of attributing correctness to moral judgments.
According to the second premise, the best way of making sense of these
features is to assume that absolutism holds for this domain. Moral thinking
and discourse display a certain unity that would be less straightforwardly
explained if forms of metaethical relativism or non-cognitivism were true.2
One can object to this argument from unity in two ways. One might
reject the first premise, denying that morality displays “absolutist” patterns
1
For discussion, see Björnsson (2013).
2
For arguments of this form, see e.g., Brink (1989: ch. 2); Huemer (2005: chs.
2–3); Lyons (1976: 19–20); McNaughton (1988: 39–41); Sayre-McCord (2006: 42);
Shafer-Landau (2003: chs. 2–3); Smith (1994); Streiffer (2003: ch. 1).
Disagreement, Correctness, and Metaethical Absolutism 161
of disagreement. Or one might reject the second premise, either denying

that absolutism can make good sense of the unity of moral discourse, or,
as is more common, arguing that non-absolutist accounts can. My con-
cern here is to develop the latter sort of objection, an objection typically
associated with quasi-realist defenses of non-cognitivism,3 but lately pur-
sued more generally.4 Instead of focusing directly on the moral domain,
I first look at how some aspects of unity can be found in paradigmati-
cally non-absolutist domains, involving judgments of personal taste and
likelihood. I then proceed to provide independently motivated accounts of
attributions of agreement, disagreement, correctness, and incorrectness that
can explain both why absolutist domains display all aspects of unity and
why certain non-absolutist domains display some. Against this backdrop,
I provide preliminary reasons to think that the unity of moral discourse, to
the extent that moral discourse displays unity, can be given a non-absolutist
explanation.
7.1 METAETHICAL ABSOLUTISM
As I have indicated, a central part of my strategy in the following sections

is to draw lessons from attributions of disagreement and correctness in the
domains of personal taste and likelihood. Unfortunately, the semantics for
concepts and predicates of taste and likelihood is notoriously contested. To
allow us to distill the lessons without getting entangled in these debates, this
first section introduces a way of understanding absolutism that stays clear of
controversial talk about truth, content, or propositions expressed.
As I will understand it here, absolutism is a claim about moral judgments
and moral concepts, not about moral language. (Here and in what follows,
I use “judgment” to denote either the psychological act of judging whether
something is such-and-such, or the psychological state resulting from that act;
I will use “claim” to denote the declarative expression of a judgment.) More
generally, what I will call “absolutism” about a concept is the claim that the
concept has non-relative satisfaction conditions that are constant across judges
and judgments. The “satisfaction conditions” of a concept expressed by some
predicate F are those conditions that an item X needs to satisfy in order for
someone’s act of judging that X is F to be internally successful. Finally, for an
act of judging whether X is F to be “internally successful” is for it to achieve
3
See e.g., Björnsson and McPherson (2014); Blackburn (1984, 1991b, 1993a);
Gibbard (2003); Hare (1970); Stevenson (1944); Tersman (2006).
4
See e.g., Björnsson and Finlay (2010); Finlay (2014: ch. 8); Plunkett and Sundell
(2013).
162 Gunnar Björnsson
what one is aiming at—what one tries to get right—merely in virtue of judg-
ing whether X is F.5 (Notice that internal success conditions are properties of
judgments, not of assertions.) To say that absolutism holds for F, then, is to
say that any two judges who are judging in good faith whether some X is F are
trying to get the same things right, or, equivalently, that their judgments have
the same (non-relative) internal success conditions. If one of them judges that
X is F and the other that X is not F, then if one is successful, the other is not;
if both make the same judgment, then they are equally (un)successful.6
Admittedly, talk about what judges are “trying to get right” when judg-
ing whether X is F or about the “internal success conditions” of their judg-
ments might seem to be nothing but a roundabout way of talking about the
truth-conditions of judgments whether something is F. That is not quite
right, however, and it will be important later that the two come apart on
occasion. There are cases in which we have reasonably clear ideas about
what judges are trying to get right, and whether they are trying to get the
same thing right, well before we decide between quite different accounts of
the semantics of the concept in question. There are also cases where we can
agree that someone’s judgment has been internally successful but where it
seems inappropriate for us to say that their judgment was true.
For illustration, consider an example involving judgments of personal
taste. Gus and Tibus both want something to enjoy on their way home from
work. Each is looking at the vending machine, trying to determine whether
there is something tasty on offer. In doing so, each is trying to get something
right, but if they have different palates, they are presumably trying to get
different things right. Each, it seems reasonable to say, is trying to determine
whether there is anything on offer that satisfies his palate (perhaps under
suitably normal conditions). (Each, like most of us, knows perfectly well that
others have other gustatory preferences, but neither would think that this is
relevant for the judgment he is making.) Because of this, it might be that if
Gus and Tibus both conclude that there is something tasty in the vending
machine, only one of them would be successful in his endeavor.7 Given this,
5
In judging whether X is F (whether Belgium has a king, say), I might of course have
a variety of goals other than getting the judgment right (to test my memory, to save face,
etc.), but those are not goals I have merely in virtue of making the judgment.
6
Here and throughout I set to the side obvious cases of vagueness and putative coun-
terexamples to the law of non-contradiction.
7
It should be noted that judgments of taste sometimes have more interpersonal or
idealizing pretensions, being concerned with whether something accords with the taste
of most people, or with suitably refined palates (cf. Doerfler (2012); Egan (2010); Loeb
(2003)). This, however, does not mean that all judgments of taste have such interpersonal
or idealizing ambitions (cf. Goodwin and Darley (2008)). In fact, I believe that the most
common kind of attribution of tastiness in particular lacks these ambitions, and this is
the kind that concerns us here.
we know enough to conclude that absolutism doesn’t hold for the kind of
tastiness concept exemplified by Gus’s and Tibus’s judgments.8 However, we
are far from determining the semantics of tastiness claims, i.e., of declarative
expressions of tastiness judgments. Perhaps such claims are best analyzed
along contextualist lines: a tastiness claim asserts the proposition that some
relevant subject(s)—typically the speaker—is disposed to have positive gus-
tatory experiences from the item in question.9 Or perhaps an expressivist
analysis is preferable: tastiness claims express but do not ascribe such a dis-
position.10 Or perhaps the proposition asserted or its truth-value depends on
the dispositions of those who assess the claim semantically, not on those of
the speaker.11 By focusing on what judges are trying to get right in making
their judgments instead of on the semantics of characteristic expressions of
such judgments, we can avoid getting involved in differences between such
analyses until they become relevant for our purposes.
The case of tastiness judgments also illustrates the possibility of attrib-
uting internal success to judgments without thinking that they are true.
Suppose that after looking at a particular item in the vending machine,
Gus exclaims, “Ah, there’s something tasty!” If the item in question is one
that we find disgusting, we will be reluctant to say that Gus was right, or
that he had said something true. But his judgment would clearly have been
internally successful if the machine contained something that would satisfy
Gus’s palate.
absolutism, then, is not understood in terms of truth, or propositions
expressed, but in terms of the satisfaction conditions of concepts, which in
turn are understood in terms of the judgment-internal success of the rele-
vant kinds of judgments, and the concepts involved. Metaethical absolutism
is absolutism about moral judgments and concepts. More specifically, I will
take it to be absolutism about concepts that are central to moral thinking
8
By contrast, absolutism might well hold for explicitly relativized tastiness judg-
ments: if Gus and Tibus were both judging whether the machine contains something
tasty for Gus, they would likely be trying to get the same thing right. In saying this,
I distinguish the concept at play when Gus and Tibus are judging whether the machine
contains something tasty from the concept at play when they are judging whether it con-
tains something that is tasty for Gus. I think of a concept expressed by a predicate F as a
kind of mechanism the operation of which constitutes the agent’s activity of judging
whether something is F. Since the two italicized characterizations of judgments in the
previous sentence pick out two extensionally different kinds of judgments, they also pick
out two different kinds of concepts. This is not to deny that the concepts are closely
related—for all I have said, they might be simultaneously at play when Gus is judging
whether something is tasty. I thank Ben Lennertz and an anonymous referee for pressing
me on this point.
9
See e.g., Glanzberg (2007); Schaffer (2011); Sundell (2011).
10
See Buekens (2011), who combines contextualist and expressivist elements.
11
See e.g., Egan (2010); Lasersohn (2005); Pearson (2013); Stephenson (2007).
and involved in paradigmatic cases of moral disagreement: concepts of

intrinsic good, moral wrongness and blameworthiness, justice, normative
reasons, and ought-all-things-considered-to-be-doneness, to mention some
of the most prominent.12
7.2 THE ARGUMENT FROM unity
Hardly anyone defends absolutism with reference to detailed studies of the

individual parties of moral disagreements and what each party is trying
to get right when making their judgment. Rather, belief in absolutism is
grounded in general and easily accessible aspects of moral judgment and
moral discourse. The most important and pervasive source of support for
absolutism is a package of interrelated features that people normally seem
to attribute to paradigmatically absolutist discourse and thinking, but not
to clear cases of non-absolutist discourse. The package can be described in
slightly different ways, but I take the set of features that I will call disagree-
ment, correctness, independence, and no relativization to cover its
core. I will refer to this set as “unity,” as it is meant to represent how we
take people to be concerned with the same issue in paradigmatically absolut-
ist discourse.
disagreement: If A claims that X is F and B that X is not F, then
in agreeing with one claim, one disagrees with the other.13 (When A
and B both agree with their own claims, they disagree with that of
the other.)
12
Just as non-absolutism about tastiness is compatible with a variety of analyses of
“tasty,” metaethical absolutism, as it is understood here, is itself compatible with different
semantic views: Kantian prescriptivism, non-naturalistic as well as naturalistic realism,
what Sharon Street (2006) calls “rigidifying antirealism,” and ideal observer or advisor
theories. It is incompatible with various forms of relativism as well as with at least some
versions of non-cognitivism. For example, Gibbard’s (2003) view seems perfectly com-
patible with the assumption that when different people are deciding whether to plan
to φ in C (i.e., judging whether φ-ing is the thing to do in C), what standards for such
planning they are ultimately trying to conform to might be quite different. However,
absolutism is compatible with a contextual semantics for evaluative and normative expres-
sions such as “good” and “ought.” Even if such terms pick out different ends or norms
in different contexts, it might be that they pick out the same ends or norms when used
in moral contexts: moral value and the moral law, say. Finally, as it stands, the definition
allows for error-theoretic versions of absolutism, according to which no positive moral
judgments are ever internally successful.
13
Again, I set to the side complications due to obvious cases of vagueness and putative
counterexamples to the law of non-contradiction.
correctness: If A claims that X is F and B that X is not F, then if

one claim is correct (true, right) the other must be incorrect (false,
wrong).14
independence: What determines the correctness of A’s claim that X is
(not) F are features of X that do not essentially depend on A’s attitudes
toward X.15
no relativization: If one is involved in disagreement about whether
X is F, one does not take answers of the form X is F relative to Y or X
is not F relative to Z to resolve the disagreement (unless a non-relative
answer can be derived from these).
Talk about “agreement” and “disagreement” can be understood in different
ways, some of which will involve the other features of unity, in particular
correctness. However, I think that there are weaker notions, captured by
most everyday attributions of agreement and disagreement. The notions I
have in mind here can be roughly operationalized as follows: B agrees (disa-
grees) with A’s claim that such-and-such if and only if B’s judgment about
the matter would make it natural to say “A said that such-and-such, and B
thinks so too (but B doesn’t think so).” They are also naturally expressed
using “yes” and “no” in reaction to claims, where the target of the accept-
ance or rejection is not intuitively metalinguistic, concerned with the
pronunciation or choice of words, as in: (1) A: “Pedants correct the [pro-
noun-ciation] of others.” B: “No they don’t. They correct the [pro-nun-
ciation] of others.” (2) A: “Burgers come with chips.” B: “No they don’t.
Burgers come with french fries.”16
It might seem plausible that, if we let “F” express any of the moral con-
cepts with which absolutism is concerned, unity represents how engaged
participants tend to understand moral reasoning and moral debate, even
in cases of deep moral disagreement. (Think for example of disagree-
ments about whether we have an obligation to help distant strangers in
need when we can do so at little cost to ourselves.) By contrast, we do
not understand recognizably non-absolutist domains as involving all
features of unity. For example, suppose A claims that the cathedral is
14
Cf. n. 13.
15
For simplicity, I’ll ignore cases where X itself involves A. The blocking of essential
dependence allows that contingent effects of A’s thinking might affect X’s F-ness. (For
example, that A thinks that it would be morally wrong to treat B in a certain way might
make it wrong to treat B in that way under certain circumstances.)
16
The notion is thus somewhat narrower than that employed by Sundell
(2011: 275–6), from which I borrow the two examples.
to the left while B denies that it is to the left, but that A and B express
directional judgments made from different spatial perspectives.17 Then
we can well agree with and deem true or correct (or disagree with and
deem wrong, false or incorrect) what both said. Furthermore, this would
be true even if A and B were all-knowing and perfectly rational creatures,
and any disagreement about whether the cathedral is to the left might be
fully resolved by determining whether it is to the left relative to a certain
perspective.
Insofar as it satisfies unity, then, moral discourse and thinking very
much looks like an absolutist domain, and looks unlike the clearest cases
of non-absolutist domains. Moreover, if the moral domain behaves in ways
characteristic of absolutist domains, it might seem reasonable to assume
that it is an absolutist domain, unless we have positive reasons to the con-
trary. (If it quacks, walks, and looks like an absolutist domain, our default
assumption should be that it is one.)
This argument can be strengthened by the following considera-
tion: unity not only seems to represent an ordinary understanding of
absolutist discourse, but an understanding that would seem perfectly
adequate given absolutism. If we are indeed concerned to get the same
things right, it makes sense to let agreement, disagreement, and assess-
ments of correctness follow these patterns. If we are not concerned to get
the same thing right, the rationale for these features of the practice seems
much less clear.
Metaethical absolutism might seem to be jointly supported, then, by
its promise to make straightforward sense of unity and by the fact that
we characteristically recognize its full package in paradigmatically abso-
lutist domains of discourse and judgment, but never in paradigmatically
non-absolutist domains. Conversely, forms of non-absolutism seem to be
undermined. Or so people have thought.18
There are of course various ways in which one might resist the argument
from unity. One is to deny that moral discourse and thinking display
unity. Empirical evidence suggests that not everyone always under-
stands moral discourse to satisfy all aspects of unity. However, patterns
of understanding characteristic of paradigmatically absolute domains
seem predominant (especially within cultures) for what most people take
to be serious and uncontroversially moral wrongdoing, and considerably
17
In this case, absolutism fails for A’s and B’s judging whether the cathedral is to the
left when we individuate kinds of judgments and corresponding concepts in a way that
ignores differences in perspective. absolutism might well hold for judgments whether the
cathedral is to the left from A’s perspective. Cf. n. 8.
18
For references, see n. 2.
stronger than for what is commonly understood as matters of taste.19

Moreover, various rejections of aspects of unity do not immediately under-
mine the argument as at least some such rejections can be understood as
expressing doubts about the feasibility of insisting upon or producing gen-
erally convincing arguments for the moral views in question, or as expres-
sions of degrees of tolerance;20 further empirical research is needed.
My focus here will instead be on the claim that absolutism provides the
best explanation of unity, to the extent that unity is indeed displayed in
the moral domain. Elsewhere I have elaborated on the familiar relativist
or non-cognitivist argument that widespread and seemingly fundamental
moral disagreement undermines straightforward absolutist explanations of
unity.21 My specific objective here is to propose what I take to be the most
promising route for non-absolutist explanations.
Non-absolutist accounts of unity can often seem problematic, relying
on strong forms of motivational internalism, or failing to make sense of
disagreement that involves no clash of attitudes.22 Moreover, one might
worry that such accounts are too disconnected from a general understand-
ing of why unity is displayed elsewhere, and inevitably ad hoc in their
appeal to mechanisms that are clearly not at work in paradigmatic absolutist
domains.23 In the absence of other strong reasons to reject absolutism, then,
one might well feel that unity gives it considerable support.
What I will argue in this chapter, however, is that non-absolutist expla-
nations of unity need not be ad hoc. Aspects of unity seem to hold not
only in paradigmatically absolute domains, but also, to a limited extent,
in paradigmatic non-absolute domains, such as those of personal taste
and epistemic modals. Moreover, given an independently plausible gen-
eral account of the mechanisms governing attributions of agreement or
19
See e.g., Goodwin and Darley (2008); Wright et al. (2012); Sarkissian et al. (2011).
20
See Wright et al. (2014).
21
See Björnsson (2012).
22
Most metaethicists these days reject strong motivational internalism (see Björklund
et al. (2012)). For discussion of problems with accounts of disagreement in terms of
clashing attitudes of the sort proposed by Stevenson’s (1944, 1963), see e.g., Gibbard
(2003: ch. 4); Dreier (2009), and Ridge (2013).
23
Simon Blackburn (1991a, 1991b), in a debate with Nicholas Sturgeon (1991),
argues that the best “projectivist” account of sameness of meaning is the mere application
of a more general account, covering ordinary descriptive discourse. Michael Ridge (2013)
puts forth an account of disagreement that is meant to be general and avoid problems
with Stevenson’s and Gibbard’s proposals. (I think that Ridge’s account is on the right
track, but argue elsewhere that it runs into problems with disagreements about taste, and
fails to capture what is ultimately driving intuitions about disagreement; cf. Björnsson
(n.d.).) The account of unity developed here could be seen as an effort to lay the ground
for a more complete and independently compelling picture in the spirit of these earlier
proposals.
disagreement and attributions of correctness or incorrectness in both abso-

lute and non-absolute domains, such mechanisms might well give rise to
the full range of unity independently of whether absolutism holds for that
domain. This leaves open the possibility of an independently motivated and
entirely non-ad hoc non-absolutist account of unity in the moral domain,
and one that does not rely on strong forms of motivational internalism.
7.3 ASPECTS OF unity IN NON-ABSOLUTIST

DISCOURSE: THE CASE OF TASTY AND LIKELY
We have already noted that elements of unity are violated by claims about
whether something is to the left, and the same is true about claims employ-
ing a variety of context-dependent locutions, including “local,” “ready,” and
“tall.” But judgments expressed by other paradigmatically non-absolutist
locutions—in particular epistemic modals and various predicates of personal
taste—seem to satisfy several aspects of unity even on occasions where judges
are trying to get different things right. Understanding why these judgments
display such patterns will tell us what to expect in the case of moral judgments.
In this section, we will consider one example each of two classes of
predicates: predicates of personal taste, and epistemic modals. Throughout,
I will appeal to phenomena that have been adduced in support of
assessor-relativism in these areas, but as before I will remain neutral as to
whether relativistic analyses of these phenomena are preferable to contextu-
alist or expressivist analyses.
7.3.1 Tastiness
Start with predicates of personal taste, exemplified by tasty. Suppose that our
friends Gus and Tibus have each decided to determine whether Marmite is
as tasty as some say, or as vile as others claim. Each has a taste, and each
expresses his judgment, in soliloquy:
Gus: “That’s surprising. Marmite’s tasty.”
Tibus: “What a disappointment. Marmite’s not at all tasty.”
We have already said that absolutism is violated for tastiness judgments.
When Gus judges that Marmite is tasty, he doesn’t take the reactions
and judgments of others to matter for his verdict: the ultimate arbiter is
whether Marmite would give him a pleasant gustatory experience (perhaps
under normal circumstances). Likewise, mutatis mutandis, for Tibus.24
Assuming, as before, that we are concerned with expressions of personal taste (cf. n. 7).
24
Nevertheless, it seems that Gus and Tibus disagree, and more generally
that if someone agrees with one of the two tastiness claims, he thereby
disagrees with the other. So disagreement seems to be satisfied for tasti-
ness claims, even when judges are concerned to get different things right
when making the judgments expressed.25 Relatedly, it seems a little prob-
lematic to think that both Gus’s and Tibus’s claims are true, as that might
seem to involve agreeing with both their claims. On the other hand, many
would be hesitant to think that Gus’s (or Tibus’s) claim was incorrect or
false merely on account of disagreeing with him about the tastiness of
Marmite. So correctness has at most a limited appeal. Similarly, inde-
pendence is problematic. Many of us would find it misplaced of Tibus
to deem Gus’s claim incorrect or false without taking into account Gus’s
taste disposition. Likewise, finally, for no relativization: once we have
determined for whom Marmite is and is not tasty (and under what cir-
cumstances), there seems to be nothing left of the question of whether it
is tasty.26
Notice, though, that attributions of correctness and incorrectness seem
unproblematic when all involved are assumed to have similar taste reactions
to the item in question. If Gus and Tibus knew that they have the same taste
and Tibus claimed that Marmite isn’t tasty while misremembering the taste
of Marmite, Gus could naturally say, “You’re wrong, it really is quite tasty,”
and if Tibus had claimed that Marmite is tasty, Gus could have naturally
replied “That’s true, it really is.” In such contexts, tastiness judgments might
seem to satisfy correctness.
Although the example of tastiness discourse shows that disagreement
is not unique to absolutist domains, it fails to satisfy the other aspects of
unity, except to a limited extent within contexts where an overlap of taste
dispositions is assumed. In itself, then, the case of tastiness might seem to
strengthen the argument from unity: we (or at least many of those involved
in engaged moral disagreement) understand moral discourse along the lines
of absolutist discourse, not along the lines of this clearly non-absolutist
domain.
25
This is of course part of what has motivated various relativist theories about taste
judgments (see e.g., Egan (2010); Kölbel (2004); Lasersohn (2005); Pearson (2013);
Stephenson (2007)).
26
In effect, Robin McKenna (2014) relies on this mismatch between disagree-
ment and correctness in the case of taste judgments to raise a problem for my
earlier treatment of disagreement in Björnsson and Finlay (2010). The account of
attributions of disagreement and attributions of correctness and incorrectness (truth
and falsehood) provided in sections 7.4 and 7.5 promises a principled explanation of
this mismatch.
7.3.2 Likelihood
Another family of non-absolutist predicates displaying aspects of unity
are epistemic modals, expressed in English by locutions such as “likely,”
“improbable,” “might,” “possible,” “can,” and “must.” Here we will focus on
likelihood judgments. Suppose that Basil needs to get hold of his cat Felix
to bring him to the veterinarian. Basil muses:
“Felix does sometimes wander off to the stream to fish, but his favorite
pastime is chasing birds, and most birds are at the apple orchard today.
So he is likely to be in the orchard.”
What Basil tried to get right in drawing his conclusion was a relation
between evidence available to him before the beginning of his search and
the various places where Felix might be: his question concerns where Felix
is likely to be given that evidence. He knows that someone might have bet-
ter evidence—quite a few people are out and about and might have seen
Felix—but this doesn’t affect what he is trying to get right, as he has no fea-
sible way of accessing that evidence.27 As it happens, Claudia is one of the
people who had seen Felix, as he was heading in the opposite direction of
the apple orchard a few minutes ago, towards the stream where she knows
that he occasionally goes to fish. When Claudia asks herself where Felix
might be, she concludes:
“Felix is unlikely to be in the apple orchard.”
Like Basil, Claudia knows that someone might have better evidence regard-
ing Felix’s whereabouts but lacks any feasible way of getting access to that
evidence. Her judgmental efforts, it seems, are successful if Felix is unlikely
to be in the orchard given evidence available to her. Since the evidence avail-
able to Basil and Claudia differs, absolutism doesn’t hold for their likeli-
hood judgments.
At the same time, it seems that disagreement is satisfied: intuitively, if
we were to agree with or accept Basil’s claim that Felix is likely to be in the
orchard, we would disagree with and have to reject Claudia’s claim that he
is unlikely to be in the orchard. Moreover, correctness might seem to be
satisfied in at least some contexts where absolutism is violated. Consider
the following exchange:
27
In describing this scenario, I am not denying that we sometimes make likelihood
judgments with more interpersonal or objective pretensions, just that this is what is going
on here. Cf. n. 26.
finding felix:
Alicia: “I need to find Felix. I wonder where he might be.”
Basil: “The birds are in the orchard today. He is likely to be there.”
Claudia: “I think that’s wrong, Alicia. I saw Felix head in the other direction
earlier. He is more likely to be by the stream.”
Claudia’s assessment of Basil’s initial claim seems natural enough. Claudia
thinks that Felix is unlikely to be in the orchard, and this seems to not only
force her to reject Basil’s claim, but also to negatively assess its correctness.
Moreover, correctness does not seem to presuppose absolutism: it seems
natural to assume that Claudia’s assessment was based on evidence that Basil
was not concerned to relate to when making his judgment, evidence avail-
able only to her.28
Similarly, I would expect parties of finding felix to take the other aspects
of unity to be satisfied. independence will seem to be satisfied, as Claudia’s
and Alicia’s assessments of the correctness of Basil’s judgment will take into
account any available evidence concerning Felix’s whereabouts, without
restriction to what information Basil had. Likewise for no relativization: If
Basil and Claudia agreed that Felix is likely to be in the orchard given Basil’s
evidence and unlikely to be there given Claudia’s, they would not thereby
have agreed about whether Felix is likely to be in the orchard or not.
It seems, then, that at least from the point of view of participants, some
cases of non-absolutist discourse can display all four aspects of unity. But
the picture is complicated, as not all the aspects hold universally, or from all
points of view. Consider:
hindsight:
Having talked to Basil, Claudia, and some kids, you now know the
following: (1) At nine o’clock this morning, Basil concluded, based on
long experience and knowledge about the whereabouts of the birds,
that Felix was likely to be in the orchard. (2) Independently, and at
about the same time, Claudia concluded that Felix was unlikely to
be in the orchard, as she had seen him take off in the opposite direc-
tion, towards the stream, earlier that morning. (3) As a matter of fact,
28
It might be argued that this assumption should be abandoned, and that the best expla-
nation of why correctness seems to hold in this context is that we take Basil to express
a “communal” judgment, concerned with what is likely given the information available to
either of the parties of the conversation. If he were, then absolutism would hold for Basil
and Claudia’s judgments. The analogous argument has been made by Janice Dowell (2011)
for the case of epistemic “might.” I defend the naturalness and correctness of the solipsistic
interpretation against Dowell’s argument at some length in Björnsson and Almér (2010);
Montminy (2012) independently replies along partly similar lines (also cf. Swanson (2011)).
however, and surprisingly given either Basil’s or Claudia’s evidence,

Felix was in the back yard at the time, playing with the neighborhood
kids.29
What is interesting about our situation in hindsight is this: while we know
where Felix actually was, and what evidence different people had about his
whereabouts, we do not seem to be in a position to judge whether Felix was
likely to be in the orchard. It might be natural to say that Felix was unlikely
to be in the orchard in light of Claudia’s information, but likely to be there in
light of Basil’s, but it would seem odd for us to judge, based on everything
we know, without explicit relativization to some other epistemic perspec-
tive, that Felix was unlikely to be in the orchard.
Now consider the four aspects of unity, as they will appear from our
hindsight perspective. First, disagreement might seem difficult to assess.
Exactly because we cannot judge (without relativization) whether Felix was
likely to be in the orchard, it is unclear whether we are in a position to
agree or disagree with Basil or Claudia’s judgments. Perhaps, though, it is
still true that if one were to agree with Basil’s judgment that Felix was likely
to be in the orchard—if one were to judge that Felix was likely to be in
the orchard—one would have to reject Claudia’s judgment, and vice versa.
correctness and independence seem more deeply problematic, however,
as natural assessments of the correctness of Basil’s and Claudia’s judgments
seem to relate them to their respective epistemic perspectives at the time,
rather than to information had by us now, or by others at the time of Basil’s
and Claudia’s judgments. For example, the following assessment of Basil’s
judgment, explicitly relating to his information, seems natural enough from
the perspective of hindsight:
“Basil thought that Felix was likely to be in the orchard. That was cor-
rect: given information available to him at the time, Felix was indeed
likely to be in the orchard.”
By contrast, this assessment, relying on information not available to Basil,
does not:
“Basil thought that Felix was likely to be in the orchard. That was
incorrect: as we now know from the kids, Felix was not at all likely to
be in the orchard.”30
29
Von Fintel and Gillies (2008) appeal to retrospective cases like this to undermine
evidence for assessor-relativism of the sort defended by Egan (2007), MacFarlane (2011),
and Stephenson (2007). Some of the other cases they appeal to transpose to the case of
“likely.”
30
In eliciting intuitions about cases like these, it is important to be aware of various
sources of error. One is that attributions of correctness might target the formation of a
judgment rather than the judgment formed. Another is that people might take beliefs
Finally, no relativization seems to be violated from the hindsight per-

spective. If we have answered questions of the sort “from epistemic perspective
E, where was Felix most likely to be?” for all interesting perspectives, we seem
to have exhaustively resolved any disagreements we might have had about
where Felix was likely to be.
In this section we have seen how two paradigmatically non-absolutist
domains satisfy some, but not all, aspects of unity, and sometimes only par-
tially. In itself, this might seem to reinforce the argument from unity. What
I will argue in the next three sections, however, is that when we understand
why aspects of unity are displayed in absolutist discourse and partially
displayed in non-absolutist discourse, there is room for a non-absolutist
account of unity in the moral domain.
7.4 UNDERSTANDING disagreement
The purpose of this and the following section is to propose, in outline,

general accounts of the various phenomena involved in unity. In particular,
I will sketch accounts of what it is to agree or disagree with what someone
says or thinks, and accounts of what it is to take a claim or judgment to
be correct or true, or incorrect or false. The guiding principle is that the
accounts should not only predict, but also make intelligible why all features
of unity are present in the case of paradigmatically absolutist discourse,
none are in certain kinds of non-absolutist discourse, and some are, to vari-
ous degrees, in the case of tastiness and likelihood judgments.
In this section, I propose a general account of agreement and disagree-
ment. As illustrated by cases involving tastiness and likelihood judgments,
agreeing with someone’s claim is not essentially a matter of thinking that
they have been internally successful in making the judgments they expressed.
The explanation for this, I will argue, is that agreement and disagreement
relate to the claim’s communicative function, and to the related phenomena
of accepting or rejecting the claim.
Generally, accepting or rejecting an utterance is a matter of going along
with, or ruling out going along with, the standard flow of communication
that happen to be false to be less justified (Young et al. (2010)), which can lead them to
think that the judge (Basil, say) had not adequately taken into account available informa-
tion. Yet another is that people who know whether p is the case will fail to assess the claim
that p is likely, instead assessing the claim that p, as this is an epistemically more relevant
target of assessment (cf. Almér and Björnsson (2009); Björnsson and Almér (2010)).
Still, new evidence strongly suggests that hindsight judgments follow the patterns postu-
lated here (Knobe and Yalcin (forthcoming)).
initiated by the utterance. We have ways of saying, asking, entreating, and

so forth, because such acts tend to have specific communicative effects: that
our audience believes what is said, takes on the task of answering the
question, forms intentions to comply, and so forth. To some extent, our
understanding of communicative function might be closely tied to an inde-
pendent understanding of speakers’ perlocutionary intentions. But it is pri-
marily driven by our general understanding of the words employed and
syntax, including the sentence form (declarative, interrogative, imperative),
and aspects of the context implicated by this understanding. Plausibly, this
general understanding is an understanding of the conventional communica-
tive function of the sorts of expressions involved, adapted to the characteris-
tic psychological effects on hearers of the expressions that explains why, in
general, we tend to use those kinds of expression, and under what circum-
stances.31 Even when speakers clearly do not intend us to believe a claim,
answer a question, or fulfill a request, then, we can readily understand their
utterances as utterances whose communicative function, in virtue of the
conventional function of its various elements and the context, is to elicit
judgments or intentions in hearers.
The idea now is that to accept a claim, question, or request, is exactly to do
one’s part as a hearer in the performance of its communicative function: judg-
ing that something is the case, intending to provide an answer, or intending
to comply. Correspondingly, to reject an utterance is to be in the sort of state
that most directly rules out being in the former state: judging that something
is not the case, intending not to provide an answer, intending not to comply.32
Our primary concern here is specifically with what I have called “claims”—
expressions of judgment in the declarative. The characteristic communica-
tive function of claims, I take it, is to produce or make occurrent judgments
in hearers: we characteristically make claims because they produce or make
31
See e.g., Millikan (1984); Stevenson (1944). Following Millikan (1998), I take
speech acts to be usefully categorized based on their conventional communicative func-
tion, and traditional speech act categories to closely track such functions.
Notice that the conventional communicative function of an utterance can be rela-
tional and context-dependent. For example, it is presumably a conventional function of
an utterance in the declarative (“The cathedral is to the right”) to produce or make occur-
rent a judgment the content of which depends on the words constituting the sentence as
well as context needed to fix the content of context-dependent expressions.
32
The notion of communicative function is closely related to the notion of an expres-
sion’s context change potential in dynamic semantics, and to the resultant of the force and
content of a speech act (see e.g., Stalnaker (1999)). Likewise, the notion of acceptance of a
claim employed to be introduced is closely related to the notion of acceptance employed
by Stalnaker (see e.g., (2002)). However, the use to which it is put here does not presup-
pose the adoption of that semantic framework: for all I say here, the meaning of sentences
might be analyzed in some other way.
occurrent suitably related judgments in hearers, and our understanding of

sentences in the declarative is shaped by this function.33 I thus propose that:
accept/reject: To accept a claim is to make the judgment that would
constitute doing one’s part as a hearer in fulfilling the characteristic
communicative function of that sort of claim; to reject the claim is to
make the contrary judgment.34
The basic disagreement phenomenon is then understood as involving two
claims where the acceptance of one rules out the acceptance of the other;
disagreements involving judgments are understood as propensities for disa-
greements of this basic form:
agree/disagree: Two claims are in agreement when accepting (reject-
ing) one would mean accepting (rejecting) the other; they are in disa-
greement when accepting one would mean rejecting the other. To agree
(disagree) with some claim C is to make a judgment the characteristic
declarative expression of which would be in agreement (disagreement)
with C. To agree (disagree) with some judgment J is to make a judgment
the characteristic declarative expression of which would be in agreement
(disagreement) with the characteristic declarative expression of J.35
33
Relatedly, Millikan’s (1984) argues that the “stabilizing proper function” of the
indicative mood is to produce true belief in hearers: true belief because this is what both
speakers and hearers tend to have an interest in such that this reinforces speaker and
hearer dispositions. This explanation is correct and important, but for the explanations
of acceptance, rejection, agreement, and disagreement developed here, we can bracket the
truth-value of the judgment.
34
I am assuming here that the judgment that not-P and the judgment that P are each
other’s contraries. (Dialetheists will need a somewhat different account than that offered
here; I set that complication to the side.)
One might of course resist accepting an utterance without thereby rejecting it,
namely by withholding judgment or more generally being undecided. The sort of rejec-
tion in judgment that we are concerned with here also contrasts with a metaconceptual
or metalinguistic refusal to employ a certain term or concept (because of its despicable
connotations, say). I thank Guy Fletcher for raising this complication. Another compli-
cation, raised by Michael Ridge, comes from contents that are in certain ways peripheral
to a claim, such as the implications of “even” in “Even Granny had some wine,” or par-
entheticals like that in “Lance Armstrong, who was born in Copenhagen, won the Tour
de France seven times.” To produce or make occurrent acceptance in judgment of these
contents seem to be part of the communicative function of the claims, but contrary to
accept/reject one might think it possible to reject these contents without rejecting the
claims. I lack space here for a full discussion, but I am inclined to deny that possibility:
we do reject a claim if we reject some of its peripheral contents. The reason one might
nevertheless feel that one “accepts the claim” is simply that these contents are less central
to communicative concerns: one can accept the central parts.
35
Understanding disagreement in judgment in terms of disagreement between claims
might seem roundabout, but I argue elsewhere that this idea explains why agreement and
acceptance can come apart in the case of tastiness judgments (Björnsson (n.d.)).
Consider how this applies to the case of most ordinary, non-relative, descrip-
tive claims. The characteristic communicative function of such claims, we can
assume, is to inform (or remind) hearers about a certain aspect of objective
reality. Here, the judgment that satisfies the characteristic communicative
function of the claim is simply the judgment that this aspect obtains. To accept
the claim is thus to make such a judgment, and to reject the claim is to make
the contrary judgment. Given this, agree/disagree yields disagreement for
discourse involving ordinary, non-relative, descriptive claims. If A said that X
is F and B said X is not F, and F is a non-relative descriptive predicate, then
accepting one claim means rejecting the other.36 Correspondingly, agreeing
with A’s (B’s) claim means making the judgment characteristically expressed by
saying that X is (not) F, and thus to disagree with B’s (A’s).
If this account of disagreement in paradigmatically absolutist discourse
is correct, the reason that claims expressing tastiness and likelihood judgments
follow the pattern of disagreement should be that their communicative func-
tions are relevantly similar to those of ordinary descriptive statements. So what
are those communicative functions?
Start with tastiness claims. Their characteristic function, I suggest, is to
create (or make occurrent) tastiness beliefs in hearers and corresponding
expectations of what hearers will find tasty or not, thus providing gustatory
guidance.37 It is clear that we routinely do form tastiness beliefs on the basis of
tastiness claims, and clear that speakers adjust their utterances in light of this
effect, making explicitly subjective claims (“I think that this is really tasty”)
when they have specific reasons to suspect that the audience does not share
their taste preferences and will be misled if accepting a bare tastiness claim.38
Moreover, it makes good sense that expressions of tastiness judgments would
have this function. True, differences in taste are common enough for us to only
sometimes be in a position to confidently judge that all our hearers will share
36
Ignoring, as before, complications related to vagueness and dialetheism.
37
More generally, tastiness claims function to express the speaker’s taste judgments,
which can be done to guide taste expectations, but at other times merely to convey one’s
food preferences (as when small children express their taste judgments to parents), or
perhaps to compliment the chef. However, as I have argued elsewhere (Björnsson and
Almér (2010: 23–6)), this general expressive function provides an appropriate target for
assessments and potential acceptance or rejection only rarely, as speakers tend to have
epistemic authority concerning their taste dispositions, leaving hearers without inde-
pendent grounds for accepting or not accepting that speakers have the taste dispositions
in question. For this reason, intuitions of agreement (and correctness) will typically fol-
low the guiding function.
38
In saying that we withhold claims of the form “X is tasty” when we think that
others have different taste preferences, I am not saying that we also withhold our cor-
responding judgments. (For two different explanations of why explicitly relativized
or first-personal taste judgments make for a different communicative function, see
Björnsson (2001: 101–3); Björnsson and Almér (2010: 26–7).)
our taste reaction to a particular item. But there is enough overlap for expres-
sions of subjective judgments to serve as reasonably reliable guides for others
who have not yet tasted the item.39
Given this communicative function of tastiness claims, the response
required for its fulfillment is that hearers expect to have the corresponding
taste experience from the item in question. Given accept/reject, to accept
the positive claim that X is tasty or reject the negative claim that X is not
tasty is thus to expect to have a positive taste experience from X; to accept
the claim that X is not tasty or reject the claim that it is tasty is to expect
not to have a positive experience from X. Given that accepting one of these
claims means rejecting the other and given agree/disagree, disagreement
follows for tastiness claims.40
A similar story is available for likelihood judgments and likelihood claims.
The core role of likelihood judgments is presumably to set our subjective
probabilities for various possibilities—high for possibilities we judge likely;
low for those we judge not likely—thus affecting the amount of cognitive,
emotional, and physical work that we invest in these possibilities. The sug-
gestion now is that the characteristic communicative function of likelihood
claims (i.e., declarative expressions of likelihood judgments) is to produce
the corresponding likelihood judgments in hearers, i.e., likelihood judg-
ments that would typically set (roughly) the same subjective probabilities.
A practice with claims performing this function makes good sense when
speakers often enough have better information about the matter at hand
and have thought more about the issue than have hearers. Unless speak-
ers have weighed the evidence erroneously, or hearers have relevant evi-
dence that speakers have not taken into account, hearers can improve their
epistemic situation by adopting the corresponding subjective probabilities,
now based on evidence that includes the speaker’s claim.41 Moreover, this
39
This accounts for the major difference between the tastiness case and ordinary descrip-
tive discourse: the mismatch between the judgment expressed and communicative purpose.
Taste claims function to create taste beliefs and expectations of taste experiences, but they
do something weaker than assert that others will have the relevant experiences. (To see why
the assertoric content of ordinary tastiness claims cannot plausibly be “all of us [in this con-
versation] would enjoy this,” consider first that such a claim seems to require much more
knowledge about one’s audience than does “this is tasty,” and, second, that if this were the
content, the fact that one judge, A, doesn’t like an item should provide another, B, with
grounds for rejecting C’s claim that the item is tasty, which it clearly doesn’t: in normal
conversational settings, B has such a ground only if B thinks that she doesn’t like it.)
40
The idea that agreement and disagreement about taste should be understood in terms
of the communicative function of taste claims is very similar to Egan’s (2010) proposal.
41
Recently, a number of people have tried to analyze epistemic modals in terms of
communicative function. See especially Willer’s recent (2013) proposal to understand
them in terms of their context change potential and Björnsson and Finlay’s (2010)
proposal to understand epistemic relativity in deontic modals in terms of the practical
practice would not only make sense, but seems to be our actual practice. It
is clear that hearers often respond in the relevant way to likelihood claims.
It is also clear that speakers adjust their utterances in ways relevant to this
function, expressing their likelihood judgments when epistemic guidance
is needed and withholding or making explicitly subjective their likelihood
claims when they have specific reasons to think that others have more infor-
mation about the case at hand (“He is likely to be in the orchard, I think,
but Claudia knows more”).
Assume, as seems overwhelmingly reasonable, that this is the characteris-
tic communicative function of likelihood claims. Given accept/reject, to
accept the claim that X is likely or reject the claim that X is not likely is then
to judge that X is likely; to accept the claim that X is not likely or reject the
claim that X is likely is to judge that it is not likely. Given agree/disagree,
disagreement follows for likelihood claims.42
It seems, then, that agree/disagree can account for the relevant phe-
nomena both in ordinary descriptive discourse and in the case of tasti-
ness and likelihood claims, given independently plausible ascriptions of
characteristic communicative functions to these claims. This adds to the
antecedent plausibility of the postulated connection between agreement,
disagreement, and communicative function.
7.5 UNDERSTANDING correctness , independence,

AND no relativization
Turn now from agreement and disagreement to the sort of attributions of

correctness and incorrectness (and cognates truth and falsehood, rightness and
wrongness) that are directly involved in the correctness, independence,
and no relativization aspects of unity. As in the previous section, the
task is to not only predict but also make sense of patterns of such attribu-
tions across domains of discourse. The key here, as in the previous section,
will be an understanding of the characteristic function of the claims and
judgments involved.
function (cf. Lennertz (2014)). See also Montminy’s (2012) suggestion that claims of the
form it might be that p have as their main communicative function to “weakly suggest” p,
and that acceptance, rejection, and assessments targets this “suggestive”.
42
More generally, the suggestion that the function of epistemic modals is to guide
likelihood judgments setting subjective probabilities within an epistemic situation, seems
to explain many of the problem cases that Von Fintel and Gillies (2008) raise against
assessor relativist accounts of epistemic modals.
The account to be developed builds on the notion of a fundamental

standard for judgments:
fundamental standard: S is a fundamental standard for a judgment,
J, if and only if there is evidential support for J to the extent that, and
because, there is evidence that J satisfies S.
For illustration, the fundamental standard for ordinary judgments about
non-relative features of the world is presumably that they represent the
world as it is, or correspond to the facts (leaving open how representa-
tion and correspondence are best understood). So, for example, there is
evidential support for the judgment that Hobbes wrote Leviathan to the
extent that, and because, there is evidence that the world is as represented
by this judgment. Moreover, treating correspondence as the fundamental
standard makes sense given that these judgments serve as our “map” of
an intersubjectively accessible world, a map variously employed in theo-
retical and practical cognition to serve our many concerns and interests.
Normally, our successful use of this map will rely on a systematic cor-
respondence between its constituting judgments and the relevant aspects
of the world.
Using the notion of a fundamental standard, I suggest that the rele-
vant attributions of correctness and incorrectness (and cognates) work as
follows:
correct/incorrect: We find a judgment correct, right or true (incor-
rect, wrong or false) to the extent that we take that judgment (its con-
trary) to conform to the fundamental standard for such judgments.
We find a claim correct, right or true (incorrect, wrong or false) to
the extent that we take the sort of judgment constituting acceptance
(rejection) of the claim to conform to the fundamental standard for
such judgments.43
Assume that we take correspondence to be the fundamental standard for
judgments communicated in ordinary, non-relative descriptive discourse.
Given correct/incorrect, perceived correspondence will then guide
attributions of correctness and incorrectness to such judgments, as well as
to claims the function of which is to produce (or make occurrent) such
judgments. correctness follows straightforwardly: if F is a non-relative
43
Notice that correct/incorrect provides no account of truth, falsehood, etc., only
an account of our attributions of such properties. It should be compatible with a variety
of substantive theories of truth, including minimalist theories, but it might relate in most
obvious ways to theories building on the idea that truth is the goal of rational inquiry. For
ways of building theories of truth on that idea and related platitudes, see Lynch (2009)
and Wright (1992).
descriptive predicate and A claims that X is F while B claims that X is not

F, then if one claim is correct (true, right) the other must be incorrect (false,
wrong). Likewise for independence and no relativization.
What I will argue now is that there are plausible stories to tell about
what we should expect the operative fundamental standards to be not only
for paradigmatically absolutist judgments, but also for tastiness and like-
lihood judgments. Given these stories, correct/incorrect has enough
substance to let us explain why correctness, independence, and no
relativization obtain to varying degrees for these different kinds of
judgments.
Start with tastiness claims, focusing on the positive variety (“X is
tasty,” as opposed to “X is not tasty”). The characteristic communica-
tive function of such claims, I have suggested, is not to bring about
judgments that some non-relative state of affairs obtains, but to provide
(fallible) gustatory guidance by making hearers expect that their own
palates would react favorably to the item in question. Correspondingly,
the fundamental standards for accepting or rejecting positive tastiness
claims are standards that determine what we take to support or under-
mine such expectations. Clearly, we do not in general take there to be
any such standards beyond the palates of the individual judges, func-
tioning under suitably normal circumstances. Given this, correct/
incorrect explains why attributions of correctness or incorrectness
to tastiness claims seem off when there is reason to think that par-
ties of the conversation will have different taste reactions to the item
in question: there is no one relevant standard to appeal to in the dis-
course. If there is a standard to apply here, it is judge relative, leading
to violations of correctness, independence, and no relativization.
But correct/incorrect also explains why assessments of correctness
and incorrectness seem possible in contexts where it is assumed that
all involved have matching taste reactions with respect to the relevant
foods: in such contexts, there is a common standard. Together with an
independently plausible account of the function of tastiness clams, cor-
rect/incorrect can thus explain both why correctness is problem-
atic for tastiness and why attributions of correctness and incorrectness
are nevertheless possible in some contexts.
Next consider likelihood claims. Their communicative function, I have
suggested, is to produce likelihood judgments that set the same subjective
probabilities as the judgments expressed by the claims. By correct/incor-
rect, the relevant standards for assessments of correctness, incorrectness,
etc., would thus be standards for these likelihood judgments. What stand-
ards are these? In general, we think that likelihood judgments governing
subjective probabilities should be reasonable given the evidence, and be
based on more relevant evidence rather than less. This explains why we
might take a likelihood claim to be incorrect if we can support rejection of it
with reference to evidence going beyond what is accessible to the speaker, as
Claudia does in the case of finding felix. We thus have a straightforward
explanation of why likelihood claims can conform to correctness even
when absolutism is not satisfied for the judgments involved: the funda-
mental standards are insensitive to the information accessible to particu-
lar speakers. For the same reason, independence might seem to hold: any
evidence that hearers have access to, whether available to the speaker or
not, might be relevant for criticizing or supporting acceptance or rejection
of the claim. Likewise for no relativization: To determine which likeli-
hood judgments are supported relative to the evidence available to different
judges is not yet to resolve any disagreement about whether Felix is likely
to be in the orchard. This can only be done by determining which evidence
is best.
What, though, of our willingness, in hindsight, to deem correct Basil’s
claim that Felix was likely to be in the orchard, and our reluctance to say
that it was incorrect? If a likelihood claim should be rejected if our best
evidence supports the contrary likelihood judgment, why do we not just
reject Basil’s claim? The explanation for this, I suggest, is the same as for
our difficulty to judge, in hindsight, whether Felix was likely to be in the
orchard. Since the function of ordinary, unrelativized, likelihood judg-
ments is to guide our subjective probabilities for a proposition, they have
no place once we take the truth-value of that proposition for granted.
Relatedly, likelihood claims having as their characteristic communicative
function to produce likelihood judgments have no such function in rela-
tion to hearers who already know the truth-value of that proposition. The
correctness of such claims must thus be assessed in relation to some other
epistemic perspective than that of hindsight knowledge. Since the most
salient such perspective when considering the correctness of Basil’s claim
will be Basil’s own, we are naturally led to think about whether accepting
the claim would satisfy the standards relative to the information available
in that perspective.44 Of course, the same holds, mutatis mutandis, for
our assessments of Claudia’s claim that Felix was not likely to be in the
orchard. So correctness is violated, and, by extension, so are independ-
ence and no relativization.
44
Cf. Björnsson and Almér (2010); Björnsson and Finlay (2010: 23). As noted for
other epistemic modals, related phenomena are displayed within a conversation when it
is common knowledge that one of the parties has more information but will not share it
(Von Fintel and Gillies (2008: 90); Björnsson and Almér (2010: 31–2)).
7.6 PROSPECTS FOR A NON -absolutist ACCOUNT

OF MORAL unity
Thus far I have proposed ways of understanding disagreement and attribu-

tions of correctness that are independently motivated and seem to capture
the extent to which unity is satisfied in paradigmatically absolute domains
as well as in the case of tastiness and likelihood judgments. Since space pre-
vents a fuller defense of these proposals, my objective here has been more
modest: to present and provide initial motivation of a principled ground
for assessing non-absolutist explanations of unity in the moral domain.
Suppose that agree/disagree and correct/incorrect are indeed on
the right track. Then a non-absolutist explanation of moral unity requires
two things. First, it requires that moral claims have a characteristic com-
municative function such that judgments constituting acceptance of these
claims can differ in their internal success conditions. If they do, people’s
agreement and disagreement on moral issues do not depend on absolutism.
Second, it requires that people take fundamental standards for moral judg-
ments to have consistent content and apply equally across moral judgments,
independent of variations in the internal success conditions, and indepen-
dently of assessors’ own relations to the judges. If people do, this would be
sufficient, independently of absolutism, for people to attribute correctness
and incorrectness along the pattern of correctness, while taking inde-
pendence and no relativization to be satisfied.
Both requirements seem well within the realm of the possible. Consider
the first. Although few defend stronger forms of motivational internalism, it
should be much less controversial that moral claims have as their character-
istic communicative function to bring about or make occurrent judgments
with a certain practical role. For example, it seems plausible that claims of
moral wrongness have as their characteristic communicative function to
bring about or make occurrent judgments that dispose judges to negative
social emotions towards the agent. Likewise, it seems reasonably plausible
that claims of what ought to be done under certain circumstances, all things
considered, have as their characteristic function to bring about or make
occurrent plans to act accordingly. Neither of these functional hypotheses
requires any strong form of motivational internalism. What is required is
merely that we have default expectations that these claims are made in order
that hearers make judgments with certain practical roles, and that these
expectations are embodied in our default patterns of acceptance and rejec-
tion.45 Moreover, it seems at least possible that moral claims have these
45
This assumption is in line with recent empirical studies of internalist intuitions
practical functions even though internal success conditions for moral judg-
ments vary among moral judges. The lack of agreement—extensional or
intensional—on substantial characterizations of moral properties provides
prima facie evidence that internal success conditions do in fact vary, and it
seems undeniable that we need coordination of attitudes between people
not antecedently committed to the same practical standards.
The second requirement seems equally surmountable. Non-absolutism
itself says nothing about how we conceive of the fundamental standards
against which we assess moral judgments, and forms of relativism and
non-cognitivism seem compatible in principle with the assumption that
we conceive of such standards as invariant across judges and judgments.
In fact, if moral discourse and moral thinking are shaped to support the
coordination of attitudes through the coordination of moral judgments,
as non-absolutists have often suggested,46 such invariance in standards for
judgments might be an important regulative ideal.
At least at a first glance, then, agree/disagree and correct/incor-
rect seem to leave room for non-absolutist explanations of moral unity.
Interestingly, they also suggest reasons why the moral domain might not
display unity to the fullest extent. One possibility, for example, is that peo-
ple understand the characteristic function of moral claims as restricted to an
audience within the speaker’s society, roughly in the way that the function of
likelihood claims might be restricted to an audience sharing the same epis-
temic predicament. If people do, they should find disagreement problematic
when considering the moral claims and judgments of people in other societies.
Another possibility, suggested by the case of hindsight, is that some people
see fundamental standards as relative in various ways, or apply different stand-
ards depending on whose judgment is being assessed, perhaps depending on
the society in which it is made. People who do might reject correctness
and independence while thinking that disagreement holds.47 Agree/disa-
gree and correct/incorrect thus provide tools given which unity can be
systematically explained in non-absolutist discourse to the extent that it is
actually there.
Of course, to say that some form of non-absolutism can explain unity in
the moral domain is not to say that there is one that actually does. For all
among lay people (see Björnsson et al. (2014)) but should be compatible even with
strong externalist views, such as those of Strandberg (2012) and Svavarsdóttir (1999,
2006). Practical roles might include not only being for or against something, but
also states of preferential neutrality, either all-considerations-considered, or all-moral-
considerations-considered (see Dreier (2006, 2009); Silk (forthcoming)).
46
See e.g., Björnsson and McPherson (2014); Blackburn (1993b); Gibbard (1990).
47
The folk intuitions canvassed by Sarkissian et al. (2011) might reveal such an
understanding.
I have said, absolutism might be true, and both relativist and non-cognitivist
accounts might ultimately succumb to objections other than the argument
from unity. If what I have said here is correct, however, a systematic, inde-
pendently motivated and non-ad hoc account of moral unity is in the off-
ing. This is an interesting enough prospect.48
References
Almér, A. and Björnsson, G. 2009. “Contextualism, Assessor Relativism, and
Insensitive Assessments,” Logique et Analyse 52: 363–72.
Björklund, F., Björnsson, G., Eriksson, J., Olinder, R. F., and Strandberg, C. 2012.
“Recent Work on Motivational Internalism,” Analysis 72: 124–37.
Björnsson, G. 2001. “Why Emotivists Love Inconsistency,” Philosophical Studies
104: 81–108.
Björnsson, G. 2012. “Do ‘Objectivist’ Features of Moral Discourse and Thinking
Support Moral Objectivism?” Journal of Ethics 16: 367–93.
Björnsson, G. 2013. “Quasi-Realism, Absolutism, and Judgment-Internal
Correctness Conditions,” in C. Svennerlind, J. Almäng, and R. Ingthorsson
(eds.), Johanssonian Investigations: Essays in Honour of Ingvar Johansson on His
Seventieth Birthday. Heusenstamm: Ontos, 96–119.
Björnsson, G. n.d. “Discursivism About Disagreement in Attitude,” manuscript.
Björnsson, G. and Almér, A. 2010. “The Pragmatics of Insensitive
Assessments: Understanding the Relativity of Assessments of Judgments of
Personal Taste, Epistemic Modals, and More,” in B. H. Partee, M. Glanzberg,
and J. Skilters (eds.), Formal Semantics and Pragmatics: Discourse, Context and
Models. Manhattan, KS: New Prairie Press, 1–45.
Björnsson, G. and Finlay, S. 2010. “Metaethical Contextualism Defended,” Ethics
121: 7–36.
Björnsson, G. and McPherson, T. 2014. “Moral Attitudes for
Non-Cognitivists: Solving the Specification Problem,” Mind 123: 1–38.
Björnsson, G., Eriksson, J., Strandberg, C., Olinder, R. F., and Björklund, F. 2014.
“Motivational Internalism and Folk Intuitions,” Philosophical Psychology. doi: 10.
1080/09515089.2014.894431.
Blackburn, S. 1984. Spreading the Word. Oxford: Oxford University Press.
Blackburn, S. 1991a. “Just Causes,” Philosophical Studies 61: 3–17.
48
I am grateful for feedback from audiences at the University of Connecticut,
University of Gothenburg, University of Edinburgh, Lund University, Umeå University,
the 9th Wisconsin Metaethics Workshop, and the Values in Context Workshop at the
University of Lisbon. Special thanks to Tristram McPherson, Ben Lennertz, Robin
McKenna, Jussi Suikkanen, Caj Strandberg, Ragnar Francén Olinder, John Eriksson, and
Christian Munthe for extensive comments on earlier versions, and to two anonymous
referees. Work on this essay has been supported by the Swedish Research Council, grants
2009-1507 and 2012-988.
Blackburn, S. 1991b. “Reply to Sturgeon’s ‘Contents and Causes: A Reply to

Blackburn’,” Philosophical Studies 61: 39–42.
Blackburn, S. 1993a. Essays in Quasi-Realism. Oxford: Oxford University Press.
Blackburn, S. 1993b. “How to Be an Ethical Anti-Realist,” in Blackburn, Essays in
Quasi-Realism. Oxford: Oxford University Press, 166–81.
Brink, D. O. 1989. Moral Realism and the Foundations of Ethics.
Buekens, F. 2011. “Faultless Disagreement, Assertions and the Affective-Expressive
Dimension of Judgments of Taste,” Philosophia 39: 637–55.
Doerfler, R. 2012. “A Comedy of Errors or, How I Learned to Stop Worrying and
Love Sensibility-Invariantism About ‘Funny’,” Pacific Philosophical Quarterly
93: 493–522.
Dowell, J. 2011. “A Flexibly Contextualist Account of Epistemic Modals,”
Philosopher’s Imprint 11: 1–25.
Dreier, J. 2006. “Negation for Expressivists: A Collection of Problems with a
Suggestion for Their Solution,” Oxford Studies in Metaethics 1: 217–33.
Dreier, J. 2009. “Relativism (and Expressivism) and the Problem of Disagreement,”
Egan, A. 2007. “Epistemic Modals, Relativism and Assertion,” Philosophical Studies
133: 1–22.
Egan, A. 2010. “Disputing About Taste,” in T. Warfield and R. Feldman (eds.),
Disagreement. Oxford: Oxford University Press, 247–86.
Finlay, S. 2014. Confusion of Tongues: A Theory of Normative Language.
Gibbard, A. 1990. Wise Choices, Apt Feelings. Cambridge, MA: Harvard
University Press.
Glanzberg, M. 2007. “Context, Content, and Relativism,” Philosophical Studies
136: 1–29.
Goodwin, G. P. and Darley, J. M. 2008. “The Psychology of Meta-Ethics: Exploring
Objectivism,” Cognition 106: 1339–66.
Hare, R. M. 1970. “Meaning and Speech Acts,” Philosophical Review 79: 3–24.
Huemer, M. 2005. Ethical Intuitionism. Basingstoke: Palgrave Macmillan.
Knobe, J. and Yalcin, S. Forthcoming. “Epistemic Modals and Context: Experimental
Data,” Semantics and Pragmatics.
Kölbel, M. 2004. “Faultless Disagreement,” Proceedings of the Aristotelian Society
104: 53–73.
Lasersohn, P. 2005. “Context Dependence, Disagreement, and Predicates of
Personal Taste,” Linguistics and Philosophy 28: 643–86.
Lennertz, B. 2014. “Taking ‘Might’-Communication Seriously,” Analytic Philosophy
55: 176–98.
Loeb, D. 2003. “Gastronomic Realism—a Cautionary Tale,” Journal of Theoretical
and Philosophical Psychology 23: 30–49.
Lynch, M. P. 2009. Truth as One and Many. Oxford: Clarendon Press.
Lyons, D. 1976. “Ethical Relativism and the Problem of Incoherence,” Ethics
86: 107–21.
MacFarlane, J. 2011. “Epistemic Modals Are Assessment-Sensitive,” in A. Egan

and B. Weatherson (eds.), Epistemic Modality. Oxford: Oxford University Press,
144–78.
McKenna, R. 2014. “Disagreeing About ‘Ought’,” Ethics 124: 589–97.
McNaughton, D. 1988. Moral Vision: An Introduction to Ethics. Oxford: Blackwell.
Millikan, R. G. 1984. Language, Thought, and Other Biological Categories: New
Foundations for Realism. Cambridge, MA: MIT Press.
Millikan, R. G. 1998. “Language Conventions Made Simple,” Journal of Philosophy
95: 161–80.
Montminy, M. 2012. “Epistemic Modals and Indirect Weak Suggestives,” Dialectica
66: 583–606.
Pearson, H. 2013. “A Judge-Free Semantics for Predicates of Personal Taste,” Journal
of Semantics 30: 103–54.
Plunkett, D. and Sundell, T. 2013. “Disagreement and the Semantics of Normative
and Evaluative Terms,” Philosopher’s Imprint 13: 1–37.
Ridge, M. 2013. “Disagreement,” Philosophy and Phenomenological Research
86: 41–63.
Sarkissian, H., Park, J., Tien, D., Wright, J., and Knobe, J. 2011. “Folk Moral
Relativism,” Mind and Language 26: 482–505.
Sayre-McCord, G. 2006. “Moral Realism,” in D. Copp (ed.), The Oxford Handbook
of Ethical Theory. Oxford: Oxford University Press, 39–62.
Schaffer, J. 2011. “Perspective in Taste Predicates and Epistemic Modals,” in A.
Egan and B. Weatherson (eds.), Epistemic Modality. Oxford: Oxford University
Press, 179–226.
Silk, A. Forthcoming. “How to Be an Ethical Expressivist,” Philosophy and
Phenomenological Research.
Stalnaker, R. 1999. Context and Content: Essays on Intentionality in Speech and
Thought. Oxford: Oxford University Press.
Stalnaker, R. 2002. “Common Ground,” Linguistics and Philosophy 25: 701–21.
Stephenson, T. 2007. “Judge Dependence, Epistemic Modals, and Predicates of
Personal Taste,” Linguistics and Philosophy 30: 487–525.
Stevenson, C. L. 1944. Ethics and Language. New Haven: Yale University Press.
Stevenson, C. L. 1963. Facts and Values: Studies in Ethical Analysis. New Haven: Yale
University Press.
Strandberg, C. 2012. “A Dual Aspect Account of Moral Language,” Philosophy and
Street, S. 2006. “A Darwinian Dilemma for Realist Theories of Value,” Philosophical
Studies 127: 109–66.
Streiffer, R. 2003. Moral Relativism and Reasons for Action. London and
New York: Routledge.
Sturgeon, N. L. 1991. “Contents and Causes: A Reply to Blackburn,” Philosophical
Studies 61: 19–37.
Sundell, T. 2011. “Disagreements about Taste,” Philosophical Studies 155: 267–88.
Svavarsdóttir, S. 1999. “Moral Cognitivism and Motivation,” Philosophical Review

108: 161–219.
Svavarsdóttir, S. 2006. “How Do Moral Judgments Motivate?” in J. Dreier (ed.),
Contemporary Debates in Moral Theory. Oxford: Blackwell, 163–81.
Swanson, E. 2011. “How Not to Theorize About the Language of Subjective
Uncertainty,” in A. Egan and B. Weatherson (eds.), Epistemic Modality.
Tersman, F. 2006. Moral Disagreement. Cambridge: Cambridge University Press.
Von Fintel, K. and Gillies, A. S. 2008. “CIA Leaks,” Philosophical Review 117: 77–98.
Willer, M. 2013. “Dynamics of Epistemic Modality,” Philosophical Review
122: 45–92.
Wright, C. 1992. Truth and Objectivity. Cambridge, MA: Harvard University Press.
Wright, J. C., Grandjean, P. T., and McWhite, C. B. 2012. “The Meta-Ethical
Grounding of Our Moral Beliefs: Evidence for Meta-Ethical Pluralism,”
Philosophical Psychology 26: 336–61.
Wright, J. C., McWhite, C. B., and Grandjean, P. T. 2014. “The Cognitive
Mechanisms of Intolerance: Do Our Meta-Ethical Commitments Matter?” in
T. Lombrozo, S. Nichols, and J. Knobe (eds.), Oxford Studies in Experimental
Philosophy, Volume 1. Oxford: Oxford University Press, 28–61.
Young, L., Nichols, S., and Saxe, R. 2010. “Investigating the Neural and Cognitive
Basis of Moral Luck: It’s Not What You Do but What You Know,” Review of
Philosophy and Psychology 1: 333–49.
8
Grounding the Autonomy of Ethics
Barry Maguire
8.1 INTRODUCTION
In their disputes with nihilists, subjectivists, revisionary expressivists, and other

disreputable characters, metaethical realists and quasi-realists often employ
arguments that appeal, perhaps implicitly, to certain theses about the autonomy
or isolation or distinctness of the ethical domain. T. M. Scanlon, G. A. Cohen,
Derek Parfit, Simon Blackburn, and Ronald Dworkin have all employed argu-
ments of this sort. Much is at stake. Scanlon and Parfit assume that the defence
of non-naturalism depends upon a proper understanding of autonomy. Cohen
argues that the debate between liberals and socialists turns on an issue closely
related to autonomy. Dworkin argues, from a premise about the autonomy of
the ethical domain, that ethical nihilism is not merely false, but incoherent.1
Oddly, these theorists rarely pause to engage directly with the literature
on ethical autonomy—beyond a passing reference to A. N. Prior’s famous
paper. Were they to do so, they would find, for the most part, theorists
vigorously debating various logical autonomy theses.2 Such theses maintain
that certain logical relations do not obtain between ethical and non-ethical
sentences, for instance that no non-ethical sentences logically entail an ethi-
cal sentence.
I think this focus on logical autonomy is a mistake.3 The thesis so impor-
tant to our metaethicists is not a logical thesis but a metaphysical one.
1
Cf. Scanlon’s discussion of A. N. Prior and Mark Schroeder in his (2014); Cohen’s
argument against Rawlsian constructivism in his (2008); Parfit’s argument against Frank
Jackson’s reductionism in his (2011); Blackburn’s arguments against the charge of revi-
sionism in his (2006) and elsewhere; Dworkin’s (1996) and (2012).
2
For example: all of the twenty-seven papers in the most recent volume on ethical
autonomy (Pigden (2010)) defend or oppose some logical autonomy thesis.
3
Moreover it is a mistake encouraged by a misreading of Hume. In the famous passage
in the Treatise, he insists: ‘ … that a reason should be given; for what seems altogether
Grounding the Autonomy of Ethics 189
Roughly, the relevant metaphysical autonomy thesis maintains that ethical

facts are not fully grounded in non-ethical facts.
I motivate this shift from logical to metaphysical autonomy in two main
ways. Firstly I show that the specific metaphysical characterization that I
favour both avoids and also helps to explain the success of the counterexam-
ples to logical autonomy theses. Secondly I show that this characterization
has fruitful upshots in ethics, metaethics, and ethical methodology.
We can distinguish two different goals that one might have when defending
an autonomy thesis. One goal is to defend some characterization of autonomy
as a characterization of autonomy, i.e., as an interesting and plausible articula-
tion of the thesis that theorists are, or are often, or perhaps should be, arguing
about when they are arguing about autonomy. This is what we are up to here.
Another goal is to argue that this characterization is true. I’m strongly inclined
to believe that ethical autonomy on this metaphysical characterization is true.
However I will not defend this here. Most of the objections to the charac-
terization of autonomy that I will consider are objections to the characteriza-
tion as a characterization, rather than objections that assume that it is a good
characterization of autonomy, but insist that it is false. In fact, since some
challenges of the latter sort presuppose something like the characterization
I will defend, they provide further indirect support for this characterization.
This chapter has five main sections. All autonomy theses, just like all
nihilistic theses, make certain background assumptions about which facts
and propositions count as ethical—that is, they presuppose certain taxo-
nomic theses. We begin in section 8.2 with some preliminaries about the
relationship between taxonomy and autonomy. In section 8.3 we discuss
logical autonomy theses and a host of problematic counterexamples. In
section 8.4 we introduce our metaphysical autonomy theses, spend some
time on some details, and present some of their advantages. In section 8.5
we consider some apposite challenges to our metaphysical autonomy the-
ses, using these challenges to draw out various further implications of this
conception of ethical autonomy. In section 8.6 we discuss various implica-
tions of this metaphysical conception of autonomy for some of the debates
mentioned above. In particular, we gain insight into the the ethical domain
itself, which appears to be (at least partly) structured by various grounding
relations. Moreover the account of autonomy I propose would provide us
inconceivable, how this new relation can be a deduction from others, which are entirely
different from it’. Hume’s use of the word ‘deduction’ suggests to our ears a logical rela-
tion. However, Hume did not use the word ‘deduction’ to pick out a specifically logical
relation, and nor is it plausible from the context, or the general project of the Treatise,
that Hume had a specifically logical thesis in mind. Thanks to Don Garrett for a helpful
discussion of this point.
190 Barry Maguire
with an expansive conception of the ethical, and correspondingly minimal

conception of the ‘metaethical’—insofar as this refers to a domain that does
not overlap with the ethical. This lends support to the idea that many puta-
tively ‘metaethical’ disputes—e.g., about the explanatory direction between
different kinds of normative facts—are simply abstract ethical disputes. In
turn this lends support to various kinds of opposition to revisionary meta-
ethical theses.
8.2 PRELIMINARIES: AUTONOMY AND ETHICAL

TAXONOMY
Autonomy theses maintain that there is some specific relation such

that facts or propositions from two different categories do not stand in
that relation to each other. More strongly, they maintain that there is
some relation such that facts or propositions from some category do not
stand in that relation to facts or propositions from any other category.
An autonomy thesis requires some specification of the relation in ques-
tion, for instance whether classical entailment, metaphysical grounding,
epistemic justification, or whatever, and some minimal account of the
categories with which it is concerned.
We have these two tasks in the theory of ethical autonomy. One concerns
closure: figuring out which operations and relations preserve ethicality. This
is our topic, and we will turn to it shortly. Another is to distinguish ethical
from non-ethical facts or propositions by distinguishing ethical from non-
ethical constituents of facts and propositions. We don’t need to take much
of a stand on this issue, but a few words will make things clearer as we
proceed.
There are various different kinds of ethical constituent. We won’t be
concerned much with differences between normative (e.g. ‘ought’, ‘imper-
missible’) and evaluative notions (e.g. ‘good’, and ‘bad’), or between thin
(‘right’, ‘required’) and thick notions (‘dumpy’, ‘graceful’). It is not easy
to draw the line between thin and thick notions or between thick notions
and non-evaluative notions. These difficulties blur the distinction between
the ethical and non-ethical in one sense. For today, this is somebody else’s
problem.
One simple method for distinguishing ethical and non-ethical facts
is unavailable. Not all facts with ethical constituents are ethical facts.
Foot-binding was considered obligatory for court dancers during the
Song Dynasty. This fact has an ethical constituent but it is not an ethi-
cal fact. It is a fact about what was considered to be ethical at the time.
However, clearly some facts with ethical constituents are ethical facts. The least
controversial often consist in some act-token or state of affairs instantiating
some ethical property or falling under some ethical operator, such as <Ariel
Castro’s treatment of the girls was impermissible> or <you should donate 10%
of your pre-tax income to charity>. Since we are not directly concerned with
this part of ethical taxonomy, we’ll help ourselves to the ethicality of such
uncontroversially ethical facts and propositions.
A final preliminary. Logical and metaphysical autonomy theses most
likely concern different categories of relata. Logical connectives relate sen-
tences or statements, grounding relates objects or facts or perhaps proposi-
tions. It will be convenient to pick something such that it is plausible that
that thing can stand in both logical and metaphysical relations. To save
words, I’ll default to talk about ethical and non-ethical propositions.
8.3 LOGICAL AUTONOMY THESES AND SOME

COUNTEREXAMPLES
Almost all of the literature directly concerned with ethical autonomy focuses
on the attempt to defend or reject different logical autonomy theses.4
Start with simple logical autonomy, the thesis that no non-ethical
propositions entail an ethical proposition. Take any non-ethical
proposition—that Susan Rumplebottom won the 2009 Gloucestershire
Cheese Rolling competition—and call it R. Let M be the ethical proposi-
tion that Castro’s treatment of the girls was impermissible. (R & ~R) entails
M, our ethical proposition. So simple logical autonomy is false.
A more interesting counterexample to simple logical autonomy is due
to Arthur Prior (1960). He argued as follows:
1. R entails (R ∨ M)
2. ((R ∨ M) & ~R) entails M
3. Either (R ∨ M) is ethical or not.
4. If so, then 1 constitutes a counterexample to simple logical autonomy.
5. If not, then 2 constitutes a counterexample to simple logical autonomy.
6. Therefore simple logical autonomy is false.
Premises 1 and 2 are incontestable. If (R ∨ M) is ethical, premise 4 follows
directly from 1. If (R ∨ M) is not ethical, then (so long as we assume that a
4
I stick with classical logic throughout. For an interesting discussion of relevance
logic in connection with some of our topics, see Samuel (n.d.).
192 Barry Maguire
conjunction with non-ethical conjuncts is non-ethical) premise 5 also fol-

lows. Hence it seems we should reject simple logical autonomy.5
In response to this argument, various theorists opt to modify the auton-
omy thesis. Charles Pigden (1989, 2010) has argued that we get a problem
with explosion and disjunctive syllogism because in such cases the ethical
expressions in the conclusions are contingently vacuous.6 (This is easiest to
see with premise 1 in the Prior argument as presented above.) Pigden then
defines an alternative principle, namely non-vacuous logical auton-
omy: no non-ethical propositions entail a non-vacuous ethical proposition.7
M in premise 1 of Prior’s argument occurs vacuously, since R entails the dis-
junction of R and any proposition you like. Assuming R is true, the disjunc-
tion will be true. Similarly with explosion. (R & ~R) entails any proposition
whatsoever. Hence neither constitutes a counterexample to non-vacuous
logical autonomy.
There are two problems with Pigden’s alternative principle. Firstly, as
Pigden freely admits, it does not capture anything interesting about ethics.
Non-vacuous logical autonomy ‘is commonplace and not confined to eth-
ics’ (1989: 7). Propositions about hedgehogs are also autonomous in this
same sense. No non-hedgehog-involving proposition non-vacuously entails
any hedgehog involving proposition. Pigden says that ‘logical autonomy is,
5
In subsequent work Lloyd Humberstone (1982, 1995) and David Lewis (1988)
have shown that if one accepts certain logical taxonomic principles, one ends up commit-
ted to the view that all propositions are in every category. The argument is as follows. Let
C be a putatively isolated subclass of propositions. S is any proposition. Now suppose:
equivalence: if S and S* are classically equivalent, and S is in C, then S* is in C.
converse-implication: if S entails S* and S* is in C then S is in C.
disjunction: if S1, …,Sn are in C, then the disjunction of S1, …,Sn with any other
proposition is also in C.
negation: if S is in C, then not-S is in C.
Now the proof. Suppose that S is in C, and that S* is some arbitrary proposition. By
negation, not-S is in C. Hence by converse-implication, (S & S*) and (not-S & S*)
are in C. Then by disjunction ((S & S*) or (not-S & S*)) is in C. This is equivalent to
S*, so S* is in C. Since S* is arbitrary, it follows that all propositions are in C.
If one were to accept all these taxonomic principles for the ethical domain, then if any
proposition were ethical, all propositions would be ethical. It would follow that both
ethical autonomy and ethical nihilism were be trivially false. So any full ethical taxonomy
will need to reject at least one of the Humberstone/Lewis principles.
6
Here is Pigden’s account of contingent vacuity (for sentences): ‘An expression E is
contingently vacuous in the conclusion of a valid inference if the inference would remain
valid if E were replaced by any expression whatsoever of the same grammatical type’
(1989: 134).
7
It is unclear whether Pigden holds the view that ethicality is a relation a conclusion
bears to premises—and hence that no non-ethical propositions logically entail an ethical
proposition, or whether he thinks that some non-ethical propositions logically entail
an ethical proposition, but none do so non-vacuously. The former is problematic, for
reasons that will emerge later.
in itself, rather trivial’ (1989: 21). But this gives us some reason to doubt
that we have finally captured the important thesis that so many philoso-
phers have taken Hume to have brought to our attention.
More pertinently, non-vacuous logical autonomy faces several poten-
tial counterexamples. There are valid arguments from seemingly non-ethical
premises to seemingly ethical conclusions.
Consider first the following argument, due to Toomas Karmo (1988: 253):
1. Everything that Alfie says is true.
2. Alfie says that it is impermissible to starve the Irish.
3. Therefore it is impermissible to starve the Irish.
The conclusion seems ethical. The second premise is non-ethical. We’ll
shortly discuss whether the first premise is ethical. But for now it is worth
noting that here we clearly have a valid argument with arguably non-ethical
premises and a non-vacuous ethical conclusion.8
Now consider the following argument from Stephen Maitzen (2010: 293):
1. At least one ethical proposition is true.
2. If at least one ethical proposition is true, then torturing innocent chil-
dren is impermissible.
3. Therefore torturing innocent children is impermissible.
Again clearly the conclusion is ethical and the argument is valid. Maitzen
argues that the two premises are both non-ethical. The point for now is
that the move from simple logical autonomy to non-vacuous logical
autonomy doesn’t help with this putative counterexample. Moreover the
reason why not is fairly clear. The premises intuitively and non-vacuously
guarantee the conclusion. There is nothing logically problematic with these
arguments. What is ‘wrong’ is that their premises don’t explain their conclu-
sions. These are promising counterexamples to logical autonomy theses, but
they don’t get a foothold on metaphysical autonomy theses—or so I’ll argue.
8.4 METAPHYSICAL AUTONOMY THESES
I’ll now suggest that there is a version of Hume’s principle which is plausi-
ble, and which not only avoids all these counterexamples, but also helps to
explain which counterexamples work and why.
8
T. M. Nelson presents a structurally similar argument in his (1995: 555).
194 Barry Maguire
We start with some remarks about the character of relations within the ethi-
cal domain. Here’s Ronald Dworkin (2012: 31):
… someone asking herself whether it would be wrong to leave an unhappy marriage
might reflect on more general issues about what people owe other people they have
asked to trust them, for instance, or about the moral responsibilities children bring.
One natural interpretation here is that facts about it being wrong to leave an
unhappy marriage would obtain in virtue of facts about what one owes to
someone who trusts one, and in virtue of facts about the moral responsibili-
ties children bring—together with various non-ethical facts about the circum-
stances of the marriage, the ages of the children, and so forth.
G. A. Cohen suggests a similar explanatory structure in ‘Facts and
Principles’:9
Suppose someone affirms the principle that we should keep our promises (call that P)
because only when promises are kept can promisees successfully pursue their projects (call that
F). Then she will surely agree that she believes that F supports P because she affirms …
that we should help people to pursue their projects.
The idea, construed as a metaphysical thesis, is that particular ethical facts
obtain in virtue of more general ethical facts together with pertinent non-ethical
facts. The ‘in virtue of’ or equivalently the ‘grounding’ relation is an explana-
tory relation between metaphysical entities. We can use this characterization
of the relations that distinguish the structure of relations between facts within
the ethical domain, to characterize the sense in which facts within the ethical
domain are autonomous. This gives us the following metaphysical characteri-
zation of the ethical autonomy:
metaphysical autonomy: No ethical fact is fully grounded just by
non-ethical facts.
metaphysical autonomy is a plausible characterization of the important
sense in which ethics is autonomous, if it is. Consider a few contentious is–
ought transitions:
1. The invasion of Iraq contravened international law, therefore the inva-
sion of Iraq was wrong.10
2. Individuals are motivated by personal gain, therefore the principles of
justice allow inequalities so long as these improve the positions of the
worst off.11
9
Cohen (2008: 234). We’ll ignore the epistemological gloss that Cohen and Dworkin
often give these claims. For more on epistemic or methodological autonomy and its rela-
tionship to metaphysical and logical autonomy, see my (forthcoming).
10
This is the motivating example in Campbell Brown’s (2014).
11
This is a mock-up of the case central to Cohen’s objection to Rawlsian constructivism.
3. Jones uttered the words ‘I hereby promise to pay you, Smith, five dol-
lars’, therefore Jones has an obligation to pay Smith five dollars.12
4. Ronnie wants to dance, therefore Ronnie has a reason to dance.13
metaphysical autonomy diagnoses the error that all four disputants would
intuitively be making, if ethics were indeed autonomous. Their premises
may partially ground their conclusions, but they do not fully ground their
conclusions. Moreover metaphysical autonomy offers some guidance: we
should look around to see whether any plausible principles underlie these
arguments, for instance some principle of the form ‘It is wrong to violate
international law,’ or, ‘One ought to keep one’s promises.’
For completeness I also propose a companion principle, namely:
converse metaphysical autonomy: Any fact partly grounded by an
ethical fact is an ethical fact.
The first principle rules out the metaphysical possibility of any view accord-
ing to which some ethical fact is fully grounded just by non-ethical facts.
This ensures that if ethical facts are grounded at all, they are grounded by
at least one ethical fact. Consequently there will be ethical facts of some
kind (presumably ethical principles) that ground other facts but that are
not themselves grounded. (It is common to say of facts with this profile
that they are fundamental.14) The second principle is more contentious. To
borrow a metaphor from Hartry Field, the idea is that ethical fluid flows
unceasingly upwards, from ethical grounds to whatever they ground or
partly ground.
But before I clarify some of the details in these theses, and present a more
general defence, let me discuss the notion of ‘ground’ in some more detail.
8.4.1 On the Grammar, Logic, and Nature of Ground

We can formulate grounding claims by means of an operator like ‘because’
that takes sentences to make a sentence. For instance we would say ‘The
triangle is equilateral because its three sides are the same length.’ As Kit
Fine has argued (2001, 2012) this operational approach has the advantage of
ontological neutrality. It does not commit to an ontology of facts, proposi-
tions, or relations. On an alternative predicational approach, ‘grounds’ is a
relational predicate that takes designators for facts or propositions to make
sentences. This is my preferred approach. For instance we would say that
12
This is a truncated version of Searle’s argument (1964).
13
Schroeder (2007). For critical discussion, see Scanlon (2014).
14
For instance see Bennett (2011).
196 Barry Maguire
‘The fact that the triangle is equilateral obtains in virtue of the fact that the
triangle’s three sides are all the same length.’ Or equivalently we can say ‘The
fact that the triangle’s three sides are all the same length grounds the fact
that the triangle is equilateral.’ The scrupulous are invited to translate what
follows into operational language.15
The next question concerns the category of the relata of the reasons rela-
tion. According to the intuitive notion, grounding is a metaphysical rela-
tion between worldly items rather between linguistic or representational or
theoretic entities. We assume that the relevant worldly items are facts or
true propositions rather than objects. We will remain agnostic about the
relations between objects instantiating or exemplifying properties, states
of affairs, facts, and true propositions.16 I use the word ‘fact’ in such a way
as to allow that principles are facts, for instance universal facts of the form
‘necessarily (for some specified kind of necessity) for any x, if x is F then x
is G’.17
We will assume that grounding is factive. Non-obtaining facts cannot
ground anything. False propositions cannot ground anything. However, it
will be useful to have a way of talking about what would have grounded
what if things had been different. For instance we can compare two epis-
temically possible situations, in one of which utilitarianism is true and
in the other of which some simple deontological theory is true, and ask
whether the fact that doing something would maximize net pleasure in the
world would together with the relevant ethical principles ground the fact
that you ought to do that thing. For at least some actions, these different
principles will yield different results. Or more simply we can change differ-
ent features of the consequences of the action itself. We will allow ourselves
to talk freely about such counterfactual grounding. However, we do not
allow inter-world grounding: a would-be fact in one possible world cannot
ground a would-be fact in another possible world.18 Statements about what
15
For instance, non-cognitivists articulating a conception of autonomy may prefer
to avoid the predicationalist approach. I’m borrowing this terminology from Fabrice
Correia (2010: 253). Compare the choice between operator and predicational approaches
in tense logic.
16
If you prefer to have propositions as the relata of the grounding relation you will
need to ensure that propositions are rather finely grained, for instance to allow that the
glass contains water obtains in virtue of its containing H2O. The propositionalist about
grounding may also have to heed the remarks about factivity to come in the main text;
see Fine (2012: 16).
17
Cohen assumes that principles are not facts. This is a terminological difference. He
also thinks that grounding is a relation between representational items of some kind.
We part ways on this. For an excellent characterization of principles as facts attributing
properties to kinds, see Knut Skarsaune (this volume).
18
Compare Fine (2012: 16).
grounds what are implicitly relativized to some world or set of worlds in

which the grounds and the grounded facts obtain.
I will assume that grounding is strongly irreflexive, strongly asymmetric,
and transitive. Neither A nor any list of facts including A grounds A. If A on
its own or with any other facts grounds B, then neither B nor any list of facts
including B grounds A.19 Transitivity is a little more controversial, given the
fact that the grounding relation is an explanatory relation. But transitivity
still seems plausible, and it will usefully allow us to appeal beyond immedi-
ate to mediate grounds,20 so we will assume it in what follows.
The most essential feature of the grounding relation is the fact that it
is an explanatory relation. When A grounds B, the fact that A obtains
explains the fact that B obtains. Since explanation in the relevant sense is
non-monotonic, grounding is non-monotonic.
8.4.2 On the Modal Status of Ground

The explanatory connection between grounds and what they ground is the
most distinctive feature of the grounding relation. It is common to add a
modal claim, that grounds necessitate what they ground.21 But there remain
different views about what kind of necessitation is involved.
Let’s start with a common way of speaking, which I think we should
reject. Some philosophers talk about natural facts ‘grounding’ ethical facts,
when they really mean that the relevant natural facts together with relevant
ethical principles ground the ethical facts. For instance in one context T. M.
Scanlon says that ‘the fact that a resort is pleasant is a reason to visit it …
and the fact that a discovery casts light on the causes of cancer is a reason
to applaud it and to support further research of that kind. These natural
properties provide a complete explanation of the reasons we have for reacting
in these ways’ (1998: 97; my italics). However when Scanlon is attending to
our question directly he insists that these natural facts only ground ethical
facts together with further ‘pure’ ethical facts. In Scanlon’s view these pure
ethical facts are not grounded by any non-ethical facts (see his 2014: 41 and
following). Scanlon’s considered view is that given that certain pure ethical
19
For challenges, see Ichikawa-Jenkins (2011) and Fine (2012).
20
For the mediate/immediate grounds distinction, see Fine (2012: 19).
21
This principle about ground is extremely widespread (e.g. see Rosen (2010) and
Fine (2012)). For a helpful discussion of different kinds of conditionality to which
grounding facts (facts of the form A grounds B) might be subject, see Bader (forth-
coming), especially the distinction between enablers and grounding principles. See also
Chudnoff (2013).
198 Barry Maguire
facts obtain, certain natural facts will ‘explain’ (in a more colloquial sense)
certain ‘mixed’ ethical facts.
This brings us to the word ‘fully’ in the metaphysical autonomy prin-
ciple. Clearly ethical facts are often partly grounded by non-ethical facts.
The fact that it is impertinent to slurp my noodles obtains partly in virtue
of the fact that I am having dinner with the Queen. But this non-ethical
fact only partly grounds the ethical fact. We additionally need some facts
about conventions (perhaps themselves also non-ethical facts) and some
fact about the impertinence of ignoring these conventions (an ethical prin-
ciple or more general ethical fact).22
There is a further question about the modal status of the ‘pure’ ethi-
cal facts, or what are more naturally thought of as ethical principles. The
necessity involved must be more than a mere contingent generalization.
We can follow Kit Fine in distinguishing normative necessity and met-
aphysical necessity.23 It is a theoretically open question whether, when
some ethical grounding facts obtain—for instance <the fact that x-ing
would maximize pleasure and that you ought to do whatever would maxi-
mize pleasure grounds the fact that you ought to x>— such grounding
facts obtain with normative or metaphysical necessity. This modal distinc-
tion provides us with more precise versions of our autonomy theses. We
have two main options: an autonomy thesis based on normative necessity
and an autonomy thesis based on metaphysical necessity, perhaps even
understood in terms of essences. I will leave both options on the table for
current purposes.
Let me close this discussion of modal issues with a remark about super-
venience, by which I mean the (metaphysical, rather than conceptual)
supervenience of mixed ethical facts on non-ethical facts.24 Suppose we
22
There is an interesting question about what distinguishes principles that are part of
the grounds, from principles that provide grounds for the fact that the grounds ground
what they ground. For instance what distinguishes the following two views. (i) The fact that
x-ing causes pain grounds the fact that it is wrong to x, and that fact, the grounding fact,
is grounded by the fact that it is always wrong to cause pain. (ii) The fact that x-ing causes
pain together with the fact that it is always wrong to cause pain together ground the fact
that x-ing is wrong. (See also Bader (forthcoming).) This difference won’t matter much to
us, since we assume that grounding is transitive, and our autonomy theses are not restricted
to immediate grounds. Notice that naturalistic analyses seem to play the role of principles
in some explanations. This suggests that either the principles expressing the naturalistic
analyses are themselves ethical principles (cf. Scanlon’s (2014: ch. 2) claim that desire-based
theories of reasons can be plausibly construed as substantive ethical principles) or else meta-
physical autonomy (though a good characterization of autonomy) is false.
23
Fine (2002). See also Rosen (n.d.).
24
Here I have in mind the so-called explanatory challenge posed by supervenience.
Cf. Blackburn (1971, 1985).
maintain quite generally that grounds necessitate what they ground. And
suppose we argue specifically that certain ethical principles obtain, either
with normative or metaphysical necessity. These principles are precisely in
the business of grounding mixed ethical facts when combined with perti-
nent non-ethical facts. These theses entail the supervenience of the mixed
facts on the non-ethical facts, with either normative or metaphysical neces-
sity. Consequently, on this (sensible) view, there is no puzzle about how to
explain supervenience.25
8.4.3 Extending to Other Autonomy Theses

Firstly we are in a position to state various more specific ethical autonomy
theses. For instance perhaps you are a naturalist about value, thinking that
facts about value are fully grounded in facts about well-being, which are
in turn fully grounded in social scientific facts. But you are not a natural-
ist about deontic facts, since you don’t think that deontic facts are fully
grounded in facts about value. Then you would think that value facts are
not autonomous but deontic facts are. Or what about the divine com-
mand theorist, who thinks that deontic facts are fully grounded in facts
about God’s will? Divine command theory is consistent with the thesis
that no deontic facts are fully grounded just by natural facts, but clearly
incompatible with the thesis that no deontic facts are fully grounded just
by non-ethical facts.
Now notice that structurally analogous principles would also apply to
other autonomy theses (in particular to some of those discussed by Gillian
Russell 2010). It is plausible that no universal fact is fully grounded just by
some particular facts. It is plausible that no necessity fact is fully grounded
just by some contingent facts. It is plausible that future fact is fully grounded
just by facts about the past.
As further support for metaphysical characterizations of autonomy
theses, over logical characterizations, note that all of these autonomy
theses would be trivially true if we were working with something like
non-vacuous logical autonomy. We would be prevented, by subopti-
mal characterizations of the relevant theses, from engaging with interesting
philosophical questions.
25
Hence the business of explaining supervenience is only problematic for hardcore
particularism. It does seem consistent with hardcore particularism that ethical proper-
ties could be just scattershot around unsystematically; which is odd. I say more about
particularism in section 8.5.4.
200 Barry Maguire
8.4.4 On the Counterexamples to Logical Autonomy

metaphysical autonomy also provides us with a simple response to all the
counterexamples to logical autonomy, since in those counterexamples
the premises do not purport to provide grounds for their conclusions.
This is clearly the case with the argument from explosion. (R & ~R)
entails but does not ground M.
Now think about Prior’s argument. Start with the first premise. R entails
(R ∨ M). In this case, the premise entails and also grounds the conclusion.
By hypothesis R is non-ethical and M is ethical. Is the conclusion ethical? We
are not quite yet in a position to answer this, since there are two cases to con-
sider: one in which M obtains and one in which M does not obtain. (R ∨ M)
has two full grounds in the case in which both R and M obtain. This is quite
common. The fact that there is at least one dog in the office is fully grounded
in the fact that Tink the Bloodhound is in the office and also fully grounded
in the fact that Maggie the Mastiff is in the office (they are friends).
Suppose that M does not obtain. Then (R ∨ M) is fully grounded just
by R, hence fully grounded just by a non-ethical fact, and hence, by meta-
physical autonomy, (R ∨ M) is not ethical.
Now suppose that M does obtain along with R. In this case (R ∨ M)
has two full grounds, one ethical and one non-ethical. Is (R ∨ M) ethical
or non-ethical?26 Notice that metaphysical autonomy is silent about this.
metaphysical autonomy is concerned with cases in which ethical facts
have grounds. It says nothing about the converse case, in which an ethical
fact is among the grounds of some other fact. In order to provide a more
complete characterization of autonomy, we turn to the additional principle
concerned with ethical facts as grounds:
converse metaphysical autonomy: Any fact partly grounded by an
ethical fact is an ethical fact.
By ‘partly’ here we have the usual idea in mind, that ethical facts are
grounded by non-ethical facts together with ethical principles or more gen-
eral ethical facts. This principle uses an expansive conception of ‘ethical’,
which ranges over pure ethical facts, such as fundamental principles, ‘mixed’
ethical facts, such as facts explained by principles together with the natural
facts which would figure in their antecedents, and also ‘partly’ ethical facts,
such as conjunctions or disjunctions of pure or mixed ethical facts with
non-ethical facts.
26
One shortcoming with this terminology is that it suggests that a fact cannot be both
ethical and non-ethical. This is misleading. A fact may be ethical and non-ethical, e.g., in
precisely such a case as the one to which this note is appended. It would be clearer that
this is unproblematic if we had used ‘descriptive’ rather than ‘non-ethical’ throughout.
I avoid this to remain neutral about the nature of the non-ethical.
With this principle in place, we can say that if M obtains, then M

grounds (R ∨ M) and hence that (R ∨ M) has an ethical fact among its
grounds, and so we have no counterexample to our metaphysical autonomy
theses. In this case our taxonomy vindicates this counterexample to simple
logical autonomy, for it is still the case that R is non-ethical, and that R
entails (R ∨ M). On the other hand, if M does not obtain, then (R ∨ M) is
not ethical, and hence again we have no counterexample.
What about the second premise, that ((R ∨ M) & ~R) entails M? In any
world in which ((R ∨ M) & ~R) obtains, (R ∨ M) is not grounded by R
and so it is grounded by M. Since grounding is irreflexive, it is not also the
case that ((R ∨ M) & ~R) grounds M. Hence the premise entails but does
not ground the conclusion. Hence this is not a candidate counterexample
to our metaphysical autonomy theses.
So disjunctive syllogism does not furnish counterexamples to metaphysical
autonomy or converse metaphysical autonomy. We can add that in the
Maitzen and Karmo examples it is not at all plausible that the premises provide
grounds for their conclusions. The fact that at least one ethical proposition is
true is not among the grounds for any particular ethical proposition—on the
contrary, the grounding relation would go the other way. The fact that Alfie
believes that some moral proposition is true doesn’t ground the truth of that
proposition, unless some kind of divine command theory is true, and Alfie
is divinely in command. Hence those counterexamples to logical autonomy
theses are not counterexamples to these metaphysical autonomy theses. We’ll
consider the ethicality of the various premises as we proceed.
It is worth mentioning that we are now entitled to reject one of the more
innocuous of the Humberstone/Lewis principles, namely equivalence: if
S and S* are classically equivalent, and S is [ethical], then S* is [ethi-
cal]. Compare R with (R ∨ (R & M)). They are logically equivalent. R is
non-ethical. But in the world in which M obtains, M is one of the grounds
of (R & M) and of (R ∨ (R & M)), which by converse metaphysical
autonomy is ethical. This is an interesting upshot of the view.27
8.5 CHALLENGES TO OUR METAPHYSICAL

AUTONOMY THESES
We can now discuss some challenges to our metaphysical autonomy the-

ses. These are challenges to our theses qua autonomy theses, not challenges
that assume that our theses provide good characterizations of autonomy
and challenge their truth. As against metaphysical autonomy we have
Thanks to Kit Fine for the example.

27
202 Barry Maguire
some facts that seem to be ethical but which have non-ethical and no ethi-
cal grounds. As against converse metaphysical autonomy we have some
facts that have ethical grounds but which do not seem to be ethical. We’ll
take them in turn.
8.5.1 Challenges to metaphysical autonomy

First we’ll consider challenges to metaphysical autonomy. These are cases
in which we have some fact in some world which seems to be ethical, but
that has no ethical grounds, and even though some ethical principles obtain
in that world. We can consider two classes of such cases. First, we have facts
which seem ethical but which obtain trivially in virtue of non-ethical facts,
and not in virtue of the true ethical theory. Second, we have propositions
which seem ethical but which are in fact false, and hence lack grounds alto-
gether. We’ll take them in turn.
We can start with Stephen Maitzen’s second premise, that if at least one
ethical proposition is true, then torturing innocent children is impermis-
sible. Maitzen notes that ethical nihilists can consistently accept this prem-
ise if they regard it as trivially true for having a false antecedent.28 This
is consistent with our metaphysical autonomy theses. We can continue
to insist that this proposition is ethical in a world in which the anteced-
ent obtains. This reply to this example is the same as our reply to Prior’s
argument—which you might expect given the fact that the material condi-
tional can be expressed with disjunction and negation.29
A similar approach enables us to deal with a challenge from Peter Vranas.30
Suppose there are no police officers in the post office. The fact that there
are no police officers in the post office grounds the fact that every police
officer in the post office is morally required to accept bribes. Moreover the
fact that there are no police officers in the post office plausibly fully grounds
the latter fact in some world. (There will be some worlds in which this fact
has an ethical ground, but this isn’t one of them.) The fact that there are
no police officers in the post office is not an ethical fact. So if the fact that
every police officer in the post office is morally required to accept bribes is
an ethical fact then we have a counterexample to metaphysical autonomy.
But it is clear how we should reply. We know from above that not all facts
with ethical constituents are ethical facts. We cannot simply assume that
28
He says the same about the following proposition: ‘if catastrophic global warming is
a genuine threat, then we ought to do something about it’ (2010: 323).
29
There is a difficult question about how to classify the necessitated form of Maitzen’s
conditional. I omit discussion for lack of space.
30
In discussion. See also Vranas (2010).
the latter fact—the fact about police officers being morally required to take
bribes—is ethical. Moreover in cases like this, in which some fact is fully
grounded trivially by some non-ethical fact, it is intuitively plausible that
the fact is not ethical. (Later on we’ll say the same about facts of the form
‘it is not impermissible to x’ in worlds in which ethical nihilism is true.)
Here’s the second kind of case.31 Suppose that in the actual world the
proposition that early stage abortion is impermissible is false. Grounding is
factive. So since this proposition is false it doesn’t have grounds, and neither
does it ground anything. So it doesn’t seem as though it will be classified
as ethical according to our autonomy theses. But the proposition that early
stage abortion is impermissible is plausibly ethical, both in general, and in
our world.
The objector has changed the subject. Our metaphysical autonomy the-
ses do not purport to provide a full ethical taxonomy, and in particular they
do not purport to provide a taxonomy of ethical propositions.32
Still, there are some things we can say about ethical propositions. Firstly
some propositions have ethical constituents. These are ‘ethical’ in one
sense—precisely in the sense that they have at least one ethical constituent,
but this is not the sense of ‘ethical’ that we are working with. Secondly some
propositions are ethical in every world in which they are true.33 Call such
propositions essentially ethical. There is plausibly no world in which
early stage abortion is impermissible, but in which this fact does not have
ethical facts among its grounds. A weaker version of this thesis is that some
propositions, had they been true, would (or the fact to which they corre-
spond would) have obtained in virtue of some ethical fact. We can call such
propositions ethical hereabouts. This is weaker because it is consistent
with there being some distant possible worlds in which the proposition is
true and not ethical. The important thing to notice is that a proposition
might be essentially ethical or ethical hereabouts and yet not ethical
in this world, since it is not true in this world.
31
Thanks to an audience at the Northern Institute of Philosophy for a discussion of
this objection.
32
It is also worth noting that on some accounts of facts, negative facts can have ethical
grounds. I discuss this further in section 6.1.
33
I beg for your patience here as we move between fact-talk and proposition-talk. It is
simplest to assume that facts are true propositions. I’m optimistic that the general points
to be made here can be translated into your preferred metaphysics of objects, properties,
facts, propositions, and truth.
204 Barry Maguire
8.5.2 Challenges to converse metaphysical autonomy

Remember Alfie from earlier on. Everything he says is true. Suppose Alfie
restricts his conversation to non-ethical matters. Perhaps he tells you about
the distances between various planets. Then it is plausible that the fact that
everything Alfie says is true is non-ethical. But suppose that later on Alfie
pronounces upon many ethical matters, for instance that early abortion in
cases of serious foetal abnormality is impermissible, that one ought to be
sincere, that one should give 10% of one’s gross income to charity. Then,
plausibly, the fact that everything Alfie says is true is ethical. For in the rel-
evant case, the grounds for this proposition would include various ethical
facts. This is all consistent with converse metaphysical autonomy.
The same seems to be true of the fact that there is at least one ethical fact.
That latter fact obtains in virtue of any ethical fact. Hence by converse
metaphysical autonomy it is ethical. This also seems plausible.
However, consider a related challenge. Any ethical fact will also ground
the fact that there is at least one thing instantiating a property. Hence by
converse metaphysical autonomy this latter fact would be ethical. But
the fact that there is at least one thing instantiating a property doesn’t seem
to be an ethical fact. We can generalize this worry. Nearly every fact is a
ground for the fact that there is at least one thing instantiating a property.
We could entertain a generalization of converse metaphysical autonomy
that applies to any category, namely: for any category of fact K, no non-K
fact is partly grounded by any K fact. It would follow from this generalized
principle that the proposition that at least one proposition is true would
belong to every category one proposition about which was true.
Perhaps you think this really is an objection to the generalized princi-
ple. But it strikes me as utterly harmless. The more specific the grounded
proposition, the fewer the categories to which it will belong and the more
plausible the generalized version of the principle will appear—and this gen-
eralized principle is much stronger than we need. Also bear in mind that
we are employing a more expansive conception of ‘ethicality’ that includes
‘partly ethical’ and ‘mixed ethical’ facts.
I owe to Jack Woods a more worrisome objection. Imagine you are play-
ing a game, in which the winner is the one who ends up with the most
points. The way to score points is to do anything ethically impermissible
between 5 p.m. and 6 p.m. on the day of the game. Then it may be that
kicking a cat between 5 p.m. and 6 p.m. would win you a point in virtue of
the fact that cat-kicking is ethically impermissible and the facts about the
game just mentioned. And yet the fact that kicking a cat between 5 p.m.
and 6 p.m. would win you a point doesn’t seem like an ethical fact.
The premises here are ethical, and they certainly ground the conclusion.
Here I have to dig in, and insist that this fact about the game is an ethi-
cal fact. We let the theory decide cases like these.34 This is similar to the
fact that what Alfie said is true, when he says it is wrong to kick dogs. But
here’s another example that might make this reply sound more plausible.
Compare two theories about the nature of law, one of which maintains
that ethical facts are always among the grounds of legal facts, the other
which denies that ethical facts are always among the grounds of legal facts.
Plausibly legal facts of the first kind are also ethical facts, and plausibly legal
facts of the second kind are not.
8.5.3 Taxonomic Essentialism

The metaphysical autonomy principles involve a commitment to the fol-
lowing thesis: that the taxonomic category of a fact varies with its grounds
(or rather, the category of an instance of a fact-type varies with its grounds).
We have seen various examples of this:
• ‘All police officers in the post office are morally required to take bribes’ is
ethical when it is grounded by some ethical theory and non-ethical when
it is grounded fully and uniquely by the contingent fact that there are no
police officers in the post office.
• ‘Everything that Alfie said is true’ is ethical if Alfie made some ethical
claims and non-ethical if he did not.
• ‘Either it is permissible to abort in cases of serious foetal abnormality
or Susan Rumplebottom won the 2009 Gloucestershire Cheese Rolling
competition’ is ethical when it is grounded by the first disjunct, even if
also grounded by the second disjunct, and non-ethical when grounded
only by the second disjunct or not grounded by either disjunct.
We can break this thesis down into two claims. The first is the thesis that
the taxonomic category of a proposition is contingent. The second is the
specific thesis about what the taxonomic category of a fact is contingent
upon, namely the grounding relations that it stands in.
Here is Stephen Maitzen characterizing the first claim (2010: 302):
the contingency thesis asserts that a given proposition’s classification as moral or
non-moral can vary across possible worlds; the essentiality thesis, on the other hand,
asserts that a proposition’s status as moral or non-moral is essential to it, invariant
across the possible worlds in which it exists.
I owe this reply to Gideon Rosen.

34
206 Barry Maguire
Maitzen presents two arguments for the second option, which he calls tax-
onomic essentialism.35 Here’s Maitzen’s first argument (2010: 302):
What we mean is up to us, but whether what we mean is true is up to the world. But
if what we mean is up to us—if it’s determined by our communicative intentions,
collective or otherwise—then surely the taxonomic status of what we mean is up to
us too, since what kind of thing we mean depends on what it is, in particular, that
we mean. But the contingency thesis makes the kind of thing we mean—not the
wide content of our utterance (as semantic externalists already insist) but its very
taxonomic category—depend on something besides our communicative intentions,
namely, the way the world is.
Both premises are contestable. As Maitzen indicates, semantic externalists
will reasonably reject the first premise. It is plausible that what we mean by
‘water’ is partly determined by what water is. Similarly it is plausible that
what we mean by ‘right’ or ‘good’ is determined by what right is and what
good is, for instance, perhaps to be good is to have properties that make
positive responses appropriate; perhaps to be right is to be value maximiz-
ing. But more importantly, it wouldn’t obviously follow from the fact that
what we mean is up to us that the taxonomic category of propositions is
up to us. For as we have already seen, it is plausible that the proposition
that what Alfie said is true is ethical when he said something ethical and
non-ethical otherwise. But the meaning of this proposition doesn’t change.
Similarly we can theoretically distinguish cases in which legal facts are or
are not grounded in ethical facts, or mental facts are or are not grounded
in physical facts, or mathematical facts are or are not grounded in logical
facts. Unless we subscribe to some grand metaphysical idealism, these mat-
ters will not be decided by our communicative intentions. But it strikes
me as plausible that legal facts are ethical facts if and only if they are partly
grounded in ethical facts, and mutatis mutandis for psychological facts and
mathematical facts.
Here’s Maitzen’s second argument (2010: 303):
The contingency thesis makes us implausibly ignorant of the correct classification
of disjunctions such as (GR) Goldbach’s Conjecture is true, or Rothenberg’s set-
ting his son on fire was morally wrong, since we don’t, and perhaps can’t, know the
truth-value of one of the disjuncts.
Strictly speaking we are concerned with the grounds for the disjunction
rather than its truth-value, though the two are related. It is true that
according to our metaphysical autonomy theses, we won’t know the taxo-
nomic category of something until we know its grounds. However, this
35
We can formulate a fact-analogue of taxonomic essentialism appealing to
fact-types instead of propositions.
doesn’t seem so bad. We won’t know the taxonomic category of mental

or mathematical facts until we do some philosophy. Plausibly we won’t
know whether ethical facts are natural facts until we do some philosophy.
So also plausibly we won’t know whether such a disjunction is mathemati-
cal or ethical until we do some philosophy. Perhaps we will never know.
But I see no reason to accept that a fact or proposition has a taxonomic
category only if we are in position to know that it has that category, unless
one accepts some such thesis in the philosophy of language as the one
mentioned above. Hence this objection relies upon the previous one,
which anyone disposed to reject taxonomic essentialism would also be
disposed to reject.
Let’s now turn to the more specific thesis that taxonomic category is
contingent upon facts about grounding. Maitzen rightly notes that it
would beg the question in the context of a dispute about simple logi-
cal autonomy simply to insist that the ethical status of a conclusion is
contingent upon the ethical status of its premises in the following way: a
conclusion is ethical only if it is not logically implied by non-ethical prem-
ises. But it does not analogously beg the question to insist that the ethical-
ity of a fact is contingent upon its grounds or what it grounds. For this
further thesis has been motivated independently. Moreover the best puta-
tive counterexamples to the metaphysical autonomy theses (such as Mark
Schroeder’s 2007) accept that these theses provide a decent characteriza-
tion of autonomy. This vindicates rather than undermines the taxonomic
status of these theses.
8.5.4 Particularism
A second worry about the idea that taxonomic category of a fact depends on
its grounds has to do with ethical particularism. Recall that grounds neces-
sitate what they ground. This has modal implications: if facts A ground
fact B in one situation, then facts A will ground fact B in any other situ-
ation. Imagine a hardcore particularist who denies that ethical facts ever
have grounds of this sort. Does this position put pressure on the grounding
strategy?
No. We still distinguish the fact that x-ing would realize S, for some
maximally specific x and S, from the fact that <the fact that x-ing would
realize S> is a reason for you to x. The former fact doesn’t ground the latter
fact. If hardcore particularism is true then there are no general principles
relating these two kinds of facts, such as a fact of the form, whenever x-ing
would realize S that fact is a reason for you to x. Instead the facts about rea-
sons themselves (the latter facts) have no grounds. This is clearly compatible
with metaphysical autonomy.
208 Barry Maguire
If hardcore particularism is true, our autonomy theses might also be true,

but as it were they would have less work to do. They would still be doing some
work, since there would still be various facts grounded by these facts about
reasons. For instance it would still be the case that there is at least one ethical
fact. converse ethical autonomy would entail that this fact is ethical.
8.6 IMPLICATIONS OF METAPHYSICAL AUTONOMY
8.6.1 Negation and Nihilism

Suppose you ask a reflective atheist whether the proposition that there is no
God is a theological proposition. She might not be exactly sure what you
are asking, but she might be willing to assent. Suppose you asked another
reflective atheist whether there are any true theological propositions. He
might not be sure exactly what you are asking, but he might be willing
to deny that there are. Now imagine asking either one of these reflective
atheists both of these questions at once. Clearly he or she would not assent
to both of them together. Why? The two questions are determinate only
if one implicitly presupposes certain taxonomic theses. But if we assume
that the proposition that there is no God is a theological proposition, then
no reflective atheist—indeed, no half-witted atheist—would insist that all
theological propositions are false.
Similarly, there are some characterizations of nihilism that together with
certain taxonomic theses entail that nihilism is trivially false. Consider the
combination of:
negation: The negation of any ethical proposition is ethical.
And:
prima facie nihilism: there are no true ethical propositions.
If the negation of any ethical proposition is ethical, then for any unconten-
tiously ethical proposition M, the proposition that it is not true that M
is ethical, by negation. But we assume either M is true or M is not true.
So either way at least one ethical proposition is true. Hence prima facie
nihilism is false. In fact we get the same result with weaker negation: for
at least one ethical proposition, the proposition that that proposition is not
true is an ethical proposition. weaker negation also entails the negation
of prima facie nihilism.
This argument is not likely to convince anyone of its conclusion. What it
rather shows is that we need to reassess either the taxonomic thesis (negation)
or the characterization of nihilism (prima facie nihilism).
We now have a reason to reject negation, or rather the fact-hood ver-

sion, according to which the negation of any ethical fact is ethical. For
if there are negative facts, and negative facts have grounds, clearly these
grounds will be quite different from the grounds for the non-negated fact.
Here’s a speculative suggestion. Suppose that some version of utilitarianism
is true, i.e., some principle of the following form obtains: for all actions x,
one ought to x if and only if x-ing would bring about more net happiness
than any available alternative to x. Furthermore, suppose that this version of
utilitarianism includes the following normative bridge principle: x is imper-
missible only if it is not the case that one ought to x. Furthermore, suppose
that in some case of serious foetal abnormality, an abortion will bring about
more net happiness than any alternative. In this situation, given this ethical
theory, it is not impermissible to perform this abortion. Importantly, the
fact that it is not the case that it is impermissible to perform this abor-
tion plausibly obtains in virtue of the facts about the ethical theory and
the non-ethical facts about the bringing about of net pleasure. Hence, by
converse metaphysical autonomy, this negative fact is ethical. In such a
case, the fact that it is not the case that the abortion is impermissible is an
ethical fact.36
Nihilism logically entails the fact that it is not the case that abortion is
impermissible. It is a separate question whether nihilism grounds this fact,
and if not, what does. Nihilism is logically inconsistent with any ethical
ground for this fact. But importantly the nihilist denies impermissibility by
saying that this abortion has no ethical properties—it is neither impermis-
sible, permissible, nor required. The nihilist is motivated by the idea that
the world does not contain any ethical properties at all. To quote Richard
Joyce: ‘The nihilist denies that there is anything morally permissible about
[performing the abortion] with just as much gusto as she denies that there
is anything morally wrong with doing so’ (2013).37 So assuming that nihil-
ism is true, in this case the fact that it is not the case that the abortion is
impermissible is not an ethical fact. It is another example of a non-ethical
fact with an ethical constituent.
36
J. S. Mill’s harm principle provides a nice case of a positive ethical ascription of
permissibility.
37
This is from a recent review of David Enoch’s defence of ‘robust realism’. Here’s
Gilbert Harman making this point: ‘According to the moral nihilist, “nothing is ever
right or wrong, just or unjust, good or bad” ’ (Harman 1977: 11). Here’s Michael Smith:
‘ … [according to a nihilist] the world contains no moral features at all: not the feature
of being obligatory, not the feature of being forbidden, and not the feature of being per-
missible either. This is all to say that external sceptics deny that any moral qualities exist,
including the quality of permissiveness’ (2010: 512).
210 Barry Maguire
8.6.2 Characterizing Revisionary Metaethical Positions

In order to avoid the foregoing argument from negation, nihilists some-
times restrict their thesis to atomic nihilism: no atomic ethical proposi-
tions are true.38 However, there are problems with this option. Sometimes,
even though atomic nihilism is plausibly true, there are some true propo-
sitions that are plausibly ethical, and hence nihilism is plausibly false. This
suggests that atomic nihilism is not the best characterization of nihilism.
Imagine the following situation. The following proposition is true: it
is impermissible to kick dogs. No other principles of this kind are true.
However, there are no dogs. Alas, the last dog died two months ago. Were
there any dogs, it would be impermissible to kick them. But there aren’t, so
it isn’t. There is nothing such that it is impermissible to kick it. Hence in this
world, at this time, there are no instantiated ethical properties, and hence
there are no true atomic ethical propositions. However, we feel inclined to
say that nihilism is not true in this world.
We can make the same move again. Suppose it isn’t impermissible to kick
dogs in some world. Perhaps the dogs in this world are infertile unless they
receive a stout kick. There are still no other high-level principles that obtain
in this world. Still, a lower-level principle obtains, which says that it is
impermissible to kick dogs unless it is in their best interest. This lower-level
principle also seems ethical.
A naïve nihilist would want to deny that these dog-kicking principles are
true, or that consequentialism is true, or that it is impermissible to break
one’s promises, or that whatever causes pleasure without hard work is evil,
with just as much gusto as she would deny that it is impermissible to break
some particular promise or kick some particular dog.
We need further taxonomic principles to tell us which propositions
are ethical. Here we have an argument to the effect that we should accept
taxonomic principles that tell us that principles such as these—principles
that would have been the grounds for atomic ethical propositions, if the
non-ethical facts had been different—are ethical. These considerations sug-
gest that such principles are ethical even if they do not in fact ground any-
thing in some world. So let robust nihilism be the thesis that there are no
atomic ethical facts and no ethical principles either. I submit that this is an
interesting and plausible conception of nihilism.
It is plausible that any true principle that plays or would play this role of
grounding atomic ethical facts is ethical. This provides support for Simon
Blackburn’s well-known contention that subjectivistic principles relating
38
Charles Pigden is the main exemplar of this view, see his (1989) and contributions
to (2010).
ethical claims to psychological facts are ethical as opposed to being non-

ethical metaethical claims.39 Moreover this provides support for T. M.
Scanlon’s interpretation of the ‘Humean theory of reasons’ as a substantive
ethical theory (2014). Any principles that purport to provide grounds for
ethical claims, whether or not they should be classified as ‘metaethical’ prin-
ciples, are also ethical principles.
8.6.3 The Classic Argument for Nihilism

Now let’s turn to the classic argument for nihilism.40 It has two premises.
Premise one maintains that there are ethical facts only if there are facts of
type K, and premise two insists that there are no facts of type K. Perhaps K
is ‘causally inert facts’, or ‘queer facts’, or ‘categorical facts’. This won’t really
matter for our purposes. The conclusion is that there are no ethical facts. We
can present the first premise in the form of a principle:
the nihilist’s principle: for any x, x is impermissible only if there
are facts of type K.
For simplicity we will assume that if there are no impermissibility facts then
there are no other plausibly ethical facts (e.g. facts about right and wrong,
virtue, moral requirements, etc.).
Ronald Dworkin’s provocative suggestion is that this principle is ethical.
If that were true, then the argument for nihilism would undermine itself,
since the conjunction of the premises and the conclusion would be incon-
sistent. Dworkin’s argument for the claim that the nihilist’s principle is
ethical is a burden-pushing, no-relevant-difference argument. The idea is
that there is no relevant difference between the kicking-dog principle and
the nihilist’s principle. Since the kicking-dog principle is ethical, the
nihilist’s principle is ethical—or so Dworkin argues.
However there are various relevant differences.
Firstly, the dog-kicking principle states sufficient conditions. But the
nihilist principle is a necessity principle: it states necessary conditions
for anything’s being impermissible. It is analogous to principles such as
ought-entails-can.
39
Cf. Blackburn (1993: 4) and elsewhere—e.g., the review of Dworkin (Blackburn
1996). The ‘protected contexts’ strategy for carving out theoretical room to distinguish
Blackburn’s view from a straightforward realist position accepts—or at least is consistent
with—the thesis that dependency claims like these are ethical claims.
40
For versions of this argument, see Joyce (2001) and Smith (2010). Cf. also Mackie
(1977: 48). His ‘argument from queerness’ has a metaphysical premise and an epis-
temological premise. We can characterize his epistemological premise as a separate
sub-argument for the second premise in the classic argument presented in the main text.
212 Barry Maguire
Necessity principles cannot ground ethical facts with just non-ethical

facts. They need some companion sufficiency principle. But now look at
our metaphysical autonomy thesis. It maintains that whenever some
ethical fact is grounded by some other facts, those facts include at least one
ethical fact. metaphysical autonomy does not and cannot discriminate
within these grounds. In the case of sufficiency principles we can work out
that a principle is ethical by elimination. For example the principle that it
is impermissible to kick dogs, together with the fact that this creature is a
dog, grounds the fact that it is impermissible to kick this creature. Here the
sufficiency principle is the only plausible ethical fact among the grounds.
But since necessity principles never ground ethical facts without some suf-
ficiency principle, our taxonomic principles underdetermine whether they
are ethical. At best, further taxonomic principles would need to be articu-
lated and defended, which are capable of drawing plausible distinctions.41
However, there is a more serious problem with Dworkin’s proposal.
Neither nihilism nor the nihilist’s principle was ever supposed to
ground any positive or negative facts about impermissibility. The nihilist
does not mean to suggest that the absence of queerness makes it the case
that all atomic ethical propositions are false. For the nihilist may be quite
open to there being other problems with ethics. She is not committed to the
idea that ethical principles obtain by some kind of default. the nihilist’s
principle simply logically entails that, for any x, it is not the case that x is
impermissible. This completely side-steps our autonomy theses.
8.6.4 Arguing Mooreanly Against Revisionists

Nihilism is not incoherent. But it is surely ethically objectionable. As Simon
Blackburn has emphasized, highly revisionary theses which make facts
about right and wrong dependent on contingent psychological states are
also ethically objectionable.
There is one particularly interesting upshot of insisting upon character-
izing autonomy metaphysically (and in particular with rejecting negation).
All sorts of ethical and non-ethical propositions will be logically inconsistent
with each other. In particular, any old nihilistic thesis is going to be logically
inconsistent with very many extremely plausible ethical claims.
41
the nihiilist’s principle is also what you might call a sledgehammer principle: it
rules all of them out at once. This does seem like a relevant difference between the nihil-
istic premise and other non-nihilistic views like consequentialism and, importantly, sub-
jectivism. However, it is important to note that the principle that unless God exists,
nothing is impermissible, also has this feature. It is also a sledgehammer principle. But
that divine command principle is plausibly ethical.
This picture fits well with the way in which Dworkin, Cohen, Scanlon,
and Blackburn want to react to revisionary metaethical principles, that is,
metaethical principles that entail substantive ethical claims which fail to
cohere with our best ethical reflection. Consider the following characteristic
passage from Ronald Dworkin (1996: 117):
Let us accept, for the sake of argument, that we are forced to choose between the
following two propositions. (1) Human beings have a special though sometimes
fallible faculty of judgment that enables us to decide which moral claims to accept
or reject, a capacity whose malfunctioning may sometimes result only in moral mis-
judgement with no spillover impairment of other cognitive activity. (2) There is no
moral objection to exterminating an ethnic group or enslaving a race or torturing a
young child, just for fun, in front of its captive mother. Which should we abandon?
On these terms, clearly we should accept 1 and reject 2. Now consider a
choice between, on the one hand, rejecting some extremely plausible ethical
claim, and on the other hand rejecting the conjunction of the nihilist’s prin-
ciple together with the metaphysical thesis that the relevant facts of kind K
do not obtain. Clearly we should hang on to the extremely plausible ethical
claim. This is another way to interpret what Dworkin was getting at in his abor-
tion argument. When one is arguing for the negation of an ethical claim one
might not thereby be arguing for an ethical claim. But one may nevertheless be
arguing for a claim that is extremely implausible. Indeed, no matter what you
think about the ethics of abortion in cases of serious foetal abnormality, you
will probably agree that the interlocutor who maintains that abortion is not
impermissible in such circumstances, because nihilism is true, has by far the
least plausible position. This is because by committing herself to nihilism, she
thereby commits herself to denying so many other extremely plausible claims.42
References
Bader, R. Forthcoming. ‘Conditions, Modifiers, and Holism’, in E. Lord and B.
Maguire (eds.), Weighing Reasons. Oxford: Oxford University Press.
Bennett, K. 2011. ‘By Our Bootstraps’, Philosophical Perspectives 25: 27–41.
Blackburn, S. 1971. ‘Moral Realism’, in J. Casey (ed.), Morality and Moral Reasoning.
London: Methuen, 101–24.
42
Special thanks to Gideon Rosen and Jack Woods for many helpful comments
and objections. Thanks also to Derek Baker, David Faraci, Kit Fine, Meghan Flaherty,
Boris Kment, Adam Lerner, Errol Lord, Sarah McGrath, Tristram McPherson, Carla
Merino-Rajme, Andreas Müller, L. A. Paul, Carlotta Pavese, Charles Pigden, Kristin
Primus, Karl Schafer, Mark Schroeder, Whitney Schwab, Michael Smith, Daniel Wodak,
two anonymous referees, and audiences at the Northern Institute of Philosophy and the
Madison Metaethics Workshop.
214 Barry Maguire
Blackburn, S. 1985. ‘Supervenience Revisited’, in I. Hacking (ed.), Exercises in

Analysis: Essays by Students of Casimir Lewy. Cambridge: Cambridge University
Press, 47–67.
Blackburn, S. 1993. Essays in Quasi-Realism. Oxford: Oxford University Press.
Blackburn, S. 1996. ‘Comments on Dworkin’. Available at: <http://www.brown.
edu/Departments/Philosophy/bears/symp-dworkin.html>.
Blackburn, S. 2006. ‘Must We Weep for Sentimentalism?’ in J. Dreier
(ed.), Contemporary Debates in Moral Theory. London: Blackwell, 144–60.
Brown, C. 2014. ‘Minding the Is-Ought Gap’, Journal of Philosophical Logic
43: 53–69.
Chudnoff, E. 2013. Intuition. Oxford: Oxford University Press.
Cohen, G. A. 2008. Rescuing Justice and Equality. Cambridge, MA: Harvard
University Press.
Correia, F. 2010. ‘Grounding and Truth-Functions’, Logique et Analyse 53: 251–79.
Dworkin, R. 1986. Law’s Empire. Cambridge, MA: Harvard University Press.
Dworkin, R. 1996. ‘Objectivity and Truth: You’d Better Believe It’, Philosophy and
Public Affairs 25: 87–139.
Dworkin, R. 2012. Justice for Hedgehogs. Cambridge, MA: Harvard University Press.
Fine, K. 2001. ‘The Question of Realism’, Philosopher’s Imprint 1: 1–30.
Fine, K. 2002. ‘The Varieties of Necessity’, in T. S. Gendler and J. Hawthorne (eds.),
Conceivability and Possibility. Oxford: Oxford University Press, 253–82.
Fine, K. 2012. ‘Guide to Ground’, in F. Correia and B. Schneider (eds.), Metaphysical
Grounding: Understanding the Structure of Reality. Cambridge: Cambridge
Harman, G. 1977. The Nature of Morality. New York: Oxford University Press.
Humberstone, L. 1982. ‘First Steps in Philosophical Taxonomy’, Canadian Journal
of Philosophy 12: 467–78.
Humberstone, L. 1995. ‘A Study in Philosophical Taxonomy’, Philosophical Studies
83: 121–69.
Jenkins, C. S. 2011. ‘Is Metaphysical Grounding Irreflexive?’ The Monist 94:
267–76.
Joyce, R. 2013. ‘Review of Taking Morality Seriously’, Ethics 123: 365–9.
Karmo, T. 1988. ‘Some Valid (But No Sound) Arguments Trivially Span the
Is-Ought Gap’, Mind 97: 252–7.
Lewis, D. 1988. ‘Statements Partly About Observation’, Philosophical Papers
17: 1–31.
Mackie, J. 1977. Ethics: Inventing Right and Wrong. London: Penguin.
Maguire, B. Forthcoming. ‘The Autonomy of Ethics’, in T. McPherson and D.
Plunkett (eds.), Routledge Handbook of Metaethics.
Maitzen, S. 2010. ‘Moral Conclusions from Non-Moral Premises’, in C. Pigden
(ed.), Hume on Is and Ought. Basingstoke: Palgrave Macmillan, 290–309.
Nelson, T. M. 1995. ‘Is It Always Fallacious to Derive Values from Facts?’
Argumentation 9: 553–62.
Pigden, C. 1989. ‘Logic and the Autonomy of Ethics’, Australasian Journal of

Pigden, C. (ed.) 2010. Hume on Is and Ought. Basingstoke: Palgrave Macmillan.
Prior, A. N. 1960. ‘The Autonomy of Ethics’, Australasian Journal of Philosophy
38: 199–206.
Rosen, G. 2010. ‘Metaphysical Dependence: Grounding and Reduction’, in B.
Hale and A. Hoffmann (eds.), Modality: Metaphysics, Logic, and Epistemology.
Rosen, G. (n.d.). ‘Normative Necessity’, manuscript.
Russell, G. 2010. ‘In Defence of Hume’s Law’, in C. Pigden (ed.), Hume on Is and
Ought. Basingstoke: Palgrave Macmillan, 151–61.
Samuel, J. (n.d.). ‘Grounding as a Relevant Relation’, manuscript.
Scanlon, T. M. 1998. What We Owe to Each Other. Cambridge: Harvard University
Press.
Schroeder, M. 2007. Slaves of the Passions. Oxford: Oxford University Press.
Searle, J. 1964. ‘How to Derive “Ought” from “Is” ’, Philosophical Review 73: 43–58.
Smith, M. 2010. ‘Dworkin on External Scepticism’, Boston University Law Review
90: 509–20.
Vranas, P. 2010. ‘Comments on “Barriers to Implication” ’, in C. Pigden (ed.),
Hume on Is and Ought. Basingstoke: Palgrave Macmillan, 260–7.
9
Irreducibly Normative Properties
Chris Heathwood
Those who maintain that normative or evaluative properties cannot be

reduced to, identified with, or analyzed in terms of natural properties have
difficulty explaining what these properties are. Stephen Finlay characterizes
the problem in the following passage:
On the nonnaturalists’ view … reality has ‘brute, inexplicable’ normativity, which
cannot be explained in motivational or other natural terms. This inexplicability is
twofold: we cannot explain what normativity is in nonnormative language, and
neither can we explain why the fundamental normative truths hold (e.g., why the
fact that pain hurts counts in favor of preventing it).
(Finlay 2007: 24)
I have argued elsewhere that no metaethical theory—naturalist,
non-naturalist, or otherwise—can explain why the fundamental norma-
tive truths hold (Heathwood 2012). In this chapter, I attempt to address
the other “inexplicability problem” for normative non-naturalism: that of
explaining what normativity is in non-normative language. I don’t claim
to be giving a complete characterization of normativity in non-normative
terms, such as an identification of normativity with some natural phenome-
non. To do that would presumably be to abandon non-naturalism. Instead,
I put forward a substantive thesis about normative properties that, if true,
goes some way towards elucidating their nature in non-normative terms.
At a first pass, the view is this:
normative properties are those such that, to attribute one to some-
thing is, in virtue of the nature of the property attributed, necessarily
to commend or condemn that thing.
It characterizes normativity in terms of the natural phenomenon of per-
forming certain familiar speech acts. The idea is that in merely reporting
some of the facts of the world, we can’t help but get ourselves involved in
the further business of evaluating—of commending, recommending, con-
demning, and so forth—when the facts that we are reporting are among
Irreducibly Normative Properties 217
the normative facts of the world. And this is due not to any contingent
practices or conventions of ours (beyond whatever is required to make the
assertion and attribute the property), but to the nature of the property we
have attributed.
In what follows, I further explain the initial problem and provide addi-
tional background (section 9.1); I clarify and refine the proposed solution
(section 9.2); I address some objections (section 9.3); and I describe further
explanatory work that the hypothesis can do, both for the non-naturalist
and more generally (section 9.4). Our topic includes normative facts nar-
rowly construed—as when someone ought to do something—as well as
evaluative facts—as when some state of affairs would be good in itself. For
simplicity, I group both under the label “normative.” The thesis is meant to
cover both “positive” normative facts, as in the above examples, as well as
“negative” cases, such as when someone has a reason against doing some-
thing or when some outcome would be intrinsically bad. I’ll often speak
only of one or the other of the positive or negative cases, even when what
I say applies to both. Since these “thin” normative notions will be enough
to occupy us, I set aside discussion of how the theory might be extended to
so-called thick evaluatives.
It is not my aim here to be giving positive arguments for the existence
and instantiation of irreducibly normative properties. It is rather to be offer-
ing a theory about what such properties would be like. The view is supposed
to enable non-naturalists to deflect an objection to or complaint about their
theory: that the theory posits a class of properties whose natures are myste-
rious and ineffable. However, as we will see, critics of non-naturalism can
accept the account too, even as part of an argument against non-naturalism.
9.1 A PROBLEM FOR NORMATIVE

NON-NATURALISM: WHAT IS NORMATIVITY
OR VALUE?
According to normative non-reductionism, there are normative properties

and facts—facts such as that people ought to be more kind or that the world
would be better if people were—and these facts are sui generis: that is, they
are not identical to any facts that we can express or adequately understand
using terms from some other domain. This view has appeal. For surely there
are some normative facts, such as the examples above, and it doesn’t seem,
at least prima facie, that when we assert some such fact, we are stating a fact
that we could just as well state using non-normative language—as when, in
stating that the earth is covered in water, we could just as well state that fact
218 Chris Heathwood
in chemical terms, by saying that the earth is covered in H2O. The fact that
people ought to be more kind does not at least appear to be the same fact
as any fact expressible in non-normative terms, such as that people would
be motivated to be more kind if they had full information, or that people’s
being more kind would increase preference satisfaction. Rather, the norma-
tive facts about any situation would seem to be further facts about it, and
the properties they involve thus irreducibly normative.
Non-naturalists hold, further, that these sui generis normative facts are
themselves not natural facts about the world. In saying that, they usually
mean one or more of the following: that the facts are not causally effica-
cious, that they are not discoverable wholly empirically, or that they are
not the sorts of facts the natural sciences investigate. These claims also seem
reasonable. Normative facts don’t seem observable with the senses, even
indirectly, nor required to causally explain any non-normative events in
the world.
I am here just remarking on the initial appearances, not on the ulti-
mate truth of the matter. For these initially plausible views face well-known
problems. Non-naturalists, for example, have difficulty explaining how we
can come to know normative facts, or even grasp normative properties, if
these facts and properties don’t interact causally with our brains. And all
non-reductionists have difficulty explaining why the normative facts cannot
vary independently from the non-normative facts, given their view that the
normative facts are further facts about any situation.
Reductive naturalists, who hold that normative facts are identical to cer-
tain natural facts with which we are already familiar, appear to have an
easier time explaining normative knowledge and supervenience. Reductive
naturalists avoid another problem as well: that of saying what normative
properties are, or of explaining the nature of normativity or value. Their
reductionism delivers this automatically. To illustrate, according to a sim-
ple reductive hedonism, the property of being intrinsically good just is the
property of being a state of pleasure; and according to a simple Humean
theory of reasons, to have a reason to do something just is to be motivated
to do it. These reductive theses tell us, respectively, what intrinsic value and
normative reasons are.1
Since non-naturalists resist any identification of these phenomena with
any natural phenomena, they have difficulty saying what their irreducibly
normative properties are, or are like. They can say what they are not like:
they are not causally efficacious; they are not empirically discoverable. But
1
I am not suggesting that reductive naturalism has an easier time giving a correct or
satisfying account here, just that, unlike non-reductionism, it comes with a ready-made
answer to our question.
we’d like to know something by way of positive characterization. For one

thing, these negative characterizations don’t distinguish irreducibly nor-
mative properties from other potentially non-natural properties, such as
modal, mathematical, or logical properties.
By way of positive characterization, non-naturalists typically simply
repeat the normative notions we were wanting some account of, and con-
cede that no other kind of positive characterization is possible. G. E. Moore
writes:
If I am asked ‘What is good?’ my answer is that good is good, and that is the end of
the matter. Or if I am asked ‘How is good to be defined?’ my answer is that it cannot
be defined, and that is all I have to say about it. (1903: §6)2
Derek Parfit is similarly resigned to accepting the inexplicability:
If words like ‘reason’ and ‘ought’ neither refer to natural features, nor express our
attitudes, what could they possibly mean? Non-reductive realists, as I have con-
ceded, do not give helpful answers to these questions. (2006: 330)
More recently, Parfit acknowledges that this opens his view up to the objec-
tion we are considering here:
I admit that, when I say that we have some reason, or that we should or ought to
act in a certain way, what I mean cannot be helpfully explained in other terms …
Williams suggests that the phrase ‘has a reason’ does not have any such intelligible,
irreducibly normative … sense. When he discusses statements about such … rea-
sons, Williams calls these statements ‘mysterious’ and ‘obscure’, and suggests that
they mean nothing. Several other writers make similar claims.3 (2011: 272)
I hope to offer something to blunt the complaint that irreducibly norma-
tive properties are wholly mysterious and obscure. Now, I cannot deny that
some mystery and obscurity will remain even if my view is correct. And of
course other problems, such as concerning knowledge and supervenience,
will remain. But I believe the proposal here makes for some measure of pro-
gress in explaining the nature of normativity on the non-naturalist view.4
2
By “What is good?,” Moore surely means, What is goodness? He of course has sub-
stantive, informative answers to the question, What things are good?
3
Williams asks, “if [an agent] becomes persuaded of this supposedly [irreducibly nor-
mative] truth [that he has a reason to do a certain thing], what is it that he has come to
believe?” (Williams 1995 [1989]: 39). And as Finlay notes, “many philosophers remain
unsatisfied with the thought that normativity might be brute and inexplicable” (Finlay
2010: 8).
4
One might wonder to what extent this problem for non-naturalism is also a problem
for other forms of non-reductionism, especially non-reductive naturalism (the view that,
while normative properties cannot be analyzed non-normatively, they are themselves nat-
ural properties). For reasons that I lack the space to explain, I believe that the complaint
does apply to non-reductive naturalism, but less acutely.
220 Chris Heathwood
9.2 A SOLUTION: IRREDUCIBLY
NORMATIVE PROPERTIES AS ESSENTIALLY
COMMENDATORY PROPERTIES
We use words to describe reality, but we do many other things with them as
well. By uttering certain words in the right context, we can thank someone,
make an offer, condemn an act. Speech acts are a familiar, natural phenome-
non. Also familiar is that sometimes, in performing a speech act of a certain
kind, we thereby perform another speech act. If I say, “I have a car,” I have
described reality as being a certain way; I have performed a description. If
certain other things are true of the circumstances—for example, if you had
just said, “I need a ride to the store”—then, in saying, “I have a car,” I might
also be offering you a ride. In simply describing things as being a certain
way, I can also make an offer.
Typically, and perhaps even in all other cases, which other speech
acts, if any, a person performs in performing a description requires the
existence of certain background conditions beyond whatever is required
to make the description. The semantic meaning of the assertion is not
enough to give rise to other kinds of speech act. But what is interesting
about normative properties, I claim, is that if a person attributes one
to something, thus performing a description, she can’t help but also be
commending or condemning the thing. Normative and evaluative prop-
erties, if irreducible, have this special feature: if someone says sincerely
that something in the world has one of these properties, she, of necessity,
due to the nature of these properties, rather than due to background
conventions and other conditions, involves herself in more than mere
description of the world. The nature of the property is such that it makes
her commend or condemn, praise or criticize, speak positively or nega-
tively, speak for or against. The properties are at once descriptive—as,
trivially, any genuine property must be—and evaluative. We can charac-
terize this as the view of normative properties as essentially commenda-
tory properties.
This hypothesis, if true, should go some way towards assuaging critics of
non-naturalism who are mystified as to what these irreducibly normative
properties are supposed to be. We are all familiar with commending and
condemning; we all do it, no matter our metaethical predilections. These
irreducibly normative properties are interesting, according to our hypothe-
sis, because they are inherently such as to make us do it, whether we want to
or not, whenever we merely attribute one to something. That is something
substantive and interesting about their nature; it distinguishes them from
non-normative properties; and it distinguishes them from other properties
whose nature and existence is contested in philosophy, such as modal, mental,

mathematical, and logical properties.
To be sure, the claim is not that normative properties are such that if some-
thing has one, we ought to commend or condemn it. Such a thesis would
not be characterizing normativity non-normatively. Nor is the claim that
commending—a contingent, interest-relative practice of human beings—itself
forms part of the nature of a putatively objective, stance-independent property.
The normative properties don’t themselves commend; only people can do that.
It is rather that the properties are “commendatory,” which is to say that they
have a certain power: the power to make us commend when we merely attrib-
ute one to something.
I don’t mean “make” in a causal sense, as when a parent, concerned with
politeness, makes his child commend a friend, or when a red object makes us
experience a sensation of red. The relationship is rather a constitutive one. In
attributing a normative property, we thereby commend. To use terminology
J. L. Austin (1962) introduced, it is an illocutionary rather than a perlocution-
ary act. This makes it plausible that this power to commend and condemn
gives us some insight into the nature of the underlying normative property. If
the relation were merely causal, then, since “anything can cause anything,” we
couldn’t claim that the commending gave us any insight into the property’s
intrinsic nature. But since the relation is a much stronger relation—indeed, an
internal relation—we can plausibly claim this.
Speech act theorists have developed other categories and distinctions to help
us understand their object of study. One is the distinction between direct and
indirect speech acts. In saying, “I have a car,” in the earlier example, I was,
directly, making a description, and, indirectly, making an offer. Suppose I say,
“Martin is a good man.” On my view (as well as most others), I am making a
description. On my view (as well as most others), I am also thereby making a
commendation. But what is the status, on my view, of this commendation? Is
it a direct or an indirect speech act?
Typically, and perhaps even in all other cases, whether a person has per-
formed an indirect speech act (in addition to whatever direct speech act she
has performed) is not settled by the semantic content of the utterance, or by
“what is said.” Additional conventions, intentions, and knowledge of these by
the parties involved may be required.5 I am claiming that no such background
5
According to Searle:
In indirect speech acts the speaker communicates to the hearer more than he actually
says by way of relying on their mutually shared background information, both linguistic
and nonlinguistic, together with the general powers of rationality and inference on the
part of the hearer.
(1979: 31–2)
222 Chris Heathwood
conventions, intentions, and knowledge are required to turn a normative asser-

tion into a commendation or criticism. I am suggesting that it is settled—with
an important possible exception to be accommodated shortly—by the seman-
tic content of the assertion. It is because the speaker is saying that a certain
thing has a certain normative property that she is now, whether she intends to
or not, commending or condemning the thing.
For this reason, perhaps we should say that the act of commending or con-
demning that a person performs in attributing a normative property is direct
rather than indirect. It is certainly “less indirect” than stock cases of indirect
speech acts, which involve mediation by the extra-semantic phenomena. On
the other hand, the view says that in attributing a normative property, we
thereby commend. The commendation is parasitic on the description, and is
explained by it. Thus the commendation is “less direct” than the description.
Does it matter what the answer is here, and, more generally, how well
the phenomenon I am postulating fits into accepted speech-act-theoretic
categories? I suppose it would be nice if there were recognized cases that
behaved like the phenomenon I am postulating, but I’m not sure it matters
much. I don’t think it should be much of a surprise if the phenomenon pos-
tulated here turns out to be unusual, or even unique. It is invoked to explain
something unusual, and indeed unique. And granting that the thesis is con-
troversial and novel, we should not have expected speech act theorists to
have used the phenomenon it postulates to guide the construction of their
theories. The issue of the oddness of the phenomenon will come up later
when it comes to explaining how it can be used to account for the queerness
of irreducible normativity.
None of this is to affirm that our phenomenon (of commending due to
the content of an assertion) can occur wholly absent any of the contextual
features required for ordinary speech acts to occur. For one thing, in order
simply for a description to occur certain conditions must obtain (e.g., cer-
tain beliefs and intentions may need to be present in the speaker). And of
course for our words to have the meanings they do requires all manner of
conditions. What is being claimed here is that, once we have whatever is
required for a person to be performing the speech act of genuinely describ-
ing something as having a normative property, nothing else is required for
According to Green:
Whether, in addition to a given speech act, I am also performing an indirect speech act
would seem to depend on my intentions … What is more, these intentions must be fea-
sibly discernible on the part of one’s audience. Even if, in remarking on the fine weather,
I intend as well to request that you pass the salt, I have not done so. I need to make that
intention manifest in some way.
(2009: §3.4)
the further speech act of commendation to occur; rather, what explains why
the further speech act occurs is the nature of the property attributed.
9.2.1 Contrast with Motivational Judgment Internalism

The view of normative properties as necessarily commending properties
should not be confused with any form of motivational judgment internal-
ism, the view that normative judgment entails motivational pro-attitudes
of some kind on the part of the judger (at least for some class of judgers).
It is no part of the view here that when one asserts, say, that one ought
to do some act available to one, and thereby, according to the hypothesis,
commends one’s doing it, one must have some motivation to do it, or any
kind of favorable non-cognitive attitude towards the act at all. A person can
commend something even when he has no such attitudes, just as a person
can thank someone or apologize to someone even when the person doesn’t
feel at all thankful or apologetic.
It is worth spelling out this comparison further. We can distinguish dif-
ferent grades of these speech acts. Consider apology. There is fully insin-
cere apology, as when someone is being sarcastic. No apology occurs there.
Among genuine apologies, we can distinguish at least two kinds. There are
high-grade apologies, in which the apologizer feels genuine remorse. This is
the best kind of apology. But there is a lower-grade variety as well, in which
there are no feelings of remorse, but a genuine apology still occurs. Suppose
I wrong you. I feel guilty about it initially, but, as happens, these feelings
subside. Although I can no longer muster any emotions or disfavorable atti-
tudes about the incident, I still know that what I did was wrong, and this
prompts me to say to you, “I apologize for doing that.” I have apologized
to you, despite lacking the attitudes or feelings that might make it an ideal
apology.6
6
Cf. Austin (1962: 10):
It is gratifying to observe … how excess of profundity, or rather solemnity, at once paves
the way for immodality. For one who says ‘promising is not merely a matter of uttering
words! It is an inward and spiritual act!’ is apt to appear as a solid moralist standing out
against a generation of superficial theorizers: we see him as he sees himself, surveying the
invisible depths of ethical space, with all the distinction of a specialist in the sui generis.
Yet he provides Hippolytus with a let-out, the bigamist with an excuse for his ‘I do’ and
the welsher with a defense for his ‘I bet’. Accuracy and morality alike are on the side of
the plain saying that our word is our bond.
I’m not sure whether Austin would call the apology described above “insincere,” but
it is pretty clear that he would not deny that I have apologized; the apology is not, in his
terminology, “void.” See Austin (1962: 40).
224 Chris Heathwood
Something similar seems true of normative utterance. A person who sin-

cerely attributes a normative property to something—they are not being
sarcastic, they really think the thing has this feature—thereby commends or
condemns it. If he lacks appropriate motivational states or attitudes towards
the thing, this may mean that something less than ideal is going on. Perhaps
whenever we genuinely believe, say, that some act was wrong, we should
have a disfavorable attitude towards it. But if things aren’t ideal, and we lack
the attitude (perhaps we are callous, or tired, or under heavy sedation), but
still believe that the act was wrong, and so describe it as such, a genuine
condemnation has still occurred.7
Thus the view defended here is no form of motivational judgment inter-
nalism.8 Later (section 9.4.3), I indicate how the view can get us what moti-
vational judgment internalism has often been relied upon to deliver: an
account of the “essential practicality” of normativity.
9.2.2 Comparison with Hybrid Theories

The idea that making a normative utterance inherently involves an act of
commending or condemning is often associated with non-cognitivism.
R. M. Hare claims that “the primary function of the word ‘good’ is to com-
mend” (1952: 127). But, as a theory about the nature of normative proper-
ties and facts, the view here is not a kind of non-cognitivism. More popular
these days than pure non-cognitivist theories, however, may be hybrids of
these with cognitivist theories. I arrived at the view here through a problem
in normative metaphysics: the problem of the nature of irreducibly nor-
mative properties. But the view turns out to have implications concerning
whether a “hybrid theory” in metaethics is true. The view is in fact a kind of
hybrid theory, though of a less-discussed variety.
To begin, we should distinguish theses about normative thought, or judg-
ment, from theses about normative utterance. The former tell us what kind
of mental states normative judgments are; the latter tell us what we are doing
when we make normative utterances. According to cognitivist theories of
normative judgment, normative judgments are cognitive states—in particu-
lar, beliefs. According to non-cognitivist theories, they are non-cognitive
7
Copp (2009: 173–4) affirms a similar view.
8
Thomson (2008: 54) similarly dissociates speech acts like commending from the
having of positive attitudes:
it is one thing to perform the speech act of praising a thing and quite another to have any
thing that would ordinarily be regarded as a favorable attitude towards the thing praised.
states, such as desires. According to hybrid theories of normative judgment,

normative judgments are composite states consisting of both.
One way to characterize normative utterances is in terms of the mental
states they are thought to express. Thus, one kind of hybridism about nor-
mative utterance is the view that a declarative normative utterance expresses
both a cognitive and a non-cognitive state. But we can also characterize
normative utterances behaviorally rather than psychologically—that is, in
terms of which kinds of speech act they are instances of rather than accord-
ing to which kinds of mental state they express. According to one such view,
declarative normative utterances are assertions or descriptions, and nothing
more. We can call this descriptivism about normative utterance. The oppo-
site view, non-descriptivism, is the view that grammatically declarative nor-
mative utterances are not in the business of describing reality, and instead
do something such as prescribe or commend.
The theory defended here about the nature of normative properties
attempts to get at their nature by advancing a thesis about what we are
doing when we attribute normative properties to things, that is, when we
make declarative normative utterances. Thus, while it has no direct implica-
tions regarding normative judgment or thought, it does have direct implica-
tions regarding normative utterance. The view is a version of a less-discussed
form of hybridism about normative utterance: a hybrid of what I have
called descriptivism and non-descriptivism. For it holds that declarative
normative utterances necessarily do something descriptive—they attribute
normative properties to things—and something non-descriptive—they
commend or condemn. Unlike on some other forms of hybridism, the
non-descriptive and descriptive elements are necessarily connected on my
view, in that making a normative description entails making a commenda-
tion or condemnation.9
Often hybrid theories in metaethics have naturalistic motivations, such
as to inject normativity, or something like it, into a naturalistic realist meta-
ethic. But the view defended here is motivated instead by a desire to be able
to explain, to some extent, what normativity might be if it is non-natural
and irreducible. Thus more common forms of hybridism and my view
begin from quite different motivations, even if we end up in similar places.10
9
For reasons I lack the space to explain, this enables the theory to avoid some prob-
lems faced by other hybrid theories, such as, for example, the one discussed in Schroeder
(2009: 268–71).
10
What about normative thought? One view that fits naturally is that whenever we
believe that something has a normative property, we engage in something like a private
mental act of commendation (if there are such things). Other intriguing ideas that I wish
I could explore here are (i) that of explaining why having a normative belief entails mak-
ing a commendation by appeal to the idea that a belief counts as a normative belief only
if it is also a commendation; (ii) a related thesis about concept mastery that a person
226 Chris Heathwood
9.2.3 Why “Commending” and “Condemning”?

I state the thesis using the somewhat archaic language of “commending”
and “condemning.” Why these terms?
One way to put the guiding thought of the theory is that attributions of
normative properties involve us in a kind of practice. One way to character-
ize the practice is as one of evaluation. This term, however, can make the
theory sound vacuous, as the theory can then be put as the view that evalu-
ative properties are properties the attributions of which are evaluations. And
there may be a temptation to hold that evaluations themselves are simply
attributions of evaluative properties. I’m not sure the temptation to char-
acterize evaluation in this way is justified, but we can sidestep the issue by
choosing a different practice, or at least a different term. Thus, I’m looking
for a term that stands for a practice that can occur in contexts other than
the attribution of normative properties. In this way it would be a practice
that we have some independent familiarity with and grasp of. Since the
theory is supposed to shed some light on the nature of normative proper-
ties, it is helpful if our understanding of the phenomenon that is acting as
the explanans not be wholly parasitic on the phenomenon it is called into
service to elucidate.11
I would also like to find terms that can cover all the different kinds of
(thin) normative properties that we attribute: a term that covers evalua-
tive properties, as in “It’s good to be loved,” along with narrowly norma-
tive properties, as in, “You ought to go”; a term that covers attributions to
objects in different ontological categories, such as actions, states of affairs,
propositions, and people; a term that covers mild as well as severe normative
strengths; and a term that covers non-verdictive—that is, prima facie or pro
tanto—normative judgments.
“Commend” and “condemn” do this reasonably well, though perhaps
not perfectly. “Commend” is quite natural for evaluative statements. As
Hare notes, the OED characterizes “good” as “the most general adjective of
commendation” (1952: 79). “Commend” is less natural for narrowly nor-
mative statements. If I say, “you ought to go,” it would be more natural to
say that I am recommending that you go rather than that I am commending
your going. But I think that if we think about it, we will agree that we are
commending something whenever we are recommending it. We are prais-
ing it, applauding it, taking our hat off to it, giving it a thumbs up.
qualifies as grasping some normative concept only if he is capable of engaging in these

mental commendings; and (iii) to what extent this can help non-naturalism explain nor-
mative concept acquisition.
11
I return to this in section 9.3.2.
“Condemn” brings with it an additional complication. It may not be

quite the opposite of “commend,” since it may imply a certain severity of
criticism.12 It is also not clear that it can correspond to non-verdictive judg-
ments. I therefore choose “condemn” with the conditional proviso that if in
fact “condemn” does not properly apply to the attribution of milder norma-
tive properties (such as in, “He has some reason not to want that”), then
I stipulate a wider sense for it, for the purposes of the theory, a sense that
makes “condemn” the literal opposite of “commend.”13
Although “commending” and “condemning” seem to me to do a well
enough job at filling these bills, I am not wedded to them. What I am wed-
ded to is explaining the nature of normative facts by appeal to the speech
acts, beyond description, that asserting these facts necessarily involves us in.
If it turns out that “commending” and “condemning” are not adequate, I’m
hopeful either that some other terms are better, or that we can understand
the phenomenon I have in mind well enough—especially in light of the
present discussion—even if no term of English happens to be just right
for it.
9.2.4 Refining the Thesis

So far we have been working with the “basic idea” of the theory. We can put
that as follows, giving it a name now:
NP1: Normative properties are those such that, to attribute one to
something is, due to the nature of the property, necessarily to com-
mend or condemn that thing.
Note that this amounts to a necessary and sufficient condition—being
essentially commendatory is both necessary and sufficient for being a nor-
mative property (some earlier formulations may have suggested only the
necessary condition). This basic idea faces some potential counterexamples.
9.2.4.1 Unknowing Attributions of Normative Properties

Suppose your favorite property is, appropriately enough, intrinsic good-
ness, although I don’t know this. You tell me that a certain thing has your
favorite property. I report this to a third party, though, again, I don’t know
what property I am attributing. In reporting this to the third party, have
12
Thanks to Guy Fletcher here.
13
Thomson (2008: 54, 77) uses the unfamiliar term “dispraise” to describe what we
are doing when we call something bad. This term might, for my purposes, work just as
well as “condemn.”
228 Chris Heathwood
I attributed intrinsic goodness to the thing? It would seem so. Have I com-
mended this thing? Not obviously so.
Since perhaps it is also not obvious that I have not commended the
thing (there is independent reason to think that we can commend without
knowing it), consider another example. Suppose there is a race of rational
creatures spying on us from another planet. They become interested in
a certain use of our word “good” (when it is used to attribute intrinsic
goodness to things). They have no idea what the word means or what
phenomenon it signifies, but they are able to see that it is a predicate, and
thus suspect that it stands for some property. A whimsical member of their
community proposes that they incorporate this meaning of “good” into
their language, with the stipulation that whenever one of them applies it
to something, one attributes to this thing the same property, whatever it
is, that we humans are attributing when we apply it to something.14 Next,
suppose that after some time, certain confused members of this alien race
begin to believe that they have some insight into the nature of the property
this word expresses, and so begin genuinely to believe, of certain things,
that these things have this property. When they say that certain things
have the property, they would seem to be attributing intrinsic goodness
to it. But when they do this, are they thereby commending these things?
The pull to answer “No” in this case of community-wide ignorance may
be made even stronger if we stipulate that these aliens themselves have
no conception of value and, further, have no practice of commending or
condemning (although some may wonder whether these additional stipu-
lations make for a genuinely possible case).
This example might refute NP1. But I don’t believe it calls for wholesale
abandonment of its general idea. Rather, we can use the insight the example
provides to devise a better formulation of the general idea. Consider
NP2: Normative properties are those such that, to attribute one
knowingly to something is, due to the nature of the property know-
ingly attributed, necessarily to commend or condemn that thing.
In order to attribute a property knowingly to something, one has to know
which property one is attributing. This requires some degree of grasp of
the property. If you don’t “get” normative reasons or intrinsic value, you
can still attribute them to things, by using words learned from others who
do get it. One can “latch onto” these properties without understanding
them, as the aliens did in the example above, but one cannot attribute them
knowingly to things without understanding them. I set aside the question
of just what level of understanding of the property is required, other than
14
This example is similar to a case in Eklund (2013: §3).
to say that we probably don’t want to require perfect grasp—perhaps that

never happens—and likewise don’t want the requirement to be so lax that
the attributors in the cases above count as grasping.
NP2 is very much of a piece with NP1. The basic idea of my view is that
normative properties get us into the business of performing certain speech
acts. But of course they don’t do this completely on their own. We need to
meet them partway, by getting ourselves into a certain relation with them.
NP1 had it that all we have to do is attribute the properties. But attribution
is cheap, and the examples above suggest that more is required. We have to
know what we’re attributing in order for the properties to be able to turn
our attribution into a commendation or condemnation.15
9.2.4.2 Disjunctive and Comparative Properties

A second kind of counterexample is based on problems concerning certain
kinds of normative properties—in particular, disjunctive normative proper-
ties and comparative normative relations.16 I group them together because
they may admit of a single solution.
Consider this remark: “This is either good or bad, though I don’t know
which.” Maybe no attribution of a normative property takes place here;
perhaps that happens only when the speaker takes a stand as to which
it is, good or bad. If so, then there is no counterexample. But it’s also
possible that there are disjunctive properties, and, further, that a disjunc-
tive property each of whose disjuncts is a normative property is itself a
normative property. If so, then the speaker of this remark does knowingly
attribute a normative property, the property of being either good or bad.
But it doesn’t seem that the speaker is either commending or condemning
anything.
Being built up out of other normative properties, disjunctive normative
properties are non-basic. The simplest solution is thus to restrict the thesis
to one about basic normative properties, as follows:
NP3: Basic normative properties are those such that, to attribute one
knowingly to something is, due to the nature of the property know-
ingly attributed, necessarily to commend or condemn that thing.
15
An alternative possible way to deal with such cases—though perhaps it amounts
to the same view in the end—is to require that the property be attributed directly, as
discussed in Roberts (2013). The problem cases above would be cases of indirect property
attribution.
16
The problem concerning disjunctive normative properties was brought to my atten-
tion by Matt Chrisman. Several audience members, including Noah Lemos, have raised
the worry about comparative judgments.
230 Chris Heathwood
Since being either good or bad is not a basic normative property, NP3
avoids the implication that saying, “This is either good or bad, though
I don’t know which,” is to commend or condemn something. Nor does
NP3 leave the nature of these non-basic normative properties mysteri-
ous, since non-basic normative properties are, by definition, analyzable
in terms of the basic normative properties, properties whose nature NP3
elucidates.
Next consider comparative normative judgments, such as that it’s bet-
ter to suffer a paper cut than a migraine. To state this fact may not be to
attribute a normative property to something, but surely the normative
relation attributed is something that the general approach here should
want to shed light on. One plausible way for the theory to do this is to
assimilate the case of these comparative normative assertions to the dis-
junctive case above, and hold that comparative normative relations—such
as in our example above—are non-basic, and reducible to absolute,
non-comparative, normative properties. This approach requires no revi-
sion to NP3.
To illustrate, we might say that “x is intrinsically better than y” means
that x has a certain intrinsic value, n; y has a certain intrinsic value, m; and n
is greater than m (where “n” and “m” range over real numbers). Claims such
as that x has an intrinsic value of n will correspond to commendations when
n is positive and condemnations when n is negative. (When n is zero—that
is, when we say that something has no intrinsic value—no normative prop-
erty is attributed.)
Another promising strategy is to hold that such utterances involve speech
acts that are the comparative analogs to commending and condemning.
Thus, to say that it’s better to suffer a paper cut than a migraine is to com-
mend paper cuts relative to migraines (it may also be to condemn migraines
relative to paper cuts). Judith Thomson accepts a view like this about bet-
terness relations. She holds that when we say, “Smith is a better chess player
than Jones,” we praise Smith “relative to Jones” (2008: 61).17
17
Another potentially problematic case is that of rights claims (thanks to Daniel
Wodak for raising this point). The claim that fetuses have a right to life is surely a norma-
tive claim. Are we commending fetuses when we say this? Maybe. Note that, instead of
saying that fetuses have rights, some people mean to convey more or less the same idea by
saying that fetuses have intrinsic value, and this claim seems commendatory. Note also
that we seem to be positively evaluating fetuses if we claim that they have rights, and so
we are engaging in the kind of speech act I am ultimately after here (even if “commend-
ing” isn’t a perfect word for it (see section 9.2.3)). A final point here is that rights claims
may be equivalent to certain claims about obligations. If so, then the fact that some being
has a certain right just is the fact that it is wrong to treat this being in certain ways; and
wrongness is straightforwardly covered by the theory.
9.3 THREE OBJECTIONS

9.3.1 Commending the Bad
Having presented and refined the theory of normative properties as
essentially commendatory properties, I would now like to address three
important objections. The first is similar to a familiar concern for both
non-cognitivism and motivational judgment internalism. Imagine a cadre
of devils interested in discovering what would we be bad precisely to do it.
One devil says to another, “I recommend that you do this, since it would
be very bad indeed.” My theory commits me to saying that, in attributing
badness to this act, the devil is condemning it.18 But in fact he is attributing
badness to it precisely to commend it.
A familiar response to this kind of case maintains that the devil isn’t really
saying that the act would be bad to do, but is instead using an “inverted
commas” or “scare quotes” sense of “bad” (Hare 1952: 124–5; Smith
1994: §§3.3–3.4). According to this idea, the devil doesn’t really judge that
the act in question would be bad—he’s recommending it after all. What he
is really saying is something like this: “I recommend that you do this, since
it would be what most people call ‘very bad’.” Since such a remark does not
involve the devil in attributing actual badness to anything, if this is what his
original remark really means, it would be no counterexample to the theory.
The “inverted commas reply” is an interesting strategy for non-cognitivists
and motivational judgment internalists, but it is a non-starter for normative
realists who want to accept the view of normative properties as essentially
commendatory. Non-cognitivists don’t believe in normative properties.19
Their account of normativity locates it in our language and thought rather
than in the world. But my theory is for those who believe in properties that
are themselves normative. If these properties are real, they are there for the
devil to learn about, and knowingly attribute to things. And that is just
what he has done in the example.
Nonetheless, I don’t believe that the objection ultimately succeeds. A
plausible case can be made for the view that the devil is in fact condemn-
ing the action he knows is bad. We can begin by noting that any asser-
tion, whether in language naturalistic or normative, can be used to perform
18
The theory as formulated (NP3 above) doesn’t strictly speaking imply this. To gen-
erate the implication, we need to make explicit what was surely already implicit: that to
attribute positive normative properties is to commend and to attribute negative normative
properties is to condemn.
19
I cannot discuss the alleged possibility that non-cognitivists might believe in nor-
mative properties and facts.
232 Chris Heathwood
almost any kind of speech act, given the right conventions and context. In
particular, if you know that your audience is interested in finding some-
thing with a certain feature, you can commend or recommend to them
something simply by pointing out that it has this feature. This holds even if
the feature is badness. But the theory of normative properties as essentially
commendatory is compatible with this. The theory describes one way that
we can commend or condemn, but allows for all manner of other ways that
this can occur, such as the way just described. Thus, while it is obvious that,
in pointing out that the act is bad, the devil is thereby recommending it,
this fact is in no tension with the theory. What is incompatible with the
theory is the claim that, in pointing out that the act is bad, the devil is not
also thereby condemning it. The objection may be implicitly assuming that
if one is commending something by describing it in a certain way, one can-
not also be condemning it by describing it in that way. But such an infer-
ence has not been justified.
So instead of deriving as a lemma the claim that the devil is not con-
demning the act by pointing out that it is bad, the objection must just
assert this as a premise. Against this, the theory of normative properties as
essentially commendatory must maintain that the devil involves himself in
a kind of conflict of speech acts. On the one hand, he is recommending the
act in calling it bad, since his audience is interested in finding an act that
would be bad to do. On the other hand, he is also condemning the act,
since he has said sincerely that it would be bad to do.
My defense of the idea that the devil is in fact condemning the act has
two parts, one negative, one positive. The negative part exposes a poor rea-
son for thinking that the devil is not condemning the act. According to this
thought, the devil must not be condemning the act in question because
the devil has no disfavorable attitudes towards it. But, as discussed earlier,
a person can genuinely commend or condemn without having the corre-
sponding attitudes, just as a person can genuinely apologize even if he’s
unable to feel remorse. Sympathy for the devil objection may be rooted in
this mistaken view of commending and condemning.
More positively, there are reasons to think that the devil is in fact con-
demning the act in pointing out that it would be bad to do. Here is a
simple argument for this. To say that an act would be bad to do is to say
something bad about it. To say something bad about an act is to (verbally)
evaluate it negatively. To (verbally) evaluate an act negatively is to condemn
it.20 These intuitively plausible principles imply that the devil has indeed
20
I include the term “verbally” because it is possible to negatively evaluate an act just
in thought, and it’s not clear whether this is a kind of condemnation. See footnote 10.
condemned the act that he has said would be bad to do. Note that this
argument does not presuppose my theory. Those who reject the theory
of normative properties as essentially commending properties can accept
the argument. Consider, for example, the view that it is nothing about the
property of badness itself that makes attributions of it condemnations, but
something about our mode of representing or expressing this property that
makes attributions of it condemnations (the common analogy with slurs is
helpful here). This naturalist-friendly theory can agree with the plausible
idea that to say that an act would be bad to do is to say something bad about
it, to evaluative it negatively, and to condemn it.
Finally, it may be helpful to note that similar speech act conflicts occur in
other contexts. Judith Thomson, who defends views about attributions of
goodness that are in some ways similar to mine, gives the following example:
We have to grant in any case that it is possible to both praise and dispraise a person
in saying some words about him. If I am a professor of mathematics, and my letter
of recommendation for my graduate student for a teaching position at Greatorex
University consisted entirely of the words “He is good at doing arithmetic,” then
I have both praised and dispraised the student. I have praised him, since writing “He
is good at doing arithmetic” is praising him. But the context in which I wrote those
words makes it the case that I also dispraised him.
(Thomson 2008: 56)
Similarly, the devil has both praised and dispraised the act. The devil dis-
praised it, since saying, “It would be very bad” is dispraising it. But the con-
text in which he said those words makes it the case that he also praised it.
9.3.2 An Unhelpful Tautology

According to another objection, the theory of normative properties as
essentially commendatory sheds no light on normativity, as it is supposed to
do, because it is covertly tautologous. It is covertly tautologous because the
best account of what it is to commend something is that it is to attribute a
positive normative property to it. My view would thus ultimately be saying
no more than that the normative properties are those such that when you
attribute one to something, you can’t help but be attributing a normative
property to it.
But the account of commending on which this objection relies is doubt-
ful. For we often commend without attributing normative or evaluative
properties to things. I might commend a bicycle simply by pointing out that
it is made of carbon, but being made of carbon is not a normative property.
And we can commend without attributing a property at all. When someone
says, “I commend you for your efforts,” they are, as Austin would say, not
234 Chris Heathwood
reporting on a commendation, but indulging in one (Austin 1962: 5). We

can also commend by giving a thumbs-up or a high-five. A ballplayer might
commend his teammate out of habit by giving him a high-five without hav-
ing come to the view that the teammate did anything good.
These cases might be dealt with by understanding or formulating the
target theory of commendation in a less restrictive way. The account could
hold that to commend is to attribute a normative property either explicitly
or implicitly.21 Perhaps all the cases I cite above are cases of implicit nor-
mative property attribution. I have doubts about that, but I don’t think
it matters, since another case shows pretty clearly that one can commend
without even implicitly attributing a normative property. This is the case of
the committed normative nihilist. We can imagine such a person engaging
in the business of commending while simultaneously making her nihilism
conspicuous. Upon receiving a request for advice, she might say this:
“As you know, I’m a normative nihilist: I don’t believe that anything has any norma-
tive properties. And so in particular, I don’t believe that there is anything that you
ought to do. Nonetheless, I’m happy to give you some advice: I recommend that
you donate $20.”
The nihilist is commending a certain course of action, but she appears to
have succeeded in refraining from attributing any normative properties to
it, even implicitly.
9.3.3 Might Some Natural Property Be Essentially

Commendatory?
Our final objection targets the sufficiency claim of the theory. NP3 says
that being essentially commendatory is both necessary and sufficient for
being a basic, irreducibly normative property. But might some natural,
non-normative property also be essentially commendatory, in conflict
with NP3?
It of course won’t do to find some property that plays this role contin-
gently, as many natural properties do on account of our interests or conven-
tions. But might some natural property, presumably one that is intimately
related to the very phenomenon of commendation itself, be essentially such
as to play this role? I have claimed, and indeed relied upon the fact, that
commendation is a natural phenomenon. So let “N” stand for a natural
property—perhaps a rather complex one—the instantiation of which is suf-
ficient for a commendation to occur. Might this property be what we are
21
I am grateful to Gunnar Björnsson here.
looking for: one such that, to attribute it to something is, due to the prop-
erty’s nature, necessarily to commend that thing?22
No, for there is a difference between a property (i) being such that if it
obtains, a commendation has occurred, and (ii) being such that if someone
attributes it to something, a commendation has occurred. (i) is true of N,
but the objection requires a property of which (ii) is true. To illustrate, the
state of affairs involving the nihilist saying, “I recommend that you donate
$20,” is sufficient for a commendation to occur. But my reporting this
natural fact—my saying, “The nihilist said, ‘I recommend that you donate
$20,’ ”—is not itself to commend or recommend anything. Whatever natu-
ral property I attribute in stating this natural fact, while it is a natural prop-
erty the instantiation of which is sufficient for a commendation to occur,
is not a natural property such that, to attribute it to something is itself to
make a commendation.
The objector here is looking for a natural fact such that merely stating
this fact is sufficient for a commendation to occur. What about the nihil-
ist’s utterance itself: “I recommend that you donate $20”? Making such
utterances is sufficient for making a commendation. However, these “per-
formatives” are not the reportings of facts; they are not true or false (Austin
1962: 6). Since no property is attributed, a fortiori no essentially commend-
atory natural property is attributed.
But perhaps some variation on the canonical performative form will pro-
vide an example of a kind of utterance that is at once descriptive and essen-
tially commendatory. If I say, “I commend you for your efforts,” or simply,
(C1) “I commend you,”
I have not described my commending you; I have done it. This is like the
nihilist’s performative. But suppose I say,
(C2) “You are hereby commended.”
(C2) is certainly a commendation, at least ordinarily, just as an utterance
of, “You are hereby warned not to come closer,” is a warning. But is (C2)
also a description, and hence a property attribution, where the property in
question is the property of being commended?
I find this hard to decide. (C2) may be just another way of saying (C1),
in which case it, too, would not be true or false. On the other hand, its
grammatical form seems rather descriptive, for it is identical in form to the
following:
22
For ease of presentation in what follows, I omit the qualification that the attribution
must be knowing.
236 Chris Heathwood
(S2) “You are hereby spoken to.”

This remark, however, is grammatically odd—it sounds like a funny hybrid
of a performative and a description. We may therefore not want to draw any
conclusions from it. We can repair it, however, as follows:
(S3) “You are hereby being spoken to.”
(S3) is perfectly grammatical, and, when uttered to someone, is always
true. It is thus a description. So let’s consider its potentially-commendatory
cousin:
(C3) “You are hereby being commended.”
(C3) is as clear a case of a description as (S3). But is (C3) true, whenever it
is sincerely spoken to someone? That is, must a sincere utterance of (C3) be
a commendation, just as a sincere utterance of (C1) is always a commen-
dation? I think this is also hard to decide. Must we be thanking someone
anytime we sincerely tell him, “You are hereby being thanked”? Is that even
ever a thanking? How would you feel if that was the thanks you got?
Even if we decide that any sincere utterance of (C3) must be a com-
mendation, the objection to NP3 faces another problem, this one more
decisive. The objection aims to show that a certain natural property, the
one attributed in (C3), is essentially commendatory. What property is this?
One natural thought is that it is a certain relation, the relation that obtains
between an utterer, an utterance, and an utteree just in case the utterer is
commending the utteree by means of the utterance. Call this relation “C.”
The problem is this: not all attributions of C are commendations, and thus
C is not essentially commendatory. That not all attributions of C are com-
mendations is shown by the fact that one can attribute C to someone else, as
in, “Look over there: Mabel is commending Abel by means of the utterance
she is making.”23
A natural rejoinder to this reply is to hold that the indexical elements in
(C3) are essential to its being a commendation. The objector could thus
hold that whenever I say to someone,
(C3) “You are hereby being commended,”
I am self-attributing C, and it is self-attributions of C that are essentially
commendatory. That might be true, but this doesn’t refute NP3, which
says nothing about self-attributions. To self-attribute natural property C
23
Strictly speaking, I should say that one can attribute C, not to someone else, but
to a certain set of things, perhaps an ordered triple, consisting of utterer, utterance, and
utteree.
may necessarily be to make a commendation, but refuting NP3 requires

finding a property such that merely attributing it is necessarily to make a
commendation.24
When I was first thinking about this problem for NP3,25 I suspected
that we very well might find a case of an essentially commendatory natural
property. I was thinking that the theory would then need to be weakened
to a mere necessary condition. A weakened theory would still, I believe,
reveal an interesting and substantive facet of the essence of irreducible
normativity—even if a certain natural property shared this facet, too. Such
a theory would also still distinguish normative properties from other puta-
tively non-natural properties—such as modal, mathematical, and logical
properties—since none of these is a candidate for being essentially com-
mendatory. It would have had the drawback that some of the claims in
the next section, about further work the theory can do, would have to be
qualified. But, in the final analysis, no such qualifications seem necessary.
It seems that no natural property is such that, to attribute it to something
is, due to the nature of the property, necessarily to commend or condemn
that thing.26
24
To avoid testing the reader’s patience, I have spoken somewhat loosely here. In
uttering (C3), I am not strictly speaking self-attributing C, but rather putting myself in
the “utterer slot” of C, my listener in the “utteree slot,” and (C3) itself in the “utterance
slot.” This isn’t self-attribution because the relation is attributed to a group—perhaps
an ordered triple—of people and things, of which I am a member. We might call this
“us-attribution.” But whatever we call it, the problem remains: we may have succeeded
in finding a natural property such that anyone who “us-attributes” it to a certain ordered
triple thereby commends a certain member of that triple, but this is no counterexample
to NP3, which says nothing about us-attribution.
25
Versions of which have been suggested by Justin D’Arms, Mike Ridge, Brian
Tackett, and Mark Heller.
26
Two final points. First, if non-reductive naturalism is true, there very well might be
natural properties that are essentially commendatory, namely, the normative properties
(which, on this view, are natural properties). (I say “might” because it isn’t certain: it
may require that these properties be essentially inexpressible in non-normative terms,
something that in fact strikes me as doubtful.) But I believe this possibility is dialecti-
cally irrelevant. My project is (mainly) to assume non-naturalism, and then defend an
account of the nature of irreducible normativity. So I am taking it for granted in this
section that non-reductive naturalism is false. (And note, for what it’s worth, that even
if non-reductive naturalism is true, there is still an apparently true thesis that corre-
sponds to my thesis in this section, namely, that no non-normative property is essentially
commendatory.)
Second, and relatedly, one might be tempted by the thought that if reductive natu-
ralism is true, then there will be a natural property that is necessarily commendatory,
namely, the natural property that, according to the true form of reductive naturalism, is
identical to a certain basic normative property. In reply to this, I repeat the point about
dialectical irrelevance. But I also note that it doesn’t even seem true. If we attribute this
property under its naturalistic guise, or using naturalistic vocabulary, it is not plausible
to think that such attributions must always be commendations. Even if pleasantness is
238 Chris Heathwood
9.4 FURTHER WORK THE HYPOTHESIS CAN DO
The view of normative properties as essentially commendatory not only helps

non-naturalists explain the nature of these properties, it does further interest-
ing work as well: (a) it provides for an account of the “queerness” of normative
properties, one superior to other accounts; (b) it explains why reductionism
fails, in a way friendly to non-naturalism (as opposed to non-cognitivism);
and (c) it can help deflect arguments against non-naturalism from the
“action-guidingness” or “essential practicality” of normativity.
9.4.1 Characterizing Queerness

The theory of irreducibly normative properties as essentially commendatory
makes for an alternative, and I believe superior, characterization of the sense
in which irreducibly normative properties are unusual. J. L. Mackie famously
declares normative properties to be unacceptably “queer” (1977: 38–42).
One central aspect of this queerness, Mackie suggests, is their power to moti-
vate anyone acquainted with them (1977: 40). Mackie is here implying that
non-naturalists about value may be committed to some kind of motivational
judgment internalism. But, as many point out, this isn’t so. Non-naturalists
can, and many do, plausibly reject motivational judgment internalism.27
Nonetheless, irreducibly normative properties, if they exist, do seem to have
some kind of “oomph” lacked by ordinary properties. In my view, it is not their
ability to make us comply with the facts in which they figure, or to motivate
us even slightly to do so. It is instead their ability to make anyone who merely
attributes one knowingly to something thereby to be engaging in a practice that
goes beyond mere description—the practice of commending or condemning. It
does this whether the attributor wants to be doing so or not. It can indeed seem
rather odd that simply reporting, as disinterested scientists do, that some object
in the world has some property, can force one into the business of making an
evaluation, but this, according to the view here, is what normative properties
do. And this is a reasonable sense in which they are queer.
What is queer is not the notion that using normative language involves one
in making recommendations, commendations, etc. I’m not sure any meta-
ethical view would deny that normative language can do this. What is queer is
that, when it comes to irreducibly normative properties or facts, one can’t help
but get oneself involved in the business of making a recommendation simply
identical to goodness and we all know this, I still don’t believe that to describe something
as pleasant must always be to make a commendation.
27
See, e.g., Shafer-Landau (2003: ch. 6).
by stating a fact.28 It is that these properties are simultaneously descriptive and

evaluative: the facts involving them simultaneously describe and evaluate the
world, so to speak, and thus get us to engage in acts of evaluation merely by
describing things. This account of normativity’s queerness supports Mackie’s
contention that these properties or facts “would be entities … of a very strange
sort, utterly different from anything else in the universe” (Mackie 1977: 38).
It is not our business here to investigate whether irreducible normativity’s
queerness is enough, or even any, reason to doubt its existence. I am not here
arguing for non-naturalism but addressing an objection to it, the objection
that non-naturalism has no plausible account of what these non-natural,
irreducibly normative properties are supposed to be. As this section illus-
trates, the account offered is an account that even those who would reject
non-naturalism can accept, and may even welcome, as it helps underwrite an
argument against non-naturalism: the argument from queerness. Mackie’s
argument from queerness (or the strand of it we are focusing on here) seems
to many to fail, because it falsely encumbers non-naturalism with some-
thing like motivational judgment internalism. If the account of queerness
provided by my view is correct, however, opponents of non-naturalism have
an argument from queerness that avoids at least this defect. The theory of
normative properties as essentially commendatory may thus be something
of a double-edged sword for non-naturalists, helping them to shed light on
the nature of their unseen postulates, but in so doing, revealing something
suspect about them. I don’t believe this is any reason to think that the theory
of normative properties as essentially commendatory is mistaken, though it
does admittedly pull against its original motivations.29
9.4.2 Explaining Why Reductionism Fails

Non-naturalists believe that attempts to reduce the normative to the natural
are bound to fail. That is, they believe that no identification of normative
properties or facts with properties or facts that we can express in naturalistic
language (or any non-normative language, for that matter) can succeed.
Why should that be?
28
The “irreducible” here is important. If reductive naturalism is true, one can
attribute normative properties without evaluating, by attributing them under their
naturalistic guise.
29
Michael Pendlebury rejects realism on the following grounds, and in terms that
strike me as suggestive of the view here: “One reason why I am inclined toward normative
expressivism is that I do not understand how a factual proposition could have the prop-
erty that anyone who is committed to it thereby takes a normative stance” (2010: 185).
Pendlebury may be rejecting realism because it is committed to a kind of queerness not
unlike the sort I am describing here.
240 Chris Heathwood
Some explanations for why reductive naturalism is bound to fail do not

fit comfortably with non-naturalism. In the context of explaining why
they believe that the open-question argument against reductive naturalism
is compelling, Darwall, Gibbard, and Railton offer one such explanation.
They claim that any naturalistic reduction will seem implausible due to
(i) the fact that “Attributions of goodness appear to have a conceptual link
with the guidance of action,” combined with (ii) “our seeming ability to
imagine, for any naturalistic property R, clear-headed beings who would
fail to find appropriate reason or motive to action in the mere fact that R
obtains (or is seen to be in the offing)” (1992: 117). In other words, reduc-
tive naturalist theories are bound to fail because, while normative prop-
erties are essentially action-guiding, no natural properties are essentially
action-guiding, where a property is essentially action-guiding just in case
attributions of it (or, more plausibly, certain special attributions of it) neces-
sarily imply some motivation to act on the part of the attributor.
This explanation, which essentially appeals to motivational judgment
internalism, should make non-naturalists uneasy. “For,” as Darwall, Gibbard,
and Railton point out, “it appears no easier to see how an appropriate link
to motivation or action could be logically secured if we were to substitute
… ‘sui generis, simple, nonnatural property Q’ for ‘naturalistic property R’ ”
(1992: 118). Darwall, Gibbard, and Railton conclude that non-cognitivism,
which can straightforwardly accommodate the action-guidingness they
posit, is “the real historical beneficiary of the open question argument,” or
of the untenability of reductive naturalism (1992: 119).
But if non-naturalists adopt the theory of normative properties as essen-
tially commendatory properties, they can provide an alternative explana-
tion for why reductive naturalism is implausible, one that is friendlier to
non-naturalism. Whereas Darwall, Gibbard, and Railton explain the unten-
ability of reductive naturalist theories by appeal to the alleged essential
action-guidingness of normative properties, non-naturalists can do so by
appealing to the essential “commendatoriness” of normative properties, a
feature that, I have been arguing, they have reason to believe in anyway. The
reason—or at least a reason—that normative properties can never be identi-
fied with any natural property is that, whereas normative properties are essen-
tially commendatory, no natural property is. No natural property is such that
any possible attribution of it to something is also a commendation.30
30
This is one of the remarks that would need qualification if I am wrong that no natural
property is essentially commendatory (see section 9.3.3). The qualification would be that
the recherché natural property, attribution of which is sufficient for a commendation—the
property I called “C” above—has no promise at all as a reduction base for normative prop-
erties. Some other claims in this section could be similarly qualified.
9.4.3 An Account of the “Essential Practicality”

of Normativity
In a more recent paper, Stephen Finlay describes another common com-
plaint against non-naturalism: “Other objections to sui generis normative
facts and properties proceed from the thought that there is something
essentially practical about normativity” (2010: 336). This objection is
implicit in Darwall, Gibbard, and Railton’s remark about how a prop-
erty’s being non-natural and sui generis makes it no more essentially
action-guiding than any ordinary natural property. To indulge in some
serious metaphor, here may be one way to put the idea of essential practi-
cality: normative claims are not simply inert, sterile descriptions of reality,
as naturalistic claims are; normative facts have a kind of “oomph” about
them, or perhaps a kind of “glow.” As Mackie puts it, they have a “to-be-
doneness” built into them (Mackie 1977: 40). Just what is this essential
practicality?
As Finlay notes (2010: 336), it “has often been understood in moti-
vational terms,” and is thus often taken to require the doctrine of moti-
vational judgment internalism, a doctrine not especially friendly to
non-naturalism. This is why the essential practicality of normative facts
looks hard to accommodate on non-naturalism, and on realism more
generally.
But if the theory of normative properties as essentially commendatory
is true, non-naturalists have an alternative way to account for the essen-
tial practicality of normativity, or to deliver what motivational judgment
internalism is often relied upon to deliver. The essential practicality of
normative properties lies in the fact that they make those who attribute
them knowingly to involve themselves in speech acts beyond mere attri-
bution. Normative properties’ “oomph” consists, at least in part, in the
fact that we can’t help but commend or condemn, just by attributing
them. Normative facts get us to do certain things—not to want to comply
with them if we merely believe them, but to commend or condemn if we
merely state them.
What’s more, explaining the essential practicality of normativity in
terms of the speech acts that normative assertions involve us in rather
than in motivational terms has independent advantages—not just for
non-naturalists but for other realists as well. Motivational judgment inter-
nalism is quite controversial; irrespective of its implications for metaethics
and considered on its own merits, it strikes many as empirically dubi-
ous. My explanans is, on its own merits, less doubtful. It is more plau-
sible that, in making a normative assertion, we must be commending or
242 Chris Heathwood
condemning, than it is that, in making a normative judgment, we must

have some motivation.31
9.5 CONCLUSION
My purpose has been to show that non-naturalists can answer the

complaint that they cannot tell us what the properties they posit are
like, other than either negatively or simply by repeating the normative
notions. I have claimed that a certain thesis is available to them that ena-
bles them to characterize irreducible normativity positively and substan-
tively in non-normative language. I have tried to clarify and refine the
thesis, and to defend it against objections. I have also outlined some of
the other work I believe it can do. I hope all of this gives non-naturalists
good reason to accept the thesis.
A certain kind of worry might be lingering: How could there be a
property that is like this? How could there be a property such that, due
to its special nature, knowingly attributing it to something necessarily
constitutes commending that thing? First, note that if my view is right,
this question really amounts to the question, How could there be nor-
mative properties? That’s a good question! Although normative facts are
among the most familiar of facts, it is not hard to get into a frame of
mind in which they seem astonishing, or impossible. So perhaps it is no
surprise if my account of the nature of irreducible normativity preserves
its strangeness, as indeed I emphasized earlier (section 9.4.1). Second,
although I might not be able to explain how there could be essentially
commendatory properties, it may help reduce our perplexity if we are
persuaded that the thesis might actually be true (on the assumption that
non-naturalism is true). If one is persuaded that the view has no clear
counterexamples, then one agrees that anytime anyone is knowingly
attributing an irreducibly normative property, that is sufficient for him
to be making a commendation. Since the making of the commendation
doesn’t depend on additional contingencies of context or convention, it
must simply be a result of what one is saying about the thing one is talk-
ing about: that is, it must be something about this property. To review all
this isn’t to explain how the view could be true, but appreciating that the
31
Judith Thomson, who is no friend of non-naturalism, agrees. She puts the point
in terms of favorable attitudes more generally rather than motivation in particular
(Thomson 2008: 54). For an overview of the controversy surrounding motivational judg-
ment internalism, see Björklund et al. (2012).
view does indeed seem true (if non-naturalism is true) may help reduce
one’s bafflement.32
References
Austin, J. L. 1962. How to Do Things with Words. Oxford: Clarendon Press.
Björklund, F., Björnsson, G., Eriksson, J., Olinder, R. F., and Strandberg, C. 2012.
“Recent Work: Motivational Internalism,” Analysis 72: 124–37.
Copp, D. 2009. “Realist-Expressivism and Conventional Implicature,” Oxford
Studies in Metaethics 4: 167–202.
Darwall, S., Gibbard, A., and Railton, P. 1992. “Toward Fin de siècle Ethics: Some
Trends,” Philosophical Review 101: 115–89.
Eklund, M. 2013. “Evaluative Language and Evaluative Reality,” in S. Kirchin (ed.),
Thick Concepts. Oxford: Oxford University Press, 161–81.
Finlay, S. 2007. “Four Faces of Moral Realism,” Philosophy Compass 2: 820–49.
Finlay, S. 2010. “Recent Work on Normativity,” Analysis 70: 331–46.
Green, M. 2009. “Speech Acts,” in E. Zalta (ed.), The Stanford Encyclopedia of
Philosophy (online Spring 2009 edn.).
Hare, R. M. 1952. The Language of Morals. Oxford: Oxford University Press.
Heathwood, C. 2012. “Could Morality Have a Source?” Journal of Ethics & Social
Mackie, J. L. 1977. Ethics. New York: Penguin.
Moore, G. E. 1903. Principia Ethica. Cambridge: Cambridge University Press.
Parfit, D. 2006. “Normativity,” Oxford Studies in Metaethics 1: 325–80.
Parfit, D. 2011. On What Matters, Volume Two. Oxford: Oxford University Press.
Pendlebury, M. 2010. “How to Be a Normative Expressivist,” Philosophy and
Roberts, D. 2013. “It’s Evaluation, Only Thicker,” in S. Kirchin (ed.), Thick
Concepts. Oxford: Oxford University Press, 78–96.
32
This essay has benefited from the feedback of quite a few people. I’d like to thank
Andrew Alwood, David Barnett, Gunnar Björnsson, Paul Bowman, Gwen Bradford,
Dan Brigham, Campbell Brown, Matt Chrisman, Christian Coons, Tim Crane, Justin
D’Arms, Matti Eklund, Aaron Elliott, David Faraci, Stephen Finlay, Guy Fletcher,
Mark Heller, Adam Hosein, Anthony Kelley, Uri Leibowitz, Noah Lemos, Hallvard
Lillehammer, Eden Lin, Don Loeb, John Maier, Kris McDaniel, Michael Pendlebury,
Jason Raibley, Mike Ridge, Jon Robson, Michael Rubin, Brian Tackett, Pekka Väyrynen,
Preston Werner, Eric Wiland, Daniel Wodak, and two anonymous referees for Oxford
Studies in Metaethics. Special thanks to Gunnar Björnsson, who was my commenta-
tor at the 2013 Central APA. I also thank audiences at the 2012 Madison Metaethics
Conference, the 2012–13 Princeton Workshop in Normative Philosophy, the 2013
Central APA, the University of Cambridge, the University of Nottingham, the University
of Edinburgh, Syracuse University, and the Center for Values and Social Policy at the
University of Colorado Boulder. Most of the work on this essay was completed during a
fellowship at the University Center for Human Values, Princeton University; I thank the
Center for that support.
244 Chris Heathwood
Schroeder, M. 2009. “Hybrid Expressivism: Virtues and Vices,” Ethics 119: 257–309.

Searle, J. 1979. Expression and Meaning. Cambridge: Cambridge University Press.
Shafer-Landau, R. 2003. Moral Realism. Oxford: Oxford University Press.
Thomson, J. J. 2008. Normativity. Chicago: Open Court.
Williams, B. 1995 [1989]. “Internal Reasons and the Obscurity of Blame,” in
Williams, Making Sense of Humanity. Cambridge: Cambridge University
Press, 35–45.
10
How to Be a Moral Platonist
Knut Olav Skarsaune
This chapter will develop a form of non-natural realism about value, arrived
at by way of reflection on the so-called “supervenience objection” to the
view. Though I will argue the objection fails as an objection, I think it is a
first-rate challenge; one that forces us to come back with a clearer and hope-
fully better statement of our position.
The supervenience objection has been discussed and developed so widely
that it is probably quixotic to try to give a statement that will satisfy every-
one. One reason to hope many will find their favourite version addressed
is that I take up, not one, but two objections. The bulk of the chapter
responds to Simon Blackburn’s (1971, 1985) challenge, which is to explain
why it is analytic that normative properties supervene on descriptive prop-
erties. In the last section, I take up a more recent, metaphysical challenge,
which asks how normative properties can be fundamental and at the same
time supervene on other properties. Does not that violate Hume’s Dictum,
the ban on “necessary connections between distinct existences”?
The heart of my response to Blackburn is an idea I got from Kit Fine.1
It is that normative predicates express subtly different senses when they are
applied to particular things and to kinds, respectively. The kind-applying
senses are basic and primitive, while the particular-applying senses are
defined in terms of them. A particular thing is goodpar, for example, just in
case it is a token of a goodkin kind. But then, since two descriptive twins are
tokens of exactly the same kinds, they must either both be goodpar or both
not be. Supervenience falls out of the definitions of the particular-applying
normative concepts.
This account is a cognitivist analogue of R. M. Hare (1952) and Allan
Gibbard’s (1990) non-cognitivist accounts. They also analyse ascrip-
tions of a normative predicate to a particular in terms of a general com-
mitment (in the case of Hare, a universal prescription, and in the case of
1
In discussion. Many thanks to Fine for the idea; all mistakes, of course, are mine.
246 Knut Olav Skarsaune
Gibbard, endorsement of a norm). My account applies the same structure

in a truth-conditional framework. In honour of Hare, I call it cognitive
universalism.
I do not try to motivate cognitivism over non-cognitivism in this chap-
ter. My main message is that cognitivists of all stripes—even Moorean
non-naturalists—can explain the analyticity of supervenience in the same
elegant way that Hare and Gibbard have shown us.
In the last part of the chapter I turn to metaphysics. I argue that
non-natural realists should just repeat, at the level of properties, the expla-
nation of supervenience that cognitive universalism provides at the level
of concepts. Just as they should say that the basic normative concepts
apply to kinds, so they should say that the basic normative properties are
second-order; they do not apply to particular things, but rather to kinds,
or perhaps to properties. I argue, in other words, that non-natural realists
should be platonists.
10.1 BLACKBURN’S CHALLENGE
Blackburn’s challenge is to explain why normative judgments exhibit a

peculiar combination of independence and dependence vis-à-vis descriptive
judgments. Let us start with the independence. From G. E. Moore (1903)
onwards, non-natural realists have said that at least some normative con-
cepts are primitive, indefinable, basic. Perhaps some of them can be defined
in terms of others, perhaps they form an interdefinable cluster, but no nor-
mative concept can be defined in entirely non-normative terms. Prominent
candidates for primitiveness include good and bad (Moore); being a rea-
son for (Parfit, Nagel, Scanlon); and ought (Broome, Wedgwood). I want
to be neutral on this issue, and will switch liberally between different nor-
mative concepts in my exposition.
The independence at issue is more than primitiveness. red, for example,
could be primitive and yet not have the relevant kind of independence from
descriptive concepts, being one of them itself. But non-natural realists fur-
ther maintain that normative concepts are different in kind from descriptive
concepts. Descriptive concepts are used to judge how things are; normative
concepts are used to judge how they should be; and these are different kinds
of judgment.
The upshot is that descriptions never entail evaluations. There is no
description D such that
(1) a is D.
How to Be a Moral Platonist 247
entails
(2) a is good.
Or as Blackburn puts it: ‘There is no moral proposition whose truth is
entailed by any proposition ascribing naturalistic properties to its subject’
(1993: 116).
So far independence. But there is also a close connection between nor-
mative and descriptive judgment. As far as I know, every prominent non-
natural realist accepts that it is in some sense a conceptual truth that things
have their value in virtue of being the way they are descriptively, and that
value therefore supervenes on descriptive properties.2
Speaking now in my own voice, I believe the best way to make this pre-
cise is as follows. What I want to say is not that any grand supervenience
principle is itself analytic, but rather that ordinary English sentences and
inferences that exemplify certain principles are analytic.3 Keeping in mind
throughout that descriptive likeness includes both properties and relations,
I take it that the inference from
(2) a is good.
(3) b is descriptively exactly like a
to
(4) b is good.
is analytic, and likewise
(5) If Ted is good, then it is impossible to be just like Ted in every descrip-
tive respect and not be good.
Sentence (5) exemplifies
2
Notice that Blackburn uses “naturalistic” to describe the supervenience base. While
I have no problem with this, it tends to invite needless distractions having to do with
divine command theory. I therefore use “descriptive” instead. In order to head off
Sturgeon’s (2008) worries, let me stress that I am not referring to the properties picked
out by this or that descriptive vocabulary. “Descriptive property” is meant to convey
a metaphysical notion, probably primitive, but sometimes elucidated with the phrase
“ways things can be”.
3
A sentence “exemplifies” a principle just in case its negation contradicts the princi-
ple; an inference exemplifies it just in case the premises together with the negation of the
conclusion contradict it. Of course, in this sense the supervenience principles exemplify
themselves, but as we shall see, they are not “ordinary English sentences”, hence they fall
outside the analyticity claim.
Strong normative-descriptive supervenience: for all possible x, y: if x and

y are alike in every descriptive respect, then x and y are alike in every
normative respect.
while the inference from (2) and (3) to (4) exemplifies both strong and
Weak normative-descriptive supervenience: for all x, y in the same pos-
sible world: if x and y are alike in every descriptive respect, then x and
y are alike in every normative respect.
Strictly speaking, the principles themselves are not analytic, because they
are not true in virtue of the meanings of the words that figure in them alone;
perhaps they are in some sense true in virtue of the meanings of “good”,
“bad”, etc., but these words do not figure in the principles. There, rather,
we find the term of art “normative”, and I doubt that its meaning guarantees
supervenience. But speaking loosely, I will say that both weak and strong
supervenience are analytic, in the derivative sense that ordinary English sen-
tences and inferences that exemplify them are analytic.
Are they analytic in this sense? Here is Blackburn’s comment:
One thing, then, that must be established in defending this part of the argument
is that if somebody claimed, say, that an action was absolutely identical in every
respect with another, except that it was much worse; or that a feature of character
like courage had changed in no way in its nature, relations, consequences, but yet
was of much less value than formerly; it would be a logical and not merely a moral
mistake that had been made. (1993: 116)
R. M. Hare:
we cannot say ‘P is exactly like Q in all respects save this one, that P is a good pic-
ture and Q not’. If we were to say this, we should invite the comment, ‘But how
can one be good and the other not, if they are exactly alike? There must be some
further difference between them to make one good and the other not.’ Unless we at
least admit the relevance of the question ‘What makes one good and the other not?’
we are bound to puzzle our hearers; they will think that something has gone wrong
with our use of the word ‘good’. Sometimes we cannot specify just what it is that
makes one good and the other not; but there always must be something. Suppose
that in the attempt to explain our meaning we said: ‘I didn’t say that there was any
other difference between them; there is just this one difference, that one is good and
the other not. Surely you would understand me if I said that one was signed and the
other not, but that there was otherwise no difference? So why shouldn’t I say that
one was good and the other not, but that there was otherwise no difference?’ The
answer to this protest is that the word ‘good’ is not like the word ‘signed’; there is a
difference in their logic. (1952: 81)
I will not try to defend these analyticities, or analyticity in general, in this
chapter. I will only report that I share Hare and Blackburn’s intuitions. If
an interlocutor repeatedly flouted supervenience, by calling some things

“good” and other things “not good” on the basis of no descriptive difference
whatever, then I would be left with nothing else to think than that her word
“good” meant something other than mine. It would be like talking with
someone who said both “x knows that p” and “p is false”.
The dialectical situation is this: I will shortly present Blackburn’s objec-
tion, namely that non-natural realists cannot explain these analyticities, and
then respond to his objection by offering an explanation. It is somewhat
ironic that, if I should succeed perfectly, my reward will be to get in trouble
with those who deny the analyticities in the first place, perhaps because they
reject the notion of analyticity altogether. But c’est la vie; one has to start
theorizing from some conception of the data, as best one can discern them.
The stage is set for Blackburn’s objection:
if A has some naturalistic properties, and is also good, but its goodness is a distinct
further fact not following from the naturalistic features, and if B has those features
as well, then it follows that B is also good. And this is a puzzle for the realist, because
there is no reason at all, on his theory, why this should follow. If the goodness is, as
it were, an ex gratia payment to A, one to which A is not as a matter of logic entitled
in virtue of being as it is in all naturalistic respects, then it should be consistent to
suppose that although goodness was given to A, it was not given to B, which merely
shares the naturalistic features that do not entail the goodness. … Supervenience
becomes, for the realist, an opaque, isolated logical fact for which no explanation
can be proffered. (1993: 118–19)
The objection is that non-natural realists have no good explanation of the
analyticity of supervenience, exemplified by entailments like the one from
(2) and (3) to (4).4 The view that normative predicates are primitive or
unanalysable seems to prevent us from pointing to any feature of their
4
Blackburn later (1985) restated the argument as follows: Suppose we judge a thing
to be good. We are then committed to there being some collection of descriptive prop-
erties and relations that underlie its goodness. Put all these together in a big, conjunc-
tive property F. Include in F also all its normatively relevant negative properties, that is,
include not being G, if being G would have destroyed its goodness. The big descriptive
property F, then, suffices for goodness, or so we think. If descriptions never entail evalu-
ations, there will be conceptually possible worlds in which the F’s are not good, as well
as worlds in which the F’s are good. But since every normatively relevant property and
relation, both positive and negative, is included in F, there will not be any conceptually
possible “mixed worlds”, in which some of the F’s are good and others not.
This strikes Blackburn as odd. If there are conceptually possible worlds in which the
F’s are good, and ones in which they are not good, then why are there no conceptually
possible worlds in which, say, half of them are good? Blackburn thinks it is implausible to
have such a “ban on mixed worlds” without giving any explanation for it.
The account I will offer meets this restated argument in exactly the same way as the
original version, so I will not discuss it any further in the text.
meaning that would produce this entailment. The only option seems to be
to take supervenience as a further primitive; to say that it is just a basic fact
about the meanings of normative terms that they behave this way. That is
what Allan Gibbard, for example, thinks the non-natural realist has to say:
A non-naturalistic “moral realist” can present certain features of ethical concepts
as brute truths: that, for example, whether an act is right or wrong depends on its
natural properties. … Such a theorist, though, offers no explanation at all of the
features of moral and other normative concepts. My aim in this book is to render
normative concepts unmysterious, to explain those features of ethical concepts that
such a non-naturalist can only treat as brute. (2003: 20)
My own theory explains much that non-naturalism takes as brute features of the
non-natural realm. If the good exercises its own sovereignty, why does goodness
depend on natural fact? That’s just the way the concept works, the non-naturalist
must be reduced to saying: it just does. (2003: 184)
I agree with Blackburn and Gibbard that this would be a weak position. It is
not an inviting view to maintain that there are just three things to say about,
say, the concept good, namely (a) it is primitive, (b) it is not descriptive,
and (c) it cannot apply to one but not the other of two descriptive twins. If
the concept is primitive, where does this restriction come from?
However, I believe (c) has an underlying explanation, to which I now turn.
10.2 COGNITIVE UNIVERSALISM
I begin with weak supervenience, the rule about evaluation of worldmates.

I will add an hypothesis in section 10.5 that will take us from the weak to
the strong principle.
The account begins by noticing that normative predicates can gram-
matically apply both to particulars (dated, non-repeatable things) and to
kinds (timeless things that can have instances). We can say that Florence
Nightingale was good or that altruism is good; we can say that the Iraq war
was wrong or that larceny is wrong. One might think this is merely a surface
phenomenon, because normative claims about kinds are covertly quantifi-
cational. For example, “Altruism is good” might be analysed as saying that
all, or typical, or that in general, instances of altruism are good. On such
analyses, notice, some quantifier binds a variable, the variable ranges over
particulars, and the predicate is applied to the variable. After the sentence
has been interpreted, then, the normative predicate is applied to particulars
also in general claims.
I will defend a different account of such claims, on which the normative
predicates are genuinely applied to kinds. For example, “Altruism is good”
does not say that all/typical/in general instances of altruism are good. It says
that altruism, a kind of motivation, is itself good.
Before I explain this proposal, I should mention that we are not com-
pelled to treat all such claims the same way. It could be that some are quan-
tificational and others kind-referring, or that some allow both readings. It
seems plausible that many non-normative uses of “good” or “bad” should be
analysed as quantificational generic claims or “generics”. For example, per-
haps “Apple pie is good” should be analysed: in general, if x is an instance of
apple pie, then x is good. Perhaps some apparently kind-referring normative
claims are also quantificational generics, or at least permit such readings. The
account I will propose is perfectly compatible with this possibility, albeit in
a roundabout way.5 But for ease of exposition I will present a “clean” view
on which the relevant claims are always kind-referring.
The proposal, then, is that normative predicates can apply both to par-
ticulars and to kinds, not just superficially but also after the sentence has
been interpreted. The next question is how to explain the logical relation-
ship between claims about kinds and claims about particulars. This issue
may take a moment to register, given how used we are to quantificational
analyses of general claims. Explaining logical relations between general and
particular claims is just what quantifiers do. For example, taking us from
“All F’s are G” together with “a is F” to “a is G” is just what “all” does. But
suppose “Murder is wrong” has the simple logical form wrong(murder).
Suppose further that “a is an instance of murder” and “a is wrong” have the
logical forms instance(a,murder) and wrong(a), respectively. In that case
we lose the logical relation between these claims; the three formulas in small
caps are logically independent of each other.
This is an appropriate result in some other cases. Claims of the form
“Kind K is F” do not always imply anything about whether individual K’s
are F. For example, the Monsanto Company has patented the genetically
modified corn Genuity VT Triple PRO. Which seems to make
(6) Monsanto owns Genuity VT Triple PRO.
true. But nothing follows about who owns the kind’s instances; (6) is com-
patible with Monsanto being sold out of Genuity …, all the physical corn
thus being the property of others.
5
Once I distinguish the pro tanto and all-things-considered readings (discussed
below), I am personally not able to get an exception-allowing, generic reading of “Lying
is wrong”, but suppose we want one. On my account it would be genx [lying(x)]
(wrongpar(x)), which, given the definition that will shortly be offered in the text, would
in turn be analyzed as genx [lying(x)] (∃K [token(x,K) & WRONGkin(K)]). “gen” is a
quantifier that means, roughly, “in general”.
But in the normative case, there are clearly logical relations between
claims about kinds and claims about particulars. It would be incoherent
to say that murder is wrong but that each murder has no tendency to be
wrong. If this is not explained by a silent quantifier in “Murder is wrong”,
then how is it explained?
A natural idea would be to reinstate the quantificational approach at the
level of lexical semantics, in the following way: to say that “wrong”, when
applied to a kind, does not express the concept wrong, but the concept
wrong*, which in turn has the definition being such that all one’s
instances are wrong. In other words, the sentence would not have a quan-
tifier in it, its logical form would just be wrong*(murder). But the defini-
tion of wrong* would in turn have a quantifier in it, which generates the
entailment from “Murder is wrong” and “a is a murder” to “a is wrong”.
Or perhaps one would prefer to define wrong* as being such that, in
general, one’s instances are wrong, in which case “Murder is wrong”
together with “a is a murder” would provide some kind of defeasible sup-
port for “a is wrong”.
However, for reasons I will give later, I believe we should instead explain
the logical relations from the opposite direction; we should take normative
predicates to express simple senses when they are applied to kinds, and com-
plex senses when they are applied to particulars.
Concerning the kind-applying senses, in other words, I suggest that
non-natural realists should carry on saying the things we have always said
about normative concepts (primitive, irreducible, simple, basic, fundamen-
tal). But for the particular-applying senses, we should not say those things.
These senses are, on the contrary, complex and definable. Not, though, in
terms of non-normative concepts, but in terms of the kind-applying ones.
For example, let goodpar and goodkin be the particular- and kind-applying
senses, respectively, of “good”. Then we can define the former in terms of
the latter, as follows:
CU: goodpar(x) ⟷def ∃K [token(x,K) & goodkin(K)]
A particular is goodpar just in case it is a token of a goodkin kind.
The variable “K” ranges over descriptive kinds. War, for example, is an
event kind; eating bananas is an act kind, being happy is a kind of mental
state. Importantly, there is no restriction on how general or specific the
kinds are. The range of K includes eating bananas while sitting on a train
passing by a lake, and even kinds that cannot be expressed in English at any
length.
However, we should not include so-called haecceitic kinds in the range
of K. A kind is haecceitic if it concerns a specified individual. For example,
buying Mary a bucket of roses and moving to Dallas are haecceitic kinds. The
motivation for this restriction is that normative concepts do not permit mere
haecceitic differences to make a normative difference. In other words, we
should understand “alike in every descriptive respect” in our supervenience
principles to mean “alike with respect to all qualitative descriptive proper-
ties and relations”. My impression is that that is what Hare, Blackburn, and
most participants in the subsequent debate have intended.6
Like properties, kinds can have instances/tokens. For example, the Thirty
Years War is a token of the kind war, and I am an instance of the kind Homo
sapiens. A kind and a particular stand in the tokening-relation just in case the
particular is an instance of the kind.
The right-to-left direction of CU predicts that, if goodkin applies to a
kind, then goodpar applies to every instance of the kind. That might seem
too strong. For example, we might be inclined to accept both
(7) Knowledge is good.
and
(8) Oedipus’s learning that Jocasta was his mother was not good.
But this is not a counterexample. The explanation is that “good” has two
senses along another dimension. It is widely agreed that normative predi-
cates can express both pro tanto and all-things-considered senses.7 And we
6
One exception is Matthew Kramer (2009: ch. 10). He points out that many reli-
gious believers think being pleasing to God is normatively relevant, but they can hardly be
accused of conceptual confusion. He also imagines a man who favours acts that benefit
France, but not because France has any interesting descriptive properties, or because he
is French or stands in any other interesting relation to France.
I do not find these examples convincing. The example from religion is misleading,
because religious believers think God’s opinion matters because he has certain qualitative
properties (power, wisdom) and stands in certain qualitative relations to them (has cre-
ated them, cares about them). It is not clear that anyone thinks God’s haecceitic identity
makes a difference, so that, for example, we could have another qualitatively identical
world, with a qualitatively identical creator, but the people over there have no reason to
obey their creator.
The thought-experiment with the Francophile is also weak. We need to imagine two
qualitatively identical countries, France and Schmance, and our man must stand in the
same qualitative relations to both. So he did not, for example, spend the summers of his
youth in one of them; nor does he have different feelings towards them. He knows all
this, but still, on the basis of no other difference whatsoever, he calls acts that benefit
France “good” and acts that benefit Schmance “not good”. Then I would simply repeat
Hare’s intuition from section 10.1: I would be left with nothing else to think than that
“something has gone wrong with his use of the word ‘good’ ”.
7
If an act is pro tanto wrong, then it has something wrong about it, even though it
may also have something right about it. If it has more wrong than right about it, then it
is all-things-considered wrong. Some writers distinguish pro tanto concepts from prima
facie concepts (the difference is not important for my purposes here); if both exist, then
we can apply CU to each of them.
must distinguish these before we apply CU. For example, in (7) we have
pro-tanto-goodkin, and in (8) we have all-things-considered-goodpar.
Under those readings, (7) and (8) are jointly compatible with CU.
This is not an ad hoc move: the distinction between pro tanto and
all-things-considered senses is independently motivated. And the predic-
tion that, if a normative concept applies to a kind, then the correspond-
ing particular-applying concept applies to every instance, is confirmed by
intuition. For example, if you think (8) is true, then you will not get (7) to
be true if you force the all-things-considered reading. And if you think
knowledge is pro tanto good, you will not get (8) to be true if you force the
pro tanto reading.
So far I have discussed “good”, but I think parallel analyses apply at least
for “bad”, “right”, “wrong”, “just”, “unjust”, “fair”, and “unfair”, or more
accurately for both the pro tanto and the all-things-considered senses of
these.8 CU, then, is a general recipe for defining particular-applying senses
of normative predicates in terms of their kind-applying senses.
CU has the same structure as R. M. Hare’s universal prescriptivism (1952),
and Alan Gibbard’s norm-expressivism (1990). These accounts also analyse
ascriptions of a normative predicate to a particular in terms of a general
commitment; in the case of Hare, a universal prescription, and in the case
of Gibbard, endorsement of a norm. In both cases, the general commitment
is existentially quantified.9 What the new account achieves is to make that
insight of Hare and Gibbard’s available in a truth-conditional framework.
In a nod to Hare, I will call it cognitive universalism.
The view now presented, I will proceed to give some reasons to accept
it. The first is that it explains the analyticity of weak supervenience, and
as we shall see in section 10.5, also that of strong supervenience, when we
combine it with a further hypothesis about normative predicates. But let
us start with the weak. The challenge is to explain things like the analytical
entailment from
(2) a is good.
(3) b is descriptively exactly like a.
to
(4) b is good.
8
I think analyses similar in spirit, but perhaps different in detail, apply for “rea-
son” and “ought”. And I suspect that aesthetic predicates like “beautiful” behave in this
way too.
9
For Gibbard, to say that a particular act is rational is to express acceptance of some
norm that permits it. For Hare, to say that a particular act is right is to command every-
one to act likewise in like circumstances (without specifying the act or the circumstances).
The explanation is this. From (2), by UC, it follows that there is some kind,
let us call it “L”, such that a is a token of L, and L is goodkin. From (3) and
our definitions of kind and of the tokening-relation, it follows that b is a
token of exactly the same kinds as a. So in particular, b is a token of L,
which, recall, is goodkin, and so by UC,
(4) b is good.
It should be straightforward to see how this generalizes to the other norma-
tive predicates, and also to the case where a is not good (in which case it
follows that b is not good either).
We have, in effect, replaced a situation where we had a primitive concept
good, and a brute conceptual necessity (supervenience), with a situation
where we have a primitive concept goodkin and a defined concept goodpar.
Weak supervenience just falls out of the definition of goodpar.
10.3 BEDFELLOWS
It might seem extravagant to posit a systematic duplication of predicate

senses, just in order to explain the analyticity of supervenience. But in this
section, I will present independent linguistic evidence for this duplication.
I believe we should posit it anyway; not just for normative predicates but
indeed across language.
The issue is tied to another feature of the account just given; its division
of labour between sentence semantics and lexical semantics. The work of
explaining truth-conditions and inferential relations between sentences is
not done in a single step, at the level of logical form; a significant part of the
job is relegated to lexical semantics. I will begin by saying something more
to motivate and explain this general approach.
It is nearly uncontroversial that kind-selecting predicates like “extinct”,
“widespread”, “invented by”, and “legalize” only take kinds as subjects.
Consider
(9) Skateboarding was invented by bored surfers.
(10) Cain invented fratricide.
The predicates are true of the act kinds in question, but are not true of
any of their instances. For example, “invented by Cain” does not apply
to any of the particular fratricides. Granted, we can come up with a more
complicated quantificational story that captures the truth-conditions
(roughly, Cain committed a fratricide at some time t, and there were no
fratricides before t). Similarly, maybe we can come up with quantificational

truth-conditions for
(11) Mehmed II legalized fratricide.
(What makes it difficult is that (11) could be true even if there never were
any fratricides, but we do not want it to be vacuously true in that case.) If
we multiply examples like these, a defender of the quantificational approach
would have to produce ever more complicated hypotheses about logical
form. That is not in itself a damning objection—in one sense, these claims
do have complicated meanings. But the problem is that the complexity
does not seem to lie with the sentences, but rather with the predicates. We
will capture what is going on better if we assign sentences like (11) a rather
simple logical form, and instead assign “legalize” a rich lexical semantics.
This point is reinforced by the fact that the inference from (10) and
(11) to
(12) Mehmed II legalized something Cain invented.
seems like a very simple inference. But on the quantificational approach
it would not be, because all these sentences would have hugely complex
logical forms.
The natural view to take of these examples is that, as far as sentence seman-
tics is concerned, NP’s occurring together with kind-selecting predicates
really do refer to kinds. For example, the logical forms of (11) and (12)
are just
(11*) legalized(Mehmed, fratricide)
(12*) ∃x [legalized(Mehmed, x) & invented(Cain, x)]
This view does not imply that (11) is true because Mehmed II performed
some mysterious act on an abstract object (the act kind fratricide). Clearly,
what ultimately makes (11) true is that he performed certain particular
acts; perhaps he signed a document or made an announcement in the town
square. Similarly, (9) is true because a gang of bored surfers performed some
particular, pioneering acts of skateboarding. That is how you legalize, or
invent, an act kind. But the point is that this information belongs in the
lexical semantics of the predicates, not in the logical form of the sentences.
In other words, understanding what these claims ultimately demand about
the occurrence of particular acts in the world is not a matter of under-
standing the logical form of the sentences, but of understanding what the
predicates mean.
The upshot is that we must distinguish between two levels of
truth-conditions. First we have the results of semantic analysis of sentences;
we can think of the results of this analysis as sentence truth-conditions or logi-

cal form. But it would be a mistake to think of these as giving us a picture
of what the world must be like in order for the sentence to be true. We
must check with lexical semantics first, to see whether any of the items that
occur in logical form should be further analysed, before we get to what we
can call worldly truth-conditions.10 So for example, because we understand
what “widespread” means, we understand that the worldly truth-condition
of “Bedbugs are widespread in Brooklyn” does not involve some abstract
object being spread out over Brooklyn, but rather that there be concrete
bedbugs at many locations across the borough.
Once we employ this division of labour between sentence and lexical
semantics, and make the corresponding distinction between logical form
and worldly truth-conditions, it becomes extremely plausible that ordinary
particular-applying predicates, like “have four legs” or “fifty feet tall” are
also capable of expressing kind-applying senses. Consider
(13) Panthera tigris has four legs.
(14) There was a dinosaur species that was fifty feet tall.
It is extremely plausible that “has four legs” in (13) is predicated of the kind
Panthera tigris, and that “fifty feet tall” in (14) is predicated of a variable that
the rest of the sentence relates to a dinosaur species.
But (13) does not say that tiger-kind has four legs in the same sense
as Bo the dog. Roughly, what (13) says about tiger-kind is that its geneti-
cally normal, unmolested tokens have four legs. Likewise, what (14) says about
the dinosaur species is that its healthy adult tokens were about fifty feet tall.
Someone who does not understand this does not understand the claims
being made. But since this is not encoded in logical form, the work must be
done by the predicates instead.
The view, then, is that the logical forms (ignoring time) are
(13*) have-four-legs*(Panthera tigris)
10
Related distinctions are drawn, in a similar context, by Koslicki (1999) and Leslie
(2008).
The division of labour I discuss here, between sentence and lexical semantics, should
not be confused with the division of labour between semantics (as a whole) and meta-
physics. How to draw that line is another huge subject. But I take it that there are clear
cases on either side of the line. For example, understanding what kind-selecting predi-
cates demand about particular tokens is clearly on the semantic side. Someone who does
not understand that “The Dodo is extinct” is true iff there once were Dodos, but none
now, simply does not understand what “extinct” means. On the other hand, reduction
of truths about chairs to truths about atoms arranged chairwise is clearly on the meta-
physical side. One cannot get this reduction just on the basis of understanding concepts.
(14*) ∃x [dinosaur species(x) & fifty-feet-tall*(x)]

where have-four-legs* is the concept being such that one’s genetically
normal, unmolested tokens have four legs, and fifty-feet-tall* is
the concept being such that one’s healthy adult tokens are about
fifty feet tall.
The view is not that all these predicate senses are lexicalized. Plausibly,
the usual case is that the particular-applying senses are lexicalized, while
suitable kind-applying senses are constructed ad hoc, in order to get rea-
sonable interpretations. For example, if I say “The English Setter weighs
between 40 and 50 pounds”, you construct a suitable kind-applying sense
of “weigh between 40 and 50 pounds” on the spot, in order to get a sensible
interpretation.
This account predicts that, if we apply a predicate to a particular-referring
and a kind-referring NP at the same time, the result should be at least
mildly zeugmatic. Zeugmas are not ungrammatical, but feel a bit like puns,
in that a word is used to mean two different things at the same time. Since
the particular-applying and kind-applying senses are so closely related, the
effect is not likely to be as strong as in “He took his hat and his leave”, but
it should be perceptible. As indeed
(15) Bo and Panthera tigris both have four legs.
(16) Equus zebra and uncle Bob live in Africa.
testify.
I have given some reasons to think it is a common phenomenon that
predicates express different senses when they are applied to particulars and
to kinds, respectively. And I have explained the general framework I think
this kind of view should be located in. Turning now to normative language
in particular, I will present some evidence that normative predicates can be
applied to kind-referring NP’s (I assume it is clear that they can be applied
to particular-referring NP’s, as in “Florence Nightingale was good”).
Since “widespread” is a kind-selecting predicate, “murder” refers to a
kind in
(17) Murder is widespread.
But then it seems that “wrong” is applied to a kind in
(18) Murder is widespread and wrong.
(19) Murder is widespread even though it is wrong.
Likewise, we can apply “good” to a variable that the rest of the sentence
relates to a kind:
(20) Some kinds of charity are good.

There is independent evidence, then, that normative predicates can be
applied both to particulars and to kinds, not just superficially but in logi-
cal form. The main reason to think that they express different senses in the
two uses has already been given: otherwise it would be hard to explain the
logical relations between the two sorts of claim. In addition, and finally,
I suggest there is a feeling of zeugma when we apply a normative predicate
to kind-referring and particular-referring NP’s at the same time:
(21) The war in Iraq and sexual harassment are wrong.
(22) Philanthropy and Warren Buffett are virtuous.
10.4 WHY THE KIND-APPLYING SENSES ARE BASIC
There remains, though, a deep contrast between normative language, on

the one hand, and at least most descriptive language, on the other. Even if
kind-reference is fairly common at the level of sentence semantics or logical
form, in the descriptive case it will at least usually be “analysed away” in
lexical semantics, so that, when we get to worldly truth-conditions, we are
left with only particulars and their properties and relations.
For example, in the case of kind-selecting predicates like “invent” and
“widespread”, even though these are applied to kinds in logical form, this
kind-reference is analysed away in lexical semantics. Just by knowing what
“widespread” means, we know that what it is for a kind to be widespread
is for it to have instances in many, scattered locations. So the worldly
truth-conditions of “widespread”-claims are about particulars and their
properties and relations.
There may also be areas of descriptive language that do not work like
this; areas where kind-reference and predication of properties to kinds per-
sist all the way to worldly truth-conditions. I am neutral on this issue. But
I will argue that normative language works in this second way. In fact, nor-
mative worldly truth-conditions are always about kinds, and never about
particulars.
This is because the kind-applying senses of normative predicates are
basic. So a normative claim about a kind, like “Murder is wrong”, will
not be reduced in terms of particulars at any level of semantic analysis. Its
logical form is wrongkin(murder) and its worldly truth-condition is also
wrongkin(murder).
On the other hand, applications of normative concepts to particulars
are analysed away at the level of lexical semantics. However, they are given
mixed worldly truth-conditions, in part descriptive, in part normative. For

example, “Bob is good” has the logical form goodpar(Bob). Applying the
definition of goodpar, we get the worldly truth-condition
∃K (token(Bob,K) & goodkin(K)).
Bob is a token of some descriptive kind or other, such that that kind is
goodkin. So the descriptive part of the truth-condition is still about the par-
ticular, Bob. But the normative part, saying that the relevant kind is goodkin,
ascribes a normative property to a kind, not to a particular.
Cognitive universalism, then, predicts that normative judgments about
kinds are basic, in the following sense: the worldly truth-conditions of nor-
mative claims are either purely about kinds, or else mixed, but with the
normative part being about kinds. Normative language bottoms out in truth-
conditions about kinds. I have already given the main reason to believe this;
it explains the analyticity of (weak) supervenience. But we can also give
an independent argument for this aspect of the view. It is plausible that
normative claims about kinds are in this way basic, because they are episte-
mologically basic in a parallel way. Moral epistemology also bottoms out in
judgments about kinds, as I will now explain.
Let us start with a simple observation. In descriptive enquiry, we typically
go from judgments about particulars to judgments about kinds. So, for
example, we might do ornithology in roughly the following fashion:
(A) This bird sings in the morning and that bird sings in the morning
and yonder bird …—and come to think of it, they are all robins! So it
seems robins sing in the morning.
But notice how backwards it would be to try to do ethics in a similar way:
(B) This act is wrong and that act is wrong and yonder act …—and by
golly, they are all sexual harassments! So it seems sexual harassment
is wrong.
The direction of epistemic justification is the reverse in the normative case:
we go from general normative judgments, and empirical judgments about
particulars, to normative judgments about those particulars. For example,
we go from the general judgment that using a position of power to pressure
someone into bed is wrong, and the empirical judgment that that is what x
did to y, to the judgment that what x did to y was wrong.
Granted, there are complications. We do sometimes arrive at normative
judgments about kinds through investigation of their instances. For exam-
ple, we probably do not have a direct intuition that appointing relatives to
government jobs is wrong. We arrive at this judgment by looking at coun-
tries where that practice is common, and observing its typical effects. But
this is not a counterexample to the point I am making. For what we then

investigate about these instances is not their normative, but their descriptive
properties. For example, we see that they lead to inequality of opportunity
and loss of general utility. But our judgment that general utility and equal
opportunity are good are in turn judgments about kinds, and not based on
normative judgments about particulars.
The extreme case of this epistemic structure is pure hedonic utilitarianism.
For the committed utilitarian, normative enquiry will be a lot like empirical
enquiry; particular cases will be decided by investigating their effects, and
judgments about kinds will be a lot like inductive generalizations. But at the
bottom of all this epistemic activity is the single ur-judgment that an act is
right if and only if it maximizes the balance of pleasure over pain. This is a
judgment about kinds, and is not in turn arrived at by inference from par-
ticular cases. Take it away and the whole superstructure evaporates.
The claim, then, is that all normative justification bottoms out in judg-
ments about kinds. That may seem to contradict the popular view that
moral epistemology proceeds by the method of reflective equilibrium; going
back and forth between principles and cases, trying to find intuitively attrac-
tive principles that yield intuitively attractive verdicts about cases, revising
both kinds of intuitions as we go. But not really. For notice that “cases” here
are not really particulars; they are instead narrowly circumscribed kinds. For
example, we might well have as a fixed point in our normative reasoning that
the Srebrenica Massacre was wrong. But this fixed point is not really that
a particular act de re, the Srebrenica Massacre, was wrong. For imagine we
became convinced by some internet conspiracy theory that the “massacre”
is in fact a fabrication of Western propaganda. That would make us revise
our normative beliefs about the event, and perhaps about particular politi-
cal leaders, etc. But these “local” changes would not ramify into our wider
normative belief system. We would still believe that killing eight thousand
people on account of their ethnicity is wrong, and this belief would continue
to play the role in our reflective equilibrium that we had hitherto (mislead-
ingly) ascribed to the belief that the Srebrenica Massacre was wrong.
The same point applies to everyday uses of our moral sensibility. Suppose
for example that you witness a man subjecting another to some kind of
humiliating treatment for no good reason. You have an immediate gut reac-
tion telling you that what the first man is doing is wrong, and this in turn
leads you infer that it is wrong to humiliate others needlessly. One might
think that here, surely, is a case where a normative verdict about a particular
case supports a general normative conclusion. But that would again be mis-
leading. For suppose you learn that the two men were in fact actors rehears-
ing a play. That would lead you to revise your normative verdict about that
particular act (de re). But it would not lead you to conclude that it is OK to
humiliate others needlessly after all; the lesson you learned about that still
holds good.11
In general, what matters to your normative thinking is not really your
reaction to particular cases de re, but your reaction to the descriptive proper-
ties you think these cases have, that is, to the kind you think they instanti-
ate. What particular cases can do is to make this or that kind salient to us, by
making a token salient, but it is our verdict about the kind that plays a role
in our reflective equilibrium, not our verdict about the token. In this way,
beliefs about particulars are epiphenomenal in our normative belief system;
they are supported by but do not support beliefs about kinds. Reflective
equilibrium is reached when our beliefs about more general kinds fit with
our beliefs about more specific kinds.
Normative enquiry, then, has exactly the structure we should expect if
cognitive universalism is true. Since normative claims about kinds are not
generalizations over their instances, they are not justified in the way gen-
eralizations are justified, from premises about instances. Instead, they are
justified either by some kind of direct conviction about the kind (say, that it
is wrong to kill people on account of their ethnicity), or else by their coher-
ence in a network of judgments about kinds. And just as normative claims
about particulars, in worldly truth-conditions, factor into a descriptive com-
ponent about the particular, and a general normative component, so they
are justified by evidence about the descriptive properties of the particular,
and a general normative judgment or principle.
10.5 WHY STRONG SUPERVENIENCE IS ANALYTIC
Earlier, I postponed discussion of strong supervenience and its analyticity.

My formulation of the principle was:
Strong normative-descriptive supervenience: for all possible x, y: if x and
y are alike in every descriptive respect, then x and y are alike in every
normative respect.
What I take to be analytic, again, are not such principles themselves, but
ordinary claims and inferences that exemplify them. Since strong superveni-
ence coordinates the normative properties of individuals in different worlds,
we have to look to modal or counterfactual claims in order to find sen-
tences that exemplify the strong but not the weak principle. By their nature,
then, the relevant claims will be rather complicated in their non-normative
This point is from Fine (2005).

11
elements, which makes it more difficult to say whether they are analytic.
Indeed, there may be no sharp line between what is analytic and what is just
very unarguably true. That being said, I would agree that
(5) If Ted is good, then it is impossible to be just like Ted in every descrip-
tive respect and not be good.
is analytic.12 But cognitive universalism does not in itself predict this. The
antecedent entails that there is some kind T, such that Ted is a token of T,
and T is goodkin. The consequent, in effect, says that no one, in any possible
world, is T but not goodpar. But from T’s being goodkin in the actual world,
it does not obviously follow that T is goodkin in all worlds. And if there is a
world where T is not goodkin, then CU allows people there to be T but not
goodpar. Before I explain why this does not happen, I will set the stage by
discussing a different example, from mathematics.
Syntactically, “47 is a prime number” is in the present tense. So one might
think its truth-condition is that 47 is prime at the time of utterance. After all,
that is the semantical contribution the present tense usually makes. But it
seems to me that someone who says
(23) 47 is prime today, but on Thursday it will not be.
is showing clear signs of conceptual confusion. Assuming they understand
the rest of the sentence, they apparently do not understand “prime” ade-
quately. In other words, it is plausible that
(24) If a number is prime, then it is always prime.
is analytic. The explanation is that “prime” is a tenseless predicate; it may
accept tense syntactically speaking (“was prime”, “will be prime”), but it
does not really have tense semantically speaking. Exactly how to cash this
out depends on your general semantical framework; perhaps the proposi-
tion expressed by “47 is prime” does not have a time parameter, or time
argument. In any case, the result is this: if you fully understand “47 is
prime”, then you know that you are not supposed to evaluate it with respect
to any particular time. The result is that (23) is analytically false and (24)
analytically true; but unobviously so, due to the subtle explanation of their
analytic truth/falsehood. That seems about right.
Suppose, next, that someone says
(25) 47 is prime, but if Gore had won the Florida recount, 47 would not
have been prime.
12
Like before, read “descriptively alike” to include both properties and relations.
That assertion would also seem to signal conceptual confusion. Given a yes/
no choice, again, I would say that
(26) If a number is prime, then it is necessarily prime.
is analytic. The explanation is that “prime” is what Fine (2005) calls an
“unworldly” predicate. This is the modal analogue of tenselessness. Again,
exactly how to cash it out depends on one’s general semantical framework;
perhaps the proposition expressed by “47 is prime” does not have a world
parameter or world index. In any case, the result is this: if you fully under-
stand “47 is prime”, then you know you are not supposed to evaluate it
with respect to any particular world. This makes (25) analytically false and
(26) analytically true, but in an even more subtle and unobvious way than
before; which again seems like a welcome prediction.
The explanation, I propose, of the analyticity of strong supervenience is
cognitive universalism plus the fact that the basic, kind-applying senses of
normative predicates are tenseless and unworldly. The analyticity of (5) is
then explained as follows. From the antecedent, it follows by UC that Ted is
a token of some kind, let us call it “T”, such that T is goodkin. Since “good-
kin” is tenseless and unworldly, T is goodkin in every world. But then, by UC,
any possible thing that is T is also goodpar, so the consequent is established.
The tenselessness and unworldliness of the kind-applying senses is
easy to miss, because the words we use to express them also express
particular-applying senses, which are both tensed and worldly. Or more
accurately, the particular-applying senses are “bastards”, having a worldly
part (the descriptive part) and an unworldly part (the normative) in their
definition.
Another reason one might think normative claims about kinds are sensi-
tive to time or world is that we often make such claims in a “parochial” way,
leaving some background conditions implicit. For example, we think
(27) Slavery is wrong.
is true and obviously so. But it seems possible for slavery to be right; suppose
we come across an alien, intelligent species for which freedom is torture and
servitude bliss, and which does not even have the potential to be happy in
any other way. It seems it would be OK, perhaps a duty, to enslave them.
It is important to notice, however, how we react to this kind of example.
My reaction is to clarify what I mean by (27).13 When I now assert (27),
I do not intend “slavery” to cover this kind of case; what I intend to say is
13
As mentioned in section 10.2 and explained in n. 5, my account can accommodate
exception-allowing, generic readings for claims like (27). I cannot feel any such reading
of (27), but if you can, restrict what I say in this section to the kind-referring reading.
that enslavement of humans and other creatures that are capable of autonomy
without agony, is wrong.
There is no point in lengthy clarifications of precisely which act kind
we are talking about when our interlocutor understands it perfectly well
anyway. There may be bizarre but metaphysically possible scenarios where
slavery, rape, genocide, etc., would be OK; but it would only waste time
and strain our interlocutor’s patience to enumerate outlandish exceptions
that are irrelevant to the moral problems we are dealing with.
The crucial point, again, is this: if such a faraway scenario is brought to
attention and made relevant, we treat it as calling for clarification of the
claim we made, for example that slavery is wrong. We do not stick to the
original claim but reserve it for the actual world; we do not conclude that
slavery can be right but is actually wrong.
A careful look at how we react to such examples, then, indicates that the
kind-applying normative concepts are indeed unworldly. Which, together
with cognitive universalism, explains the analyticity of strong superveni-
ence. I suggest in conclusion that these explanatory advantages, and the
structure of moral epistemology, make cognitive universalism an attractive
account of normative language.
10.6 THE METAPHYSICAL CHALLENGE
There is another supervenience challenge facing non-natural realism. We

now have an explanation of why it is analytic that normative properties
supervene on descriptive properties. But, ironically, non-natural realists
also have to explain why it is true that normative properties so supervene.
That might seem confused: analyticity entails truth, so why is there still a
problem?
Because what is analytic is that the properties picked out by normative
predicates supervene on descriptive properties. But it does not follow that
non-natural normative properties so supervene. That only follows if we
assume that normative predicates pick out non-natural normative proper-
ties. But if we can give no independent assurance that non-natural norma-
tive properties are well-behaved in respect of supervenience, and in fact have
some reason to doubt it, then we have no right to this assumption.
Cognitive universalism, in other words, is a theory about English;
it explains why the normative part of our language behaves a certain
way. Now we need to explain why the stuff in our theory behaves in the
corresponding way.
Why doubt it? The concern is that non-natural realists claim that norma-
tive properties are fundamental and non-descriptive. They cannot be reduced
in terms of our desires, beliefs, attitudes, or anything else; in fact they are
properties of a different kind from all others. But on the other hand, given
supervenience, they are each going to have a cointensional (necessarily coex-
tensional) descriptive property. Take for example goodness: just look at all
the good things in all the possible worlds, and put each of their profiles of
descriptive properties and relations together in a long, disjunctive property
D. Given supervenience, goodness and D are cointensional: every possible
good thing is also D, and every possible D thing is also good.14
This result collides with Hume’s Dictum: the principle that there be
no “necessary connections between distinct existences”. It is not obvious
exactly what that means (cf. Wilson 2010), but in the present context it
boils down to a ban on distinct, cointensional properties. In other words: if,
in every possible world, all the Fs are Gs and vice versa, then F and G are the
same property. For if F and G are really different properties, then surely it is
possible for something to be F without being G or vice versa?
Hume’s Dictum is especially plausible for fundamental properties. If F is
a fundamental property, then a thing x’s being F does not consist in some-
thing else being the case with x; x’s being F is just a basic fact. And likewise
for x’s being G, if G is also fundamental. But if x’s being F is just a basic
fact, then there should be another world that is otherwise like the given one,
except that x is not F.15 And if x’s being G is another basic fact, then it would
be strange if it were impossible to remove x’s F-ness without removing its
G-ness. Thus, even looking just at a single thing x, it seems like it should be
possible for fundamental properties to come apart.
The worry, then, is that we have no good explanation of why non-natural
values or reasons would supervene on descriptive properties, and that, sup-
posing they do supervene, they will violate (a version of ) Hume’s Dictum, by
having cointensional descriptive properties (from which they are neverthe-
less supposed to be distinct). So if non-natural realism is to be plausible, it
needs to explain why non-natural reasons or values supervene on descriptive
properties, and explain it in a way that either gets around Hume’s Dictum,
or else makes it plausible that (the relevant version of ) the Dictum is false.
14
See Jackson (1998: ch. 5) for a rigorous statement of this point.
15
It may be that x would need some other property from F’s contrast class. For exam-
ple, if F is a shape, then x would take some other shape H. But if G is some other fun-
damental property (by hypothesis not a shape), then it would be strange if this change
would have to result in the removal of x’s G-ness.
10.7 MORAL PLATONISM
Many seem to think the non-natural, irreducibly normative properties that

Moore, Parfit, Nagel, etc., posit take particular things as their bearers. For
example, that Moore posited a non-natural property (“goodness”) that is
instantiated by things like Bill’s pleasure at time t from looking at picture x,
or Bob’s knowing at t that p. Or that the irreducibly normative reasons Parfit
posits are relations between particular facts and responses available to par-
ticular agents. For example, that the fact that Bill is in pain is a reason for
Bob to give him an aspirin.
If that is the view, then it will indeed be mysterious why these properties
should supervene on descriptive properties. Suppose we have, as a basic,
irreducible fact, that the fact that Bill is in pain (and Bob is around with a
spare aspirin), is a reason for Bob to volunteer the aspirin. Then why would
it be impossible for Susan and Tracy, say, to find themselves in a similar
predicament, but without Tracy having a similar reason? (Supposing, again,
that the reason-givingness of Bill’s pain is a basic fact, not deriving from
some underlying, general normative fact.)
But I see no reason why the view should take that form. The main intel-
lectual motivation behind non-natural realism has always been a strong
commitment to certain pre-theoretical, first-order normative beliefs. And
now that cognitive universalism has shown that these beliefs bottom out in
truth-conditions about what kinds of things have value or provide reasons,
it is only natural to direct one’s metaphysical commitment accordingly: to
facts about what kinds of things have value or provide reasons.
And if we take up that metaphysical commitment, then there is no
need to posit, in addition, irreducible normative properties that take par-
ticulars as their bearers. Instead, we can give a reductive account of the
particular-applying normative properties. For example, we can say that the
property goodness (the one that takes particulars as bearers) is just the prop-
erty being a token of a good kind.
On this reductive view, we can explain why the values of particular things
supervene on their descriptive properties. Whether or not a particular thing
is good will depend on two things: (i) which (descriptive) kinds it tokens,
and (ii) which (descriptive) kinds are good. The first part supervenes on
the thing’s descriptive properties in a real and obvious way: it is a token
of a given kind just in case it instantiates the corresponding property. The
second part, concerning which kinds of things are good, also supervenes
on descriptive properties, but in a trivial, uninteresting sense. Facts about
what kinds of things are good, non-natural realists say, are necessary, and
necessary facts trivially supervene on everything (there can be no difference
in the necessary facts without a difference in the Y-facts, for any Y you like,
because there can be no difference in the necessary facts, period).
The view requires metaphysical commitment not only to non-natural
normative properties, but also to kinds to serve as their bearers. If we have
kinds in our metaphysics anyway, that is of course no problem, but not
everybody does. However, suppose we have properties in our metaphysics,
for independent reasons. Then we can suit the view to our liking, as fol-
lows: instead of saying the metaphysically basic fact that makes “causing
needless pain is wrong” true is that the kind causing needless pain has the
property wrong, we can say it is that the property causes needless pain has
the property wrongmaking. Rather than posit the kind pleasure to instanti-
ate the property good, we can posit the property pleasant to instantiate the
property goodmaking. And so on. For each basic, normative concept F that
applies to kinds, we posit a corresponding F-making property that applies
to properties.
On this second view, the particular-applying property goodness will just
be the property having a goodmaking property. Particular-applying wrongness
will just be the property having a wrongmaking property. And so on. The
explanation of supervenience proceeds as before.
I do not mean to suggest that there is a deep metaphysical difference
between properties and kinds. There seems to be a shallow difference,
marked by the linguistic phenomena above; the kind murder is wrong, but
the property murderhood is not wrong (but wrongmaking). I am not com-
pletely sure what to think about this, but it seems to me that the two ver-
sions are different ways of spelling out the details of a single underlying
view. The underlying view is that the basic normative properties are second-
order; they take universals, rather than particulars, as their bearers. Hence
the name “moral platonism”. It is this broad view I want to defend.16 For
ease of exposition, I will revert to the kind-based view, but my discussion
applies equally to the property-based view.
Moral platonism will, to be sure, leave something unexplained, and it will
contain some necessary connections between distinct properties. But it is
crucial to see exactly what is left unexplained, and exactly what these neces-
sary connections are like. What is left unexplained is not supervenience,
but rather (some of the) facts about which kinds of things are good, bad,
wrong, etc. For example, that causing needless pain is wrong; that happi-
ness is good; that suffering is bad. Some of these facts may be explainable in
16
The view is suggested by Mackie (1977: 23, 41) and superbly defended by Forrest
(1986).
terms of the others, but some of them are going to be basic, and admit of
no further explanation.
Everyone agrees it is a desideratum on metaethical theories that they
should explain why the values of particular things supervene on their
descriptive properties. Non-natural realism can do that, using the reductive
account of particular-applying normative properties, and appealing to facts
about the values of kinds. So the question is whether it is OK to leave these
latter facts unexplained. In other words, is it also a desideratum on meta-
ethical theories that they should explain why suffering is bad, or why happi-
ness is good, and so on? To me these seem like good places for explanations
to end. But it is hard to argue about where explanations should end; so let
us just record that the view developed here will have such commitments.
But what about the necessity of these facts? Will they not give us “neces-
sary connections between distinct existences”? Yes they will, but it is impor-
tant to see exactly what these connections are like. Actually they come in
two forms. First, particular-applying normative properties will have coin-
tensional descriptive properties, like goodness and D, as discussed above.
But particular-applying goodness is not a fundamental property; it is just
the property being a token of a good kind. That it necessarily co-occurs with
D is just a trivial consequence of its definition, given that facts about the
values of kinds are necessary.
So the interesting necessary connections are these latter facts them-
selves. Take agony and badness, for example. The necessary connection
between them is not co-occurrence, but instantiation. It is not that agony and
(kind-applying) badness are instantiated by the same things; it is that agony
itself instantiates badness, that agony is bad. In other words, the necessary
connection is that a first-order universal has a second-order universal.
There are other examples of necessary connections between first- and
second-order universals. Crimson, for example, has the property of being
a shade of red. And it has that property in every world, or in every world
in which it exists, on an Aristotelian view of universals. That is not strange
at all; of course we need not check, with any given world, to see whether
crimson is a shade of red there.
Necessary relations between first- and second-order universals, then, can-
not be ruled out as a matter of principle. We have to look at each case and
see if a given first- and second-order universal are contingently or necessar-
ily related. In the case of determinables and determinates, it is plausible that
the connections are necessary, because they hold in virtue of the nature of
the universals themselves. Being a shade of red is part of what crimson is.
The normative case is not like that. It is, for example, not part of what
agony is that it is bad. But it is still intuitively plausible that we do not have
to check with a given world to see whether agony is bad there. Or whether,
say, treating another as a mere means is wrong there. If we keep in mind

that we are considering the basic normative facts, they are intuitively not
contingent.
But why? I suggest the non-natural realist should again look to the
semantics for guidance. The kind-applying normative concepts, I argued,
are unworldly; the propositions they figure in are not to be evaluated with
respect to this or that world. The metaphysical analogue of this is what Fine
calls “transcendental” facts; facts that hold, not in every world, but indepen-
dently of all the worlds (2005: ch. 10).
Fine makes a distinction, in other words, within the class of what is usu-
ally thought of as the necessary truths. The necessary truths proper are those
that engage with each world, but in such a way as to come out true every
time. “P or not-P” is one example; in some worlds it holds because “P” is
true, in other worlds because “not-P” is. Another example is “Nothing is
both red and green all over”. Whichever world it is evaluated at, it looks at
every object in that world, but never finds a counter-instance.
The transcendental truths, on the other hand, do not even engage the
different worlds; there is nothing in the worlds that they answer to for their
truth. Mathematical claims are good examples: “Seven is a prime number”
can only in a degenerate sense be said to be true “in” a given world, for there
is nothing about any of the worlds that makes it true.
The distinction between necessary and transcendent truths depends on a
distinction between two ways of thinking about possible worlds. On what
we can call the tractarian conception, a world is a totality of facts. The actual
world is everything that is the case; a merely possible world is a way eve-
rything could have been. On this conception, there will be no distinction
between necessary and transcendental truths; the fact that seven is prime,
for example, is a part of every world.
On the second conception, a world is a totality of concrete substances
and their attributes. If you specify all the concrete substances there ever are,
and all the properties and relations they have throughout their careers, then
you have specified a world. Let us call this the substantive conception of a
world. The actual world, in this sense, is just the physical universe, but sub-
stantive worlds need not be physical; some have angels and spooks in them.
The tractarian notion of possible worlds and the corresponding notion
of necessity are useful for many purposes, but I suggest that they are not
helpful in understanding the content of certain philosophical views. When
the mathematical platonist says that mathematical facts are necessary, she is
not well understood as saying that they hold, again and again, in each pos-
sible world. The mathematical facts hold once and for all, as it were; they
belong to an “invariable framework” within which possibilities play out
(Fine 2005: 326). To make the point vivid, we can imagine God pondering
which world to create. His alternatives are the substantive possible worlds,
not the tractarian ones. Seven being a prime number is not somehow a fea-
ture of every alternative, it is not a feature of any of them. The mathematical
facts are already there, before he creates anything.
Likewise, when non-natural realists say the basic normative facts are nec-
essary (Parfit 2011: II, 489; Enoch 2011: 146; Scanlon 2014: 39–41), they
are not well understood as saying that these facts hold, again and again, in
every world. They are part of the invariable framework. God may decide
which world to create, but he does not get to say how good it will be if cre-
ated. The basic normative facts are already there.
The strong supervenience of particular-applying normative properties on
descriptive properties is a natural consequence of this view. Whether or
not the view violates Hume’s Dictum depends on how we understand the
Dictum, once the worldly/unworldly and necessary/transcendental distinc-
tions have been made. One could combine moral platonism with a version
of the Dictum, restricted to necessary relations between worldly properties
(properties that figure in worldly facts), or between fundamental worldly
properties.
But a proponent of Hume’s Dictum could reasonably say that the spirit
of the Dictum requires a ban on both necessary and transcendental connec-
tions between distinct properties. In that case the present view contradicts
her principle. But someone who is otherwise inclined to accept transcenden-
tal facts is not likely to worry that they connect “distinct existences” (what
else would they connect?). So Hume’s Dictum turns out to be a side issue;
the big question is whether to accept transcendental facts in the first place.
Not surprisingly, then, the case for moral platonism opens up into the
case for platonist metaphysics in general. If mathematical, logical, and/or
modal facts are best understood as transcendental, then moral platonism
has impressive allies. I leave it to others to defend the other platonisms;
here I will only point out that one can also argue in the other direction.
If non-natural realism is the best account of value, and if, as I have argued
here, thinking about supervenience takes us from non-natural realism to
platonism, then one might simply conclude that there are transcendental
facts because there are values.
References
Blackburn, S. 1971. ‘Moral Realism’, in J. Casey (ed.), Morality and Moral
Reasoning: Five Essays in Ethics. London: Methuen, 101–24.
Blackburn, S. 1985. ‘Supervenience Revisited’, in I. Hacking (ed.), Exercises in

Press, 47–68.
Enoch, D. 2011. Taking Morality Seriously: A Defense of Robust Realism.
Fine, K. 2005. Modality and Tense. Oxford: Oxford University Press.
Forrest, P. 1986. ‘Supervenience: The Grand-Property Hypothesis’, Australasian
Gibbard, A. 1990. Wise Choices, Apt Feelings. Cambridge, MA: Harvard
University Press.
Hare, R. M. 1952. The Language of Morals. Oxford: Oxford University Press.
Jackson, F. 1998. From Metaphysics to Ethics. Oxford: Oxford University Press.
Koslicki, K. 1999. ‘Genericity and Logical Form’, Mind and Language 14: 441–67.
Kramer, M. 2009. Moral Realism as a Moral Doctrine. Malden, MA: Wiley-Blackwell.
Leslie, S.-J. 2008. ‘Generics: Cognition and Acquisition’, Philosophical Review
117: 1–47.
Mackie, J. L. 1977. Ethics: Inventing Right and Wrong. Harmondsworth: Penguin.
Moore, G. E. 1903. Principia Ethica. Cambridge: Cambridge University Press.
Scanlon, T. M. 2014. Being Realistic About Reasons. Oxford: Oxford University Press.
Sturgeon, N. 2008. ‘Doubts about the Supervenience of the Evaluative’, in
R. Shafer-Landau (ed.), Oxford Studies in Metaethics, Volume 4. Oxford: Oxford
Wilson, J. 2010. ‘What is Hume’s Dictum, and Why Believe It?’ Philosophy and
11
Explaining the Quasi-Real
Jamie Dreier
This chapter is about whether Quasi-Realism gains any advantage over

Robust Realism with respect to a certain well-known problem: the problem
of explaining supervenience. The simplest way it might have an advantage is
by having an explanation for the phenomenon that Robust Realism cannot
explain. That, I believe, is the traditional view of the dialectic situation. But,
there are other ways. A Quasi-Realist might deny the claims of superveni-
ence. Or, Quasi-Realism might have no explanation for a claim it endorses,
but have a more satisfactory story about why there is no explanation.
I will start with a summary of what the supervenience problem is sup-
posed to be (sections 11.1 and 11.2). I will only be able to lay out the varie-
ties of supervenience and give some reason for thinking that a particular one,
a kind of Strong Supervenience, holds between the moral and the natural.
In section 11.3 I will mention a few recent worries about supervenience,
but I will not try to address the worries in great detail. The point of this
chapter is not to work out exactly what to think about the seriousness of
the supervenience problem, but only to see how the prospects for Quasi-
Realists differ from the prospects for Robust Realists.
Section 11.4 recounts the history of expressivist thinking about super-
venience. The gist is that the supervenience problem was a challenge raised
by expressivists (or their intellectual ancestors) against Robust Realists, with
the idea that expressivism had an excellent explanation of the phenome-
non and realism had none. This explanation involves what I will call the
“expressivist sidestep,” a favorite maneuver of expressivists by which they
avoid answering the traditional questions of metaethics and give satisfying
answers to different questions—questions that arise when we look at the
same phenomena “side on”—instead.
Section 11.5 explains what Quasi-Realism is supposed to be, as I see it,
and how (again, according to me) it is distinguished from Robust Realism;
clear thinking about these questions brings the big problem out in the open.
The big problem is that Quasi-Realists have not provided any explanation at
274 Jamie Dreier
all for the phenomenon of supervenience. They are, it appears, in the same
boat as Robust Realists, with respect to this problem (and, as I’ll explain,
some other similar ones).
Section 11.6 maps out the possible paths for Quasi-Realists to travel in
pursuit of Quasi-Explananda, and gives what I think are some reasons to be
optimistic, but does not run the quarry to ground. I end with a suggestive
analogy and a synopsis of the state of the dialectic.
11.1 THE SUPERVENIENCE PROBLEM
When something has a moral property, it also has a natural property whose
possession necessitates the moral one. This is a kind of supervenience of
the moral on the natural. The supervenience relation comes in a number of
strengths and versions. I will lay out a crash course in the varieties, and settle
on one for our purposes, and along the way settle a few other details. Then
I will say why the supervenience of the moral on the natural is supposed to
be a problem for a certain kind of moral realism. To end the section, I will
mention some doubts and worries about the supervenience of the moral on
the natural and try to say why they are not of great concern here, but I will
not address them seriously.
Supervenience is a relation between families of properties. Properties of
one kind supervene on properties of another kind when there cannot be
a difference of the first kind without there being a difference of the sec-
ond kind. Cannot, because supervenience is a relation of necessitation. In
1952, R. M. Hare (1991) introduced metaethics to the supervenience of
the evaluative on the natural, or as he often put it on the descriptive. A pair
of paintings, hanging side by side, could not differ in their goodness with-
out differing in their non-evaluative description. It is a somewhat vexed
question whether the moral supervenes on the natural—whether superveni-
ence holds between the family of moral properties and the family of natural
properties—not least because it is hard to delineate the family of natural
properties. For our dialectical purposes it won’t be important to specify the
family that is subjacent, as Hare (1984) put it. We will suppose that there
is a family of natural properties, that it includes the properties studied in
the sciences, and that it is closed under Boolean combination (so that, for
example, if P is in the family and Q is, then so is P&Q).
Hare plainly had in mind a wider class than the moral; his paintings were
to be evaluated in aesthetic terms, for instance. And nowadays metaethics
seems not to be a very natural subject if its explananda are restricted to the
moral realm proper. Philosophers tend to think the mysteries and problems
Explaining the Quasi-Real 275
of accounting for moral language, thought, and phenomena are shared by

the normative in general, or at least the practically normative. What needs
explaining, it now appears, are the categorical oughts and unconditional
goods, whether moral, aesthetic, prudential, or whatever. Still, I will stick in
this chapter to the supervenience of the moral. The aim is to make trouble
for moral realism, and then for quasi-realism. If, as most suspect, moral
realism comes together with normative realism, as a package, then there will
be trouble for normative realism too, and then for its quasi counterpart. In
any case, there is likely to be as much trouble delineating the family of the
normative as there is in marking out the family of the moral.
As I said, supervenience comes in a number of varieties. For most of
them it has seemed helpful to speak in terms of possible worlds. One kind
of supervenience, in fact, is about worlds. A family, α, globally supervenes
on another family, β, just in case there is no pair of possible worlds that
differ in some α respect without differing in some β respect. Some people,
myself included, worry about modal intuitions that can be evoked only by
quantification over possible worlds. Our home concepts are concepts of
necessity, possibility, and impossibility, and our philosophical discussions of
possible worlds are derivative from those home concepts. We are on more
solid ground if we can recover from any talk of possible worlds the natural
idiom that we grasp in common sense. For this reason I prefer to avoid
using global supervenience in an argument.
The other varieties can be captured by variations in the following formula.
⃞ (∀F in α)(∀x)[Fx →(∃G in β)(Gx & ⃞ (∀y)(Gy → Fy))]
This is Strong Supervenience. If we omit the second necessity operator we
have Weak Supervenience. (See Kim 1984 for some canonical formulations
and articulations.) And we can adjust strengths by interpreting the opera-
tors in different ways: nomic necessity, metaphysical necessity, conceptual
necessity.
We will be taking α, the supervening class, to be the moral properties,
and β, the subjacent class, to be the natural. The first operator, with widest
scope, will be interpreted as conceptual necessity, and the second as meta-
physical necessity. I believe this is the strongest relation that in fact holds
between the moral and the natural. What matters dialectically is whether it
does hold, not whether there is anything stronger that holds, so let me give
a couple of examples to try to convince you that it does.
First, here is a way of grasping what the formula says when interpreted
as I’ve just suggested. It says that a certain claim holds as a matter of con-
ceptual necessity. What claim? This one: that when something has a moral
property, it also has a natural property that necessitates the moral one. That
necessitation is expressed in the formula, by the subformula consisting of the
276 Jamie Dreier
metaphysical necessity operator and the conditional formula in its scope.

Necessitated conditionals express necessitation.
Now the examples. Jackie Robinson was a good man, on the whole. Now
consider his complete naturalistic description, including all of his relational
(natural) properties, and bundle their conjunction together and call it “D”
(Robinson’s description). Suppose someone thinks that another man could
have lived just as Robinson did, satisfied his every description, borne the
property D, while being rather less good than Robinson was. What would
we think of this suggestion? With Hare, I believe this evaluator would have
to be confused. We could ask, “In what way do you think this other man
would be worse? What is it about him, as you imagine him, that makes him
worse?” Our supervenience-flouter can give no answer. There is nothing
else about the imagined Robinson counterpart in virtue of which he is (or
would be) worse, according to the evaluator. But surely this makes no sense.
It is not just an odd ethical perspective, as if someone thought that another
man just like Jackie Robinson in all respects but with very slightly shorter
hair would have been a worse man. That would be an odd view, but per-
haps intelligible.1 To be an employment of our concept of moral goodness,
though, surely something has to be said about the respects in which a man
could fail to be as good as Jackie Robinson was.
What about me? Could I have lived a worse life than I in fact have? No
doubt; and I could have lived a better one, morally better. I remember tell-
ing a lie I should not have told. I could have lived a better life had I not
lied on that occasion. Could I have lived a better life in no natural respect
different from my actual life? I don’t see how. I don’t even see what someone
could be thinking, who thought I could have. And the “could” here is not
really an ability modal, as it might appear—I could have lived a better life
by finding a cure for arthritis, even though I never had the ability to find a
cure for arthritis. I could have lived that better life if I had been able to find
the cure.
I think these examples support Strong Supervenience. My life could not
have been better had it not been naturalistically different. No one could be
worse than Jackie Robinson without being different in some natural way.
And these couldn’ts, which express metaphysical necessity, seem to be true of
conceptual necessity, in that we would meet with confusion and not moral
opprobrium if we were to insist otherwise.
1
But perhaps not; see Foot (1958) for examples showing how doubtful it is that a
person has a moral view at all when the content of the view comes apart too far from
familiar ones.
11.2 THE EXPLANATORY CHALLENGE FOR

ROBUST REALISM
Robust Realism is just my name for non-naturalist moral realism. It was

Moore’s view.2 And although it was unfashionable for most of the twentieth
century, it is now shared by Shafer-Landau (2003), Huemer (2005), Cuneo
(2010), and Enoch (2011), among others.3 Moral facts and properties,
according to Robust Realists, are mind-independent, not constructed from
our perspectives, and they are not natural facts and properties nor reducible
to natural facts and properties.
Robust Realists have a problem explaining why the moral supervenes on
the natural. For they think these are distinct families of properties—that is,
the properties in the moral family are wholly distinct from the properties
in the natural family. They are not constructs of the natural properties, they
are not combinations of or reducible to the natural properties. But super-
venience is a necessary relation. It says that each morally good action has a
natural feature that necessitates its goodness, and each unjust institution has
an underlying descriptive nature that necessitates its injustice, and so on. The
question is, how can the items in the one class necessitate the items in the
other, if they are distinct families of properties?
The matter is different for Naturalist Realism. Constructivists like Street
(2012) have an explanation par excellence of why moral features are neces-
sitated by natural features: they are nothing but constructs of the natural
features themselves. They necessitate their consequential evaluative proper-
ties as parts together necessitate the whole they compose. And naturalistic
rationalists like Smith (1994) and so-called Cornell realists like Sturgeon
(2006) likewise provide smooth explanations, according to which logic,
analysis, or identity can explain the necessity that pins the moral to the
natural. The problem arises only for metaethical theories that take the moral
realm to be distinct from the natural one; or maybe it’s better to say that
the question of explanation arises for all theories, but has a clear and easy
answer for naturalist views.
2
I think it is not as clear as often supposed exactly what Moore’s view was; see Dreier
(2006) for my best try at working it out. Still, Moore serves as a good paradigm for the
kind of view I have in mind.
3
Perhaps including Parfit (2011) and Scanlon (2014), but there is some question of
whether it’s correct to use “Robust” to describe their realism. I won’t take a stand on the
interpretive question here. If these new realists are not robust, we might consider whether
they can end up with a better explanation of supervenience, or whether instead they just
end up in the same boat, as I argue is true of Quasi-Realism.
278 Jamie Dreier
So this was the state of the dialectic for some decades since Blackburn
(1971), or maybe even since Hare (1991). Robust Realism has an explan-
atory burden, which it will have trouble discharging, and other theories
thereby gain an advantage if they have no such trouble. But in recent years
some new doubts have arisen. The next section describes these problems
and offers my response. It is too brief to be satisfying, I am sorry to say, but
a full exploration will have to await a separate occasion.
11.3 WORRIES ABOUT MORAL SUPERVENIENCE
Matthew Kramer (2009) has suggested that supervenience is not a meta-

physical fact or conceptual truth at all, but rather a substantive, if quite
plausible, moral doctrine. It therefore requires a specifically moral expla-
nation, and the work to be done is work for first-order moral theorizing.
Kramer proceeds mainly by looking deeper into what R. M. Hare said
about which kinds of combinations of claims are ruled out by superveni-
ence, and arguing that what would be wrong with someone who insisted on
such combinations would be a fault of fairness, or a kind of moral arbitrari-
ness or partiality. Kramer may be quite right about at least some of Hare’s
claims. But as my examples (of Jackie Robinson and myself ) seem to show,
there is also a basic conceptual truth, deriving from the shape of the concept
of a moral property, that in turn entails a relation of metaphysical neces-
sitation connecting the natural with the moral. Someone who insisted that
my life could have been better—that there is a metaphysical possibility of
its being better, even with no natural difference from my actual life—seems
to be confused, and not merely unfair. Even if Kramer is right that some
similar-sounding claims can be given a substantively moral account, there is
also the metaphysical relation to be explained.
A second worry is that once the family of natural properties is expanded
enough to ensure that Strong Supervenience is indeed a conceptual truth,
it will be also a trivial truth, a kind of theorem of modal logic. Note first
that we have to decide whether spatio-temporal relations and locations are
in the natural family. On the one hand, they had better be. Someone who
holds that temporal proximity matters to moral decision and evaluation
is not merely confused about the concepts. It is a substantive moral claim
that we owe as much to distant future generations as we owe to next year’s
population, or that we have no duties to our neighbors that we don’t have
to men and women half way around the world. But then on the other
hand, if spatio-temporal relations and locations are included, then super-
venience seems trivial. No two persons or actions can share all natural
properties without sharing all moral properties, but for the trivial reason
that no two persons can share all natural properties! You and I differ in our
spatio-temporal location … and necessarily so.
Strong Supervenience is not trivialized by this observation, in fact. For
although no two actual persons share a location at a time, there are cer-
tainly two distinct possible persons, each of whom could have been right
here, right now (though not together, in the same possibility). Strong
Supervenience says that these distinct persons do not share all natural prop-
erties unless they also share all moral properties. And that is not trivial. For
instance, you yourself could have been right there, where you are and when
you are, even though you had the name “Rumpelstiltskin.” (Other natural-
istic, relational properties of yours would have had to differ, of course.) The
worry can be pressed further,4 but the threat of trivialization can be held at
bay by clear thinking.
The third problem is also a kind of trivialization worry. I will call it the
Concept Defense. And I have to give a somewhat longer treatment of it,
because the upshot of my response to this problem will be important in
the sections to follow. The Defense arises in the context of Tim Scanlon’s
non-naturalism about reasons, and specifically the view that moral and
other normative features of our world are non-natural because the property
of being a reason is non-natural. Brad Hooker and Philip Stratton-Lake
(2006) have suggested that the supervenience of this normative non-natural
property, being a reason, needs no metaphysical explanation, because it has
a purely conceptual explanation. There is something plausible about this
idea. After all, Strong Supervenience is a conceptual necessity. Conceptual
necessities are explained by the nature of concepts. So it is hard to see how
there can be a special problem for metaethicists with a particular view of
the metaphysics of moral properties. Whatever kind of explanation there
is available to other metaethical views should presumably be available to
Robust Realists. Hooker and Stratton-Lake are discussing Scanlon’s view
that the fundamental normative notion is the notion of being a reason.
They write, “We do not see a difficulty here. It is a conceptual truth that
if you have a reason to care about A, then there must be something that
provides the reason. This something is what the reason supervenes on”
(Hooker and Stratton-Lake 2006: 164). Their way of putting it is somewhat
indirect—the issue is not what the reason supervenes on, but what the prop-
erty of its being a reason supervenes on—but the point is that it is a feature
of the concept (of being a reason) that it supervenes on the non-normative
features of the situation and the agent. In the context, Stratton-Lake and
4
Thanks to Krister Bykvist for making all of this clear to me.
280 Jamie Dreier
Hooker are urging the advantages of Scanlon’s approach over Moore’s, but
I think they are mistaken.5 A Robust Realist who took goodness or wrongness
or some other moral concept as the basic one, or who took no particular
moral concept as more basic than any other, could surely offer the same
explanation of supervenience: these concepts have built into them that their
properties supervene on the non-moral ones.
Russ Shafer-Landau (2003: 86) made a similar point:
Assume for now that it is a conceptual truth that moral facts/properties/relations
are supervenient ones. The problem, then, [w]‌ould be that competent speakers of a
language can conceive of a world in which the base properties that actually underlie
particular moral ones fail to do so. But there is no mystery here, since people can
conceive of many things that are not metaphysically possible. If certain base proper-
ties metaphysically necessitate the presence of specified moral properties, then the
conceptual possibility that they fail to do so reveals only a limitation on our appre-
ciation of the relevant metaphysical relations. There is no deep explanatory puzzle
resisting resolution here.
The main idea is that apparently metaphysical necessities can often be
explained conceptually. Consider Planet–Star Necessity:
Planet–Star Necessity
It is necessary that every planet orbits a star. (This really is a matter of
stipulation, since the International Astronomical Union changed the
definition of “planet” in 2006 by fiat, and ordinary speakers seem to
defer to expert panels like the IAU.) And all planets are massive, round
bodies. But if someone complained that we cannot by our words make
it necessary that all those massive round bodies orbit the sun, they
would be missing the point. The necessity is conceptual: to count as a
planet a big round thing has to orbit a star.6
The IAU did not have to first check to be sure that the planet–star connec-
tion really is necessary. Why, then, should we think that once the conceptual
truths of moral attribution are made fully explicit, there is any remaining
necessitation to explain?
Here again is Strong Supervenience:
⃞ (∀F in α)(∀x)[Fx →(∃G in β)(Gx & ⃞ (∀y)(Gy → Fy))]
Hooker and Stratton-Lake were, I think, distracted by the first necessity
operator. Keep your eye on the second one. Suppose it is a conceptual truth
that the α properties are necessitated by the β properties. Still, we would
need an explanation for why they are. Necessitation seems to require an
5
For a thorough discussion, see Olson (2004).
6
I think I first got this analogy, in a different context, in conversation with Jon Tresan.
explanation. To see this, consider the concept of an Anankat, which I have

just made up.7
Anankats
An anankat is a cat that is necessarily on a mat. When you see a cat on
a mat, you may suspect it is an anankat, but unless it is on that mat
with metaphysical necessity, it is not an anankat.
There are no anankats. But suppose Archy is convinced that there are some.
They are generally identifiable, he thinks, by their stripes. Spying Mehitabel
on a mat and observing her stripes he concludes that she is an anankat.
Archy may still wonder how it can be true that anankats are on mats with
metaphysical necessity. What prevents them from hopping up? Whence
the metaphysical glue? How do anankats determine mats? Archy’s situation
seems very similar to ours, if we are Robust Realists about morality. We
think there are instances of moral injustice, and we think injustice is deter-
mined, metaphysically, by underlying non-moral properties. We should
wonder as Archy does. It is no answer to Archy’s question if we point out
that just as a matter of the concept of an anankat, all of them are on their
mats with metaphysical necessity.
Likewise, it is a bad answer to the question, “How does the property
underlying the very-goodness of St. Francis necessitate his very-goodness?”
to tell us that it is a moral property and so it’s conceptually necessary that
a non-moral property necessitates it. The parallel to the story of Cain and
Abel is exact, except of course that the necessity there was causal and here
metaphysical.
Finally, some new arrivals on the non-naturalist scene seem puzzled by
the demand for an explanation of supervenience. I think both Parfit (2011)
and Scanlon (2014) fit this mold, but I’ll use the latter to articulate the
attitude. Scanlon thinks there are some fundamental reason facts, facts
about what a person would have reason to do in various specific circum-
stances, which are necessary facts. The covariation between the natural and
the moral, then, is explained by these fundamental necessities, along with
the particular circumstances that particular persons are in as matters of con-
tingent fact. The reason anyone in your exact circumstances shares your
moral reasons is simply that your circumstances necessitate your having the
reasons you have.
So much is clear enough. But is there any explanation of this necessi-
tation? Your circumstances are fully natural (we may suppose), and com-
pletely characterized by your natural properties. The further property of
7
Inspired by McPherson (2012), and also by Marquis (1990).
282 Jamie Dreier
your having a reason to read carefully, or to pay your taxes, or whatever,

is non-natural, according to Scanlon. And these properties are necessarily
connected, though fully distinct. How? Scanlon has no answer to this ques-
tion. But he thinks he needs none. As Scanlon sees the issue, it is entirely
obvious and a part of common sense that having reasons is necessitated by
the fine enough details of our circumstances, and it is fully consistent with
his non-naturalist view. Why does it need any explanation?
The Supervenience Challenge is not supposed to be a refutation of
Robust Realism. It is not supposed to show an incoherence in the view or
an entailment that common sense tells us is obviously false. It is supposed
to show that there is a feature of morality, of our moral experience, that
Robust Realism cannot explain. That is a cost of the metaethical view. For
other views can explain it. Metaethical theories are supposed to explain
the interesting, striking features of moral thought and talk. Of course, not
everything is explicable. Some things are brute facts. So, it is not a decisive
objection to a metaethical theory that there is a feature of morality that it
cannot explain. But, it is a cost. And a theory had better not posit too many
brute, inexplicable facts, or it is simply a failure as a philosophical theory.
So here is the upshot of the supervenience of the moral on the natural. It
comprises a metaphysically necessary connection between moral properties
and natural properties. Some metaethical theories have excellent explana-
tions of this necessary connection. Robust Realist views seem to have a
problem. It may turn out that there is a good explanation to be provided by
Robust Realism, but there is a prima facie problem, at least. There is some-
thing to be explained, and some theories explain it, and it is not clear how
Robust Realism can give a good explanation. In the next section, I will set
out the traditional understanding of how expressivism conceives of the fact
of supervenience, and what expressivists have said by way of explaining it.
11.4 THE EXPRESSIVIST SIDESTEP
After Hare, the most influential deployment of the supervenience argument

is Simon Blackburn’s, first offered in his “Moral Realism.” I won’t bother
with the further details of how Blackburn used supervenience against moral
realism.8 According to Blackburn, and thence the conventional wisdom, his
own expressivism (not yet so-called) had a completely satisfactory explana-
tion of supervenience:
8
See Dreier (1992) for my sorting out of that argument; I still think most of the argu-
ment in that paper is correct.
If we turn again to the anti-realist explanation of (S), we can see [it has] the feature
that an attitude—the attitude of moral approval—is said to have certain properties,
and this by itself is the truth of which (S) is … a propositional reflection. Thus, the
moral attitude is said to be necessarily held because of the naturalistic properties of
its objects, and the statement of supervenience, made in terms of which differences
entail which others, is a realistic-appearing way of putting the view that difference
in moral attitudes to two things must, logically, be justified by differences in beliefs
about them.
(Blackburn 1971: 120)
In moral language, according to Blackburn, we express our moral
attitudes—complexes of emotion and commitment. The supervenience
constraint, then, is really a constraint on how we “moralize.” Part of the
language game of moral discussion involves maintaining like attitudes
toward like cases—we approve or disapprove or resent things for their nat-
ural properties, so the reappearance of those same natural properties calls
again for the same attitudes. Someone who flouted this constraint would
show that she did not understand the game, and so did not possess the
concept. The “propositional reflection” is Blackburn’s characterization of
the form of language we use in our moralizing. We speak in declarative sen-
tences when we express our moral attitudes, and the sentences have a gram-
mar that suggests—but perhaps misleadingly—that we are talking about
some independently existing moral reality. The conceptual supervenience
constraint shows up in this propositional reflection as a necessary connec-
tion between that apparent independent reality and the natural world. But
really the only necessity involved is the institutional, conceptual necessity
of our practice.
Later Blackburn put it like this:
[W]‌e could say that in the moral case as well, when we deal with analytically pos-
sible worlds, we are dealing with the beliefs we have about competence: in this case
the belief that the competent person will not flout supervenience. But this belief
is explained only by the further, anti-realist nature of moralizing. If moralizing
were depicting further, moral aspects of reality, there would be no explanation of
the conceptual constraint, and hence of our belief about the shape of a competent
morality.
Blackburn’s account of supervenience here exemplifies what I’m calling the
Expressivist Sidestep. Here’s how it works.
We might be puzzled about the nature of morality, and seek out a sea-
soned professional to set us right. “What is moral wrongness?” we ask.
Some metaethicists have a “straight” answer, but expressivists have none.
“There’s nothing helpful I can say about that,” they reply. “If you want to
284 Jamie Dreier
know what makes things morally wrong, you’ll have to ask a normative
moral theorist, and if you’re asking me about the moral properties them-
selves, I’m afraid you are just laboring under a mistaken presupposition.”
How disappointing. “But I can tell you something else that will let you
understand what’s really going on here. I can tell you what we are doing
when we call things ‘wrong’. And I can tell you what we are thinking when
we think that things are wrong, and when we disagree with others, and so
on.” And they proceed to do so. Expressivists think there is nothing much
to say about the nature of the moral realm and its elements, except of
course by moral theory. They think the illuminating way to understand our
life of moral talk and thought is by stepping off to the side and looking at
the phenomena from there. Instead of thinking about wrongness, we think
about thoughts and talk of wrongness. Everything we wanted to explain
turns out to have a much clearer explanation from this side-on vantage.
Here is the formulation of supervenience in Gibbard (2003: 90):
Two acts in two possible situations differ in being okay or not only if they differ,
somehow, in their prosaically factual properties. That is to say, for any two possible
situations s1 and s2, we have the following: only if act a1 in s1 differs factually from act
a2 in s2 will it be okay to do a1 in s1, though not okay to do a2 in s2.
Gibbard’s view works like this. People make plans, which in Gibbard’s tech-
nical sense means that they form intentions or preferences for what to do
and how to feel (and, indeed, what to believe) in various circumstances.
These contingency plans they can then express in normative language, and
that’s what normative language is for. For you think that in Sherlock’s situ-
ation one ought to pack one’s bags to prepare to escape Moriarty is for you
to plan to pack your bags if in that situation. For you to think it is “not
okay” to pack in Sherlock’s situation is for you to have plans that rule out
packing in that situation. What can be okay or not okay is an act in a situ-
ation, as the quotation above suggests. In Thinking How to Live, Gibbard
gave his own explanation of the supervenience constraint. He showed that
the constraint is a kind of theorem of the conceptual apparatus he develops,
in two stages.
First, Gibbard shows that (in his semantics) each person is committed
to all claims that would be true in each maximal specification of her plans
and beliefs together. The idea is that the plans of ordinary human beings
like us are incomplete and underspecified. We may, if we’ve been reflecting
on the Sherlock Holmes stories, have formed a contingency plan for what
to do if in Sherlock’s situation; but many people have not, of course, and
there is little practical cost to having no view about what to do if pursued
by Moriarty. Similarly we may have no belief about whether the number
of stars in the Milky Way is odd or even. But we can imagine maximally
opinionated superhumans who have definite opinions about all “matters
of prosaic fact,” like how many stars there are in which galaxies, and so
have doxastic states that can be represented by particular, individual pos-
sible worlds. And we can imagine maximally opinionated hyperplanners,
too, who have perfectly detailed hyperplans for what to do when in each
and every imaginable situation. What you are committed to, in your plan-
ning, is not just what you explicitly plan but what you would plan in every
hyperplan that fully specifies your actual plans. These might be commit-
ments that haven’t occurred to you, because you haven’t worked out all the
implications of your plans. So that is the first step: we are committed to
everything planned and believed by the maximally opinionated hyperplan-
ners whose plans are further specifications of what we’ve already planned
and what we already believe.
Second, Gibbard shows that each hyperplanner is committed to super-
venience. This is not too hard to see. It is of the nature of plans that the
contingencies for which they are plans are features of the situation that are
at least in principle recognizable.
A planner, after all, must identify acts in terms of their prosaically factual proper-
ties: a plan, say, always to do whatever is the thing to do is no plan at all. A hyper-
plan can take the infinite form, in situation S1 do the act with property P1, in
situation S2 do the act with property P2, and so on. From this we can construct the
grand property, having the property P1 in S1, P2 in S2, and so on. Call this property
Pn; the plan, then, is in any possible situation, to do something with this grand prop-
erty Pn. In a hyperdecided state, this shows, one accepts that there is a property that
constitutes being okay to do—namely Pn. And this property is constructed, finitely
or infinitely, out of factual properties.
(Gibbard 2003: 96)
Since the hyperdecided planner, no matter what her plans, accepts that
there is a natural (“factual”) property that constitutes being okay to do,
she accepts supervenience. Each of us must be committed to whatever all
hyperplanners accept, since we are committed to whatever all hyperplanners
who completely specify our own plans are committed to. So, each of us is
committed to supervenience.
This is a more elaborate and more rigorous explanation than Blackburn’s,
but it is in the same mode. To explain supervenience, we do not try to find a
metaphysical relation between properties. We look for features of the logic,
the concepts we deploy in our moral talk and thought, that account for the
a priori nature of our commitment to supervenience. That’s the Expressivist
Sidestep.
286 Jamie Dreier
11.5 THE SAME BOAT: QUASI-REALISM GETS NO

ADVANTAGE OVER ROBUST REALISM
The expressivist “side-on” accounting of supervenience looks good.

I thought it was good, for a long time, and that it was one of the important
respects in which expressivists had an advantage over Robust Realists. But
I now think I was mistaken. There is no advantage, and the theoretical situ-
ation has been misunderstood. I will begin by laying out my understanding
of the Quasi-Realist program. Then I’ll rehearse a fairly familiar worry that
arises if Quasi-Realism is successful, namely, that it becomes difficult to
distinguish it from Robust Realism. I’ll offer my own answer to this worry,
which I call the Explanation Explanation. At that point it will be easy to say
why the traditional expressivist accounts of supervenience give the view no
advantage over Robust Realism.
Quasi-Realism, named by Simon Blackburn, is the program of starting
with expressivist foundations and reconstructing the realist-sounding lan-
guage of ordinary moral talk.
I call the enterprise of showing that there is [no mistake in ordinary moral lan-
guage]—that even on antirealist grounds there is nothing improper, nothing ‘dis-
eased’ in projected predicates—the enterprise of quasi-realism. The point is that it
tries to earn, on the slender basis, the features of moral language (or of other com-
mitments to which a projective theory might apply) which tempt people to realism.
The essays in Blackburn (1993) elaborate on the project, particularly the
introduction and “Attitudes and Contents.” I want to say just a bit about
why one might think a Quasi-Realist approach is useful, and that will help
me explain what its extent should be. So first, the point of the program, as
I see it, is that many metaethicists want to avoid an error theory. They think
the conclusion that nothing is morally right or morally wrong is too dra-
matic, that it sounds like a nihilistic moral view, as on some interpretations
Nietzsche espoused. Such conclusions are beyond the purview of metaeth-
ics, according to a kind of pragmatic tradition that fits well with expressivist
theorizing. So the completion of the expressivist view should leave our main
moral thoughts and assertions largely as they were; it should explain what
we are up to, but also insofar as possible it should vindicate our practice,
not, certainly, by offering up a justification for our actual moral views but
by making sense of our methods, by showing how when we are mistaken,
we are mistaken in an ordinary moral way and not because of some enor-
mously false presupposition of everything we ever think or say. Since our
ordinary talk contains many realist-sounding elements, the completion of

the expressivist explanation of moral language must be Quasi-Realist.
Notice that the project is not, as is sometimes said, to mimic everything
that a Robust Realist says. There is no reason for an expressivist to try to do
that. According to expressivists, Robust Realists have a metaethical, meta-
physical philosophical view, and much of it is not true. So the idea that they
would want to mimic everything Robust Realists say is pretty strange. No,
the point is to vindicate ordinary moral thought and language, not to vin-
dicate heavy-duty metaphysical theorizing. There are some things Robust
Realists will say that Quasi-Realists will deny, then.
But what, exactly? Here are a few suggestions. Maybe Robust Realists
assert and Quasi-Realists deny that moral statements have truth-values. But
that would be to abandon the program. For ordinary talk is quite commit-
ted to some moral judgments’ being true and others false—for example, it
is true that slavery is unjust, and many of the moral judgments asserted by
televangelists are false. If some are true and some are false, then of course
they have truth-values. So maybe Robust Realists assert and Quasi-Realists
deny that there are moral facts. But it is true that slavery is unjust. And if it
is true, surely, it is a fact that slavery is unjust. What else is there to facts? So
Quasi-Realists agree that there are moral facts.
Talk of truth and of facts admits of a minimalist account. Minimalism
about truth, as for instance advanced by Horwich (1998), tells us that we
know everything there is to know about truth once we have in our posses-
sion the collected instances of the schema
“S” is true iff S
and know how to work out their implications. Expressivism tells us how to
use normative sentences; it has some hard tasks to execute if it is to succeed,
but that’s the program. So insofar as it is successful in these tasks, it can
bring minimalism on board, and vindicate all our talk of truth in ethical
contexts.
Minimalism about propositions says that the collected instances of
the schema
“S” expresses the proposition that S
likewise equip you with all you need to know about propositions. The
expression “the proposition that” serves as a logical-grammatical device to
form noun phrases out of sentences, helpful for generalizing. Since ethi-
cal sentences are meaningful (and embed grammatically into that clauses),
there are, trivially, ethical propositions. And as we noted, facts come free,
without further metaphysical cost, with true propositions.
288 Jamie Dreier
We can add minimalism about properties. The collected instances of

the schema
x has the property of being F iff x is F
together embody all there is to know about properties. There is a property
for each intelligible predicate, but no ontological commitment involved
in accepting these minimal, deflated properties. Once we understand that
“wrong” and the like behave logically and grammatically as predicates, there
is no further question of whether there is a property of being wrong.9
These elements of minimalism are, on the one hand, very helpful for
Quasi-Realism. They show how to vindicate fully our ordinary talk of truth,
facts, properties, propositions, perfectly consistent with expressivist scru-
ples, or as Blackburn said, “on the slender basis.” But, on the other hand,
all this minimalism also threatens to make Quasi-Realism indistinguish-
able from Robust Realism. The main things we thought Robust Realists
might say, to distinguish themselves from Quasi-Realists, it turns out that
Quasi-Realists endorse too. Or, if you prefer to see the burden placed on
the other side, the main things we thought Quasi-Realists would deny, to
distinguish themselves from the Robust kind, are things they shouldn’t deny
after all.
So that’s the problem.10 Here is how I suggest it can be solved. We want
to focus not on the things we might say when we are morally engaged (like
that slavery and apartheid share a moral property), but on the sorts of
things we want to say when we are specifically engaged in metaethics. And,
I think, what we do in metaethics is to provide explanations. So to distin-
guish Quasi-Realism from Robust Realism, we should focus on the kinds of
explanations the theories offer.
Suppose Julia believes that the experience of pleasure is intrinsically
good. What is it about Julia in virtue of which this is true of her? The kind
of explanation I want to focus on is this in virtue of kind. And here we will
find a difference between Robust Realism and the Quasi-kind. The Robust
Realist will advert to intrinsic goodness, in providing the explanation of the
facts about Julia in virtue of which she believes the experience of pleasure
is intrinsically good. There is some kind of doxastic relation that she bears
to intrinsic goodness and to pleasure, and these together constitute and
explain, in the in virtue of sense, her belief. But the Quasi-Realist does not
agree. Julia’s state of mind, according to Gibbard, is fully constituted by her
9
Perhaps. I in effect deny this claim in “The Normative Explanation of Normativity,”
a work in progress.
10
I spell this out more fully, and elaborate on the answer that follows in the text here,
in Dreier (2004).
“prosaic factual” beliefs (which have no normative contents) along with her
plans. Goodness has no explanatory role to play.11
I’ll call this the Explanation Explanation. What distinguishes
Quasi-Realism from Robust Realism is that in the latter, but not the for-
mer, the properties and facts that form the subject matter appear in the
best explanation of our beliefs and assertions with that subject matter as
content. The moral facts and properties are part of the explanation for what
it is about us in virtue of which we have moral beliefs and make moral
assertions, according to Robust Realism; according to Quasi-Realism, on
the other hand, although we certainly do make moral assertions and have
moral beliefs, and these are sometimes true, and factual, and so on, the
moral facts and properties play no role in explaining what it is about us in
virtue of which we count as saying these things and having such thoughts.
So far, so good. But now it will be easy to say why the standard story,
according to which expressivism has a satisfying account of superveni-
ence and Robust Realism has none, must be wrong. Quasi-Realists will,
naturally, agree that the moral supervenes on the natural. They agree that
there could not have been a man just like Jackie Robinson in all naturalistic
respects, but a little worse, or a lot better, or morally different. Because of
their minimalism about property talk, they agree that Robinson had a prop-
erty, perhaps hard to specify precisely, of being good to some degree, and
that this property is one that he could not have lacked unless he lacked some
natural property he did in fact have. So, they must agree that he had a prop-
erty that is necessitated, metaphysically, by his natural properties. Okay, so
what? Quasi-Realists are happy to say these things, perhaps. But then we ask
for the explanation for this metaphysical necessitation. What shall they say?
Will they give us Gibbard’s story about why all planners are committed
to supervenience? Or Blackburn’s story about why we regard it as consti-
tutive of competence with moral concepts that we respect the rule that
we always apply like moral concepts to naturalistically like cases? They are
entitled to offer these explanations. But they explain the wrong thing. These
explanations are explanations of why we accept supervenience, or why we
are committed to it. They do not explain supervenience itself. They do not
explain the metaphysically necessary connection; they explain why we are
all constrained to believe it.
Well, that’s how the Expressivist Sidestep works, after all, and on the
whole it has seemed to be a very promising strategy. Why is it no good
here? It’s no good because the problem that supervenience poses for Robust
11
Unfortunately, matters have become much more complicated with the publication
of Gibbard (2012). I try to address this complication in “The Normative Explanation of
Normativity.” I just ignore these complications in what follows.
290 Jamie Dreier
Realists is a problem of explanatory deficit. The problem is that the moral

supervenes on the natural, that they are connected by a metaphysical neces-
sity, and the Robust Realist has no good explanation of this connection.
But now it turns out that the Quasi-Realist has no explanation of the con-
nection. Of course, the Quasi-Realist has an explanation of something
else. Quasi-Realism has an explanation of why we are committed to super-
venience. That is something conceptual. Quasi-Realism has an explanation
of why supervenience is a kind of conceptual truth, something to which
we are committed no matter what our moral outlook. But, recall, some
Robust Realists make the same claim. That was the gist of the Conceptual
Defense: the supervenience of the moral on the natural is a conceptual
truth, requiring only a conceptual and not a metaphysical explanation. We
rejected the Conceptual Defense (and I noted that our reasons for rejecting
it would be more important than our reasons for rejecting the other chal-
lenges; this is why). The Quasi-Realist and the Robust Realist are in the
same boat. They recognize the relation of supervenience, of metaphysical
necessitation between the moral properties and the natural ones, and they
have no explanation of it. Quasi-Realism gains no advantage over Realism
on this score.
Contrast an older view, R. M. Hare’s, in the same broad camp as expres-
sivism. Hare never accepted supervenience as I have formulated it. He
accepted a formal mode counterpart, a principle about what it is intelligible
to say. Hare seems to have held the Quinean view that all necessity is lin-
guistic or conceptual. So Hare is in good shape with his explanation. He
has an explanation, in the same vein as Blackburn’s and Gibbard’s, of why
we follow the linguistic rule and hold others to it, and he doesn’t need any
explanation of the material mode statement of supervenience, because he
just denies it. Hare is not a true Quasi-Realist. He just denies that there
are any moral (or other evaluative or normative) properties at all; no moral
facts, no moral truths. The Same Boat verdict derives from the full-scale
Quasi-Realism that embraces minimalism.
11.6 THE WAYS FORWARD
When challenged to explain something in a philosophical domain, I think

we have some options. The most straightforward, of course, is to offer an
explanation. Naturalists, including many constructivists, do offer an expla-
nation of moral supervenience, so it’s not as though we can’t even imagine
what could count as an explanation. But the prospects for a Quasi-Realist
expressivist explanation of supervenience itself—not its formal mode
analogue of commitment to following the rule that accepting superveni-

ence embodies, but the necessary connection between the families of
properties—does not look very good. Quasi-Realists, it appears, cannot
offer the explanation embraced by naturalists. For naturalists explain super-
venience by positing identity between the particular naturalistic properties
in question—the ones doing the necessitation—and the moral properties.
And Quasi-Realists, like Robust Realists, do not accept any such identity.12
At any rate, it is, I have argued, no better than the prospects for a Robust
Realist. As far as the project of offering a straight explanation of superveni-
ence is concerned, Quasi-Realists are in the same boat with Robust Realists.
They gain no advantage on this score.
Another option, though a disappointing one, is the quietist line of
admitting that there is no explanation, but resting content with internal
consistency and agreement with common sense. As I said above, it is not
fatal to a theory to admit failure in one explanatory project. Here, again,
Quasi-Realism might just be in the same boat with Robust Realism. The
cost of explanatory failure might just be borne, if the theory is otherwise
good enough. But these options do not distinguish Quasi-Realism from
Robust Realism.
There is a third way. We might, when challenged to explain something,
confess that we cannot but offer a kind of higher-order explanation. Maybe
Quasi-Realists cannot explain supervenience, but can instead say something
persuasive about why it needs no explanation; why the apparent explana-
tory deficit is no deficit at all, or not a troubling one. It is this option that
seems most promising, to me at least, for Quasi-Realists. After all, what
distinguishes Quasi-Realism, according to my own account, is the differ-
ent role it assigns to the moral features of the world in our philosophical
explanations. To put it glibly, moral facts and properties are explanatorily
lightweight. Maybe that means the phenomena in which they participate
do not require metaphysical explanations. We will have to spell this out
more clearly.
This third path of explaining why no explanation is required itself bifur-
cates into two. One way of proceeding is to say that no explanation is
needed because the phenomenon is a kind of illusion. We can’t explain,
12
Actually, this line is more complicated than I am making it out. Michael Ridge has
suggested to me that there is a quite plausible Quasi-Naturalist line, according to which
the expressivist account of the point and expressive function of moral language is correct,
and no moral vocabulary need occur in the best explanation of our moral judgment, but
a kind of identity or claim of constitution can still be embraced. Indeed, it may be that
this is the best way of understanding Allan Gibbard’s view. I pursue a similar idea in “The
Normative Explanation of Normativity,” but here I will leave it to one side and continue
to assume that Quasi-Realism takes the form of Quasi-Non-Naturalism.
292 Jamie Dreier
because there isn’t really anything to explain. When Quasi-Realists say that
there are moral properties, and that they are necessarily correlated with
natural properties, they mean nothing more than … Something simple
and undemanding—I will try to finish the thought in a moment. I am
not optimistic about this fork, but some will find it more amenable than
I do, and I’m not averse to it—if it turns out to work I will be pleased, not
disappointed. The other fork in this path follows the idea that moral facts
and properties are of a different kind from natural ones, and because of
their nature they are particularly undemanding when it comes to explain-
ing their superficially striking patterns. This line takes the robustness of
Robust Realism seriously. There is an intelligible respect in which some
properties are robust, and their metaphysical features and relations stand in
need of explanations, while other properties are wispy and insubstantial and
explanatorily undemanding. Let me now develop these two ways of trying
to make good on the claim that Quasi-Realists can explain why superveni-
ence needs no explanation.
My Explanation Explanation of the difference between Quasi-Realism
and Robust Realism could be understood this way:
if we need F to explain what it is to believe that something is F, then
(that’s evidence that) there is F-ness; otherwise we have other reasons
to speak as if there were, but Strictly Speaking there isn’t.
Maybe we should understand it this way. Then the interpretation of
Quasi-Realism is that according to it, there aren’t really any moral proper-
ties, even though it is fine to speak as if there are. There are some contexts
in which this makes perfectly good sense. It is fine to speak, in certain con-
texts, as if there were a Santa Claus, even though there isn’t; and perhaps it
is fine to speak as if there were an average American woman, even though
there isn’t one. Some ways of thinking about ontology encourage this sort
of distinction. Quine suggested that the items in our true ontology are the
ones we are prepared to quantify over when we are in a serious philosophi-
cal mood, and not simply speaking with the vulgar. Those who like Quine’s
approach might then think of Quasi-Realists as metaethicists who are happy
to speak with the vulgar, and who (at least claim to) have shown that there
is no harm in doing so, but who do not really believe in moral properties.
In that case, it is perfectly sensible also for Quasi-Realists to deny that there
is, really, any metaphysical necessitation to explain, when it comes to moral
supervenience. They could think of supervenience as articulating a rule, but
not a truth. And they can give an explanation for why we do and should
follow the rule, but be content to have no explanation for what they could
say only a Robust Realist believes.
Here is why I am not very happy with this suggestion. Ask a Quasi-Realist
whether slavery is unjust and there is no doubt what answer you’ll get. Of
course slavery is unjust, what a question! We’re not error theorists! We’re fine
upstanding moralizers!
Now if we can use the minimalist schema
x has the property of being F iff x is F
we can conclude that slavery has the property of being unjust. No fur-
ther premise is needed—just the premise that slavery is unjust along with
the minimalist schema. A Quasi-Realist could try taking it back. He could
deny that slavery is unjust, or at least decline to assert it. But this looks like
the abandonment of Quasi-Realism. So instead, the Quasi-Realist could
wait until a serious philosophical moment, when he’s wearing his serious
metaphysicians hat, and then take it all back. “Slavery is not really unjust,
you know—I mean, at least, it is not the case that slavery is unjust, not to
say that it is just, either. It’s just okay to speak like that, with the vulgar.”
But again, this looks like the abandonment of the position in favor of error
theory. (Most error theorists think it’s fine to say the false things that so
many ordinary moralizers say.)
So, finally, a Quasi-Realist could decline to use the minimalist schema.
And that might not be so bad. She could say, “The minimalist schema is
fine for non-committal talk of properties, but do not mistake that talk for
serious metaphysical talk of serious metaphysical properties. It takes more
than the use of a grammatical predicate to commit us to the serious meta-
physical properties.” I must admit that I sometimes think this is right. Oh,
you meant, properties; I thought you just meant properties. Only I have
the sinking feeling that we don’t know what this means. What, exactly, is
the difference? What else is the robust metaphysical sort of property, besides
what we are committed to just by the minimalist schema along with the use
of the grammatical predicate?
Well, my own Explanation Explanation might answer that question.
It might be that the robust metaphysical properties are precisely the ones
that pull their own weight in certain kinds of explanations. They are the
ones that appear in the best explanation of the beliefs and assertions that
have them (the properties) in their contents. Then Quasi-Realists are dis-
tinguished precisely by their view that moral properties are not robust and
metaphysical in exactly this sense. That would be gratifying.
But now we are no longer taking the first fork I mentioned. We are taking
the second fork. The criterion for a property’s being metaphysically signifi-
cant and robust is the role it plays in the explanation of our beliefs about
it. So, there are just two kinds of properties. The claim is now that the
lightweight kind have some especially undemanding status when it comes
294 Jamie Dreier
to explanations. The claim is no longer that these properties do not really

exist. Of course they exist!
That is why I am not comfortable on the first fork. But my reasons are
far from decisive, and the first fork should still be regarded as potentially
open. Now I’ll turn to the second fork. On this fork, the interpretation of
my Explanation Explanation would be this:
the criterion for whether a property is (say) robust is whether it is
needed to explain beliefs that involve it. Only robust property
match-ups demand explanations.
This idea seems very close to a satisfying answer to our question. When a
property is not robust, call it lightweight. What makes a property light-
weight is that it does not participate in a certain central sort of explanation.
It is, so to speak, a loose cog in the metaphysical explanatory machinery. So
it is not too surprising that its correlations and necessitations don’t them-
selves have any metaphysical explanation. Interactions between the flicker-
ing images on the cinema screen typically have storyline explanations when
those images are taken to be (or represent) characters in a narrative, but they
do not really interact with one another causally. And lightweight properties
may be necessarily correlated with robust natural ones, but in the spirit of
minimalist property talk we might feel that they do not really bear meta-
physical relations to one another.
Suggestive, but not really satisfying, I think. What we know about light-
weight properties is that they play no role in the best explanation of what
it is to have beliefs about them. Why should this entail that necessary con-
nections between them and robust properties need no explanation? We are
in the right neighborhood for an explanation (for why there is no explana-
tion) but I don’t see how to reach the destination. Yes, lightweight proper-
ties (by definition) do not carry explanatory weight of a certain kind, but
why don’t their striking necessary match-ups with moral properties need an
explanation?13
I am afraid I do not have anything conclusive to say about the overall
problem. I do have a suggestive and I think interesting analogy, so I will
close with that.
13
I reached just this point, too, in Dreier (2012). That paper is also about the issue
of explaining the quasi-real, the case in point being what Sharon Street calls the cosmic
coincidence between our normative beliefs and the normative truth. A good and satisfy-
ing answer to the problem raised in this essay would, I am confident, give a fully satisfy-
ing answer to the problem raised in that one.
11.7 EXPLANATIONS IN FICTIONS: AN ANALOGY
In Ian Fleming’s James Bond novels, the hero frequently finds himself in
a perilous, almost hopeless situation. Swimming in a tank of killer sharks,
strapped to a table with the prospect of being split in two by a laser, his
chances of escape are slim. And yet, he escapes, each time. Over and over he
needs a great stroke of luck, or each of a series of events to fall just his way, if
he is to get out of the deadly jam, and over and over he just manages to pull
it off. What an astounding sequence of coincidences! There must be some
explanation. I wonder what it could be. Do the gods smile upon 007? Are
there unseen forces guiding the apparently random events? No, of course
not. It is no part of the stories that he is protected by supernatural forces.
Of course, Bond possesses unusual physical prowess, unrivaled will power,
and a steely constitution. But his skin-of-the-teeth escapes are still largely a
matter of luck, which is to say, they have no overall explanation.
Or do they? Maybe there is this explanation: Ian Fleming needs Bond for
the next novel. And his books would be dull if Bond never got into appar-
ently impossible jams. So, Fleming’s devices require this cosmic coincidence
of escapes. Maybe that explains them.
No, that’s not right. Ian Fleming’s literary (and commercial) needs do not
explain Bond’s narrow escapes. Fleming does not even exist in Bond’s world!
It is easy to confuse two questions, but they really are two separate ques-
tions. One is a question about James Bond and his predicaments. The other
is a question about the existence of stories of a certain kind. If we use “[F]‌”
as the in the fiction operator, as e.g., in Lewis (1978), we can distinguish the
questions as follows:
1. [F]‌why does Bond escape?
2. Why [F]‌does Bond escape?
The first question can’t be answered with facts about Ian Fleming, since
Fleming does not exist in the fiction and so cannot be part of any explana-
tion. Facts about Fleming can answer the second question, of course. The
first question, presumably, has no answer at all. It’s just luck. (I suppose
that’s an answer; I mean, there is no explanation for Bond’s amazing series
of escapes.) The second question gets an answer which, properly under-
stood, helps us feel better about the lack of any answer to the first question.
Sometimes it is very important to a story that a series of odd events
turns out to have a satisfying explanation. Suppose you read a detective
story saturated by strange clues, like the fact that each victim was wearing
a striped hat, that the banks were robbed only on Tuesday mornings, or
296 Jamie Dreier
that the sergeant in charge of the case and the butler’s sister each grew up
in Swansea and later moved to Market Basing. In the end, it turns out that
these coincidences have no significance at all. They were just coincidences.
This is a terrible detective story. But, that’s because of the conventions of
detective fiction—silly red herrings are not playing fair. As a general rule,
there can be pure unexplained coincidences in stories if they advance the
plot or add entertaining elements.
Perhaps the analogy is clear enough. The fictional world of James Bond
is like the “projected” world of moral facts, according to Quasi-Realism.
Things happen in this image created by our words and thoughts; the
things themselves do not depend in any ordinary causal or counterfactual
way on us (we don’t even exist in most fictional worlds), but the struc-
ture of the stories does. There is no explanation for Bond’s escapes, or
the necessary connection between the moral and the natural. There is an
explanation for why there is a story in which Bond is so lucky, and there
is an explanation for why we have an expressive practice in which our
moral property attributions are tied to the natural properties of things.
The explanations we get are not explanations of the object-level phenom-
ena, of course, but maybe they can relieve us of the felt need for such an
explanation.
Expressivists accuse Robust Realists of lacking an explanation for a
striking metaphysical necessity, the one entailed by the supervenience of
the moral on the natural. They seem to be right. But, it turns out, they
are in the same boat. The explanations they can offer are explanations for
something else, more like explaining why Ian Fleming wrote a story in a
certain way than like explaining why a secret agent escaped. If they are
Quasi-Realists, expressivists must admit the existence of the explanandum.
They can claim that it is a quasi-explanandum, but then the question is
why this is supposed to help. Why do quasi-explananda need no real expla-
nation? Because being Quasi-Real is not a way of being real? Or because
Quasi-Properties are explanatorily lightweight, in that they do not bear
explanatory weight, and so are explanatorily undemanding, in that their
connections do not need explanation? I have unfortunately been unable to
resolve this complicated question. I hope I have at least made it look like
a serious question.
References
Blackburn, S. 1971. “Moral Realism,” in J. Casey (ed.), Morality and Moral
Reasoning: Five Essays in Ethics. London: Methuen, 101–24.
Blackburn, S. 1984. Spreading the Word: Groundings in the Philosophy of Language.
New York: Oxford University Press.
Blackburn, S. 1985. “Supervenience Revisited,” in I. Hacking (ed.), Exercises in

Press, 47–68.
Cuneo, T. 2010. The Normative Web: An Argument for Moral Realism.
New York: Oxford University Press.
Dreier, J. 1992. “The Supervenience Argument Against Moral Realism,” Southern
Dreier, J. 2004. “Meta-Ethics and the Problem of Creeping Minimalism,”
Dreier, J. 2006. “Was Moore a Moorean?”, in M. Timmons (ed.), Metaethics after
Moore. Oxford: Oxford University Press, 191–207.
Dreier, J. 2012. “Quasi-Realism and the Problem of Unexplained Coincidence,”
Analytic Philosophy 53: 269–87.
Enoch, D. 2011. Taking Morality Seriously: A Defense of Robust Realism.
Foot, P. 1958. “Moral Beliefs,” Proceedings of the Aristotelian Society 59: 83–104.
Gibbard, A. 2012. Meaning and Normativity. Oxford: Oxford University Press.
Hare, R. M. 1984. “Inaugural Address: Supervenience,” Proceedings of the Aristotelian
Society Supplementary Volume 58: 1–16.
Hare, R. M. 1991 [1952]. The Language of Morals. Oxford: Oxford University Press.
Hooker, B. and Stratton-Lake, P. 2006. “Scanlon vs. Moore on Goodness,” in
M. Timmons (ed.), Metaethics after Moore. Oxford: Oxford University Press, 149–68.
Horwich, P. 1998. Truth. New York: Oxford University Press.
Huemer, M. 2005. Ethical Intuitionism. New York: Palgrave Macmillan.
Kim, J. 1984. “Concepts of Supervenience,” Philosophy and Phenomenological
Kramer, M. H. 2009. “Supervenience as an Ethical Phenomenon,” in Kramer,
Moral Realism as a Moral Doctrine. Oxford: Wiley-Blackwell, 304–64.
Lewis, D. 1978. “Truth in Fiction,” American Philosophical Quarterly 15: 37–46.
McPherson, T. 2012. “Ethical Non-Naturalism and the Metaphysics of
Supervenience,” in R. Shafer-Landau (ed.), Oxford Studies in Metaethics, Volume
7. Oxford: Oxford University Press, 205–34.
Marquis, D. 1990. Archy and Mehitabel. New York: Anchor Books.
Olson, J. 2004. “Buck-Passing and the Wrong Kind of Reasons,” Philosophical
Quarterly 54: 295–300.
Smith, M. A. 1994. The Moral Problem. Oxford: Blackwell.
Street, S. 2012. “Coming to Terms with Contingency: Humean Constructivism
about Practical Reason,” in J. Lenman and Y. Shemmer (eds.), Constructivism in
Practical Philosophy. Oxford: Oxford University Press, 40–59.
Sturgeon, N. L. 2006. “Ethical Naturalism,” in D. Copp (ed.), The Oxford Handbook
of Ethical Theory. Oxford: Oxford University Press, 91–121.
Index
abilities normative properties and

obligation and commendation of 231–3
acting for the right reasons Balaguer, M. 89, 91 n29, 94, 98 n47
27–8, 33–42 basic normative properties 229–30
‘ought’ 44–50 beliefs
absolutism see metaethical absolutism broad and narrow 2, 8–10
acting for the right reasons 33–42 and moral uliefs ix, 14–23
Credit 37, 39 and moral testimony 2–7, 9
Explanatory Condition 34, 39 and moral uncertainty 54, 55, 56,
intelligibility reasons 40–1 59–62 69–73, 75–8
and moral worth 37 n18 false moral views 56, 57, 60, 62–8,
new information problems 42–50 74, 75, 78
Physiological Ability Condition see also mathematical beliefs; moral
33, 45–6 beliefs
Right Reasons Ability Condition 33–8, Benacerraf, P. 85–6, 87 n17, 98
40, 41, 46 Bengson, J. 86 n13, 96 n42
Sensitive No Accident 36, 37, 39 Bennett, K. 195 n14
Sensitivity Condition 34, 39 Berker, S. 91 n27, 115, 116 n14
Advice Problem 43 Björklund, F. 4 n5, 167 n22, 242 n31
Almér, A. 171 n28, 173 n30, 176 n37, Björnsson, G. 42 n24, 50
176 n38, 181 n44 n31, 109 n6
Alston, W. 117 n16 metaethical absolutism xi, 160–84
alternative grounding views Blackburn, S. xiii, xiii n2, 1 n2, 8 n13,
on the normativity of morality 137, 8 n14, 161 n3, 167 n23, 183
138, 154, 155, 156 n46, 278
apology 223 on autonomy 188, 198 n24,
Arpaly, N. 36 n15, 37 n18, 67 210–11, 213
n20, 68 n21 conceptually possible worlds 249 n4
atomic nihilism 210 Essays in Quasi-Realism 286
Austin, J.L. 221, 223 n6, 233–4, 235 “Moral Realism” 282–3, 285
authoritative normativity 148 supervenience 289, 290
autonomy and moral judgements 7 challenge 245, 246–50, 253
autonomy theses xi–xii, 188–213 blameworthiness
and deontic facts 199 acting for the right reasons 38
the ethical and the metaethical 190 moral uncertainty 56, 56 n4, 60, 77
grounding relations 189, 195–7 false moral views 62–8, 75
Humberstone/Lewis principles 192 BonJour, L. 125
n5, 201 Boyd, R. 82 n3, 104
nihilism xii, 208, 210–12, 213 Bratman, M. 121 n21
taxonomy and autonomy 189, Brink, D.O. 1 n1, 82 n3, 160 n2
190–1, 210 Broome, J. 107 n3, 145, 154
see also logical autonomy; metaphysical Brown, C. 194 n10
autonomy Buekins, F. 163 n10
Azzouni, J. 91 n29
Casey, J.
Bader, R. 197 n21, 198 n22 Morality and Moral Reasoning xiii n2
badness categorical reasons/oughts 141 n8
300 Index
Chihara, C. 82 n3 converse metaphysical autonomy 195,

Chisholm, R.M. 125 200, 201
Chrisman, M. 229 n16 challenges to 204–5, 208, 209
Chudnoff, E. 197 n21 Copp, D. 1 n2, 224 n7
claims and moral unity 173, pluralist-teleology 137
174–8, 182–3 rationality and moral authority
Clarke-Doane, J. xi, 145–58
justification and explanation in “Why be Moral?” question 135
mathematics and morality correctness 161, 164, 165, 169, 170–1, 172
x, 80–100 non-absolutist moral unity 182, 183
Coady, C.A. 1 understanding 178–81
cognitive universalism 246, 250–62 Correia, F. 196 n15
normative judgments about kinds as Cuneo, T. 123 n25, 127 n30, 277
basic 259–62
strong supervenience 254, 262–5 Dancy, J. 40 n21
truth-conditions 246, 255–7, 259–60, Darley, J.M. 162 n7, 167 n19
262, 263, 267 D’Arms, J. 237 n25
cognitivism Darwell, S. 240, 241
defining 1 Darwin, C. 92, 93
moral judgement theory ix, 1–23 Davidson, D. 8, 8 n12
beliefs and testimony 2–7, deflationary view of obligation 31–2
9–10, 12–13 deliberative indispensability
broad and narrow beliefs 2, Enoch’s Robust Realism and 107–9
8–10, 14, 15 epistemic justification of xi, 104–31
moral testimony 3–7, 9–10, 12–13, instrumentalist indispensability
14–16, 22, 23 118, 122
moral uliefs ix, 2, 10, 14–23 intrinsic indispensability 118
moral understanding 10–14, 22 practical deliberation 108, 109, 111,
non-cognitivism 1–2, 8 n13, 14 n24, 22 121, 123, 126
redefining 21–3 see also Indispensabilism
Cohen, G.A. 188, 194, 196, 213 deliberative obligations 27–8, 28–32, 44,
Cohen, S. 119 45–7, 49–50
Colyvan, M. 87, 99 descriptive properties
commendatory properties of judgments about kinds 261
normative properties as 220–30 normative properties supervening on
objections to the theory 231–7 245, 246–50
communicative function in descriptivism
disagreement 173–8 about normative utterance 225
comparative normative judgments 230 desire
competence-rationality 143 broad and narrow sense of 8
Conceptual Defense of supervenience disagreement 161, 164, 165–6, 169,
279–81, 290 170, 172
“condemning” and normative non-absolutist moral unity 182, 183
properties 227 understanding 173–8
constructivism disjunctive normative properties 229–30
deliberative indispensability and dispraise 227 n13, 233
128, 131 Doefler, R. 162 n7
and moral supervenience 290 Douven, I. 117
and naturalist realism 277 Dowell, J. 42 n24, 55 n3, 171 n28
context change potential Dreier, J. xiii, xiii n3, 138, 167 n22,
in the communicative function 183 n45
174 n32 explaining quasi-realism 273–90
contingent claims ‘The Normative Explanation of
a priori justification of 121 n23 Normativity’ 288 n9, 289 n11
Index 301
Driver, J. 4 n4 epistemic responsibility 125–6
Dworkin, R. 104, 188, 194, 211, 213 epistemic teleology 115
epistemology
Egan, A. 162 n7, 163 n11, 169 n25, and moral uncertainty 70–4
172 n29, 177 n40 ethical facts 105 n1
Eklund, M. 228 n14 challenges to converse metaphysical
Ellis, B. 91 n29 autonomy 204–5
empirical justifications counterexamples to logical
of mathematical beliefs 98–9 autonomy 200–1
empiricism deliberative indispensability and
Quinean 82, 83–4 104, 109, 111, 120, 121, 122,
Engel, P. 9 n15 127–8, 131
Enoch, D. xi, 271 ethical and non-ethical facts 200 n26,
“An Outline of an Argument 209, 210–11, 212
for Robust Metanormative metaphysical autonomy thesis
Realism” 104–5 194–5, 197–9
defense of Indispensabilism 106–12 negation of 209
IBE (inference to the best explanation) taxonomy of 190–1
112, 113, 117, 126 taxonomic essentialism 206–7
Pragmatic (vindication of ethical hereabouts 205
Indispensabilism) 111–12, 129 ethical normativity 106
case against 112–24 ethical reasoning
and epistemic responsibility 125 and deliberative indispensability 119
robust realism 104–5, 107–9, 126–7, ethical responsibility 125
128, 130, 131, 209 n37, 277 evidence
Taking Morality Seriously 104 beliefs and 2–7, 9 n16, 14–15
epistemic deliberation 46–7 expressivism
epistemic justification expressivist sidestep 273, 282–5, 289–90
alternating vindicating accounts moral belief 8 n14
of 124–6 and quasi-realism 286–7, 288
deliberative indispensability and
xi, 104–31 Fantl, J. 115
distinctiveness of 106 fiction, explaining quasi-realism in 295–6
and Enoch’s defense of Field, H. 82 n3, 84, 86, 87 n17, 89, 92,
Indispensabilism 106–12 94, 98, 195
Enoch’s Pragmatic account of 111–12 Fine, K. 195, 196 n16, 197 n19–21, 198,
case against 112–24 201 n27, 245, 262 n11
and ethical justification 113, 116 “transcendental” facts 270
explanatory indispensability in 117 “unworldly” predicate 264
foundationalism and 111, Finlay, S. 42 n24, 50 n31, 161 n3, 169
122–3, 126–7 n26, 177 n41, 181 n44
moral judgements 23 on normative properties 216, 219
moral uncertainty 75–8 n3, 241
norms of 106 n2 Fitzpatrick, W. 66 n17
sources of basic 110 Fletcher, G. 227 n12
testimony as a source of 122–3 Foot, P. 141, 276 n1
Truth-Directness 106, 107, 113–17, Forrest, P. 268 n16
121–4, 125, 126–8, 129–30 foundationalism
epistemic modals (likelihood judgments) and epistemic justification 111,
161, 170–3 122–3, 126
epistemic normativity 106, 126 functionalist psychology
epistemic rationality deliberative indispensability and 118 n18
and Indispensabilism 105 fundamental standard for
and moral judgements 23 judgments 179–80
302 Index
Gauthier, D. 135, 140, 149 Horwich, P. 287

Gendler, T. 17 n25 Huemer, M. 160 n2, 277
Gibbard, A. 1 n2, 8 n13, 89 n24, 115 Humberstone, L. 192 n5
n12, 161 n3, 164 n12, 167 Hume, D. xi–xii, 188–9 n3, 193
n22, 183 n46, 240, 241 Humean theory of reasons 211
analyticity of supervenience Hume’s Dictum 245, 266, 271
245–6, 250 Husi, S. 109 n6
norm-expressivism 254 Hussar, K.M. 4 n5
quasi-realism 288–9, 290 hybrid theories of normative
“Thinking How to Live” 284–5 judgments 225
Gibbons, J. 29 n4, 38 n19, 40 n21 Hyman, J. 38 n19
Gilles, A.S. 172 n29, 178 n42, 181 n44
Glanzberg, M. 163 n9 Ichikawa-Jenkins, C. 197 n19
global supervenience 275 independence 164, 165, 169, 171, 172
Glover, J. 21 n28 Blackburn’s supervenience
Gödel, K. 83 n5 challenge 246–7
Goldman, A. 124 non-absolutist moral unity 182, 183
Goodwin, G.P. 162 n7, 167 n19 understanding 178–81
Graham, P.A. 26 n1, 27, 32 n8 Indispensabilism 129–31
Greco, J. 125 abandoning 126–9
Greene, J.D. 4 n5 alternative vindicating accounts of 124–6
Green, M. 222 n5 case against Enoch’s Pragmatic 112–24
Grimm, S. 11 n18 Enoch’s defense of 106–12, 129
Guerrero, A. 54 n1, 57 n5, 58 n6, 63 n9 alternatives to 107
rejection of 106, 123–4
haecceitic kinds 252–3 Tempting Idea of 105, 106, 107, 109,
Haidt, J. 4 n5 128, 129, 130–1
Hare, R.M. 161 n3, 226, 231, 278 virtues of 105–6
supervenience 245–6, 248, 253, 274, see also deliberative indispensability;
276, 278, 290 epistemic justification
universal prescriptivism 254 internalists
Harman, E. accepting pure testimony 6 n10
“Ethics is Hard. What Follows?” 66
n18, 68 n22 Jackson, F. 30 n6, 32 n8, 55 n3, 188
irrelevance of moral uncertainty x, 53–78 n1, 266 n14
Harman, G. x, x n1, 20 n27, 80, 81, Jones, K. 4 n4, 14 n24
104, 209 n37 Joyce, R. 83–5, 86 n13, 88, 95 n38,
Harman’s Objection to moral beliefs 99, 139
82–3, 83–4, 89 n23, 95–6, 100 on nihilism 209, 211 n40
Harris, P.L. 4 n5 justificatory challenge
Hart, W. 87 n16 for mathematical realism x,
Hawthorne, J. 88–9 n36, 115 80–3, 87, 99
Heathwood, C. justificatory explanations of obligation 34
normative properties xii, 216–43
Heller, M. 88–9 n36, 237 n25 Kamtekar, R. 20
Hills, A. Kant, I. 140
cognitivism about moral judgement categorical imperative 144
ix, 1–23 Groundwork 4 n6, 134–5
moral understanding 10–14 Kantianism and moral uncertainty 54
hindsight 172, 183 Metaphysics of Morals 4 n6
Hooker, B. 279–80 on moral testimony 4 n6
Hopkins, R. 4 n4, 7 prescriptivism 164
Horgan, T. 8 n14, 9, 16, 91 n29 Karmo, T. 193
Hornsby, J. 38 n19, 40 n21 Keller, S. 77 n35
Index 303
Kim, J. 275 deliberative indispensability and
Kitcher, P. 11 n18 epistemic justification x–xi, 104–31
Knobe, J. 173 n30 Maguire, B.
knowledge grounding the autonomy of ethics xii,
moral knowledge 10–14 188–213
pragmatic encroachment about 115 Maitzen, S. 193, 202, 205–7
Köbel, M. 169 n25 Markovits, J. 36 n15, 37 n18, 67
Kohlberg, L. 4 n5 n20, 68 n21
Kolodny, N. 32 n8, 48 n28, 55 Marquis, D. 281 n7
n3, 107 n3 mathematical beliefs
Korman, D. 94 Benacerraf-Field challenge 85–7, 92,
Korsgaard, C.M. 104, 128 n31, 135, 94 n35, 98, 100
140, 149, 150 empirical justifications of 98–9
Koslicki, K. 257 n10 genealogical debunking
Kramer, M. 253 n6, 278 arguments for 99
Kratzer, A. 48 n28 Indispensability Thesis 82–3, 83, 84–5,
Kyanvig, J. 11 n18 86–7, 91–2, 92 n30, 94, 98, 99
justificatory challenge for x,
Lackey, J. 77 n35 80–3, 87, 99
Lasersohn, P. 163 n11, 169 n25 reliability challenge for x, 80, 84, 85,
Lehrer, K. 119 87, 89–90, 99–100
Lemos, N. 229 n16 safety of 94
Lenman, J. 109 n6 sensitivity of 92 n30, 94 n35
Lennertz, B. 178 n41 Merricks, T. 91 n27
Leslie, S.-J. 257 n10 metaethical absolutism xi, 160–84
Lewis, D. 192 n5, 295 and absolutism 163–4
likelihood judgments 161, 170–3, blocking of essential dependence
177–8, 182 165 n15
communicative function of 180–1 challenges to 201–8
Lipton, P. 11 n18, 117 hindsight 172
local rationality 73 n27, 74 n28 motivational internalism 167, 168
Lockhart, T. 54 n2, 57 n5, 58 n6 satisfaction conditions 161–2
Loeb, D. 162 n7 see also moral unity
logical autonomy 188–9, 191–3 metaphysical autonomy 188–9, 189–90,
counterexamples to 200–1 193–201
non-vacuous 192–3, 199 challenges to 202–3
simple 191–2, 193, 201, 207 ethical facts 194–5, 197–9
Lord, E. ethical hereabouts 205
obligation and acting for the right grounding claims 195–7, 203
reasons ix–x, 26–51 modal status of ground 197–9
Lynch, M.P. 179 n43 implications of 208–13
Lyons, D. 160 n2 negation 208–9, 212
particularism 207–8
McDowell, J. 12 n21 taxonomic essentialism 205–7
MacFarlane, J. 32 n8, 42 n28, 55 n3, see also converse metaphysical
172 n29 autonomy; nihilism
McGrath, N. 115 metaphysical challenge to supervenience
McGrath, S. 4 n4, 7 246, 265–6
McKenna, R. 169 n26 metaphysical grounding
Mackie, J.L. 135, 140, 211 n40, 238, rationality and 136–7, 153
239, 241, 268 n16 metaphysical necessities
McNaughton, D. 160 n2 supervenience and 280–1
McPherson, T. 161 n3, 183 n46, metaphysics of obligation 44, 45–7
281 n7 Milgram experiments 19–21
304 Index
Millikan, S. 174 n31, 175 n33 aspects of unity in non-absolutist

Mill, J.S. 209 n36 discourse 168–73
minimalism non-absolutist account of 182–4
and quasi-realism 287–8, 290, 293 Moser, P. 116
minimal moral belief 8 n14 motivational judgment internalism
Moller, D. 54 n1, 57 n5, 58 n6 223–4, 241
Montmarquet, J. 125 n27 commending the bad 231
Montminy, M. 171 n28, 178 n41 and reductive naturalism 239, 240
Moore, G.E. 26 n1, 27, 219, 246, 267, 277 My Lai massacre 21 n28
moral authority
rationality and xi, 145–58 narrow beliefs 2, 8–10
moral beliefs naturalism
core 93–4 non-reductive 219 n4, 237 n26
genealogical debunking arguments for quasi-naturalism 291 n12
83–5, 93, 96, 98 reductive 218, 237 n26
justificatory challenge for x, 82–3, 87 natural properties
Modal Security 95–8, 100 as essentially commendatory 234–7
ordinary object beliefs and 90–1, 92, necessary and transcendental
94–5, 96 n41 truths 270–1
safety of 92–5, 96 negation 208–9, 212
sensitivity of 87–92, 96 Nelson, T.M. 193 n8
moral epistemology Neta, R. 29 n4
and judgments about kinds 260–1 Nietzsche, F. 286
moral judgement theory see cognitivism nihilism xii, 208, 210–12, 213
moral knowledge 10–14 atomic 210
moral motivation classic argument for 211–12
rationality and the lack of 151–2 and natural properties 255
moral platonism 267–71 normative 234
moral relativism 7 and quasi-realism 286
moral sensibility robust 210
and cognitive universalism 261–2 nihilist’s principle 211, 212, 213
moral subjectivism 6–7 non-cognitivism
moral supervenience see supervenience about moral judgement 1–2, 8 n13, 14
moral testimony see testimony n24, 22
moral truths 76 n31 metaethical absolutism 160, 161
explanations of 11 n19 metaphysical autonomy 196
moral uliefs see uliefs and normative predicates 245–6
moral uncertainty x, n52– 7, 53–78 normative properties 224–5
Actualism 58–62, 68, 69–73 commending the bad 231
and epistemology 70–4 rationality doctrine (RD) and 134
and non-moral beliefs 67–8 non-descriptivism
and non-moral uncertainty 61, 62 n8 about normative utterance 225
objective moral truths 55 non-naturalism
subjective moral truths 55 and irreducible normativity 237
Uncertainism 52–7, 69–78 n26, 242–3
Epistemic Uncertainism 71, 73–4 and motivational judgment
false moral views 56, 57, 60, 62–8, internalism 241
74, 75, 78 normative xii, 216, 217–19, 220–1
moral claims 55 norrmative xii, 216, 217–19
two revisions of 75–8 quasi-naturalism 291 n12
moral understanding 10–14, 22 queerness of normative
moral unity properties 239
argument from unity 160–1, 164–8 and reductionism 238, 240
Index 305
non-naturalist realism second-order universals 268–9
and explanation of supervening on descriptive properties
supervenience 281–2 xii, 245, 246–50, 267–8
metaphysical challenge to unknowing attributes of 227–9
supervenience 246, 265–6 see also non-naturalism; normativity;
moral platonism 267–71 speech acts
and the supervenience objection normative realism 275
xii, 245–50 normative reasons
see also cognitive universalism; robust and normative oughts 142
realism for obligation 27, 28–9
non-reductionism normative statements
normative 217, 219 subjective and objective 55, 55 n3
non-vacuous logical autonomy 192–3 normative thought 225–6 n10
no relativization 164, 165, 169, 171, 173 normative truths 90–1
non-absolutist moral unity 182 normative utterances 224–5, 235
understanding 178–81 normativity
normative facts 217–18 authoritative 148
normative nihilism 234 essential practicality of 238, 241–2
normative predicates formal vs robust 111 n10
applied to particulars and kinds xii, and normative properties as
245, 250–9 commendatory 233–4
kind-applying senses as basic 259–62 and rationality 135, 139–46
kind-applying senses as Basic Linkage Thesis (BLT) 135–6,
unworldly 264–5 146–52, 157
metaphysical challenge to concept of normativity 140–2
supervenience 246, 265–6 concept of rationality 142–6
supervenience objection to norm-expressivism (Gibbard) 254
245, 246–50 Nozick, R. 88 n19, 125
see also cognitive universalism
normative properties xii, 216–43 objective moral claims 55, 56, 70 n24
basic 229–30 obligation 26–51
as second-order 246 deflationary view of 31–2
as commendatory properties deliberative 27–8, 28–32, 44,
220–30, 242 45–7, 49–50
“commending” and metaphysics of 44, 45–7
“condemning” 226–7 objectivism about ix, 26–7, 28, 30,
objections to 231–7 42, 47, 51
comparison with hybrid and ability conditions 33, 34–5,
theories 224–5 38, 45–6
contrast with motivational judgment perspectivalism on ix–x, 26–7, 28–9,
internalism 223–4, 231, 239, 30–2, 50–1
240, 241 and ability conditions 35–42
disjunctive 229–30 and deliberation 46–7
grasping 228–9 new information problems 42–4
normative judgments 226 and 'ought’ claims 49
comparative 230 possession of reasons for 27,
and descriptive judgments 28–9, 38–42
245, 246–50 semantics of 44, 48–50
positive and negative normative facts 217 simple ignorant cases 29–30
queerness of irreducible xii, sophisticated ignorant cases 29, 30–2,
222, 238–9 43–4, 47 n27
reductive account of subjective obligations 31, 32
particular-applying 267–8 see also acting for the right reasons
306 Index
Olinder, R.F. 109 n6 queerness of normative properties xii,

Olson, J. 280 n5 222, 238–9
ought-fundamentalism Quine, W.V.O. 81–2
and normativity 140, 141–2 quasi-realism and 292
Owens, D. 9 n15 Quinean empiricism 82, 83–4
Parfit, D. 30 n6, 140–1, 144–5, 188, Railton, P. 240, 241

219, 246, 267, 271, 281 rationality
robust realism 277 n3 concept of 142–6
partially reductionist views moral authority and xi, 145–58
on the normativity of morality 137–8, Rationality Doctrine (RD) xi, 134–58
141 n9, 145 n14, 155 arguments against 152–6
particularism from other options 153–4
metaphysical autonomy 199 parity argument 154–5
n25, 207–8 Basic Linkage Thesis (BLT) 135–6,
past obligations problem 42–3 146–8, 157
Pearson, H. 163 n11, 169 n25 motivations for 148–50
Pendlebury, M. 239 n29 objections to 150–2
perceptual beliefs failure of 138–9, 157
reliability challenge for 89–90 formulation and arguments 136–9
performance-rationality 143–5 linkage thesis 135, 136, 146, 147
phronesis 12–13 and the normativity of
Pigden, C. 188 n2, 192–3, 210 n38 morality 157–8
Plunkett, D. 161 n3 normativity and rationality
deliberative indispensability 135, 139–46
and epistemic justification practical rationality 139
x–xi, 104–31 Reasons Linkage Thesis (RLT)
practical oughts 141 147, 148
practical reason realism
moral authority 135, 141, 142, 144, justificatory challenge for mathematical
145, 147, 150, 151, 154, 157 realism 80–3
predictionalist approach naturalist 277
to metaphysical autonomy 195–6 normative 275
primitivism rationality doctrine (RD) and 134
Blackburn’s supervenience supervenience and moral realism
challenge 246–50 274, 275
rationality and moral authority 137, see also non-naturalist realism;
138, 154, 155, 156, 157 quasi-realism; robust realism
Prior, A.N. 188, 191, 200 reasons-fundamentalism
Pritchard, D. 11 n18, 93 n32 and normativity 140, 141–2, 145 n15
prudential uncertainty 57 reductionist views
Putnam, H. 81 on the normativity of morality 137–8,
154–5, 157
quasi-naturalism 291 n12 reductive naturalism 218, 237 n26
quasi-non-naturalism 291 n12 why reductionism fails 238, 239–40
quasi-realism 286–96 reflective equilibrium
Dreier’s explanation of xiii, 273, 286, and normative thinking 261, 262
289, 290–4 Regan, D. 30 n6
in fiction 295–6 reliabilism
expressivist sidestep 273, epistemic justification and 124–5
282–5, 289–90 reliability challenge
robust realism and xiii, 273–4, 279, for D-realism/mathematical realism x,
286–90, 291–4 80, 84, 85, 87, 89–90, 99–100
Index 307
responsibilism simple logical autonomy 191–2, 193,
epistemic justification and 125–6 201, 207
Ridge, M. 1 n2, 167 n22, 175 n34, Sinclair, N. 8 n14
291 n12 Sinnott-Armstrong, W. 88 n21
rights claims Skarsaune, K.O. xii, 196 n17
and comparative normative properties supervenience objections 245–71
230 n17 Blackburn’s challenge 245, 246–50
Roberts, D. 229 n16 cognitive universalism 250–62
robust nihilism 210 Smith, H. 55 n3
robust realism Smith, M. 22 n30, 57 n5, 68 n23, 138,
Enoch’s defense of 104–5, 127, 128, 139, 147 n18, 149, 150, 160 n2
130, 131, 209 n37 on nihilism 209 n37, 211 n40
explanatory challenge for 277–8 on normative properties 231
quasi-realism and xiii, 273–4, 279, Smith, M.A. 277
286–90, 291–4 Sosa, E. 29 n4, 117 n16, 125
supervenience and 281 speech acts
Rosen, G. 63 n11, 64 n12, 66 n15, apology 223
136, 198 n23, 205 n34 attribution and normative
Ross, J. 27, 30 n6, 42, 54 n1, 54 n2 properties 229
Ross, W.D. 83 n5 commendatory 220, 221–4
Roush, S. 125 “commending” and “condemning” 227
Ruse, M. 88 n29, 93 and comparative normative
Russell, G. 199 judgments 230
conflict of 232, 233
Sabini, J. 20 normativity of 216–17
Samuel, J. 191 n4 positive attitudes and 224 n8
Sarkissian, H. 167 n19, 183 n47 practicality of normativity in 241
Sayre-McCord, G 160 n2 see also normative properties
Scanlon, T.M. 104, 140 n6, 188, 195 Stalnaker, R. 174 n32
n13, 197–8, 211, 213, 271 Stanley, J. 115
robust realism 277 n3, 280 Steiner, M. 86
supervenience and 281–2 Stephenson, T. 163 n11, 172 n29
Schaffer, J. 163 n9 Stevenson, C.L. 161 n3, 167 n22,
Schapiro, T. 128 n31 174 n31
Schechter, J. 84, 89 n24, 89 n25, 110, Stich, S. 116
111, 114 n11 Strandberg, C. 183 n45
Schoenfeld, M. 74 n28 Stratton-Lake, P. 279–80
Schroeder, M. 1 n2, 8 n13, 188 n1, Street, S. 84, 85 n11, 89 n24, 90–1, 93,
207, 225 n9 95 n37, 96 n40, 164 n12, 277
Searle, J. 195 n12, 221 n5 quasi-realism 294 n13
self-attributions Streiffer, R. 160 n2
of natural properties 236–7 strong supervenience 254, 271, 275–6
sense experience 81 n2 analyticity of 262–5
Sepielli, A. 54 n1, 54 n2, 70 n26, normative-descriptive 248
73 n27, 74 n28, 75 n29, properties of 278–9, 280–1
76 n32 Stroud, S. 77 n35
Setiya, K. 22 n30 Sturgeon, N. 82 n3, 88 n22, 167 n23
Shafer-Landau, R. 144, 160, 238 n27, Sturgeon, N.L. 277
277, 280 subjective moral claims 55, 56, 70
Shah, N. 115 n12 n24, 74–5
Sider, T. 88–9 n36 substantive conception of possible
Silk, A. 88 n22, 183 n45 worlds 270–1
Silver, M. 20 Sundell, T. 161 n3, 163 n9, 165 n16
308 Index
supervenience xii–xiii truth-conditions

cognitive universalism 250–62 cognitive universalism 246, 255–7,
Conceptual Defense of 279–81, 290 259–60, 262, 263, 267
expressivist sidestep 273, 282–5
global 275 uliefs ix, 2, 10, 14–23
metaphysical autonomy 198–9 and N-beliefs 17–21
metaphysical challenge to 246, 265–6 uncertainty see moral uncertainty
moral platonism 267–71 understanding
normative-descriptive 245, 246–50 moral understanding 10–14, 22
problem 273, 274–6 universal prescriptivism (Hare) 254
properties of 274–5, 275–6, utilitarianism
278–9, 281–2 and moral uncertainty 54
quasi-realism and the explanation normative enquiry and 261
problem 273, 289–90, 290–4
robust realism 277–8 Van Inwagen, P. 91 n27
worries about moral Van Roojen, M. 1 n1
supervenience 278–82 vegetarianism 4 n5
see also strong supervenience; weak Velleman, J.D. 8 n14, 9 n15, 17 n25,
supervenience 115 n12
Svavarsdóttir, S. 183 n45 virtuous persons
Swanson, E. 171 n28 and moral judgements 21 n29, 23
Sylvan, K. 29 n4, 34 n11 Von Fintel, K. 172 n29, 178 n42,
181 n44
Tackett, B. 237 n25 Vranas, P. 202
tastiness judgments 161, 162–3, 164
n12, 167, 168–9, 173, 175, weak supervenience 275
176–7, 182 cognitive universalism 250–5, 260
communicative function of 180 normative-descriptive 248
taxonomy and autonomy Weatherson, B. 57 n5
ethical taxonomy 189, 190–1 Wedgwood, R. 9 n15, 115 n12
nihilism 210 White, R. 119 n19
taxonomic essentialism 205–7 Willer, M. 177 n41
Tersman, F. 161 n3 Williams, B. 4 n6, 8, 219
testimony Williamson, E. 29 n4
moral judgements 3–7, 9–10, 12–13, Williamson, T. 9 n15, 46 n26
14–16, 22, 23 Wilson, J. 266
moral uncertainty 76, 77 Wodak, D. 230 n17
as a source of epistemic Wolf, S. 63 n11
justification 122–3 Woodward, J.B. 11 n18
Thomson, J.J. 26 n1, 27, 43, 224 n8, Wright, C. 116, 179 n43
227 n13, 230, 242 n31 Wright, J.C. 167 n19
Timmons, M. 8 n14, 9, 16
tractarian conception of possible Yalcin, S. 173 n30
worlds 270–1 Young, L. 173 n30
transcendental and necessary
truths 270–1 Zagzebski, L. 11 n18, 125 n27
Trogdon, K. 110 n7 Zimmerman, M. 63

Landau Metaethics 10

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Landau Metaethics 10

Hochgeladen von

Copyright:

Verfügbare Formate

OXFORD STUDIES IN METAETHICS

1 Cognitivism about Moral Judgement 1

Gunnar Björnsson is Professor of Philosophy at the Department of

facts, these facts cannot, on Lord’s view, contribute to the determination

Simon Blackburn’s seminal 1971 paper on moral supervenience2 put the

When someone tells me that what I am proposing to do is immoral,

1.1 COGNITIVISM AND NON-COGNITIVISM

The second makes a positive claim:

1.2 BELIEF, MORAL JUDGEMENT, AND

Cognitivism is an appealing theory of moral judgement. Like beliefs,

most responsible way of forming moral beliefs: more responsible than using

1.3 BROAD AND NARROW BELIEF

A broad belief is a representational state, a propositional attitude. It has a

Moral judgements that should not respond to pure moral testimony do

1.4 MORAL UNDERSTANDING

By moral understanding, I mean understanding why p where p is some

1.6.1 Is a Ulief a Kind of Cognitive State At All?

1.6.2 Why Distinguish Uliefs and N-Beliefs? Why Not

1.6.3 What Is the Relationship between Moral Uliefs

1.6.3.1 The Milgram Experiments

be. It is hard—much harder than we predict—to form a clear ulief that

1.7 COGNITIVISM REDEFINED

Cognitivists about moral judgement have typically claimed that moral

And correspondingly, non-cognitivism should be defined either as the

Secondly, I raised a question earlier about whether the ‘ideal’ of form-

Kant, I. 1991. Metaphysic of Morals, trans. M. Gregor. Cambridge: Cambridge

Some moral theorists—we’ll call them objectivists about obligation—hold

2.2 DELIBERATION, OBLIGATION,

question: Which obligation ought I satisfy? The ‘ought’ in this question is

2.3 ABILITIES, ACTING FOR THE RIGHT REASONS,

It is close to uncontroversial that our abilities can constrain our obligations.

to be within one’s perspective. Thus, objectivism is false. Here’s the argu-

2.3.1 In Defense of the Right Reasons Ability Condition

This is plausible, in part, because in Bad Husband it is an accident that

Credit is a hypological notion. Thus, since the deontic doesn’t necessarily

reasons. The problem is that intelligibility comes cheap. It is because of this

2.4 NEW INFORMATION PROBLEMS

My argument for perspectivalism notwithstanding, there are some powerful

2.4.1 Two New Information Problems

2.4.1.1 The Past Obligations Problem

2.4.1.2 The Advice Problem

she ought to do with more information. It is especially plausible that when

2.4.3 Abilities, Obligation, and ‘Ought’

2.4.3.1 New Information as a Problem for the Metaphysics

sometimes deliberatively obligated to withhold belief. To take the easiest

2.4.3.2 New Information and the Semantics of ‘Ought’

relativized to the best information available. What our deliberative obliga-

Cf. Bjornsson and Finlay (2010).

3.1 MORAL UNCERTAINTY

moral uncertainty (and specific moral credences) are crucially relevant to

mistake: the mistake of thinking that Uncertaintism is true. Once we see

My third observation is that the Uncertaintist is committed to the fol-

3.2 THE ALTERNATIVE TO UNCERTAINTISM

In this section, I will discuss the implications of my argument.

wouldn’t be a minor wrong to Sue, so it’s not an elaboration of the initial

3.3 DO FALSE MORAL VIEWS EXCULPATE?

In this section, I will defend premise 2 of my main argument.

My main argument in this chapter holds that Uncertaintism is committed

poisoning.) Anne’s blamelessness is fully explained by the first principle,

they have violated procedural moral obligations regarding learning the

3.4 IS THERE REALLY DISAGREEMENT BETWEEN

Uncertaintism is a view about how a person should behave; consider

Mary’s case also provides a counterexample to the following more general