OXFORD STUDIES IN METAETHICS

Oxford Studies
in Metaethics
Volume 9

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 2014
The moral rights of the authors have been asserted
First Edition published in 2014
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: 2006234688
ISBN 978–0–19–870929–9
ISBN 978–0–19–870930–5 (pbk.)
As 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.

00_oxfordhb-9780198709299.indd iv 2/22/2014 3:40:32 PM

 Contents

List of Contributors vi
Introduction vii

1 From Thought to Action 1

Jonathan Dancy
2 Cognitivism about Practical Rationality 18
John Brunero
3 Vindicating Practical Norms: Metasemantic Strategies 45
Hille Paakkunainen
4 Debunking Evolutionary Debunking 76
Katia Vavova
5 No Coincidence? 102
Matthew Bedke
6 Moral Skepticism and Moral Disagreement in Nietzsche 126
Brian Leiter
7 Moral Vagueness: A Dilemma for Non-Naturalism 152
Cristian Constantinescu
8 Relax? Don’t Do It! Why Moral Realism Won’t Come Cheap 186
Sarah McGrath
9 Wrong Kinds of Reason and the Opacity of Normative Force 215
Justin D’Arms and Daniel Jacobson

Index 245
 List of Contributors

Matthew Bedke is Associate Professor of Philosophy, University of British

Columbia
John Brunero  is Associate Professor of Philosophy, University of
Missouri-St. Louis
Cristian Constantinescu is Lecturer, Birkbeck College, University
of London
Jonathan Dancy is Professor of Philosophy, University of Texas, and
Research Professor of Philosophy, University of Reading
Justin D’Arms is Professor of Philosophy, the Ohio State University
Daniel Jacobson is Professor of Philosophy, University of Michigan
Brian Leiter is Karl N. Llewellyn Professor of Jurisprudence and Director
of the Center for Law, Philosophy & Human Values, University of
Chicago
Sarah McGrath is Assistant Professor of Philosophy, Princeton University
Hille Paakkunainen is Assistant Professor of Philosophy, Syracuse
University
Katia Vavova is Assistant Professor of Philosophy, Mount Holyoke
College
 Introduction
Russ Shafer-Landau

The current volume of Oxford Studies in Metaethics amply illustrates the

breadth of work being pursued in our field today. Jonathan Dancy starts
things off with the first in a trio of chapters here on practical reasoning and
rationality. Dancy initially poses a long-standing question: Can action be
the conclusion of practical reasoning? He eventually offers an affirmative
answer, but only after modifying the question so as to shed the original of
some contentious assumptions. One of the virtues of his view, as Dancy
sees it, is that it enables him to offer an account of theoretical reasoning
that retains a perfect structural similarity to the account of practical reason-
ing on offer. Dancy devotes the remainder of his chapter to replying to two
worries: first, that only beliefs, and never actions, can be the conclusion of
reasoning, and second, that practical reasoning can generate at most only
intentions to act, rather than actions themselves.
John Brunero then offers a detailed assessment of cognitivism about
practical rationality. Such cognitivism states that rational requirements
governing intentions can be explained by rational requirements governing
beliefs. Brunero’s focus is primarily on the prospects for cognitivism about
one such rational requirement—Means-Ends Coherence, which requires that
one intend the means one believes to be necessary for achieving one’s ends.
Perhaps the natural cognitivist thought here is what Brunero calls the Strong
Belief Thesis: the claim that intending to do something involves believing
that one will do it. But Brunero argues that such a claim is either false or is
unable to do the explanatory work the cognitivist needs it to do. He then
considers versions of cognitivism that abandon the Strong Belief Thesis, but
argues that the most plausible of these still suffer from fatal flaws.
Next up:  Hille Paakkunainen’s sophisticated effort to link facts about
concept-possession with rational requirements and permissions. Paul
Boghossian and Christopher Peacocke have argued that if it is a condition
of possessing a concept C that one must be disposed to update one’s beliefs
in accord with a norm N, then beliefs updated in accord with N are thereby
epistemically warranted. Peacocke calls such strategies of vindicating norms
“metasemantic.” Paakkunainen offers a new twist on this strategy, by con-
sidering whether it can be deployed in support of practical norms—in this
viii Russ Shafer-Landau

case, those that regulate updating intentions, rather than beliefs. She is cau-
tiously optimistic about this new route. Focusing on the concept of ought
to Φ, she argues that we can validly get from premises about the condi-
tions of possessing this concept to the conclusion that updating intentions
en­kratically is rationally permissible. She then shows how the argument can
generalize to apply to other normative concepts, and to be extended so as to
generate rational requirements as well as permissions.
In just the past six or seven years, a minor cottage industry has grown up
around the question of whether the evolutionary origins of our moral senti-
ments and judgment-forming faculties have skeptical implications. Most
authors agree that the problem is most acute for the moral realist. In a
happy pairing of chapters, Katia Vavova first seeks to debunk the debunkers,
while Matthew Bedke seeks to formulate and defend the sharpest debunk-
ing argument designed to make trouble for the realist.
Vavova presents two central anti-realist debunking arguments and claims
that each one fails, though for a different reason. The first of these claims
that we cannot rationally maintain our evaluative beliefs if we have no good
independent reason for thinking them true. Evolutionary considerations are
meant to show that we indeed lack any such reason. But proceeding in this
way, argues Vavova, will lead to a global skepticism that is both implausible
in its own right, and contrary to the aims of the anti-realist debunkers, who
sought to identify a special problem for evaluative beliefs. Alternatively,
debunkers might rely on the claim that if we have good reason to suspect
error on the part of our beliefs, then we cannot rationally maintain them,
and then proceed to argue that evolutionary considerations do provide such
reason when it comes to our evaluative beliefs, though not to (most of ) our
non-evaluative ones. Vavova here accepts the general epistemic principle,
but identifies problems when it comes to applying it as debunkers would
like. She considers debunking efforts that aim at all of our evaluative beliefs,
just our moral beliefs, and, finally, only our deontological ones. She finds
distinct flaws with each of these debunking efforts.
Bedke pursues the thought that, if moral realism were true, then the
truth of our moral beliefs would be so coincidental as to undermine any
presumptive justification they enjoy. He takes it as a striking fact about the
ongoing discussion of the topic that both realists and anti-realists agree on
the central premises of the best debunking argument, yet disagree about
their epistemic implications. Much of the value of Bedke’s discussion lies in
its careful attention to making precise the nature of the putative defeater,
which he identifies as a kind of insensitivity that he calls “the obliviousness
of normative belief to non-natural fact.”
Our next two authors also do their best to make trouble for the moral
realist. Brian Leiter enlists some ideas and passages from Nietzsche’s work to
 Introduction ix

develop a new anti-realist argument from moral disagreement. Such argu-

ments have long been the stock in trade of those who harbor doubts about
the objectivity of morality. Standard versions rely on the incompatibility of
norms across cultures, or the diversity of intuitive responses to concrete cases
within cultures. As Leiter reads him, Nietzsche’s argument from disagreement
begins with a point about philosophical failure—namely, that no rational con-
sensus has been secured on any substantive, foundational proposition about
morality. On his reading of Nietzsche, the best explanation of this lack of
consensus is moral skepticism. If that is so, however, then how to explain the
many thousands of lifetimes devoted to trying to develop an adequate moral
theory? The answer is as one would expect from Nietzsche: philosophers can
still construct dialectical justifications for moral claims, because the prem-
ises of different justifications will answer to the psychological needs of at
least some philosophers and thus be deemed true by some of them. Such an
explanation of philosophical activity is of course perfectly compatible with
the absence of any moral reality that might be accurately represented by
our moral theories. Does this kind of debunking argument overgeneralize
to other areas in which there is persistent philosophical disagreement? Leiter
concludes by considering the matter and returning a negative reply.
Cristian Constantinescu then takes up a set of worries that have not
been much discussed in metaethics—those to do with metaphysical vague-
ness. There is of course a large literature in philosophy about the nature
and implications of vagueness. But metaethicists have had relatively lit-
tle to say about the matter. In his chapter, Constantinescu targets non-­
naturalist moral realists. After providing many considerations that support
the view that moral predicates are vague in the same ways that other famil-
iar, sorites-susceptible predicates (“red,” “tall,” “heap”) are, Constantinescu
argues that non-naturalists are faced with a dilemma. They can either
reject the appearances and argue that moral predicates (and the proper-
ties they allegedly designate) are perfectly sharp in determining their exten-
sion, or follow the appearances and allow that they are vague. There is
trouble either way. On the latter horn, vague properties seem to threaten
the realist’s commitment to the objectivity of moral standards, the super-
venience of moral properties, and the non-reductionist nature of those
properties. Alternatively, sharp properties are said to raise problems for the
non-naturalist’s commitment to supervenience and to moral rationalism.
Though Constantinescu does not claim to have shown that these difficulties
are insurmountable, he does argue that considerations of vagueness sub-
stantially limit the things non-naturalists can consistently say about moral
properties, facts, and reasons.
Sarah McGrath, while not advancing an anti-realist argument per se,
seeks to make trouble for moral realists of a certain stripe—namely, those
x Russ Shafer-Landau

she calls “relaxed realists.” Such moral realists—among whom are numbered
Ronald Dworkin, Thomas Nagel, Derek Parfit, and T. M. Scanlon—accept
a non-naturalistic interpretation of morality, but argue that the standard
metaphysical and epistemological costs associated with non-naturalist real-
ism can be easily jettisoned, so long as we adopt a “non-metaphysical”
understanding of moral reality. In defending such a view, much is often
made of certain parallels between mathematics and morality. McGrath
lays these similarities out quite carefully and then proceeds to argue that
such parallels will not support the sort of relaxed realism that is lately com-
ing into fashion. She concludes her essay with an extended discussion of
Dworkin’s realism, arguing that it fails in ways that indicate the poverty of
the relaxed approach to moral realism generally. If moral realism is to be
vindicated, it will have to be along more “robust” lines that take on more
substantive metaphysical commitments.
Justin D’Arms and Daniel Jacobson wrap up this edition with a new
take on a problem that has occupied the attention of metaethicists for a
couple of decades now. The problem besets analyses that understand nor-
mative concepts or properties in terms of reasons for having attitudes. So,
for instance, one might think that something’s goodness can be understood
as, or in terms of, a reason to favor it, love it, be attracted to it, etc. The
problem—now known as the Wrong Kind of Reasons (WKR) problem—is
that there can be (typically pragmatic) reasons to have these attitudes, even
when those reasons fail to indicate the presence of the analysandum. If, in a
far-fetched scenario typical of those presented in the literature, we were told
that the only way to avert disaster is to love someone we know to be despic-
able, then there would be a reason (intuitively of the wrong sort) to develop
such affection even though there was nothing lovable, admirable, etc. about
the person. D’Arms and Jacobson draw our attention to a new set of WKRs
whose normative force is opaque. In such cases, it is quite clear that some
consideration bears on whether or not to feel (e.g.) shame, pity, or amuse-
ment, but unclear just how it does so—specifically, whether the considera-
tion helps to make the object of the attitude shameful, pitiable, or funny.
D’Arms and Jacobson argue that this phenomenon is much more wide-
spread than has been recognized. It occurs in many real-life examples, and
has hitherto unappreciated implications for many other areas of philosophy.
If they are right, then the discussion of WKRs has thus far been focused far
too narrowly. Following their lead will mean expanding the scope of such
discussions, as existing solutions to the WKR problem have failed to attend
to the opacity of normative force.
All of the chapters included in this volume are based on talks given at
the ninth annual Wisconsin Metaethics Workshop, held in Madison in
September 2012. I’d like to acknowledge the work of the event’s program
 Introduction xi

committee:  Matthew Chrisman, Jamie Dreier, Mike Ridge, Dave Sobel,

Mark Timmons, and Pekka Väyrynen. They did a great job of selecting a
slate of speakers from among an extremely competitive field. I’d also like to
express my gratitude to the two reviewers for OUP. I had hoped to be able
to thank them publicly, but they remain anonymous even to me. My edi-
tor, Peter Momtchiloff, worked his usual magic at enlisting reviewers will-
ing to write page after page of helpful commentary. Thank you, Peter—for
that, and for your constant support of this series. Thanks also to the OUP
production team, who, as always, fulfilled their commitment to the highest
standards of excellence.
 1
From Thought to Action
Jonathan Dancy

In this chapter I start by laying out an account of the nature of practical

reasoning, or deliberation. I then give an analogous account of theoretical
reasoning. I turn finally to consider, and I hope rebut, the various reasons
that have been advanced for thinking that no such account of deliberation
is possible. In these thoughts, I take for granted the notion of acting for a
reason. As I understand it, many cases of acting for a reason do not involve
practical reasoning at all. For there to be reasoning involved, there has to
be some more complex train of thought involved than we normally find
in such simple activities as answering the telephone, where one’s reason
is simply that someone is calling—though of course one does sometimes
deliberate before answering.

1.  Can an Action be the Conclusion of

Practical Reasoning?

If I had to answer this question I suppose that my answer would be yes. But
it seems to me that the question is better avoided. Those who ask it tend to
start from the idea that belief is the conclusion of theoretical reasoning, and
they then wonder whether reasoning that starts in the same sort of place,
from premises, could result in action in the same sort of way that it results
in belief. To this question the answer is probably no. But all that this shows
is that there is something wrong with the question, and indeed with the
picture of theoretical reasoning that tends to accompany it.
In fact we often see the distinction between theoretical and practical
reasoning defined in these terms: theoretical reasoning is reasoning whose
proper conclusion is belief; practical reasoning, if there were any such thing,
would be reasoning whose proper conclusion is action. But what does it
mean to say that belief is the conclusion of theoretical reasoning? Consider
the grid.
2 Jonathan Dancy

Belief p
Belief If p then q
Belief q

There is one sense in which the conclusion of the reasoning is q, which is

not a belief at all. The distinction between premises and conclusion is drawn
entirely on the right-hand side of the grid. And of course q is not a belief;
it is a thing believed, or to be believed. Can we say that the belief that q is
also (and in a different sense) the conclusion of reasoning? If we say this, we
are thinking of the case in which the reasoner is someone who believes that
p and that if p then q, and who comes by reasoning to believe that q. But of
course one can reason from things that one does not believe, and one can
draw a conclusion from such things, in which case one would probably not
believe the conclusion either. One could also reason from things that one
does believe to a conclusion which one sees to follow from them, without
yet accepting, that is, believing, that conclusion. One sees the force of the
inference, but one wants to mull things over for a while; perhaps one might
abandon one of the premises if the alternative is to accept this conclusion.
So has one then avoided concluding that q? One sees that the conclusion is
that q (and so in that sense one has drawn the conclusion), but one has not
yet concluded that q.
There is another difficulty derived from the fact that actions take time.
When, we might ask, is the supposed practical conclusion drawn? There
seem to be only three possible answers to this question: one draws the con-
clusion at the beginning of the action, as one goes along somehow, or only
at the end, when all is done. That none of these answers is straightforward
leads me to believe that there is something wrong with the question.
A further conceptual difficulty arises when we try to map action onto the
grid above, which is designed for inference from belief to belief. The third,
bottom line (the ‘conclusion’ line, as it were) requires a distinction between
the belief (the mental state, I suppose) and the thing believed, which is not
a mental state. However we understand this distinction (the locus classicus
for it is Alan White’s seminal 1972 article), it does seem that there is some
such distinction to be drawn. But if we were to map action onto the same
grid, we would need an analogous distinction between the action, or doing,
on the left, and what is done on the right. And it is not at all clear to me
that this distinction is in good health. An immediate worry is that the thing
done seems to inherit too many of the properties of the doing of it. With
belief, we can find lots of properties that the thing believed can have but the
 From Thought to Action 3

believing of it can’t, and vice versa. For instance, the believing occurs in time
but the thing believed probably doesn’t. The believing is done by someone
but the thing believed is not—and so on. It is not at all so easy to find such
differences between the doing and the thing done. If the doing takes time,
the thing done does too. If the doing is done by someone, so is the thing
done (I say more on this in Dancy 2008).
So there are all these tangles, which seem to me to be worth avoiding if
we can—and indeed we can. When someone comes to believe that q as a
consequence of reasoning, their so believing stands in some relation to the
considerations adduced in the reasoning that preceded it, and to which it is
a response. We do not need to specify that relation in order to ask the ques-
tion whether an action can stand in just that relation to the considerations
adduced in the reasoning that preceded it, that led to it, and to which it is
a response. This question seems to me to be in good health, and it avoids
entirely all the difficulties we had in making sense of the question whether
a belief can stand as the conclusion of theoretical reasoning—difficulties
which are only mirrored in the question whether an action can stand as the
conclusion of practical reasoning. These latter questions should be avoided
rather than answered in their present form. Or if we cannot avoid them, we
retain the option of saying that action cannot be the conclusion of practical
reasoning, but that belief cannot be the conclusion of theoretical reasoning
either, so that little is lost, and the possibility of practical reasoning not
impugned by any contrast with theoretical reasoning.

2.  The Better Question

So now we have our better question:  can an action stand to practical

reasoning in the same relation that a belief can stand to theoretical rea-
soning? And this question can itself be restated in slightly amplified form
thus: can an action stand to the considerations rehearsed in the delibera-
tion that leads to it, and to which it is a response, in the same relation
as that in which a belief can stand to the considerations rehearsed in the
reasoning that leads to it, and to which it is a response? The considera-
tions here are to be understood as featuring on the right-hand side of the
grid above; they are the premises, not beliefs but things believed, or at
least things that can be believed. This restated question does not assume
that all theoretical reasoning leads to belief. (It might lead to suspicion, or
doubt.) And to this restated question, my answer is yes. Action and belief
are on a par, as far as this is concerned.
In fact I think that this answer is easy. The notion of a conclusion with
which we started, but which we left aside in our new formulation of the
4 Jonathan Dancy

question, is a normative notion. It is the notion of a proper conclusion,

not just of whatever comes last in some process. Many things conclude, or
close off, deliberation without offering to be the conclusion of that delibera-
tion: death and boredom are two such things. So the relevant notion of a
conclusion is itself normative, and that offers us yet a further way of asking
our question, which now becomes: is there a suitable normative relation in
which both belief and action can stand to the considerations rehearsed in
the reasoning that leads to them, and to which they are a response? Now it
seems to me that there is an answer to this question that involves no strain
or difficulty. It involves an appeal to a normative relation with which we are
(I would claim) all very familiar, even though we may not be able to provide
any analysis of it. This is the relation of favouring. A  reason is a consid-
eration that favours some response. So when someone deliberates well and
then acts accordingly, the action done is the one favoured by the considera-
tions rehearsed in the deliberation, taken as a whole. It is a response to those
considerations as together calling for or favouring it. And this is perfectly
analogous to theoretical reasoning, when someone forms a belief as the
belief most favoured by the considerations adduced as premises. The belief
here is not the thing believed but the believing of it. The thing believed is
not a response; it is the believing of it that is a response, and can be favoured
or disfavoured by the considerations adduced.
This bald statement of the analogy obviously needs a lot of supporting
detail, some of which I will now attempt to provide.
The crux of the matter is the favouring relation. This relation obtains
between states of affairs and responses. Those responses can be beliefs,
actions, emotions, suspicions, doubts, intentions—anything that there can
be a reason for. And the things that do the favouring are features of the
situation (in the practical case, where there is a situation to respond to), or
states of affairs. (It may be that favourers are features under a description.)
They are such things as that it is a fine sunny day after a lot of rain, which
favours getting out and about. These features of the situation, or states of
affairs, are not propositions. No proposition can favour anything. Even if
there are propositions, and one of them is that it is a fine sunny day after a
lot of rain, what favours getting out and about is not that proposition, but
the state of affairs that the proposition represents. Propositions cannot call
for or favour responses. So the considerations that I referred to just above,
the ones adduced, are not propositions; they are states of affairs, or aspects
of the situation, that the deliberator considers, or brings to bear, in deciding
what to do.
This, at least, is how it is for practical reasoning, or deliberation. The
deliberator looks for that response or course of action that is most favoured
by the situation taken in its various aspects, and does that action in that light.
 From Thought to Action 5

Now in putting matters in this way, I have ignored two facts. The first of
these is that what is favoured by the considerations adduced is not a particu-
lar action, but acting in a certain way. Reasoning and deliberation can do
no more than offer a sort of blueprint that informs the way in which we act
when we act in its light. So it is not really true that the deliberator looks for
‘that response or course of action which is most favoured by the situation’;
it would be better to say that the deliberator looks for a way of responding
that best fits the nature of the situation. And when he does respond, his
response will be of the sort favoured by the relevant considerations, though
of course it will be of many other sorts as well, about which the relevant
reasons have nothing direct to say (e.g. which foot to start off from). I will,
however, sometimes ignore this delicacy in what follows.
Some people think that the mere fact that practical reasoning takes us,
not to particular acts, but only to types of action, is already sufficient to
reveal a rift between practical and theoretical reasoning. They suppose that
theoretical reasoning takes us to particular beliefs, not to beliefs of some
type, and in this they are presuming that belief is, as it were, particularized
by its content. The content is not usefully thought of as a type, since there
is no relevant distinction between different contents of that type. But to
allow the particularity of the content of belief does nothing to show that
what is favoured by the relevant considerations in theoretical reasoning is
itself particular in any sense that distinguishes the practical from the theo-
retical. What is favoured in theoretical reasoning is not the thing believed,
nor the proposition that is (sometimes thought of as) the content of that
believing, but the believing of that thing believed. And since such believings
are themselves particulars, different believings of the same thing believed
can differ in all sorts of ways that are not relevant to the question whether
this or that believing is of the sort favoured by the considerations adduced.
Believing is something that can start at a certain time, and can be more or
less committed or tentative, but the considerations adduced in the reason-
ing say nothing about when or how enthusiastically one should accept the
relevant ‘thing to believe’. It seems to me, therefore, that at the end of the
day both practical and theoretical deliberation take us to types rather than
to tokens, and no relevant distinction emerges from considerations to do
with the distinction between type and token.
The second thing I ignored is the fact that people often deliberate from
misconceptions. Things are not always as they suppose. But it does not fol-
low from this that what they are deliberating from are propositions rather
than states of affairs. They take the considerations that they bring to bear
to be the case, to be states of affairs, and deliberate accordingly. If they are
wrong about this, what they are deliberating from are supposed states of
affairs, not real ones. But a supposed but non-actual state of affairs is not
6 Jonathan Dancy

the same thing as a false proposition. Of course a merely supposed state of

affairs cannot favour anything, any more than a proposition can. But for
misinformed deliberators it is like this: there is this state of affairs, and it
calls for this response. What they deliberate from is not the case, but they
may still be correctly tracking what that situation would have favoured had
it been the case, and deciding to act accordingly. In such a case, they will
have reasoned well, but since they started in the wrong place they may well
end up doing the wrong thing, or a wrong thing anyway.
There is a second standard mistake that it is easy to make on this point.
When people deliberate from a misconception, they cannot be deliberating
from a state of affairs, from something that is the case. And we might be
tempted to infer from this that when they are lucky enough to be dealing
with something that is the case, they cannot be responding to that, but
must (as they are in the unsuccessful case) be responding to something that
falls short of that—and a proposition is a tempting candidate for this role.
This inference should be firmly resisted (I spend a long time on this in
Dancy 1995).
How then are we to characterize theoretical reasoning? There is con-
siderable temptation to say that in the theoretical case we are reasoning
from proposition to proposition. But this temptation, which would intro-
duce an enormous gap between theoretical and practical reasoning, must
be resisted. Take the case where the reasoner believes the premises of his
reasoning, and things are as he supposes. Let us take the grid with which
I started as representing such a case. We need here to track two relations.
The first relation goes down the right-hand side of the grid, and this one
is a relation between propositions (or at least there is nothing wrong with
casting it as such for present purposes). The proposition that q must be
true if the proposition that p and the proposition that if p then q are both
true. But the conclusion, q, is not favoured by those propositions. It is
itself a proposition, and a proposition is not a response, and cannot be
favoured by anything. But though q is not favoured by [p and if p then
q], believing that q is favoured by something. This is where the second
relation, that of favouring, comes in. What is it that favours believing
that q? Believing that q is favoured, not by the proposition that p and
the proposition that if p then q, nor by one’s believing that p and believ-
ing that if p then q, but by its being the case that p and its being the case
that if p then q. These two things, taken together, constitute a reason to
believe that q—a conclusive reason, indeed. It is these matters of fact, or
rather this complex matter of fact, that favours believing that q. And if it
is not the case that p, or not the case that if p then q, then maybe nothing
favours believing that q (though the reasoner may suppose otherwise, and
believe accordingly).
 From Thought to Action 7

In this way it emerges that action and belief are on a par as far as the
force of reasoning is concerned. That force (if it is worth calling it a force at
all, something that could be disputed) is the force of the favouring relation.
Practical reasoning has the same sort of force as does theoretical reasoning.
There is no such thing as a sort of logical force that we can find on one side
and not on the other.
There is more to be said here. For there is an explanatory connection
between our two relations, the relation between propositions and the rela-
tion between considerations and response. It is the former relation that
explains the latter relation in the toy example we have been using. The
explanation of the fact that the complex state of affairs that p and that if p
then q favours believing that q is given by the relation between the three
propositions, that if the first two are true the third must be true as well.
This fact explains why believing that q is the (or a) proper response to the
complex state of affairs represented by the premises. And similar explana-
tions will be available for all cases of formally valid deductive reasoning.
There will be logical necessity in these cases, but there is no such thing as
logical force. The only force available is the normative force of the favour-
ing relation. (What we are to say about non-deductive theoretical reasoning
remains to be seen.)
No such explanation is available in the practical case. If that it is a fine
sunny day after a lot of rain favours getting out and about, the explanation
of this is not to be given by finding some relation between propositions.
(How then is it to be given? This is a good question, but it is one that
I am not going to try to answer here.) The important point is that, despite
the differences in the ways the various favourings are explained, what is
explained is the same sort of thing both times. The belief and the action are
both favoured by the considerations rehearsed in the reasoning, and the rea-
soner comes so to believe, or so to act, in response to those considerations,
taking them to be matters of fact. In this respect the analogy is perfect.
There are no doubt all sorts of differences between belief and action, but
those differences don’t seem to matter for our purposes here.
We can bring all these things together by thinking about moral reason-
ing. Moral reasoning is theoretical reasoning, reasoning to a moral belief,
the belief that some course of action is right, wrong, permissible, and so
on. There will be considerations that favour that belief. Take a case where
the belief is that one ought to give the money back immediately. The con-
siderations that favour so believing are, let us say, that the person who lent
us the money has an immediate need for it, and that we have the money
easily available. But these considerations also favour acting in a certain way,
namely giving the money back immediately. They favour both belief and
action. We might then ask what explains these favourings. The favouring
8 Jonathan Dancy

of the action is not explained by relations between propositions, as I have

already said. But in this non-deductive case, the favouring of the belief is
not explained by such relations either. It must be explained by something
else. And one candidate explanans is the fact that those same considera-
tions favour action—in fact they not only favour it, they make it morally
required. It is because they make it morally required that they also speak in
favour of believing it to be morally required. The practical relation is what
drives here, and the theoretical relation follows; but both are favouring rela-
tions. We will see further such examples of what we might call the primacy
of the practical in what follows.

3.  The Irrelevance of Various Differences

between Belief and Action

The complications that I have just adumbrated should not be allowed to

detract from the considerable naturalness of the broad picture that has
emerged. I urged above that various differences between belief and action
should not be taken to destabilize that picture. The remainder of this chap-
ter is an attempt to respond to those who suppose otherwise.
It is natural to think that there is more of a gap between thought and
action than there is between thought and thought, so that reasoning that
takes one from thought to action is somehow harder to conceptualize than
is reasoning that takes one from thought to thought. I heard Richard Moran
give a nice elaboration of this idea, one whose aim was to defuse it by break-
ing it up into various strands and defusing each one, strand by strand. He
suggested that what looks like one large gap is in fact four gaps. We can
call these the temporal gap, the metaphysical gap, the volitional gap, and
the worldly gap. The temporal gap is the fact that the reasoning may be
over long before the action gets done. The metaphysical gap is that belief is
mental or abstract while action is physical or concrete. The volitional gap is
that the reasoning may be fine but the action that gets done may not be the
one that the reasoning recommends. The worldly gap is that reasoning takes
place inside but action takes place outside.
These supposed gaps are not as worrying as they may seem. (What fol-
lows is my own response to them.)
The temporal gap, properly understood, applies as much to belief as to
action. I can, after all, see the force of an inference but think that I don’t
have to decide which way to jump (accept the conclusion or abandon a
premise) until tomorrow, or even that I would do better to put things off
because I am too tired this evening to think straight. (It would be wrong
 From Thought to Action 9

to try to close the temporal gap by saying that practical reasoning leads
immediately to the formation of intention, and only mediately to action;
but I deal with attempts to insert intention in this sort of way in a later
section.)
The metaphysical gap is overblown. Action is more than physical move-
ment (see the fourth gap below) and belief is not abstract rather than con-
crete; if I believe that Kate is currently in Paris, the object of my belief, what
I believe, is as particular as one could wish, and so is my believing of it.
What I believe is what is the case, and what is the case is not abstract rather
than concrete.
The volitional gap does indeed exist. The agent may decide not to do
the action favoured by the considerations adduced, even when he correctly
identifies that action. He may decide to do some other action instead. But
I would suppose that the same thing can apply to belief. If so, this gap does
not undermine the analogy. I return to this point below.
Finally, the worldly gap is overblown. Action can only be held to take
place ‘outside’ rather than ‘inside’ if we think of action as mere locomotion.
But action is not mere locomotion, nor even locomotion with a special sort
of cause. To act (I here assert, ignoring swathes of complications) is to cause
a change. When one moves, the sort of change that one causes is a motion
of one’s body, locomotion. The action is the causing of that locomotion,
not the locomotion caused, and the relevant sort of causing is not ‘outside’
rather than ‘inside’.
But even if we allow ourselves to be persuaded that these supposed gaps
are not obstacles, there remain rather different considerations which will
be much harder to dispose of (even if on occasion the supposed gaps will
reappear).

4. Practical Reasoning as Theoretical

Reasoning with a Practical Conclusion

Joseph Raz is of the view that nothing else than belief can stand in the
sort of relation to considerations adduced in reasoning that belief can stand
in (Raz 2011: ch. 7, esp. section 2). Practical reasoning, if it is genuinely
reasoning, can therefore be nothing more than theoretical reasoning with
a practical conclusion. By ‘practical conclusion’ here I mean that practical
reasoning is reasoning to such conclusions as that I have most reason, or
ought, now to V. These conclusions are practical, I suppose, but only in a
weak sense. The crucial point is that V-ing itself cannot stand in that sort of
relation to considerations adduced in reasoning.
10 Jonathan Dancy

Raz is influenced by some obvious considerations. It is possible to delib-

erate, to reach a conclusion, and to fail to implement that conclusion—to
fail to act accordingly or even to form the intention so to act. If one gets as
far as the conclusion that one ought to V, and then fails to V, or to come
to intend to V, this is no doubt some sort of failure or defect. But here we
are talking about insincerity, weakness of purpose, an inability to stick to
one’s guns—that sort of thing. These are failures but they are not failures of
rationality, or defects in our reasoning. Reasoning can take us only to the
recognition of some course of action as called for by the considerations we
adduce—that is, to coming to believe that this is what we ought to do.
These things are true enough, but they are not the whole truth. The same
applies to belief itself, as I have already suggested. It is possible to follow the
reasoning where it goes, to draw the appropriate inference, to see what fol-
lows from what, and not yet to accept that things really are the way it seems
that they must be. Reasoning can take us to the appropriate thing to believe,
but it cannot itself take us to believing it. Believing requires more than
reasoning ability. The mother who cannot accept her son’s guilt, despite
recognizing the strength of the evidence, is just one well-known example of
this. So we should draw the same conclusion about belief, that reasoning
itself cannot take us all the way to belief. More is required, and the person
who refuses yet to accept that which he can perfectly well recognize as fol-
lowing from things he already accepts is not defective in reasoning ability;
the defect lies elsewhere.
We could then say that the unwillingness to accept what one sees that
one has conclusive reason to accept is a form of irrationality, one that lies
beyond any ability or inability to reason appropriately. But then the same
could be said about the unwillingness to do what one sees that one had
conclusive reason to do. That too has often been called a form of irrational-
ity—practical irrationality—but it takes us beyond any defect in reasoning
ability.
On this account it looks as if the answer to our main question would be
yes, but in a rather unexpected and disappointing sort of way. The relation
in which action stands to the considerations adduced in reasoning can be
just the same relation that belief can stand in, but neither action nor belief
would be thought of as the conclusion of reasoning, or even as intrinsically
connected to the reasoning that precedes it, because the reasoning process
is exhausted when one has merely seen the connection between ‘premise’
and ‘conclusion’.
There is a second issue about Raz’s picture, which is that it seems to
over-intellectualize the process of practical reasoning. Raz supposes that,
whatever happens thereafter, all practical reasoning passes through a theo-
retical conclusion. First we decide what we have most reason to do, and
 From Thought to Action 11

then we decide whether to do it. But this picture is not obligatory, and it
seems a rather heavy-handed picture of deliberation, which is after all a
pretty commonplace affair. My own view is that I can adduce considera-
tions, deliberate, and act accordingly without needing to form an interme-
diate conclusion that this or that course of action is the one I have most
reason to pursue. The notion of a reason need not appear explicitly in my
thought, because to respond to something as a reason is not, and does not
require, believing it to be a reason. (Just as to respond to someone as an
authority does not require believing her to be an authority.) That belief is an
extra stage, which need not occur.
There are two ways to move from the considerations adduced, because
those considerations favour two things at once. They favour acting in this
way rather than that, and they favour believing that this way of acting
is the one for which we have most reason. Nothing in this tells us that
the only rationally appropriate passage is to the action favoured via the
belief favoured; nor are we told that the passage to the belief favoured
should pass through another belief to the effect that this belief is indeed
favoured. (Regress lies this way—but one might wonder what stops the
regress-generating move.) So Raz’s picture, which has reasoning as passing
always through (and for him, stopping at) the theoretical conclusion, is less
than obligatory.
Raz could perhaps retreat at this point and claim only that reasoning to
action is tighter if it passes through belief. That is to say that someone who
selects a course of action in the light of the considerations he adduces in
deliberation would be proceeding more rationally if he were first to recog-
nize that this course of action is the one most favoured, and then to select it
for implementation. But even if one were to allow this, which I would not,
it would do little to tell us that action cannot stand to the considerations
adduced in the sort of relation that belief can.
Returning then to the simple idea that there are two things favoured
by the considerations adduced, the relevant action and the belief that that
action is most favoured by those considerations, let us ask about the proper
relation between these. Both of them are favoured by the relevant consid-
erations. So those considerations favour acting and they favour believing
that they are reasons to act. Which of these, if either, is the primary rela-
tion? I suggest that, if anything, the theoretical relation is secondary; it is
itself explained by the practical relation. That is, these things are reasons
to believe that one ought to act in such and such a way because they are
reasons to act in that way. We distort the focus of our reasons if we suppose
that the considerations adduced are primarily reasons to believe that they are
reasons to act, and only secondarily reasons to do the actions. Going down
the theoretical path, though not inappropriate, is not somehow required by
12 Jonathan Dancy

the nature of the reasons to which we are responding. Nor should we say
that in going down the theoretical path we are really only making explicit
what is implicit in the direct move from reasons to action. I am not a master
of the distinction between explicit and implicit, but I would hope that not
everything to which we are somehow rationally committed (as in respond-
ing to certain considerations as favouring this course of action we may be
said to be committed to their having the status of reasons) is something that
we are already doing, even if only implicitly.
So far then I see nothing that stands in the way of supposing that action
and belief stand in much the same relations to the considerations adduced
in reasoning. But Raz is also influenced by another phenomenon, one that
has much exercised him in the past, which is the fact that practical reason-
ing can throw up several courses of action as equally favoured by the rel-
evant considerations. We can, that is, conclude that we have equally good
reason to do any of three things, and at that point reasoning has given out,
leaving us with a simple choice. The course of action we select is certainly
favoured by the reasoning, but so were the other options.
Michael Bratman suggested to me that Raz’s point is far more damaging
than Raz himself recognizes. Raz supposes that reasoning does often enough
serve up one option as the one most favoured by the relevant considerations;
it is merely that on occasions it does not succeed in doing that. Bratman
suggests that, since the particular act to be done is always underspecified
by the reasoning, and we are given only a blueprint to which any eventual
action must conform, all the actions that would so conform are equally
recommended by our reasoning. So the phenomenon of underdetermina-
tion is ubiquitous. And if that is so, Raz has no need for a further argument
to the effect that what we are to say about the special cases of underdeter-
mination should apply also to cases of determination. And it would be no
response to him to argue that any such further argument is going to be
defective. One might, for instance, hope to limit the damage caused by
underdetermination by suggesting (as is often suggested in discussions of
the argument from illusion) that we should not allow our account of the
bad case to infect what we say about the good cases. But if Bratman is right,
no such response is available because there are no good cases.
But Bratman is not right. The sort of underdetermination that Raz is
thinking about is just a different phenomenon from the sort that impresses
Bratman. For Raz can allow that deliberation only produces a blueprint;
his point is that on occasion it produces two blueprints, when what we
were hoping for is just one. The fact that any blueprint needs to be, as
one might put it, fleshed out as we move from plan to action is not of any
moment. The relation between blueprint and action done is indeed worth
calling underdetermination; but if it is, we should not also use that term
 From Thought to Action 13

for the case where there are two blueprints. For if there are two blueprints,
neither of them determines, even partly, what is to be done. The sort of
partial determination that is all that a blueprint can achieve is something
that no blueprint achieves if there is a second, different blueprint compet-
ing with the first. For where there is competition of this sort, nothing is yet
determined.
This returns us to underdetermination in Raz’s sense. The question is
what we can learn from this phenomenon. It is true that in this case reason-
ing does not succeed in taking us to the action done. It stops at something
that looks theoretical, the belief that there are multiple equally acceptable
courses open to us. But this is because the normal route, from considera-
tions to action, is stymied. There is not one course of action that is most
favoured by the considerations, but there is one belief that is most favoured.
So we believe that, and this is as far as reasoning can take us; all that then
remains for us to do is to select one of the options left in play. And that sort
of selection is not normally necessary, because in the normal case the selec-
tion is done for us by our reasoning, or perhaps I should say that we have
already done it in that reasoning. As far as I can see, there is nothing here
that should unsettle the thought that, where the considerations adduced
favour one course of action over all others, that action stands in the same
sort of relation to those considerations as does, or would, the belief that
things are so.

5.  Reasoning to an Intention

John Broome is of the opinion that we can reason to a belief and reason to
an intention, but we cannot reason to an action. The nearest that reasoning
can get us to action is by leading us to form an intention. Whether we then
go on to act accordingly is just another issue. So for Broome (2002) the
formation of an intention can stand in the same relation as does belief to the
considerations adduced in reasoning, but that is as far as reasoning can go.
One thing that is influencing Broome here is the thought that we can-
not be rationally required to do what we may fail to do through no rational
fault of our own. Action, therefore, cannot be rationally required because
although every reason speaks in favour of V-ing, I may fail to V because you
lock me up and prevent me. Broome presumes that this argument does not
also apply to the formation of intention. Suppose that reasoning requires
me at least to form the intention to V, but that you distract me so that
I don’t get round to doing that before it is time to knock off for the day,
or prevent me from doing it by playing very loud music so that the only
14 Jonathan Dancy

thing I can think about is getting away from here as soon as possible, and
by the time I have done that it is too late for the action anyway. There was
no rational fault in me, it seems, but still I  failed to form the intention
that I was rationally required to form. There is an assumption, then, that
somehow intention is not subject to the sort of interference that can pre-
vent action, and I don’t see that this should be so. The idea that intention is
something we can always achieve if we set ourselves to do so seems odd to
me; it is as if forming intentions is terribly easy, somehow.
Consider the moral analogy of the thought that is influencing Broome.
I can be morally required to do something tomorrow which I will fail to do
through no moral fault of my own. Perhaps I will die tonight in my sleep.
Perhaps I will try to do it but fail despite my best efforts. Does this sort of
thing show that I was really only morally required to try, not to succeed?
Even if we allow that I can only be rationally required to do things that
I cannot fail to do without rational fault, we can question whether delibera-
tion always throws up its conclusion as rationally required of us. Could not
deliberation recommend a course of action without requiring it? To some
extent our answer to this question will depend on whether we think that
we are under a general requirement to do whatever we have most reason to
do. My own view about this is that it is too heavy-handed. It is especially
heavy-handed if the requirement concerned is a rational requirement. Why
should the failure to do what one accepts one has most reason to do always
be a rational failure (that is, one that is a sign of partial irrationality) rather
than some other sort of failure? Someone who fails to do what they accept
they ought to do may be morally weak without being rationally weak. And
what if I knew, really, that my velvet jacket doesn’t go with these trousers,
and that I should be wearing something else, but just thought I would try
something new for a change? This seems to be evidence of a sartorial lapse
rather than a rational one.
I think that the driving thought on the other side is that if we are deal-
ing with deliberation, and deliberation is just practical reasoning, failure to
carry through on our deliberation has just got to be rational failure, because
reasoning is a rational practice. My response to this is that moral reasoning is
expressive of our rationality, but it is also expressive of our moral character,
and a failure of moral reasoning (special pleading, perhaps) is often much
more a moral failing than a rational one. That is to say, if asked whether
someone caught out in special pleading has shown defects of rationality or
of morality, I would prefer the latter if I had to choose.
I now return to the suggestion that forming an intention is something
we can always do if we decide to do it. This at least is entirely up to us, so
that if we fail to do it, we are liable to charges of a certain sort whatever the
cause of our failure. My own view about this, as I said above, is that it is
 From Thought to Action 15

exaggerated. What if I have done my reasoning, and just before I form the
relevant intention someone distracts me until it is too late, or something
turns up that is more important and needs my immediate attention? This,
it seems to me, is a case where I fail to form the relevant intention through
no rational fault of my own—indeed I am not at fault rationally at all. (The
same, of course, applies to belief.)
There is an idea that, even if you cannot bring yourself to do the action,
you can at least always bring yourself to form the intention to do it, the
weakness being left for later when you fail to carry through with that inten-
tion. And this might tempt some to say that rationality can only require
the formation of intention. There may be difficulties about actually doing
the thing intended, which we have to allow; so we restrict the demands of
rationality to that which those difficulties do not threaten. This train of
thought seems to me no stronger than the familiar thought that one can
only be morally required to do what one cannot fail to do through no moral
fault of one’s own. The standard conclusion from this is that we can only
be morally required to try, not to succeed. Success, after all, is not up to
us—at least not normally—but trying is. Myself, I doubt this last. Michael
Smith has reported to me that there are certain moments in brain surgery
(for which the patient has to be awake) in which patients report, not just
that they cannot move their hand (say), but that they have lost the ability
even to try to move it. But leaving that aside, there is another way of coping
with the phenomena here, which is to say that what is morally required of
one is not to try to do it but rather to do it if one can. This seems to me to
be much more realistic. It is true that the logical form of this requirement is
interesting, and not entirely clear. But it is certainly distinct from a require-
ment to do it which you are only under if you can do it; the conditionality
is within the scope of the requirement. And we cannot say that you are mor-
ally required to do everything that you cannot do. But whatever the require-
ment to do it if you can does in fact mean, it is distinct from a requirement
to try to do it, and it seems to me to be a much more promising tool for
understanding the relevant phenomena.
Finally, in this section, I turn to the implied contrast between intention
and action. Broome’s position implies this contrast because he supposes,
as we used to put it, that reasoning can take us to an intention but not to
an action. Now this distinction seems to me to derive from concentrating
on the notion of a prior intention—from thinking, that is, about inten-
tion that one forms some time before the action intended is to be done,
and which may not eventually be successfully implemented. That, rather
than the action intended, is the sort of thing that deliberation can require
of us, apparently. And I have to admit that, if this scenario is the right one
to press, it does seem hard to suppose that deliberation, having taken us to
16 Jonathan Dancy

the formation of the relevant intention, has not somehow been used up,
so that the eventual performance of the action lies beyond its reach. But
what if we think of the matter not in terms of prior intentions, but in terms
of intentional action, of the intention that is in the action when one acts
as one intends (rather than simply as one intended)? Take any case where
the action is to be done now. In such cases, it is much harder to separate
the intention from the action. The one seems to begin and end where the
other begins and ends. What sense would there be in these cases of insist-
ing that deliberation can only take us to the intentional side of the inten-
tional action? What on earth could that mean? And if one separates out the
intentional side of the intentional action, what is then left to be the action
that remains, the residue? Myself, I doubt that anything much remains at
all—or if it does, it will be mere bodily movement, or motion, which is to
say movednesses rather than movings. But this is not the sort of thing we
should be contrasting with intention; these are not actions at all, but mere
events. The action intended is not in this sense a mere bodily motion; it
is a full-scale action, shot through with purpose, and not metaphysically
distinct from the intending of it. So the idea that reasoning might take one
to intention and not to action seems to require, in this sort of case, a most
peculiar distinction between intention and action.
With these remarks I am working round to the suggestion that the idea
that reasoning can only take us to prior intention, not to action, is an unsat-
isfactory half-way house between Raz’s view that reasoning can only take
us to belief, and my view that it can take us to action. (We should bear in
mind all the while the official formulations of these points in terms of the
sorts of relations in which belief, intention, and action can stand to the
considerations rehearsed in the reasoning.) Where the sort of intention we
are dealing with is that involved in intentional action, in acting with an
intention, there seems to be no half-way house position available at all. In
such cases, we really have only the belief, on one side, and the action on the
other to choose between.
And once we see this point for the case where action succeeds delib-
eration immediately, what room remains for us to say that, in the cases
where it doesn’t, deliberation can only take us to the intention and not
all the way out to the action? It is true that the prior intention can exist
in full glory though the poor action never gets to be done at all. But the
fact that when the deliberation is done the time for action is not yet come
should not be allowed to alter the focus of the deliberation, which is not
on intention-forming but on the action to be intended. (I am not talking
here about elaborate planning, but about the simpler case where I form
the intention to go shopping tomorrow after my visit to the dentist.) The
focus of deliberation is the action to be done, and the prior intention is
 From Thought to Action 17

little more than a staging post for cases where delay is inevitable for one
reason or another. This point is the same as one that we saw in discus-
sion of Raz’s views. There is an order of explanation here: the intention
is required because the action is. The reasons we consider in delibera-
tion favour the action, primarily, and favour forming the intention to do
that action only derivatively. They are reasons to intend to act because
they are reasons to act. This is the primacy of the practical again. And
we should not suppose that our reasons to intend to act are, as it were,
rational reasons, or that they are reasons that we ignore on pain of irra-
tionality, while the reasons to act are practical reasons (e.g. moral ones).
Things cannot be carved up in this sort of way. It is not as if rationality
calls on us to intend to act and morality (or something else) calls on us to
do the action intended. If we are required to intend to act, this is because
we are morally required to act (if we can), and the time for action has
not yet come.

6. Conclusion

What I  have tried to do in this short chapter is to present the outlines

of a theory of deliberation according to which practical reasoning is no
more peculiar than is theoretical reasoning. The supposed mystery about
Aristotle’s practical syllogism is no mystery at all; it only looks mysterious
because one takes it in the wrong way in the beginning. Having presented
the theory, I then argued against other views which suppose that genuinely
practical reasoning, that is, reasoning to action, is impossible.

References
Broome, J. 2002. ‘Practical Reasoning’, in J. Bermúdez and A. Millar
(eds), Reason and Nature:  Essays in the Theory of Rationality, 85–111.
Oxford: Clarendon Press.
Dancy, J. 1995. ‘Arguments from Illusion’, Philosophical Quarterly, 45(181)
(Oct.): 421–38.
Dancy, J. 2008. ‘Action in Moral Metaphysics’, in C. Sandis (ed.), New Essays on the
Explanation of Action, 398–417. Basingstoke: Palgrave Macmillan.
Raz, J. 2011. ‘Practical Reasoning’, in his From Normativity to Responsibility, ch. 7
(esp. sect. 2). Oxford: Oxford University Press.
White, A. R. 1972. ‘What We Believe’, in N. Rescher (ed.), Studies in the Philosophy
of Mind, 69–84. APQ monograph series, 6. Oxford: Blackwell.
 2
Cognitivism about Practical Rationality
John Brunero

Cognitivism about practical rationality is the view that some rational

requirements governing intentions can be explained by the rational require-
ments governing beliefs. Cognitivists tend to focus on two requirements of
practical rationality in particular: a consistency requirement on intentions,
and a requirement of means-ends coherence. Cognitivism about these prac-
tical requirements might seem promising. After all, we’re very comfortable
speaking of consistency and coherence requirements governing beliefs. And
many have thought that intentions involve beliefs in some way. And so it’s
natural to think that the rational requirements governing those involved
beliefs might explain the rational requirements governing those intentions.
To get a sense of how such an explanation might go, consider how a
cognitivist might explain a simplified version of a consistency requirement
governing intentions:
Intention Consistency:  Rationality requires that [if one intends to X,
then one does not intend not to X].1
The cognitivist could first point to a similar consistency requirement gov-
erning beliefs:
Belief Consistency: Rationality requires that [if one believes that P, then
one does not believe that ~P].
And she could then defend a thesis about the way intentions involve beliefs:
Strong Belief Thesis: Intending to X involves believing that one will X.
If the Strong Belief Thesis is true, whenever one intends to X and intends
not to X, one has inconsistent beliefs about what one will do: a belief that
one will X and a belief that one will not X.  This suggests that Intention

1
 These requirements are wide-scope requirements in that the scope of “requires”
ranges over a conditional. See Broome (1999).
 Cognitivism about Practical Rationality 19

Consistency can be explained by Belief Consistency and the Strong Belief

Thesis: it’s irrational to have inconsistent intentions because, in doing so,
you’ll have inconsistent beliefs about what you will do, which is irrational.
This is a simplified version of a consistency requirement on intentions
since rationality requires not just that our intentions be consistent with
one another, but also that they be consistent with our beliefs. For example,
I would be irrational if I intended to travel to New York this afternoon and
intended to travel to San Francisco this afternoon, while believing that if
I  travel to New  York this afternoon, I  won’t travel to San Francisco this
afternoon. But the cognitivist employing the Strong Belief Thesis will have
no trouble here since the associated beliefs—that I’ll travel to New York this
afternoon, that I’ll travel to San Francisco this afternoon, and that if I travel
to New York this afternoon, I won’t travel to San Francisco this afternoon—
are jointly inconsistent.
Cognitivist approaches to practical rationality have both their defend-
ers (including Gilbert Harman, David Velleman, Jay Wallace, and Kieran
Setiya) and their critics (most notably, Michael Bratman).2 In this chap-
ter, I’m going to follow the trend in the literature and focus in particular
on the prospects for cognitivism about means-ends coherence (also often
referred to as “instrumental rationality”). Means-ends coherence, roughly
speaking, requires that we intend the means we think are necessary for
achieving our ends. For instance, a requirement of means-ends coher-
ence prohibits me from intending to travel to New York this afternoon,
believing I’ll travel to New York this afternoon only if I intend to buy
an airplane ticket, but not intending to buy an airplane ticket.3 Someone
who finds herself with this prohibited combination of attitudes could
escape in three ways: coming to intend the means of buying the ticket,
giving up her end of traveling to New York this afternoon, or revising her
instrumental belief. So far as the requirement of means-ends coherence
is concerned, there is no specific way one must proceed. For this reason,

2
 See Harman (1976, 1986), Velleman (1989, 2007), Wallace (2006), and Setiya
(2007). See Bratman (1987, 2009a, 2009b).
3
  The instrumental belief refers to an intention thought to be necessary for achieving
an end. Note that Means-Ends Coherence should not be understood to apply to every
believed necessary condition for achieving an end. For instance, consider expected side
effects. Suppose I intend to grade fairly, but believe I’ll do so only if I hurt the feelings
of some students. Surely I’m not rationally required to intend to hurt their feelings (or
revise my other attitudes). But I am rationally required to form those intentions I think
necessary for my grading fairly. Suppose I think I’ll grade fairly only if I intend to blind
the papers. In that case, it does seem that I’m required to intend to blind the papers (or
revise my other attitudes).
20 John Brunero

the requirement is often formulated as a wide-scope requirement, in that

“requires” ranges over a conditional:
Means-Ends Coherence: Rationality requires that [if one intends to E,
and believes that one will E only if one intends to M, then one intends
to M].4
This formulation is no doubt in need of further refinement.5 But since the
details won’t matter here—except for a brief point I’ll make in §2.3—we
can work with this rough formulation.
I’ll start (§1) by considering the prospects for a cognitivist account of
Means-Ends Coherence that appeals to the Strong Belief Thesis. I’ll argue
we should reject such accounts since the Strong Belief Thesis is either false
or it’s restricted in such a way that it can’t do the explanatory work the cog-
nitivist needs it to do. I then (§2) consider the prospects for a cognitivist
account that doesn’t appeal to the Strong Belief Thesis, but instead to some
weaker thesis about the connection between intention and belief. I  start
with Wallace’s cognitivist account, present two objections to it, and then
present an account that’s similar but avoids these objections. I defend the
key assumptions of this account, and argue it’s the best available cognitivist
account of Means-Ends Coherence.
However, the best isn’t good enough. I  argue (§3) that, while cognitiv-
ists might be able to show that every instance of means-ends incoherence
involves incoherence in belief, they haven’t established that the theoretical
requirements explain the practical ones. I  argue that there are two reasons
to doubt cognitivism’s explanatory thesis. First, we should expect a theory
that explains Means-Ends Coherence to also be able to explain closely related
practical requirements, but cognitivism seems unable to do this. Second, if we
consider all the relevant theoretical requirements that apply, and not simply
select out some and ignore others, cognitivism issues false predictions about
the rationality of ways of escaping from a state of means-ends incoherence.

1.  Cognitivism with the Strong Belief Thesis

Some philosophers have developed cognitivist explanations of Mean-Ends

Coherence using the Strong Belief Thesis.6 Let’s consider how such an
explanation might go. Suppose I’m instrumentally incoherent: I intend to
travel to New York this afternoon, believe I’ll get there only if I intend to
4
  See Broome (1999).
5
  See Kolodny and Brunero (2013: §3), and Setiya (2007: 667–8).
6
  See Setiya (2007:  663–71). Setiya’s account differs slightly from the sketch given
here, in ways that won’t matter for the following discussion.
 Cognitivism about Practical Rationality 21

buy an airplane ticket, but don’t intend to do this. According to the Strong
Belief Thesis, I believe
(1) I will travel to New York this afternoon.
If we pair this belief with my instrumental belief
(2) I will travel to New York this afternoon only if I intend to buy an air-
plane ticket, and apply the theoretical rational requirement

Closure: Rationality requires that [if one believes that P, and that if P then Q,
then one believes that Q]
then rationality requires that our instrumentally incoherent agent either
give up his belief (1), which, by the Strong Belief Thesis, would involve his
not intending to travel to New York this afternoon, and so would involve his
escaping instrumental incoherence, or give up his belief (2), which would
also constitute an escape from instrumental incoherence, or come to believe
(3) I intend to buy an airplane ticket.
This gets us quite close to a cognitivist explanation of Means-Ends
Coherence, but not all the way there, since it seems possible for one to
believe one intends to do something without actually intending to do it.
But if one can close this gap by showing how it is independently theoreti-
cally irrational to have false beliefs about one’s intentions in this context,
then one would have a cognitivist explanation of Means-Ends Coherence
in terms of Closure and the theoretical requirement not to have false beliefs
about one’s intentions in this context.7 However, this cognitivist explana-
tion will only be as plausible as the Strong Belief Thesis. The remainder of
this section will be devoted to the question of whether that thesis is true.

1.1  The Strong Belief Thesis

According to the Strong Belief Thesis, intending to X involves believing one
will X. But this need not preclude one from also believing one’s belief could
be false. Just as I could believe that my friend Dana will be in New York this
afternoon, but also believe that I could be wrong about this, when I intend
to travel to New York this afternoon, and so believe that I will, I could also
believe that I could be wrong about this. The Strong Belief Thesis is com-
patible with one’s being a fallibilist about the involved beliefs.8

7
  For an attempt to close this gap, see Setiya (2007: 670–1). For arguments that the
gap remains open, see Bratman (2009a: §4), and Brunero (2009: §1).
8
  See Harman (1986: 92).
22 John Brunero

However, one might still worry that the Strong Belief Thesis is vulnerable
to counterexamples. Critics of the thesis usually present cases involving an
agent intending to do something difficult, where, it is alleged, the agent
intends to X, but doesn’t believe that she will X. Some examples include
a golfer intending to sink a difficult putt, an amateur basketball player
intending to make a half-court shot, and someone intending to leap across
a wide gap. Defenders of the Strong Belief Thesis aren’t persuaded by these
alleged counterexamples, and they usually reply by denying that these really
are cases in which one intends to sink the putt, make the shot, leap across
the gap, and so forth.9 Since the Strong Belief Thesis has been discussed
for many years now, with some very capable philosophers being convinced
by these alleged counterexamples, and other equally capable philosophers
not being convinced by them, we might get the impression that the cogni-
tivist wouldn’t face significant costs in proceeding on the assumption that
the Strong Belief Thesis is true.10 I think this impression is mistaken. I’ll
argue that those argumentative maneuvers needed to save the Strong Belief
Thesis from these counterexamples, if successful, undermine the cognitiv-
ist’s broader explanatory project. I’ll argue that the cognitivist employing
the Strong Belief Thesis thus faces a dilemma: she must either concede that
the Strong Belief Thesis is false or maintain that it’s true but unable to play
the explanatory role that the cognitivist needs it to play. Either way, cogni-
tivist explanations employing the Strong Belief Thesis fail. If this is right,
the prospects for cognitivism with the Strong Belief Thesis are not as good
as they might initially appear.
So, let’s now turn to some alleged counterexamples. First, let’s consider
the case of someone who intends to lift a heavy log that has fallen onto his

9
  One way to proceed would be to consider our linguistic intuitions about agential
expressions of an intention (or lack of an intention) in such cases. Suppose the golfer
makes the difficult putt. On the one hand, if she were to say, “I intended to make that,”
she comes across as bragging (Harman 1986: 91). However, this might merely be a fea-
ture of Gricean conversational pragmatics; notice that the implicature of overconfidence
can be canceled by, for instance, her saying, “I intended to make that, but really didn’t
think I  would.” And there’s also some linguistic evidence pointing in the other direc-
tion: it would sound odd for the golfer to say, “I had no intention of making it” (Adams
2007: 151–2). In this chapter, we can avoid putting much weight on such arguments.
Our question is more manageable: can the Strong Belief Thesis be true in a way that lends
support to the cognitivist explanation of Means-Ends Coherence? I think we can answer
this question in the negative without having to rely on linguistic intuitions of this sort.
10
  For instance, the Strong Belief Thesis is endorsed by Hampshire and Hart (1958),
Grice (1971), Harman (1976, 1986), Velleman (1989, 2007), and Davis (1984). But it
is opposed by Audi (1973), Davidson (2001), and Mele (1992); Bratman (1987) also
expresses skepticism about the thesis, and develops an account of intention that doesn’t
rely on it. For a very helpful overview of possible views of the intention–belief connec-
tion, see Adams (2007: esp. 143–7). Adams also opposes the Strong Belief Thesis.
 Cognitivism about Practical Rationality 23

front porch. Plausibly, he intends to lift the log but doesn’t believe that he
will. It’s not that he believes he won’t; he’s simply agnostic about whether
he will. Second, let’s consider Bratman’s famous case of an absent-minded
cyclist who intends to stop by the bookstore on the way home, but, aware
of his tendency to go on “autopilot” once on his bike, doesn’t believe that
he will. Again, he’s agnostic about whether he’ll succeed in doing what he
intends to do.11 (While the first example involves one having doubts about
whether one’s actions will produce the intended result, the second involves
one having doubts about whether one will perform the requisite actions at
the appropriate time.)
One way to defend the Strong Belief Thesis against such counterexam-
ples, suggested by Gilbert Harman, is to argue that the alleged counterex-
amples don’t involve intentions to do the act believed to be difficult, but
only intentions to try to do the act.12 For instance, on Harman’s suggestion,
the man intends to try to lift the heavy log, and the absent-minded cyclist
intends to try to stop by the bookstore. We need to be careful about what’s
being claimed here. It’s plausible to think that when a speaker says, “I intend
to try to X,” she’s expressing the thought that she intends to X, and the “try
to” is there to express her doubts about success. But this view, of course,
would be of no help to the Strong Belief Thesis, since this view concedes
that the Strong Belief Thesis is false. For Harman’s suggestion to save the
Strong Belief Thesis, it must be the case that in the alleged counterexamples
what one intends is to try to X—and, moreover, one has this intention to try
to X without also intending to X.
Harman’s move saves the Strong Belief Thesis, but it’s of little help to the
cognitivist attempting to explain Means-Ends Coherence. Let’s suppose we
follow Harman in thinking that the man really only intends to try to lift the
log. Intuitively, if this man were to believe that he’ll lift the log only if he
bends his knees when he lifts, and were to fail to intend to bend his knees
when he lifts, he would be criticizable as means-ends incoherent. But he
might think that bending one’s knees, while necessary for lifting the log,
isn’t necessary for trying to lift the log. After all, we could suppose that the
last time he didn’t bend his knees, he tried and failed to lift the log, but
didn’t fail to try to lift the log. So, if his intention is merely to try to lift the
log, he is no longer criticizable as means-ends incoherent in failing to intend
to bend his knees. He’s no longer failing to intend means believed necessary
for achieving his end. But, intuitively, this is the wrong thing to say; surely
he is criticizable for being means-ends incoherent here.

11
  Bratman (1987: 32).
12
  Harman (1986: 90–4).
24 John Brunero

The same point could be made about Bratman’s cyclist. He may believe
that intending, at the relevant time, to turn left on Bookstore Lane is a
necessary means to stopping by the bookstore, but not believe it’s a necessary
means to trying to stop by the bookstore. After all, he may know that the last
time he tried to stop by the bookstore, he failed to stop by precisely because
he didn’t intend, at the relevant time, to turn left on Bookstore Lane. But
he didn’t fail to try to stop by the bookstore.13 Since intending to turn down
Bookstore Lane isn’t believed by him to be a necessary means to carrying
out his intention to try to stop by the bookstore, he isn’t means-ends inco-
herent in failing to form that intention. But that’s the wrong result, since
surely he is means-ends incoherent in failing to form that intention.
The problem here is that there is, intuitively, a requirement of means-ends
coherence that applies to the cyclist and the log-lifter. If the Strong Belief
Thesis is false, the cognitivist can’t account for this requirement in the
standard way (as in the sample explanation of Means-Ends Coherence given
above). But if we save the Strong Belief Thesis by saying that these cases
involve only intentions to try, we still are unable to explain this require-
ment, since the agent, while believing that forming certain intentions (to
bend at the knees; to turn down Bookstore Lane) is necessary for achieving
an end, might not believe that forming these intentions is necessary for try-
ing to achieve those ends. And so what should count as a case of criticizable
means-ends incoherence isn’t so counted by the cognitivist.
Another way to save the Strong Belief Thesis from these counterex-
amples, originally suggested by Harman but developed by Velleman in
Practical Reflection, is to argue that “intend” is ambiguous, and the Strong
Belief Thesis holds for only one sense of “intend.” Velleman writes:
The words ‘intention’ and ‘intend’ are thus ambiguous. They are used to denote,
on the one hand, the agent’s attitude toward outcomes that are settled, from his
perspective, at the close of deliberation and, on the other hand, his attitude toward
outcomes whose pursuit is the topic of his deliberation but whose attainment is
not thereby settled. In other words, they are used to denote both plan-states and
goal-states of the agent.14
On this view, the log-lifter would have the goal of, but not plan on, lifting
the log, since the outcome isn’t “settled from his perspective.” But since the
Strong Belief Thesis is restricted to the sense of “intend” denoting “plan-
states,” such examples are not counterexamples to the thesis.

13
  In this case, he doesn’t go to the bookstore because of his own forgetfulness, rather
than because of any external impediment. But we don’t want to say that if the relevant
failure is “internal” then one doesn’t make an attempt. That would rule out the possibility
of one’s trying to, say, solve a math problem in one’s head, or remember an anniversary.
14
  Velleman (1989: 112). See also Harman (1986: 93–4).
 Cognitivism about Practical Rationality 25

But this is of little help to the cognitivist, since Means-Ends Coherence gov-
erns intentions in both senses—that is, in both the “plan-state” and “goal-state”
senses of “intend.” For instance, the log-lifter would be instrumentally inco-
herent in intending (in the “goal-state” sense) to lift the log, believing that he’ll
lift the log only if he intends to bend his knees, and not intending to bend
his knees. The cognitivist saves the Strong Belief Thesis, but at the cost of no
longer being able to explain why Means-Ends Coherence applies in every case
in which it does.

1.2  Arguments for the Strong Belief Thesis

So far, I’ve argued that we should reject the Strong Belief Thesis—or, at least,
reject the idea that the thesis can be both true and useful to a cognitivist expla-
nation of Means-Ends Coherence. In the remainder of this section, I’ll con-
sider and assess some arguments that have been given for the Strong Belief
Thesis. In particular, I’ll focus on some recent work by David Velleman, who
argues that we need to accept the Strong Belief Thesis if we are to explain
certain characteristic functions of intentions, and explain why intentions are
governed by rational requirements like Means-Ends Coherence and Intention
Consistency. I’ll consider four arguments, and explain why none of them are
persuasive.
First, Velleman reminds us of Elizabeth Anscombe’s view that the natural
expression of an intention to X is “I am going to X.” It seems that if one is
sincerely asserting this, then one believes that one will X. As Velleman observes,
the assertion, “I am going to X, but I  don’t believe that I  will” appears to
be an instance of Moore’s paradox.15 One might think this is some evidence
for the Strong Belief Thesis—namely, that without the Strong Belief Thesis,
Moore-paradoxical assertions would be licensed. But I don’t think this is right,
since Anscombe’s remark could be evidence for the Strong Belief Thesis only
if it is understood as a universal generalization—that is, as something like,
“For all intentions, the natural expression of the intention is ‘I am going to
X’ ”—and that generalization seems to be shown to be false by the very same
examples that challenge the Strong Belief Thesis. For instance, for the person
intending to make the difficult putt, “I am going to make it” doesn’t seem to
be the natural expression of his intention. Something expressing less confi-
dence—like “I am going to try to make it”—would better reflect his appre-
ciation of the difficulty involved in making the putt, and hence be the more
natural thing to say in this context. And no Moore-paradoxical results emerge

15
  Velleman (2007: 206–7).
26 John Brunero

if this less confident assertion is paired with one’s asserting, “But I don’t believe
that I will.”16
Let’s now turn to a second argument from Velleman—one central to
our concerns here—which holds that the Strong Belief Thesis is needed to
explain Means-Ends Coherence. Velleman writes:
Why, for example, should an agent be rationally obliged to arrange means of car-
rying out an intention, if he is agnostic about whether he will in fact carry it out?
Suppose that I  form an intention to fly to Chicago next Tuesday, well knowing
that I often forget to take trips that I have planned. (I am even more forgetful than
Bratman.) Buying a ticket for my flight to Chicago will turn out to have been a
waste of money if I  forget to take the trip. . . . But why should I  be categorically
required to invest in means whose benefits I am not yet prepared to believe in? If
I am still entertaining the possibility that a ticket will go to waste, why shouldn’t
I weigh its expected benefits against those of alternative investments?17
The idea here seems to be that if we think one could intend to fly to Chicago
without believing one will do so, then we have no explanation for why there
would be a rational requirement to intend the means of buying a ticket, for
doing so should be viewed by the agent as a potential waste of money given
his agnosticism about whether he’ll go.
Velleman might be right that there would be no categorical requirement
for the agnostic traveler to intend to buy the ticket. But it doesn’t follow
from this that Means-Ends Coherence doesn’t apply to the agnostic traveler.
Means-Ends Coherence is a wide-scope requirement; what it requires is that
one either intend the means, or abandon the end, or give up one’s instru-
mental belief. It doesn’t follow from one’s not being required to intend the
means that this disjunctive requirement isn’t in place.
Why is Means-Ends Coherence formulated this way? One reason is that
it often happens that we recognize an end of ours as imprudent, immoral,
or otherwise unreasonable. In such cases, it seems perfectly rational for an
agent to give up his end. But if we understood Means-Ends Coherence as a
narrow-scope requirement—a categorical requirement to intend the means,
as Velleman puts it—we would have to concede that one who abandons an
end instead of intending the means violates this rational requirement. But
that seems to be the wrong result; abandoning the end instead could involve
no irrationality whatsoever. And Velleman’s example seems to be precisely

16
 As an anonymous referee correctly observed, the defender of the Strong Belief
Thesis could deny that one intends to X in such cases, and insist that one only intends to
try to X—a strategy we’ve already rejected as unhelpful to the cognitivist—and then insist
that it’s still the case that the natural expression of an intention to X is “I am going to
X.” However, this shows that the argument from Anscombe’s remark will be convincing
only to those who already accept a controversial response to the alleged counterexamples.
17
  Velleman (2007: 205).
 Cognitivism about Practical Rationality 27

the kind of case that motivates the wide-scope formulation in this way. In
being agnostic about one’s success in getting to Chicago, and seeing the sig-
nificant expense involved in buying the ticket, the agent would be perfectly
rational in giving up the end instead of intending the means.
There are two points to make about Velleman’s argument. First, even in
cases where one isn’t agnostic, but instead believes one will carry out what
one intends, it could be rational for one to abandon one’s end instead. For
instance, one could realize that the expense involved in intending the means
is so great that, even if one achieves one’s end, it isn’t worth it. In such cases,
it would be false to say that there’s a categorical requirement to intend the
means. So, insisting on the Strong Belief Thesis won’t be enough to block
this allegedly bad result.
Secondly, and more importantly, this allegedly bad result isn’t actually
a bad result. There’s no cost in our having to say that, in Velleman’s exam-
ple, there’s no narrow-scope requirement to intend the means. There would
be a cost in saying that there’s no applicable requirement of Means-Ends
Coherence. But it doesn’t follow from there being no narrow-scope require-
ment to intend the means that no wide-scope requirement of Means-Ends
Coherence is in place.18
Velleman offers a third argument for the Strong Belief Thesis:  if the
Strong Belief Thesis were not true, then we would be unable to explain how
intentions function so as to coordinate behavior. Velleman writes:
When an intention coordinates behavior, the agent and his associates proceed on
the assumption of its being executed—which would be an odd way to proceed if
the agent himself were agnostic on the question. If I  am agnostic as to whether
I will be in Chicago, why should anyone plan or act on the assumption of my being
there. And why should anyone hesitate to plan or act in ways inconsistent with that
eventuality.19
In short, Velleman argues that, since we already believe that intentions play
a role in coordinating behavior, and the Strong Belief Thesis is necessary to

18
 Bratman (2009b:  §8) also observes Velleman’s mistake of understanding
Means-Ends Coherence as a narrow-scope requirement to intend the means. In reply-
ing to Velleman’s argument, Bratman goes on to develop the idea that intentions have a
distinctive aim, in much the same way that belief is sometimes thought, by Velleman and
others, to have a distinctive aim. In particular, Bratman suggests that “intentions aim at
the coordinated control of action that achieves what is intended” and this explains why
Means-Ends Coherence is a rational requirement. Exploring this would take us too far
afield. And we need not do so to show that Velleman’s argument doesn’t succeed. Rather,
we can simply note that a crucial premise of Velleman’s argument—namely, that if there’s
no requirement to intend the means, then Means-Ends Coherence does not apply—is
false and so this argument for the Strong Belief Thesis is unsound.
19
  Velleman (2007: 206).
28 John Brunero

explain how intentions play that role, we should believe in the Strong Belief
Thesis as well.
Velleman’s argument identifies two specific coordinating roles for inten-
tions: intentions coordinate the behavior of both the agent herself and the
behavior of her associates. However, it’s not clear how this latter coordinat-
ing role provides support for the Strong Belief Thesis, since what matters for
such coordination is what the associates believe, not what the intending agent
herself believes, about whether the intention will be successfully executed.
But the Strong Belief Thesis is a thesis about what the intending agent her-
self believes.
Putting aside this worry for a moment, it’s not clear that when an inten-
tion coordinates behavior, one’s associates always “proceed on the assump-
tion of its being executed,” though they may sometimes do. Suppose we are
teammates on a basketball team and you intend to make a shot, and, aware
of your intention, I position myself for a rebound. I’m clearly not proceed-
ing on the assumption that you’ll execute your intention. Had I proceeded
on that assumption, I would have hurried back down the court to set up
on defense. Or suppose I’m in the stands watching the game, and someone
offers me a bet on whether you’ll make the shot. I’d be irrational to pro-
ceed on the assumption you’ll successfully execute your intention. I should
instead consider the probability of your doing so and calculate the expected
utility of accepting the bet. And much the same goes for the agent herself.
When one coordinates behavior with one’s own intentions, one need not
proceed on the assumption that one’s intention will be successfully exe-
cuted. The shooter might position herself for a rebound. And she might not
bet the farm on making the shot.
So, it’s a mistake to characterize the coordinating role of intentions
merely in terms of how agents (the actor herself or her associates) plan on
the assumption that the intention will be successfully executed. Sometimes
we do this. But sometimes we don’t, and we instead consider the probability
that the intention will be successfully executed. So, with this in mind, we
should return to our question: is the Strong Belief Thesis needed to explain
the coordinating role of intentions?
Now it seems as though some weaker thesis about the intention–belief
connection might actually do a better job of explaining the coordinat-
ing role. Consider, for instance, Robert Audi’s view of intention. Audi
doesn’t accept the Strong Belief Thesis, but thinks that intending to X
involves believing that X is more likely than not.20 (According to Audi’s
view, both Shaquille O’Neal and Wilt Chamberlain intended to make free
throws—barely—whereas Ben Wallace hoped, but did not intend, to make

20
  Audi (1973).
 Cognitivism about Practical Rationality 29

them—assuming all were aware of their appalling free throw percentages.21)

On this view, sometimes one will intend to X and believe one will X,
whereas other times one will intend to X and merely believe the probability
of one’s X-ing is greater than (.5).
Audi’s view—which I’m not endorsing here—seems to do a better job
explaining the coordinating role of intentions than the Strong Belief Thesis
does. Since it allows that sometimes one intends to X and believes one will X,
it can account for all those cases in which we plan on the assumption that one
will X. But since it also allows that one can intend to X without believing one
will X, it can also account for some cases where one plans in light of the prob-
ability that one will X. So, it turns out that Velleman is wrong, and the Strong
Belief Thesis isn’t needed to explain the coordinating role of intentions.22 Once
we consider that not all coordination involves proceeding on the assumption
of successful execution, this argument for the Strong Belief Thesis fails.
Of course, Velleman might deny that these really are intentions. He
might say that in cases where the behavior of oneself or others is coor-
dinated based on some assessment of probabilities, rather than on an
assumption of successful execution, we’re not dealing with an agent’s inten-
tions—maybe only intentions to try, or intentions in the “goal-state” sense
of the term. One odd thing about this possible response (and Velleman’s
argument in general) is that whether something counts as an intention to
X shouldn’t depend upon whether or not someone else plans on the assump-
tion that I will X. What does their planning have to do with my beliefs?
Additionally, such a reply would employ strategies we’ve already rejected
as unhelpful to the cognitivist. Also, such a reply would appear to rig the
argument from the start: it would be using the Strong Belief Thesis to deter-
mine what counts as an intention, thereby narrowing what counts as the
relevant coordinating phenomena to be explained, and then saying that we
should believe in the Strong Belief Thesis because it’s necessary to explain

21
  Wallace shot 42% from the line for his career, whereas Wilt shot 51% and Shaq
53%. Mele (1992) uses the example of a free throw shooter to argue against Audi’s view.
See esp. pp. 131–2, 136–7. If we look at certain commonly accepted functions of inten-
tions—specifically, the functions of initiating and guiding an agent’s actions, coordinat-
ing behavior, terminating practical reasoning, etc.—then we won’t notice a difference
between, say, O’Neal and Wallace shooting a free throw. So, it seems odd to say that one
intends and one doesn’t.
22
  Holton (2008) makes the related point that an appeal to partial, as opposed to
all-out, belief would allow for a response to Velleman’s coordination argument for the
Strong Belief Thesis. (Audi’s view is that intention involves an all-out belief that success
is more likely than not, not a partial belief in success.) Holton argues that intentions,
coupled with partial beliefs in success, can play a role in coordinating the behavior of self
and others. (As Holton correctly observes, one’s informing others of one’s uncertainty of
success facilitates that coordination.)
30 John Brunero

the relevant coordinating phenomena. It wouldn’t give us an independent

reason for believing in the Strong Belief Thesis to begin with.
Velleman offers a fourth argument for the Strong Belief Thesis, arguing
that we can’t explain a consistency requirement on intentions unless the
Strong Belief Thesis is true:
But why should my intentions be subject to a requirement of consistency if I can
remain cognitively uncommitted to their truth? If I am agnostic as to whether I’ll be
in Chicago on Tuesday evening, why should my plans for Tuesday evening have to
be consistent with my being there?23
Velleman specifically asks why he shouldn’t also make dinner plans some-
where local for Tuesday evening, knowing he won’t be able to both be in
Chicago and keep the dinner reservation.
The first thing to note is that we should distinguish cases of inconsist-
ency in intentions, which is irrational, from cases of contingency planning,
which need not be irrational. If I doubt I’ll succeed in going to Chicago, it
may be perfectly rational to put in place a backup plan that involves dinner
reservations at a local restaurant.24 We aren’t tempted to describe such cases
of contingency planning as involving an intention to eat at the restaurant;
rather, we would say that one intends to eat there if one stays in town.
And this contingency plan isn’t inconsistent with one’s planning to be in
Chicago instead.
Additionally, even if such contingency planning involved inconsistency,
this wouldn’t give us grounds for accepting the Strong Belief Thesis, since
we engage in contingency planning even when we believe we’ll succeed in
our intentions. For instance, when I intend to clear the small pond on the
easy par 3 and believe I will, I may still carry an extra ball just in case I’m
wrong. So, adopting the Strong Belief Thesis won’t help one avoid this sup-
posed inconsistency in intentions.
So, let’s assume we’re not dealing with a case of contingency planning,
but rather a case of someone who intends to be in Chicago and intends
to eat at a local restaurant, knowing he can’t do both. Velleman asks “why
should my intentions be subject to a requirement of consistency if I can
remain cognitively uncommitted to their truth?” There are two ways to
understand this question. On one reading, Velleman is suggesting that, if
intentions did not involve cognitive commitment, then there would be no
consistency requirement governing intentions. But since there is a consist-
ency requirement governing intentions, it follows that intentions do involve

23
  Velleman (2007: 206).
24
  Of course, one wouldn’t want to invite others to dinner without telling them this is
just a contingency plan—that would risk rudeness.
 Cognitivism about Practical Rationality 31

cognitive commitment. On a second reading, Velleman is asking what

explanation can be given of a consistency requirement on intentions if it’s
not the cognitivist one.
Start with the first reading. I  don’t think Velleman is right that there
wouldn’t be a consistency requirement on intentions if the Strong Belief
Thesis were false. Let’s assume it’s false. Let’s assume again that Audi’s
weaker view of intention is correct, so that intending to X involves believ-
ing X is more likely than not. It doesn’t follow that we’re now rationally
permitted to have inconsistent intentions. Consider an example. Suppose
I  know that Farmer Joe grows five fruits (apples, oranges, plums, pears,
and peaches) and every day selects at random only four of those fruits to
bring to the Farmer’s Market. If Audi’s view is correct, in order to intend to
buy apples at the Farmer’s Market, I must believe it’s more likely than not
I’ll buy apples. But that doesn’t present a barrier to my intending to buy
apples; I can here intend to buy apples and believe it’s more likely than not
that I will. Likewise, I can intend to buy oranges, since I believe it’s more
likely than not that I’ll buy oranges. The same goes for the other fruits. But
now suppose that I tell you, knowing that Joe brings only four fruits to the
market, that I intend to buy apples, intend to buy oranges, intend to buy
plums, intend to buy pears, and intend to buy peaches. You would rightly
accuse me of irrationality in having these five intentions, while believing
Joe will bring only four fruits. This example suffices to show that Velleman
is wrong: even if the Strong Belief Thesis is false, one could still have inten-
tions that are rationally criticizable for their inconsistency.
On the second reading, Velleman is challenging his opponent to provide
some other explanation of the consistency requirement on intentions if it’s
not going to be the cognitivist one. This is a hard question. We’d have to
provide an account of intentions that shows why having intentions whose
contents are inconsistent (or inconsistent with the contents of one’s beliefs)
is irrational—and, of course, do so without appealing to the involvement
of beliefs. Moreover, we’d have to explain how intentions differ from other
attitudes, such as desires, where having that attitude towards inconsistent
contents need not involve irrationality. There is, after all, nothing irrational
about both wanting to go for a smoke (because it’s pleasant) and wanting
not to go for a smoke (because it’s healthy), but there is something irra-
tional about both intending to go for a smoke and intending not to go for
a smoke. That’s a task too complicated to take on here, but it’s also one we
need not pursue. Our project is to evaluate the cognitivist’s explanation
of the requirements of practical rationality, not to propose an alternative
theory. And, in any case, the lack of an alternative theory of the consist-
ency requirement on intentions wouldn’t constitute a convincing reason for
accepting the Strong Belief Thesis.
32 John Brunero

In summary, although the Strong Belief Thesis might be a philosophically

respectable but controversial thesis, I think we should be much less confident
that it could both be true and successfully employed in a cognitivist explana-
tion, since those maneuvers needed to save it from counterexamples (saying
one merely intends to try, or that “intend” is ambiguous) make it unsuitable
to play the necessary role in the cognitivist explanation. And the various argu-
ments I’ve considered in favor of the thesis aren’t convincing as they stand.

2.  Cognitivism without the Strong

Belief Thesis

Perhaps the Cognitivist could account for Means-Ends Coherence without

relying on the Strong Belief Thesis. I’ll start by introducing an account that’s
similar to one proposed by Jay Wallace, but able to avoid some difficul-
ties facing Wallace’s account, and then I’ll defend the key assumptions of
that account. I’ll argue that this is the best available cognitivist account of
Means-Ends Coherence.

2.1  Towards a Cognitivist Account without Strong Belief

Wallace’s account relies on a weaker thesis about the connection between
intention and belief: intending to X involves believing X is possible. On this
thesis, the person who intends to travel to New York will believe
(4) It is possible for me to travel to New York.
Wallace assumes that his instrumental belief takes the following form:
(5) It is possible for me to travel to New York only if I intend to buy an
airplane ticket.
Now suppose that the means-ends incoherent agent, who doesn’t intend to
buy the ticket, is “minimally self-aware” and so believes
(6) It is not the case that I intend to buy an airplane ticket.
In that case, the means-ends incoherent agent, Wallace argues, would have
inconsistent beliefs. He argues that if you are means-ends incoherent, you
will be
left in effect with the following incoherent set of beliefs (assuming you are minimally
self-aware): the belief that it is possible that you do x, the belief that it is possible that
you do x only if you also intend to do y, and the belief that you do not intend to do y.
The incoherence of these beliefs is a straightforward function of the logical relationship
 Cognitivism about Practical Rationality 33

among their contents, suggesting that the normative force of the instrumental principle
can be traced to independent rational constraints on your beliefs—in particular, to
constraints on certain combinations of beliefs . . .25
This cognitivist explanation of Means-Ends Coherence doesn’t require the
Strong Belief Thesis.
However, one might have worries about this weaker thesis as well. Couldn’t
one intend to X while being agnostic about whether one can X? Suppose
I intend to go shopping this evening, but I’m not sure whether the only shop-
ping mall in town is open, and so I’m agnostic about whether I can go shop-
ping. It’s not that I believe I can’t go shopping; I just don’t believe I can.26 It
might be poor planning on my part to form that intention before finding out
whether the mall is indeed open, but it seems possible to do so.
Additionally, one might have some doubts about the logical form of the
instrumental belief in (5), which may not be obvious. Perhaps what one
believes is that the combination of traveling to New York and not intending
to buy a ticket is not possible, so that the logical form would be:
~◇ (N & ~T)
where “N” is “I travel to New York” and “T” is “I intend to buy a ticket.” On
Wallace’s view, the inconsistency of the beliefs is “a straightforward function
of the logical relationship among their contents” which he takes to be:
(4′) ◇ N
(5′) ◇ N → T
(6′) ~T
But if the logical relationship among their contents were instead
(4′) ◇ N
(5″) ~◇ (N & ~T)
(6′) ~T
then the contents of the involved beliefs wouldn’t be inconsistent. (To see
that there’s no inconsistency here, consider another example:  one might
consistently believe Hank isn’t going to Nashville, isn’t going to Tennessee,
can go to Nashville, can go to Tennessee, but can’t go to Nashville and not
go to Tennessee.27)

25
  Wallace (2006: 106).
26
  An anonymous referee has suggested, plausibly, that to intend to go shopping, one
must have at least some evidence that one can go shopping. But that still falls short of
saying one must believe one can go shopping.
27
  Note that it won’t help Wallace to claim intending to travel to New York this after-
noon involves believing (4″) ◇ (N & ~T). That would indeed generate an inconsistency
34 John Brunero

Here’s a reason for thinking Wallace has misidentified the logical form
of the instrumental belief. Suppose he is right that the logical form of the
instrumental belief is ◇N → T, the contrapositive of which is ~T → ~◇N.
Possibility and necessity are defined in terms of one another, such that ~◇N
→ □~N, and so it follows from these two claims that ~T → □~N. But this
seems wrong. This reading licenses us to detach the claim that □~N via
modus ponens from ~T. But my not traveling to New York this afternoon
is surely believed to be a contingent matter, not a necessary one. What
I  should conclude is that I’m not traveling to New  York this afternoon,
not that it’s necessary that I’m not traveling to New  York this afternoon.
I don’t take my not traveling to New York this afternoon to be necessary;
rather, I only take it to be necessary that it’s not the case that I both travel
to New  York this afternoon and not buy a ticket. So, it seems better to
understand the logical form of one’s instrumental belief as □ (~T → ~N),
instead of ~T → □~N.28
Perhaps there’s a way to defend some specific conception of possibility
that allows us to avoid these two objections to Wallace’s account.29 But it
might be easier to develop a version of cognitivism that preserves the central
insights of Wallace’s approach, but avoids the difficulties that come with
talk of possibility. I’ll here outline such an account, and then defend its

in belief with (5″), ~◇ (N & ~T), but that would make one’s belief (6′) irrelevant to
the inconsistency in belief. One’s not intending the means, and so believing one doesn’t
intend the means, wouldn’t matter—you’ve got an inconsistency regardless. But, if we’re
trying to explain Means-Ends Coherence, whether one intends the means should matter.
28
  See Hughes and Cresswell (1996: 14–16).
29
 Perhaps we should understand the possibility involved here as epistemic possibility,
so that our subject believes, for instance, in (4), that for all he knows, it’s possible that
he travels to New York. (Thanks to Mike Titelbaum for this suggestion.) Since it seems
mere belief, not knowledge, is relevant to the alleged inconsistency in (4)–(6), the relevant
sense of possibility would be doxastic possibility, which Chalmers helpfully defines as
follows: “A scenario [a ‘maximally specific way the world might be’] is doxastically pos-
sible for a subject if and only if it is not doxastically ruled out by any of the subject’s
beliefs” (2011:  62–3). This might circumvent our two objections. First, it’s plausible
that our agent believes that going shopping isn’t doxastically ruled out by any of his
beliefs. Second, the detachment of □~N isn’t as worrisome, since this should be under-
stood as stating that N is ruled out by some of the subject’s beliefs. However, one might
worry that, once we establish this as the relevant sense of “possibility” at work, we should
worry about Wallace’s intention–belief thesis. Is it true that one can intend to X only if
one believes X is doxastically possible? Perhaps not. Indeed, cases when one is aware of
one’s own instrumental irrationality—specifically, when one is aware that one doesn’t,
but must, intend the means, if one is to X—might be cases where a subject doesn’t believe
X-ing isn’t ruled out by any of his beliefs. Wallace’s thesis now appears false. Moreover,
if Wallace’s thesis were true, we’d have to say that the subject doesn’t really intend to X,
and so there is no instrumental irrationality for him to be aware of. That’s implausible.
(But for further discussion of the notion of possibility at work in Wallace’s account, see
Wallace 2006: 114–17.)
 Cognitivism about Practical Rationality 35

components against some objections.30 First, let’s follow Wallace in assum-

ing that the instrumentally incoherent agent who is minimally self-aware
will believe
(7) I do not intend to M.
Second, let’s assume that the agent’s instrumental belief takes the
following form:
(8) If I do not intend to M, I certainly will not E.
If we apply Closure, then rationality requires that he believe
(9) I certainly will not E,
or revise his other beliefs. Third, let’s assume a thesis about the connec-
tion between intention and belief that’s far weaker than any we’ve discussed
so far:
Very Weak Belief Thesis: Intending to X involves not believing that one
certainly will not X.
According to this thesis, someone who believes (9) does not intend to E.
On this account, whenever one is means-ends incoherent, one has beliefs
that violate Closure. Moreover, if a means-ends incoherent agent were to
come to comply with Closure, he would make revisions that would also
remove him from his state of means-ends incoherence. We’re assuming that
the means-ends incoherent agent believes both (7) and (8). So, he should,
according to Closure, either come to believe (9), which, by the Very Weak
Belief Thesis, will involve his not intending to E (thereby removing him
from a state of means-ends incoherence), or give up his instrumental belief
(8)  (thereby removing him from a state of means-ends incoherence), or
give up his belief (7), which, we’ll assume for now, will involve him form-
ing the intention to M (thereby removing him from a state of means-ends
coherence).
This account differs from Wallace’s account in at least two ways. First,
instead of holding that intention involves having a certain belief, it holds
that intention involves not having a certain belief. Second, like the account
we considered in §1, it employs Closure, rather than a consistency require-
ment on beliefs. I think this account is preferable since it avoids the two
objections to Wallace’s account. But challenges could still be raised about
the Very Weak Belief Thesis. Additionally, both this account and Wallace’s
account face an obvious objection: what if someone were to falsely believe

30
 This account is a slight variation on the account that’s presented in §3.4.3 of
Kolodny and Brunero (2013).
36 John Brunero

she intends the means? Such a person, it appears, wouldn’t violate any theo-
retical requirements of rationality, but would be instrumentally incoherent
in not intending the means. I’ll first provide some defense of the Very Weak
Belief Thesis, and then turn to that objection.

2.2  The Very Weak Belief Thesis Defended

Rather than hold that an intention to X involves having some belief (that
one will X; that it’s more likely than not one will X; that it’s possible to
X), we’re holding merely that intending to X involves not believing one
certainly will not X. But even this view has been challenged. Anscombe has
argued that “in some cases, one can be as certain as possible that one will
do something, and yet intend not to do it.”31 She provides three examples.
First, consider someone hanging on a ledge who intends to hold on, but
“is as certain as possible that he must let go and fall.” Second, consider a
prisoner who intends not to break down under torture but is “as certain as
possible that he will break down.” Third, consider St Peter, who has it on
highly reliable authority that he’ll deny he knows Christ three times, but
intends not to do so.
There’s a common feature to Anscombe’s examples:  the agent knows
that he can carry out his intention for some period of time, but is certain
that after some point, he’ll no longer be able to do so (though he may not
have any precise view about when that point is). For instance, the man on
the ledge believes he can hold on for some time, but is certain his fingers
will eventually give out. And the prisoner thinks he’ll initially resist the
demands of his torturers, but is certain his willpower will eventually fade.
The relevant question here concerns how we should construe the inten-
tions involved. If we say, for instance, that the person on the ledge intends
to hold on for some time, or intends to hold on as long as he can, then we
wouldn’t have a counterexample to the Very Weak Belief Thesis; he wouldn’t
be intending to do something he believes he certainly won’t do. For it to be
a counterexample, he would have to intend to hang on as long as he can and
then some, or perhaps intend to hang on for some specific period of time for
which he is certain he won’t hang on. But once we specify what the relevant
contents of his intention must be, if it’s to be a genuine counterexample to
the Very Weak Belief Thesis, it becomes easier to deny that these really are
his intentions.32 So, I  don’t think any of Anscombe’s examples provide a
conclusive counterexample to the Very Weak Belief Thesis.

31
  Anscombe (1957: 94).
32
  I here disagree with Holton (2008), who is convinced by these examples.
 Cognitivism about Practical Rationality 37

Anscombe does give an argument immediately preceding her introduc-

tion of the examples. In this passage, she is explaining the plausibility of
someone saying, “I am going to [X]‌unless I don’t” when one intends to X
but has doubts about whether one will X. She writes:
“I am going to . . . unless I do not” is not like “This is the case, unless it isn’t”. It has
an analogue in estimates of the future: “This is going to happen . . . unless it doesn’t”.
(Someone may prevent it.) This could be said even of an eclipse of the sun; because
the verification of predictions awaits the event—and the sun might blow up before
the eclipse. It is for this reason that one can be as certain as possible that one will do
something, and yet intend not to do it. (1957: 93–4)
Anscombe is correct that we can’t verify that some predicted event occurs
until after that event occurs. But it’s unclear how it follows from this that one
can be as certain as possible one will X and yet intend not to X. Moreover,
if this is Anscombe’s motivation for rejecting the Very Weak Belief Thesis,
it’s unclear why she would limit herself to examples that share the common
feature mentioned above (where one believes one will succeed for some time
but is certain one will eventually fail), since this line of reasoning would
support one’s intending any future event that one is as certain as possible
one will not do, since, for any future event, one’s failure is not yet verified.
For instance, I could intend to jump to the moon, or intend to stop the
solar eclipse. By Anscombe’s reasoning in this passage, we would have to
concede such intentions are possible.
So, Anscombe’s arguments against the Very Weak Belief Thesis aren’t
conclusive. Is there anything that could be said in favor of this thesis? I’ll
here give two reasons to think it’s true, both of which are grounded in a
plausible account of the functions of intentions.
First, one function of intentions is the settling of deliberation. In delib-
erating about what to do in some future situation, one can now form an
intention about what to do then, and carry the results of that deliberation
forward until the time of action.33 There are obvious advantages to being
able to do this. It would take a lot of time and resources to continually
deliberate up until the time of action. By forming an intention, we set-
tle on some course of action, and remove the need to continually deliber-
ate, thereby saving deliberative resources. (Of course, sometimes we reopen
questions that are settled, especially if we receive new information.) But it’s
unclear how one could settle deliberation in this way when the course of
action one settles upon isn’t recognized as a live option. It seems that if my
deliberation has settled on my X-ing, but I am certain I will not X, there
will be strong pressure to reconsider and instead decide upon a live option.

33
  See Bratman (1987: 2).
38 John Brunero

Second, intentions also function to give direction to further planning.34

An agent with an intention will be disposed to form sub-plans for achieving
her end. For instance, when I intend to write a paper, I’ll be disposed to
form further plans at the appropriate times to facilitate my writing it (set-
ting aside time, gathering relevant books from the library, etc.). And my for-
mation of these sub-plans will be guided by my beliefs about how effective
various means will be in realizing that end. For instance, if I think a library
book by Parfit will help me write my paper, but another book by Williams
would be of no use, I’ll plan on checking out the Parfit book, and not the
Williams book. But if I’m certain that I won’t do what I intend to do, it’s
hard to see how one would be disposed to engage in further sub-planning.
Since I’m certain I won’t achieve my ends, any further sub-planning toward
that end should be seen as ineffective in achieving my end, and so should
be treated in the same way I treat the idea of getting the book by Williams
that I  won’t need for my paper. In other words, since, by my lights, no
sub-plan will effectively facilitate the achievement of my end, I would not
be disposed to engage in any sub-planning.

2.3  Unknown Failures to Intend

Both of the accounts discussed in §2.1 face an obvious objection. What if
one didn’t intend the means, but didn’t believe this? For instance, suppose
I never notice that I don’t intend to buy a ticket to New York. Or suppose
I falsely believe instead that I do intend to buy a ticket. I violate Means-Ends
Coherence, but yet I  don’t have the relevant belief that would allow for
either of these cognitivist explanations of Means-Ends Coherence to get off
the ground.
It won’t help to deny the possibility of such ignorance and false beliefs.35
Nor could we easily claim such a person would be irrational, if we under-
stand rationality as a matter of coherence among one’s attitudes, since an
agent’s being unaware of, or having false beliefs about, the absence of an
intention need not involve any incoherence. Perhaps the cognitivist’s best
bet is simply to concede that one isn’t instrumentally irrational in such
cases. In other words, someone can get off the hook, as far as the charge
of instrumental irrationality goes, just by being ignorant of, or having false
beliefs about, her not intending the means. This response might not be as
far-fetched as it sounds. After all, we can also get off the hook by being

34
  See Bratman (1987: 3).
35
  See Schwitzgebel (2010) for an overview of some of the psychological evidence and
philosophical arguments concerning the possibility of mistaken beliefs about our own
psychology.
 Cognitivism about Practical Rationality 39

ignorant of, or having false beliefs about, the necessary means to our ends.
Additionally, precise formulations of the requirement will aim to take into
account rational self-trust.36 It’s not irrational not to intend the means to
one’s ends when one trusts one will intend the means in the future. For
instance, suppose I intend to buy a house next year, and believe that to do
so, I must intend to fill out some legal paperwork. I don’t currently intend
to fill out that paperwork, but I trust that I will when the time comes. It
doesn’t seem right to accuse me of irrationality. To get off the hook, as far
as the charge of instrumental irrationality goes, I can form some (perhaps
false) beliefs about my future intentions. But, if we allow that false beliefs
about one’s future intentions get one off the hook, why not also allow that
false beliefs about one’s current intentions can get one off the hook as well?
Perhaps this response to the objection isn’t convincing. If it isn’t, I don’t
think it poses any threat to the thesis being defended here: that this cog-
nitivist account of Means-Ends Coherence is better than any other avail-
able cognitivist account, especially accounts that employ the Strong Belief
Thesis, since those accounts will also face a version of this same objection.
Recall the account from §1. Couldn’t I come to believe (3)—that I intend
to M—without intending to M? This objection is just as much an objection
to the account of Means-Ends Coherence based on the Strong Belief Thesis.
And so it doesn’t challenge my contention that the account sketched in §2.1
is the best available cognitivist account of Means-Ends Coherence.

3. Problems for Cognitivism’s

Explanatory Claim

But I don’t think the best available cognitivist account is good enough. It
might be true that whenever one is means-ends incoherent, one also has
beliefs that violate requirements of theoretical rationality. But cognitivism
makes a further claim: those theoretical requirements explain the practical
ones. However, one might doubt this explanatory claim. One might think
that, even though every violation of Means-Ends Coherence involves a vio-
lation of requirements of theoretical rationality, the latter requirements don’t
explain the former. I’ll give two reasons for doubting the explanatory claim.
First, note that cognitivism seems to be unpromising for explaining
practical rationality in general. For instance, cognitivism doesn’t seem to
be a plausible strategy for accounting for the consistency requirement on
intentions once we reject the Strong Belief Thesis. Weaker views about
the intention–belief connection seem unable to do the trick. For instance,
36
  See, for instance, Setiya (2007: 668).
40 John Brunero

as Bratman observes, if we work with Wallace’s view that intending to X

involves believing it is possible to X, we can’t explain what’s wrong with
intending to X and intending not to X since the associated beliefs (it is pos-
sible to X; it is possible not to X) are perfectly consistent with one another.37
If we took Audi’s view of intentions, the associated beliefs would be incon-
sistent (X-ing is more likely than not X-ing; not X-ing is more likely than
X-ing) but such a view doesn’t help once we expand our stock of intentions.
It’s irrational to intend to A, intend to B, and intend to C, while believing
that, if one As and Bs, one will not C. But the associated beliefs (A-ing is
more likely than not A-ing; B-ing is more likely than not B-ing; C-ing is
more likely than not C-ing) need not be inconsistent. (If the odds of A-ing
and the odds of B-ing are each only slightly better than a coin toss, then it
wouldn’t be more likely than not that one both As and Bs, and so it could
still be that C-ing is more likely than not C-ing.)
But besides consistency constraints on intentions and beliefs, there is
much more that rationality requires of us. For instance, rationality also
prohibits one from believing one ought to ϕ but not intending to ϕ. But
no one has given a cognitivist explanation of this Enkratic Requirement.
Additionally, Means-Ends Coherence is narrowly restricted to our beliefs
about necessary means to our ends. But we might think that practical ration-
ality extends beyond this, and has something to say about beliefs about
non-necessary means, or about our preferences or values.
It seems that cognitivism can only explain a small part of the relevant
phenomena in need of explanation. One might wonder what explanation
can be given of these other requirements of practical rationality, and whether
that same explanation could then be given for Means-Ends Coherence. If we
can find another theory that can explain Means-Ends Coherence and other
requirements of practical rationality, we would have reason to prefer that
theory to cognitivism. Considering other views would take us too far afield,
but we should at least register a suspicion about a theory that can explain
only a small part of the relevant phenomena needing to be explained.
But there is a second, more serious, worry about the cognitivist’s explana-
tory claim. The worry is that the cognitivist’s explanation appears to work
only because she considers some applicable theoretical requirements while
ignoring others. Once we consider all the applicable theoretical require-
ments, it no longer appears plausible to think that Means-Ends Coherence is
explained by the requirements of theoretical rationality. Consider someone
who violates Closure by believing P, P→Q, but not believing Q. When one
notices one holds this irrational combination of attitudes, one should look
to one’s reasons for belief in order to determine the appropriate direction of

37
  See Bratman (2009b: §3) and Ross (2009: 245).
 Cognitivism about Practical Rationality 41

response. Let’s assume that P→Q is a fixed background belief. If one thinks
there are really strong reasons for believing P, and relatively weak reasons for
not believing Q, then one should escape this irrational state by coming to
believe Q. However, if one thinks there are strong reasons not to believe Q,
and relatively weak reasons for believing P, one should revise one’s beliefs
in the modus tollens direction instead, and cease to believe P. In short, one’s
assessment of the relevant reasons for belief determines the appropriate
direction of revision.38
But now consider our means-ends incoherent agent who, on our cog-
nitivist account, believes he doesn’t intend to buy an airplane ticket, and
believes he certainly will not travel to New York if he doesn’t intend to buy
an airplane ticket, but does not believe that he certainly will not travel to
New York. Let’s suppose that he considers the relevant reasons for belief.
Presumably, the evidence that he doesn’t intend to buy an airplane ticket
will be rather strong. Introspection isn’t foolproof, but it will normally pro-
vide very strong evidence that he doesn’t intend to buy an airplane ticket—
and there doesn’t seem to be any evidence on the other side. Taking into
account the requirements of theoretical rationality governing the direction
of revision, it seems that in all but the most unusual cases, revision should
proceed in only one direction: coming to believe he certainly will not travel
to New York (thereby ceasing to intend to travel to New York). Cognitivism
thus predicts an asymmetry in the direction of response.39
But, as Means-Ends Coherence is usually understood, there is no such
asymmetry. (Indeed, if there’s any temptation toward thinking there’s an
asymmetry, it would be toward proceeding in the other direction: coming
to intend the means.) Rather, one could escape a state of means-ends inco-
herence by either abandoning the end, or by coming to intend the means.
This symmetrical feature of Means-Ends Coherence is unexplained by the
cognitivist account, and so the cognitivist account is inadequate.

38
  I’ll here avoid the complicated task of specifying the principles of rationality gov-
erning such revision. But I do think such principles are principles of rationality, not prin-
ciples of reason. One’s assessment of the reasons could be mistaken. For instance, it could
be that a person who thinks there are strong reasons to believe P is mistaken, and there
are really strong reasons to believe ~Q instead. Here, rationality requires him to revise his
attitudes in such a way that he comes to have a belief not well-supported by reasons. (A
similar point could be made for the practical case.)
39
  Consideration of such asymmetries have played an important role in arguments
against the wide-scope formulation of some rational requirements, including, most not­
ably, the Enkratic Requirement. See esp. Schroeder (2004) and Kolodny (2005). It is
argued that such asymmetries put pressure on us to construe the Enkratic Requirement
as a narrow-scope requirement to intend to do what we believe we ought to do. If cogni-
tivism is true, we should expect to feel the same pressure toward construing Means-Ends
Coherence as a narrow-scope requirement (to abandon the end). That we don’t feel such
pressure is a reason to think cognitivism is false.
42 John Brunero

In summary, when we consider the requirements of theoretical rationality

in an appropriately holistic way—that is, we don’t select one requirement
and ignore how it interacts with other relevant rational requirements—then
we see that the requirements of theoretical rationality offer a poor expla-
nation of Means-Ends Coherence. Theoretical rationality requires that we
respond to states of incoherence by revising in light of our assessment of
the relevant reasons for belief, which are provided by the evidence, while
practical rationality requires that we respond to states of incoherence by
revising in light of our assessment of the relevant reasons for action. These
rational requirements may not push in the same direction. Specifically,
as I’ve argued, the theoretical requirements issue the false prediction that
we should almost always respond to instrumental incoherence by giving
up on our ends. So, the cognitivist, while perhaps establishing that viola-
tions of Means-Ends Coherence involve violations of theoretical rationality,
hasn’t established that the requirements of theoretical rationality explain
Means-Ends Coherence.

4. Conclusion

In this chapter, I’ve considered the merits of some possible ways of develop-
ing a cognitivist account of Means-Ends Coherence. I’ve argued that we
should reject accounts that rely on the Strong Belief Thesis, since that thesis
is either false or unable to do the explanatory work the cognitivist needs
it to do, and the arguments for that thesis aren’t convincing. I’ve argued
that the cognitivist would do better to work with an account modeled on
Wallace’s account, but which employs a weaker belief thesis, avoids talk
of possibility and necessity, and employs Closure instead of a consistency
requirement. However, in the end, I’ve argued that there is much about
practical rationality that cognitivism can’t explain, and the explanation it
gives of Means-Ends Coherence appears to work only because it considers
some theoretical requirements in abstraction from others.40

40
  Work on this chapter developed out of research for the SEP entry on instrumental
rationality, co-authored with Niko Kolodny, and no doubt benefited much from Niko’s
contributions to that project. Thanks also to Waldemar Rohloff, Dana Tulodziecki,
Eric Wiland, and two anonymous referees for helpful comments on this chapter. And
thanks to audiences at the University of Missouri, and, especially, the 2012 Wisconsin
Metaethics Workshop.
 Cognitivism about Practical Rationality 43

References
Adams, F. 2007. “Trying with the Hope,” in M. Timmons, J. Greco, and A. Mele
(eds), Rationality and the Good, 143–62. Oxford: Oxford University Press.
Anscombe, E. 1957. Intention. Oxford: Basil Blackwell.
Audi, R. 1973. “Intending,” Journal of Philosophy, 70(13): 387–403.
Bratman, M. 1987. Intentions, Plans, and Practical Reason. Cambridge, MA: Harvard
University Press.
Bratman, M. 2009a. “Intention, Belief, and Instrumental Rationality,” in D.
Sobel and S. Wall (eds), Reasons for Action, 13–36. Cambridge:  Cambridge
University Press.
Bratman, M. 2009b. “Intention, Belief, Practical, Theoretical,” in S. Robertson
(ed.), Spheres of Reason, 29–62. Oxford: Oxford University Press.
Broome, J. 1999. “Normative Requirements,” Ratio, 12(3): 398–419.
Brunero, J. 2009. “Against Cognitivism about Practical Rationality,” Philosophical
Studies, 146(3): 311–25.
Chalmers, D. J. 2011. “The Nature of Epistemic Space,” in A. Egan and B.
Weatherson (eds), Epistemic Modality, 60–107. Oxford: Oxford University Press.
Davidson, D. 2001. “Intending,” in Essays on Actions and Events, 83–102.
Oxford: Clarendon Press, 2nd edn.
Davis, W. 1984. “A Causal Theory of Intending,” American Philosophical Quarterly,
21(1): 43–54.
Grice, P. 1971. “Intention and Uncertainty,” Proceedings of the British Academy,
57: 263–79.
Hampshire, S., and Hart, H. L. A. 1958. “Decision, Intention, and Uncertainty,”
Mind, 67: 1–12.
Harman, G. 1976. “Practical Reasoning,” Review of Metaphysics, 29(3): 431–63.
Harman, G. 1986. Change in View:  Principles of Reasoning. Cambridge,
MA: MIT Press.
Holton, R. 2008. “Partial Belief, Partial Intention,” Mind, 117(1): 27–58.
Hughes, G. E., and Cresswell, M. J. 1996. A New Introduction to Modal Logic.
London: Routledge.
Kolodny, N. 2005. “Why be Rational?” Mind, 114(455): 507–63.
Kolodny, N., and Brunero, J. 2013. “Instrumental Rationality,” in Edward N. Zalta
(ed.), The Stanford Encyclopedia of Philosophy. Stanford, CA: Stanford University
Press, spring 2013 edn.
Mele, A. 1992. “Intention, Belief, and Intentional Action,” in Springs of Action,
128–53. Oxford: Oxford University Press.
Ross, J. 2009. “How to Be a Cognitivist about Practical Reason,” in R. Shafer-Landau
(ed.), Oxford Studies in Metaethics, iv. 243–82. Oxford: Oxford University Press.
Schroeder, M. 2004. “The Scope of Instrumental Reason,” Philosophical Perspectives,
18: 337–64.
Schwitzgebel, E. 2010. “Introspection,” in Edward N. Zalta (ed.), The Stanford
Encyclopedia of Philosophy. Stanford, CA: Stanford University Press, fall 2010 edn.
44 John Brunero

Setiya, K. 2007. “Cognitivism about Instrumental Reason,” Ethics, 117(4): 649–73.

Velleman, J. D. 1989. Practical Reflection. Princeton: Princeton University Press.
Velleman, J. D. 2007. “What Good is a Will?,” in A. Leist and H. Baumann (eds),
Action in Context, 193–215. Berlin: de Gruyter.
Wallace, R. J. 2006. Normativity and the Will. Oxford: Clarendon Press.
 3
Vindicating Practical
Norms: Metasemantic Strategies
Hille Paakkunainen

Some ways of updating belief have more epistemic merit than others.
Paul Boghossian (2000, 2001, 2003)  and Christopher Peacocke (2000,
2004) have defended varieties of the view that the epistemic merit of cer-
tain ways of updating belief is explained by facts about the conditions of
possessing certain concepts. In particular, they argue that if it is a condi-
tion of possessing a concept C that one must be “willing,” in a sense to be
explored, to update one’s beliefs in accord with a putative norm N, then
beliefs updated in accord with N are (defeasibly) epistemically warranted
in virtue of this fact.1 Let us say that a norm N is a “legitimate” epistemic
norm when updating beliefs in accord with N has the relevant sort of epis-
temic merit: it leads to defeasibly warranted belief. Let us say that a strategy
for explaining the legitimacy of a putative norm N is a strategy for “vindi-
cating” N. And following Peacocke, let us call strategies of vindicating N
by appeal to conditions of concept-possession “metasemantic.”2 Might a
parallel metasemantic approach be made to work in vindicating practical
norms, norms for updating intentions?
After rejecting some blind alleys, this chapter argues for a qualified “yes.”
I argue that there is a valid metasemantic route from premises about the con-
ditions of concept-possession to conclusions about which putative norms
for updating intentions are legitimate. The sense in which metasemanti-
cally vindicated practical norms are “legitimate” is, in the first instance, that
updating intentions in accord with such norms is rationally permissible. It
follows that there cannot be rational requirements not to update intentions

1
 For the language of “willingness,” see Boghossian (2003:  280)  and Peacocke
(2000: 265, 267).
2
 Peacocke (1993, 2004:  172). The literature often glides between talk of condi-
tions of semantic understanding and conditions of concept-possession. See Williamson
(2003: 272) for a brief defense of this practice in the context of metasemantic arguments.
I will only talk in terms of concept-possession.
46 Hille Paakkunainen

in these ways, or to update intentions in conflicting ways.3 We can thus

derive at least some of the content of any acceptable theory of practical
rationality from assumptions about the conditions of concept-possession.
I further examine, in a more tentative spirit, how we might be able to derive
rational requirements, not just rational permissions, from suitable premises
about concept-possession.
I will not here defend any premises about the conditions of
concept-possession of the sort that the metasemanticist needs. I merely note
that some such premises have seemed plausible to some philosophers. My
interest is in showing what such premises entail about rationality. For exam-
ple, it has seemed plausible to some that, to possess the concept ought
to Φ (where “Φ” stands for an action), one must be disposed to intend
to do what one judges one ought.4 Given such a premise, we can derive
an enkratic norm according to which agents are rationally permitted—or
perhaps more strongly, rationally required—to update their intentions to
match such ought-judgments.
§1 explains and motivates the basic metasemantic idea, and outlines the
argument to come.

1.  The Metasemantic Idea: Basics and

Motivation

Boghossian formulates the core metasemantic thesis as the following

“meaning-entitlement connection” (2003: 280):
MEC Any inferential transitions built into the possession conditions
of a concept are eo ipso [defeasibly] entitling.5
§§2–5 examine arguments for MEC or a practical parallel. Here I explain
what MEC says and how something like it might apply in the practical case;
and I briefly motivate attempting to defend it.
The idea that some type of inferential transition, in accord with a puta-
tive norm of inference N, is “built into” the possession conditions of a con-
cept C is, roughly, the idea that one must be in some sense “willing” to
engage in N-type inferences in order to possess C.  For Boghossian, such

3
  The notion of a rational permission is thus different from Boghossian’s “defeasible
warrant.” See §1. I discuss the relevant notions of rational requirement and permission
in some detail there.
4
  See e.g. Gibbard (2003: esp. Introduction and pp. 152–8 of ch. 7). For a different
development of the idea, see Wedgwood (2007: esp. chs 1, 4, and 7). For resistance, and
many more references, see Schroeter (2005). I use small caps for concepts throughout.
5
  For the qualification about defeasibility, see Boghossian (2003: 281).
 Vindicating Practical Norms 47

willingness is a dispositional matter: willingness to infer in accord with N is

being disposed to so infer, in conditions in which N applies.6 Boghossian’s
central example is the concept conditional and inferences from premises
of the form p, p→q to q (MPP-ish inferences). For Boghossian, a disposi-
tion to engage in MPP-ish inferences (in relevant conditions) is “constitu-
tive” of possessing conditional, in the sense of being both necessary and
sufficient for possessing it, and of being that in virtue of which one possesses
it.7 The exact shape of the requisite MPP-ish inferential disposition depends
on the specific inferential role that conditional must play in the mind of
any thinker possessing it. For instance, it may be no part of the requisite dis-
position that one be disposed in all circumstances to explicitly draw out all
of the consequences of one’s beliefs of the form p, p→q; it may be enough
that one be disposed to do so when prompted in certain ways.8
MEC claims that inferential transitions that are in this way “built into”
the possession conditions of a concept are thereby “entitling.” The idea of
an “entitling” inferential transition is, in the epistemic context, the idea of
a type of inference such that, if you infer this way, then you are genuinely
entitled to the beliefs you end up with by so inferring. There is room for
nuance here: some “entitling” inferences may merely transfer entitlement,
from independently justified premise-beliefs to conclusions drawn on their
basis. Here, one’s conclusion is inferentially justified only given independ-
ent justification for the premises. Other types of “entitling” inference may
generate entitlement, even without independent justification for the prem-
ises. For example, it may be that a surface’s seeming red to one is not itself a
case of warranted belief (or of warranted anything), but that such perceptual
seemings nonetheless put one in a position to move to an (at least defeasibly)
warranted belief that the surface in question is red. In a loose sense of “infer-
ence” in which any process of updating belief that is subject to epistemic
norms is a type of “inference,” updating one’s beliefs to match one’s percep-
tual appearances is, intuitively, an entitlement-generating type of inference.
(It is a further question whether its being an entitlement-generating type of

6
  See e.g. 2003: 279–80; 2001: 258.
7
 2003:  279; 2001:  258. Boghossian thinks that other logical connectives are
analogous.
8
  Further, the disposition may be sensitive to conditions, if any, in which a conclusion
of the form q does not follow from premises of the form p, p→q. So one cannot dis-
miss Boghossian’s claim about the conditions of possessing conditional by pointing to
the possibility that someone like Vann McGee might sincerely deny the general validity
MPP as a truth about implication. (See Williamson 2003: 251–2 on this as a proposed
counterexample.) I will continue to ignore the distinction between truths about implica-
tion and rules of inference, assuming with the literature that, while the two are distinct,
we can easily formulate inference rules that roughly correspond to truths about logical
implication.
48 Hille Paakkunainen

inference is explained by facts about what it takes to possess certain con-

cepts, such as _is red.)
MEC is formulated in terms of “entitlement,” which sounds like a relatively
weak normative notion: a permission or perhaps a recommendation, but at
any rate not a requirement. However, when the metasemantic idea is formu-
lated in terms of warrant-transfer instead of “entitlement,” we can start to
see how concept-constituting inferential transitions, if warrant-transferring,
might also yield requirements to believe. Suppose I  have ironclad proof
that p and that p→q. Then if MPP is warrant-transferring, I also have very
strong warrant for the conclusion that q; and if there is no independent
evidence that not-q, this warrant might have the strength of a require-
ment to believe.9 In the practical case, one might analogously attempt to
show that certain principles are reasons-transferring because dispositions to
infer in accord with them are concept-constituting. For example, the fol-
lowing seems like a fairly plausible reasons-transfer principle for updating
intentions:
RM/RE  Given (some, or decisive) object-given reasons to intend an
end, one also has (some, or decisive) object-given reasons to intend to
take the necessary means to that end.10
In contrast, the following seems quite implausible as a general reasons-
transfer principle:
RS/RE  Given (some, or decisive) reasons to intend an end, one also
has (some, or decisive) reasons to intend to stop others from pursuing
their ends.
Applied to this case, the practical metasemanticist’s claim would be that we
can explain why RM/RE is a genuine reasons-transfer principle by appeal
to the conditions of concept-possession. If possessing the concept end, and/
or necessary means requires being disposed to update intentions in accord
with RM/RE, then RM/RE is thereby reasons-transferring. And depending
on the strength of reasons transferred, RM/RE might help yield practical
9
 Cf. Horwich (2005:  157, 163)  and Schechter and Enoch (2006:  §4.3), who
complain that metasemantic strategies at best account for epistemic entitlements, not
obligations.
10
  RM/RE is restricted to “object-given” reasons so as to rule out cases in which one’s rea-
sons to intend an end aren’t also reasons to pursue that end, as in Kavka’s (1983) toxin case.
Object-given reasons are reasons stemming from what would be the case if one actually
performed the action that is the object of one’s intention. In contrast, “state-given” reasons
to intend are reasons stemming from the good consequences of having the intention-state.
In Kavka’s puzzle, a demon promises to give me $1,000,000 if, at noon, I intend to drink
a lethal toxin at 3pm. Here I have state-given reason to intend to drink the toxin, but no
object-given reason to so intend. (I will have won the money before 3pm.) On state-given
v. object-given reasons, see e.g. Parfit (2011: i. 50) and Schroeder (2012).
 Vindicating Practical Norms 49

obligations as well as permissions or recommendations. In contrast, the

thought would be, RS/RE has no analogous claim to legitimacy.
How does all this connect to our present concern with “rationality”?
Although Boghossian doesn’t put it this way, it seems natural to think of dis-
positions to infer in warrant-transferring ways as facets of epistemic ration-
ality: the rationality characteristic of psychologies that function well in the
pursuit of truth, knowledge, and/or justified belief.11 Practical rationality
can likewise be thought of as a characteristic of a kind of well-functioning
psychology. “Practical rationality” is of course a term of art; and it would
be contentious to say that practical rationality involves the pursuit of some
goal analogous to e.g. knowledge. (Say, the good.) Still, it seems harmless
to suppose that, just as updating beliefs in accord with warrant-transferring
principles is a facet of epistemic rationality, updating intentions in accord
with reasons-transfer principles might be a facet of practical rationality.12
And plausibly there are other facets, too. The most uncontroversial examples
of practical rationality are examples of consistency among attitudes, where
such consistency has, at least prima facie, nothing to do with what norma-
tive reasons one has for what. For example, it seems irrational to simultan­
eously have inconsistent intentions, such as the intention to Φ at time t and
the intention to not Φ at time t—and this regardless of whether one has any
reason for or against either of these intentions. Such irrationality is a viola-
tion of a so-called “wide-scope” rational requirement, of the form
Rationality requires that [If you X then you Y],
where “X” and “Y” stand for either having or lacking an attitude like inten-
tion, belief, or desire, at some time or times.13 Some philosophers have
defended wide-scope enkratic requirements, and wide-scope means/ends
requirements:

11
  I do not presuppose that any particular goal is the primary epistemic goal. The state-
ment in the text is just to give a very general characterization of one worthwhile topic one
might mean to refer to by “epistemic rationality.”
12
  It is of course a difficult question what it is to be disposed to update intentions
“in accord” with such a principle. Does it involve the disposition to believe that one has
certain kinds of reasons to intend, when one has them? Or is it the disposition to intend
the necessary means to one’s ends, indiscriminately? I put these questions aside here. As
I explain below, my focus will be on rational permissions and requirements of a differ-
ent sort, not reasons-transfer principles. I raise reasons-transfer principles here because
they are the closest analogue to Boghossian’s concern with warrant-transfer, and are thus
helpful for starting to see how the metasemantic idea might apply in the practical case.
13
  If the times of both X and Y are the same, the requirement is synchronic. If the
times are different, the requirement is diachronic. See Kolodny (2007: 373) for a detailed
example of how even a putative diachronic requirement might be formulated as having
wide scope.
50 Hille Paakkunainen

ENR(wide)  Rationality requires that [If you believe that you ought
to Φ, then you intend to Φ].14
M/E(wide)  Rationality requires that [If you intend to Ψ, and you
believe that you can Ψ only by Φ-ing, then you intend to Φ].
These requirements have “wide scope” in the sense that they govern combi-
nations of attitudes. Someone who believes that she ought to Φ can come
to conform to ENR(wide) either by adopting the intention or by giving up
the belief. In contrast, putative “narrow-scope” requirements of rationality,
of form “If you X, then you are rationally required to Y,” cannot be. Here
is a putative narrow-scope enkratic requirement, and a parallel means/ends
requirement:
ENR(narrow)  If you believe that you ought to Φ, then you are
rationally required to intend to Φ.
M/E(narrow)  If you intend to Ψ, and you believe that you can Ψ
only by Φ-ing, then you are rationally required to intend to Φ.
In addition, some ways of updating intentions might be rationally permit-
ted but not required. Here is a putative narrow-scope enkratic permission:
ENP(narrow)  If you believe that you ought to Φ, then you are
rationally permitted to intend to Φ.
Part of what is in question in the theory of practical rationality is what
the different facets of well-functioning practical psychologies are, and how
they relate to each other.15 The metasemantic idea is to try to derive at least
some of the content of a theory of practical rationality from premises about
what it takes to possess certain concepts. For instance, if a disposition to
update intentions in accord with ENR(narrow) is a condition of possessing
ought to Φ, then according to the metasemanticist, one is either genu-
inely permitted or required to update intentions in this way. This would
vindicate the disposition to follow ENR(narrow) as part of being practi-
cally rational.16 Our question is how the inference from premises about the

14
  See e.g. Broome (2007). Time constraints are relevant to more precisely formulat-
ing which requirement is at issue, but this will do for illustration.
15
  e.g. there is a lot of literature on whether rational requirements correspond to, or
provide, normative reasons. See e.g. Kolodny (2005), Broome (2008).
16
 Notice that talk of “updating intentions in accord with” ENR(narrow) is most
naturally heard as being about a diachronic process, and so the relevant version of
ENR(narrow) is most naturally read as a diachronic requirement. But the metaseman-
ticist might also try to vindicate ENR(narrow) read as a synchronic requirement, by
arguing that if the disposition to bring oneself to conform to this synchronic requirement
is constitutive of possessing ought to Φ, then conforming to the synchronic require-
ment is part of being practically rational. I will not track these different possibilities in
the text, and will often speak of “inferential dispositions” as concept-constituting, along
with the literature, even though this language may not easily fit the idea that dispositions
 Vindicating Practical Norms 51

conditions of concept-possession to claims about genuine permissions or

requirements is supposed to go.
I introduce these distinctions between types of principle here to clarify
the territory, and to give an initial sense of how the metasemantic idea might
apply in the practical case. For now, we can set these distinctions aside until
§4; §§2–3 examine extant metasemantic arguments, which make no use of
them. My focus will eventually be on rational permissions and requirements
that don’t assume that one has any normative reason for the attitudes that
the relevant permissions or requirements concern. I won’t defend the pos-
sibility of metasemantic vindications of reasons-transfer principles.
But why attempt metasemantic vindications of either epistemic or practi-
cal norms, whatever precise form such norms may take? The proof of the
pudding will be in the eating; and if a metasemantic vindication of some
norm works, this is in a way more interesting than any general motivation
we can give for attempting such vindications. Nonetheless, we can see why
the metasemantic idea should look initially attractive. For there seems to be
some important connection between conceptual competence and rational-
ity. For instance, it seems that rationality cannot require one to misuse a
concept.17 Further, it seems that it would have to be a misuse of a concept
to incorrectly follow a concept-constituting inferential rule.18 If so, then
rationality cannot require one to incorrectly follow a concept-constituting
inferential rule. Of course, it does not yet follow that rationality requires
or even permits one to correctly follow concept-constituting inferen-
tial rules. For it may be that rationality requires one to not follow some
concept-constituting inferential rule at all. But such a requirement to not
follow a concept-constituting rule would entail that merely possessing cer-
tain concepts involves one’s having an irrational disposition, in the sense of
a disposition to infer in ways in which one is rationally required not to infer.
The idea that merely possessing some concept is irrational, either epistemi-
cally or practically, is at least uncomfortable. A sound defense of MEC would
eliminate the discomfort, vindicating the intuitive thought that merely pos-
sessing a concept cannot, in and of itself, constitute irrationality.19

to conform to synchronic requirements might be concept-constituting. And while I for-

mulate my arguments in §§4–6 in terms that most naturally fit the idea of diachronic
requirements and permissions on updating intentions, I think the arguments work, with
minor modifications, for putative synchronic requirements and permissions as well.
17
  This seems to be so even for “gruesome” concepts that pick out gerrymandered cat-
egories. Rationality might require one to not use a gruesome concept, but this is different
from a requirement to misuse it. (See later in the chapter.)
18
  A “concept-constituting” inferential rule is just a rule that one must be disposed to
infer in accord with in order to possess some concept.
19
  Can’t there be “defective” concepts whose possession leads one to make bad infer-
ences—to update one’s attitudes in irrational ways? What about objectionable concepts
such as pejoratives? I address these topics in §§5–6. Cf. Boghossian (2003: 282–5).
52 Hille Paakkunainen

One might object that it is not uncomfortable to suppose that possess-

ing a concept might be irrational if we accept the inferentialist thesis that
concept-possession is a matter of having inferential dispositions. For infer-
ential dispositions clearly can be irrational.20 However, the present point is
that, when considered independently from the inferentialist thesis, the idea
that merely possessing a concept might constitute irrationality seems quite
unintuitive. If inferentialists are stuck thinking that some concepts can’t be
possessed without being irrational, this seems like a problem for inferential-
ism. A sound defense of MEC would eliminate the problem. My aim here
is to see how MEC might be defended, given suitable inferentialist premises
about concept-possession.
Of course, another alternative is to give up inferentialist commitments
about concept-possession. Although some philosophers have defended
inferentialist views about the conditions of possessing e.g. ought to Φ,21
I personally doubt whether any inferentialist premise of the requisite sort
is true. But if my arguments here stand, then in order to resist the meta­
semantic strategy, it is such premises about concept-possession that we must
attack. The connection between concept-possession and rationality is in
good order.
§2 examines Boghossian’s argument for MEC, arguing that it fails to
deliver the goods. §3 examines a different argument by Ralph Wedgwood
(2007). Wedgwood’s argument gets at something important, but falls short
in crucial ways. §4 then mounts a better argument for MEC—or better,
for a practical analogue regarding rational permissions to update inten-
tions in concept-constituting ways. §5 considers objections, and §6 consid-
ers whether we could derive not just rational permissions but also rational
requirements from premises about concept-possession.

2.  Boghossian on MEC: The Argument from

Blamelessness22

Boghossian asks:  Why do some inferential transitions transfer epistemic

warrant while others don’t? One answer is “simple externalism,” on which
an inference transfers warrant if and only if (a) it is truth-preserving, and
(b) the subject is justified in believing its premises, independently of her

20
  Thanks to an anonymous reviewer for putting this objection to me.
21
 See n. 4.
22
  Much of the discussion in this section is in terms of epistemic warrant-transfer since
this is Boghossian’s focus. But I draw connections to the case of rational permissions or
requirements for updating intentions where appropriate.
 Vindicating Practical Norms 53

justification for believing the conclusion.23 However, Boghossian argues,

(a) and (b) are insufficient for warrant-transfer. An inference can be
truth-preserving and made on the basis of independently justified belief in
its premises, yet make an irresponsible leap that fails to warrant the thinker
in believing the conclusion. (Boghossian’s example is an irresponsible but
truth-preserving leap from justified belief in a claim of form “x, y, z, and n
are whole numbers and n is greater than 2,” to belief in a claim of form “xn +
yn is not equal to zn.”)24 For Boghossian, the lesson is that inferences must be
epistemically responsible or “blameless” to be warrant-transferring.25 What
is blameless inference?
One proposal is “simple internalism”:  a thinker infers blamelessly if
and only if she is in a position to “reflectively know,” or at least justifiably
believe, that her premises do warrant believing her conclusion.26 However,
Boghossian argues that this condition is too intellectually demanding. On
simple internalism, a child could not gain any inferentially justified beliefs
via seemingly simple MPP-ish inferences without being antecedently jus-
tified in beliefs to the effect that her premises warrant her conclusion.27
We need a different explanation of blameless inference. Here is where
concept-constitution enters into the picture for Boghossian:
Suppose it’s true that my taking A to be a warrant for believing B is constitutive of
my being able to have B-thoughts (or A-thoughts, or both, it doesn’t matter) in the
first place [since constitutive of my possessing the concepts that figure in either A or
B, or both]. Then doesn’t it follow that I could not have been epistemically blame-
worthy in taking A to be a reason for believing B, even in the absence of any reason
for taking A to be a reason for believing B? [. . .] If inferring from A to B is required,
if I am to be able to think the ingredient propositions, then it looks as though so
inferring cannot be held against me, even if the inference is [one which I have no
further reason to consider correct, or entitling]. (2003: 279)
Boghossian concludes that concept-constituting inferential transitions are
“blind yet blameless,” in the sense of being “entitling without our know-
ing, or being able to know, anything about them” (2003: 276–7, 279). It
is precisely this “meaning-based explanation of blameless blindness” that
Boghossian claims MEC states (2003: 280). This is Boghossian’s argument
for MEC.

23
 2003: 268.
24
 2003: 268–70.
25
  Boghossian (2003: 270).
26
 Boghossian (2003:  270). I  have modified Boghossian’s statement of the simple
internalist view slightly.
27
 Boghossian further presses Carroll-style regress worries against simple
internalism: 2003: 273–6.
54 Hille Paakkunainen

As these last remarks suggest, Boghossian seems to ultimately identify

blamelessness with being entitling. Indeed, as stated, MEC entails that
concept-constitution alone is sufficient for (defeasible) entitlement. If being
warrant-transferring is a way of being entitling, then it is not clear whether
truth-preservingness is even necessary for warrant-transfer on Boghossian’s
view.28 I  come back to truth-preservation below. But what to make of
Boghossian’s appeal to blamelessness?
Blameless inference is epistemically responsible inference (Boghossian
2003:  270). And epistemic responsibility seems to be a matter of
non-culpability:  roughly, one does as well as one could reasonably be
expected to do under the circumstances.29 One does not, for example,
reason carelessly or recklessly, engage in wishful thinking, or self-servingly
disregard evidence that stares one in the face. Now, it is plausible to think
that individual instances of inference must be blameless in this sense
in order to transfer warrant. The trouble is that this is not sufficient for
warrant-transfer. For one can infer blamelessly in accord with norms that
are not warrant-transferring, as the following example from James Pryor
illustrates:
Imagine a hapless subject who is taught bad epistemic standards. For instance, sup-
pose he makes mistakes when engaging in tricky statistical reasoning. Let’s say the
standards he’s been taught fail to distinguish between the likelihood that a test will
yield a false negative and the likelihood that the test will yield a false positive. Our
hapless subject does his best to apply these standards, but unfortunately he is not
intellectually capable of discerning their defects. Furthermore, he has lost all mem-
ory of his childhood and so doesn’t recall how it was he first acquired these stand-
ards. Cases of this sort naturally prompt two judgments. First, the subject’s beliefs
are epistemically defective in some important way. But second, the subject cannot
be held culpable for those defects. (2001: 114)
Pryor’s “hapless subject” might reason very carefully and responsibly in
accord with the bad rules he was taught. He may also have blamelessly
acquired and retained those rules. Yet he reasons with “bad epistemic stand-
ards.” Surely warrant-transfer has something to do with the type of inference
one makes, where a “type” of inference is individuated by the inference rule
that one follows. At any rate, Boghossian’s own aim is to justify particular
rules as epistemically good or legitimate ones to follow. But since one can

28
  Though at 2003: 278, Boghossian explicitly leaves it open that truth-preservation
may be necessary for warrant-transfer. Notice that MEC is also compatible with thinking
that only truth-preserving inferences could be concept-constituting. Truth-preservation
might thus come back through the back door to constrain epistemic entitlement. This is
Peacocke’s (1993) view; but Boghossian rejects it at 2003: 281.
29
  For blamelessness as “non-culpability” and as meeting reasonable expectations, see
Pryor (2001: 114–15).
 Vindicating Practical Norms 55

blamelessly reason in accord with illegitimate rules, as well as blamelessly

acquire and retain the disposition to do so, the appeal to blamelessness
is too general to separate legitimate rules from the rest.30 (We can con-
struct a practical parallel:  a practically hapless but blameless subject who
has been subjected to strict teaching that one is required to update one’s
intentions to reflect not one’s own judgments about what one ought to do,
but instead one’s judgments about what one’s mother would wish one to
do. This putative requirement may be blamelessly internalized and care-
fully employed by the hapless subject, but this does not make it a genuine
rational requirement.)
If this is right, then the appeal to blamelessness is insufficient to show
a particular type of inference to be warrant-transferring. But this suggests
a response on behalf of Boghossian. What is wrong with Pryor’s hap-
less but blameless subject is that his inferences fail to preserve truth, or
probability, in the right ways. Perhaps truth-preservation is after all nec-
essary for warrant-transfer; and together with blamelessness, it might
be sufficient. If concept-constituting inferential transitions are blame-
less because concept-constituting, as Boghossian claims, then such tran-
sitions, if truth-preserving, would transfer warrant. And their being
concept-constituting would help to explain why they transfer warrant, if the
property of being concept-constituting ensures the blamelessness of such
transitions.31
However, we need to be careful about what the relevant claim about
concept-constitution is. It matters that what is constitutive of possess-
ing a concept is not any particular instance of a type of inference. This
would be far too implausible as an account of the conditions of possessing
some concept; nor does Boghossian hold it. Nor does the type of infer-
ential transition somehow itself constitute a subject’s grasp of a concept.
(It is unclear what this would mean.) Rather, what is supposed to be
concept-constituting is the disposition to engage in inferences of a certain
type. If being concept-constituting entails blamelessness, then anyone can

30
  Likewise it is possible to reason irresponsibly or carelessly in accord with “good”
standards. For instance, one might reason carelessly in accord with MPP, from premises
of the form p, p→q to q*, where q* is very close to q but different enough to not be
supported by p and p→q. It is of course a difficult question what distinguishes incorrect
reasoning in accord with MPP, caused by e.g. recklessness or inattention to the task, from
correct reasoning in accord with some subtly different rule, MPP*. Still, there seems to
be such a distinction.
31
  It is somewhat harder to see what the analogy to truth-preservation might be in
the practical case. One answer is: truth-preservation. If intentions constitutively involve
beliefs, then requirements for updating intentions might be, at bottom, requirements
governing belief. This is the view of “cognitivists” about practical reason; see e.g. Ross
(2009), Setiya (2007).
56 Hille Paakkunainen

blamelessly have a disposition to engage in N-type inferences if that disposi-

tion is constitutive of possessing some concept C. But it does not yet follow
that any instance of actually performing an inference of type N is blame-
less—not even if it is possible to reason responsibly in accord with N, in
the sense of doing so with due care and attention. For it is possible to have
dispositions that one ought to mask, subvert, or even attempt to eradicate,
perhaps precisely because they are irrational ones.32 If the expectation to
mask the disposition is reasonable, then manifesting the disposition might
be culpable even if having or acquiring the disposition is not.
We thus lack an explanation of why the fact that a particular disposi-
tion is concept-constituting and therefore blameless should entail that
manifesting the disposition is itself blameless, in any sense corresponding
to warrant-transfer. After all, perhaps there are some concepts that we
should not employ, even if we are blameless in having them, and even if
it is possible to employ them carefully and in this sense responsibly. This
is, in a way, just the problem we started with: the problem of explaining
why there can’t be some concept-constituting inferential dispositions that
one should not manifest at all. (Or that one is rationally required not to
manifest.)
I have not shown that there is no hope for the argument from blameless-
ness. But I cannot see how to improve it; nor do I see how an appeal to
blamelessness might fare better in the practical case. So let us try a different
tack. A  more promising line of argument is opened up by reflection on
some of Ralph Wedgwood’s work.

3.  Wedgwood’s Perfectly Rational Being

Wedgwood (2007) gives a general argument that concept-constituting

inferential dispositions must be rational ones, in the sense of being disposi-
tions to engage in inferences that are themselves rational, or in accord with
“principles of rationality.”33 Since Wedgwood understands the rationality
of a disposition by reference to its being a disposition to infer in accord
with “principles of rationality,” there is no danger that the rationality of
concept-constituting inferential dispositions might here drift apart from the
rationality of individual inferences that manifest those dispositions. So this
is a good starting point.
Now, Wedgwood’s own ambition is not to give metasemantic vindica-
tions of specific epistemic or practical norms. His interest is rather in

32
  Cf. Williamson (2003: 255–6).
33
  For Wedgwood on “rational disposition,” see 2007: 161–5.
 Vindicating Practical Norms 57

giving a general argument that thinkers possess concepts in virtue of the

inferential dispositions they have, and that, whatever these dispositions
are specifically, they must be rational ones. This is all in aid of show-
ing that facts about thinkers’ having concepts, and about having atti-
tudes employing these concepts, cannot be stated “without mentioning
normative properties and relations” (172); and that consequently “the
normative is at least as metaphysically fundamental as the intentional”
(175). I won’t address the details of Wedgwood’s larger project here. My
interest is in Wedgwood’s argument for the claim that, whatever the
concept-constituting inferential dispositions are, they must be rational
ones. If Wedgwood’s argument for this claim works, the metasemanticist
can co-opt it for her own purposes.34
For example, if we did have good reason, independent of the assump-
tion that enkratic inferences are rational ones, for thinking that a disposi-
tion to engage in enkratic inferences is constitutive of possessing ought to
Φ, then we could plug in Wedgwood’s argument that concept-constituting
inferential dispositions must be rational ones, and thereby explain, without
circularity, why the disposition to engage in enkratic inferences is itself a
rational one.35 Of course, calling the enkratic disposition a “rational” one
leaves it thus far unspecified whether updating one’s intentions enkratically
is rationally required or merely rationally permitted. But we would have at
least made some progress.

34
 MEC looks to state a metaphysical dependence claim, to the effect that if
there are any inferential transitions that are concept-constituting, then those transi-
tions are entitling and not just brutely so:  rather, they are entitling in virtue of being
concept-constituting. (Cf. Boghossian 2003: 278.) Thus understood, the fundamental
task for an argument for MEC is to spell out how this metaphysical dependence of the
normative property of entitlement on the seemingly non-normative property of being
concept-constituting goes. However, if Wedgwood’s larger project were sound, would
this mean that there is no metaphysical dependence of the normative on facts about
concept-possession? No: Wedgwood’s claim is that there is interdependence (175). Such
interdependence is consistent with the thought that particular dispositions are rational
because of being concept-constituting. It is just that this would not be a reduction of these
dispositions’ property of being rational to their property of being concept-constituting,
since one would in turn need to appeal to the dispositions’ rationality in order to explain
their being concept-constituting; and such circular explanations are not reductive. For
Wedgwood on “reduction,” and on why not all constitutive accounts that are genuinely
explanatory of the nature of a thing need to be reductive, see 2007: 136–7. I continue to
assume in the text that, unlike Wedgwood, the metasemanticist is interested in giving a
non-circular explanation of the rationality of concept-constituting dispositions.
35
  What might an explanation of the concept-constitutingness of a particular disposi-
tion look like, if it did not make any assumptions about the rationality of the disposition?
See e.g. Gibbard (2003).
58 Hille Paakkunainen

So what is Wedgwood’s argument that concept-constituting inferential

dispositions must be rational ones? His argument appeals to the idea of a
perfectly rational being:
On the one hand, it seems very hard to see how any thinker could possess a con-
cept without having any disposition to use the concept in any rational way:  even
extremely irrational thinkers, who are constantly committing such fallacies as affirm-
ing the consequent or denying the antecedent, are generally disposed also to accept
certain basic rational inferences, such as inferences that have the form of modus pon-
ens. On the other hand, it seems quite easy to see how this concept could be possessed
by a thinker who had no disposition to use the concept in any irrational way. After all,
it seems that any concepts that you have could be shared by a perfectly rational being
who had no irrational dispositions at all. (For example, the perfectly rational being
would need to possess these concepts in order to ascribe attitudes to you accurately,
and to diagnose the various confusions and irrationalities that mar your thinking.) So
that perfectly rational being would have to possess those concepts in virtue of some
of her rational dispositions (since she has no irrational dispositions). This makes it
plausible that we also possess those concepts in virtue of some of our rational disposi-
tions, and not in virtue of our irrational dispositions. (2007: 168)
As stated, this argument does make assumptions about the rationality of
specific inferential dispositions, but only for illustrative purposes.36 The
basic idea of the argument can be stated without these assumptions. Thus
the metasemanticist can make use of the argument without worrying that,
in doing so, she engages in circular reasoning, explaining the rationality of
a specific concept-constituting disposition in part by appeal to the assump-
tion that it is indeed a rational one. At bottom, Wedgwood’s argument
seems to be just that, since a perfectly rational being could possess any
concept that we less than perfectly rational beings possess, no concept can
have among its possession-conditions an irrational disposition. For per-
fectly rational beings do not, by definition, have any irrational dispositions.
What to make of the argument? Wedgwood does not give us much reason
to think that perfectly rational beings must be able to possess any concept that
we possess. He claims that the perfectly rational being would “need to possess
these concepts in order to ascribe attitudes to you accurately,” as well as to “diag-
nose” various ills about the psychologies of those less rational. But why must a
perfectly rational being be able to accurately ascribe attitudes to her less rational
brethren, or to diagnose their psychological ills? One could see why these abili-
ties are required if the perfectly rational being had to be omniscient. But it is
not at all clear why omniscience should follow from being perfectly rational.

36
  Wedgwood also assumes that a disposition to update one’s intentions enkratically
is a rational disposition (2007: 25). And he says that the enkratic disposition “might” be
essential to possessing the concept ought to Φ (2007: 170).
 Vindicating Practical Norms 59

In response, one might suggest that it is not that the perfectly rational
being must be omniscient, but rather just that it is odd to think that perfect
rationality could, in and of itself, bar a thinker from knowing certain facts.
In particular, if perfect rationality consists, in part, in having all the epis-
temically rational dispositions, then shouldn’t such thinkers be in a better
position to know any fact than would be an imperfectly rational thinker
who has some epistemically irrational dispositions? Yet the supposition that
a concept C requires an epistemically irrational disposition to possess it
would make facts knowable only by means of an employment of C—if
there are any—unknowable for perfectly rational epistemic agents.
Perhaps something similar can be said to explain why concept-constituting
inferential dispositions cannot be practically irrational. Since a perfectly
rational being lacks all practically irrational dispositions, she lacks any con-
cept C that it takes a practically irrational disposition to possess. But then
if there are any facts that are knowable only by means of an employment
of such a concept C—a big if, to be sure—then the perfectly rational being
cannot know those facts, either. Her practical rationality deprives her of
the possibility of knowing certain facts—in this case, facts that a practically
irrational person could presumably know, if she were at least epistemically
rational. And this, again, would be rather odd.
I think this argument may be getting at something right. But it is unsat-
isfying as it stands. Why think that, for any concept C that it takes an irra-
tional disposition to possess, there might be some facts that are knowable
only through an application of C? Further, why should perfectly rational
beings, just as such, be in a position to know every fact, even if they have all
the epistemically rational dispositions? Knowing some facts might require
too long a process of reasoning for the knowledge to be achievable in the
perfectly rational being’s lifetime.37 Knowing others might require ability to
fly like a hawk. (“This is what it is like to fly like a hawk.”) Being perfectly
rational does not guarantee having all the resources and abilities required to
know all the facts. Why then should it guarantee access to all the concepts?
After all, it takes a contingent sensory modality to have e.g. the demonstra-
tive concept this shade, for some particular shade of red in front of one.
Why should the nature of being perfectly rational guarantee access to such
contingent conceptual resources?

37
  Cf. McGrath (2010). Unless, of course, we assume that the perfectly rational being
also lives for an infinitely long time, or perhaps is outside of time altogether. But again, it
is unclear why we should assume this. Wedgwood does not claim that he is appealing to
the notion of some sort of a supernatural deity. Indeed, such an appeal would introduce
its own problems. Why think that such a deity needs dispositions of reasoning at all,
instead of knowing everything intuitively?
60 Hille Paakkunainen

One might retort that it is one thing to say that some perfectly rational
being might contingently lack the resources to know certain facts. It is another
to say that being perfectly rational would in and of itself necessitate lacking
certain concepts, and thus would necessitate lacking the knowledge of facts,
if any, that are knowable only by means of those concepts. And the latter is
what is at issue if some concepts can only be possessed by having irrational
dispositions. But it is still unclear why it should be worrisome if some facts
are necessarily unknowable for a perfectly rational being, if it is not worri-
some that some facts are contingently unknowable for such a being.
Wedgwood does make a further remark that might seem helpful. He says
that possessing a concept is a “cognitive power or ability—not a cognitive
defect or liability” (2007: 169). I think there is something right about this.
But the remark is cryptic. Wedgwood does not explain why, even if having
a concept is a cognitive “power or ability” along one dimension, it cannot be
a cognitive defect or liability along another. Suppose that a concept makes it
possible for one to entertain new thoughts one could not have entertained
before. Suppose the concept even makes it possible for one to know new
facts. We have still not explained why it cannot at the same time be a liabil-
ity along another dimension, in that it disposes one to reason badly.
I think Wedgwood’s argument is getting at something important. But
if the foregoing is right, it does not quite show that concept-constituting
inferential dispositions must be rational ones. Recall, too, that it does not
tell us whether the “rational” dispositions at issue are dispositions to infer
in ways that are rationally required, or in ways that are merely rationally
permitted. The arguments of §§4–6 hope to do better.

4. Perfect Rationality and

Non-Accidentality

Let us adopt a notion of “rational dispositions” of thought as dispositions

to think in rational ways, or to have rational combinations of attitudes.
“Ways of thinking” here covers e.g. ways of updating intentions and ways
of updating beliefs. Suppose we want to show that updating one’s inten-
tions enkratically is “rational,” and so that the disposition to so update
intentions is also “rational.” We need to specify which dispositions we have
in mind, exactly; and relatedly, what we mean by “rational.” Do we mean
to show that one is rationally required to match one’s intentions to one’s
ought-beliefs? Or merely that one is rationally permitted to do so? Further,
are the relevant requirements or permissions of wide or narrow scope? In
this section, I argue that both wide and narrow-scope permissions to update
one’s intentions enkratically can be derived from suitable assumptions about
 Vindicating Practical Norms 61

concept-constitution: in this case, from the assumption that a disposition to

update one’s intentions in accord with the relevant permission is constitu-
tive of possessing ought to Φ.38 §6 considers how the argument might be
expanded to cover requirements as well.
Consider the following putative enkratic permissions:
ENP(narrow)  If you believe that you ought to Φ, then you are
rationally permitted to intend to Φ.
ENP(wide)  Rationality permits that [If you believe that you ought
to Φ, then you intend to Φ].
I will run the argument in terms of ENP(narrow), but consider its applica-
tion to ENP(wide) and other rational permissions as we proceed.
Suppose the metasemanticist thinks that the disposition to update
one’s intentions in accord with ENP(narrow)—call this disposition
D[ENP(narrow)]—is constitutive of possessing ought to Φ.39 And she wants
to say that, because of this, ENP(narrow) states a genuine rational permis-
sion: ENP(narrow) is true.40 How to get from the premise to the conclusion?
Suppose, for reductio, that ENP(narrow) is false, even though the disposi-
tion to follow it (to update one’s intentions in accord with it) is constitutive
of possessing ought to Φ. If ENP(narrow) is false, then one is not permit-
ted to intend to Φ when one believes that one ought to Φ. But if one is not
permitted to intend to Φ when one believes that one ought to Φ, then it
must be that one is required to not intend to Φ when one believes that one
ought to Φ. For wherever there is no requirement to the contrary, there is a
permission: permissions come cheap. We can put the point by saying that if
ENP(narrow) is false, then an Opposing Rule, to the effect that when one
believes that one ought to Φ, one is rationally required not to intend to
Φ, is true. (More generally, this will be so for any rational permissions. If a
putative permission is false, this is because an opposing requirement is true.
For example, if ENP(wide) is false, it must be that one is rationally required
to not have the combination of attitudes it involves.)
So far, so good. Whence the reductio? The basic idea is this. If ENP(narrow)
is false though concept-constituting, while the Opposing Rule is true, then

38
  Recall n.  16:  I  formulate my argument in a way that is most naturally read as
concerning diachronic enkratic permissions. But I think essentially the same argument
would work, with minor modifications, for synchronic enkratic permissions.
39
  We can bracket for now the question what such a disposition would look like, and
how it would differ from e.g. a disposition to update one’s intentions in accord with
ENR(narrow) (from §1). I address this briefly in §6.
40
 I am assuming that the metasemanticist will not want to argue from the
concept-constituting status of D[ENP(narrow)] to the rational permissibility of some
other way of updating intentions.
62 Hille Paakkunainen

the following principles, No Accident, and No Irrationality, can’t be true

together.
No Accident  It is part of being perfectly rational to non-accidentally
comply with all of the rational requirements that apply to one.
No Irrationality  It is part of being perfectly rational to have no
irrational dispositions—no dispositions to update one’s attitudes in
irrational ways.
But both No Accident and No Irrationality are true. Whatever else is true
of the nature of perfectly rational agents, they surely lack irrational disposi-
tions, and their rationality equips them to non-accidentally comply with
the rational requirements that apply to them. So ENP(narrow) can’t be false
though concept-constituting. This is how we can derive the conclusion that,
if ENP(narrow) is concept-constituting, then it is true. (Mutatis mutandis
for other rational permissions.)
That is the basic idea. So why think that if ENP(narrow) is false but
concept-constituting, and the Opposing Rule is true, then No Accident and
No Irrationality can’t be true together?
First consider what the perfectly rational being’s non-accidental compli-
ance with rational requirements must be like. The compliance is not ran-
dom: rather, there is a tendency to comply. Moreover, the tendency itself
is not merely a statistical tendency with no explanation. Rather, it seems
that the tendency of a perfectly rational being’s intentions to be as they are
required to be must be the manifestation of a rational disposition. Which
rational disposition? The obvious answer is:  the disposition to update
one’s intentions in accord with whatever the requirement in question is.
In the case of the Opposing Rule requirement, this would be the disposi-
tion to update one’s intentions in accord with the Opposing Rule: call this
disposition D[OR].
It is of course a difficult question what exactly a disposition to update
one’s intentions in accord with a rule’s dictates is, apart from external con-
formity with the rule’s dictates. But an outline answer is that the rule must
somehow structure one’s thinking in order for one to be disposed in the
relevant way. How, then, does the Opposing Rule structure one’s think-
ing when one is disposed to update one’s intentions in accord with it? The
Opposing Rule says that when one believes that one ought to Φ, one is
rationally required to not intend to Φ. In order for the Opposing Rule to
structure one’s thinking, it is not enough that one just ends up, never mind
how, being in the output states that the Opposing Rule would require one
to be in, were one to have ought-beliefs. We don’t attribute the disposition
to comply with the Opposing Rule to people merely on the basis of occa-
sions on which they fail to have an intention to Φ. (Notice that since the
 Vindicating Practical Norms 63

Opposing Rule is schematic, Φ-ing could be anything. And since there are
lots of intentions we all constantly fail to have, attributing the Opposing
Rule disposition to people would be far too easy.) Instead, to attribute the
disposition, we need some evidence that, even given beliefs to the effect
that one ought to Φ, for any Φ, one still tends to lack the intention to
Φ.41 Compare: ending up with beliefs that q is no evidence on its own of
a disposition to engage in MPP-ish reasoning. What is evidence of such a
disposition are the output beliefs one ends up with when one has the input
beliefs p, p→q.
Of course, it is possible for dispositions to be masked, so that their mani-
festation is prevented. Certainly we shouldn’t count each failure to be in the
output state given the input state as evidence against having the disposition.
It may even be that some dispositions can always be masked, so that they are
never manifested; though this is more contentious. But when a disposition
is masked, its input states don’t produce the output state they would regu-
larly—if unmasked—produce. This doesn’t get us away from the fact that,
to attribute a disposition to a person, we must be able to attribute to her the
relevant input states. In particular, to attribute the D[OR] disposition to an
agent, that agent must have had, and be in a position to have again, beliefs
to the effect that she ought to Φ.
But now consider what having, or being in a position to have, the input
beliefs requires. In the case of the disposition to engage in MPP-ish reason-
ing, having the input beliefs requires having the concept conditional. In
the case of D[OR], having the input beliefs requires having ought to Φ.
So having D[OR] requires having ought to Φ.42
We are now ready to complete our reductio. Recall the metasemanti-
cist’s starting assumption that to possess ought to Φ, one must have the
disposition D[ENP(narrow)]. Under our assumptions, this means that in
order to possess ought to Φ, one must have an irrational disposition. Yet

41
  Notice that the Opposing Rule is, in effect, a rule prescribing akrasia.
42
  One might object that it is possible to lose the concept ought to Φ in between
having instances of belief that one ought to Φ; and yet retain the disposition D[OR]
all along. If so, then one can have D[OR] while lacking ought to Φ. But this seems
extremely implausible. In any case, if one does lose the concept, then one is not at that
moment in a position to have the input beliefs, for having those beliefs requires having
the concept. So it is unclear why we shouldn’t say that losing the concept means losing
the disposition whose input beliefs are in question. (Notice that the reverse needn’t be
true: losing D[OR] needn’t mean losing ought to Φ. For there might be other disposi-
tions that allow one to have ought to Φ, or perhaps one can have ought to Φ without
having any particular disposition at all. It is only if a disposition is constitutive of hav-
ing ought to Φ, as we are supposing D[ENP(narrow)] to be, that losing the disposi-
tion would entail losing the concept. The present point, however, is just that losing the
concept would mean losing all dispositions whose input states require one to have the
concept.)
64 Hille Paakkunainen

possessing ought to Φ is a condition of non-accidentally complying with

the Opposing Rule requirement, for it is a condition of having D[OR].
So to be perfectly rational—to comply non-accidentally with one’s rational
requirements—one must have ought to Φ; but this entails having an irra-
tional disposition. The result is this: If ENP(narrow) is false but constitu-
tive of having ought to Φ, then the truth of No Accident would require
the falsity of No Irrationality. But both No Accident and No Irrationality
are true. So it can’t be that ENP(narrow) is false but constitutive of having
ought to Φ. This is how we can derive the conclusion that, if ENP(narrow)
is concept-constituting, then it is true.
The argument is general. It applies not just to ENP(narrow) and the con-
cept ought to Φ, but to any rational permission, whether of wide or nar-
row scope, such that the disposition to update intentions in accord with it is
constitutive of possessing some concept C. For any rational permission P we
suppose to be false, there is going to be a matching opposing requirement,
OR. And to comply non-accidentally with OR, one must be in a position
to have the very attitudes that OR concerns. Those attitudes, in turn, must
be the very same attitudes that P concerns. (Otherwise OR wouldn’t require
one not to have the combination or sequence of attitudes that P would per-
mit: OR wouldn’t really be an opposing requirement.43) And if the disposi-
tion to comply with P is constitutive of having C, it must be that the input
and/or output attitudes relevant to P are attitudes whose contents employ
C. (One couldn’t plausibly claim that e.g. D[ENP(narrow)] is constitutive
of possessing _is red.) Since the attitudes relevant to P must also be the
attitudes relevant to OR, it follows that the attitudes relevant to OR must
employ C as well. This is enough to get the argument going. The disposi-
tion to comply with OR, D[OR], would again require having C; but hav-
ing C would require having D[P]‌; and D[P] is supposed to be an irrational
disposition, a disposition to update intentions in ways that one is required
not to do.
If this is right, it follows that any permissive rules, whether of nar-
row or wide scope, must be true if concept-constituting. It remains to
be seen whether rules stating rational requirements must also be true if
concept-constituting. §6 considers how the argument might be expanded
to cover rational requirements as well. But first, §5 considers the most press-
ing objections to the foregoing.

43
  Although it might of course happen to issue requirements that come into practical
conflict with the original permissions.
 Vindicating Practical Norms 65

5.  Objections and Responses

Objection 1
The argument of §4 goes wrong in supposing that the Opposing Rule
must concern the very same combination or sequence of attitudes that the
would-be permission concerns. For example, we can oppose the permission
of ENP(narrow) by the simple requirement “Don’t intend to Φ!” There is
no need for an opposing rule of the form “If you believe that you ought
to Φ, then you are rationally required to not intend to Φ.” So one can
non-accidentally comply with a rule opposing the would-be permission,
without thereby having to have the concept ought to Φ.44

Response
If the putative requirement “Don’t intend to Φ!” is supposed to range over
all actions Φ, then it seems absurd. It would amount to a requirement to
not intend anything, not a requirement governing what to intend or how
to update one’s intentions. (In contrast, the Opposing Rule “If you believe
that you ought to Φ, then you are rationally required to not intend to Φ”
leaves one free to intend to do whatever one doesn’t believe one ought to
do. The rule merely constrains one’s intentions, telling one to give up inten-
tions to do things that one comes to believe one ought to do.) On the other
hand, suppose that “Φ” in “Don’t intend to Φ!” stands for a particular type
of action. It certainly seems possible that intending some types of action is
strictly prohibited, just as such. But what sort of prohibition is supposed to
be in question?
One option is that the proposed requirement states that one always has
decisive normative reasons to not intend to Φ; so that one ought not to so
intend. However, it is a substantive and disputed question whether rational
requirements governing combinations or sequences of attitudes correspond
in any clear way to normative reasons for action or intention.45 I mean to leave
open here the possibility that one might have decisive reasons not to intend
something that one is rationally permitted, or even rationally required, to
intend. To be relevant, then, the putative requirement “Don’t intend to Φ!”
had better be a putative rational requirement in this restricted sense, not
the claim that there are decisive normative reasons to not intend to Φ. But

44
 Thanks to Owen King and an anonymous reviewer for raising versions of this
objection.
45
 See n. 15.
66 Hille Paakkunainen

rational requirements seem generally to govern combinations or sequences

of attitudes, not single attitudes. If there are single attitudes that can be
irrational to have on their own, they must be attitudes with contradictory
contents:  for example, it seems irrational to believe that (p & ~p), or to
intend to (Φ at t & ~Φ at t). We can plausibly suppose that, if any enkratic
disposition is constitutive of possessing ought to Φ, the contours of that
disposition are circumscribed so as to not lead from beliefs that one ought
to both Φ and not-Φ to self-contradictory intentions. The exact shape of
the enkratic disposition that is supposedly required for possessing ought
to Φ depends on the specific inferential role that ought to Φ must play
in the mind of any thinker possessing it. There is no reason why the meta­
semanticist must suppose that it is part of the inferential role of ought to
Φ to lead agents who believe they are facing contradictory obligations to
have self-contradictory intentions.46
So it seems to me that, to state a rational requirement worth worrying
about, the Opposing Rule must after all concern the very same combi-
nations or sequences of attitudes that the would-be permission concerns.
And if the rest of the argument stands, this means, as before, that in order
to non-accidentally comply with the Opposing Rule, an agent must have
ought to Φ, and so must have the irrational disposition D[ENP(narrow)].47

Objection 2
ENP(narrow) could be false even if concept-constituting, because one
might be required to give up the concept ought to Φ altogether, or to
not have it in the first place. This is not an opposing rational requirement
governing the same combinations of attitudes that ENP(narrow) governs.
Rather, it is a demand to give up or not have the concept ought to Φ, and
thus to give up having, or not have, any attitudes that deploy that concept.
Notice that the objection is not that one might be rationally required to
not manifest a disposition that one has. That is the worry that has been
with us all along, and that the reductio above is supposed to counter (by
showing that concept-constituting dispositions must be rational ones, in
the sense of being dispositions to update intentions in rationally permissible
ways). Rather, the present objector admits that if I have the concept, and so
the disposition, then I am rationally permitted to update intentions in the

46
  Cf. the discussion in §1 of the exact contours of the MPP-ish inferential disposition
supposedly required for possessing conditional (p. 47 and n. 8).
47
  A further possibility is that one is under a rational requirement to not intend to Φ
whenever one, say, desires to Φ; and one either happens or tends to desire to Φ whenever
one believes that one ought to Φ. But this route to opposition with ENP(narrow) would
likewise require beliefs that one ought to Φ; and so, would require having ought to Φ.
 Vindicating Practical Norms 67

ways I am disposed to do. If I have the disposition, then I am permitted to
manifest it. What the objector denies is that I am thereby (unconditionally)
rationally permitted to update my intentions in the relevant ways, for she
denies that I am rationally permitted to have the concept.48

Response
The idea of being rationally required to lack a concept is odd. It is hard to
see how the idea of such a requirement relates to the sense of “rationality” in
play with requirements governing combinations or sequences of attitudes.
In any case, if there were rational requirements to lack concepts, it is hard
to see how one might non-accidentally comply with them, even if one is
perfectly rational. Suppose A is perfectly rational, and has never in fact had
the concept C that she should avoid. It is hard see how A might make sure
to avoid learning C, when she does not yet know what C is. Nor can A even
believe that she is to avoid acquiring concept C: this belief would require
having C. But if even a perfectly rational being can only accidentally avoid
violating a putative rational requirement to not acquire C, then this is not
really a rational requirement.
In counter-response, one might suggest that it is possible to know or
justifiedly believe that one should avoid acquiring a particular concept, if
someone reliable tells one to avoid something in the vicinity. For instance,
this reliable helper to maintaining one’s rationality might say: “Don’t asso-
ciate with those people,” or more to the point, “Don’t try to understand
what they’re saying,” or “Try not to understand what they’re saying.”49 But
it is hard to see how one might ensure that one won’t understand, or won’t
come to understand via a different route. Perhaps a perfectly rational being
can ask the reliable helper to tie her to the mast, making sure to keep bad
influences out. But it seems that a perfectly rational being’s tendency to
non-accidentally comply with rational requirements shouldn’t depend on
the contingent availability of such highly informed helpers. If it’s in the
nature of being perfectly rational to non-accidentally comply with rational
requirements—as per No Accident—then a perfectly rational being’s
rational dispositions should on their own equip her to non-accidentally
comply with rational requirements.50

48
  Thanks to an anonymous reviewer for raising this objection.
49
  Thanks to Kris McDaniel, Jamie Dreier, and Shyam Nair for raising versions of
this objection.
50
  Of course, even a perfectly rational being’s dispositions might not always ensure
compliance: dispositions can fail to be manifested. But the point is that, in the good cases
when obstacles to manifestation are not present, the rational dispositions will suffice for
compliance on their own.
68 Hille Paakkunainen

This still leaves open the possibility that one might be rationally required
to relinquish having the concept C, if one does have it. However, it is not
clear how this coheres with the claim, which the present objector was sup-
posed to grant, that if one does have the concept, then one is rationally
permitted to manifest the concept-constituting disposition. (Manifesting
that disposition involves having C.) In any case, the idea of a rational
requirement to relinquish a concept is as odd as the idea of a requirement
to avoid acquiring it. If I were under a requirement to relinquish a concept,
I would not know how to begin satisfying this requirement. I might acci-
dentally forget C. But this is not to the point: again, rational requirements
are supposed to be such that it is in principle possible to comply with them
non-accidentally. I might consciously and quite purposely renounce a con-
cept: I might consider its way of carving up the world to be misleading or
objectionable in some way, and resolve not to think in its terms. Pejorative
concepts like boche are plausibly like this. But renouncing a concept as
objectionable need not mean—and plausibly doesn’t mean—losing the
concept. Even after renouncing the concept as objectionable, I  can still
understand xenophobes’ objectionable thoughts that employ the concept.
Indeed, it is because I continue to understand xenophobes’ thoughts that
I can continue to knowledgeably view them as objectionable, and continue
to knowledgeably renounce the concept. If renouncing pejorative con-
cepts had to result in losing those concepts, then one couldn’t continue to
knowledgeably renounce them. Rather, having renounced them, one would
cease to understand them. This seems false to the phenomena. I conclude
that renouncing a concept isn’t relinquishing it.51 It is still unclear how one
might non-accidentally comply with this alleged rational requirement to
relinquish a concept.52
A final suggestion is that a perfectly rational being would simply be dis-
posed to relinquish any concept that she is required to relinquish: that is how
she non-accidentally relinquishes it, however the disposition works exactly.
Of course, unlearning the concept would put her at risk of accidentally
learning it again. But were she to accidentally relearn the concept, her dispo-
sition to relinquish it would soon ensure that she unlearns it. However, this
would trap the perfectly rational being in an odd see-saw loop of unlearning
and accidental relearning. It seems better to think of rational requirements,

51
  Thanks to Guy Fletcher for very helpful discussion of pejorative concepts.
52
 Perhaps one could go to a very specialized neurosurgeon who could zap the
circuits that sustain this specific concept. But a perfectly rational being’s ability to
non-accidentally comply with rational requirements shouldn’t depend on the contingent
availability of such specialized neurosurgeons. (And how expensive would these surgeons
be? Could only the rich be rational? Or would the state help one be rational?)
 Vindicating Practical Norms 69

if we can, in terms that don’t trap perfectly rational beings—the beings best
placed to comply with rational requirements—in such see-saw loops.53

Objection 3
The argument of §4 goes wrong in assuming that a perfectly rational being
must non-accidentally comply with the Opposing Rule requirement. It is
true that, as No Accident claims, a perfectly rational being non-accidentally
complies with all the rational requirements that apply to her. But the
Opposing Rule requirement need not apply to every perfectly rational
being. In particular, it does not apply to a perfectly rational being who
lacks the concept ought to Φ. For if one lacks this concept, then one
is not yet in a position to have the types of attitude that the Opposing
Rule requirement governs. And surely this means that the Opposing Rule
requirement does not even apply to one. You might protest that perfectly
rational beings can’t lack concepts. But surely they can. Just as one can
be perfectly rational while contingently lacking the demonstrative concept
this shade, for some particular shade of red, likewise one can be perfectly
rational while lacking the concept ought to Φ. In effect, then, one can
continue to be perfectly rational so long as one has neither the disposition
D[OR] nor D[ENP(narrow)].

Response
Consider first the claim about application. Suppose a young child lacks
the concept ought to Φ. She could not, then, have beliefs to the effect
that she ought to Φ until she acquired this concept. Does this mean that
requirements governing how to update intentions in the light of such beliefs
don’t apply to her? It is not clear that it does. Compare: I am not now in
a position to violate traffic laws in Syracuse. I am sitting in my office, not

53
  There is another, different objection in the area that is separate from our present
concerns. According to this different objection, just as one can possess pejorative con-
cepts while renouncing thinking in their terms, one can likewise continue to possess
the concept ought to Φ while renouncing thinking in its terms; and if the renouncing
is effective, this results in losing ought-beliefs, and might thereby also result in losing
the allegedly concept-constituting enkratic disposition. This would be an objection to
the metasemanticist’s premise about the conditions of possessing ought to Φ. And it
is not my aim here to assess that premise. For what it’s worth, I  think that it is not
plausible that xenophobic dispositions are required for grasping xenophobic concepts,
for reasons closely related to the reasons why it seems possible to renounce xenophobic
concepts while possessing them. If the metasemanticist thinks that ought to Φ is differ-
ent in this regard, or that it can’t be renounced, then this is something that her theory of
concept-possession had better explain.
70 Hille Paakkunainen

doing the sort of thing that those traffic laws govern. Still, they apply to me.
And they would apply to me even if, by contingent coincidence, I could not
right now engage in the activities they govern. (Perhaps I have no idea how
to drive, though I could learn.) What is true is that neither traffic laws nor
rational requirements apply to non-rational beings, such as lampposts. But
these rules do seem to apply to beings of the sort who are in general capable
of engaging in the activities that the rules govern.54 I don’t have to wait to
actually get into a car and learn how to drive in order for the laws about how
to drive to apply to me. Rather, the laws do apply, and if I don’t get into a
car, I conform to the laws governing driver behavior by default. Likewise,
it seems, I  don’t have to wait to actually have ought-beliefs in order for
requirements governing ought-beliefs to apply to me. Rather, as long as
I don’t have any ought-beliefs, I conform to the requirements by default.
Of course, my conformity to the rules here is merely external, not the
manifestation of a disposition to follow the relevant rules. I cannot have dis-
positions to update intentions in relation to ought-beliefs until I learn the
concept ought to Φ, just like I cannot have dispositions to drive a certain
way until I learn at least the rudiments of driving. Still, I don’t see why this
should mean that the relevant requirements don’t apply to me before I learn
to drive, or learn the concept.
Still, there is the question whether perfectly rational beings must
non-accidentally conform to rational requirements that apply to them, via
dispositions to follow those requirements. That is an important claim in the
argument of §4, since it is what led us to say that the perfectly rational being
needs to have the concept ought to Φ, if the Opposing Rule requirement
applies to her. And the present objector says that perfectly rational beings
can lack concepts, thus lacking the associated dispositions; and still remain
perfectly rational.
However, we must be careful about which intuitions lead us to think
that a perfectly rational being might lack certain concepts. Consider again
the young child who lacks the concept ought to Φ. As before, she is not
in a position to have ought-beliefs. Certainly we don’t want to say that
she is violating, or disposed to violate, any rational requirements governing
ought-beliefs. She is not irrational in this sense. Rather, she is conform-
ing to the requirements by default. If she lacks irrational dispositions, then
we might say that she is, in one sense, “perfectly rational.” However, this
does not mean that she is “perfectly rational” in a different, fuller sense.

54
  One might put the point by saying that the relevant rules quantify over everyone in
a certain relatively wide class—the class of beings of the sort that are capable of the sorts
of activities that the rules govern. The rules apply to everyone in that class in the sense
that the rules address themselves to everyone in the domain of quantification. Thanks to
Kim Frost for helpful discussion here.
 Vindicating Practical Norms 71

In this fuller sense, lacking perfect rationality need not consist in having
an irrational disposition: it can also consist in lacking a rational one—for
instance, a disposition to conform to rational requirements that apply to
one. And the child lacks a rational disposition: the disposition to update
intentions in accord with the Opposing Rule. She lacks this disposition at
least until she gains the concepts requisite to have the attitude-types that
the requirement governs. In this fuller sense of perfect rationality, one can-
not be perfectly rational—one cannot have all the dispositions to comply
with rational requirements that apply to one—without having the concepts
requisite for having these dispositions.
With this distinction between senses of “perfect rationality” to hand, let
us stipulate that No Accident concerns perfect rationality in the fuller sense.
(This is plausible in any case.) Let us also stipulate that No Irrationality
concerns perfect rationality in the fuller sense. (Though it also concerns
perfect rationality in the more modest sense of merely lacking irrational dis-
positions.) With these stipulations, the argument of §4 still goes through.
Not only does the Opposing Rule requirement apply to the (fully) per-
fectly rational being; but that being must also be disposed to comply with
the Opposing Rule. And this entails, again, that she must have the con-
cept ought to Φ. So, as before, No Irrationality and No Accident cannot
both be true, if the concept-constituting disposition D[ENP(narrow)] is an
irrational one.
(Notice that this still leaves it open that there are some concepts that per-
fectly rational beings, even in the fuller sense, can lack. It is just that those
concepts had better not be needed for having dispositions to comply with
the rational requirements that apply to one.)
These seem to me to be the worst objections facing the argument of
§4. If my responses stand, that argument looks to be on fairly strong foot-
ing. But that argument only purported to show that any permissive rules,
whether of narrow or wide scope, must be true if concept-constituting.
What about rules stating rational requirements? Must these also be true if
concept-constituting?

6.  Expanding the Argument: Rational

Requirements

Suppose that the disposition constitutive of possessing ought to Φ is

not D[ENP(narrow)], but instead D[ENR(narrow)]. One might well ask
what this supposition means:  what distinguishes these dispositions from
each other? At least in part, they are distinguished by the different sorts
of responses to situations that they tend to produce when manifested.
72 Hille Paakkunainen

Someone disposed to follow ENP(narrow) might, when believing that she

ought to Φ, either intend to Φ or not, depending on how she feels. This is
compatible with full manifestation of the disposition D[ENP(narrow)]. In
contrast, someone disposed to follow ENR(narrow) will intend to Φ when
she believes that she ought to Φ, if her disposition is manifested to the full.
We could likewise describe different patterns of response corresponding to
yet further dispositions.55
Suppose, then, that the disposition constitutive of possessing ought to Φ
is D[ENR(narrow)]. Under what conditions might the concept-constituting
requirement ENR(narrow) be false, so that one is not required to intend
the things that D[ENR(narrow)] would lead one to intend? There seem
to be three cases. In the first case, ENR(narrow) is false because one is
merely permitted to intend as it claims, not required. In the second case,
ENR(narrow) is false because one is required to not intend the things that
following ENR(narrow) would lead one to intend. And in the third case,
ENR(narrow) is false because it is rationally supererogatory or recom-
mended (if there is such a thing) to intend as it claims, not required.56
In the second case, the argument of §4 applies. There will be an opposing
requirement such that, to non-accidentally comply with it, one must have
the concept ought to Φ; meaning that one must also have the putatively
irrational disposition D[ENR(narrow)]. The rest of the argument is as it was
with ENP(narrow). Just like ENP(narrow), ENR(narrow) couldn’t be false
but concept-constituting if its falsity were due to an opposing requirement.
What about the other cases? Take the first case, in which one is per-
mitted but not required to do as the concept-constituting but false rule
says. Here updating intentions in the way that D[ENR(narrow)] would
lead one to update them is rationally permissible. It remains possible that
ENR(narrow) is false even though concept-constituting. But at least mani-
festing the disposition D[ENR(narrow)] wouldn’t lead one to update inten-
tions in ways that one is rationally required not to do. The initial animating
danger (from the end of §1) would be averted: concept-constituting dispo-
sitions couldn’t be irrational, in the sense of being dispositions to update
intentions in ways that one is rationally required not to do. So it can’t be

55
  The general strategy of individuating putatively rational dispositions here is similar
to Michael Smith’s (2003) strategy for individuating ways of falling short of rationality,
by means of differential responsiveness.
56
  Thanks to Brad Cokelet for suggesting this third case. You might think there is a
fourth case: ENR(narrow) is false because there are no rational requirements. However,
my argument is not an existence proof that there must be rational requirements if there
are facts about concept-constitution. Instead, it is an account of what the content of the
domain of rational permissions and requirements must be on the assumption that certain
claims about concept-possession are true, if there is such a domain.
 Vindicating Practical Norms 73

irrational to possess a concept, or to update intentions in accord with its

concept-constituting rule. Finally, the same reasoning applies in the third
case. If the concept-constituting rule is false because one is not required but
merely recommended to update intentions as it says, then again we avert the
initial animating danger, despite the falsity of the concept-constituting rule.
Notice that if this is right, then there can’t be any “defective” concepts, in
the sense of concepts that in and of themselves dispose thinkers who possess
them to update their attitudes in irrational ways. That is a good result for
the metasemanticist. However, one might still worry. For we have not quite
shown that if a putative rational requirement like ENR(narrow) is constitu-
tive of possessing a concept, then ENR(narrow) must actually be a rational
requirement. It might just be a rational permission or recommendation,
contrary to what the rule itself says. In response, I  have no knockdown
argument that concept-constituting rules couldn’t in this way subtly mislead
about their actual rational import. However, there is at least something very
strange about thinking that a putative norm’s status as concept-constituting
is enough to earn it some type of legitimacy, but that the concept itself is
somehow wrong, as it were, about what type of legitimacy this is.
There is more to be done to tie up this last thread. But I tentatively con-
clude that the prospects for expanding the metasemantic argument to cover
rational requirements are promising.

7. Conclusion

I have been arguing that starting from suitable inferentialist premises about
what it takes to possess concepts, we can validly derive rational permissions
and, more tentatively, rational requirements. Recall that I have not argued
for any such inferentialist premises. As I  noted, e.g. Allan Gibbard and
Ralph Wedgwood have both defended such premises about ought to Φ
in different ways. There is at least some plausibility to such premises. If my
argument here stands, it is these premises about concept-possession that
we must attack in order to resist metasemantic vindications of practical
norms.57
There is a different strand of thinking about the relationship between
ought to Φ and intentions to Φ that the above argument does not
address. Adapted to our present concerns, so-called “moderate internalism”

57
 Notice that nothing in my argument depends on whether judgments that one
ought to Φ are cognitive or non-cognitive states. To get the argument going, it is enough
that such judgments involve deploying the concept ought to Φ. Thanks to an anony-
mous reviewer for a helpful remark on this.
74 Hille Paakkunainen

about ought-judgments holds that an individual can have ought to Φ

without having a disposition to update one’s intentions to match one’s
ought-judgments; but that this is possible for individual agents only if
they are part of a community of agents who do have both the concept
and the disposition. An individual’s having the concept then still requires
that someone around her has the relevant disposition; but that someone
need not be the individual herself.58 Might such “moderate internalism”
about the possession conditions of ought to Φ also serve to vindicate the
concept-constituting enkratic norm that the community around the dissent-
ing individual are disposed to follow? This is work for another occasion.59

References
Boghossian, P. 2000. “Knowledge of Logic,” in P. Boghossian and C. Peacocke (eds),
New Essays on the A Priori, 229–54. Oxford: Clarendon Press.
Boghossian, P. 2001. “How are Objective Epistemic Reasons Possible?” Reprinted
in P.  Boghossian, Content and Justification:  Philosophical Papers, 235–66.
Oxford: Clarendon Press, 2008.
Boghossian, P. 2003. “Blind Reasoning.” Reprinted in P. Boghossian, Content and
Justification: Philosophical Papers, 267–87. Oxford: Clarendon Press, 2008.
Broome, J. 2007. “Wide or Narrow Scope?” Mind, 116: 359–70.
Broome, J. 2008. “Is Rationality Normative?” Disputatio, 11: 153–71.
Gibbard, A. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Horwich, P. 2005. “Meaning Constitution and Epistemic Rationality,” in P. Horwich
(ed.), Reflections on Meaning, 134–73. Oxford: Oxford University Press.
Kavka, G. 1983. “The Toxin Puzzle,” Analysis, 43: 33–6.
Kolodny, N. 2005. “Why Be Rational?” Mind, 114: 509–63.
Kolodny, N. 2007. “State or Process Requirements?” Mind, 116: 371–85.
McGrath, S. 2010. “Moral Realism without Convergence,” Philosophical Topics,
38: 59–90.
Parfit, D. 2011. On What Matters. Oxford: Oxford University Press.
Peacocke, C.  1993. “How are a Priori Truths Possible?” European Journal of
Philosophy, 1: 175–99.

58
  See van Roojen (2010: 499–500) for a recent formulation of this idea.
59
  I am very grateful to the audience at the Madison Metaethics Workshop for chal-
lenging discussion, and especially to Jamie Dreier, Jennifer Morton, Shyam Nair, and
Andrew Sepielli for questions that by increments prompted large changes in my argu-
ment. Guy Fletcher, Kim Frost, André Gallois, Owen King, Kris McDaniel, and two
anonymous reviewers for Oxford Studies in Metaethics each deserve huge thanks for read-
ing and commenting on previous drafts. Special thanks to Owen King, whose careful
questioning helped me tremendously in clarifying my argument. I  would also like to
thank Kieran Setiya for extensive feedback on early incarnations of the central ideas in
this chapter, and Kathryn Lindeman for helpful discussion of related issues.
 Vindicating Practical Norms 75

Peacocke, C.  2000. “Explaining the a Priori:  The Programme of Moderate

Rationalism,” in P. Boghossian and C. Peacocke (eds), New Essays on the A Priori,
255–85. Oxford: Clarendon Press.
Peacocke, C. 2004. The Realm of Reason. Oxford: Clarendon Press.
Pryor, J. 2001. “Highlights of Recent Epistemology,” British Journal for the Philosophy
of Science, 52: 95–124.
Ross, J. 2009. “How to Be a Cognitivist about Practical Reason,” in R. Shafer-Landau
(ed.), Oxford Studies in Metaethics, iv. 243–82. Oxford: Oxford University Press.
Schechter, J., and Enoch, D. 2006. “Meaning and Justification: The Case of Modus
Ponens,” Noûs, 40(4): 687–715.
Schroeder, M. 2012. “The Ubiquity of State-Given Reasons,” Ethics, 122: 457–88.
Schroeter, F. 2005. “Normative Concepts and Motivation,” Philosophers’ Imprint,
5: 1–23.
Setiya, K. 2007. “Cognitivism about Instrumental Reason,” Ethics, 117: 649–73.
Smith, M. 2003. “Rational Capacities, or:  How to Distinguish Recklessness,
Weakness, and Compulsion,” in S.  Stroud and C.  Tappolet (eds), Weakness of
Will and Practical Irrationality, 17–38. Oxford: Clarendon Press.
van Roojen, M. 2010. “Moral Rationalism and Rational Amoralism,” Ethics,
120: 495–525.
Wedgwood, R. 2007. The Nature of Normativity. Oxford: Oxford University Press.
Williamson, T. 2003. “Understanding and Inference,” Proceedings of the Aristotelian
Society: Supplementary Volume, 77: 249–93.
 4
Debunking Evolutionary Debunking
Katia Vavova

1.  The Evolutionary Challenge

Worries about the compatibility of evolution and morality are not new—
even Darwin had them. A number of recent arguments revive these con-
cerns. These evolutionary debunking arguments take the following form: you
just believe what you do because you evolved to, therefore you’re not justi-
fied in believing what you do. They typically target evaluative realism: the
view that evaluative facts are attitude-independent—that what is valuable is
valuable whether or not we happen to value it.1
The worry is that just as evolutionary forces shaped our eyes and ears,
so they shaped our evaluative attitudes. But, the debunker argues, we have
no reason to think that these forces would track the attitude-independent
evaluative truths that the realist posits.2 Worse yet, we seem to have a good
reason to think that they wouldn’t:  evolution selects for characteristics
that increase genetic fitness—not ones that correlate with evaluative truth.
Plausibly, the attitudes and judgments that increase a creature’s fitness come
apart from the true evaluative beliefs. If this is so, then it seems that evolu-
tionary forces have had a distorting effect on our evaluative attitudes. The

1
  This understanding of realism follows the evolutionary debunking literature. Similar
definitions can be found in metaethics more generally (see Shafer-Landau 2005:  15
on “stance-independence”). For present purposes, evaluative propositions are of the
form: that X is a normative reason to Y, that one should or ought to X, that X is good,
valuable, or worthwhile, that X is morally right or wrong, and so on. Evaluative attitudes
include (conscious or unconscious) beliefs in evaluative propositions, as well “as desires,
attitudes of approval and disapproval, unreflective . . . tendencies such as the tendency to
experience X as counting in favor of or demanding Y,” etc. (Street 2006: 110).
2
  From here on I’ll drop the “attitude-independent” qualifier on evaluative attitudes
or truths.
 Debunking Evolutionary Debunking 77

debunker concludes, insofar as we are realists and insofar as the evolution-

ary facts are thus-and-so, we are not justified in our evaluative beliefs.
Evolutionary debunking arguments are sometimes meant to establish
just this: evaluative skepticism. Other times the skeptical conclusion is in
the service of the greater goal of undermining evaluative realism. In either
case, the debunker must first establish that learning about the evolutionary
origin of our evaluative beliefs gives us, qua realists, good reason to worry
about our evaluative beliefs. I will argue that the considerations she puts
forth cannot give us such reason. I will conclude that there is little hope
for distinctly evolutionary debunking arguments. This is bad news for the
debunker who hoped that the cold, hard scientific facts about our origins
would undermine our evaluative beliefs.

2.  The Debunker’s Argument

“[T]‌here can hardly be a doubt,” Darwin speculated, that if we had

evolved under the same conditions as hive-bees, “our unmarried females
would . . . think it a sacred duty to kill their brothers, and mothers would
strive to kill their fertile daughters; and no one would think of interfer-
ing” (1871: 73). If instead we had evolved as lions did, Street argues, males
would have “a strong unreflective evaluative tendency to experience the kill-
ing of [other’s] offspring . . . as ‘demanded by the circumstances’.” Not only
would females lack an “unreflective tendency to ‘hold it against’ a male
when he killed her offspring,” but would tend to become “receptive to his
advances soon afterwards” (2006: 121).
These observations are meant to support this counterfactual: if we had
evolved differently, we would have believed differently—our evaluative
beliefs, in particular, would have been different. In turn, this counterfactual
is meant to support the claim that the content of human evaluative judg-
ments has been “tremendously influenced . . . by the forces of natural selec-
tion” (Street 2006: 121).
The debunker hopes to use this story to undermine our evaluative beliefs.
We cannot rationally maintain our opinions about good and bad, right
and wrong, reasons and values, she argues, once we realize from where they
came. The debunker thus aims to get somehow from
influence. Evolutionary forces have influenced our evaluative beliefs.
to
revision. We cannot rationally maintain our evaluative beliefs.3

3
 Street doesn’t say that we cannot rationally maintain belief, but rather that
we “should suspend belief ” (forthcoming:  2). I  think we mean the same thing here.
78 Katia Vavova

To be sure, influence is not equally worrying for everyone. Anti-realists

take the evaluative truths to be attitude-dependent—somehow a function
of our (actual, ideally rational, etc.) beliefs and desires. Since anti-realists
hold that our values determine what is valuable, they needn’t worry from
where those values came. Realists are more vulnerable. Since they take the
evaluative truths to be independent of our beliefs and desires, they are com-
mitted to the possibility of evaluative error: what we value and what is valu-
able can come apart.
Some varieties of realism are importantly different and may be better
placed to dismiss the debunker. I  won’t explore that here. First, under-
standing the debunker’s challenge doesn’t require digging into the details
of realism. The evolutionary story is at least initially worrying for anyone
who holds that the true evaluative beliefs come apart from the adaptive
evaluative beliefs.4 Second, since I will present structural problems with the
debunker’s challenge, my strategy should be one that realists of any stripe
may deploy in self-defense.
Let us grant then that some form of evaluative realism is the target, and
assume for the sake of argument that the true evaluative beliefs come apart
from the adaptive evaluative beliefs. Given this much, the question is how
to get from influence to revision. To seal this gap, we need to know what
is the epistemic significance of the evolutionary story for our evaluative
beliefs.
In the next sections, I will consider two ways of filling in the debunker’s
story.5 I will extract valid arguments to revision from both. The first, which
Street suggests, is compelling, but too strong for the debunker’s purposes.
It collapses her challenge into a more general skeptical challenge. The sec-
ond is more promising and the right way to understand distinctly empirical
debunking arguments.

Presumably Street’s “should” is an epistemic one. What we “should” epistemically believe

could depend on what we actually believe, what our evidence supports, etc. Here it won’t
matter exactly how we understand this “should” or the relevant notion of “rational”
because Street’s argument proceeds by first trying to establish a lemma that I will argue
she cannot.
4
  This is in contrast with the claim that the challenge is best understood as aimed at
non-naturalist or non-reductive realists (e.g. Bedke MS: 1). The challenge may be more
formidable for this particular variety of realism, but a more minimal commitment suf-
fices to get it going.
5
  There is textual evidence for both readings, though I do not know of others who dis-
tinguish them. For the first see Street (MS and Forthcoming); for the second see Street’s
talk of distorting influences in her 2006.
 Debunking Evolutionary Debunking 79

3.  Do We Have Good Reason to Think

We’re Right?

The evolutionary debunker claims that in some sense of “evolved” and

in some sense of “belief,” we evolved to hold our evaluative beliefs. The
thought is that just as “creatures inveterately wrong in their inductions
have a pathetic but praiseworthy tendency to die before reproducing their
kind” (Quine 1969: 126), so creatures with deep-rooted inclinations to kill
themselves and their offspring tend to have quite short evolutionary histo-
ries. Given that different evaluative tendencies can have “extremely differ-
ent effects on a creature’s chances of survival and reproduction,” we should
expect “over the course of our evolutionary history, relentless selective pres-
sure on the content of our evaluative judgments” (Street 2006: 114).
This is the evolutionary story. The debunker doesn’t suggest, implaus­ibly,
that evolution directly shaped our more sophisticated evaluative beliefs.
The evolutionary story is meant to undermine directly only more basic and
less controversial beliefs,6 like the belief that the fact that something would
promote one’s survival is a reason in favor of it, or that we have greater obli-
gations to help our own children than complete strangers. But the evolu-
tionary story is also meant to undermine indirectly the rest of our evaluative
beliefs, including our much more sophisticated judgments. If our belief that
we have reason to avoid inflicting unnecessary suffering goes, so does the
moral theory that rests, partly, on it. Hence, the debunker concludes: “our
system of evaluative judgments is thoroughly saturated with evolutionary
influence” (Street 2006: 114).
This is the empirical claim. No one, not even the debunker, thinks it
conclusive.7 So, why take it seriously? Because the philosophically interest-
ing question is not whether some empirical claim is true, but what follows
about the rationality of our beliefs if something like it were true. This ques-
tion has implications for our epistemology and our metaethics, but it is also
of practical interest. Even if the evolutionary debunker fails, some of our
other beliefs might reflect some other suspect influence. We need to know
how to respond to such evidence if, or when, we do get it.
Grant the evolutionary story for argument’s sake. Why should it worry
us? Because if it is true, the debunker argues, then the best explanation for
why we hold the evaluative judgments we do is that they are adaptive.8
6
  Or some sort of proto-belief states or tendencies (Street 2006: 115).
7
  Cf. Street (2006: §3). For reasons to think that the evolutionary story is “a long way
from even beginning to fill out the empirical details needed to fully secure” these prem-
ises, see the just quoted Kahane (2011: 111), Sliwa (MS), and FitzPatrick (Forthcoming).
8
  Cf. Street (2006) on the adaptive link account.
80 Katia Vavova

And this explanation is epistemically unflattering: that we evolved to hold a

judgment is no reason to think that it is true.
The debunker then asks: knowing just about the evolutionary origin of
our evaluative beliefs and nothing else, do we have reason to think that
those beliefs are true? We know that, by hypothesis, evolution selects for
adaptive beliefs regardless of their truth. So it may be that the evaluative
beliefs we should hold are such-and-such, but that the ones we do hold are
this-and-that, because the latter are adaptive and the former aren’t. Our
evaluative beliefs may, then, be massively mistaken and our origin story
gives us no reason to think that they are not.9
This is Street’s suggestion. Since we evolved to hold our evaluative beliefs,
we have no reason to think they are true. Rationality requires we have good
reasons for thinking our beliefs are true. So we cannot rationally maintain
our evaluative beliefs. Skepticism follows.
This version of the debunker’s story relies a principle like this:
no good. If you have no good reason to think that your belief is true,
then you cannot rationally maintain it.
Street explicitly endorses a principle like this. She argues that it captures the
difference between being hypnotized to believe that Hayes was the nine-
teenth US president and learning it in school (Forthcoming: 2). In the for-
mer case you have no reason to think that the process by which you gained
your belief would have led you to form true beliefs. We don’t typically think
that magicians use their powers of hypnosis for good—to implant in their
victims true beliefs about US history. Competent high school teachers, on
the other hand, are concerned with just this task. The explanation of your
historical beliefs in terms of hypnosis is thus undermining; the one in terms
of education is vindicating.
Street argues that evolution is more akin to a careless hypnotist than a
teacher.10 We have no good reason to think that selective pressures would
push us toward the truth. Learning about the influence of evolutionary
forces on our evaluative beliefs should thus undermine those beliefs.
Many have found this puzzling, insisting that we have plenty of good
reasons to think our evaluative beliefs are true. Even if evolution caused us
to believe that “pain and injury are bad, and that we have strong reasons to
promote the survival and well-being of ourselves and our children,” Parfit
writes, “these beliefs are not badly mistaken, but correspond to some of

9
  There are two relevant ways of understanding “mistaken” here. On the first, a belief
is mistaken just in case it is false. On the second, a belief is mistaken just in case it is not
supported by the believer’s evidence. What sort of mistake does the debunker point to?
That’s for her to say. I will follow much of the literature and focus on the first. This mostly
won’t matter for my purposes, but I will make a note when it does.
10
  I agree, though I’ll soon argue that this principle doesn’t capture these differences.
 Debunking Evolutionary Debunking 81

the independent normative truths. Pain is bad, and we do have strong rea-
sons to promote the survival and well-being of ourselves and our children”
(2011: 533). Discussing an analogous case, Dworkin wonders what the fuss
is about. Why shouldn’t we, he writes, “count it as a piece of luck—a special
example of what Bernard Williams has called moral luck [that our adaptive
beliefs and the true ones] here coincide?” (Dworkin 1996: 125).
Other defenses of realism begin with similarly substantive moral assump-
tions: that pain is bad, that survival is good, that we have rights, and so on.11
Street argues, however, that such assumptions are illegitimate in this context.
To presuppose the truth of particular evaluative judgments is to presuppose
exactly what the evolutionary story is meant to bring under scrutiny. This is
“trivially question-begging,” Street argues. Our reasons for thinking that our
judgments are true cannot simply assume “the very thing called into ques-
tion,” namely the truth of those judgments (Street MS: 15–16).
Whatever we think of the best version of this response, we should grant
that there is something prima facie fishy about it. This is most evident in
Dworkin. He begins by granting that evolution has been a suspicious, epis-
temically no-good influence on our evaluative beliefs. He then insists that
we happened to have gotten things right. After all, we believe we have rea-
son to take care of our kids, and we are right in so believing. How lucky that
the adaptive beliefs and the evaluative truth here coincide!
If the onus is on us to demonstrate that we are not mistaken, we cannot
simply insist that our beliefs are true and count ourselves lucky. We would
be like the dogmatist who reasons that since he knows that p, any evidence
he gets against p must be misleading, so he can ignore it.12 We cannot safe-
guard our beliefs from defeating evidence like this. Nor can we dismiss the
debunker’s challenge so easily.
We can now see what the debunker thinks we need if we are to avoid
her challenge: a reason to think that we are not mistaken in our evaluative
beliefs that doesn’t simply presuppose the truth of those beliefs. This reason
is, in some sense, independent of what is called into question.13

11
  Wielenberg’s (2010) response assumes that we have rights. Enoch’s (2010) assumes
that “survival or reproductive success (or whatever else evolution ‘aims’ at) is at least
somewhat good” (2010: 18). Dworkin repeatedly insists that we can just count ourselves
lucky (1996, 2011). Parfit earlier claims that moral beliefs can be justified by their intrin-
sic credibility (see his 2011: 490). I won’t say more about these here. I take them up in
my MS b.
12
  Cf. Harman (1973: 148) and Kripke (2011: 49).
13
  This independence requirement is crucial to the debunker’s argument, and yet has
no defense in the debunking literature. Elga (2007), Christensen (2007), and others
explicitly endorse similar independence requirements for disagreement. White ques-
tions them (2010: 588–9). More must be said about what counts as independent, how
to set aside what is not, and how to characterize this “setting aside” formally. These
82 Katia Vavova

This explains why the debunker asks us to bracket our evaluative beliefs—
even those that we know or rationally believe—and to focus only on the
origin story. If we do not do this, we stack the deck in our own favor. The
danger, of course, is that if we do, then we may well lack reason to think our
beliefs aren’t mistaken.

3.1 Why no good Is No Good

The debunker thus needs a “good” reason to be an appropriately “independ-
ent” reason. This stringent understanding allows the debunker to dismiss
Parfit et al. and claim that we have no good reason to think our evaluative
beliefs are right. But if we understand “good” reason this way here, we must
understand it in the same way in no good. This, I will now argue, entails a
skepticism far more pervasive than the debunker ever intended.
Start with an explicit statement of this version of the argument.
1. influence. Evolutionary forces have influenced our evaluative beliefs.
2. We have no good reason to think that our evaluative beliefs are true. [1]‌
3. no good. If you have no good reason to think that your belief is true,
then you cannot rationally maintain it.
4. revision. We cannot rationally maintain our evaluative beliefs. [2, 3]
Every premise in this argument is controversial. I granted the first, and I will
momentarily grant, for argument’s sake, that it somehow entails the second.
Do not worry that this concedes too much to the debunker. Such generosity
will not give the game away. Focus instead on the third premise. no good
seems compelling because it raises a familiar sort of skeptical challenge. But
it also collapses the debunker’s challenge into that more ambitious one for
which no empirical premise is necessary and which undermines much more
than evaluative realism. To see this, consider:
Perception. We come to hold beliefs about our manifest surroundings
on the basis of signals that hit our sensory organs.
Unless we are skeptics, we should grant that sensory perception is a per-
fectly good belief-forming method. Ceteris paribus, if you perceive that p,
you are rational in concluding that p. Do we have good reason to think
that perception would lead us to true beliefs about our surroundings? Not
if “good” reason is understood as an appropriately independent reason: for

questions have been little addressed in the literature and I won’t be remedying that here.
Though rough, the characterization here suffices. I think independence requirements are
plausible, though I won’t argue for this here. See my MS a and MS b.
 Debunking Evolutionary Debunking 83

if we set aside all that is in question, we must set aside all beliefs gained by
perception. This includes all scientific beliefs, like the belief that evolution-
ary theory is true. Without those, we cannot evaluate the rationality of
beliefs formed by perception. We can test the reliability of a particular sense
modality by granting the reliability of others. We can test our eyes against
our ears, and so on. But if we cannot rely on any of our senses, we have
nothing with which to evaluate reliability. We have set aside too much.
This might just be what the skeptic aims to demonstrate: that our jus-
tifications eventually run out and our beliefs ultimately rest on nothing.
This, however, was never the debunker’s point. She aimed to undermine
a particular, limited set of our beliefs using good scientific evidence that
they are mistaken. no good commits her to much more. If this argument
works, it undermines all that we believe and the evolutionary premise drops
out. Worse yet, if we aren’t justified in believing anything, then everything is
awful, but there is no special problem for the evaluative realist.
Some have argued that the evolutionary story is not essential to the argu-
ment. This is only true in an uninteresting sense:  any suspect influence
could do the job. It needn’t be evolution. But an empirical claim of some
sort is essential—this is the distinctive feature of such arguments.14
This isn’t always made clear. Elsewhere Street begins by pointing to the
phenomenally low “odds that among all the possible coherent normative
systems, one’s own is the right one” (MS: 21). Since there are infinitely many
possible coherent normative systems, she argues, it would be a “striking
coincidence” if one’s own normative system happened to be the correct one
(MS: 21).15 Given that “one has no non-trivially-question-begging evidence
that one’s own system is the right one,” it is unreasonable to conclude that it
is (MS: 21). Street thus concludes that we have no good reason to think that
our evaluative beliefs are roughly on-track, for we have no reason that does
not assume the very thing called into question: the truth of those beliefs.
This version of the debunker’s challenge brings nothing new to the table.
It demands that we demonstrate that we aren’t massively mistaken about
morality. Legitimate or not, this is not the debunker’s demand.16 It is just an
instance of a general skeptical worry, suspiciously similar to this one:
Possibility of Error. Some possible states of belief are coherent and
stable—they look fine “from the inside”—and yet are mistaken. There

14
  Cf. Bedke (MS: 3) and Street (2006: 155).
15
  Bedke presents the challenge this way:  as that of explaining this striking coinci-
dence. He does think an empirical premise is necessary, however, so it isn’t obvious which
way he goes.
16
  I argue for this in my MS b, first presenting the explanatory demand and then
distinguishing it from the debunker’s.
84 Katia Vavova

are infinitely many of these and just one that is right. Furthermore,
we have no good reason to think we’re not in such a state. So it would
be unreasonable for us to be confident that we’re not in such a state.17
This challenge doesn’t and needn’t rely on empirical claims. You are asked
to justify your entire body of belief—and, on the relevant understanding of
“good reason,” you must do it without presupposing the truth of any of the
beliefs that have been called into question. But all of your beliefs have been
called into question, so the skeptic asks you to put them all aside. She then
asks: have you one good reason to think that your beliefs are true? You do
not, of course. And it isn’t because you have some reasons, but they aren’t
any good. The problem is that once you put aside all that you believe, you
don’t have any reasons left.18 You do not even have beliefs, so how could
you have reasons?19
This challenge can be raised against any subject matter. It isn’t peculiar to
the evaluative, it isn’t uniquely a problem for realism, and it can be raised
without empirical premises. If the debunker accepts no good, she commits
herself to the legitimacy of this reasoning. She thus ends up with the con-
clusion that we should all—regardless of our metaethics—suspend judg-
ment about everything. But that was never her goal.
Focusing on the many coherent evaluative states that we might be in is
thus misleading. That there are many such states, and that we have no good
reason to think we are in one of the good ones may be a problem, but it isn’t
the debunker’s problem. Her aim is to show, I will now argue, that we have
good reason to think that we are in one of the bad states.

4. Why good Is Good

What is the epistemic significance of the evolutionary story for our evalua-
tive beliefs? I argued that it couldn’t be that it leaves us with no good reason
to think we are not massively mistaken about the evaluative. If we under-
stand a “good reason” as we must, to avoid begging any questions, then

17
 Elga (MS: 7).
18
  Do you have anything left with which to even comprehend the skeptic’s question?
That is another difficulty. There is a more general anti-skeptical strategy in this spirit,
most commonly attributed to Wittgenstein (1969). Wright (2004) develops a view in
the same spirit. My goal is not so ambitious. It is simply to distinguish skeptics from
debunkers.
19
  Of course, there is a sense of “reason” on which I can have one even if I do not
or cannot believe I have one. For the record, here and throughout, I will use “having a
reason” and “believing you have a reason” interchangeably.
 Debunking Evolutionary Debunking 85

we certainly lack such reason. But we lack it for our entire body of beliefs.
While that may be a problem, it isn’t the debunker’s problem. So her point
cannot be that we lack good reason to think we’re right.
What is her point? It has something to do with the epistemically unflat-
tering picture the evolutionary story paints. What is epistemically unflatter-
ing, however, isn’t that we cannot independently establish that these beliefs
are right. Rather, it is that in learning this story about the origin of our
evaluative beliefs, we get good reason to think that our beliefs are wrong.
Since evolutionary forces select for adaptive beliefs—and not true ones—
evolution is a bad, potentially distorting influence on our evaluative beliefs.
On this alternative line of thought, the problem is not that we cannot dis-
miss the possibility of error—it is that good scientific evidence makes this
possibility more probable.
This version of the debunker’s argument is distinct from traditional skep-
tical arguments since it rests on an empirical claim. It is more selective than
traditional skeptical arguments because it targets all and only the suspi-
ciously influenced beliefs. The epistemic principle it relies on is:
good. If you have good reason to think that your belief is mistaken,
then you cannot rationally maintain it.20
The difference between good and no good is subtle but crucial. Roughly,
it is the difference between taking our beliefs to be innocent until proven
guilty and taking them to be guilty until proven innocent. no good requires
you to launch a defense on behalf of your belief; good requires you to hear
out the prosecution. Both of these principles can be used to formulate a
valid debunking argument, but the debunker should accept good only.
The debunker’s point is that evidence of evolutionary influence is evi-
dence of error. When we get such evidence, we must accommodate it with
appropriate revision. This is exactly what good expresses. It rightly shifts

20
  The caveat from n. 9 is relevant here. I use “mistaken” to mean “false” here, but
these principles could be formulated in terms of rationality, justification, or evidential
support. e.g.
good*. If you have good reason to think that your belief is not supported by your evi-
dence, then you cannot rationally maintain it.
This is more controversial. Christensen (2011), Elga (2007), and Vavova (MS a)
defend principles along these lines. Kelly (2005) and Weatherson (MS) reject them.
They argue that “higher-order” evidence about p—evidence about your evidence for p—
should not affect your “first-order” attitude about p. There might be nothing wrong, on
their views, in believing both that p and that your evidence does not support p. So they
would reject good*. They could still accept good, however, for that commits them to
something weaker and more plausible: that you cannot rationally believe both that p and
that p is false.
86 Katia Vavova

the burden to the debunker. It isn’t up to us to show her that we aren’t

mistaken. It is the debunker’s job to show us that we are mistaken. good
reflects this dialectic and provides a plausible link between the discovery
that a belief reflects the influence of a suspect process and the conclusion
that we cannot rationally maintain that belief.
Earlier we granted, for the sake of argument, that we have no good
(independent) reason to think our evaluative beliefs are not mistaken.
With no good, this entailed that we could not rationally maintain our
evaluative beliefs. If we accept good only, the debunker must do more.
Our lack of good (independent) reason to think our evaluative beliefs are
right leads nowhere without something like no good. The onus is now
on the debunker to show that the evolutionary story supports something
stronger. She must do more than merely demand an explanation and watch
us squirm. She must show us that we have good reason to think that our
evaluative beliefs are mistaken.21
A good reason is here, as before, an appropriately independent one. Your
evaluation of whether you have good reason to think that you are mistaken
about p should not rely on p or on the evidence or arguments on which p
is based. This is for the same reason as before: to block a certain kind of
question-begging response. If I can take for granted that pain is bad and
survival is good, then I have a quick and easy explanation for why evolution
is concerned with exactly the attitude-independent moral truths.
The independence requirement is also important here for another reason.
Since the onus is now, rightly, on the debunker to give us evidence of error,
this evidence should be good evidence we can recognize as such. It should
follow from our other beliefs about reasons and evidence. But notice how
odd it would be for her to rely on the beliefs she does not allow us to rely
on—the ones we are supposedly mistaken about. Her argument would be
something like this one: p is probably false, but it entails q, so you should
believe q. The debunker cannot simply rely on the beliefs that are supposed
to be mistaken—the very same ones she won’t let us take for granted. She
must build her case upon solid, independent grounds. She thinks she can,
but I will argue to the contrary.22

21
  There is some evidence for this reading (cf. n.  5). See esp. Street (2006) where
she often talks of the “distorting” Darwinian forces having led us off-track, or “having
pushed us in evaluative directions that have nothing whatsoever to do with the evalua-
tive truth” (121). I do not think anyone is consistent on which way to understand the
evolutionary debunker’s challenge: like this or as a more general skeptical challenge. My
MS b more thoroughly defends this interpretation of the dialectic.
22
  Cf. Street (MS) where she argues that the particular normative assumptions in ques-
tion are not needed for either raising or responding to the challenge.
 Debunking Evolutionary Debunking 87

Consider first this revised version of the argument:

5. influence. Evolutionary forces have influenced our evaluative beliefs.
6. mistaken. We have good reason to think that our evaluative beliefs are
mistaken. [1]‌
7. good. If you have good reason to think that your belief is mistaken,
then you cannot rationally maintain it.
8. revision. We cannot rationally maintain our evaluative beliefs. [2, 3]
Every premise of this argument is also controversial, but good is weaker
and more plausible than no good. It provides a framework within which
the debunker can pose an appropriately selective and distinctive challenge.
It is at least possible to construct the right kind of debunking argument.
The action is now with the second premise: have we, realists, been given
good reason to think that our evaluative beliefs are mistaken? I will examine
three evolutionary debunking arguments, which aim at a different set of our
evaluative beliefs. I will argue that in all three, the debunker fails to give us
good reason to think we are mistaken. Since we can then reject the second
premise, we aren’t pushed into evaluative skepticism.

5.  Debunking Evaluative Realism

The most familiar evolutionary debunking argument targets moral realism,

and aims to undermine our beliefs about what we have reason to do. I will
start with a more ambitious argument, which aims to undermine evaluative
realism wholesale: not just our beliefs about what we have reason to do, but
also our beliefs about what we have reason to believe. This debunker thus
targets realism about both practical and epistemic reasons.23
To see how the trouble is supposed to arise, consider our belief that fre-
quency facts like
[tigers] the fact that all previously encountered tigers were
carnivorous,
give us reason to believe inductive claims like
[next tiger] the next tiger we encounter will also be carnivorous.
It is clear why we evolved a tendency to form beliefs like [next tiger]
on the basis of frequency facts like [tigers]:  if we hadn’t, tigers would

23
  Cf. Street (2009).
88 Katia Vavova

have eaten us. But why did we evolve to take frequency facts like [tigers]
as reasons to believe facts like [next tiger]?24 Is it because grasping this
attitude-independent normative truth was itself adaptive? Unlikely, Street
argues: natural selection favored a tendency to take considerations of truth
to bear on what to believe “not because it constituted a perception of an
independent fact about reasons, but rather simply because it guided the
formation of creatures’ beliefs in ways that turned out to be advantageous
for the purposes of survival and reproduction—in particular, because it got
them to believe things that turned out to be true, or at least roughly true,
about tigers and much else” (Forthcoming: 17).
In other words, we wouldn’t believe that [tigers] is a reason for believ-
ing [next tiger] if concluding [next tiger] on the basis of [tigers]
weren’t to our evolutionary benefit. Since evolution has no interest in the
attitude-independent epistemic truth, the beliefs it influences are likely to
be mistaken. Insofar as we are realist, the debunker argues, and continue to
maintain that what is epistemically valuable is valuable whether or not we
value it, we seem pushed to skepticism.
This argument rests on the claim that the same kinds of considerations
meant to undermine beliefs like we have reason to take care of our children
would also undermine beliefs such as we have reason to believe this rather
than that on this evidence. Even as she launches a formidable defense of this
claim, arguing both that evolutionary forces influenced our beliefs and that
this should worry us, Street admits that this case is much harder to make.
Grant her the first bit again (namely, influence) and ask: if evolution
had shaped our beliefs about epistemic reasons, would this give us a good
epistemic reason to worry about those beliefs? I will argue that it does not
and it cannot, for there is a deep structural problem with an argument this
ambitious.
The debunker aims to give us good reason to believe that we cannot
trust our beliefs about reasons for belief. But this itself—what the debunker
wants to give us—is a reason for belief. So we cannot trust it. We are there-
fore not permitted to take for granted the very thing we need to call our
evaluative beliefs into question. This is because, recall, the debunker must
give us good independent reason that is, by our own lights, reason to think
we are mistaken. But on this version, what we are supposed to be mis-
taken about includes, crucially, epistemic principles about how to revise our
beliefs in light of evidence. We need to take for granted the truth of good
and mistaken. Both of these claims, however, are about what we have rea-
son to believe, which is exactly what we’re supposed to be mistaken about.

24
  I assume here a view on which taking [tigers] to be a reason to believe [next
tiger] is something more than merely having the disposition to infer one from the other.
 Debunking Evolutionary Debunking 89

The debunker thus faces a dilemma. She may relax her standards for
what counts as a “good” reason, or she may maintain them. If she maintains
them, then she cannot give us good reason to think we are mistaken about
the evaluative. In short, this is because to evaluate we must rely on the eval-
uative. But in aiming to debunk all of our evaluative beliefs, the debunker
leaves us with nothing with which to evaluate whether those beliefs have
been debunked.
If instead the debunker relaxes her understanding of “good reason,” then
good is back. But so are our other beliefs about epistemic reasons, like the
belief that [tigers] really does give us reason to believe [next tiger], and
so on. And if we are allowed these assumptions, then the question-begging
response Street blocked is open again.
There is a natural response available to the debunker here. She could
reply that her point is dialectical, not skeptical. Though some debunkers
are skeptics or nihilists, others, like Street, are not. They do not really aim
to debunk our evaluative beliefs—they think those are true. Instead, they
aim to debunk realism. The skeptical conclusion is only for the purposes
of reductio, for these debunkers. It follows from realism and science, they
argue, and it is absurd. We cannot give up science, so realism must go.
Unfortunately, this response won’t do. Even if the debunker does not
ultimately endorse the skeptical conclusion, she must still show that it fol-
lows from realism and the evolutionary story. It is only if she can dem-
onstrate this that she has what she needs for her reductio. To do so, the
debunker must give us realists good reason to think we are mistaken, if
evolutionary theory is true. I have argued that the debunker is in principle
incapable of providing evidence of such global error. The reductio thus can-
not go through. mistaken is false. We do not have good reason to think we
are mistaken. The evolutionary story, at least, hasn’t given us any.
Such is the fate of the debunker who attacks evaluative realism wholesale.
Perhaps it isn’t surprising that this most ambitious debunker failed in just
this way. There are well-known puzzles about whether we can revise, or
even be anything short of certain of, our most fundamental principles of
belief revision.25 But perhaps the debunker can sidestep these difficulties
and avoid such a fate, if she can narrow her target.

6.  Debunking Moral Realism

There is more hope for the debunker who aims only at moral realism. Since
she does not target our beliefs about epistemic reasons, both good and

25
  See Field (MS a, MS b), and Lewis (1971).
90 Katia Vavova

mistaken are potentially in play. The question is whether she can actually
establish the latter—whether she can use her evolutionary story to give us
good reason to think we are mistaken about morality. There are two impedi-
ments in her way.
The first is that the debunker must show that evolution causes trouble for
our moral beliefs only—that there is some disanalogy between this argu-
ment and the previous one. But the two arguments are presented as exactly
analogous (Street 2009). If the debunker cannot narrow down her target in
a principled way, this less ambitious argument collapses into the previous,
thereby sharing its fate.
The second is that even an appropriately narrowed challenge calls too
much into question. Since it targets all of our moral beliefs, we are left
knowing nothing about morality. But how can we tell if we are likely to be
mistaken about morality, if we know nothing about it? This concern will
occupy the rest of this section. To see it more clearly we need to zoom in to
the first inference of the argument.26
So far, we have either granted or glossed over the move from influence
to mistaken. Now we must look closer, for mistaken simply doesn’t follow
without, at least, reason to be suspicious of the purported influence. As
Street puts it:
genealogical information by itself implies nothing one way or another about
whether we should continue to hold a given belief. Rather, in order validly to draw
any conclusions about whether or how to adjust one’s belief that p, one must assess
the rational significance of the genealogical information, locating it in the context
of a larger set of premises about what counts as a good reason for the belief that
p. (Forthcoming: 2)
Kahane (2011) suggests, as a possible supplementary premise, that evolu-
tion is an “off-track” process since, by hypothesis, it doesn’t track the atti-
tude-independent evaluative truths.27 So long as we think that the adaptive
beliefs come apart from true beliefs, we can accept this premise. Expanding
the argument thus we get:
1. Evolutionary forces select for creatures with characteristics that increase
fitness.

26
  In fact, the previous debunker faces an exactly analogous problem:  if we know
nothing about the evaluative, how can we tell we are likely to be mistaken about it?
27
  Bedke rightly warns that a process being off-track “is ambiguous between the claim
that the process has been shown to be unreliable and the claim that explanations for the
process do not aver to the target facts” (MS: 4–5). I think the debunker should be claim-
ing something more like the former. The latter claim is more akin to the aforementioned
explanatory demand, which I take up and distinguish from the debunker’s in my MS b.
 Debunking Evolutionary Debunking 91

2. The true evaluative beliefs and the adaptive evaluative beliefs

come apart.
3. Evolutionary forces are off-track:  they do not track the evaluative
truth. [1, 2]
4. influence. Evolutionary forces have influenced our evaluative beliefs.
5. off-track. Off-track forces have influenced our evaluative beliefs. [3, 4]
If the debunker can establish off-track, she is a short step from mistaken.
After all, an off-track influence pushes your beliefs in directions having
“nothing whatsoever” to do with the truth. Reason to think your belief
reflects the influence of an off-track process thus looks like good reason to
worry about the truth of that belief. If the above argument gives us good
reason to think that our evaluative beliefs reflect an off-track influence, then
it seems that we have good reason to think that those beliefs are mistaken.
good then takes the debunker home:
6. mistaken. We have good reason to think that our evaluative beliefs are
mistaken. [5]‌
7. good. If you have good reason to think that your belief is mistaken,
then you cannot rationally maintain it.
8. revision. We cannot rationally maintain our evaluative beliefs. [6, 7]
We’ve granted influence and good. We could resist the inference from
off-track to mistaken, but we shouldn’t. It isn’t so controversial: it doesn’t
say that learning about an off-track influence should all-things-considered
worry you; just that it gives you a reason to worry.
Focus instead on off-track. To get there, the debunker needs P2: the
claim that the evaluative truths and the adaptive beliefs come apart—that
there isn’t any helpful overlap between these two sets. Why should the real-
ist accept this? Can’t she point to an apparently obvious overlap? Pain is bad,
survival is good, and these are exactly the things evolution tracks! It may
not track the evaluative truth directly, but evolution tracks it indirectly, by
selecting for features with which it correlates (cf. Parfit et al.).
Street hoped to block this move. Our beliefs that pain is bad and survival
is good are exactly the sorts of beliefs we would expect evolution to lead us
to, whether or not they were true. A legitimate response to the debunker’s
challenge, Street argued, cannot just assume the very things called into
question. We must set aside the suspect beliefs and independently evaluate
whether we have good reason to think we are mistaken.
The problem here is that our entire body of moral beliefs is suspect. It
follows that we must set all of our moral beliefs aside, if we are to block such
question-begging responses. We cannot, then, simply assume that we have
92 Katia Vavova

reason to avoid pain—that morality is about what is good for us, and that
needlessly throwing ourselves off of cliffs just isn’t that sort of thing. These
assumptions aren’t appropriately independent. Taking them for granted
threatens to stack the deck against the debunker. I will now argue, how-
ever, that taking these assumptions off the table threatens to undermine the
debunker’s argument.
Recall that we are meant to be getting good reason to think that we
are mistaken about morality. But we cannot determine if we are likely to
be mistaken about morality if we can make no assumptions at all about
what morality is like. I argued that the debunker’s challenge threatens any-
one who holds that the attitude-independent moral truths do not, in any
helpful way, coincide with the evolutionarily advantageous beliefs—anyone
who accepts P2. But even to make this crucial judgment, that these two
sets do not have the same contents, we need to know something about the
contents of those sets—what they are or what they are like.
Compare:  I  cannot demonstrate that I  am not hopeless at interacting
with external objects in my manifest surroundings without knowing some-
thing about what those objects and surroundings are like. Likewise, I can-
not show that I am not hopeless at understanding right and wrong without
being allowed to make some assumptions about what is right and wrong.
If we can make no moral assumptions, then we cannot get P2: the claim
that the true evaluative beliefs and the adaptive evaluative beliefs come
apart. Now, I think P2 is plausible, and probably you do too. Certainly any
realist should believe it. However, we find P2 plausible against the back-
ground of our substantive moral beliefs. For example, we believe it is wrong
to discriminate against someone on the basis of race. At the same time,
there are evolutionary explanations of racism, on which it is adaptive to be
suspicious of those who do not look like you. In this case, then, the adap-
tive belief and the true moral belief come apart. Thus, to believe P2, one
must also believe that the evaluative beliefs are such-and-such, while the
evaluative truths are this-and-that. But if we cannot take for granted any of
our beliefs about the evaluative truths, then we cannot infer that they come
apart from the adaptive beliefs.
Again the debunker faces a dilemma. She may relax her standards for
what counts as a “good” reason, or she may maintain them. If she relaxes
them, she cannot give us good reason to think we are mistaken. Worse yet,
if we are permitted to assume that pain is bad, etc., then we can give her
good reason to think we are not mistaken and her purportedly undermining
story vindicates our evaluative beliefs.
If, instead, the debunker maintains her standards, she blocks such
responses. But she also blocks herself. If we cannot make any moral
assumptions—not even that pain is bad—then morality could be about
 Debunking Evolutionary Debunking 93

anything.28 To hold that the moral truths do not coincide with the adap-
tive judgments, we must assume something about what those moral truths
are, or are like. If we may assume nothing about morality, then morality
could be about anything. And if morality could be about anything, then
we have no idea what morality is about. So we have no reason to think that
the attitude-independent truths and the adaptive beliefs don’t overlap. But
without that, we have no sense of what the chances are that we are mistaken.
Therefore, we cannot get to the conclusion that we probably are mistaken.29
Not, at least, via an evolutionary story.

7.  Debunking Deontology

The third debunking argument aims to undermine neither realism nor our
entire body of moral beliefs. It targets a restricted class of those beliefs: those
based on deontological intuitions.30 This should be the most promising
argument yet. Leaving intact most of our belief system gives this debunker
an abundance of resources with which to construct her challenge.
Unfortunately, this debunker’s evolutionary story is either idle or too
strong. On the first point: worries about the targeted intuitions arise inde-
pendently and are not worsened when supplemented with an origin story.
On the second point: even if we lack other reason to worry, we should be
reluctant to rely on an evolutionary story. It just isn’t selective enough.
But first, the argument. It begins with a sociological observation: most
think it permissible to divert a trolley away from five people toward one,
but impermissible to push one in front of a trolley to save five. Why the
discrepancy? We are killing one person in both cases, after all. The answer,
of course, is evolutionary. Pushing the one, rather than diverting the trol-
ley onto the one involves “up close and personal” violence of the sort that,
unlike button pushing or lever pulling, has been around for a long time
(Greene 2008: 43). Evolution selects for negative responses to this direct
way of killing; it doesn’t select for similarly negative responses to more indi-
rect ways of killing. But the fact that “I have killed someone in a way that
28
 You might worry here that we are even talking about morality any more. The
debunker assumes that morality really could be about anything—it is conceptually poss­
ible that morality is about throwing ourselves off of cliffs and causing each other pain. I’m
not so sure about this. Cuneo and Shafer-Landau (MS) argue that some of the very basic
moral claims (like that pain is bad) are conceptual truths: if we don’t have them we don’t
have our concept of “morality.” This seems right to me, but I won’t explore it further here.
29
  I expand on this discussion in my MS b.
30
  Here I  follow Greene:  deontological judgments are those “in favor of character-
istically deontological conclusions (e.g. ‘It’s wrong despite the benefits’),” and mutatis
mutandis for consequentialist judgments (2008: 39).
94 Katia Vavova

was possible a million years ago, rather than in a way that became possible
only two hundred years ago” is morally irrelevant (Singer 2005:  348). If
our deontological intuitions have this suspect origin, then we should worry
about the beliefs we rest upon them. They are likely to be mistaken. The
debunker concludes that we can only trust our utilitarian judgments, which
come from our uncontaminated “rational intuition” (Singer 2005: 350–1).
Two questions arise for this debunker. First, did we need an evolution-
ary story to make us worry about these particular intuitions? Second, why
should we think that our consequentialist intuitions are less suspect?
On the first point. It is true that we feel a greater pull to help the nearby
needy than the distant needy. Greene says: “the only reason that faraway
children fail to push our emotional buttons is that we evolved in an envi-
ronment in which it was impossible to interact with faraway individuals”
(2008:  76). This should make us uncomfortable, he argues, if we think
we are justified in ignoring the distant needy. For it was just an accident
of evolution that we are “emotionally insensitive to their plight” (Greene
2008: 76).
Recognizing that we are emotionally responsive to only nearby suffering
should worry us, but for more familiar reasons.31 What, after all, is the moral
difference between the drowning child in front of you and the starving child
across the world? Our intuitive judgment that we may be selectively altru-
istic in these ways is already under pressure in the same way our judgments
about trolley cases are under pressure. Try as we might, we can’t seem to
find satisfying reasons for these diverging judgments. But this problem is a
distinctly moral one. We can recognize it from the armchair—no empirical
origin story is necessary.
Likewise with the intuition that it is impermissible to push one person
off of a bridge to save five. Many of us feel this quite strongly. Many of us
also believe that there is no morally significant difference between killing
by lever-pulling and killing by person-pushing. Recognizing that we are
making a distinction without a difference should already make us quite
confident that we are making a mistake. Furthermore, we are rightly more
confident in this judgment than we are in any origin story. Learning, then,
that we evolved to make this distinction without a difference shouldn’t fur-
ther increase our confidence that we are making a mistake:  it is already
maxed out. The evolutionary story is thus, at best, an idle premise in this
argument.32
Why doesn’t this just mean that this debunker is lucky—that the under-
mining of our deontological beliefs is overdetermined? This is where the

31
  Singer (1972).
32
  Some of these considerations reinforce similar points made by Berker (2009).
 Debunking Evolutionary Debunking 95

second point comes in. Suppose we lack this other reason to worry, so the
full weight of the conclusion falls on the evolutionary story. Wouldn’t we
then have good, evolution given, reason to worry about our deontological
beliefs? Only if the evolutionary story could be employed selectively against
only our deontological intuitions, and it cannot.
First, there isn’t an uncontroversial evolutionary explanation of our altru-
istic tendencies.33 Without one we cannot accept this debunker’s influence
claim. Second, if we had such an explanation, it should debunk more than
our belief in selective altruism: “if a disposition to partial altruism was itself
selected by evolution, then the epistemic status of its reasoned extension
[impartial altruism of the sort utilitarians promote] should also be suspect”
(Kahane 2011: 119). This echoes a claim Street (2006) makes for a differ-
ent purpose: that rational reflection cannot correct for evolution’s unsavory
influence. If our most basic moral intuitions are infected, and they are the
starting points for our moral reasoning, then any result of that reasoning
will also be infected. Even Greene recognizes this point, succinctly putting
it thus: “garbage in, garbage out” (2008: 116). Why, then, isn’t he worried?
Surely he should be. For if these considerations are right, then Greene’s
argument targets utilitarian beliefs too. It thus collapses into the previous,
more ambitious argument. This is bad for two reasons. First, it is no longer
an argument against our deontological beliefs only, as it was intended to be.
Second, as an argument against all of our moral beliefs, it suffers from the
same trouble as the previous one.
This debunker thus faces a different dilemma. She either relies on an evo-
lutionary claim or she doesn’t. If she doesn’t, she accomplishes nothing new
philosophically. She merely reiterates one side of the same old debates about
drowning children and runaway trolleys. If the debunker instead decides to
rely on an evolutionary claim, she presents a new argument, but it isn’t a
good one. The considerations she cites undermine more than she intended,
and the argument collapses into the more ambitious and less promising
one. This is why the evolutionary bit is, at best, idle, and at worst, too pow-
erful, for this debunker’s purposes.

8.  Some General Lessons

I argued that we have reason to worry about each of the available evolution-
ary debunking arguments. This doesn’t show that evidence of an off-track
influence could never give us good reason to think we are mistaken. On the

  Cf. Okasha (2009).

33
96 Katia Vavova

contrary, I think it often can.34 What the foregoing shows is the limits of
such arguments, and so, more generally, the limits of our ability to get evi-
dence of our own error. These limits are not of our cognitive architecture.
Our ability to acquire evidence of our own error is not limited because we
are, say, bad at recognizing such evidence. These limits arise out of the way
that such evidence works. Let me explain by extracting two lessons from
the above:  one about debunking arguments and one about undermining
evidence more generally.

8.1  Is Evolution the Problem?

I argued that the most modest of the available debunking arguments
fails: that insofar as it poses an epistemic threat it does so without an evolu-
tionary story and adding one threatens to weaken it. I now want to suggest
that this situation will be quite typical of evolutionary debunking argu-
ments more generally.
Consider a debunking argument that should work, if the empirical facts
are as they need to be. Studies show that, on the basis of only CV perusal,
people are much more likely to favor job candidates with stereotypically
white names over candidates with stereotypically black names.35 This is
so even though the CVs are otherwise identical:  same content, different
names. If these studies are right, then we have reason to worry about our
judgments in such situations.
We could try to give an evolutionary explanation to diagnose the worry.
Whether it works depends on the empirical details. Is there a sufficiently
selective evolutionary story available? Notice, however, that we needn’t
investigate this. We can establish a problematic pattern of CV selection
much more easily than we can establish the influence of evolutionary forces
on these particular beliefs. The empirical details are far less clear and the
debunking is far less promising in the evolutionary case. Given that we do
not take a candidate’s name or race to be a relevant hiring consideration,
this pattern of selection is sufficient cause for concern. Given everything else
we believe, we have good reason to think that we are making a distinction
without a difference here. Insofar as we can control our confidence in the
quality of this candidate over that one, we should revise it when it fits this
troublesome pattern. If we cannot revise it, then we should judge our con-
fidence that Emily is better suited for this job than Lakisha to be less than
fully rational and take whatever correcting measures we can. We have good
independent reason to think we are mistaken.

  See my MS a.
34

  Bertrand and Mullainathan (2004).

35
 Debunking Evolutionary Debunking 97

The upshot here is that both of these more modest debunkers are better
off without the evolutionary bit of their arguments. This will be typical
for evolutionary debunkers. Their stories just cannot provide an appropri-
ately selective argument that targets all and only the intended beliefs. This
might not be surprising. It can seem as if an evolutionary story can be told
about any of our beliefs. Advances in the relatively young field of evolution-
ary psychology might change this. Currently, however, things do not look
promising for the evolutionary debunker. She is better off dropping the
evolutionary story altogether.

8.2  On Undermining
I have argued that all three debunking arguments fail to give us reason to
worry about our beliefs. This assumes, of course, that the burden is on the
debunker to give us such a reason to revise our beliefs. It also assumes that
rational belief revision works a certain way. I should make this background
picture explicit, though it is not controversial. Two minimal assumptions
guide the foregoing thoughts:
A1. A reason is a reason, and evidence is evidence, only against a backdrop
of beliefs we take for granted.
Consider: the sound of water drops on my office window is typically evi-
dence that it is raining. Suppose, however, I believe that the college gardener
is out to get me, so he regularly aims the garden hose at my window in the
hope of flooding my office. Water drops on my window, relative to these
background beliefs, is not evidence of rain. It is evidence that the gardener
is at it again.
The second assumption is this:
A2. The undermining power of a reason or a piece of evidence is not
all-or-nothing.
Hearing a trusted colleague say that the gardener is in the shed is a good
reason to think that the gardener is in the shed. But seeing him there, with
my own eyes, putting the hose away, may be a better reason to think so.
Just as you can get various strengths of reasons for thinking that the
gardener is out in the shed, you can get varying strengths of reasons for
thinking that you are mistaken about some p. The stronger your reason
for thinking that you are mistaken, the more substantial revision you will
probably have to make. The strength of this reason will depend on what
you have to go on. This ties in with the first assumption: the more substan-
tial the body of beliefs you can take for granted, the more potential you’ll
have for getting a good reason. If, for example, you cannot take for granted
98 Katia Vavova

the trustworthiness of testimony, you won’t be able to get testimonial evi-

dence that the gardener is out in the shed. Things would be worse yet if you
couldn’t take for granted the trustworthiness of your own eyes and ears, or
even your powers of reasoning.
These observations about how evidence works apply to the evidence the
debunker presents in the following way. Evidence of error is a piece of evi-
dence like any other: the better ground you have from which to evaluate the
evidence, the more potential you’ll have for getting good reason to revise.
This is important because, recall, the debunker requires you to set aside the
targeted beliefs when evaluating her challenge. You must not take those for
granted if you are to avoid stacking the deck in your own favor. The more
substantial the body of beliefs that the debunker calls into question then,
the less substantial your independent ground will be. And the less substan-
tial the independent ground is, the worse the resources for both presenting
and evaluating evidence of error.
Notice, however, that having many uncontested beliefs does not guar-
antee having “substantial” independent ground, since these beliefs might
be quite superficial or otherwise irrelevant. Perhaps I have memorized the
phonebook, so that I have a large number of true beliefs. These are useless
for determining if I am likely to be mistaken about evaluative matters. Nor
is this independent ground of help to the debunker: she cannot make her
case on phone numbers alone. What determines whether we have good
ground from which to evaluate the debunker’s challenge is thus not how
many beliefs are appropriately independent, but whether the right sorts of
beliefs are appropriately independent. These include, at least, beliefs about
rationality, evidence, and belief revision.
These assumptions about evidential support combined with the good
principle of belief revision suggest something like the following rule
of thumb.

The Inverse Rule of Debunking. The potential strength of a debunking

argument is inversely proportional to its ambition.

The “strength” of the debunking argument has to do with how extensively

we must revise. The “ambition” of the debunker’s argument refers to how
much it targets.
Again, what matters in determining the strength of a debunking argu-
ment, or the undermining evidence it provides, is not how many of your
beliefs it calls into question but whether it leaves you enough of the right
sorts of beliefs with which to evaluate the evidence that has been put before
you. What matters for the ambition of the argument is also thus not how
 Debunking Evolutionary Debunking 99

many beliefs it calls into question, but whether it calls the right sorts of
beliefs into question, namely:  those necessary for evaluating the relevant
evidence.
This rule of thumb thus issues in the following prediction. The most
ambitious debunker, she who aims to undermine all we believe, has the
lowest chance of success. The evidence she aims to give has no undermin-
ing potential. This is because, by calling all of our beliefs into question, she
leaves us nothing with which to question.
A moderately ambitious debunking argument may be able to cast some
doubt on the targeted beliefs. The extent to which it succeeds depends on
what exactly is called into question, and thus how substantive are the rel-
evant independent grounds. The debunker who aims to debunk your most
fundamental beliefs, those on which everything else you believe rests and
with which you judge what to believe—she is out of luck. Her challenge
just is the ambitious one in disguise. If she aims at a fairly superficial set of
your beliefs, she has a decent chance of undermining them.
The most modest debunker thus has the best prospects. She aims low,
but she may score high. This result makes sense. A comparison might help
show this.
Consider three disagreements. You disagree with Anne about the permis-
sibility of abortion. You agree on other moral and political matters. You
disagree with Beth about the permissibility of abortion, but also about a
myriad of other moral and political matters. You disagree with Clarisse
about the permissibility of abortion. But you also disagree about every other
moral matter. Clarisse is a psychopath.
Your disagreement with Anne has the most undermining potential; your
disagreement with Clarisse the least. Generally, the more common ground
you share with someone, the more significant their disagreement may be.
This is because the more common ground you share, the more independent
ground you have from which to get evidence of your error. You have much
independent ground on which to evaluate your disagreement with Anne;
you have none with Clarisse.
Since evidence of a suspect belief influence is also evidence of error,
we should expect the same pattern:  evidence of a more pervasive belief
influence should be less worrying. This may seem counterintuitive, but
it is a good result. It makes sense that the most modest, targeted sort
of debunking argument should be the most effective, if it works. That
sort of argument provides me with good reason to think I  am wrong
about some p. This good reason is good by my own lights: it follows from
my other beliefs about reasons and evidence. This is the kind of reason
the debunker must provide. It is also, I  have argued, the kind that the
100 Katia Vavova

distinctly evolutionary debunker cannot provide. This is good news for

the realist. Whatever her epistemic troubles, this scientifically grounded
one is not of them.36

References
Bedke, M. MS. “No Coincidence?” draft for Wisconsin Metaethics Workshop 2012.
See Chapter 5, this volume.
Berker, S. 2009. “The Normative Insignificance of Neuroscience,” Philosophy and
Public Affairs, 37(4): 293–329.
Bertrand, M., and Mullainathan, S. 2004. “Are Emily and Greg More Employable
than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination,”
American Economic Review, 94(4): 991–1013.
Christensen, D. 2007. “Epistemology of Disagreement:  The Good News,”
Philosophical Review, 116(2): 187–217.
Christensen, D. 2011. “Disagreement, Question-Begging and Epistemic
Self-Criticism,” Philosophers Imprint, 6: 1–22.
Cuneo, T., and Shafer-Landau, R. MS. “The Moral Fixed Points.”
Darwin, C. 1871. The Works of Charles Darwin, xxi. Albany, NY: NYU Press, 2010.
Dworkin, R. 1996. “Objectivity and Truth: You’d Better Believe it,” Philosophy and
Public Affairs, 25(2): 87–139.
Dworkin, R. 2011. Justice for Hedgehogs. Cambridge, MA: Harvard University Press.
Elga, A. MS. “Lucky to be Rational,” draft as of 6 June 2008: <http://www.prince
ton.edu/~adame/papers/bellingham-lucky.pdf>.
Elga, A. 2007. “Reflection and Disagreement,” Noûs, 41(3): 478–502.
Enoch, D. 2010. “The Epistemological Challenge to Metanormative Realism: How
Best to Understand it, and How to Cope with it,” Philosophical Studies,
148(3): 413–38.
Field, H. MS a. “The Revision of Logical Laws.”

36
  Ancestors of this chapter were given at the First Biennial Mentoring Workshop
for Women in Philosophy, the University of Sydney, the Australian National University,
the Australian Association of Philosophy 2012 meeting, the Ninth Annual Wisconsin
Metaethics Workshop, Union College, and the University of Edinburgh. Thanks to the
organizers for the opportunity to share my work and to the audiences for their help-
ful comments. Many thanks also to John Bengson, Lee Bowie, Sarah Buss, David
Christensen, Terence Cuneo, Tom Dougherty, Adam Elga, Catherine Elgin, Jyl Gentzler,
Alexander George, Daniel Greco, Toby Handfield, Caspar Hare, James Harold, Justin
Horn, Sophie Horowitz, Hilary Kornblith, Dustin Locke, Elisa Mai, Sam Mitchell,
Joseph Moore, Alejandro Pérez Carballo, David Plunkett, Paolo Santorio, Nishi Shah,
Paulina Sliwa, Judith Jarvis Thomson, Jonathan Vogel, Kenneth Walden, Angela Ruohan
Wang, Tom Wartenberg, Roger White, two anonymous referees from Oxford University
Press, and the students in my Mount Holyoke College seminar on moral epistemology.
Thanks especially to Nishi Shah for many conversations on this topic, including those
from which this chapter emerged.
 Debunking Evolutionary Debunking 101

Field, H. MS b.  “Logic, Normativity, and Rational Revisability,” from his 2008
Locke Lectures.
FitzPatrick, W. Forthcoming. “Debunking Evolutionary Debunking of Ethical
Realism,” Philosophical Studies.
Greene, J. 2008. “The Secret Joke of Kant’s Soul,” in W. Sinnott-Armstrong (ed.),
Moral Psychology, iii. The Neuroscience of Morality: Emotion, Brain Disorders, and
Development, 35–79. Cambridge, MA: MIT Press.
Harman, G. 1973. Thought. Princeton: Princeton University Press.
Kahane, G. 2011. “Evolutionary Debunking Arguments,” Noûs, 14(1): 103–25.
Kelly, T. 2005. “The Epistemic Significance of Disagreement,” in J. Hawthorne and
T. Gendler Szabo (eds), Oxford Studies in Epistemology, i. 167–96. Oxford: Oxford
University Press.
Kripke, S. 2011. “On Two Paradoxes of Knowledge,” in Philosophical Troubles:
Collected Papers, i. 27–51. Oxford: Oxford University Press.
Lewis, D. 1971. “Immodest Inductive Methods,” Philosophy of Science, 38(1): 54–63.
Okasha, S. 2009. “Biological Altruism,” in The Stanford Encyclopedia of Philosophy:
<http://plato.stanford.edu/archives/win2009/entries/altruism-biological/>.
Parfit, D. 2011. On What Matters, ii. Oxford: Oxford University Press.
Quine, W. 1969. “Natural Kinds,” in Ontological Relativity and Other Essays, 114–
38. New York: Columbia University Press.
Shafer-Landau, R. 2005. Moral Realism: A Defence. Oxford: Oxford University Press.
Singer, P. 1972. “Famine, Affluence, and Morality,” Philosophy and Public Affairs,
1(3): 229–43.
Singer, P. 2005. “Ethics and Intuitions,” Journal of Ethics, 9(3–4): 331–52.
Sliwa, P. MS. “Darwin and the Moral Realist.”
Street, S. MS. “Objectivity and Truth: You’d Better Rethink it.”
Street, S. 2006. “A Darwinian Dilemma for Realist Theories of Value,” Philosophical
Studies, 127: 109–66.
Street, S. 2009. “Evolution and the Normativity of Epistemic Reasons,” Canadian
Journal of Philosophy, 39, supplement 1: 213–48.
Street, S. Forthcoming. “Does Anything Really Matter or Did We Just Evolve to
Think So?” in A. Byrne, J. Cohen, G. Rosen, and S. Shiffrin (eds), The Norton
Introduction to Philosophy. New York: W. W. Norton.
Vavova, K. MS a. “Irrelevant Influences.”
Vavova, K. MS b. “A Dilemma for the Darwinian Debunker.”
Weatherson, B. MS. “Do Judgments Screen Evidence?”
White, R. 2010. “You Just Believe That Because . . .,” Philosophical Perspectives,
24(1): 573–615.
Wielenberg, E. 2010. “On the Evolutionary Debunking of Morality,” Ethics,
120(3): 441–64.
Wittgenstein, L. 1969. On Certainty, ed. G. E.  M. Anscombe and G. H. von
Wright. Oxford: Basil Blackwell.
Wright, C. 2004. “Wittgensteinian Certainties,” in D. McManus (ed.), Wittgenstein
and Scepticism, 22–54. London: Routledge.
 5
No Coincidence?*
Matthew Bedke

1. Introduction

I will be discussing the following Coincidence Argument.

(1) Normative beliefs are about sui generis, causally inefficacious, norma-
tive facts. (Non-Natural Purport1 )
(2) Causal [/Evolutionary] forces pushed us toward forming our norma-
tive beliefs and having the justifying grounds we have for those beliefs,
but not because those beliefs represented any normative facts.2 (Lazy
Normative Facts)
(3) There are many conceptually possible arrangements of non-natural,
normative facts, including the absence of any, that are consistent with
the causal[/evolutionary] facts and their influence on normative beliefs
and their justifiers. (Many Conceptual Possibilities)

*
  Thanks to the audience at the Wisconsin Metaethics Workshop for helpful com-
ments. Special thanks to Justin Clarke-Doane, David Silver, Chris Stephens, Terence
Cuneo, and David Copp for helping the project along.
1
  This is meant to preclude non-cognitivist treatments of the discourse. Whether it
includes “non-metaphysicalist” views like Parfit’s is trickier. I think that it does so long as
quietism is not a veiled form of non-cognitivism.
2
  The “because” clauses are causal-explanatory. The justifying grounds I have in mind
are intuitions and the beliefs that feed into reflective equilibrium, though the Argument
is not wedded to this epistemology. Whatever we base our normative beliefs on (in virtue
of which they are meant to be justified), the explanations for why we have those bases do
not aver to the truths of the normative beliefs they putatively support. Also, this premise
basically grants that one horn of Street’s Darwinian Dilemma is least desirable, namely,
the one that holds that evolutionary forces pushed us toward certain normative beliefs
because those beliefs are true. Non-naturalists should be with me on this.
 No Coincidence? 103

(4) In relatively few of the possibilities in (3) do our normative beliefs rep-
resent normative facts (if such there be).3 (Rare Alignment)
(5) If (1), (2), (3), and (4), it would be an epistemic coincidence were we
caused to form normative beliefs that represent the normative facts.
(Conditional Coincidence)
(6) It would be an epistemic coincidence were we caused to form norma-
tive beliefs that represent the normative facts. (Coincidence)
The epistemic upshot is meant to be this:  when the Coincidence
Argument is seen to apply to some set of our normative beliefs,
Coincidence defeats any justification antecedently enjoyed by those
beliefs. I take the Argument to roughly capture what exercised me in my
(2009) and to perhaps more roughly capture the worries that Gibbard
(2003:  ch. 13), Joyce (2001, 2006, forthcoming), Ruse (1986), Street
(2006, 2008), and others are keen to push, and Enoch (2011:  ch. 7),
Fitzpatrick (forthcoming a, b), Huemer (2005: 214–19), Kahane (2010),
Schafer (2010), Shafer-Landau (2012), Skarsaune (2011), Wielenberg
(2010), and others are keen to resist.
Of course, premises (1)–(4) are open to debate. But it is striking that
those who accept these premises do not agree on whether a defeater
threatens. Both sides try to make their case largely through metaphor and
an­alogy, where skeptics suggest partners in crime with clear cases of epis-
temic defeat, and non-skeptics suggest innocence by association with clear
cases of epistemic acquittal. My aspiration is to sort this out. This chapter
is an extended attempt to see clearly what might be epistemically troubling
about (1)–(4).
Let me proceed as follows. First, I will comment on the Argument and
how it is related to similar arguments in the vicinity. Second, I will con-
sider various attempts to bring the threat of epistemic defeat into sharper
relief. Most of those attempts will be found wanting. In section 4, however,
I articulate a principle—obliviousness—that does a better job. The problem
with (1)–(4) is that they make normative beliefs oblivious to the normative
facts (if such there be), where obliviousness is something like insensitivity
of belief, justification, and explanation to fact.

3
  This basic idea has been put in terms of belief-forming processes failing to be reli-
able, beliefs failing to track the truth, or sheer coincidence of belief and fact, but all of
these rely on some range of alternatives against which the evaluations of reliability, track-
ing, and coincidence are made.
104 Matthew Bedke

2.  The Coincidence Argument and

Coincidence Arguments

I have formulated the Argument in terms of my preferred “cosmic coin-

cidence” version of it. The bracketed alternative sticks to evolutionary
debunking arguments that are more common in the literature. The evo-
lutionary trimmings are inessential, however, for even if there are no sig-
nificant evolutionary influences at play for some normative beliefs,4 there
will still be some complete causal explanation for why we have or tend to
have the normative beliefs that we do. This explanation does not aver to
non-natural normative facts, for those facts have no causal powers. That
is part of the point of calling them non-natural. And noting this should
generate as much a problem for normative non-naturalism as does an evo-
lutionary debunking argument. Even better, it need not rely on potentially
controversial evolutionary explanations.
The Argument is more encompassing than some other debunking argu-
ments. Some debunking arguments are aimed at moral beliefs. But the basic
worry easily generalizes to all normative beliefs insofar as they concern
non-natural matters. Similarly, Joshua Greene and Peter Singer have lately
offered selective debunking arguments that target deontology-friendly nor-
mative judgments (see Greene et  al. 2008; Singer 2005). They typically
rely on premises about what differences are morally relevant differences, or
premises about the inferiority of judgments when and because they issue
from certain cognitive processes, and argue that deontology-friendly judg-
ments are sensitive to morally irrelevant differences, or issue from suspect
cognitive processes. The Argument is not so narrowly focused. It takes aim
at consequentialism-friendly beliefs and indeed all other normative beliefs
insofar as they concern non-natural matters.
Yet other debunking arguments hold that all substantive normative
beliefs arise by processes, or in contexts, that are generally distorting or
contaminating (Sinnott-Armstrong 2007; Street 2006). That sort of argu-
ment is fairly encompassing, but it fails to cast Non-Natural Purport, Lazy

4
  See Parfit (2011: 534–42). He thinks that the belief that P’s likely truth is a reason to
believe that P is no more advantageous than the belief that P is likely true, for the latter
suffices to produce the belief that P. He also thinks it was not advantageous to believe that
we have reasons to promote the survival and hedonic well-being of ourselves and our children,
to avoid agony, etc., for we are sufficiently motivated to do these things without the aid
of beliefs about reasons for them. Regarding the Golden Rule, he thinks it is hard to see
how evolutionary forces helped to instill belief in it. Parfit is probably not considering all
the relevant selective pressures. Be that as it may, the more general cosmic coincidence
worry is immune to such controversies.
 No Coincidence? 105

Normative Facts, Many Conceptual Possibilities, and Rare Alignment in

leading roles.
Some moral epistemologists seem to endorse something like the
Argument, but occasionally try to rely on fewer premises. Sharon Street, for
example, does not talk of conceptually possible normative facts in her 2006
paper except to say it is conceptually possible that pain counts in favor of
that which causes it (2006: 148). In her 2008 paper, however, the range of
conceptually possible arrangements of normative fact plays a central role in
her summary of the Darwinian Dilemma. There, what is too coincidental is
that “evolutionary pressures affected our evaluative attitudes in such a way
that they just happened to land on or near the true normative views among
all the conceptually possible ones” (2006: 208–9).
Richard Joyce (forthcoming) is less sanguine about appeal to a range of
possibilities. He does not think “counterpossibilities” are needed to estab-
lish the claim that a process is not truth-tracking. He simply emphasizes
the claim that evolutionary explanations for our normative beliefs and
belief-forming processes do not aver to normative facts. For him, this lack of
explanatory role seems to suffice to establish the absence of truth-tracking.
So he sees a quicker route to our conclusion that only goes through Lazy
Normative Facts. I do not. Showing that truth-tracking wasn’t selected for
is not yet to show that normative beliefs actually fail to track truth, that it
would be improbable (coincidental) were they to represent the facts, that
the processes producing our beliefs do not reliably output true beliefs, or
some such.5 As Joyce himself notes, the causal (or evolutionary) stories by
themselves are silent about what the normative facts are—their explanations
do not aver to real normative facts—and so they are silent on whether or not
evolutionary or causal forces have distorted our normative beliefs or pushed
them toward conformity. For this reason we should be careful when we
say that normative beliefs are the product of a non-truth-tracking process.
This is ambiguous between the claim that the process has been shown to be
unreliable and the claim that causal explanations for the process do not aver
to the target facts. Our premises only make use of the second claim.
Sometimes moral epistemologists drop reference to non-natural norma-
tive facts (Street 2006) or indeed to real facts (Joyce forthcoming). Though
one might try to construct a coincidence-type argument that applies to
normative naturalism or irrealist views, non-naturalism is a nice test case for
seeing whether there is an epistemic problem at all. Non-naturalism makes
our job easier by separating out the normative facts from the facts that could
enter into causal explanations, so we can focus on the relationship between
(1)–(4) and an alleged epistemic defeater. If we can bring the defeater into

5
  See also Copp (2008: 194–6).
106 Matthew Bedke

relief, then others are free to craft analogous arguments for these other
meta-normative positions. I am skeptical that the sort of defeater at play
for non-naturalists extends to other meta-normative views. But my main
concern is with non-naturalism as a test case.
One last remark on the Argument. I have stated its epistemic significance
in a way that is friendly to epistemic internalism, which we can gloss as
the view that the justificatory status of S’s belief that P strongly supervenes
on S’s mental states. More specifically, I will be talking about an epistemic
assessment of how well one proceeds in making up one’s mind based on
the information to hand.6 We are all trying to make up our minds about
normative matters, and meta-normative matters, and there are certain con-
siderations we have to go on. If there is something about the world that
is in no way accessible to us as we make up our minds, we should let the
external chips fall where they may. We can still get our houses in order. If,
on the other hand, using information accessible to us we can show that
our normative beliefs are at best coincidentally true, or unlikely to track
the non-natural facts, or at best inexplicably track them, or some such,
we could only make up our minds in a procedurally justified manner by
revising some of the beliefs that generate the difficulty. The possibility that
procedurally unjustified beliefs might enjoy some other kind of justifica-
tion is cold comfort. So what follows focuses on this kind of internalist,
procedural justification, though perhaps similar worries can be cashed out
in more externalism-friendly ways.
With all this in mind, let me turn to explore some opening moves in the
debate.

3. Opening Moves

I think we can put the initial worry this way. Causal forces would push us
toward the same normative beliefs, and would push us toward having the
same justifying grounds for those beliefs, regardless of what the normative
facts turn out to be. If so, it would be coincidental should those forces hap-
pen to push us toward accurate representation of whichever normative facts
turn out to be actual. Just this much gets me worried.
Others try to ease my distress by appealing to some normative facts.
They argue that, given that the normative facts are thus and such, it is no

6
  I won’t comment on the tricky issues surrounding internalism here. Clearly, mental
states are often about things external to the mind, and the information we go on is often
information about things external to the mind. So it is not so easy to draw a line between
internal items relevant to justification and external items that are not.
 No Coincidence? 107

coincidence that some of our beliefs represent them. The salve is roughly
this: Why talk about what the normative facts could be, and the number of
conceptually possible arrangements, when we have justified beliefs about
what they are? David Enoch says that, given that survival and reproduc-
tive success are good, we can explain why beliefs that they are good would
non-coincidentally correlate with the facts, for it looks like evolution-
ary forces would have pushed us in the direction of having such beliefs
(2011: 168–75). Erik Wielenberg says that if people with certain cognitive
processes have rights, we can explain why we know we have such rights, for
evolutionary forces would have pushed us in the direction of having the cog-
nitive processes needed to be rights-bearers, and such processes would have
led us to believe that we have rights (2010: 447–52). And Knut Skarsaune
says that, if pleasure is good, we can explain why belief that it is good is
truth conducive, for evolutionary forces would have influenced us to have
this true belief (2011: 233–6). (Actually, he relies on a dilemmatic structure
to either save the realist in the above fashion, or to concede to the skeptic.)
In each case, certain normative facts would help to explain why beliefs
about those facts are not merely coincidentally correct. Enoch emphasizes
that no particular explanation given need be the one that discharges the
burden. So long as some explanation is available for the non-coincidental
correctness of normative beliefs the problem is (re)solved (2011:  171).
Ideally, there are several such explanations yielding a decent stock of justi-
fied normative beliefs, enough to ascend from there via rational inferences
to an even bigger set of justified normative beliefs and perhaps even norma-
tive theories.
It is at this point that you might wonder whether we are entitled to
rely on beliefs about what the normative facts are to get the relevant expla-
nations for non-coincidentality. It helps me to think through a familiar
an­alogy outside of normative theory. If we wonder whether our experiences
as of an external world are largely correct what we do is rely on experience,
and experience-based beliefs, to assuage our fears. We do think that evo-
lutionary forces, inter alia, have pushed us toward representation of facts
of the external world; we think that a large swath of such beliefs reliably
track truth. But these assurances are all built on the back of experience, and
experience-based beliefs. That is OK so long as we are prima facie justified
in relying on experience out of the starting gate, as it were. It can then play
a role in vindicating its deliverances, and there is nothing question-begging
about that vindication. Turning back to normativity, experience as of an
external world alone does not vindicate the thought that normative beliefs
adequately represent the normative facts. But do not some normative beliefs
enjoy prima facie justification just as some beliefs about the external world
do? If so, we can also rely on them out of the gate, just as we can rely
108 Matthew Bedke

on experience and experience-based beliefs out of the gate. So let us rely

on prima facie justified normative beliefs. In turn, we discover that causal
forces have pushed us toward representation of some normative facts. In
both cases, we do not wind up vindicating every belief about the external
world and every normative belief. And perhaps proceeding this way cannot
deliver universal debunking. But in both cases we find adequate representa-
tion of fact and the ability to further prune and revise.7
What shall we make of these opening volleys? It is as though proponents
of the Coincidence Argument—let us call them skeptics—maintain that we
must think each of some set of possible arrangements of normative fact
are equally epistemically likely unless there is reason for thinking otherwise.
They think we could break this symmetry if we had some causal explanation
for why fact and belief would adequately align. But we do not. If nothing
breaks the symmetry, adequate alignment between belief and fact is surely
coincidental. Opponents of the argument—let us call them realists—grant
that there is no purely causal explanation for why belief and fact would
adequately align. And they grant that there are lots of conceptually pos­
sible arrangements of normative fact. They point out, however, that we have
prima facie justification for believing that we are in some subset of all those
possibilities, a subset wherein the evolutionary or causal forces have ade-
quately pushed our beliefs toward alignment with fact. Ranging across this
subset, belief-forming processes are reliable enough, belief sufficiently tracks
truth, etc. For the realist, there is a mixed normative-causal explanation for
adequate alignment. And why not help ourselves to our prima facie justified
beliefs in the explanation, just as we help ourselves to beliefs justified by
experience when we generate causal explanations for the non-coincidental
correctness of perception?
I think this puts some pressure on skeptics to say more about why realists
are not entitled to rely on their prima facie justified normative beliefs to
locate them in an area of possibility space where there is adequate align-
ment. In the next few sub-sections, let me develop a couple of lines of argu-
ment of behalf of skeptics, and consider replies by realists. Things will be
looking pretty good for the realist here. It won’t be until section 4 develops

7
  Street (2008: 216–17) grants this, but wishes to draw a distinction between good
and no good accounts of reliability. I  am puzzled by her ensuing discussion. It seems
like good normative theory will help sort out which starting points are likely true, and
which belief-forming processes are reliable. I  have already cited Greene and Singer as
examples of people who start with prima facie justified normative beliefs and evolution-
ary theory to identify kinds of normative beliefs (deontology-friendly) as likely false, and
the processes that produce them as unreliable, while identifying other kinds of normative
beliefs (consequentialism-friendly) as likely true, and the processes that produce them as
reliable. I don’t mean to agree with their assessment, but just to point out a project that
parallels the partial vindication of experience by relying on experience.
 No Coincidence? 109

coincidence as obliviousness that we will see what is problematic about

premises (1)–(4).

3.1  Random or Unreliable Analogies

Here is one thing the skeptic might say. Despite the prima facie justification
of some normative beliefs, seeing that premises (1)–(4) hold is tantamount
to realizing one’s beliefs were generated randomly or unreliably. Drawing
out the thought, Street describes a case where you learn that your views
about Jupiter have been implanted in you by a hypnotist who picked them
out of a hat (2008: 214). In that case, one cannot justifiably rely on one’s
Jupiter beliefs to discover that the hat-drawing hypnotic process tracked
Jupiter truths. There are a lot of possible Jupiter facts, and it would be
too coincidental to suppose that the hat-drawing hypnotic process hap-
pened to align belief with fact. Similarly, Joyce describes a couple of cases
involving belief pills. In one, you learn that you took a pill that induced
the particular belief that Napoleon lost the battle of Waterloo (2006: 179).
In another, you learn that you took a pill that induced you to have beliefs
about Napoleon in general, where various environmental factors helped to
determine which Napoleon beliefs you formed, but where you would not
have any Napoleon beliefs at all without the pill (2006: 181). In both cases,
the relevant Napoleon beliefs have been defeated.
These are interesting cases, yet they are not clearly analogous to the
Argument. Street tells us that the hypnotist picked the Jupiter views out
of a hat, which most of us would reasonably take to be a random process,
or one generally known to be unreliable. Similarly, forming beliefs by pill
is reasonably taken to be random or unreliable. So the problem with using
these cases as damning analogies is that it might not be similarly clear that
the ex ante reasonable attitude toward the causal processes that influenced
normative beliefs is suspicion that they are random or generally unreliable.
To help see the point, consider a case where you learn that a book published
by Kendall Hunt induced your Jupiter and Napoleon beliefs. In that case we
would not assume that this is a random or generally unreliable way to form
beliefs. It is just historical information that is epistemically benign (at least
for those unfamiliar with Kendall Hunt; if that’s too loaded with epistemic
relevance for you, imagine the history is about the paper or the ink of the
book). The realist can reasonably wonder why the historical information
supplied by the Coincidence Argument is not similarly benign.
On this point, consider two kinds of skeptical case. In one case you real-
ize that it is possible that you are a brain in a vat with the same experi-
ences you now have, but where your beliefs are largely false. In another case
you realize that the entire population of earth has gone through a random
110 Matthew Bedke

selection procedure at birth to determine which half of the population will

be envatted (and handless) for the rest of their lives and which will roam
the earth unmolested. Now, maybe the first, purely modal, scenario induces
some skepticism. But the second, probabilistic, scenario is far worse. It
strikes me that Street’s Jupiter case and Joyce’s belief pill cases look more like
the second skeptical scenario, where we reasonably believe that randomiza-
tion or unreliability has intervened, whereas the Coincidence Argument
does not clearly introduce such processes and seems more analogous to
skepticism based on the mere possibility of envatment.
Having said that, I like the Jupiter and Napoleon cases, and I do think
they have features that generate the same coincidence concern as is found in
the Argument. It is just that these cases have extra features that give rise to
further epistemic difficulties. This might lead realists to reasonably dismiss
them as disanalogous and so non-probative of the normative situation. Let
me save further discussion of how these cases are relevant for section 4. As
we shall see, it is the fact that they feature oblivious beliefs that makes them
damning analogies, where obliviousness can be pulled apart from the pro-
cess that we ex ante reasonably believe to be random or unreliable. Before
getting to that, I want to discuss some other lines of argument available to
the skeptic.

3.2  Generic Skepticism?

Perhaps the skeptic should categorize the Coincidence Argument as an
instance of generic skepticism. For one way to argue that each conceptual
possibility is equally epistemically likely, despite putatively justified beliefs
that favor some possibilities over others, is to argue as follows. If your puta-
tive justification cannot discriminate between scenarios in that you would
have the same justification across the two scenarios, that putative justifi-
cation cannot favor some of those scenarios over others. In the classical
skeptical case the thought would be that the appearance as of having hands
cannot discriminate between the possibility that one is envatted (and not
handed) but made to have the appearance as of having hands, and the pos-
sibility that one has hands that reflect light into one’s eyes, etc., etc. . . . . So
these scenarios are equally epistemically likely. Neither is favored over the
other by one’s putative evidence of it appearing as though you have hands.
Mutatis mutandis, perhaps one’s putative prima facie justification for nor-
mative beliefs cannot discriminate among conceptually possible arrange-
ments of normative fact. The justifications are non-discriminatory because
we realize they are fixed by the causal ways of the world, so you would have
the same justification for your normative beliefs across the possible scenar-
ios where we hold the non-normative ways of the world fixed and vary the
 No Coincidence? 111

normative ways of the world. Perhaps one is thereby unjustified in believing

that some one of these possibilities obtains rather than others.
If that is the point, realists can reply to skeptics in the standard ways.
One could appeal to contextualist accounts to grant lack of justification
when skeptical possibilities are salient, but maintain justification when
they are not salient. This raises the interesting possibility that by making
Non-Naturalist Purport contextually salient one introduces possibilities
that need to be ruled out to have justified belief, just as raising BIV sce-
narios raises possibilities that then need to be ruled out. On such a view,
justified normative belief would ebb and flow between attainable and elu-
sive depending on whether the non-naturalist has her metaethics in view.
Alternatively, one could be Moorean, where agents that cannot discriminate
between scenarios (skeptical and non-skeptical) are justified in thinking
they are in a non-skeptical scenario. Some of these agents will be lucky in
that their external environment will cooperate while others will be unlucky
in that their external environment will not. But procedurally they are all
justified in thinking they are in the lucky scenario. Or so the thought goes.
Whatever the response, realists can relax if their normative beliefs are no
worse off than their beliefs about having hands.

3.3  Begging Questions?

So what if we let realists justifiably rely on some normative beliefs to explain
non-coincidental alignment between belief and fact? What if we leverage
these beliefs:  reproductive success is good, we have certain rights, pleas-
ure is good? Skeptics are then likely to vent some frustration. Joyce, for
instance, says that the above authors have “speculated,” “stipulated,” and
“conjectured” about the normative facts, whereas they need to make the
views “plausible” before they can debunk the debunkers (forthcoming).
Street says “It is no answer to this challenge simply to assume a large swath
of substantive views on how we have reason to live . . . and then note that
these are the very views evolutionary forces pushed us toward” (2008: 214).
Of course, skeptical frustration is misdirected if it fails to grant prima
facie (defeasible) justification for some normative beliefs. If they are not
prima facie justified, why the Coincidence Argument? Realists would lack
justification for their normative beliefs before the skeptic utters word one.
So I do not think we can charge the realist with speculation, stipulation,
conjecture, or the like until we convincingly establish defeat of that prima
facie justification.
In addition, the skeptics are at risk of arguing in circular fashion. For
it can look as though they assume one’s normative beliefs lack prima facie
justification (and hence cannot help locate one in possibility space) to argue
112 Matthew Bedke

one into a conclusion that is meant to count as a defeater for said prima
facie justification. Not a classical kind of circularity, but one that assumes
lack of justification to show lack of justification.8
Unfortunately, some realists seem to mischaracterize the nature of their
burden, and that gives the skeptics a false sense of security. Enoch suggests
that he is providing a defeater for a defeater by offering explanations of
non-coincidence that rely on normative premises (2011: 170 n. 41). But
you cannot defeat a defeater by relying on the defeated belief. That would
be like acknowledging that you are not justified in believing a wall is red
after learning it is bathed in red light half the time, but then enlisting your
belief that the wall is red, justified by how red it looks, to defeat the defeater.
We should not think of the replies on behalf of realists above and else-
where as attempts to defeat a defeater or reinstate justification. Instead, they
should be cast as expressions of incredulity that there is a defeater in the first
place. Unless realists can be made to see the defeater they are well justified
in relying on their (still) justified normative beliefs to account for various
ways in which their beliefs are non-coincidentally true. The response is not
to charge them with begging the question, but to show that there is indeed
a defeater there.9
Until that is done, realists might reasonably see the Coincidence
Argument as turning a blind eye to prima facie justification, as a prema-
ture refusal to let justified normative beliefs pare down possibility space,
as smuggling in suppressed and unjustified premises about randomness or
general unreliability, or as a recipe for generic skepticism. It is no wonder
they are not yet worried.

3.4  But Still . . . an Inexplicable Alignment?

So far we have seen explanations for why we would tend to believe P, for
some normative propositions P that we antecedently justifiably believe

8
  Cf. Schafer (2010).
9
  Another form of begging the question is purely dialectical and is not at issue. It is
the sort you get when you use premises your opponent does not justifiably share in an
attempt to convince him of some conclusion. As far as the Coincidence Argument is con-
cerned, the background project is for the realist to make up her mind about normative
non-naturalism and belief in particular normative propositions. The Argument is meant
to offer up some considerations that should make her retract her views in the face of
prima facie justification. When making up her mind she can use premises she is justified
in believing even if others do not share those views. (But, really, if we cleanly separate out
the first-order normative beliefs from the second-order metaphysical interpretation of
their contents and focus on the former, who thinks they lack prima facie justification for
some relevant normative proposition to which the realist is helping herself?)
 No Coincidence? 113

(pleasure is good, survival is good, etc.). Suppose we do this one-by-one for

many normative propositions P and theorize from there.
Perhaps there is a residual explanatory gap not yet addressed. For there is
this related, second question: Granting that we have managed to get these
normative facts right, would we tend to get the normative facts right as
such (Gibbard 2003: ch. 13)? Essentially, this is a demand that we explain
why our belief-forming processes would reliably track the normative truths
whatever they turn out to be (perhaps within a reasonable variation of possi-
bilities, but certainly across some possibilities we have no prima facie reason
to think actually obtain).10 The demand can be met in the case of percep-
tion of the external world. Not only is there actual alignment between many
beliefs and facts, but there would be alignment across a range of nearby
possibilities where the facts change a bit.
Before we ask whether the demand is met in the normative case, we
should ask whether it is a fair explanatory demand in the first place. I’m not
sure what to say. On the one hand, it is questionable that lack of explana-
tion for some kind of reliability is a defeater, for it is questionable that some
showing of reliability is needed for procedural justification. As noted above,
we do not need to show that perceptual experience, or the beliefs based on
them, reliably track truth before justifiably relying on them. More worri-
some would be a positive showing of unreliability, but the Argument does
not supply that.11
One way of developing Gibbard’s worry is not to demand an explana-
tion of reliability prior to having justified normative beliefs, but to demand
that the initial justified reliance on normative belief eventually lead to an
account of reliability. I like to think of this as probationary epistemology.
The basic idea is this: One can justifiably rely on basically justified beliefs
(a) until one comes to have justification for believing in their reliability after
sufficient inquiry, perhaps by appealing to the very beliefs enjoying proba-
tionary justification, in which case their justificatory status becomes secure,
or (b) until one fails to come across such justified beliefs about reliability
despite sufficient inquiry, in which case the probationary justification lapses
and the beliefs are no longer justified. If these are sound epistemic ideas,

10
  Regarding mathematical Platonism, Field has a similar worry that “how our beliefs
about . . . remote entities can so well reflect the facts about them” is in principle inexplic­
able (Field 2005; cf. Clarke-Doane forthcoming).
11
  Even if an explanation of reliability is needed, it is not clear how robust the reli-
ability has to be. In the case of experience of the external world, it is not literally the case
that we would reliably track truths whatever they turn out to be (hence the parenthetical
about reasonable variation). Given that, it is not clear to me that we do not get analo-
gous reliability across reasonable variation in the normative case, especially once we are
allowed prima facie justified normative beliefs to help settle how the normative facts vary
with variations in the non-normative facts.
114 Matthew Bedke

one might complain that the probationary justification of normative beliefs

has lapsed, whereas experience-based beliefs have passed their probationary
justification and earned secured justification.
This probationary epistemology is worth further exploration. But I think
there might be a quicker route to skepticism. Rather than try to show nor-
mative beliefs are unreliable, or show that we cannot explain how they
could be reliable after sufficient inquiry, perhaps the thing to focus on is
their insensitivity to the facts they are about. This is related to reliability
worries, but more narrowly focused on what one believes in nearby worlds
where one’s beliefs are false. I think the best way to develop this thought is
in terms of obliviousness, so let me turn to that now.

4.  Coincidence as Obliviousness

I want to show that premises (1)–(4) of the Coincidence Argument ensure

that normative beliefs, justifications, and explanations are robustly insensi-
tive to fact. Sensitivity is usually discussed as a condition on knowledge,
where one’s justified true belief is sensitive if (defn.) in the nearest possible
world(s) where the belief is not true, one would no longer believe it. We are
talking about procedural justification, so conditions on knowledge are not
directly relevant. Still, when making up one’s mind about whether to believe
P it does seem relevant whether, were P false, one would have believed that
P. Realizing that this is the case should cause some concern. Realizing in
addition that there is some erstwhile justifying basis J for the belief that
P that would be the same were P false should cause even more concern.
And finally, realizing that, were P false, the same explanation would hold
for why one has the justification J and belief that P should be even more
disconcerting.
Let us say, then, that the belief that P based on justification J is oblivious
to the target fact when it meets these conditions—when, were P not the
case, (i) one would believe that P, (ii) one would have the same justification
J for believing that P, and (iii) the same causal explanations for why one
believes that P and why one has justification J would hold.12 I claim that
realization that one’s belief that P is oblivious to the target fact is a defeater
for justification J. Once defeated, J cannot help to locate us in the space of
possibilities. And that is what makes adequate alignment between belief and
fact too coincidental to accept.

12
  It is not clear to me whether we need (iii) to get defeat. In classic cases of defeat,
like the red wall case discussed below, (i) and (ii) seem to suffice. But the addition of (iii)
makes the case for defeat for normative non-naturalism that much better.
 No Coincidence? 115

Premises (1)–(4) of the Coincidence Argument ensure that normative

beliefs about basic non-natural facts are oblivious. Let prima facie justi-
fication do its work, so we take ourselves to be in some world where the
causal forces conspired to adequately align normative belief with fact. So
anchored, we wonder what we would believe, and what justification we
would have, and why we would have it, were the basic normative facts other
than we believe them to be. We realize that everything we believe about the
natural world can be held fixed in these scenarios we are imagining. We
need only imagine scenarios where the basic normative facts differ or do not
exist, so that pain isn’t bad, or pleasure is not good, or some such (always
construed as a difference in non-natural fact of the matter). Were that the
case, we would have the same normative beliefs, the same justification for
them (e.g. it would still seem as though pain is bad, pleasure is good, and
so on), and the very same causal explanations for why we have those beliefs
and those justifications would apply. Our beliefs are oblivious to the facts.
Justification defeated.
Generic skeptical cases do not exhibit this defeater. Let us grant that one
has justification for believing that one has hands, and so one centers oneself
in a range of possibilities where belief and fact align. When one wonders
what one would have believed if one didn’t have hands, the nearby scenarios
to consider are those where one lost one’s hands in a tragic accident. In
those scenarios one would no longer believe that one has hands, and surely
one’s evidence for being handed would differ. So the belief that one has
hands is not oblivious to the fact. This mirrors application of sensitivity
requirements on knowledge.
Obliviousness helps to see how Street’s Jupiter case and Joyce’s belief pills
cases introduce conditions analogous to (1)–(4) of the Argument (as well as
additional epistemic difficulties already discussed). In Street’s case, for any
given Jupiter belief, if that belief were false one would still believe it, one
would have the same justification for so believing, and the same explanation
for why one has the belief and justification would hold—namely, the hyp-
notist pulled that slip out of a hat. In Joyce’s cases, for any given belief about
Napoleon, if the belief were false one would still believe it, one would have
the same justification for so believing, and the same explanation for why
one has the belief and justification would hold—namely, the pill (plus envi-
ronmental factors) induced the belief. That these processes are additionally
known to be unreliable or random is an extra difficulty. But we could have
a string of more benign historical information—about the publisher of the
book, the ink, etc.—that cumulatively meet the obliviousness conditions
and thereby constitute defeat.
My hope is that when this epistemic principle of obliviousness is spelled
out and seen to apply whenever the conditions of premises (1)–(4) above
116 Matthew Bedke

are met, it will bring the case for defeat into sharper relief. Admittedly, it is
difficult to explain why fundamental normative statuses are as they are, so
it is difficult to explain further why recognized obliviousness is a defeater.13
Nevertheless, it does seem on its face to capture an epistemic concern, and
it nicely categorizes and explains not only the above cases, but also classical
cases of defeat. Consider again that case of a red light illuminating a wall.
Initially you believe a red-looking wall is red. When you realize the wall is
illuminated with red light, what seems to make that a defeater is oblivious-
ness, or something very similar to it.
The main source of resistance in the normative case, I suspect, concerns
the necessity of the fundamental normative truths. How can we consider
what would be the case were pain not bad when we justifiably believe (a)
that pain is bad, (b) that if pain is bad, then necessarily it is bad, and so (c)
that necessarily pain is bad? However, we should not be glib about appeals
to necessitation. Whether they block application of obliviousness depends
on what kind of necessitation we are talking about and how we interpret
the subjunctives.

4.1  Obliviousness to Necessary Truths

There is nothing inherently strange about being oblivious to whatever turns
out to be a necessary truth, at least for certain necessary truths. Consider
Sally, who is justifiably convinced that the world is governed by determin-
istic laws of nature. While in a bar one night she reflects on this belief, and
forms the conditional belief that if Dropout sinks the eight ball, necessarily
Dropout sinks the eight ball. Dropout shoots and . . . sinks the eight ball.
Now Sally justifiably believes that, necessarily, Dropout sunk the eight ball
in that there is no nomologically possible world (i.e. one with the same past
and deterministic laws) where Dropout did not sink the eight ball. Can she
still intelligibly wonder what she would have believed were it not the case
that Dropout sunk the eight ball? Yes. Nomological necessity does not get
in the way. And, fortunately, Sally justifiably believes that, if Dropout hadn’t

13
  How obliviousness applies to beliefs about the future and beliefs based on enumera-
tive induction is a tricky matter. For some such beliefs, obliviousness will be a concern.
(Not for all. Consider: What would I believe, and what would be my justification for
believing it (and what explains both), were the sun not to rise tomorrow? Well, the near-
est possibility where that happens is one where the laws of nature differ, or where there
has been good evidence that sun will not rise. If so, my belief that it will rise is not oblivi-
ous.) But these will be special cases in epistemology generally, where we think that there
are grounds for justified belief—e.g. some uniformity of nature thesis—despite forms
of insensitivity. Non-naturalist realists have offered no reason for grouping normative
facts with facts about the future and the unenumerated so as to enjoy those justificatory
grounds.
 No Coincidence? 117

sunk the eight ball, she wouldn’t have believed that he did, and would have
lacked justification for believing it. Maybe he would have set her up for an
easy win. In any event, her belief is not oblivious.
Or consider Claire, who sells glacier water. She justifiably believes that
water is H2O. Moreover, she has read enough philosophy to justifiably
believe that if water is H2O, then necessarily it is H2O. So she thinks that
necessarily water is H2O. As we like to say, there is no metaphysically pos-
sible world where water is not H2O. Can Claire still intelligibly wonder
what would be the case if water were not H2O? This is a trickier case. It is
easy to hear the question along the following lines: What would be the case
if this stuff (splash it around for emphasis) were not this stuff (again, splash
it around for emphasis)? That seems unintelligible. In other words, if we
think of this necessity as secured by a special kind of reference enjoyed by
these terms, where they pick out the same referent directly and rigidly, then
perhaps the question is not really intelligible.
Suppose so. Still, non-naturalists should not take comfort in the exam-
ple. For they eschew theories of co-reference that would make our norma-
tive subjunctives as problematic as the one above about water. They think
that “bad” refers to a different property than any natural one. If they are
right, then when we ask what would be the case were pain not bad we are
not asking about what would be the case were something not itself. We are
asking about what would be the case if pain had a different normative prop-
erty than the one we take it to have, or no normative property whatsoever.
Unlike the case of water, there is no threat that the meaningfulness of the
subjunctive is ruled out by a special kind of referential relationship.
To aid discussion, it helps to distinguish two ways one might justifiably
believe the necessity of substantive normative truths. One way is derivative
and parasitic on justified beliefs in actual substantive normative truths. In
that case, one’s justified belief about which substantive normative truths
are the necessary ones is derived from (a) the general belief that norma-
tive truths are necessary, and (b) beliefs in actual, substantive normative
propositions, where the justification for (a) does not depend on having jus-
tified beliefs of type (b). For example, one might justifiably believe that
some supervenience principle holds a priori because it is analytic:14 neces-
sarily, if some object O has normative status S, necessarily any object that
is identical to O in all non-normative respects has normative status S. That
would be a general, non-substantive normative belief. If one also justifiably
believes that some episode of pain is bad, one can then infer that, necessar-
ily, anything identical to that episode of pain in all non-normative respects

14
  Though how we are justified in believing the necessities is not relevant, so long as
their justification is separate from the justification of the normative statuses of things.
118 Matthew Bedke

is bad. That is the sense in which this episode of pain is necessarily bad.15
Alternatively, one might justifiably believe that basic normative principles
are necessarily true, justifiably believe that an actual basic moral principle is
an action is right iff and because it maximizes happiness, and infer that, neces-
sarily, an action is right iff and because it maximizes happiness.
However general beliefs about the necessity of truths in a domain are
justified, the necessity of the truths does not insulate the particular beliefs
in that domain from sensitivity-type tests. To see this, set to one side the
justified general belief that normative truths are necessary and focus on
non-modal beliefs about what the normative facts are. Let us rehearse the
example about pain being bad. If pain were not bad—consider this coun-
terfactually if you like—would we still believe it to be bad, and would we
have the same justification for so believing, and would the same explanation
for why we have that belief and that justification hold? Unfortunately, the
answers are all “yes”. So that belief is oblivious, as are the stronger modal
beliefs derived from it and some general necessitarian thesis.
Let us craft an eight-ball case that features similarly oblivious beliefs,
just to convince ourselves. Imagine that Sally learns that she took a pill that
would cause her to hallucinate sinking eight balls. She now thinks the belief
that Dropout sunk the eight ball oblivious. She cannot block obliviousness
by noting that events that did take place are nomologically necessitated,
as though there is a problem even considering the obliviousness questions
because they take us to counter-possible worlds. Obliviousness of the belief
about the actual event ensures that she cannot justifiably locate herself in
the right space of nomic possibilities, so she does not know if she is asking
after counter-possibilities or genuine possibilities. To know that she must
first settle what the actual world is like, and she lacks justification for beliefs
about it insofar as they are recognizably oblivious.
Now, replace sinking the eight ball with the badness of pain, nomic
necessity with normative necessity, and the pill with premises (1)–(4) of
the Coincidence Argument, and you have an analogous problem on your
hands. The realist cannot block obliviousness by noting that normative sta-
tuses are necessitated, as though there is a problem even considering the
sensitivity-type questions because they take us to normatively impossible

15
  Some views about a posteriori knowable identities would be structurally similar. We
might know a priori that water is the stuff that meets some set of criteria associated with
the concept of water, and then discover a posteriori the nature of the stuff that actually
meets those criteria. We would then be able to infer the substantive necessity that water is
that stuff. Those who endorse this view of things should have no problem testing for the
obliviousness of the beliefs about the nature of the stuff that meets the relevant criteria.
Obliviousness of those beliefs would ensure that one cannot justifiably locate oneself in
the right space of metaphysical possibilities.
 No Coincidence? 119

worlds. Obliviousness of the belief about badness ensures that she cannot
justifiably locate herself in the right space of normative possibilities. What
is possible and what is impossible depends on what the actual world is like,
and we lack justification for beliefs about it insofar as they are recognizably
oblivious.
So far I have only discussed one way of justifiably believing in the neces-
sity of substantive normative truths. The second way of being justified is
more direct—it is not via inference with a premise concerning actual sub-
stantive normative propositions. Perhaps, for example, one is more directly
justified in thinking that necessarily pain (or an episode of pain like that)
is bad. Maybe I know directly the modal status of this synthetic claim, or
synthetic moral principles. Would that make it harder to intelligibly apply
the subjunctives that test for obliviousness?
It might seem so. For it looks like our counterfactuals have built-in
counter-possibility. But this is a pretty cheap way for a justified belief to
gain immunity from the threat of obliviousness. The necessity is still part of
the content of the belief, so it should be possible to show a belief with that
content—or any content—is oblivious or not.
To handle the situation, we can think of the obliviousness subjunctives
as asking after allodoxic possibilities, not counterfactuals. Allodoxic pos­
sibilities are false belief possibilities—they are those we can assume to obtain
contrary to what we actually believe and our justifying bases for believing
it.16 Assuming our actual beliefs are false and justifying bases misleading, we
can then emphasize the third component of obliviousness: How much of
our explanatory picture of the world would have to change to explain how
our beliefs and justifications get things so wrong?
To see how this works, consider the belief that water is H2O and all our
justification for believing it. To consider the allodoxic possibility that it is
not H2O we assume for the sake of further inquiry that it is not H2O (and
never has been), and then examine what adjustments to our explanatory
picture of the world would have to change to explain why we nevertheless
have all this justification/evidence/reason in favor of the false belief that it is
H2O. There is a lot of explaining to do, of course. Have the chemists been
lying to us? What of the chemistry I think I know that explains why water
has some of the interesting properties it has, like the fact that it expands
when it freezes? We can see that the belief that water is H2O does not meet
the obliviousness criteria, interpreted in terms of an allodoxic possibility.

16
  I prefer this way of thinking about the questions over near cousins (e.g. counter-
factuals and various ways of separating two dimensions of intension). Those alternatives
either get at slightly different questions, or get at the same questions in more confusing
ways. Also, it seems natural to ask after allodoxic possibilities with the subjective mood.
But if it bothers you, try indicatives.
120 Matthew Bedke

Neither would the modal belief that, necessarily, water is H2O. Assuming
that is false for the sake of argument, we would have a lot of explaining to
do for why we have the false belief and the misleading evidence we have.17
Now, assume it is not the case that, necessarily, pain is bad, as part of an
allodoxic possibility. Let us hold fixed the non-normative ways of the world
(as we justifiably believe them to be). Focusing on the third part of the
obliviousness criteria, what of our explanatory picture of the world needs to
change to explain how we nevertheless have all this justification/evidence/
reason in favor of the false belief that, necessarily, pain is bad? For example,
what of our causal-explanatory picture must change to explain why we have
the intuition that it is bad? Well, nothing. In the assumed scenario, there
is no additional or different explanation for why we (ex ante) justifiably
believe pain is necessarily bad. If some evolutionary explanation explains
why we think pain is necessarily bad, and explains why we have the justify-
ing bases for so believing, that same explanation holds under the allodoxic
possibility that pain is not necessarily bad. This belief is oblivious.
We avoid this result if we deny that these allodoxic possibilities are con-
ceptual possibilities, or hold that our justifications constitutively depend
on the normative facts of the matter (no facts, no justifications). I do not
see how the former can be squared with non-naturalist realism, whereby
basic, substantive normative beliefs are about stance-independent facts that
cannot be known merely by reflection on concepts. And the latter cannot
be squared with procedural justification and causal closure of the natural
world, for when making up our minds about normative matters we do not
have the non-natural facts to go on, and if we did the natural world would
not be causally closed. I see no other way of avoiding the defeat.

4.2  Better Safe than Sensitive?

Nevertheless, there are some things that worry me about obliviousness. First,
it is a cousin of sensitivity requirements on knowledge, which have been
called into question. I take some comfort in the fact that non-obliviousness
is a condition on procedural justification with a few more bells and whistles
than sensitivity. But these days some prefer a safety condition on knowledge
in lieu of a sensitivity condition, in part because of apparent difficulties
when applying sensitivity to modal truths. A belief that P is safe if (defn.) in
many(/all) nearby possibilities where one believes P, P.

17
  Similarly, if you wonder about being envatted as the actual state of things, you
have a lot of explaining to do. A good deal of your beliefs and justifications regarding the
external world would be explained not via interaction with the external world, but by
interaction with some systematically deceptive device.
 No Coincidence? 121

I have suggested that the modal concerns might be overstated.

Nevertheless, it would be nice for the skeptic were normative beliefs in
non-natural facts unsafe as well as oblivious. The problem here is figuring
out which possibilities are nearby. Are possibilities where the (necessary)
normative facts are not as we believe them to be nearby? Maybe. One meas-
ure of nearness is how much of our ordinary explanatory picture must be
shifted to adequately explain what is going on in a target possible world.
BIV worlds are obviously far off on this metric. But in worlds where the
(necessary) normative facts are other than we believe them to be nothing in
our causal-explanatory picture of the world need be shifted. So maybe our
normative beliefs are unsafe insofar as they are about non-natural facts. If
they are unsafe and insensitive in the ways described above, call our nor-
mative beliefs robustly oblivious to the facts. Realization that our beliefs are
robustly oblivious is an even better defeater than mere obliviousness.
Unfortunately, I am not at all confident about what the metric of near-
ness is for safety checks. I worry that safety involves a nearness metric that
is far too deferential to the beliefs we are testing to provide epistemic checks
with real bite. So I  will not pursue robust obliviousness any farther. The
skeptic’s clearest case rests on mere obliviousness.

5.  The Moorean Reply

Here is another thing that worries me about obliviousness. While I  find

obliviousness a compelling defeater I am not inclined to abandon my nor-
mative beliefs. Pain is bad, torturing people at random is wrong, P’s likely
truth is a reason to believe that P, and so on. But it is important to bear in
mind that the Coincidence Argument does not imply that we are unjusti-
fied in believing these things, but rather that we are unjustified in believing
these things insofar as they are about non-natural normative facts. For the
argument relies on a premise about non-natural purport. This bears directly
on a Moorean reply to the skeptical conclusion. Let me turn to it.
It often helps to notice that one is not forced to accept the conclusion of
a valid argument populated by premises one already accepts. There is always
the option of rejecting one of the premises, and this is the rational thing to
do when one is more justified in believing the negation of the conclusion
than one is justified in believing the conjunction of the premises. Parfit
makes this move in relation to evolutionary debunking arguments. His pre-
ferred formulation is the following:
(1) [Our normative] beliefs were often advantageous, by causing us to
have true worldly beliefs which helped us to survive and reproduce.
122 Matthew Bedke

(2) Because these normative beliefs were advantageous, natural selec-

tion made us disposed to have them.
(3) These beliefs would have had the same effects whether or not they
were true.
Therefore
(4) These beliefs would have been advantageous whether or not they
were true.
Therefore
(5) Natural selection would have disposed us to have these beliefs
whether or not they were true.
(6) We have no empirical evidence for the truth of these beliefs.
(7) We have no other way of knowing whether these beliefs are true.
Therefore
We cannot justifiably believe that these beliefs are true. (2011:
512, 525–6)
In both the epistemic and practical cases, one of his responses is that we are
more certain or justified in normative claims—e.g. when a belief is likely
true, we have reason to believe it (p. 521), and torturing children for fun
is wrong (p. 544)—than we are of the premises of the skeptical argument.
Parfit identifies (7) as a weak link (in addition to premise (2)). But it is
important to note that (3), and so also (4) and (5), relies on a suppressed
premise: that normative beliefs are about non-natural facts.18 Without this,
there is little reason to think that normative beliefs would be selected for
or caused regardless of their truth. If the fact that an action is right just is
the fact that an action maximizes happiness, and the belief that an action is
right just refers to this fact, it might not be adaptive to believe that actions
are right regardless of whether they maximize happiness. At least, I would
need to hear more to be convinced.
The import of a non-naturalist suppressed premise is this. One can get a
lot of mileage out of an apparent threat to beliefs no one is willing to aban-
don, e.g. that torturing people at random is wrong. But if the skeptical argu-
ment relies on the premise that these beliefs are about non-natural entities a
clear way out of the problem is to abandon the offending meta-ethical posi-
tion. We can leave Parfit’s Moorean data intact and reject non-naturalism.19
This also helps us to see that the argument is not self-defeating. It does
presume a certain kind of normative significance:  certain considerations
available to the mind count as epistemic defeaters. But the argument only

18
  Things are complicated by Parfit’s “non-metaphysicalism.” I  think we can safely
put this in the non-naturalist camp so long as it is not a veiled form of non-cognitivism.
19
  See also Bedke (2009: 205).
 No Coincidence? 123

undermines that claim when conjoined with a meta-normative claim about

Non-Natural Purport.
Even better, our reaction to the Coincidence Argument might itself pro-
vide some evidence that our normative beliefs are not about non-natural
facts. Normally, we expect beliefs about matters of fact to go out of exist-
ence when we receive defeaters. Perhaps failures to obey this expectation can
be explained away on a case-by-case basis. But it would be much harder to
explain why all beliefs about a certain domain of fact (the normative) are
systematically recalcitrant in the face of what we take to be good defeaters
for those beliefs. This, I submit, is the situation the realist faces when she
realizes that her normative beliefs are systematically recalcitrant in the face
of the coincidence-as-obliviousness defeater. The fact that she is not the
least inclined to abandon her normative commitments in the face of the
defeater is some evidence that these normative commitments are not beliefs
about non-natural properties after all.20
So it is looking as though resistance to the Coincidence Argument is fueled
primarily by steadfast commitment to one’s normative beliefs, and this can
lead some to think there must be something wrong with the Coincidence
Argument. But that position feels compelling only because we have not
clearly separated out normative commitments and meta-normative options.
The least justified premise is the one regarding normative non-naturalism.
And it becomes even less justified when one realizes that one’s norma-
tive beliefs are recalcitrant in the face of a defeater for justified belief in
non-natural fact.

6. Conclusion

I have tried to sort through some of the key moves in the coincidence litera-
ture. To my mind, the skeptical case is best expressed in terms of oblivious-
ness, and we have seen that the modal status of the target domain offers no
absolution.
One thing I have not addressed is how the points might generalize to,
say, mathematical Platonism. I think that extensions of the argument are
problematic. We can intelligibly wonder what would be the case were pain
not bad. A meta-normative theory then steps in to inform the subjunctive.
Under non-naturalism we may interpret the subjunctive one way, where
we consider counterfactuals or allodoxic possibilities about properties and
facts entirely distinct from any natural ones (though supervenient upon
them). Naturalism would have us interpret the subjunctive differently, as

  See also Bedke (forthcoming).

20
124 Matthew Bedke

concerning natural properties and facts, in which case it might be as prob-

lematic as subjunctives about water not being H2O. And such subjunctives
are just difficult on a non-cognitivist meta-normative view.
When we turn to mathematics we do not get similarly intelligible first-order
questions to be glossed in different ways by different meta-mathematical
positions. I’m not sure what we are asking when we ask what would be the
case if 2+2 did not equal 4 (without an alternative conceptual schema that
makes sense of this, and thereby changes the subject). So I’m not sure that
we get so far as to ask what a Platonic gloss would be, or a structuralist gloss,
or a naturalist gloss, or what have you. Sure, it is intelligible to ask what
would be the case were some Platonic objects different, or if they enjoyed
different relations with the natural world than we think they do.21 But given
the status of the first-order mathematical claim, and the unintelligibility of
its negation on first-order (non-meta-mathematical) grounds, it is very hard
to see how it could be about Platonic objects. At least, this is a problem the
mathematics case runs up against that the normative case does not.22
Be that as it may, when we follow the argument in the normative case we
see a defeater on the horizon. The natural reaction is to dig in our heels on
the normative commitments. This reaction makes most sense if the norma-
tive commitments are not beliefs about non-natural facts. So, at the end of
the day, we get a defeater for our normative beliefs insofar as they are about
non-natural facts, and some evidence that they are not about non-natural
facts after all.

References
Bedke, M. 2009. “Intuitive Non-Naturalism Meets Cosmic Coincidence,” Pacific
Philosophical Quarterly, 90(2): 188–209.
Bedke, M. Forthcoming. “A Menagerie of Duties? Normative Judgments are Not
Beliefs about Non-Natural Properties,” American Philosophical Quarterly.
Clarke-Doane, J. 2012. “Morality and Mathematics: The Evolutionary Challenge,”
Ethics, 122: 313–40.
Clarke-Doane, J. Forthcoming. “Moral Epistemology:  The Mathematics
Analogy,” Noûs.
Copp, D. 2008. “Darwinian Skepticism about Moral Realism,” Philosophical Issues,
18(1): 186–206.

21
  Cf. Clarke-Doane (2012).
22
  Even if there is parity with the mathematical case, I think a similar resilience of
mathematical belief in the face of the defeater can be brought to bear as some evidence
that those beliefs are not about Platonic objects after all.
 No Coincidence? 125

Enoch, D. 2011. Taking Morality Seriously:  A  Defense of Robust Realism.

Oxford: Oxford University Press.
Field, H. 2005. “Recent Debates about the a Priori,” in T. Gendler and J. Hawthorne
(eds), Oxford Studies in Epistemology, i. 69–88. Oxford: Clarendon Press.
Fitzpatrick, W. Forthcoming-a. “Debunking Evolutionary Debunking of Ethical
Realism,” Philosophical Studies.
Fitzpatrick, W. Forthcoming-b. “Why there is No Darwinian Dilemma for Ethical
Realism,” in M. Bergmann and P. Kain (eds), Challenges to Religious and Moral
Belief from Evolution and Disagreement. Oxford: Oxford University Press.
Gibbard, A. 2003. Thinking How to Live. Cambridge, MA: Harvard University Press.
Greene, J. D., Morelli, S. A., Lowenberg, K., Nystrom, L. E., and Cohen, J. D.
2008. “Cognitive Load Selectively Interferes with Utilitarian Moral Judgment,”
Cognition, 107(3): 1144–54.
Huemer, M. 2005. Ethical Intuitionism. New York: Palgrave Macmillan.
Joyce, R. 2001. The Myth of Morality. New York: Cambridge University Press.
Joyce, R. 2006. The Evolution of Morality. Cambridge, MA: MIT Press.
Joyce, R. Forthcoming. “Evolution, Truth-Tracking, and Moral Skepticism,” in B.
Reichardt (ed.), Problems of Goodness: New Essays on Metaethics. Bonn: Bernstein
Verlag.
Kahane, G. 2010. “Evolutionary Debunking Arguments,” Noûs, 45(1): 103–25.
Parfit, D. 2011. On What Matters. Oxford: Oxford University Press.
Ruse, M. 1986. Taking Darwin Seriously. Oxford: Basil Blackwell.
Schafer, K. 2010. “Evolution and Normative Scepticism,” Australasian Journal of
Philosophy, 88(3): 471–88.
Shafer-Landau, R. 2012. “Evolutionary Debunking, Moral Realism and Moral
Knowledge,” Journal of Ethics and Social Philosophy, 71:  <http://www.jesp.org/
PDF/Evolutionary_Debunking_final.pdf>.
Singer, P. 2005. “Ethics and Intuitions,” Journal of Ethics, 93(4): 331–52.
Sinnott-Armstrong, W. 2007. Moral Skepticisms. New York: Oxford University Press.
Skarsaune, K. O. 2011. “Darwin and Moral Realism:  Survival of the Iffiest,”
Philosophical Studies, 152(2): 229–43.
Street, S. 2006. “A Darwinian Dilemma for Realist Theories of Value,” Philosophical
Studies, 127(1): 109–66.
Street, S. 2008. “Reply to Copp:  Naturalism, Normativity, and the Varieties of
Realism Worth Worrying About,” Philosophical Issues, 18: 207–28.
Wielenberg, E. J. 2010. “On the Evolutionary Debunking of Morality,” Ethics,
120(3): 441–64.
 6
Moral Skepticism and Moral
Disagreement in Nietzsche
Brian Leiter

Almost everyone agrees that Nietzsche is a skeptic about the objectivity of

morality,1 but beyond that point, disagreement abounds as to the grounds
for this skepticism, its scope, and its implications for the semantics of moral
judgment. In this essay, I will set out a systematic view on the first two ques-
tions (concerning the grounds and scope of his skepticism), building on
some prior work (Leiter 2000, 2002: 136–55).2 I will assume throughout
that Nietzsche’s skepticism about the objectivity of morality is not sim-
ply a special instance of the skepticism that is sometimes associated with
his doctrine of perspectivism—that is, I will assume that it is not simply
an instance of generalized skepticism about our knowledge of the world
or a global skepticism about truth. There is probably a modest consensus
now among Anglophone interpreters of Nietzsche—including Clark (1990,
1998), Janaway (2007), Poellner (2001), Richardson (1996), and myself
(1994, 2002:  268–79)—that whatever exactly “perspectivism” means, it
does not and can not entail a general skepticism about the objectivity of
knowledge or truth. I shall not argue for that position here, however, or for
my growing suspicion that, in the end, Nietzsche does not have a coherent
or well-motivated set of general epistemological views.3 What I shall argue

1
  In Leiter (2000) I critique earlier efforts to show that Nietzsche’s putative doctrine
of the will to power grounds a kind of Nietzschean value realism.
2
  I will bracket here semantics, though I continue to believe (cf. Leiter 2000) that it
is anachronistic to saddle Nietzsche with a semantic view, as e.g. Hussain’s interesting
fictionalist reading does (Hussain 2007). Hussain (2013) appears to come around to my
view on this issue.
3
  I take that to be the real lesson to emerge from those Clark calls “the Stanford
school” (meaning Anderson 1998 and Hussain 2004), who call attention to the influence
of strands in 19th-cent. neo-Kantianism and positivism on Nietzsche, though without
drawing the conclusion that seems most warranted, namely, that Nietzsche’s amateur
reflections on questions of general metaphysics and epistemology probably betray more
confusion than insight in the end.
 Moral Skepticism in Nietzsche 127

here is that we can adduce independent grounds for Nietzsche’s skepticism

about the objectivity of morality and that these grounds are of independent
philosophical interest.

1.  The Scope and Grounds of Nietzsche’s

Value Skepticism

Is Nietzsche a skeptic about the objectivity of all value judgments? And to

the extent he is skeptical about the objectivity of value, what is it exactly that
he is denying?
In earlier work (Leiter 2002: 45–7, 106–12), I argued that the first ques-
tion should be answered in the negative: that is, I claimed that Nietzsche
could not be skeptical about the objectivity of all value judgments because
he had to admit the objectivity of judgments of prudential value. The rea-
sons given then now seem to me mistaken, and in ways that bear on the
argument here.
Nietzsche’s central objection to morality—or to what I call “morality in
the pejorative sense” (hereafter MPS), to pick out that cluster of values that
is the actual target of his critique—is that its cultural prevalence is inhos-
pitable to the flourishing of the highest types of human beings, namely,
creative geniuses like Goethe, Beethoven, and Nietzsche himself. Nietzsche
argues for this conclusion on the basis of a speculative moral psychology
that shows how agents who took seriously the norms of MPS would, in fact,
be unable to realize the kinds of excellence we associate with geniuses like
Goethe and Beethoven. If this is Nietzsche’s argument, then it might seem
that at the core of his critique of MPS is a judgment about prudential value
(i.e. about what is good or bad for an agent), namely, the judgment that
MPS is bad for certain persons because it is an obstacle to their flourishing.
And if that judgment were not objectively true, then Nietzsche’s critique of
MPS might seem to have no force.
Commitment to the objectivity of prudential value is not, of course, an
ambitious position. Railton dubs it “relationalism” (1986a) and suggests
that we “think of [non-moral or prudential] goodness as akin to nutritive-
ness.” Just as not all nutrients are good for all kinds of creatures, so too
not everything is prudentially good for everyone: to use Railton’s standard
example, cow’s milk is prudentially good for calves, but not for human
babies. So, too, what is good for the herd may be bad for the higher men,
and vice versa. Many of Nietzsche’s favorite Greek philosophers, the
Sophists, already recognized the objectivity of judgments of relational value
(see Leiter 2002: 45–6), and that might also lend support to the interpretive
128 Brian Leiter

hypothesis that Nietzsche accepts the same view. Indeed, as Railton notes,
“realism with respect to non-moral [or what I am calling prudential] good-
ness . . . [is] a notion that perfect moral skeptics can admit” (1986b: 185).
And Nietzsche is, indeed, a “perfect” moral skeptic, or so I  shall argue,
since he clearly holds that moral value (valuations of what is good or bad
simpliciter or non-relationally) is not objective. So, for example, while the
judgment that MPS is bad for higher human beings might be objectively
true, the judgment that MPS is disvaluable simpliciter or should be defeated
because it is bad for higher human beings is not.
It now seems to me, however, that Nietzsche’s position does not even
require the objectivity of judgments of prudential value. It does, to be sure,
have to be objectively true that MPS values prevent nascent Goethes from
becoming Goethes, but that causal claim need entail no evaluative assess-
ment about whether that is a good or bad outcome. Nietzsche presumably
expects the readers “suited” to his insights to view this outcome as bad for
Goethe, but all he needs for the force of his critique is the truth of the
causal claim that MPS values have certain kinds of effects. That judgments
of prudential value need not be objective is fortunate given the argument
from disagreement for value skepticism discussed below.
What, then, is involved for Nietzsche in denying the “objectivity” of what
is morally right and wrong, morally good and bad? I have been purposely
vague so far about whether the issues are semantic, metaphysical, and/or
epistemological; indeed, as I have argued elsewhere (Leiter 2000), Nietzsche
has no discernible semantic view at all. Here we will concentrate on the
metaphysical and epistemological issues. On the reading I  will defend,
Nietzsche is a moral skeptic in the precise sense of affirming the metaphysi-
cal thesis that there do not exist any objective moral properties or facts (I
will refer to this hereafter as simply “skepticism about moral facts”).4 From
this it will, of course, follow that there is also no moral knowledge, but it is
the argument for the metaphysical thesis that is crucial for Nietzsche.
Now it seems obvious that some of Nietzsche’s skepticism about moral
facts is simply skepticism about a kind of Platonism about value. Plato, to

4
  I would like what follows to be compatible with a number of different theses about
what the metaphysical objectivity of moral facts would consist in, and, in any case, do not
want to derail the discussion in the text into a characterization of objectivity. (For some
discussion, see Leiter 2007: 258–61.) Briefly, we may say that moral facts are metaphysi-
cally objective if their existence and character does not depend on what persons believe,
have reason to believe, or (perhaps) would have reason to believe under ideal conditions
about them. (Alex Silk suggests to me that perhaps Nietzsche thinks moral facts are
attitude-dependent facts, and I suspect much of the argument that follows is compatible
with that possibility, with the caveat that, as an empirical matter, Nietzsche thinks the
relevant attitudes vary significantly among persons. But I am skeptical that, in the end,
this will turn out to be a correct way of rendering Nietzsche’s view.)
 Moral Skepticism in Nietzsche 129

be sure, does not think there is a special problem about the objectivity of
value, since he thinks values are objective in the same way all Forms are.5
A Form, says Plato, “is eternal, and neither comes into being nor perishes,
neither waxes nor wanes” (Symposium 211a). In the Phaedo, he calls them
“constant and invariable” (78d) while in The Republic he refers to them
as “the very things themselves . . . ever remaining the same and unchanged”
(479e). Forms are, in the words of the Symposium, “pure, clear, unmixed—
not infected with human flesh and color, and a lot of other mortal non-
sense” (211a).
Many of Nietzsche’s skeptical-sounding passages appear to involve denials
of this kind of Platonism about value. So, for example, Zarathustra declares:
Verily, men gave themselves all their good and evil. Verily, they did not take it, they
did not find it, nor did it come to them as a voice from heaven. Only man placed
values (Werte) in things to preserve himself—he alone created a meaning for things,
a human meaning. Thus he calls himself “man,” which means:  the esteemer (der
Schätzende).
To esteem is to create (Schätzen ist Schaffen): hear this, you creators! . . . Through
esteeming alone is there value (Wert): and without esteeming the nut of existence
would be hollow. . . . (Z i.15)
Similarly, writing in his own voice in The Gay Science, Nietzsche observes
that, “Whatever has value in our world now does not have value in itself,
according to its nature—nature is always value-less, but has been given value
at some time, as a present—and it was we who gave and bestowed it” (GS
301). Of course, many realists about value might be happy to acknowledge
that “without esteeming, the nut of existence would be hollow”; as Railton,
for example, puts it, “In a universe without subjectivity [i.e., without crea-
tures for whom things matter], there is no value either” (1986a: 18). Yet
Nietzsche goes further than this when he suggests that it is we who give
things their value, though even on this score there are arguably some “real-
ist” views, such as the sensibility theories of McDowell and Wiggins, com-
patible with this projectivist rhetoric. In any case, if Nietzsche’s only target
were the metaphysics of Platonism about value, Nietzsche’s skepticism
might not worry a lot of contemporary philosophers—though it is perhaps
worth emphasizing that a kind of Platonism about value appears to remain
central to most cultural and religious traditions, so his skepticism on this
score is hardly trivial.
In a range of other passages, Nietzsche emphasizes that moral judgment
involves a kind of projective error, and here it is especially important to note
that the emphasis is not on value simpliciter, but on moral value. So, for

5
  I here confine attention to the theory of Forms of the middle books.
130 Brian Leiter

example, in Daybreak, he notes that just as we now recognize that it was “an
enormous error” “when man gave all things a sex” but still believed “not that
he was playing, but that he had gained a profound insight,” so, too, man
“has ascribed to all that exists a connection with morality (Moral) and laid
an ethical significance (ethische Bedeutung) on the world’s back,” which will
“one day” be viewed as meaningful as talk about “the masculinity or femin-
ity of the sun” (3). So, too, in Human-All-Too-Human, Nietzsche compares
religious, moral, and aesthetic judgment with astrology:

It is probable that the objects of the religious, moral (moralisch) and aesthetic experi-
ences (Empfinden) belong only to the surface of things, while man likes to believe
that here at least he is in touch with the heart of the world (das Herz der Welt); the
reason he deludes himself is that these things produce in him such profound hap-
piness and unhappiness, and thus he exhibits here the same pride as in the case of
astrology. For astrology believes the heavenly stars revolve around the fate of man;
the moral man (moralische Mensch), however, supposes that what he has essentially
at heart must also constitute the essence (Wesen) and heart of things. (4)

Just as the astrologist thinks that there are astrological facts (about man’s
future) supervening on the astronomical facts about the stars—when, in fact,
there are only the stars themselves, obeying their laws of motion—so too the
“moral man” thinks his moral experiences are responsive to moral proper-
ties that are part of the essence of things, when, like the astrological facts,
they are simply causal products of something else, namely our feelings. As
Nietzsche puts it, moral judgments are “images” and “fantasies,” the mere
effects of psychological and physiological attributes of the people making
those judgments, attributes of which they are largely unaware (D 119).
As I  argued in Leiter (2002:  148–9), these kinds of remarks suggest a
“best explanation” argument for anti-realism about moral value:  the best
explanation for our moral experiences is not that they pick out objective
moral features of phenomena, but rather that they are caused by facts about
our psychological make-up:  for example, ressentiment or what Sinhababu
(2007) has recently dubbed “vengeful thinking” to describe the mechanism
by which “slavish” types come to believe strength, nobility, and wealth con-
stitute what is “evil.” If the best explanation of our moral judgments appeals
only to psychological facts about us, and need make no reference to objec-
tive moral facts, then we have reason to be skeptical about the existence of
moral facts.
Whether or not that argument is successful—interpretively or philosoph-
ically—is an issue I propose to bracket here. It now seems to me that there
is another set of considerations that underwrite Nietzsche’s moral skepti-
cism, and that these considerations are of independent philosophical inter-
est. Nietzsche does, on this account, rely on explanatory considerations, but
 Moral Skepticism in Nietzsche 131

not with respect to our moral experiences per se but rather with regard to
the phenomenon of moral disagreement. Moral disagreement has long been
a data point invoked by skeptics about morality, but Nietzsche’s approach
is a bit different.6 For what he calls attention to is not “ordinary” or “folk”
moral disagreement, but rather what seems to me the single most important
and embarrassing fact about the history of moral theorizing by philosophers
over the last two millennia:  namely, that no rational consensus has been
secured on any substantive, foundational proposition about morality.7 By
a “foundational” moral proposition about morality, I am thinking of, for
example, deontological or utilitarian theories which specify the criteria in
virtue of which concrete or particular moral judgments are thought war-
ranted: so e.g. “it is wrong to break this promise” is a concrete moral judg-
ment, while “the wrong-making feature of an action is its effect on utility” is
a foundational proposition. With regard to such foundational propositions,
the history of moral philosophy is the history of intractable disagreement. Is
the criterion of right action the reasons for which it is performed or the con-
sequences it brings about? If the former, is it a matter of the reasons being
universalizable, or that they arise from respect for duty, or something else?
If the latter, is it the utility it produces or the perfection it makes possible?
If the former, is utility a matter of preference satisfaction (as the economists
almost uniformly believe) or preference satisfaction under idealized circum-
stances—or is it, rather, unconnected to the preferences of agents, actual or
idealized, but instead a matter of realizing the human essence or enjoying
some “objective” goods? And perhaps a criterion of right action isn’t even
the issue, perhaps the issue is cultivating dispositions of character conducive
to living a good life. And here, of course, I have merely canvassed just some
of the disagreements that plague Western academic moral theory, not even

6
  Loeb (1998) comes closest, and I have benefited from and will reference his discus-
sion in what follows.
7
 Parfit (2011a) is the most notable recent attempt to show otherwise, though (with
the exception of Nietzsche) he really only canvasses the views of “friends of Derek,” and
tries to show that, in fact, they all agree. Discussing Parfit would constitute a separate
paper, but for pertinent doubts, see Schroeder (2011). Notable for our purposes is that
Parfit shares Nietzsche’s intuition that failiure to converge on moral truths would under-
mine the purported objectivity of moral thought (2011b: 571), and he correctly recog-
nizes the need to explain away his apparent disagreement with Nietzsche, “since he is the
most influential and admired moral philosopher of the last two centuries” (2011b: 571).
His discussion of Nietzsche is problematic in several respects, but that is a topic for a dif-
ferent occasion. More surprisingly, in discussing moral disagreement, Parfit dismisses dis-
agreement about foundational questions out of hand, saying only that “we would expect
there to be more disagreement about” this, and that it is enough that theories agree about
“which acts are wrong” (2011b: 554). It is hard, though, to see how disagreement about
why an act is wrong is not a very serious kind of moral disagreement, especially since such
disagreements typically explain disagreements about other particular cases. See n. 12.
132 Brian Leiter

touching on non-Western traditions, or radical dissenters from the mainstream

of academic moral theory, such as Nietzsche himself.
This persistent disagreement on foundational questions, of course, distin-
guishes moral theory from inquiry in the sciences and mathematics, not, per-
haps, in kind, but certainly in degree. In the hard sciences and mathematics,
intellectual discourse regularly transcends cultural and geographic bounda-
ries and consensus emerges about at least some central propositions.8 How to
explain the failure of moral theory to achieve anything like this? That is the
question, to which Nietzsche proposes a skeptical answer—or so I shall argue.
But first let us make explicit the structure of this skeptical argument before
returning to Nietzsche’s texts.

2.  Arguments for Moral Skepticism from

Disagreement

Standard “best explanation” arguments for moral skepticism focus on the fact of
moral judgment, and claim that the best explanation of such judgments is not
the objective moral features of the situation to which the moral agent putatively
responds, but rather psychological and sociological factors that cause the agent
to give expression to the particular moral judgment. In the version of this argu-
ment I have defended (Leiter 2001), the central problem with explanations of
our moral judgments that appeal to the existence of objective moral facts is that
they fail to satisfy demands of consilience and simplicity that we expect from
successful explanatory theories. Moral explanations fail along the dimension of
consilience because they posit facts—“moral” facts—that are too neatly tailored
to the explanadum (they are, as I shall say, explanatorily “narrow”), and that
don’t effect the kind of unification of disparate phenomena we look for in suc-
cessful explanations. They fail along the dimension of simplicity because they
complicate our ontology without any corresponding gain in explanatory power
or scope.9 The latter claim is, of course, crucial to the anti-realist argument. For

8
  Justin Clarke-Doane has pressed on me the possibility that disagreement in math-
ematics is also deep and perhaps intractable; for some discussion see Clarke-Doane
(forthcoming). I am not sufficiently expert in the mathematics to properly evaluate this
intriguing thesis, though it does seem in tension with all the sociological evidence about
mathematics, i.e. the cross-cultural and apparently progressive convergence on a host of
fundamental propositions of mathematics, including in set theory, one of Clarke-Doane’s
primary examples. See Jech (2002).
9
  Some moral realists claim that moral properties are just identical with or super-
venient upon the non-moral natural properties that figure in the alternative explanations
of moral judgments. But a claim of identity or supervenience cannot—in isolation—save
moral realism against the explanatory argument, for we must earn our right to such claims
by both (a) vindicating the identity/supervenience thesis on non-explanatory grounds;
 Moral Skepticism in Nietzsche 133

if it were true that without moral facts we would suffer some kind of explana-
tory loss, then moral explanations (and moral realism) would be in the same
metaphysical boat as the postulates of any of the special sciences: physics can’t,
after all, do the explanatory work of biology, which is why, by “best explanation”
criteria, we can admit biological facts into our ontology.10
Needless to say, no a priori considerations can demonstrate that there
will never be an explanatory loss from eliminating moral facts from our best
account of the world. Two sorts of considerations, however, may make us
skeptical of the realist’s claim. First, outside the contemporary philosophi-
cal debate, we do not find scholars in other disciplines actually concerned
with explanatory questions trying to do any explanatory work with moral
facts. Philosophers should perhaps remember that while, for example, there
are Marxist historians using broadly “economic” facts to explain historical
events, there is no school of “Moral Historians” using moral facts to do any
significant explanatory work.
A second ground for skepticism about moral explanations is more spe-
cific: namely, that the actual candidates proferred in the literature are, by
and large, not very promising. Some moral explanations are just patently
vacuous—think of Sturgeon’s well-known claim that, if asked to explain
Hitler’s behavior, we might appeal to his moral depravity, which sounds to
me more like a repetition of the question than an explanation—but even
more ambitious moral explanations (like those put forward e.g. by David
Brink, Joshua Cohen, and Geoffrey Sayre-McCord) do not withstand scru-
tiny, as I have argued in detail elsewhere (see Leiter 2001).11
Now the skeptical argument that concerns us will differ along three key
dimensions from the more familiar kinds of “best explanation” arguments
for moral skepticism just noted. First, what is at issue is not what we might
call “raw” moral judgments, as in Harman’s famous flaming pussycat case,
where someone witnesses young hoodlums dousing a cat and setting it on

and (b) vindicating the added theoretical complexity involved in these theses by dem-
onstrating that they produce a gain in consilience or some cognate epistemic virtue (e.g.
explanatory unification). I have argued (Leiter 2001) that they do not.
10
  More precisely, non-reductive moral realists want to defend moral explanations in
a way akin to Jerry Fodor’s famous defense of the autonomy of the special sciences: they
want to claim that there are distinctive “groupings” and generalizations in moral explana-
tions that cannot be captured by a more “basic” explanatory scheme or science. Just as
nothing in physics captures the distinctive categories and generalizations of economics
and psychology, so too biology and psychology are supposed to miss the distinctive gen-
eralizations of moral theory.
11
 Peter Railton’s work (e.g. 1986b) invokes a much richer form of historical explana-
tion, but involves both a controversial reforming definition (itself defeasible on simplicity
grounds) and a controversial set of Marxian theses about the mechanisms of historical
change. His version of the best explanation story would require separate attention.
134 Brian Leiter

fire and reacts by judging the act morally wrong or reprehensible. Instead,
our data points consist of philosophical theories about morality that purport
to license particular judgments by answering foundational questions. A philo-
sophical theory, for purposes here, is a systematic account of the foundations
of correct moral judgment and action based on reasons and evidence that pur-
ports to be acceptable to (some or all) rational agents (depending on the under-
lying view of rationality). Second, the explanatory question concerns not any
particular philosophical theory, but rather the fact that there exist incompatible
philosophical theories purporting to answer foundational questions. And they
are not simply incompatible philosophical theories: the disagreements of moral
philosophers are amazingly intractable. Nowhere do we find lifelong Kantians
suddenly (or even gradually) converting to Benthamite utilitarianism, or vice
versa. So the “best explanation” argument asks: what is the best explanation for
the fact that philosophical theories, in the sense just noted, reach different and
quite intractable conclusions about foundational matters? Nietzsche’s skepti-
cal answer will be that the best explanation is that the psychological needs of
philosophers lead them to find compelling dialectical justifications for very dif-
ferent basic moral claims, and there are no objective moral facts to stand in the
way of satisfying those psychological needs. (We will set out this position more
systematically shortly.) Third, consilience and simplicity are again theoreti-
cal desiderata to be weighed in comparing explanations, but their interaction
with moral realism is different: the claim at issue will be that skepticism about
morality is part of a more consilient and simpler explanation for the existence
of incompatible philosophical theories of morality than is the assumption that
there are objective facts about fundamental moral propositions, but that com-
peting philosophical theories of morality fail to converge upon them.
In short, what makes Nietzsche’s argument from moral disagree-
ment especially interesting is that, unlike most familiar varieties, it does
not purport to exploit anthropological reports about the moral views of
exotic cultures, or even garden-variety conflicting moral intuitions about
concrete cases (such as abortion or the death penalty). Instead, Nietzsche
locates disagreement at the heart of the most sophisticated moral philoso-
phies of the West, among philosophers who very often share lots of beliefs
and practices and who, especially, in the last century, often share many of
the same judgments about concrete cases.12 Yet what we find is that these

12
  It is important to see that convergence on concrete cases (which is almost always cet-
eris paribus) does not defeat the argument. I suppose no one would think that Mussolini
and Roosevelt really converge on the same moral truths just because they both agree
about the concrete question that normally the trains should run on time. Moral philoso-
phers—at least the conventional kind who subscribe to the propositions in question—are
surely less far apart than Mussolini and Roosevelt, but that does not alter the fact that
 Moral Skepticism in Nietzsche 135

philosophers remain locked in apparently intractable disagreement about

the most important, foundational issues about morality.13
Let us now look at the evidence that Nietzsche advances this argument,
before considering some of the possible objections to it.

their apparent agreement on suitably general and hedged “concrete” moral propositions
belies real disagreements, which come out as soon as we press on the concrete cases.
13
 It may be useful to distinguish the argument at issue here from some related
skeptical-sounding arguments based on the phenomenon of disagreement. One is
“the so-called pessimistic induction on the history of science,” as Philip Kitcher calls it
(1993: 136) (or the skeptical meta-induction as Putnam earlier dubbed the same phe-
nomenon). Here is Kitcher’s statement of the skeptical position:
Here one surveys the discarded theories of the past; points out that these were once
accepted on the basis of the same kind of evidence that we now employ to support our
own accepted theories, notes that those theories are, nevertheless, now regarded as false;
and concludes that our own accepted theories are very probably false. (1993: 136)
This basic argumentative strategy might, indeed, seem to have some force against theories
of morality. After all—so the argument would go—many earlier claims about moral-
ity were based on the same kinds of evidence about what is “intuitively obvious” that
underlie contemporary Kantian and utilitarian theories. Yet we now regard intuitions
about e.g. the obvious moral inferiority of certain classes of people as social or cultural or
economic artifacts, not data on which we might base a moral theory. Is it not possible—
especially with the often surprising results about diversity of intuitions being adduced by
experimental philosophers—that the intuitions undergirding our current moral theories
will also turn out to seem equally unreliable, and so our moral theories false?
This strategy of skeptical argument is easily rebutted, however. To start, many of the
racist and sexist claims of earlier moral theories were based not on intuitions, but on
putatively empirical claims:  Aristotle’s views about “natural” slaves, for whom slavery
was supposed to be in their non-moral interest, or Kant’s disparaging remarks about
Africans, depended on armchair psychological and sociological hypotheses that are not
factually accurate. Indeed, the kind of response to the skeptical induction that Kitcher
develops on behalf of the scientific realist would seem to help the moral realist as well.
For Kitcher says that, in fact, “more and more of the posits of theoretical science endure
within contemporary science” (1993: 136), and, indeed, that our earlier mistakes (which
we now recognize as such) fall into a recognizable pattern, so that we can see where and
why we are likely to have gone wrong in the past, and thus be more confident that we are
not replicating those mistakes in our current theories.
So, too, the moral realist might claim that the mistakes made by earlier moral theo-
rists also fall into a discernible pattern, typically consisting in failing to include within
the moral community—the community of persons with moral standing—people who
belonged there because of false assumptions about those persons that admit of straight-
forward historical, sociological, and economic explanations. Thus, on this story, what
we learn from the history of failures in past moral theories is precisely that we should be
especially skeptical about excluding some persons (or, not to prejudge the issue, some
sentient creatures!) from the category of beings with moral standing. Of course, as every-
one knows, the criteria of moral standing remain hotly contested, a fact to be exploited
by the skeptical argument I will attribute to Nietzsche.
136 Brian Leiter

3.  Nietzsche’s Version of the Skeptical

Argument

There are a set of remarks about moral philosophy and moral philoso-
phers in Nietzsche about which scholars rarely comment, but which bear
directly on the argument for moral skepticism at issue here. This passage is
representative:
It is a very remarkable moment: the Sophists verge upon the first critique of moral-
ity (Moral ), the first insight into morality:—they juxtapose the multiplicity (the
geographical relativity) of the moral value judgments (Moralischen Werthurtheile);—
they let it be known that every morality (Moral) can be dialectically justified; i.e.,
they divine that all attempts to give reasons for morality (Moral) are necessarily
sophistical—a proposition later proved on the grand scale by the ancient philoso-
phers, from Plato onwards (down to Kant);—they postulate the first truth that a
“morality-in-itself ” (eine Moral an sich), a “good-in-itself ” do not exist, that it is a
swindle to talk of “truth” in this field. (WP 428; KSA xiii. 14[116]).
This is a Nachlass passage, but it has many analogues in the published cor-
pus and is of a piece with a general picture Nietzsche has of the discursive
pretensions of philosophers. Consider his derisive comment in Beyond Good
and Evil about Kant’s moral philosophy, which he describes as “[t]‌he . . . stiff
and decorous Tartuffery of the old Kant, as he lures us on the dialecti-
cal bypaths that lead to his ‘categorical imperative’—really lead astray and
seduce” (BGE 5). Kant’s “Tartuffery” and Spinoza’s “hocus-pocus of math-
ematical form” in his Ethics are simply, Nietzsche says, “the subtle tricks

Now in the context of scientific realism, Kitcher wants to draw a stronger conclusion
against the skeptic, namely, that we are actually entitled to a kind of “optimistic induc-
tion” from the fact that since every successor theory “appears closer to the truth than” the
theory it displaced “from the perspective of our current theory,” to the conclusion that
“our theories will appear to our successors to be closer to the truth than our predecessors”
(1993: 137). But the moral theorist cannot avail himself of a similar “optimistic induc-
tion,” and for a reason that will be important to the skeptical argument here: namely, that
it is not the case that e.g. later deontological theories view earlier utilitarian theories as
getting closer to the moral truth than their utilitarian ancestors, and vice versa.
More recently, there has been a lively debate among philosophers about the episte-
mological implications of disagreement among what are usually called “epistemic peers.”
What is standardly at issue in this literature is whether or not the fact of such disa-
greement should lead us to adjust the degree of credence an agent assigns to his own
beliefs (see e.g. Christensen 2007 and Kelly 2005 for contrasting views). By contrast,
the skeptical argument at issue here aims for a metaphysical conclusion via an abductive
inference: namely, that the fact of disagreement about X is best explained by there not
being any objective fact of the matter about X. As I read it, the disagreement literature
to date does not weigh the epistemic import of a successful abductive inference for this
kind of skepticism.
 Moral Skepticism in Nietzsche 137

(feinen Tücken) of old moralists and preachers of morals (Moralisten und

Moralprediger).” As Nietzsche explains it:
They all pose as if they had discovered and reached their real opinions through the
self-development of a cold, pure, divinely unconcerned dialectic . . . while at bot-
tom it is an assumption, a hunch, a kind of “inspiration”—most often a desire of
the heart that has been filtered and made abstract—that they defend with reasons
sought after the fact. They are all advocates who don’t want to be called by that
name, and for the most part even wily spokesmen for their prejudices which they
baptize “truths.” (BGE 5)
Later in the same book, Nietzsche notes that moral philosophers “make one
laugh” with their idea of “morality as science,” their pursuit of “a rational
foundation for morality,” which “seen clearly in the light of day” is really
only a “scholarly form of good faith in the dominant morality, a new way
of expressing it.” Pointing at Schopenhauer’s attempt to supply a rational
foundation for morality, Nietzsche says “we can draw our conclusions as
to how scientific a ‘science’ could be when its ultimate masters still talk
like children” (BGE 186). The real significance of the claims of moral phi-
losophers is “what they tell us about those who make them” for they are “a
sign-language of the affects” (BGE 187), betraying things about the psycho-
logical needs and condition of those who make them.14
How do these considerations, elliptical as some of them are, support a
skeptical conclusion about the objective existence of moral facts or proper-
ties? Recall the passage with which we began: Nietzsche claims that the key
insight of the Sophists into morality was that “every morality (Moral) can
be dialectically justified; i.e., they divine that all attempts to give reasons
for morality (Moral) are necessarily sophistical—a proposition later proved
on the grand scale by the ancient philosophers, from Plato onwards (down
to Kant)” (WP 428). The Sophists, on this account, advance two closely
related claims: (1) that “every morality can be dialectically justified” and
(2) that “all attempts to give reasons for morality are necessarily sophistical,”
where “sophistical” is obviously meant to have the pejorative connotation
that the apparent dialectical justification does not, in fact, secure the truth
of the moral propositions so justified. The purported dialectical justifica-
tion can fail in this way if either it is not a valid argument or some of the

14
  In fact, Nietzsche thinks this last point applies quite generally, not only to moral
philosophers. He frequently describes (see e.g. D book 1 or GS 335) moral judgments
as caused by certain feelings, learnt through a combination of customary practices and
parental influence, while the moral concepts and reasons people offer for these judgments
are merely post-hoc (cf. D 34).
138 Brian Leiter

premises are false.15 But, then, what is the force of the claim that “every
morality can be dialectically justified”? It must obviously be that every
morality can have the appearance of being dialectically justified, either
because its logical invalidity is not apparent or, more likely in this instance,
because its premises, while apparently acceptable, are not true.
Yet Nietzsche goes further when he asserts that the second claim—namely,
that “all attempts to give reasons for morality are necessarily sophistical”—is
established (“proved” (beweisen) he says) by the work of the philosophers
from Plato through to Kant (though he would presumably add, as the other
passages make clear, Schopenhauer to the list of evidence). But in what
sense do the moral philosophies of Plato, Aristotle, the Stoics, Hutcheson,
Mill, Kant, and Schopenhauer et al. establish or “prove” that “all attempts
to give reasons for morality are necessarily sophistical”? Nietzsche’s thought
must be that all these philosophers appear to provide “dialectical justifica-
tions” for moral propositions, but that all these justifications actually fail.
But that still does not answer the question of how the fact of there being all
these different moral philosophies proves that they are sophistical, i.e. that
they do not, in fact, justify certain fundamental moral propositions?
The best explanation argument sketched earlier would supply Nietzsche
an answer. The best explanation for the existence of incompatible moral phi-
losophies providing dialectical justifications for conflicting moral truths is
that (1) it is possible to construct apparent dialectical justifications for such
moral truths, because (2) given the diversity of psychological needs of per-
sons (including philosophers), it is always possible to find people for whom
the premises of these dialectical justifications seem plausible and attractive,
and (3) there are no objective moral facts offering an obstacle to the philoso-
pher satisfying his psychological needs in this way.
The alternative, “moral realist” explanation for the data—the data
being the existence of intractable disagreement between incompat-
ible philosophical theories about morality—is both less simple and less
consilient. First, of course, it posits the existence of moral facts which,
according to the more familiar best explanation argument noted earlier
(cf. Leiter 2001), are not part of the best explanation of other phenom-
ena. Second, the moral realist must suppose that this class of explanator­
ily narrow moral facts is undetected by a large number of philosophers
who are otherwise deemed to be rational and epistemically informed.
Third, the moral realist must explain why there is a failure of convergence

15
 Whether or not Nietzsche is thinking of this issue in Aristotelian terms is not
clear, though it might seem the natural candidate point of reference for a classicist like
Nietzsche, but I have found, in any case, the discussion in Smith (2007) helpful in fram-
ing the possibilities at issue.
 Moral Skepticism in Nietzsche 139

under what appear (and purport) to be epistemically ideal conditions

of sustained philosophical inquiry and reflective contemplation across
millennia. We can agree with Railton that we lack “canons of induc-
tion so powerful that experience would, in the limit, produce conver-
gence on matters of fact among all epistemic agents, no matter what
their starting points” (1986a: 6), and still note that there exists a remark-
able cross-cultural consensus among theorists about fundamental physi-
cal laws, principles of chemistry, and biological explanations, as well as
mathematical truths, while moral philosophers, to this very day, find
no common ground on foundational principles even within the West,
let  alone cross-culturally. How can a moral realist explain this? Let us
consider, now, some possibilities.

4.  Objections to the Skeptical Argument

from Moral Disagreement

Moral realists—which, for purposes here, will just mean those who deny
skepticism about objective moral facts—have developed a variety of “defus-
ing explanations” (Doris and Plakias 2008: 311, 320–1; cf. Loeb 1998 for
a useful survey and rebuttal of various strategies) to block the abductive
inference from apparently intractable moral disagreement to skepticism
about moral facts. Moral disagreement is, after all, an epistemic phenom-
enon, from which we propose to draw a metaphysical conclusion. The
“defusing” explanations of moral disagreement propose to exploit that fact,
by suggesting alternate epistemic explanations for the disagreement, expla-
nations that are compatible with the existence of objective moral facts.
We may summarize the “defusing” objections to the skeptical argument as
follows: (1) moral disagreements about concrete cases are not really intrac-
table, they merely reflect factual disagreements or ignorance, and thus
belie agreement on basic moral principles; (2) even if moral disagreements
are about basic moral principles, they are not really intractable but rather
resolvable in principle; (3) even if there are real and intractable moral dis­
agreements about foundational moral principles, these are best explained
by cognitive defects or the fact that they occur under conditions that are
not epistemically ideal:  e.g. conditions of informational ignorance, irra-
tionality, or partiality; and (4) even if there are real and intractable moral
disagreements about foundational moral principles that cannot be chalked
up to cognitive defects or non-ideal epistemic conditions, they are still best
explained in terms of differences in “background theory.” Let us consider
these in turn.
140 Brian Leiter

1. Moral disagreements about concrete cases are not really intractable, they
merely reflect factual disagreements or ignorance, and thus belie agreement on
basic moral principles. Although this was an important worry in, for example,
the response of Boyd (1988) and Brink (1989) to Mackie’s original version
(1977) of the argument from moral disagreement, it is obviously irrelevant
to Nietzsche’s version of the argument for moral skepticism, which appeals
precisely to disagreement about foundational moral principles, as exempli-
fied, for example, by the dispute between Kantians and utilitarians, among
many others. So we may set this earlier defusing explanation to one side.
2. Even if moral disagreements are about basic moral principles, they are
not really intractable but rather are resolvable in principle. This has been the
standard optimistic refrain from philosophers ever since “moral realism”
was revived as a serious philosophical position in Anglophone philosophy
in the 1980s. With respect to very particularized moral disagreements—e.g.
about questions of economic or social policy—which often trade on obvious
factual ignorance or disagreement about complicated empirical questions,
this seems a plausible retort. But for over two hundred years, Kantians and
utilitarians have been developing increasingly systematic versions of their
respective positions. The Aristotelian tradition in moral philosophy has an
even longer history. Utilitarians have become particularly adept at explain-
ing how they can accommodate Kantian and Aristotelian intuitions about
particular cases and issues, though in ways that are usually found to be
systematically unpersuasive to the competing traditions and which, in any
case, do nothing to dissolve the disagreement about the underlying moral
criteria and categories. Philosophers in each tradition increasingly talk only
to each other, without even trying to convince those in the other tradi-
tions. And while there may well be “progress” within traditions—e.g. most
utilitarians regard Mill as an improvement on Bentham—there does not
appear to be any progress in moral theory, in the sense of a consensus that
particular fundamental theories of right action and the good life are deemed
better than their predecessors. What we find now are simply the compet-
ing traditions—Kantian, Humean, Millian, Aristotelian, Thomist, perhaps
now even Nietzschean—who often view their competitors as unintelligible
or morally obtuse, but don’t have any actual arguments against the foun-
dational principles of their competitors. There is, in short, no sign—I can
think of none—that we are heading towards any epistemic rapprochement
between these competing moral traditions. So why exactly are we supposed
to be optimistic?
As grounds for optimism, many philosophers appeal to the thought (due
to Derek Parfit) that secular “moral theory” is a young field, so of course
it has not made much progress. This strikes me as implausible for a variety
 Moral Skepticism in Nietzsche 141

of reasons. First, most fields with factual subject matters have usually man-
aged to make progress, as measured by convergence among researchers,
over the course of a century—and especially the last century, with the rise
of research universities. Moral theory is, again, the odd man out, when
compared to physics, chemistry, biology, or mathematics. Even psychol-
ogy, the most epistemically robust of the “human” sciences, managed to
make progress: e.g. the repudiation of behaviorism, and the cognitive turn
in psychology in just the last fifty years. Second, Spinzoa, Hume, Mill, and
Sidgwick (among many others) may not have advertised their secularism,
but the idea that their moral theories are for that reason discontinuous with
the work of the past hundred years does obvious intellectual violence to the
chains of influence of ideas and arguments. Third, and relatedly, so-called
“secular” moral theory regularly conceives itself in relation to a history that
stretches back in time (sometimes back to the Greeks)—contrast that with
the relative youth of modern physics!—so that it becomes unclear why the
bogeyman of the deity was supposed to have constituted the insuperable
obstacle weighing down intellectual progress. Most contemporary deontol-
ogists may be atheists, for example, but it is not obvious that their atheism
enabled them to make stunning intellectual progress beyond Kant.16
If there is a reason for optimism, it will have to be sought in the next
argument.
3. Even if there are real and intractable moral disagreements about founda-
tional moral principles, these are best explained by cognitive defects or the fact
that they occur under conditions that are not epistemically ideal: e.g. conditions
of informational ignorance, irrationality, or partiality. This is, again, a famil-
iar move in the metaethical literature responding to the argument from
moral disagreement, but one must appreciate how strange it is in response
to the Nietzschean argument appealing to disagreement among moral phi-
losophers across millennia. Are we really to believe that hyper-rational and
reflective moral philosophers, whose lives, in most cases, are devoted to
systematic reflection on philosophical questions, many of whom (histori-
cally) were independently wealthy (or indifferent to material success) and
so immune to crass considerations of livelihood and material self-interest,
and most of whom, in the modern era, spend professional careers refining

16
  Alistair Norcross suggests to me that the real problem is that ethics requires reliance
on “intuitions,” and our intuitions are still strongly tainted by our religious traditions.
That seems a more plausible point, though it is unclear what criteria we are going to
appeal to in order to sort the “tainted” from “untainted” intuitions. As Nietzsche would
be the first to point out, the utilitarian obsession with sentience and suffering is, itself,
indebted to Christianity—an ironic fact, given the centrality of the wrongness of suffer-
ing to Parfit’s own moral philosophy (e.g. 2011b: 565 ff.).
142 Brian Leiter

their positions, and have been doing so as a professional class in university

settings for well over a century—are we really supposed to believe that they
have reached no substantial agreement on any foundational moral principle
because of ignorance, irrationality, or partiality?17
Ignorance seems especially easy to dismiss as a relevant consideration.
As Don Loeb puts the point:  “It seems very unlikely that the continued
existence of [the] debate [between Kantians and utilitarians] hinges upon
disagreement over the non-moral facts” (1998: 290). What non-moral facts
exactly bear on the question, for example, whether respect for the dignity of
persons or maximization of utility is the criterion of rightness?
Take a stark, and very au courant, example:  Louis Kaplow and Steven
Shavell (2002), leading law and economics scholars at Harvard Law School,
published a decade ago a massive tome arguing against the relevance of
“fairness” considerations in social policy. Their argument—I do not believe
this is an over-simplification—is, in essence, that since doing what is
“fair” is not always Pareto optimal, it is irrational to make policy based
on considerations of fairness. Kaplow and Shavell are not stupid; they are
not ignorant; they are not obviously irrational. But they do believe that if
doing what is “fair” is not Pareto optimal, it is, itself, obviously irrational.
Have Kaplow and Shavell made a factual error? A rational error? They may,
indeed, be dogmatic, but are they any less dogmatic than Kantian moral
philosophers, who rarely spend time with their hyper-utilitarian colleagues
in the economics department?
Nietzsche, in fact, presents a fine armchair test case for any thesis about
moral disagreement, since he so clearly repudiates “the egalitarian prem-
ise of all contemporary moral and political theory—the premise, in one
form or another, of the equal worth and dignity of each person” (Leiter
2002: 290). For Nietzsche is not only quite prepared, like any consequen-
tialist, to sacrifice the well-being of some for others; he often seems ready
to sacrifice the well-being of the majority for the sake of the flourishing of
his favored examples of human excellence like Goethe (Leiter 2002: 113–
36)—a view, that is, I presume, uncongenial to the vast majority of aca-
demic moral theorists! Here, then, is a stark philosophical challenge for
moral realists: “defuse” Nietzsche’s disagreement by reference to a cognitive
defect of some kind: e.g. a failure to appreciate non-moral facts or norms of

17
  Of course, Nietzsche himself does deny that philosophers, at least great ones, are
impartial—as he puts it, they are “all advocates who don’t want to be called by that name”
and “wily spokesmen for their prejudices which they baptize ‘truths’ ” (BGE 5). If moral
philosophers were to cede this point to Nietzsche, then, of course, they would have for-
feited their claim to justified moral knowledge.
 Moral Skepticism in Nietzsche 143

rationality.18 This is, of course, just a version of Hume’s famous challenge to

explain the offense to reason in preferring the destruction of the world to
a thumb prick, though in Nietzsche’s case the options are more troubling
because of the greater resonance they are likely to have for cosmopolitan
moral philosophers:  after all, if it were really true, as Nietzsche believes,
that a culture suffused with moral norms of equality really would prevent
the developments of Goethes and Beethovens, how exactly is it irrational
to prefer an inegalitarian culture that makes human excellence possible?19
Yet surely it is possible that some heretofore unrecognized cognitive
deficiencies of academic moral philosophy of the past 250  years explain
the failure of even a modicum of consensus on foundational moral prin-
ciples to emerge. Indeed, perhaps the lack of progress in moral philosophy
is proof precisely of the epistemically defective condition of the discourse
to date!20 Nothing in the argument so far rules out that possibility, but
why in the world suppose this is the correct explanation for the state of
affairs we find? Certainly no moral philosophers of the past two centuries
would want to admit to such cognitive frailties, nor is it the case that we
have any non-question-begging account of what exactly those deficiencies
might have been. As between the two explanatory hypotheses—one based
on skepticism about moral facts and one based on its denial—there is surely
an enormous burden of proof for the proponent of the latter to explain the
nature and character of the epistemic failings that have blocked access to
the moral facts.
4. Even if there are real and intractable moral disagreements about foun-
dational moral principles that cannot be chalked up to cognitive defects or
non-ideal epistemic conditions, they are still best explained in terms of differ-
ences in “background theory.” This “defusing explanation” was developed
originally against arguments from moral disagreement appealing to very

18
  I realize, of course, that “Kantians,” from Kant to Gewirth to Korsgaard, purport
to have arguments showing such positions to be irrational, but the voluminous literature
attacking their positions perhaps encourages the skeptical thought that something has
gone awry.
19
  One might observe, of course, that most philosophers do accept the egalitarian
premise, but they interpret it in ways that yield very divergent conclusions. And even
the fact that they converge on this point admits of anti-realist explanations, as I argue in
Leiter (2013).
20
 David Enoch suggests to me that perhaps philosophical tools are not the right
way of achieving knowledge of moral truths. The alternatives—e.g. reading the Bible or
intuition—are notoriously unreliable epistemic methods, however, that generate even
more disagreement than the traditional discursive methods of philosophy. In any case,
Parfittian optimism about secular, rational moral philosophy has been the default posi-
tion for philosophers, which is why it is important to make it the target here.
144 Brian Leiter

particular moral judgments. Against the familiar fact that people’s moral
intuitions about particular problems are often quite different, it is easy to
reply, as Loeb puts it, that since “all observation is theory laden . . . theo-
retical considerations will play a role in moral observations, just as they
do in any others,” and thus “differences of belief among moral reasoners
should be expected because the same information will be observed differ-
ently depending on what background theories are present” (1998: 288).
The skeptical argument from moral disagreement among systematic
moral philosophies, as Loeb himself discusses, presents two discrete chal-
lenges to this defusing explanation. First, it is quite possible for Kantians
and utilitarians to agree about the right action in particular cases, while
disagreeing about the reasons the action is right, reflecting their disagree-
ment about fundamental moral facts.21 In these cases, the disagreement we
are trying to explain is precisely the disagreement in the “background the-
ory,” and it is the surprising resilience of such disagreements, so the skeptic
argues, that calls out for skepticism about moral facts. Second, where the
disagreement about particular cases stems from differing background theo-
ries, that hardly defuses an argument from skepticism appealing to intract­
able differences about background theories. As Doris and Plakias remark, in
considering a more extreme case: “if our disagreement with the Nazis about
the merits of genocide is a function of a disagreement about the plausibility
of constructing our world in terms of pan-Aryan destiny, does it look more
superficial for that?” (2008: 321). Of course, in the Nazi case, we might
think the Nazi background theory vulnerable on other grounds (e.g. of fac-
tual error or partiality), but, as we have already noted, it is not at all obvious
how a disagreement informed by differing moral theories—say, Kantian
and utilitarian—is in any way defused by noting that the disputants dis­
agree not only about the particular case, but about the foundational moral
propositions which bear on the evaluation of the case.22

21
  Clarke-Doane has posed to me the question: why isn’t agreement on particular cases
enough? After all, we don’t let disagreement among philosophers of physics undermine
our confidence in the objectivity of physics, as manifest in the massive agreement among
scientists about particular propositions of physical theory. Of course, the “folk” do not
converge in their moral opinions the way the physicists do in theirs—that, of course, was
why defenders of moral realism first urged moving to the theoretical level where, it was sug-
gested, disagreement would evaporate. But having moved to that level, we notice, as I con-
cede in the text, that philosophers often agree on particular cases but disagree at the level of
the foundational principles. The difficulty, though, is that such agreements appear to mask
disagreement in principles, and the disagreement in principles actually does translate into
differences about particular cases, even if it is compatible with convergence on many others.
22
  A more promising suggestion, which I owe to an anonymous referee, would be to
appeal to differing background theories of rationality. This raises several issues, however.
First, it is unlikely that such disagreements suffice to defuse all moral disagreements (not
 Moral Skepticism in Nietzsche 145

5.  Has the Argument Proved Too Much?

I want to conclude the objections to the skeptical argument from disagree-

ment by considering three final worries: one interpretive—about saddling
Nietzsche with the kind of moral skepticism at issue here—and two philo-
sophical, pertaining to whether the argument sketched above has proved too
much and, relatedly, whether it is self-referentially defeating.
On the interpretive question, it seems to me that nothing has misled
readers more often about Nietzsche’s metaethical view than the volume of
his rhetoric: he writes (so the argument goes) as if there really is a fact of
the matter about his judgments about the value of human greatness and
the disvalue of Christianity and the herd and the rabble. In fact, however,
Nietzsche’s notorious rhetorical excessses make, I  think, at least as much
(perhaps even more) sense on the anti-realist picture. For if Nietzsche is a
moral anti-realist committed to the polemical project of disabusing certain
readers of their “false consciousness” about morality—their false belief that
it is good for them—then he has every reason to use all available rhetorical
devices—both rational and non-rational—to achieve that end. Indeed, rec-
ognizing that ours is a world without any objective moral truths, Nietzsche
has a special reason to write most of the time as if his own (subjective) judg-
ments of value were something other than matters of evaluative taste: for if
they can claim a kind of epistemic and practical authority to which they are
not really entitled, then they are more likely to influence belief and action,
at least among readers who view truth as practically important (as Nietzsche
supposes his readers will). Yet Nietzsche himself sometimes does admit the
“terrible truth” about the subjective character of his evaluative judgments—
as when he says that, “What is now decisive against Christianity is our taste
(Geschmack), no longer our reasons” (The Gay Science, §132) and when he
describes the “revaluation of Christian values” as an “attempt, undertaken
with every means” to bring “the counter-values (die Gegen-Werte) . . . to vic-
tory” (The Antichrist, §61)—not the “true” values or the “objectively cor-
rect” ones, but simply the opposite ones, the ones that appeal to a very
different taste.
That brings us to the final philosophical objections to the line of skepti-
cal argument explored here. The first is an objection that, no doubt, has
already occurred to everyone who has gotten this far. Is not the apparently

all utilitarians e.g. are committed only to instrumental theories of rationality). Second,
one can worry that in some cases the disagreements about rationality really constitute
part of the foundational moral disagreement, rather than standing apart from it as a
free-standing bit of the background theory.
146 Brian Leiter

intractable disagreement among moral philosophers regarding foundational

questions mirrored in many other parts of our discipline? Are not metaphy-
sicians and epistemologists also not locked in intractable disagreements of
their own? Think of debates between internalists and externalists in episte-
mology, or between presentists and four-dimensionalists in the philosophy
of time. If disagreement among moral philosophers supports an abductive
inference to denying the existence of moral facts, what, if anything, blocks
that inference in all these other cases?
Some recent writers (such as Bloomfield 2004 and Shafer-Landau
2005)  think this kind of “companions in guilt” consideration counts in
favor of moral realism, notwithstanding the disagreement among moral
philosophers. It is not entirely clear why they rule out, however, the other
natural conclusion. Nietzsche, as far as I can see, has no reason to resist it,
since he believes that, as an explanatory matter, the moral commitments
of the philosopher—at least the great philosopher—are primary when it
comes to his metaphysics and epistemology. Nietzsche writes:
I have gradually come to realize what every great philosophy so far has been: namely
the personal confession of its author and a kind of involuntary and unconscious
memoir; in short, that the moral (or immoral) intentions in every philosophy con-
stitute the true living seed from which the whole plant has always grown. In fact, to
explain how the strangest metaphysical claims of a philosopher really come about,
it is always good (and wise) to begin by asking: at what morality does it (does he—)
aim? (BGE 6)23
Since, for Nietzsche, the “morality” at which the philosopher aims is to be
explained in terms of his psychological needs and drives, and since these dif-
fer among philosophers, it will be unsurprising that there are a diversity of
moral views, and philosophical systems purportedly justifying them—and
it will be equally unsurprising that this same diversity, and intractability,
spills over into metaphysical and epistemological systems, since they are
just parasitic on the moral aims of the philosophers! Nietzsche, at least,
then has good reason to bite the skeptical bullet about much philosophical
disagreement.

23
 Nietzsche’s thesis was explicitly about the “great philosophies”—like Kant and
Spinoza—and not those “philosophical laborers” and “scholars” who possess “some small,
independent clockwork that, once well wound, works on vigorously without any essential
participation from all the other drives of the scholar” (Beyond Good and Evil, 6). Many
professional philosophers may, indeed, be laboring away at problems in a “disinterested”
way. Still, as the recent survey by David Bourget and David Chalmers (see <http://phil-
papers.org/surveys/results.pl>) brought out, there are striking, and surely not accidental,
correlations between philosophical views across different areas:  e.g. theism and moral
realism and libertarianism about free will. Even the “philosophical laborers” are not
wholly disinterested inquirers!
 Moral Skepticism in Nietzsche 147

Of course, we would need to think carefully about individual cases of

philosophical disagreement, since not all of them, in all branches of phi-
losophy, are as intractable or as foundational as they are in moral philoso-
phy. Some philosophical disagreements can, in fact, be defused fairly easily.
Thus, to take an example from one of my other fields, the debate in legal
philosophy between natural law theorists and legal positivists about the
nature of law has both an element of tractability (natural law theorists like
Finnis have, in fact, conceded most of the claims that actually matter to
legal positivism as a theory of law24) and admits, in the intractable parts,
of defusing by reference to the transparent and dogmatic religious com-
mitments of the natural law theorists on the remaining issues they refuse
to cede. In sum, the skeptical argument from disagreement among philoso-
phers may have implications beyond moral philosophy, but what precisely
they are will have to be decided on a case-by-case basis.
That still leaves a slightly different version of the worry that the argu-
ment “proves too much.” For surely most philosophers will not conclude
from the fact of disagreement among moral philosophers about the fun-
damental criteria of moral rightness and goodness that there is no fact of
the matter about these questions, as I claim Nietzsche does. But why not
think that this meta-disagreement itself does not warrant a skeptical infer-
ence, i.e. there is no fact about whether we should infer moral skepticism
from the fact of disagreement about fundamental principles among moral
philosophers, since philosophers have intractable disagreements about what
inferences the fact of disagreement supports?
Again, however, we need to be careful about the data points and the
abductive inferences they warrant. The question is always what is the best
explanation for the disagreement in question, given its character and scope.
The “meta-disagreement”—about whether disagreement in foundational
moral theory really warrants skepticism about moral facts—is, itself, of
extremely recent vintage, barely discussed in the literature.25 Even if this
chapter and the challenge in Loeb (1998), for example, succeed in mak-
ing the issue a topic of debate,26 and even if, after some critical discussion,
the meta-disagreement continues to persist, that still would not support

24
  See e.g. Leiter 2007: 162–4, including n. 42.
25
  A simple version of the argument featured at the start of MacIntyre (1981), but was
largely neglected in subsequent scholarly discussion. Three decades of moral philosophy
later, the skeptical case is even stronger.
26
  Obviously Pyhrronian skeptics have mounted challenges of this form for a very
long time, but they are not specific to ethics and they do not offer them to support
an abductive inference to a metaphysical conclusion, as Loeb and I  do. Jessica Berry
(2011) argues that Nietzsche’s point is, in fact, the Pyhrronian one—namely, to elicit a
suspension of belief—though I think this is hard to square with his rhetoric. I am also
148 Brian Leiter

the meta-skeptical conclusion that there is no fact of the matter about

whether or not disagreement in foundational moral theory supports skep-
ticism about moral facts. For before we are entitled to that conclusion, we
would have to ask what the best explanation for the meta-disagreement really
is? Surely one possibility—dare I  say the most likely possibility?—is that
those who are professionally invested in normative moral theory as a seri-
ous, cognitive discipline—rather than seeing it, as Marxists or Nietzscheans
might, as a series of elaborate post-hoc rationalizations for the emotional
attachments and psychological needs of certain types of people (bourgeois
academics, “slavish” types of psyches)—will resist, with any dialectical tricks
at their disposal, the possibility that their entire livelihood is predicated on
the existence of ethnographically bounded sociological and psychological
artifacts. Nothing in the argument here establishes that conclusion, but nor
is there any reason to think it would not be the correct one in the face of
meta-disagreement about the import of fundamental disagreement in moral
philosophy.

6. Conclusion

If disagreement in science were as profound and rampant as it is in moral

philosophy, we would expect proponents of “Intelligent Design” crea-
tionism to be lined up against evolutionary theorists in biology depart-
ments and defenders of teleological explanation to be doing battle with
the believers in mechanical causation in physics. We would expect dis-
course in physics and mathematics and chemistry to be circumscribed
by geographic and cultural boundaries, such that Japanese mathemati-
cians and Chinese physicists were engaged in a largely separate world
of intellectual discourse from their American and German counter-
parts, just like their colleagues in moral philosophy are. But everyone
outside philosophy, and at least some within it, knows that profound
and intractable disagreement about foundational moral questions is the
basic fact about the field. The fact of such disagreement—apparent to the
Sophists in antiquity, and revived as an important skeptical considera-
tion by Nietzsche in the nineteenth century—should be a live issue for
us today. As philosophers, we should forget about “folk” disagreement,

skeptical that “the passages in which Nietzsche does embrace caution, ephexis, and suspi-
cion . . . far and away outnumber those in which he sounds adamant and dogmatic” about
the non-existence of moral facts (Berry 2011: 190). She offers her alternative, Pyhrronian
reading of the disagreement passages in Nietzsche at (2011: 184–208).
 Moral Skepticism in Nietzsche 149

and instead confront the far more problematic phenomenon:  namely,

“expert” disagreement among those who devote their professional lives
to systematic and rational reflection on moral questions, and who often
share, notwithstanding this disagreement, lots of the same moral convic-
tions about concrete cases, as well as often sharing similar lifestyles and
cultural experiences. If Nietzsche is right, the best explanation for what
we find is that, when it comes to moral theorizing, it really is a “swindle”
to talk of truth in this field.27

References to Nietzsche
I have drawn on English translations by Walter Kaufmann, R. J. Hollingdale, or
Maudemarie Clark and Alan Swensen, and then made modifications based on
Friedrich Nietzsche, Sämtliche Werke: Kritische Studienausgabe in 15 Bänden, ed.
G. Colli and M. Montinari (Berlin: de Gruyter, 1980); where there is no exist-
ing English edn, the translation is my own. Nietzsche’s works are cited as fol-
lows, unless otherwise noted: roman numerals refer to major parts or chapters in
Nietzsche’s works; Arabic numerals refer to sections, not pages. I use the standard
Anglophone abbreviations for Nietzsche’s works, as follows:  The Antichrist (A);
Beyond Good and Evil (BGE); The Birth of Tragedy (BT); Daybreak (D); Ecce Homo
(EH); The Gay Science (GS); Thus Spoke Zarathustra (Z); Twilight of the Idols (TI);
The Will to Power (WP).

27
  I am grateful to John Doris and Don Loeb for extremely helpful comments on the
earliest version of this chapter. A later version benefited from presentation and discussion
at the annual History of Modern Philosophy conference at New York University in Nov.
2008; I should acknowledge, in particular, my commentator on that occasion, R. Lanier
Anderson, as well as important questions from Anja Jauernig and Ernest Sosa. I am also
grateful to Justin Clarke-Doane for very helpful written feedback on that version. The
current version benefited from discussion at the Moral Sciences Club at Cambridge
University and the Practical Philosophy Workshop at Northwestern University, as well
as philosophy workshops at McMaster University, University of Colorado at Boulder,
Georgetown University, Queen’s University in Canada, and the Graduate Center of the
City University of New York; from many instructive comments on a shorter version at
the “On the Human” blog of the National Humanities Center (I should single out espe-
cially Ralph Wedgwood, Michael Ridge, and, again, Don Loeb); and from a lively discus-
sion at the 9th Annual Metaethics Workshop at the University of Wisconsin, Madison
in Sept. 2012.
150 Brian Leiter

Other References
Anderson, R. Lanier. 1998. “Truth and Objectivity in Perspectivism,” Synthese,
115: 1–32.
Berry, Jessica. 2011. Nietzsche and the Ancient Skeptical Tradition. New York: Oxford
University Press.
Bloomfield, Paul. 2004. Moral Reality. New York: Oxford University Press.
Boyd, Richard. 1988. “How to be a Moral Realist,” in G. Sayre-McCord (ed.), Essays
on Moral Realism. Ithaca, NY: Cornell University Press.
Brink, David. 1989. Moral Realism and the Foundations of Ethics.
Cambridge: Cambridge University Press.
Christensen, David. 2007. “The Epistemology of Disagreement: the Good News,”
Philosophical Review, 116: 187–217.
Clark, Maudemarie. 1990. Nietzsche on Truth and Philosophy. Cambridge: Cambridge
University Press.
Clark, Maudemarie. 1998. “On Knowledge, Truth and Value,” in C. Janaway (ed.),
Willingness and Nothingness: Schopenhauer as Nietzsche’s Educator. Oxford: Oxford
University Press.
Clarke-Doane, Justin. Forthcoming. “Moral Epistemology:  The Mathematics
Analogy,” Noûs.
Doris, John M., and Plakias, Alexandra. 2008. “How to Argue about
Disagreement: Evaluative Diversity and Moral Realism,” in W. Sinnott-Armstrong
(ed.), Moral Psychology, ii. The Cognitive Science of Morality: Intuition and Diversity.
Cambridge, MA: MIT Press.
Hussain, Nadeem. 2004. “Nietzsche’s Positivism,” European Journal of Philosophy,
12: 326–68.
Hussain, Nadeem. 2007. “Honest Illusion:  Valuing for Nietzsche’s Free Spirits,”
in B.  Leiter and N.  Sinhababu (eds), Nietzsche and Morality. Oxford:  Oxford
University Press.
Hussain, Nadeem. 2013. “Nietzsche’s Metaethics,” in K. Gemes and J. Richardson
(eds), The Oxford Handbook of Nietzsche. Oxford: Oxford University Press.
Janaway, Christopher. 2007. Beyond Selflessness:  Reading Nietzsche’s Genealogy.
Oxford: Oxford University Press.
Jech, Thomas. 2002. “Set Theory.” Stanford Encyclopedia of Philosophy: <http://plato.
stanford.edu/entries/set-theory>.
Kaplow, Louis, and Shavell, Steven. 2002. Fairness versus Welfare. Cambridge, MA:
Harvard University Press.
Kelly, Thomas. 2005. “The Epistemic Significance of Disagreement,” in J. Hawthorne
and T. G. Szabo (eds), Oxford Studies in Epistemology, i.  Oxford:  Oxford
University Press.
Kitcher, Philip. 1993. The Advancement of Science. New York: Oxford University Press.
Leiter, Brian. 1994. “Perspectivism in Nietzsche’s Genealogy of Morals,” in
R. Schacht (ed.), Nietzsche, Genealogy, Morality. Berkeley, CA:  University of
California Press.
 Moral Skepticism in Nietzsche 151

Leiter, Brian. 2000. “Nietzsche’s Metaethics:  Against the Privilege Readings,”

European Journal of Philosophy, 8: 277–97.
Leiter, Brian. 2001. “Moral Facts and Best Explanations,” Social Philosophy and
Policy, 18: 79–101. Reprinted in Leiter (2007) (page citations are to the reprinted
version).
Leiter, Brian. 2002. Nietzsche on Morality. London: Routledge.
Leiter, Brian. 2007. Naturalizing Jurisprudence: Essays on American Legal Realism and
Naturalism in Legal Philosophy. Oxford: Oxford University Press.
Leiter, Brian. 2013. “The Boundaries of the Moral (and Legal) Community,”
Alabama Law Review, 64: 511–31.
Leiter, Brian and Sinhababu, Neil (eds). 2007. Nietzsche and Morality.
Oxford: Oxford University Press.
Loeb, Don. 1998. “Moral Realism and the Argument from Disagreement,”
Philosophical Studies, 90: 281–303.
MacIntyre, Alasdair. 1981. After Virtue. Notre Dame, IN:  University of Notre
Dame Press.
Mackie, John. 1977. Ethics: Inventing Right and Wrong. London: Penguin.
Parfit, Derek. 2011a. On What Matters, i. Oxford: Oxford University Press.
Parfit, Derek. 2011b. On What Matters, ii. Oxford: Oxford University Press.
Poellner, Peter. 2001. “Perspectival Truth,” in J. Richardson and B. Leiter (eds),
Nietzsche. Oxford: Oxford University Press.
Railton, Peter. 1986a. “Facts and Values,” Philosophical Topics, 14: 5–31.
Railton, Peter. 1986b. “Moral Realism,” Philosophical Review, 95: 163–207.
Richardson, John. 1996. Nietzsche’s System. Oxford: Oxford University Press.
Schroeder, Mark. 2001. ‘Review of Derek Parfit’s On What Matters, Volume
1 and 2’, Notre Dame Philosophical Reviews (1 Aug.):  <http://ndpr.nd.edu/
news/25393-on-what-matters-volumes-1-and-2>.
Shafer-Landau, Russ. 2005. Moral Realism:  A  Defence. Oxford:  Oxford
University Press.
Sinhababu, Neil. 2007. “Vengeful Thinking and Moral Epistemology,” in B.
Leiter and N.  Sinhababu (eds), Nietzsche and Morality. Oxford:  Oxford
University Press.
Smith, Robin. 2007. “Aristotle’s Logic,” Stanford Encyclopedia of Philosophy (14
Dec.): <http://plato.stanford.edu/entries/aristotle-logic>.
 7
Moral Vagueness: A Dilemma for
Non-Naturalism
Cristian Constantinescu

Pythagoreans, we are told, assigned the number 4 to justice. Probably par­

odying their claim, Socrates announces in the Republic that the difference
between the just and the unjust man in respect of pleasure and pain is 729
(587b–588a). Whether intended or not, the effect of such claims is utterly
comedic—almost on a par with declaring the meaning of life to be 42. By
contrast, Aristotle sounds much more sensible when, reflecting upon the
subject matter of his inquiry in the Nicomachean Ethics, he advises that
“we must be content, in speaking of [fine and just actions, and of goods in
general], to indicate the truth roughly, . . . for it is the mark of an educated
man to look for precision in each class of things just so far as the nature of
the subject admits” (1094b, 19–25). Common sense appears to side with
Aristotle here:  the very attempt to assign cardinal numbers to essentially
imprecise, indeterminate values like justice, happiness, goodness, and the
like, strikes most of us as deeply incongruous. Moral values belong to the
throbbing centre of human life and, as Mark Sainsbury reminds us, “the
throbbing centres of our lives appear to be describable only in vague terms”
(1996: 251).
I shall take up here the particular type of indeterminacy invoked by
Sainsbury: the vagueness of our moral terms. Predicates like “just,” “happy,”
“cruel,” “generous,” “good,” “permissible,” etc. are vague in much the same
way as “bald,” “tall,” “thin,” “red,” “heap,” and other paradigms of vagueness.
Paradoxes aside, this phenomenon is, to my mind, mostly benign. Unlike
other forms of indeterminacy, such as obscurity, ambiguity, or inaccuracy,
vagueness does not greatly hinder moral inquiry. Rather, it can be taken to
reflect a wholesome state of undecidedness concerning certain moral issues,
indicating that we have left the door open for further debate and nego-
tiation. More generally, acknowledging vagueness in a certain domain is
often a sign that we recognize a diversity of human goals and interests, and
 Moral Vagueness 153

that we are prepared to tolerate variation of opinion and therefore a certain

amount of disagreement within that domain.
The connection between vagueness and disagreement is easily brought
out by reflecting on the phenomenology of borderline cases (one of the
most salient symptoms of vagueness). In a borderline case, the semantic
criteria governing a predicate’s application seem to pull in opposite direc-
tions: if a is a borderline instance of the predicate “P,” then there are strong
reasons in favour of classifying a as P, and strong reasons in favour of clas-
sifying a as not-P. The clash between such reasons can be experienced inter-
nally by one speaker, leading to perplexity or ambivalence,1 or externally by
two or more speakers, leading to disagreement.
This interesting link between vagueness and disagreement has not gone
unnoticed in the literature. Aristotle himself, in the passage quoted ear-
lier, explains the relative imprecision of ethics by appealing to the fact
that “noble and just actions [and all goods, more generally] exhibit much
variety and fluctuation” (Nicomachean Ethics 1094b, 15–16). And closer
to our time, a number of philosophers seeking to defend moral realism
against the so-called “argument from disagreement” (namely, the charge
that realists cannot account for the persistence and pervasiveness of seem-
ingly intractable moral dissensions) have pointed out that ethical objectivity
is compatible with the existence of pockets of indeterminacy in our moral
frameworks, which might in turn explain the occurrence of disagreement.2
The general strategy of this defensive move (which I shall call the “vagueness
defence”) is well sketched in the following quote from Russ Shafer-Landau,
who has also provided the most sustained examination of moral vagueness
in the literature so far:3
I suggest that the apparent soundness of the central noncognitivist argument [from
disagreement] may be dispelled if we abandon an assumption long associated with
objectivism, namely, that morality is entirely determinate. . . . If we allow for moral
indeterminacy . . . , then we have a promising explanation of disagreement for the
objectivist. In those situations where perfect unanimity seems a pipe dream, the
objectivist can attempt to show that an ineliminable element of moral indetermi-
nacy exists for the situation being debated. The fact that there is no uniquely correct
assessment awaiting discovery can appropriately explain why in some cases even
idealised agents would fail to converge on the identity of a single best moral evalu-
ation. (1994: 332, 336)

1
  Ambivalence, as I use it here, is thus a form of inner disagreement, where the speaker
is pulled in opposite directions. We often express this by saying things like “I’m con-
flicted,” “I’m torn,” or “I’m in two minds about it.”
2
  See Hurley (1992); Parfit (2011: ii. 559–62); Railton (1992); Shafer-Landau (1994);
Sosa (2001); Vasile (2010); Wolf (1992).
3
  See in particular Shafer-Landau (1995).
154 Cristian Constantinescu

Recently, Derek Parfit has also concurred:

Some questions may be indeterminate, in the sense that they have no answer. That is
sometimes true, for example, of the question “Is he bald?” If some man has no hair,
he is bald. If some man has a full head of hair, he is not bald. But we cannot plausibly
assume that, in all cases between these two extremes, any man must either be, or not
be, bald. In many cases, though it is not true that some man is bald, it is also not true
that this man is not bald. Similar claims might apply to normative questions. . . . If
some normative questions are indeterminate, having no answer, this would provide
another explanation of some normative disagreements. When people disagree about
whether some act is wrong, they may mistakenly assume that this act must either be,
or not be, wrong. If these people gave up this assumption, they might often cease to
disagree. (Parfit 2011: ii. 559–60, 562)
On the face of it, this looks like a plausible suggestion. There are, as we have
seen, strong conceptual links between vagueness and disagreement, so the
vagueness defence promises to defuse the argument from disagreement: the
realist can acknowledge the possibility of intractable moral disputes, but
insist that they often arise from the vagueness of our moral terms.
Yet, despite the initial plausibility of the vagueness defence, I shall argue
in what follows that not all moral realists can readily employ this move.
On close examination, vagueness turns out to be incompatible with tenets
that are key to certain forms of moral realism. My primary target here will
be the non-naturalist realism defended by philosophers like Shafer-Landau
and Parfit. I  aim to show that by invoking moral vagueness in response
to the argument from disagreement, non-naturalists get embroiled in a
dilemma, either horn of which forces them to give up some of their central
commitments.

1.  A Sketch of the Main Argument

For simplicity, I shall use the label “non-naturalism” for the brand of realism
championed by Shafer-Landau.4 I take the following to be key tenets of this
view, which will be relevant to my argument:

4
  I focus on Shafer-Landau because he most explicitly upholds both non-naturalism
and the vagueness defence. Parfit, as we have seen, is also a good example of a
non-naturalist who appeals to moral vagueness, but the extent to which he would assent
to all of the seven theses I discuss here remains slightly unclear (see n. 5). More widely,
non-naturalism seems to have enjoyed a vigorous revival recently, as many of the current
leading metaethicists have embraced its main tenets in one form or another: see e.g. Audi
(2004); Crisp (2006); Cuneo (2007); Dancy (2006); Enoch (2011); FitzPatrick (2008);
Huemer (2005); Wedgwood (2007).
 Moral Vagueness 155

Cognitivism: Moral sentences express beliefs and are therefore truth-apt.

Correspondence: Moral sentences are true when they correspond to
moral facts.
Atomism: Moral facts are instantiations of moral properties.
Objectivism: Moral facts and properties are mind-independent.
Supervenience: Moral facts and properties supervene upon natural
facts and properties.
Non-reductivism: Moral facts and properties are metaphysically sui
generis.
Rationalism: Moral facts are intrinsically reason-giving.
All these are claims that Shafer-Landau endorses explicitly, and defends
forcefully (see Shafer-Landau 2003). I shall have more to say about some
of them in the course of my argument. For now, let me just point out
that the view that emerges by putting together such claims is a robust form
of realism:  the non-naturalist I  have in mind will not adopt a minimal-
ist conception of moral properties and facts, and will not content himself
with characterizing them merely as “those things, whatever they are, which
are picked out by our true moral judgements”; nor will he embrace a qui-
etist stance, downplaying the ontological commitments of talk of moral
facts and properties (à la Scanlon 2003 or Dworkin 2011).5 Rather, the
earnest non-naturalist will insist that moral properties are qualities which

5
 Parfit sometimes seems to adopt this strategy too in his (2011). Though he seems quite
happy to countenance talk of non-natural, irreducibly normative properties at various
points throughout the book, in §112 he puts forward a view he terms “non-metaphysical
cognitivism,” according to which: “There are some claims that are irreducibly normative
in the reason-involving sense, and are in the strongest sense true. But these truths have
no ontological implications. For such claims to be true, these reason-involving proper-
ties need not exist either as natural properties in the spatio-temporal world, or in some
non-spatio-temporal part of reality” (2011:  ii. 486). Parfit’s “non-metaphysical” view
relies on distinguishing between an ontological reading of “exist” (on which the claim
that normative properties exist comes out false), and a non-ontological reading (on which
the claim is true). I’m not sure a view that appeals to a notion of “non-ontological exist-
ence” is properly characterized as “non-metaphysical,” so much as “super-metaphysical.”
Perhaps the best way to understand Parfit’s suggestion is as a kind of quietism about
metaphysical matters. But then it becomes difficult to understand how any substan-
tive debate between naturalism and non-naturalism, of the kind Parfit himself happily
engages in, can even be had. If, as Parfit appears to suggest at times, non-naturalism is
to be characterized simply as the claim that there are irreducibly normative truths and
concepts, then his view becomes indistinguishable from the non-analytical naturalism
defended by philosophers like Gibbard (2006), according to which, although there are
irreducibly normative claims, there are no irreducibly normative facts, because normative
concepts and natural concepts signify properties of the same kind. Since Parfit opposes
Gibbard’s non-analytical naturalism, he must think that not just moral claims, but moral
properties too, are irreducibly normative. To then pull a Meinongian stunt and add that
156 Cristian Constantinescu

individuals can instantiate (as opposed to being just sets of individuals, for
instance), and which, while “resulting from,”6 “being realised by,” or “being
constituted by”7 natural properties, are nevertheless neither identical nor
reducible to natural properties.
I am going to present my argument against this type of earnest
non-naturalism in the form of a dilemma. According to the non-naturalist,
there are sui generis, irreducible moral properties. But if such properties
exist, then given the vagueness of many of our moral predicates, either
Horn 1: vague moral predicates pick out vague moral properties,
or
Horn 2: vague moral predicates pick out sharp moral properties, and
vagueness arises from a different source.
I will argue that by taking Horn 1 of this dilemma, the non-naturalist ends
up with moral properties that are either mind-dependent or reducible, in
ways that are incompatible with either Objectivism, Supervenience, or
Non-reductivism above. On the other hand, by taking Horn 2 the non-
naturalist can avoid mind-dependence and reducibility, but only by counte-
nancing instead an ontology of perfectly sharp, strongly unknowable moral
properties, which I  shall claim ultimately undercuts his commitment to
either Rationalism or Supervenience. Either way, I will conclude, the non-
naturalist cannot accommodate moral vagueness without giving up some of
his key commitments.

2.  Five Reasons to Believe in Moral Vagueness

So far, I have simply assumed without argument that moral predicates are
vague. While I find this claim absolutely intuitive, some philosophers don’t.8
Although my argument targets non-naturalists who, like Shafer-Landau,

such properties exist only in a non-ontological sense appears to me to nullify the whole
debate. The moral of the story is that the earnest non-naturalist must take moral prop-
erties seriously, as of course Shafer-Landau and the other non-naturalists listed in n. 4
all do.
6
  See Dancy (1981). Dancy takes his notion of “resultance” from Ross (1930), but
significantly sharpens it. I should note that Dancy’s view at this early stage wasn’t decid-
edly non-naturalist, as he allowed the compatibility of resultance with a relation of
token-identity between normative and natural properties.
7
  See Shafer-Landau (2003: 72–9).
8
  Dworkin e.g. has long argued that every legal question must have a perfectly deter-
minate answer (Dworkin 1977), and has recently extended this view to the ethical
domain (Dworkin 2011: ch. 5).
 Moral Vagueness 157

believe in moral vagueness, I also hope to convince those who, like Dworkin,
find this idea doubtful. My aim in doing this, of course, is to weaken the
non-naturalist position more widely:  my dilemma, I  submit, applies not
just to non-naturalists who are already committed to moral vagueness, but
to all non-naturalists, since we are all committed to moral vagueness. I shall
therefore give five interrelated reasons for thinking that at least some (and
probably most) of our moral predicates are vague. Due to their interrelat-
edness, these reasons can be expressed as a single claim: moral predicates
display all the standard symptoms of vagueness: (a) imprecise gradability;
(b) boundarilessness; (c) borderlineness; (d) tolerance; (e) soriticality.
(a) Imprecise gradability. Like vague predicates from other domains of
discourse, moral predicates are imprecisely gradable. Their gradabil-
ity means that they admit of degrees: one person can be very honest,
another slightly cruel, a third not too courageous. Furthermore, one per-
son can be more just than another, and often by much. But, pace Socrates,
it seems impossible to tell precisely by how much. The reason for this
is that justice, honesty, cruelty, courage, and all the other moral values,
simply do not appear to be the kinds of properties that could come in
discrete, cardinally quantifiable units. I must stress that the focus here
is on imprecise gradability, rather than simply impreciseness or gradabil-
ity taken separately. Famously, impreciseness is insufficient for vague-
ness: “natural number between 1 and 100” is imprecise, but not vague.
And where a property is precisely gradable, vagueness can, but does not
necessarily, arise: both “tall” and “between 100 and 200 cm high” pick
out properties which are precisely gradable (heights), but the former is
vague while the latter is not. This raises an interesting point about the
properties denoted by vague predicates: in many cases, these properties
supervene upon properties which are precisely gradable, but are not
themselves precisely gradable. Consider the predicate “bald.” Baldness
supervenes on the property of having x hairs on one’s scalp, which is
precisely gradable. But baldness itself is not precisely gradable: even if
we know precisely how many more hairs Abe has on his scalp than Ben,
it doesn’t follow that we thereby can tell how much balder Ben is (if for
no other reason then at least because other factors, like the distribution
of hair on one’s scalp, contribute to our ascriptions of baldness, thereby
making gradability imprecise). Similarly for moral properties: honesty
supervenes on the number of lies one tells. But it doesn’t follow that if
we know how many lies Abe and Ben have told so far in their lives, we
can thereby know how much more honest the one is than the other.
(b) Boundarilessness. Imprecise gradability leads to imprecise boundaries.
Because honesty does not come in precise degrees, there cannot be a
158 Cristian Constantinescu

fixed degree of honesty constituting the threshold for whether someone

counts as honest or not. Thus, the predicate “honest” fails to separate
sharply between individuals to which it applies (its extension) and indi-
viduals to which it doesn’t (its anti-extension). Suppose, for simplic-
ity, that the number of lies told were the only criterion for classifying
people as honest (while gravity, context, intentions, and countless other
factors, didn’t matter). Imagine arranging people in a continuum, based
on how many lies they told: it seems implausible to suggest that there
will be a precise cut-off point (one white lie) separating the honest from
the dishonest.
(c) Borderlineness. From boundarilessness we get to borderlineness. When
a predicate lacks a precise boundary between its extension and its
anti-extension, it also admits of borderline instances:  individuals to
which the predicate neither determinately applies, nor determinately
doesn’t apply. Along the continuum we have just imagined, there will be
people in the fuzzy area between the extension and the anti-extension
of “honest.” These people are borderline instances of the predicate: not
clearly honest, but not clearly dishonest either.
(d) Tolerance. Like paradigmatically vague predicates, many moral predi-
cates are “tolerant” with respect to small changes in the properties on
which their application is based. Consider Abe, a determinately hon-
est man: Abe may have told a few lies in his life, but they were mostly
intended to save others from embarrassment, or forms of being polite.
When it comes to the important things, Abe is always truthful. Now
add one very small lie to Abe’s history. This seems insufficient to turn
Abe from a clearly honest into a clearly dishonest person. In this sense,
the predicate “honest” is tolerant with respect to small lies.
(e) Soriticality. Finally, tolerance famously gives rise to sorites arguments: if
Abe counts as honest having told just three small lies in his fairly long
lifetime so far, then so does Abe after telling four small lies; but if Abe
counts as honest after telling four small lies, then so does Abe after tell-
ing five lies, and so on until we get the absurd conclusion that Abe is
honest despite having told a million lies.
Readers with strong Kantian inclinations may find this example uncon-
vincing. They might think, for instance, that there is a precise cut-off point
between being honest and not being honest: namely, the very first lie. The
duty not to lie is a perfect duty in Kant’s system, and that gives us an abso-
lute threshold. Nevertheless, examples of a similar kind can be generated
with ease. If causing slight pain (say, a barely perceptible pinprick) when
giving someone an injection isn’t cruel, then causing ever-so-slightly more
 Moral Vagueness 159

intense pain isn’t cruel either; but applying tolerance a sufficient number of
times would yield the absurd conclusion that causing agonizing pain isn’t
cruel. If watching hard porn isn’t harmful to Ben today, as he turns 40,
then watching hard porn wasn’t harmful yesterday; apply tolerance over
and over again, and eventually you get the conclusion that watching hard
porn wouldn’t have been harmful to Ben when he was 7.9 And lest one
should think that only thick moral concepts can be vague, consider the case
of “wrong” in the following scenario (borrowed from Sorensen 1990): it’s
definitely wrong to draw 5,000 millilitres of blood from a person for blood
tests; if so, then it’s also wrong to draw 4,999 millilitres of blood; apply this
enough times, and you get the absurd conclusion that it’s wrong to draw 1
millilitre of blood. Finally, think of “ought” in this scenario: I ought to give
£10 to charity every month; if so, then I also ought to give £10.01; but if
I apply tolerance enough times, I get the absurd conclusion that I ought to
give all my money to charity.10

3.  The Dilemma Unpacked

I hope this is enough to convince readers that moral predicates can be

vague. I turn now to the task of expounding my dilemma. If the argument
is to stick, I must show that each horn is problematic for the non-naturalist.
It will help if we fix our attention on an example. It’s Saturday morning
and Abe, who is a psychotherapist, is at home relaxing with his family. It’s
been a while since he has had the chance to do so, for work has been quite
hectic lately. Abe takes genuine pleasure in being at home with his fam-
ily, and he’s also promised he wouldn’t sacrifice family time for work this
weekend. However, Abe’s phone rings: it’s his patient, Ben, who sounds very
distraught and desperately needs to talk to him, or else “he’ll do something
stupid.” Abe knows that Ben has been very unstable recently, and so decides
to go and meet him at his office. It seems clear, given Ben’s state and despite
Abe’s promise to his family, that in this case it’s permissible for Abe to go
talk with Ben for an hour. On the other hand, spending the entire day with
Ben would clearly not be permissible:  not only has Abe made a promise
to his family, but there are other therapists at hand who could take over,

9
  Of course, in a case like this, the law does set an arbitrary cut-off point at the age of
18. But no one seriously thinks that this removes the vagueness of “harmful”: if watching
hard porn one second before midnight on the day of his 18th birthday is harmful to Ben,
so is watching hard porn one second later (despite the latter act being legally permitted,
and the former not).
10
  I thank an anonymous reviewer for pressing me on the need to refine my examples
of moral vagueness.
160 Cristian Constantinescu

not to mention that Ben should not be encouraged to develop an emo-

tional dependency on Abe. Let us say, then, that it’s clearly impermissible
for Abe to spend eight hours with Ben. Now, somewhere between one hour
and eight hours along this continuum, there will be a fuzzy area contain-
ing actions (such as Abe’s spending 120 minutes with Ben, Abe’s spending
121 minutes, and so on) which are neither clearly permissible, nor clearly
impermissible.
Eventually, Abe spends two hours with Ben (call this action of his “A”).
Let us suppose that in this case A is a borderline instance of the predicate
“permissible.” According to the non-naturalist, the predicate “permissible”
denotes an objective, irreducibly normative property, permissibility.11
What should the non-naturalist say about the property of permissibility
itself, when confronted with a borderline instance of it? It seems to me there
are only two options:
(i) A is a borderline instance of “permissible” because permissibility
is vague;
(ii) A is a borderline instance of “permissible” despite permissibility
being sharp.
This generalizes to all moral properties. Vis-à-vis any moral statement
containing a vague moral predicate, the non-naturalist can either choose to
treat that predicate as denoting a vague moral property, or else insist that
all moral properties are sharp, and consequently seek to explain predicate
vagueness in some other way. Thus we get the two horns of my dilemma: the
non-naturalist can opt for Horn 1 by committing himself to an ontology
containing vague moral properties, or for Horn 2 by countenancing only
sharp moral properties. In virtue of his allegiance to robust moral prop-
erties, the non-naturalist cannot avoid this choice. But I  will argue that
each horn leads to serious problems. Such problems, I believe, can only be
avoided by banishing moral properties altogether.
Before I go on to examine each horn in detail, a clarification is required.
It may seem that by asking the non-naturalist to opt for either a vague or
a sharp moral ontology, I am in fact arbitrarily forcing a choice between
an ontic conception that would locate vagueness “out there,” in the moral
properties, and an epistemic conception, which would insist that moral
properties are perfectly sharp and regard vagueness instead as the result of
imperfect knowledge. It may thus seem odd that the problem is framed as a
dilemma, rather than as a trilemma. After all, why would the non-naturalist
not be entitled to a semantic account of moral vagueness? This challenge,

11
  From here on I adopt the convention of using small capitals to signify properties.
 Moral Vagueness 161

however, misconstrues the nature of my dilemma. The choice I am trying to

force is based on the ontological issue of whether moral properties should
be taken as vague or sharp. On that issue, supporters of onticism and epis-
temicism are indeed sharply divided, each being forced to take either one
or the other horn. The semanticist, however, can go either way. Like Russell
(1923) or Lewis (1986a), she may insist that the vagueness infecting our
language is entirely due to the general indeterminacy of our representa-
tions: the world itself is fully determinate. In terms of our dilemma, this
would amount to choosing Horn 2. Alternatively, like Shapiro (2006) and
Hyde (2008), the semanticist may instead deplore Russell’s insistence that
the world must be crisp as a mere reflex of what Whitehead called “the myth
of independent existence”: the idea that the world can be conceived of in
separation from our representations. Instead, this latter kind of semanticist
would regard ontic vagueness as a necessary accompaniment to semantic
vagueness.
The picture we should expect is then this:  on Horn 1 of the dilemma
we’ll find both advocates of onticism and some supporters of semanticism,
countenancing vague moral properties but offering competing accounts
thereof; on Horn 2, other semanticists will agree with epistemicists that
moral properties are sharp, but disagree about the explanation of predicate
vagueness. Analysing each horn carefully will require paying close attention
to these subtle differences.

3.1  Horn 1: Vague Moral Properties

The non-naturalist gets into Horn 1 of my dilemma by choosing to account
for moral vagueness in terms of vague moral properties.12 Various accounts
of vague properties have been proposed in the general literature on vague-
ness (Sainsbury 1989; Tye 1990; Rosen and Smith 2004; Shapiro 2006;
Hyde 2008; Schiffer 2010). As a first approximation, we may start by char-
acterizing a property as vague just in case it admits of borderline instances.
Thus, baldness is vague because there are people who are borderline bald,
and in our example above permissibility is vague because it is indeter-
minate whether Abe’s action A (spending 2 hours with Ben) instantiates
it. Some authors are in fact content with this minimal account of vague
properties (Sainsbury 1989; Tye 1990; Hyde 2008). But the initial char-
acterization doesn’t suffice. For one thing, a property may have a border-
line instance despite being perfectly sharp:  for instance, if its borderline

12
  This is indeed Shafer-Landau’s strategy: in his (1994) he argues for the “worldly”
indeterminacy of moral properties, and against semantic and epistemic accounts.
162 Cristian Constantinescu

instance is a vague object.13 Moreover, since A’s being a borderline instance

of permissibility means that it is indeterminate whether A instantiates
permissibility, the attempt to define vague properties in terms of border-
line instances seems guilty of a fallacy: it moves from “It is indeterminate
whether A instantiates permissibility” to “Of the property permissibility,
it is indeterminate whether it is instantiated by A.”14 And finally, if the
minimal characterization was all we had, we couldn’t even make sense of the
differences between ontic, semantic, and epistemic theories of vagueness.
For arguably, all of these theories can accept the truism that some prop-
erties have borderline instances, but not all of them would countenance
vague properties. Thus, if it makes sense to inquire whether there really are
vague properties, then the debate between supporters of onticism, semanti-
cism, and epistemicism must be a debate over vague properties in a more
robust sense.
As I have indicated, both onticists and certain kinds of semanticists can
accommodate talk of vague properties. Let’s take these views in turn, start-
ing with semanticism.

3.1.1  The Semantic View of Vague Properties

The first theory that springs to mind under the heading of “semanticism”
is surely supervaluationism. Yet, as we shall shortly see, supervaluationists
take the properties picked out by vague predicates to be perfectly sharp, and
for this reason we shall discuss their views when analysing Horn 2 of the
dilemma. For now, we must look elsewhere for a semantic account of vague
properties.
Besides supervaluationism, the other major brand of semanticism about
vagueness is contextualism. I shall use Shapiro’s (2006) contextualist account
of vagueness for illustration, as it explicitly countenances vague properties.
Shapiro takes properties in general to be individuated by the word-usage
practices of competent speakers with regard to the predicates that express
them. A useful model for understanding these linguistic practices is to think
of them as conventions established by conversational partners who keep
logs of their conversations (this is inspired by Lewis’s notion of a “conver-
sational score”). There are certain words whose application the conversa-
tionalists always agree on. For instance, in the case of the predicate “natural

13
  Thus, it may be indeterminate whether a particular cloud weighs 1 billion kilograms
not because of any vagueness in the property weighs 1,000,000,000 kg, but because the
cloud itself lacks sharp boundaries.
14
  As Williamson points out in a similar connection, this is no more valid than the
fallacious move from “It is contingent whether the number of planets is even” to “Of the
number of planets, it is contingent whether it is even” (see Williamson 2003: 701–2).
 Moral Vagueness 163

number,” competent speakers will be able to decide, of any object they are
presented with, whether or not that object is a natural number. A determi-
nate predicate individuates a determinate property:  thus, the property of
being a natural number is sharp. On the other hand, there are also cases
where conversational scores may remain silent: the conversationalists’ previ-
ous word-use practices have not established conventions that completely
determine any future applications of such predicates. The predicate “bald”
is a typical example. We can safely assume that the linguistic conventions
governing the application of this predicate have not settled criteria of cor-
rect application for every conceivable configuration of hair on a human
being’s scalp. Therefore, there will be unclear cases, over which competent
speakers might disagree. There will also be cases about which they will feel
ambivalent: one and the same speaker may classify me as “bald” in a context
where the standard of hirsuteness is set in relation to Jimi Hendrix, and as
“not bald” in a context where I’m standing next to Billy Corgan. All of this
is, of course, very familiar. The point is that, on this account, properties are
individuated by the semantic criteria governing the use of predicates denot-
ing them, which are in turn fixed by conventions made by speakers.15
This rough characterization should suffice to suggest that the contex-
tualist approach to vague properties is not an appealing option for the
non-naturalist who wants to account for moral vagueness by appealing to
vague moral properties. On Shapiro’s account, vague properties turn out
to be judgment-dependent in a deep sense:  not only are they responsive to
certain linguistic/psychological facts about the competent speakers—they
are actually constituted by such facts. The analogue of this in the moral
realm would be a form of constructivism, grounding moral properties in the
practices of rational agents negotiating rules for mutual behaviour. But on
this view, speakers confronted with borderline instances of moral predicates
cannot be assumed to track an independent realm of properties in their
judgments: the direction of fit is Euthyphronic (from speakers to proper-
ties) rather than Socratic (from properties to speakers). Adopting such an
account of vague properties would therefore seem to be incompatible with
the thesis of Objectivism.
But perhaps we’ve gone too fast. There is an obvious rejoinder to the
preceding line of argument: while it may be true that vague properties are in

15
 This approach to vague properties may be supplemented with a psychological
account (in the manner of Schiffer 2000). After all, as Shapiro himself acknowledges
(2006:  24ff.), the picture of communicators keeping conversational scores appears to
leave one important question unanswered: what makes it the case that individual ambiva-
lence or collective disagreement are sometimes appropriate attitudes for the conversation-
alists to display? Schiffer’s account may offer the key, by explaining vagueness in terms
of belief-degrees.
164 Cristian Constantinescu

one sense judgment-dependent, this may only be the case within their bor-
derline areas. Determinate instances, on the other hand, could be regarded
as mind-independent. This is actually Shapiro’s view:

With vague predicates, judgment-dependent matters sometimes figure in how

the extension is fixed. For determinate cases, the judgment-dependent matters
in question are trumped by other factors that fix the extension. In the border-
line region, the judgment-independent features give out, and, consequently, the
judgment-dependent features of the process dominate. The predicate becomes
Euthyphronic. As I  see things, every vague predicate has (or can have) such a
Euthyphronic region. But this does not undermine the overall objectivity of the
predicate, or the discourse in which it figures. If it did, there would be precious little
objectivity anywhere. (2006: 209)

If Shapiro’s remarks here are correct, perhaps the non-naturalist can embrace
the preceding account of vague properties after all.
To see whether this will work, let us first translate Shapiro’s talk of
vague predicates in this passage into talk of vague properties. Take “bald,”
which Shapiro uses as an example of a predicate with both a Socratic and
a Euthyphronic region. Presumably this means that the property baldness
denoted by this predicate also has such “regions”: a mind-independent core,
determined entirely by objective factors, and a mind-dependent penum-
bra, determined at least in part by speakers’ conventions. The first image
that comes to mind is that of a fuzzy set of bald individuals, with differ-
ent degrees of membership:  some individuals are determinately bald and
therefore full members, while others are indeterminately bald and therefore
members to lesser degrees. For the former kind of individuals, objective
factors fix their membership status, whereas for the latter it is up to speak-
ers to decide in each case whether it is appropriate or not to count them as
members of the set. Clearly, this isn’t an understanding of vague properties
that the non-naturalist can readily embrace and apply to moral properties.
For the non-naturalist, the vague property of permissibility, for instance, is
more than just a set of actions, for there is nothing metaphysically sui generis
or intrinsically reason-giving about sets.
An alternative is to think of the different “regions” of a vague property
more literally, i.e. mereologically. Thus, a vague property could be more like
a cloud than a set. A cloud is made up of a core of particles definitely within it,
and a looser, marginal region of particles of which it’s indeterminate whether
they are parts of the cloud. At first blush, this model seems more compatible
with Shafer-Landau’s conception of moral properties as being “constituted
by,” though not identical to, natural properties (Shafer-Landau 2003: 72–9).
permissibility, for instance, could be a non-natural type-property made up
of various token-properties (permissibility-instantiations, or tropes), each
 Moral Vagueness 165

constituted or realized entirely by clusters of natural tropes.16 The difference

between tropes belonging to the property’s core and tropes in the property’s
penumbra could then be understood in terms of a further type of natural
fact belonging to the constitution of the latter, but not the former: namely,
natural facts about speakers’ linguistic conventions/attitudes. Due to being
constituted in part by such natural facts about speakers, tropes in the
penumbra of permissibility could be considered mind-dependent in the
Euthyphronic sense. However, the tropes in the property’s determinate core
would remain perfectly objective.
But there is a complication for this view. Ascriptions of vague predi-
cates are notoriously context-relative:  as noted earlier, I  can count as
determinately not-bald when compared to Billy Corgan, but indetermi-
nately bald when compared to Jimi Hendrix. The same applies, of course,
to vague moral predicates:  I  may count as indeterminately generous in
many contexts, but I’m definitely not generous when compared to Peter
Singer. Similarly, one and the same action can be definitely permissible
when compared to one set of alternatives, and indeterminately permissible
when compared to another set. On the picture we are now contemplat-
ing, this means that one and the same permissibility-trope, call it “Pi,” is
part of the property’s core from one perspective, and part of the property’s
penumbra from another. This in turn implies that Pi is constituted in one
context by a set of natural tropes, call it “Ni,” and in another context by
Ni ∪ {s1, . . . , sn}, where “s1,” . . . “sn” denote natural facts about speakers’
conventions and/or attitudes. However, this violates a fundamental con-
straint on constitution. As Ridge (2007: 342) points out, if constitution
is to do the job Shafer-Landau wants it to—namely, that of accounting
for Supervenience—then two conditions must obtain:  (i)  if a cluster of
natural tropes Ni necessarily constitutes a moral trope Pi, then whenever
Ni is present, Pi is present too; and (ii) if Ni constitutes Pi, then Ni neces-
sarily constitutes Pi. But it’s quite clear that these two conditions cannot
be met for tropes of vague properties: due to the contextual variation noted
above, one and the same moral trope will in some contexts belong to the
core, and in other contexts to the penumbra of a moral property. But then
it follows, on the present understanding of penumbral instances, that one
and the same moral trope can be instantiated by different natural tropes in
different contexts (since facts about speakers will sometimes feature, and
sometimes not, in the trope’s constitution). And this violates (i)  above,
thereby putting the claim of Supervenience in jeopardy.

16
  This is in line with Ridge’s reconstruction of Shafer-Landau’s conception of consti-
tution (Ridge 2007: 340ff.)
166 Cristian Constantinescu

No such difficulties would arise, of course, for a moral naturalist. The

naturalist could embrace Shapiro’s conception of vague properties, while
identifying vague moral properties with fuzzy sets of individuals, in the
manner contemplated above. Non-naturalism, on the other hand, appears
incompatible with a semantic view of vague properties.

3.1.2  The Ontic View of Vague Properties

The alternative approach to vague properties aims to give a purely structural
account in terms of the formal characteristics distinguishing vague from
sharp properties,17 rather than formulate individuating conditions in terms
of constitutive facts about speakers’ linguistic conventions or mental states.
This starts from the observation, already adumbrated (see n.  13), that
there are two ways in which a property can have borderline instances: either
when the property itself is vague, or when its instances are vague objects.
Consider now a sharp property: if such a property has borderline instances,
that can only happen because the objects in those instances are themselves
vague. To take our earlier example, since weighs 1,000,000,000  kg is a
sharp property, only a vague object (e.g. a cloud) could be a borderline
instance of it. In other words, sharp properties are “the properties that make
vague objects out of their vague instances” (Rosen and Smith 2004: 187).
This insight offers us a key: we could start by zeroing in on sharp properties,
and then proceed by defining vague properties negatively (as those proper-
ties which are not sharp).
To illustrate further, take the vague property tallness, and suppose Bill
is a borderline instance of it. Must there be any vagueness about Bill him-
self? Not necessarily.18 It is possible for Bill to be a perfectly determinate
body, yet a borderline instance of “tall.” Compare that to the predicate “pre-
cisely 186 cm tall.” Measurement imprecision aside, it is impossible for Bill
to be a borderline instance of this precise predicate without being vague
himself (perhaps, for instance, there’s a hair loosely attached to his head in
such a way that it is indeterminate whether or not it is part of his body, and
counting that hair may alter our measurement). Rosen and Smith’s own
example involves a comparison between the vague colour-predicate “blue”
and the precise predicate “blue-17,” which denotes a particular point-sized
region in the Munsell Colour Solid. Again, an object can be a borderline
instance of the latter predicate only by being vague itself.

17
  See Rosen and Smith (2004). The following paragraphs are based (at times only
loosely) on their approach, adding qualifications drawn from Sanford, who offered simi-
lar suggestions much earlier (Sanford 1966, 2013).
18
  Though of course there may well be, in the sense of there being at least one particle
x such that x is neither determinately inside Bill, nor determinately not inside Bill.
 Moral Vagueness 167

Rosen and Smith distinguish between vague and sharp properties by gen-
eralizing this idea. On their account, vagueness turns essentially on issues
of specificity. Properties are sharp if they are maximally specific—much
like determinate point-sized regions in the colour solid, or point-sized
units along a continuum of heights. For this reason, Rosen and Smith also
refer to sharp properties as “point-properties.” Vague properties, on the
other hand, lack specificity. Rather, they cover ranges of specific proper-
ties—much like “blue,” which covers a range of different shades, or “tall,”
which covers a range of different heights. For this reason, we may call them
“range-properties.”19
The distinction looks neat, but it does require an important qualifica-
tion. Rosen and Smith appear to overlook the fact that there are properties
which are sharp, but not maximally specific (i.e. not point-properties): e.g.
the properties denoted by “taller than 186 cm,” or “taller than 186 cm but
not taller than 189  cm.” These are imprecise, but sharp (i.e. not vague).
That some range-properties turn out to be sharp may seem to throw
some doubt on the usefulness of Rosen and Smith’s distinction. But the
issue can easily be fixed, by noting a common characteristic of such sharp
range-properties: they cover precisely bound ranges. This enables us to revise
the above characterization of vague properties, so that it now reads: vague
properties are properties that cover indeterminate or unbound ranges of
point properties.
There are good reasons to think that moral properties are vague in
the sense just specified. Like colour terms, moral predicates are general
enough to apply not just to singular instances, but to entire ranges of cases.
Moreover, these ranges are never precisely bound: there are no minimal or
maximal values of courage, kindness, permissibility, etc., as shown in §2
above. And importantly, from a metaphysical point of view this structural
account of vague properties seems less threatening than the semantic view
examined in §3.1.1, since it doesn’t rely on facts about speakers’ conven-
tions which could introduce an element of subjectivity. It might seem as if
this ontic view of vague properties could be much more promising for the
non-naturalist seeking an account of moral vagueness.
Under closer scrutiny, however, things appear more problematic. On
the present account, the lack of specificity displayed by vague properties
means that they are somehow composite, or structured:  as we have seen,
they cover fuzzy ranges, and as such are comprised of maximally specific

19
 This could also be paraphrased in terms of a distinction between determin­
ables and determinates:  range-properties are determinates and determinables, whereas
point-properties are perfect determinates (i.e. determinates but not determinables). For
more on this, see Sanford (2013).
168 Cristian Constantinescu

point-properties. This encourages the view that they are somehow conceptu-
ally derivative: i.e. capable of being broken down and understood in terms
of sharp ones. For notice that corresponding to any vague property P there
will be a set of properties {Q1, Q2, . . .} such that each Qi belonging to that
set will be (i) perfectly sharp, and (ii) necessarily sufficient for P. But then
it follows that the disjunction of all these sharp properties, Q1 V Q2 V . . . ,
will be necessarily coextensive with P. This seems intuitive: when I say that a
is blue, I locate a within a fuzzy region on the colour spectrum, while leav-
ing it open which particular shade of blue a may be; thus, my claim can be
equated to “a is B-1 or B-2 or . . . B-17 or . . . ”20
The upshot is that vague properties are essentially disjunctive: more spe-
cifically, they can always be resolved into disjunctions of sharp properties.
This is apt to spell trouble for the moral naturalist contemplating an ontol-
ogy of irreducible moral properties. For one thing, the very notion of a “dis-
junctive property” has seemed incoherent to many philosophers (e.g. Lewis
1986b; Armstrong 1978). Here’s how Armstrong puts the issue:
Disjunctive properties offend against the principle that a genuine property is identi-
cal in its different particulars. Suppose a has a property P but lacks Q, while b has
Q but lacks P. It seems laughable to conclude from these premisses that a and b are
identical in some respect. Yet both have the “property”, P or Q. (1978: 20)
But perhaps not all disjunctive properties are as gerrymandered as Armstrong
supposes they must be. There is a difference, for instance, between the fol-
lowing (putative) properties: pale yellow or bright yellow; pale yellow
or dark green; yellow or angry (Sanford 1970, 2013). While the last
of these definitely falls prey to Armstrong’s criticism, it may seem doubtful
whether the second does, and quite certain that the first doesn’t. The issue,
it may be thought, turns on resemblance. At least the first of these proper-
ties can be regarded as a real, genuinely disjunctive property, because there
are deep similarities between its disjuncts: if a is pale yellow and b is bright
yellow, then a and b are indeed “identical in some respect,” and therefore
can be properly said to satisfy the predicate “pale yellow or bright yellow.”
This may seem to provide a satisfactory response to Armstrong’s attack.
After all, most of the vague predicates in our language (“red,” “tall,” “bald,”
but also “generous,” “honest,” “just,” “courageous,” etc.) appear to “carve
nature at its joints” rather than just being artificially gerrymandered dis-
junctions (in other words, they are more like “pale yellow or bright yel-
low” than like “yellow or angry”). “Red,” for instance, is equivalent to

20
  Not in the sense that I must mean or intend that disjunction when I ascribe blue-
ness to a, of course. The claim concerns the extensions of vague properties, not their
intensions.
 Moral Vagueness 169

an indeterminate disjunction R-1 V R-2 V . . . , the disjuncts of which are

related by deep intrinsic resemblances.
Still, matters get even more complicated. For it turns out that different
red things satisfy the predicate “red” in virtue of different properties: red
light in virtue of its frequency range, red paint in virtue of the chemical
composition of its surface, red-hot objects in virtue of their temperature,
etc. (see Mellor 2012: 397). So even a seemingly natural property like red-
ness, with much more going for it than yellow or angry, may in fact turn
out to be almost as disjoint as the latter. If the real existence of properties
turns on similarities between disjuncts, then redness may not be a real
property at all.
In this respect, what goes for colour-properties also goes for moral prop-
erties, which, by virtue of being vague, turn out to be range-properties and
therefore surprisingly disjunctive. Could they, unlike colour-properties, be
shown to be real in virtue of some deep, intrinsic resemblances between the
fine-grained properties constituting their disjuncts? Hardly. Consider once
more the property of impermissibility. It is obvious that actions can satisfy
the predicate “impermissible” by virtue of very different underlying proper-
ties: some because they are acts of promise-breaking, others because they
cause physical harm, still others because they are deceitful, etc., etc. Even
thick moral predicates, which are considerably more specific, still manifest
the same lack of unity: people can be courageous by showing temerity on
the battlefield, by coping well through personal tragedies, by braving loss,
by making hard decisions, by breaking with old habits, etc.; or they can be
kind by donating money, by giving up their time, by offering a comfort-
ing word, etc. The diversity of properties falling under one and the same
moral predicate, be it thick or thin, seems almost boundless. If sameness
of property is indeed required, then the revelation that moral properties
are multiply realizable by virtue of being vague should cause the realist to
question their existence.
But at this stage, the realist has an easy rejoinder at hand: despite their
seemingly gerrymandered nature, moral properties could perhaps be
acknowledged as real based on a different ontological criterion: not their
structure, but their causal efficaciousness. In other words, we should com-
mit ourselves to moral properties if they can pull their weight in explana-
tions and predictions of moral judgments and conduct. The trouble with
this move, of course, is that it’s only available to naturalists like Jackson
(1998) or the Cornell-realists, who have long argued that moral proper-
ties can be understood as clusters of more basic, natural properties, and
that we should be realists about them so long as they pull their weight in
scientific explanations. However, for the non-naturalist this line is totally
unpromising.
170 Cristian Constantinescu

In addition, there is also the issue of reduction, which seems to lead

towards a similar conclusion. Given that, on the ontic account, vague moral
properties turn out to be co-extensive with disjunctions of sharp properties,
could this be used to reduce the former to the latter? The answer depends on
what one means by “reduction,” and philosophers have famously disagreed
here. If reduction is taken to involve the process of formulating a priori
identity-statements that would establish intensional equivalences between
the two types of properties, then it should be quite clear that vague prop-
erties cannot be reduced to sharp ones. The meaning of “blue” isn’t cap-
tured by the disjunction “B-1 or B-2 or . . . B-17 or . . .” and it took a piece
of a posteriori discovery for us to know what kinds of physical properties
underwrite our ascriptions of colours. On the other hand, philosophers like
Jackson (1998) or Gibbard (2006), who defend a weaker, non-analytical
form of naturalism, could perfectly well argue that, if vague properties
can be broken down into disjunctions of sharp properties, then the for-
mer are thereby shown to reduce to the latter by virtue of the a posteriori
extensional equivalences established. This throws doubt on the claim of
Non-reductivism, which is central to non-naturalism.
Such considerations therefore appear to establish that the ontic view
of vague properties, while friendly towards various forms of naturalism,
remains fundamentally incompatible with non-naturalism about moral
properties. There is, however, one final move left for the non-naturalist
here—and a rather effective one, too.21 On the ontic account of vagueness
under consideration, vague properties turn out to be disjunctions of more
basic, perfectly sharp properties. But what reason do we have for assuming
that the fine-grained properties constituting the disjuncts of such disjunc-
tions are natural rather than non-natural? No argument has been given so
far to support this assumption. It is true, of course, that in all the non-moral
examples considered (colour-properties, predicates like “tall,” “bald,” etc.)
the underlying sharp properties are perfectly natural (wavelengths, heights,
numbers and arrangements of hair on people’s heads, etc.). But that doesn’t
necessarily mean that the same must obtain in the moral cases. The revela-
tion that the structure of vague moral properties is disjunctive is still com-
patible with the claim that the fine-grained, sharp disjuncts are themselves
irreducibly moral, non-natural properties.
The suggestion that moral ontology may bottom out with a set of per-
fectly sharp, non-natural properties effectively moves our discussion on to
Horn 2 of the dilemma, which starts precisely by assuming that moral prop-
erties are sharp. I therefore turn now to exploring this option.

21
  I owe thanks to David Copp for suggesting this move to me.
 Moral Vagueness 171

3.2  Horn 2: Sharp Moral Properties

The non-naturalist gets into Horn 2 of my dilemma by opting for an
account of moral vagueness built around the assumption that moral proper-
ties themselves are always sharp. But there are different accounts to be given
of sharp properties, which will vary according to which theory of vague-
ness they are based upon. We have already encountered one such account,
derived from the ontic view of vagueness examined in the previous section.
I shall start by exploring this account in more detail. Another view of sharp
properties, and by far the most popular, is based on a semantic account of
predicate vagueness along the lines of supervaluationism: I explore this in
§3.2.2. Finally, the third route to sharp properties is one starting from an
epistemic account of moral vagueness, the details of which I  examine in
§3.2.3.

3.2.1  A Pointilist Moral Ontology?

Twice during our consideration of Horn 1 above, an idea has emerged
according to which the basic constituents of moral reality could be a set of
fine-grained, perfectly sharp, irreducibly moral properties. Thus, in §3.1.1
we examined Shafer-Landau’s claims about the relation between natural and
non-natural properties. According to Shafer-Landau, instantiations of moral
properties are constituted by instantiations of natural properties, but the
moral properties themselves are not identical to the natural properties that
go into their constitution. Following Ridge, I interpreted Shafer-Landau’s
claims about property-instantiations as claims about tropes. Our second
encounter with sharp moral properties came right at the very end of §3.1.2,
when we considered an account of vague properties as disjunctive proper-
ties. We then asked, on behalf of the non-naturalist: What if the fine-grained
disjuncts of such disjunctions were sharp, non-natural properties?
One important insight we have gained from our discussion of Rosen
and Smith’s account of sharp properties in the previous section is this: for
a property to be perfectly sharp, it must be either (i) a maximally specific,
point-property, or else (ii) a disjunctive property covering a precisely bound
range of such point-properties. Either way, it would seem that, on this
picture, the ultimate building blocks of moral reality would be putatively
non-natural, maximally specific moral properties. It is, again, tempting
to think of these as tropes: concrete, unrepeatable particulars, rather than
abstract, universal properties. But these pointilist moral tropes would seem
to be strange beasts indeed. It is difficult to even characterize them roughly,
for our moral language obviously does not contain terms for qualities so
finely grained. Perhaps some approximations would be: “courageous exactly
172 Cristian Constantinescu

like that person at ti” or “morally wrong precisely like my action at tj.”22
As a general strategy, we might try using Kaplan’s “dthat” operator, to yield
something along the lines of “dthat [token of M]” where M would be a term
for the kind of property to which the particular trope belongs (“courage,”
“kindness,” etc.). But the sense in which such tropes can be said to belong
to general kinds, as well as the criteria for grouping them together, remains
unclear. One obvious option, as we have seen, is to appeal to a relationship
of resemblance or similarity between tropes, based on which we could classify
them as belonging to the same kind. The non-naturalist might then argue
that, along with moral tropes—to which we must have access via direct
acquaintance or some form of intuition23—our moral experience also deliv-
ers basic facts about resemblance. So, for instance, when witnessing Carrie’s
act of saving a baby from drowning at great risk to herself, my experience
delivers not just the particular moral characteristics of the situation (the
moral tropes), but also an objective resemblance to other acts I witnessed in
the past (e.g. Dana defending a friend against aggressors), based on which
I classify all such acts as instances of “courage.”
But the plausibility of this proposal dissolves under scrutiny. As we
saw, resemblances between moral tropes belonging to the same kind (e.g.
courage-tropes) could not be resemblances between the natural tropes
constituting them. For, just like “red” in Mellor’s example, “courageous,”
“cruel,” “kind,” “right,” “good,” etc. can be instantiated by different peo-
ple or actions in virtue of a myriad of different natural properties. But if
moral-trope-resemblance isn’t natural-trope-resemblance, what can it be?
The only other option is to appeal to a primitive notion of qualitative resem-
blance between the moral tropes themselves, over and above any differences
and similarities between the natural tropes constituting them. But notice that
resemblance is never a simple, all-or-nothing affair:  a could resemble b in
some respects but not in others. When two things are similar, it seems natural
to assume that they are so because they have certain qualities in common. But
once we introduce this idea, the non-realist’s appeal to primitive tropes starts
to unravel: how are we to even begin to spell out the putatively non-natural
qualities grounding the resemblance between tropes like those exemplified by
Carrie’s saving a baby and Dana’s defending a friend, other than by saying that
they are both instances of courage? In fact, the very attempt to individuate
them qua moral tropes seems to require an appeal to the general kind they
belong to:  “courageous exactly like Carrie’s action at tj,” “courageous exactly
like Dana’s action at ti.” But in that case, it turns out after all that the general

  I thank David Copp for this suggestion.

22

  I assume that since tropes are perfectly simple and unrepeatable, they could not be
23

known via inference, either deductive or inductive.

 Moral Vagueness 173

property courage is primary to, and more basic than, courage-tropes.24

Instead of a pointilist ontology of maximally specific tropes, we’re right back
where we started: with an ontology of general properties applying across a
wide range of cases, and which display all the signs of vagueness. We are back,
that is, on Horn 1 of the dilemma. It seems that the non-naturalist must look
elsewhere for a satisfactory account of sharp properties.

3.2.2  Supervaluationism and Sharp Moral Properties

Thus we arrive at supervaluationism—arguably, the most popular account
of vagueness.25 Take a standard vague predicate like “bald,” and let Harry
be a borderline instance of that predicate. According to supervaluation-
ism, Harry is a borderline instance of the predicate “bald” because there is
no single property that “bald” denotes: rather, there are indefinitely many
sharp properties, corresponding to different precisifications of the predicate.
Vagueness, on this account, is a matter of semantic indecision: it is indeter-
minate which of these determinate properties the predicate picks out. Thus,
the supervaluationist’s world contains, at bottom, only sharp properties.26
It is easy to see, however, that supervaluationism does not hold much
promise for the non-naturalist, for a number of reasons. First, there is no
room in classic accounts of supervaluationism, such as those developed by
Fine (1975) and Keefe (2000), for any reference to properties themselves as
distinct from the extensions of predicates. That is to say, the supervaluation-
ist semantics treats properties purely extensionally, as the sets of individuals
to which predicates apply. Clearly, this is not the kind of view of properties
that will sit well with the non-naturalist’s more robust ambitions (and in
particular, with the tenet of Non-reductivism).

24
 There is, of course, far more to be said here, on both sides of the argument.
Dismissing trope-theory isn’t something one can do in a quick move like this. But I hope
I’ve done enough to at least raise some doubts about the use of this theory in tandem
with moral non-naturalism to yield what I  have called a “pointilist moral ontology.”
The non-naturalist may have more to say about the relations between moral tropes and
properties, and could perhaps use recent work on determinates/determinables, such as
Gillett and Rives (2005), to articulate a more robust view. Until such work is completed,
however, the doubts I raise are, I think, justified.
25
  Thanks to Antti Kauppinen, Tom Dougherty, and an anonymous referee for saving
me from a few significant errors I had made in my discussion of supervaluationism in
an earlier draft.
26
  For a formal explanation of why properties cannot be vague on a supervaluationist
account, see Williamson (2003: §5). Keefe (2000: 160) disagrees, but not because she
thinks anything significant hangs on whether supervaluationists take properties to be
vague or sharp. Quite to the contrary, Keefe believes that, on the extensional view of
properties embraced by supervaluationists, it makes no difference whether one says that a
predicate “P” indeterminately picks out a sharp property, or rather that “P” determinately
picks out a vague property.
174 Cristian Constantinescu

But suppose this worry could be alleviated, perhaps by grafting some

intensional view of properties onto the supervaluationist semantics. Would
this make supervaluationism available to the moral non-naturalist? The
answer, I  think, is still “No.” Assuming that supervaluationism could be
wedded to a robust conception of moral properties, the result would be a
luxuriant metaphysics, countenancing a plurality of ever-so-slightly differ-
ent moral properties, all perfectly sharp and non-natural. Returning to my
earlier example, suppose Abe’s action A (spending two hours with his patient
Ben) is indeterminately permissible. On the proposed picture, we would
have to explain this by saying that there are many sharp, ever-so-slightly
different non-natural properties of permissibility, some of which include
and some of which don’t include A as an instance, and it is indeterminate
which of those properties the predicate “permissible” picks out. Now, the
challenge is for the non-naturalist to explain the odd nature of the moral
reality depicted here. This seems difficult.
The non-cognitivist, by contrast, can appeal to a familiar strategy. She
will first point out that, on the standard supervaluationist account (without
robust properties added), the source of vagueness is semantic indecision.
But of course, as Keefe notes, this isn’t merely “a lazy indecision which
is waiting to be resolved” (2000:  155–6). No one thinks we can resolve
instances of vagueness, and dissolve all disagreements, simply by fiat. It’s not
just that some cases (among which the moral ones are paramount) will resist
all kinds of arbitrary stipulation. It’s rather that the very attempt to impose
such stipulation is inappropriate and often betrays conceptual confusion.
The question, then, is: Why do we find it so deeply inappropriate to resolve
borderline cases by fiat? Here, the non-cognitivist can appeal to her favour-
ite explanation, invoking attitudes: as I noted in the introduction, vague-
ness is intimately linked with ambivalence. There are many situations in our
practical lives when, faced with conflicts of tastes, interests, duties, or even
cultures, or perhaps simply with very difficult decisions, ambivalence seems
like the right kind of attitude, at least for a while. We do, of course, praise
people for showing resolve and the ability to react in the face of adversity.
But we also criticize those who are insensitive to the force of the conflict,
too opinionated, simple-minded, or plain reckless. We do so because we can
hear the force of all the conflicting considerations, and recognize that they
are all legitimate. We thus have the rudiments of a psychological account
of ambivalence, which could offer someone like Blackburn or Gibbard
the starting point in a quasi-realist explanation of moral vagueness.27 On

27
  The story, of course, will have to be much more elaborate. But the main ingredients
exist: Blackburn does appeal to this kind of ambivalence or open-mindedness (leaving
the door open for more than one set of values, or ways of life) in his (1984: 201), though
not directly in relation to vagueness.
 Moral Vagueness 175

this quasi-realist account, the higgledy-piggledy nature of moral properties

would be taken as a mere reflection of the fact that those properties are
projections of our attitudes onto the world.
The non-naturalist, by contrast, has precious little to say about the
unruly structure of moral properties like those characterized above. Unlike
the quasi-realist, who takes the Euthyphronic direction and explains moral
properties in terms of attitudes, the non-naturalist will have to adopt a top-
down, Socratic account. But it is unclear to me what the elements of such
an account might be.

3.2.3  Epistemicism and Sharp Moral Properties

Consider, finally, epistemicism about vagueness.28 Like supervaluationism,
this view represents the world as completely precise: there are only sharp
properties in the epistemicist’s ontology. But instead of viewing vagueness
as semantic indecision, this approach characterizes it rather as an epistemic
phenomenon: a matter of ignorance concerning the sharp boundaries of our
concepts. On this picture, our moral concepts are perfectly sharp: they draw
precise boundaries between objects to which they apply (their extensions)
and objects to which they don’t (their anti-extensions), but it is impos­
sible to know where these boundaries lie due to certain constraints on what
counts as knowledge (namely, the “margin-for-error principle”).
Take our earlier example of Abe and Ben, where we supposed that Abe’s
action A (spending two hours with Ben) was indeterminately permis­sible,
i.e. a borderline instance of permissibility. What, according to the epis-
temicist, explains our classifying certain instances of permissibility as
“indeterminate” or “borderline,” if permissibility itself, as a property, is
perfectly sharp? The answer is:  ignorance. Although there is a perfectly
sharp dividing line between actions that are permissible and actions that
are not, we cannot know where that line falls. A nanosecond is enough to
make the difference between its being permissible for Abe to spend time
with Ben and its not being permissible. Now suppose that S and S* are two
stages or time-slices of the universe separated only by one nanosecond, and
p is the statement that Abe’s action A is permissible. Suppose, moreover,
that the precise boundary falls between S and S*, such that p is true at S but

28
  Shafer-Landau explicitly rejects epistemicism as a plausible explanation of moral
vagueness in his (1994), and opts instead for an ontic account. Nevertheless, the view
may look independently plausible to other moral realists. Dworkin (2011), for instance,
explains apparent instances of moral indeterminacy in terms of ignorance or uncertainty,
and insists that there’s always a fact of the matter about what we should do. It seems
natural to interpret such claims along the lines of epistemicism. And Tim Williamson,
who has been one of the most prominent champions of epistemicism, has indicated (in
personal correspondence) that he takes a robustly realist stance on morality. It is therefore
worth considering the view’s metaethical implications in detail.
176 Cristian Constantinescu

false at S*. Could we ever be in a position to know p at S? No, because our

epistemic powers of discrimination are limited: if we are in S, and S* differs
from S only marginally, then for all we know we are in S*. It follows that if
we were entitled to claim knowledge of p at S, then we would have the same
entitlement at S*. But obviously, at S* our claim to know p would be false. If
our belief in p is to count as knowledge, we must have a sufficiently reliable
basis to discriminate between p’s being true and p’s being false. Therefore,
we are not in a position to know p at S.  More generally, in areas where
our capacity to discriminate is limited, knowledge requires a margin for
error: cases in which one is in a position to know p must not be too close to
cases in which p is false. As Williamson puts it, “[w]‌hen knowing p requires
a margin for error, the cases in which p is known are separated from the
cases in which p is false by a buffer zone, a protective belt of cases in which
p is true but unknown” (Williamson 2000: 18). For the epistemicist, what
we call the “grey,” “indeterminate,” or “borderline” area is nothing but this
buffer zone of ignorance.
On this view, then, there are determinate moral facts which are strongly
unknowable. To many, this is an unpalatable consequence. That a certain
moral theory yields unknowable normative truths or obligations is often taken
as grounds for rejecting that theory.29 But non-naturalists, like all realists in
general, have no problem countenancing unknowable facts and truths: in fact,
it’s even customary to define realism about a given domain as the claim that
truth in that domain is epistemically unconstrained (i.e. can outstrip knowl-
edge). If this holds in other domains of discourse, why would the notion of
unknowable truths be any more problematic in the moral domain?30
Thus, more needs to be said for the unknowable-truths objection to stick.
Yet, it turns out to be surprisingly tricky to put one’s finger on the problem.
One option would be to appeal to the action-guidingness of morals. The main
purpose of morality is to provide a guide for action. But how could unknow-
able moral facts ever be expected to be action-guiding? We can, of course,
act in accordance with the reasons they yield—accidentally as it were, by
guessing at them correctly. But doing so will be essentially a matter of luck.
If morality is to be genuinely action-guiding, then we should be able to act
morally not just by accident, but by following moral rules (in Kantian terms,

29
  Among those who find the idea of unknowable obligations objectionable along
such lines are Williams (1981); Sider (1995); Thomson (2008); Kramer (2009). Sorensen
defends unknowable obligations in his (1995).
30
  Shafer-Landau makes this point repeatedly in his (2003), esp. in ch. 10. On the
other hand, in his earlier (1994) he distances himself from the epistemicist’s unknowable
truths when he says:  “As a metaphysical realist, I  can countenance some unknowable
truths. But the number of such truths required by [epistemicists] is so great as to cast
doubt on the plausibility of their view.”
 Moral Vagueness 177

we could say: not merely in accordance with, but from duty). But acting from
duty involves being in a certain state of mind: knowing the contents of our
fact-given obligations. Recently, Dougherty writes in a similar vein:
[O]‌ne might hold that if it is a fact that you must take a taxi that costs less than $35.41
in order to keep a promise to meet someone at a particular time, then you must be able
to decide to take a $35.41 taxi on the basis of this fact. That is, arguably, the weakest link
one could find between ethical facts and motivation. But friends of epistemic explana-
tions of ethical vagueness should deny that even this link obtains. This is because ethical
facts that are unknowable could not guide action. (Dougherty 2013: 10)
The non-naturalist seeking to adopt an epistemic theory of vagueness is not
likely to be very impressed by such objections, though. For one thing, the
scope of unknowable facts should not be overstated:  that there are some
unknowable truths within a grey area of radical ignorance doesn’t impugn on
the whole realm of moral facts outside that area. But even focusing on facts
in the grey area itself, the non-naturalist can bite Dougherty’s bullet and sim-
ply deny any link between moral facts and motivation: after all, if moral facts
aren’t supposed to be projected, or constructed, from facts about our moti-
vational states, then why expect them to always be responsive to such states?
Call Dougherty’s taxi-involving moral fact “M.” According to Dougherty,
one might expect that if M obtains, one should be able to decide to act as M
requires on the basis of M itself. But why should that be a valid expectation?
Presumably, by “deciding on the basis of a fact” Dougherty means “deciding
on the basis of one’s knowledge of that fact.” That, although ideal, is hardly
a requirement: often, we are forced to decide solely on the basis of our beliefs
about the facts. And deciding to act as M requires based on one’s belief that
M is certainly possible in Dougherty’s scenario.
The difficulty, I believe, comes from focusing on motivation. One way
in which moral facts can be action-guiding is by producing corresponding
motives to act. But aside from motivation, action-guidingness can also mean
providing a justification, and by focusing on justification we can uncover,
I believe, a deeper problem with the notion of unknowable moral facts. The
problem is, roughly, that it is much harder to drive a wedge between moral
facts and justification than it is to separate facts from motivation.
The distinction between justification and motivation corresponds to
that between normative reasons and motivating reasons. Normative rea-
sons are justifying considerations. When we have such reasons, and act for
them, they become our motivating reasons.31 Call these “n-reasons” and
“m-reasons,” respectively. Now, according to the thesis of Rationalism, as
embraced by Shafer-Landau, Parfit, and others, moral facts are intrinsically

  Here I follow Parfit (2011: i. 37).

31
178 Cristian Constantinescu

reason-giving. What kinds of reasons are meant here? Should we think of

moral facts as intrinsically m-reason-giving? We could (some realists do), but
needn’t. It seems more plausible to interpret the thesis instead in terms of
n-reasons: moral facts intrinsically generate n-reasons; an agent has the cor-
responding m-reasons when the agent appropriately relates to her n-reasons.
Now, return to our Abe and Ben example once more. According to epis-
temicism, there is a sharp boundary between the actions that are permissible
for Abe and those that are not, and the difference can be as minute as one
nanosecond. Suppose the line between permissible and impermissible falls
between 1.44 × 1013 and (1.44 × 1013) + 1 ns. That is, it would be permis-
sible for Abe to spend four hours with Ben, but one nanosecond more and he
would be acting wrongly. It is therefore a perfectly determinate moral fact, call
it “F,” that it’s wrong for Abe to stay with Ben for (1.44 × 1013) + 1 ns. But Abe
cannot know F, due to the margin-for-error principle encountered above: any
justification that would entitle Abe to claim knowledge of F would also entitle
him to claim knowledge of the fact that it’s wrong to stay with Ben for 1.44 ×
1013 ns. But the latter isn’t a fact. So F is unknowable for Abe. Now, accord-
ing to Rationalism, F generates an n-reason for Abe to stop his emergency
session with Ben after at most 1.44 × 1013 nanoseconds (call this action “ϕ”).
However, Abe has no justification for believing he ought to ϕ, for he cannot
know that the line between permissible and impermissible falls there. We
have reached the following conclusions: there is an F-given n-reason for Abe
to ϕ, but no justification for him to ϕ. This seems almost contradictory, since
n-reasons are supposed to be justifying considerations in favour of actions. But
the contradiction is merely apparent: we can distinguish between “There is an
F-given n-reason for Abe to ϕ” and “Abe has an F-given n-reason to ϕ.” In our
case, the latter is false but the former can well be true.
It makes sense, of course, to separate one’s n-reasons from one’s actual
epistemic state in this way. But what doesn’t seem possible is to divorce
n-reasons even from a maximally improved capacity for practical rational-
ity. Thus, we can of course accept that there may be moral reasons for us to
desist from some of our current practices, but that those reasons are inac-
cessible to us, due to certain biases or errors in our judgment of which we
are unaware. But to recognize them as reasons means to accept that they
would serve as justifications for us if our reasoning abilities were improved.
What seems incoherent is the thought of an n-reason entirely divorced even
from the sound exercise of a maximally improved capacity for practical rea-
soning.32 To claim that there are reasons which couldn’t be anyone’s reasons

32
  I draw quite substantive inspiration here from Lillehammer (2002), which provides
one of the most forceful arguments I know for taking reasons to be essentially tied to the
exercise of our deliberative capacities.
 Moral Vagueness 179

seems almost vacuous. I shall express this upshot in the form of the follow-
ing epistemic constraint on normative reasons:
Epistemic constraint on reasons: If R is an n-reason for X to Φ, then R
can feature in a rational justification of the claim that X ought to Φ,
a justification which X knows or could come to know if X’s reasoning
abilities were maximally improved.
Non-naturalists can, and often do, acknowledge this conceptual link
between reasons and our maximally improved reasoning abilities. Thus,
Parfit defends an objectivist view of reasons according to which “when it
is true that we have decisive reasons to act in some way, this fact makes it
true that if we were fully informed and both procedurally and substantively
rational, we would choose to act in this way” (2011: i. 63). Similarly, in
the course of defending the notion of unknowable moral facts (though not
in the context of vagueness), Shafer-Landau concedes that “realists are not
committed to the idea that moral truths are inaccessible to absolutely ideal
epistemic agents at the Piercean limit of enquiry. Epistemically ideal agents
who have reached this limit will be fully informed. This means that they will
know all facts. Moral realists believe that some of these facts are moral ones;
so a genuinely ideal epistemic judge will know all moral facts” (2003: 17).
Now, on an epistemic account which takes vagueness to be just ignorance
concerning the sharp extensions of our moral concepts, this ignorance is
irremediable:  no improvement in our rational abilities could remove this
uncertainty, due to the margin-for-error constraints attaching to knowl-
edge. So there are radically unknowable moral facts which not even agents
with maximally improved rational capacities can come to know. If that is so,
then either those facts fail to generate n-reasons, in which case Rationalism
is false, or else there are reasons which cannot feature in an intelligible jus-
tification for anyone, violating the epistemic constraint on reasons above.
There is a reply here on behalf of the non-naturalist, which relies on
questioning the notion of a “maximally improved rationality” that I have
been using so far.33 Why assume that an ideal agent, who took every oppor-
tunity of improving and refining her reasoning abilities (both instrumental
and substantive) and knows all the facts, would still be ignorant about the
boundaries of our moral concepts? In the case of vague concepts whose
application depends on perceptual criteria (“red,” “tall,” “heap,” etc.), it
makes sense to think that even someone possessed of a perfect capacity for
reasoning would still remain ignorant about the relevant sharp boundaries,

33
  Thanks to Jen Hornsby, Michael Garnett, and Sarah McGrath for independently
alerting me to this possibility.
180 Cristian Constantinescu

because the margins-for-error at play are determined by our imperfect pow-

ers of perceptual discrimination and there is no reason to think that per-
fect rationality increases one’s ability to perceptually discriminate down to
nanometers or nanoseconds. However, the non-naturalist might hold that
things are different in the moral realm: if moral properties are non-natural,
it might be the case that we have epistemic access to them through some
form of rational intuition. But then, to imagine an agent whose rational
capacities are maximally improved just is to imagine an agent with perfect
moral discrimination. In that case, we could tie our conception of reasons
to this image of the ideal agent, and preserve Rationalism by claiming that
all facts, including the (currently) unknowable, are reason-giving.
Suppose now that Mia is such an ideal agent. Mia woke up one morning
and found that she had perfect rationality and unfailing moral discrimi-
nation. She knows now precisely when an action stops being permissible,
when a white lie turns into a guilty lie, when a foetus becomes a person,
how much kindness is too much, etc. Everyone around her still struggles, as
per usual: people have a hard time deciding what to do, when to tell a white
lie, how much to give to charity, who needs their help more, etc. Abe, for
instance, has just spent almost four hours trying to help Ben overcome his
panic attack, but has promised his family he’ll be home in the afternoon.
He doesn’t know how much more he can benefit Ben by being there, nor
how Ben’s benefit would weigh up against the distress his absence is caus-
ing at home. Mia, however, knows that the line between permissibility and
impermissibility falls precisely at 1.44 × 1013 nanoseconds. Abe is coming
up against it quite quickly, so she knows that he should get ready to go back
home. But he has no way of knowing that, and she remembers just how dif-
ficult these choices used to be. Can Mia say that there’s a normative reason
for Abe to leave soon, despite the fact that he really has no way of knowing
it? That’s probably neither here nor there. But suppose now that Abe leaves
after 1.44  × 1013–1 nanoseconds, i.e. one nanosecond before the precise
boundary between permissibility and impermissibility. Mia is happy and
praises Abe. If he’d been one nanosecond late, his action would have been
wrong. Mia would have blamed and chastized him.
Something here is amiss. Is it right for Mia to judge Abe so differently
in the two situations imagined, despite the fact that in one he leaves just
one nanosecond later than in the other? Considerations like this have led
some to argue that, in the moral realm, we should reject the epistemicist’s
idea of precise moral properties on moral grounds. In general, it seems that if
two people (actions, situations, etc.) X and Y are almost absolutely indistin-
guishable, with the exception of a minute difference with respect to one of
the considerations influencing our moral appraisals, then it would be unjust
for X to receive different treatment or to be appraised differently than
 Moral Vagueness 181

Y. But that is exactly what happens in a case of vagueness if the epistemic

account is correct: the slightest difference (one nanosecond, one nanogram,
one nano-anything) is all it takes for an option to change moral valence.
This appears to violate moral considerations pertaining to justice. Based on
such considerations, Matthew Kramer (2009) has recently argued against
vagueness-epistemicism and its attending concept of unknowable obliga-
tions, by claiming that such notions would make morality “ludicrous” and
“unfair.” Kramer works within a theoretical framework that collapses any
distinction between ethics and metaethics, insisting instead on treating the
issues of the latter domain as issues belonging to the former. Thus, the ques-
tion of whether there are any unknowable moral obligations becomes, for
Kramer, a question concerning our ethical commitments: should we coun-
tenance such obligations in our moral system? What would be the ethical
implications of doing so? Because he thinks unknowable obligations would
be unjust, Kramer argues that there can’t be any, and therefore that the
moral realist would do well to accept genuine moral vagueness, without
attempting to reduce it to uncertainty. But this move is not likely to impress
the non-naturalist, who takes the business of moral theorizing to be not that
of constructing, but of discovering moral truths. Since we don’t create moral
facts, why assume that we are at liberty to countenance or reject some of
them from our ontology on moral grounds?
I believe, however, that Kramer’s point can be restated in language
that is more likely to impress the robust non-naturalist. Consider again
Shafer-Landau’s conception of the relation between natural facts and
non-natural facts:

According to the sort of ethical non-naturalism that I favour, a moral fact super-
venes on a particular concatenation of descriptive facts just because these facts real-
ize the moral property in question. Moral facts necessarily covary with descriptive
ones because moral properties are always realized exclusively by descriptive ones. Just
as facts about a pencil’s qualities are fixed by facts about its material constitution, or
facts about subjective feelings by neurophysiological (and perhaps intentional) ones,
moral facts are fixed and constituted by their descriptive constituents. (2003: 77)

Consider Shafer-Landau’s pencil example. The pencil’s qualities supervene

upon its material constitution, meaning that there couldn’t be a change in
the pencil’s qualities without a change in material constitution. But it’s a
relatively overlooked fact that supervenience seems to also obey a quantita-
tive requirement of proportionality. This means that there cannot be a great
change in the supervening properties without a great change in the subven-
ing base. Thus, a minute change in one of the pencil’s atoms cannot make it
the case that the pencil is now rubbish when before it wrote perfectly fine;
nor can the pencil turn from long to short by losing just one nanometer.
182 Cristian Constantinescu

I  am therefore inclined to think that we should uphold the following con-

straint on the supervenience relation:
Proportionality constraint on supervenience:  If P-properties supervene
on Q-properties, then no two things can differ greatly with respect to
their P-properties without differing greatly also with respect to their
Q-properties.
Armed with this constraint, it now becomes quite evident that Kramer’s point
can be restated in metaphysical rather than purely moral terms. The fact that it’s
unjust to praise one person and blame another when the difference between their
actions was slight is, of course, a moral consideration. But the consideration is
grounded, I believe, in the proportionality constraint on supervenience: moral
responsibility, praise, and blame are concepts that supervene on the natural prop-
erties instantiated by those people’s actions. If the difference in those natural
properties was small, so must be the difference in their moral properties, too.
I conclude, based on all of this, that epistemicism about vagueness isn’t
an option for the non-naturalist seeking to preserve Supervenience. We have
examined three accounts of sharp properties in this section, and found them
incompatible with non-naturalism. It looks as if the non-naturalist can appeal
neither to theories positing vague properties, nor to accounts based on sharp
properties, in his attempt to elucidate moral vagueness.
The dilemma appears to stick.

4. Conclusions

I have argued in this paper that the moral non-naturalist seeking to counte-
nance moral vagueness in an attempt to explain moral disagreement faces a
dilemma. Non-naturalism I  have described as commitment to seven the-
ses:  Cognitivism, Correspondence, Atomism, Objectivism, Supervenience,
Non-reductivism, and Rationalism. On either horn of the dilemma, serious
problems arise for some of these theses: in various ways, vague properties seem
to threaten Objectivism, Supervenience, and Non-reductivism; on the other
hand, sharp properties raise problems for Supervenience and Rationalism. The
difficulties on each horn of the dilemma are real, and while they may not be
insuperable, they do, at the very least, drastically limit the things non-naturalists
can consistently say about moral properties, facts, and reasons. Non-naturalism
may in the end survive my dilemma, but if it does it will be a doctrine consider-
ably different from what some of its leading proponents take it to be.34

34
  I am indebted to Maike Albertzart, Tom Dougherty, Antti Kauppinen, Laura Vasile,
and two anonymous reviewers for Oxford University Press for extremely helpful com-
ments on earlier drafts of this chapter. Previous versions of the material were presented
 Moral Vagueness 183

References
Aristotle. 2009. Nicomachean Ethics, tr. D. Ross. Oxford and New  York:  Oxford
University Press.
Armstrong, D. M. 1978. A Theory of Universals: Universals and Scientific Realism, ii.
Cambridge and New York: Cambridge University Press.
Audi, R. 2004. The Good in the Right:  A  Theory of Intuition and Intrinsic Value.
Princeton: Princeton University Press.
Blackburn, S. 1984. Spreading the Word. Oxford: Oxford University Press.
Crisp, R. 2006. Reasons and the Good. Oxford: Oxford University Press.
Cuneo, T. 2007. “Recent Faces of Moral Nonnaturalism,” Philosophy Compass,
2: 850–79.
Dancy, J. 1981. “On Moral Properties,” Mind, 90: 367–85.
Dancy, J. 2006. “Nonnaturalism”, in D. Copp (ed.), Oxford Handbook of Ethical
Theory, 122–45. New York: Oxford University Press.
Dougherty, T. 2013. “Vague Value,” Philosophy and Phenomenological Research (pub-
lished online first, DOI: 10.1111/phpr.12026).
Dworkin, R. 1977. “No Right Answer?” in P. Hacker and J. Raz (eds), Law, Morality,
and Society, 58–84. Oxford: Clarendon Press.
Dworkin, R. 2011. Justice for Hedgehogs. Cambridge, MA, and London: Harvard
University Press.
Enoch, D. 2011. Taking Morality Seriously:  A  Defense of Robust Realism.
Oxford: Oxford University Press.
Fine, K. 1975. “Vagueness, Truth and Logic,” Synthese, 30: 265–300.
FitzPatrick, W. 2008. “Robust Ethical Realism, Non-Naturalism, and Normativity,”
in R. Shafer-Landau (ed.), Oxford Studies in Metaethics, iii. 159–205.
Oxford: Oxford University Press.
Gibbard, A. 2006. “Normative Properties,” in T. Horgan and M. Timmons (eds),
Metaethics After Moore, 319–38. Oxford: Clarendon Press.
Gillett, C., and Rives, B. 2005. “The Non-Existence of Determinables: Or, a World
of Absolute Determinates as Default Hypothesis,” Noûs, 39: 483–504.
Huemer, M. 2005. Ethical Intuitionism. New York: Palgrave Macmillan.
Hurley, S. 1992. Natural Reasons:  Personality and Polity. Oxford:  Oxford
University Press.
Hyde, D. 2008. Vagueness, Logic and Ontology. Aldershot: Ashgate.
Jackson, F. 1998. From Metaphysics to Ethics:  A  Defence of Conceptual Analysis.
Oxford: Clarendon Press.

at the 2012 Madison Metaethics Workshop, and at a departmental seminar at Birkbeck

College, University of London. I am grateful to audiences at both events, and in particu-
lar to David Copp, Terence Cuneo, James Dreier, Dorothy Edgington, Michael Garnett,
Jen Hornsby, Keith Hossack, Sarah McGrath, Robert Northcott, Anthony Price, Russ
Shafer-Landau, and Robert Shanklin. Hallvard Lillehammer read a very early ancestor of
this chapter many years ago, and gave me lots of insightful suggestions which provided
the impetus for much of the work done here.
184 Cristian Constantinescu

Keefe, R. 2000. Theories of Vagueness. Cambridge: Cambridge University Press.

Kramer, M. 2009. Moral Realism as a Moral Doctrine. Oxford: Wiley-Blackwell.
Lewis, D. 1986a. On the Plurality of Worlds. Oxford: Basil Blackwell.
Lewis, D. 1986b. “Against Structural Universals,” Australasian Journal of Philosophy,
64: 25–46.
Lillehammer, H. 2002. “Moral Realism, Normative Reasons, and Rational
Intelligibility,” Erkenntnis, 57: 47–69.
Mellor, D. H. 2012. “Nature’s Joints: A Realistic Defence of Natural Properties,”
Ratio, 25: 387–404.
Parfit, D. 2011. On What Matters. Oxford: Oxford University Press.
Railton, P. 1992. “Pluralism, Determinacy, and Dilemma,” Ethics, 102: 720–42.
Ridge, M. 2007. “Anti-Reductionism and Supervenience,” Journal of Moral
Philosophy, 4: 330–48.
Rosen, G., and Smith, N. J. J. 2004. “Worldly Indeterminacy: A Rough Guide,”
Australasian Journal of Philosophy, 82: 185–98.
Ross, W. D. 1930. The Right and the Good. Oxford: Oxford University Press.
Russell, B. 1923. “Vagueness,” Australasian Journal of Philosophy and Psychology,
1: 84–92.
Sainsbury, R. M. 1989. “What is a Vague Object?,” Analysis, 49: 99–103.
Sainsbury, R. M. 1996. “Concepts without Boundaries,” in R. Keefe and P. Smith
(eds), Vagueness: A Reader, 251–64. Cambridge, MA: MIT Press.
Sanford, D. H. 1966. “Red, Green, and Absolute Determinacy,” Philosophical
Quarterly, 65: 356–8.
Sanford, D. H. 1970. “Disjunctive Predicates,” American Philosophical Quarterly,
7: 162–70.
Sanford, D. H. 2013. “Determinates vs. Determinables,” in E. N. Zalta (ed.), The
Stanford Encyclopedia of Philosophy (spring 2013 edn): <http://plato.stanford.edu/
archives/spr2013/entries/determinate-determinables>.
Scanlon, T. M. 2003. “Metaphysics and Morals,” Proceedings and Addresses of the
American Philosophical Association, 77: 7–22.
Schiffer, S. 2000. “Vagueness and Partial Belief,” in E. Sosa (ed.), Philosophical Issues,
x. Skepticism, 220–57. Oxford: Blackwell.
Schiffer, S. 2010. “Vague Properties,” in R. Dietz and S. Moruzzi (eds), Cuts and
Clouds:  Essays on the Nature and Logic of Vagueness, 109–30. Oxford:  Oxford
University Press.
Shafer-Landau, R. 1994. “Ethical Disagreement, Ethical Objectivism, and Moral
Indeterminacy,” Philosophy and Phenomenological Research, 54: 331–44.
Shafer-Landau, R. 1995. “Vagueness, Borderline Cases and Moral Realism,”
American Philosophical Quarterly, 32: 83–96.
Shafer-Landau, R. 2003. Moral Realism: A Defence. Oxford: Clarendon Press.
Shapiro, S. 2006. Vagueness in Context. Oxford: Clarendon Press.
Sider, T. 1995. “Sorensen on Unknowable Obligations,” Utilitas, 7: 273–9.
Sorensen, R. 1990. “Vagueness Implies Cognitivism,” American Philosophical
Quarterly, 27: 1–14.
Sorensen, R. 1995. “Unknowable Obligations,” Utilitas, 7: 247–71.
 Moral Vagueness 185

Sosa, E. 2001. “Objectivity without Absolutes,” in A. Byrne, R. C. Stalnaker, and

R. Wedgwood (eds), Fact and Value: Essays on Ethics and Metaphysics for Judith
Jarvis Thomson, 215–27. Cambridge, MA, and London: MIT Press.
Thomson, J. J. 2008. Normativity. Chicago: Open Court Press.
Tye, M. 1990. “Vague Objects,” Mind, 99: 535–57.
Vasile, L. J. 2010. “Moral Disagreement and Moral Indeterminacy,” unpublished
doctoral dissertation, Department of Philosophy, Central European University.
Wedgwood, R. 2007. The Nature of Normativity. Oxford: Oxford University Press.
Williams, B. 1981. Moral Luck. Cambridge: Cambridge University Press.
Williamson, T. 1994. Vagueness. London: Routledge.
Williamson, T. 2000. Knowledge and its Limits. Oxford: Oxford University Press.
Williamson, T. 2003. “Vagueness in Reality”, in M. J. Loux and D. Zimmerman
(eds), Oxford Handbook of Metaphysics, 690–715. Oxford: Oxford University
Press.
Wolf, S. 1992. “Two Levels of Pluralism,” Ethics, 102: 785–98.
 8
Relax? Don’t Do It! Why Moral Realism
Won’t Come Cheap
Sarah McGrath

1. Introduction

Consider the following claims:

(1) There are objective moral truths. These truths, like normative truths
more generally, are irreducibly normative, in the sense that they are not
reducible to any set of non-normative truths.
(2) Moral properties differ from natural properties in several significant
respects. In particular, moral properties are not causally efficacious in
the way that natural properties are, are not the kinds of things that we
can investigate on the basis of ordinary perception or empirical inquiry,
and do not figure in the best explanations of why we believe what we
do about the world.
(3) Nevertheless, it is a mistake to think that these features of moral prop-
erties cast doubt either on morality or on our ability to attain moral
knowledge. One significant implication of this is that a much-discussed
test associated with Harman (1977), which invites us to treat a certain
kind of explanatory indispensability as a sort of litmus test for reality, is
not a test that morality must pass. At best, Harman’s test is useful in the
domain of natural science, but it is a confusion to try to apply it to the
moral or normative domains.
(4) Morality lacks controversial metaphysical and empirical presupposi-
tions. The moral domain thus enjoys a certain kind of autonomy, and
morality is much less vulnerable to being “debunked” by either meta-
ethical theorizing or empirical discoveries than is often assumed.
 Relax? Don’t Do It! 187

While questions might be raised about these exact formulations, I hope

that I have succeeded in sketching a general picture or perspective that is
familiar. I  will call this picture relaxed realism, in an attempt to capture
the way in which its proponents combine a commitment to realism with
a certain lack of anxiety about the status and standing of morality, despite
understanding morality in ways that might naturally encourage such anxi-
ety. Although this view is not a new one, it is fair to say, I think, that its
prominence and influence have never been greater. For example, even if we
restrict our attention to the last few years, relaxed realist themes are central
to Dworkin’s Justice for Hedgehogs (2011), Parfit’s On What Matters (2011),
and Scanlon’s Being Realistic about Reasons (2014).1
In this chapter, I want to raise some doubts about this general picture of
morality and some prominent ways of defending it. Toward this end, in the
next section I take up a comparison that is frequently invoked by relaxed
realists, and one on which they often place a significant amount of weight: a
comparison between irreducibly normative properties and truths on the one
hand, and mathematical properties and truths on the other. I argue that this
comparison is much less favorable to the relaxed realist’s cause than is often
thought. I then turn to a particularly vigorous and sustained presentation
of relaxed realism, that offered by Dworkin in Justice for Hedgehogs. Because
Dworkin’s commitment to this general picture is undiluted and uncompro-
mising, I believe that his defense of it provides a useful case study. I argue
that the lessons that emerge do not support relaxed realism.

2.  Does Mathematics Give us a Reason

to Relax?

Relaxed realists frequently compare morality to mathematics.2 A common

theme of such comparisons is this: many of the same skeptical concerns that
are raised about irreducibly normative truths (and how we could know such
truths if they exist) seem to apply equally well to mathematics and to our
knowledge of mathematics. Perhaps irreducibly normative properties would
lack causal efficacy, play no essential role in the best explanations of why we
believe as we do, and are not the kinds of things that we can discover on
the basis of ordinary perception or empirical inquiry. So long as the salient
contrast is with paradigmatically natural properties, these features are apt
to arouse suspicion about the irreducibly normative. But significantly, these

  Another influential relaxed realist is Nagel; see especially his (1997: ch. 6).

1

  See e.g. Scanlon (2014: lectures 2 and 4; 1998: 62–3); Parfit (2011: 488–510); and

2

Nagel (1997).
188 Sarah McGrath

same features seem to be shared by mathematical properties. Assuming that

we do possess mathematical knowledge, it follows immediately that the fact
that a domain is characterized by the relevant features poses no insuperable
obstacle to our attaining knowledge of it.
In fact, we can distinguish at least three related ways in which the com-
parison with mathematics might seem to bolster the relaxed realist’s case.
First, mathematical properties and truths might seem to provide ideal
“companions in guilt” for irreducibly normative properties and truths, in
the ways just described. Here, the appeal to mathematics is used to make
an essentially negative point: that a domain’s having certain features does
not preclude our having knowledge of it. Second and more constructively,
mathematics might be thought to provide a model for how we manage to
actually attain knowledge in a domain with those features.3 Finally, as we
will see, relaxed realists often take the example of mathematics to show that
Harman’s explanatory test (however exactly that test is understood) has at
best a limited range of applicability, and therefore is not a test that morality
must pass.
Given the potential significance of the mathematics comparison, it is
worth scrutinizing closely. I’ll argue that the similarities between mathe-
matical properties and moral properties have often been exaggerated when
moral properties are understood in the ways that the relaxed realist would
have us understand them—as properties that lack causal efficacy, perceptual
accessibility, and which play no essential role in the best explanations of
our beliefs about the world. The upshot, as I see it, is that the example of
mathematics is much less useful for the relaxed realist’s purposes than one
might have supposed. In fact, I’ll argue that in some ways the example of
mathematics should make us more anxious about the normative and moral
domains, if we understand those domains in the way that the relaxed realist
would have us understand them.

Harman’s Test and Explanatory Relevance

Harman (1977) invited us to compare two cases. In the first case, a scientist
judges “there goes a proton” in response to seeing a vapor trail in a cloud

3
  See e.g. Scanlon’s (2014) extended discussion of the epistemology of set theory in
the fourth of his Locke Lectures, a lecture devoted to addressing epistemological con-
cerns about irreducibly normative facts about reasons. A similar tack is taken by Parfit
(2011: 488–510). Like Scanlon, Parfit proceeds by considering the case of mathematics,
which, he assumes, is a domain in which we have knowledge of non-natural, causally
inefficacious properties and truths. Having sketched a rough account of how we manage
to attain mathematical knowledge, he proceeds to argue that a parallel story might very
well account for our ability to acquire knowledge of irreducibly normative truths.
 Relax? Don’t Do It! 189

chamber. In the second case, you judge “that’s wrong” in response to see-
ing a group of children set a cat on fire. Harman claimed that, while the
best explanation of the scientist’s judging as he does invokes a proton’s hav-
ing passed by, the best explanation of your judging that the children acted
wrongly does not invoke the wrongness of their act. Rather, we can explain
your making that moral judgment in the circumstances simply by invok-
ing facts about your psychology. A  common interpretation of Harman’s
discussion is that it proposes a kind of explanatory test, a test that morality
fails to pass. According to this line of thought, we have reason to believe in
the existence of things of a certain kind only if they play a role in the best
explanation of our observations of the world, but moral facts never play
such a role.
As noted, relaxed realists sometimes allow that the explanatory require-
ment associated with Harman is an appropriate one so long as it is applied
to the natural world. But they argue that it is not appropriately applied to
other domains, including the moral and normative domains. In this con-
text, mathematics is often cited—alongside the contested cases of morality
and the normative—as a paradigm of a domain for which the test is clearly
inappropriate. Here, for example, is Scanlon:
In the domain of natural science . . . Harman’s explanatory requirement makes good
sense in this form: we have reason to be committed to the existence of things of a
certain sort only if they play a role in explaining what happens in the natural world
(including our experience of it). But this maxim is specific to the domain of natural
science. It does not apply, as Harman’s explanatory requirement is often held to
apply, to every domain, for example to the normative domain, or to mathematics.
(2014: 26)

Similarly, Dworkin (1996: 119) concedes that Harman’s test, or something

in the near vicinity, is appropriate for “beliefs about the physical world.” But
he denies that the test is appropriate for moral beliefs, and he too explicitly
cites mathematical beliefs as among the types of beliefs for which it would
clearly be inappropriate.
In his original discussion, Harman anticipated the objection that the case
of mathematics shows that the test is not generally applicable (1977: 9–10).
As understood by Harman, the objection from mathematics runs as fol-
lows: because we cannot be in causal contact with numbers or mathematical
facts, there is no observational evidence for mathematics; therefore, math-
ematics and morality are in the same epistemological boat. In response,
Harman gestured at a broadly Quinean epistemology of mathematics,
according to which we have indirect observational evidence for math-
ematics, in a way that we do not have indirect observational evidence for
ethics. According to this line of thought, we have indirect observational
190 Sarah McGrath

evidence for mathematics because of the way in which mathematics is

bound up with physical theories that can be observationally confirmed.
This broadly Quinean picture of the way in which mathematical truths
are confirmed by observation is open to serious challenge, on the grounds
that it exaggerates the extent to which confirmation is a holistic matter
(Sober 2000). Nevertheless, I  believe that Harman was right about the
essential point:  mathematical facts sometimes play an indispensable role
in the best explanations of our observations. Moreover, that this is so does
not depend on the truth of any controversial Quinean claims about the way
in which observational evidence indirectly confirms mathematics. In fact,
I believe that there are cases in which the best explanation of our observa-
tions appeals directly to mathematical facts, and in which it is natural to cite
a mathematical fact as “the” explanation of those observations.
We can begin by noting that there are at least some cases in which the
best explanation of why physical objects are arranged in a certain way makes
essential reference to facts of pure mathematics. For example, it sometimes
makes sense to cite geometrical facts in explaining why physical objects are
arranged as they are. Here is an example that I borrow from Peter Lipton,
who uses it to make a somewhat different point:
Suppose that some sticks are thrown into the air with a lot of spin, so that they
separate and tumble about as they fall. Now freeze the scene at some point during
the sticks’ descent. Why are appreciably more of them near the horizontal axis than
the vertical, rather than in more or less equal numbers near each orientation, as one
would have expected? The answer, roughly speaking, is that there are many more
ways for a stick to be near the horizontal than near the vertical. To see this, consider
purely horizontal and vertical orientations for a single stick with a fixed midpoint.
There are infinitely many of the former, but only two of the latter. Or think of the
shell that the ends of that stick trace as it takes every possible orientation. The areas
that correspond to the near vertical are caps centered on the north and south poles
formed when the stick is forty-five degrees or less off the vertical, and this area is
substantially less than half the surface area of the entire sphere. Less roughly, the
explanation why more sticks are near the horizontal than near the vertical is that
there are two horizontal dimensions but only one vertical one. (1991: 33–4)
Here the explanandum—that at a particular moment, more of the sticks
are near the horizontal axis than the vertical axis—concerns why physical
objects are arranged in the way that they are. Nevertheless, this fact admits
of a mathematical explanation. Of course, in principle, one could try to
provide an alternative, non-geometrical explanation of the same explanan-
dum: one could (at least in principle) provide for each stick a physical expla-
nation of why, given the forces to which it was subject after it was released
from one’s hand, it ended up in the position that it did at that particular
moment. One could then conjoin all of these independent explanations of
 Relax? Don’t Do It! 191

why each stick ended up in a certain position into one long story, a story
that amounts to a sufficient condition for the obtaining of the explanan-
dum fact. But for reasons that are familiar from the general theory of expla-
nation, it’s a mistake to think that this conjunctive physical explanation is as
good as the geometrical explanation. In particular, to treat the conjunctive
physical explanation as the best explanation in this context is in effect to
treat the explanandum fact as much more fragile and contingent than it
really is. For even if the physical facts had been quite different than they
actually were, the explanandum fact would still have obtained, and this is
something that can be understood in terms of the geometrical explanation
but not in terms of the hypothesized physical explanation.4
The best explanation of the fact that the sticks end up arranged in the
way that they do is a geometrical explanation. From there, it’s a short step to
the conclusion that geometrical facts can play a role in explaining why one
believes as one does. For if one carefully observes the stick experiment, one’s
belief that the sticks are oriented in the way that they are will be explained
by the fact that sticks are arranged in that way. Given that the geometrical
facts explain the latter fact, they will also appear in a sufficiently compre-
hensive account of why one ends up in the relevant psychological state. So
there are cases in which mathematical facts figure in the best explanations
of our observational beliefs.5
It is worth noting the following points:  (1)  One can agree that math-
ematical facts sometimes play an indispensable role in the best explanations
of our observations even if one does not think that mathematical facts ever
cause anything to happen. Notably, although Lipton thinks that the best
explanation of why the sticks end up arranged as they do is a geometrical
explanation, he explicitly denies that this is a causal explanation, since, he
assumes, geometrical facts cannot be causes (1991: 34).6 (2) Similarly, one

4
  In the words of Alan Garfinkel, good explanations should be “invariant under small
perturbations of the initial assumptions” (quoted in Putnam 1975: 301; see Garfinkel
1981 for an extended defense of the idea).
5
  There are non-geometrical examples as well. Consider the way in which evolution-
ary biologists account for the highly unusual life cycle of the North American cicada.
Adult cicadas emerge either every 13 or 17  years, depending on their geographical
region; remarkably, this emergence is synchronized across all of the cicadas in a given
area. Among evolutionary biologists, there are two competing hypotheses about what
accounts for these patterns, one that emphasizes the avoidance of predators, the other of
which emphasizes the avoidance of hybridization with similar subspecies. Significantly,
both of these hypotheses make essential reference to the fact that 13 and 17 are prime
numbers, and appeal to the properties of primes as described by number theory. The case
is discussed at length by Baker (2005) who defends the claim that the number-theoretic
truths are essential to the proposed hypotheses.
6
  Indeed, Lipton employs the example in an attempt to show that not all explanations
of events or contingently obtaining states of affairs are causal explanations, pace Lewis
192 Sarah McGrath

might agree with the claim that the best explanation of why the sticks end
up arranged as they do is a geometrical explanation even if one does not
believe that one’s observing the sticks does anything to confirm or supply
evidence for the relevant mathematical facts. In short, even if we do not
presuppose the truth of a Quinean epistemology of mathematics, we should
agree that, in at least some cases, mathematical truths feature in the best
explanations of both (i) why physical objects are arranged in the ways that
they are, and (ii) why we end up holding the empirical beliefs that we do.
Insofar as one holds, as the relaxed realist does, that moral truths never play
a role in explaining either (i) or (ii), this is a significant disanalogy between
mathematical truths and moral truths.

Perceptual Accessibility
Another way in which relaxed realists think that irreducibly normative
truths differ from purely descriptive truths is in their lack of accessibil-
ity to perception. On their view, while we can have direct observational
knowledge of (many) purely descriptive truths about the world, we never
recognize irreducibly normative or moral truths on the basis of direct obser-
vation. This is another respect in which the comparison with mathematics
might seem helpful, inasmuch as we do not seem to arrive at our math-
ematical knowledge by direct observation either.7 (Notice that in Lipton’s
example, although the geometrical facts play an indispensable role in a suf-
ficiently comprehensive explanation of why we observe what we do, we do
not actually perceive the geometrical facts appealed to in the explanation
upon observing the scene.) So the example of mathematical properties and
truths might seem to be useful in dispelling suspicions about non-natural
normative properties and truths arising from their putative inaccessibility
to sense perception.
But here too, I think that there is less to the mathematics comparison
than initially meets the eye. In fact, I believe that when we look closely at the
case of mathematics, we do not find the kind of perceptual inaccessibility

(1986) and others. Baker (2005:  234)  also denies that mathematical explanations are
causal explanations.
7
  Note that even the Quinean empiricist about mathematics does not think that we
perceive particular mathematical truths by observing them; rather, on her view, the justi-
fication that is afforded to our mathematical beliefs by observation is taken to be highly
indirect. That is, for the Quinean, the empirical justification that we have for believing
truths about numbers is much like our justification for believing high level theoretical
claims about subatomic particles:  in neither case do we have any direct observational
contact with the entities in question.
 Relax? Don’t Do It! 193

that the relaxed realist thinks is characteristic of the moral and normative
domains.
Consider the following mathematical relation:
X is equinumerous with Y.

We can perceive instantiations of this relation in the world. For example, in

my household, it’s extremely important that my son and daughter receive
an equal number of cookies for dessert. When there are two cookies on
my son’s plate, and two cookies on my daughter’s plate, one can literally
see that relevant relation obtains between the two collections. The state of
affairs that consists in that relation’s obtaining between the collections is
perceptually accessible, in just the way that any number of other states of
affairs about my immediate environment are perceptually accessible (e.g.
the fact that there are now crumbs on the floor). Moreover, this is hardly an
irregular case. As Jaegwon Kim has written:
As objects of perceptual discrimination and judgment, there is nothing unusual,
uncommon or mysterious about numerical properties and relations or, more gener-
ally, mathematical properties and relations. Seeing that something is round, that
these are three green dots, that the dots over here are more numerous than those over
there, that there are more dots on the screen now than just a moment ago, and so
on are just as common, and practically and psychologically unproblematic, as seeing
that these dots are green, the dot on the left is larger and greener than the one on the
right, and so on . . . Mathematical properties do not differ in respect of perceptual
accessibility from sundry physical properties such as colors, shapes, odors, warmth
and cold. (1981: 345)

One might agree with what Kim says here, while denying that it detracts
from the usefulness of the mathematics comparison given the relaxed real-
ist’s purposes. Of course we can directly perceive states of affairs that con-
sist of physical objects instantiating mathematical and relational properties.
But what is not perceptually accessible are the abstract mathematical truths
themselves. For example, even if we can see that there are four cookies on the
table when there are two cookies on my son’s plate and two cookies on my
daughter’s plate, what we do not directly observe are the numbers 2 or 4, or
the abstract arithmetical truth that 2+2=4. And this, it might be objected,
is what a philosopher who appeals to mathematics in order to allay doubt
about the putative perceptual inaccessibility of the normative domain has
in mind.
Let’s grant for the sake of argument that we do not have direct perceptual
knowledge of any abstract mathematical truths, and that in this respect, our
knowledge of such truths resembles the knowledge that the relaxed real-
ist takes us to have of non-natural, irreducibly normative truths. Even so,
194 Sarah McGrath

I think that the mathematics comparison is ultimately unhelpful in defus-

ing worries that arise from the perceptual inaccessibility of non-natural
normative truths. For we have agreed that, even if we do not have direct
perceptual knowledge of abstract mathematical truths, we can directly per-
ceive facts that consist of concrete particulars exemplifying mathematical
properties and standing in mathematical relations to one another. (That is,
although we do not directly perceive that 2+2=4 via sense experience, we
can directly perceive that the number of cookies on my son’s plate is the
same as the number of cookies on my daughter’s plate.) What we would
expect in the moral case then, is this: although we cannot directly perceive
the truth of abstract moral claims not involving concrete particulars (e.g.
“It is wrong to torture animals for the fun of it”), we can directly perceive
moral properties and their relations when they are instantiated by concrete
particulars. That is, if morality is analogous to mathematics with respect to
its perceptual accessibility, then we would expect to be able to know on the
basis of sense perception that (e.g.) what those kids are doing is wrong, since
this state of affairs involves a moral property being instantiated by concrete
particulars. But that we can directly perceive that the hoodlums’ actions
instantiate a moral property is one of the things that is denied about the
case by relaxed realists.
The situation then, seems to be this. Although we cannot know abstract
mathematical truths on the basis of sense perception, we can see mathemat-
ical properties and relations instantiated by concrete particulars. In con-
trast, on the relaxed realist view, not only can we not know abstract moral
truths on the basis of sense perception, but irreducibly normative properties
and relations and properties can never be perceived, even when they are
instantiated by concrete particulars. But if that’s so, then the comparison
with mathematics seems ill suited to alleviating doubts about non-natural,
irreducibly normative truths that arise from their putative inaccessibility to
ordinary sense perception.
In fact, the comparison might further encourage doubts about irreducibly
normative properties so understood. This is because the class of mathemati-
cal properties turns out to be yet another class of properties some of whose
members can be detected by sense perception when instantiated by concrete
particulars. Notice that in this respect, the significant division seems to be
between the physical domain and the mathematical domain on the one
hand, and the normative domain-as-understood-by-the-relaxed-realist on
the other. Some facts about the natural world (e.g. the fact that the grass has
turned brown) are perceptually accessible, while more abstract truths about
the natural world (e.g. the laws of physics) are not. This parallels the way
in which many abstract truths of mathematics cannot be directly perceived,
although countless facts that consist of the instantiation of mathematical
 Relax? Don’t Do It! 195

properties and relations by concrete particulars are accessible to perception.

Of course, there are many mathematical properties and relations that are
not perceptually accessible, even when they are instantiated by concrete par-
ticulars in our immediate vicinity. But in this respect as well, the class of
mathematical properties resembles the class of physical properties, for there
are many paradigmatically physical properties that are perceptually inac-
cessible to us, even when they are instantiated or exemplified by concrete
particulars in our immediate vicinity. (Consider, for example, the paradig-
matically physical properties of being an electron or being a molecule.)

Causal Efficacy
Consider finally the claim that mathematical properties lack causal efficacy
and at least in this respect provide “companions in guilt” for irreducibly
normative properties as the latter are understood by relaxed realists. As was
just argued, we can (and frequently do) perceive exemplifications of many
mathematical properties, on those occasions when they are instantiated
by concrete particulars. Given that perception is a causal process, it fol-
lows immediately that mathematical properties are causally efficacious in
the relevant sense: we can be causally affected by states of affairs (events,
facts) that consist in the obtaining of mathematical relations and properties.
Here again is Kim, stating what I take to be the correct view of things with
respect to the case of mathematics:
Human perception is a causal process involving the features of the object or situ-
ation perceived and the states of our sense organs and nervous system. Just as the
character of our perceptual experience of there being a green dot is causally deter-
mined in part by the state of affairs of there being a green dot, so our perceptual
experience of there being three dots out there, or that there are more green dots than
red ones, is causally determined by there being three green dots, or there being more
green dots than red ones . . . Like any other concrete states of affairs these states of
affairs involving numerical properties are links in the pervasive causal network of the
world. In this respect there is no difference between mathematical properties instan-
tiated in physical situations on the one hand and the so-called physical properties on
the other . . . mathematical properties . . . are no worse off than such sundry physical
properties as color, mass, and volume, in respect of causal efficacy. (1981: 346–7)
But it is the contention of the relaxed realist that non-natural, irreducibly
normative properties are not causally efficacious in the way that physical
properties such as color, mass, and volume are.
Let’s recap the conclusions of this section. If what I have argued is cor-
rect, then the comparison between mathematics and morality as under-
stood by the relaxed realist stands as shown in the table.
196 Sarah McGrath

Physical Mathematical Moral Properties, as

Properties Properties Understood by the
Relaxed Realist
Do properties of the YES YES NO
relevant kind ever figure
in the best explanation
of our empirical beliefs?
Do we ever perceive YES YES NO
instantiations of such
properties via the senses?
Are properties of the YES YES NO
relevant kind (or their
instantiations) causally
efficacious?

Given these differences, I believe that the example of mathematics is much

less helpful to the relaxed realist than one might have thought.8

3.  A Case Study: Dworkin

Part I  of Dworkin’s Justice for Hedgehogs is an extended defense of the

relaxed realist picture. I want to look closely at some of his central argu-
ments, beginning with his case for the claim that, although some moral
views are objectively true, the fact that a moral view is true never plays a role
in explaining why it is held.

The Argument from Counterfactuals and

Psychological Explanations
We can begin by distinguishing Dworkin’s view from one that is signifi-
cantly more radical: the view that no opinion (moral or otherwise) is best
explained by an explanation that makes essential reference to the truth of

8
 To be clear, that conclusion is perfectly consistent with the claim that the exam-
ple of mathematics strongly bolsters the position of the moral realist in the dialectic
between moral realism and anti-realism. For many moral realists are not relaxed realists
and reject the characterization of moral properties and truths that has been assumed in
this discussion.
 Relax? Don’t Do It! 197

that opinion. Although this more radical view has sometimes been defended,
it is clear that it is not Dworkin’s own:
people’s beliefs about the physical world are often caused directly or indirectly by the
truth of what they believe, and when they are, that fact confirms the truth of their
belief. The best explanation of why I believe that it rained earlier today includes the
fact that it did rain. (2011: 71)
Moreover, Dworkin holds that, when it comes to ordinary empirical beliefs,
the lack of any possible explanatory connection between a belief and the
state of affairs that it purports to represent undermines the credibility of
that belief:
Suppose that though you believe it rained in France today, no rain in France could
possibly figure in any explanation of why you believe that . . . You would then have
no reason at all to think it had rained there. (71)
Given that Dworkin is a passionate defender of the view that moral
beliefs, like beliefs about the physical world, can be objectively true, it is
natural to expect him to tell a parallel story about the moral domain: that
in at least some favorable cases, the fact that a moral opinion is true fig-
ures in the best explanation of why it is held, and that in cases in which
there is no possible explanatory connection between the truth of a moral
opinion and its being held, that tends to undermine the credibility of
the opinion. However, Dworkin emphatically denies that the domains
are parallel in these respects. Rather, his view is that it is a deep confu-
sion to think that the truth of a moral opinion could ever figure in the
best explanation of why it is held, but that morality is none the worse
off for that.
Why should what holds for our beliefs about the physical world not
also hold for our moral beliefs? In both Justice for Hedgehogs and his earlier
paper “Objectivity and Truth: You’d Better Believe it,” Dworkin appeals to
the following idea: things are asymmetrical between our beliefs about the
physical world and our moral beliefs because the former beliefs, but not
the latter, are ones for which it makes sense to ask a certain kind of coun-
terfactual question. Specifically, it makes sense to ask of beliefs about the
physical world whether they would still be held even if the facts that they
purport to represent had been otherwise, but it makes no sense to ask the
same question of our moral beliefs. Here is a statement of this idea from
the earlier paper:
Consider Gilbert Harman’s suggestion that we cannot regard any belief as reliable
unless we think that the best causal explanation of why we hold it refers to the state
of affairs it describes. In some form, this test does seem appropriate to beliefs about
198 Sarah McGrath

the physical world. . . . But nothing in the content of moral (or aesthetic or math-
ematical or philosophical) opinions invites or justifies such a test. On the contrary,
the content of these domains excludes it, because an adequate causal explanation of
a belief includes showing that the belief would not have occurred if the alleged cause
had not been present, and we cannot understand or test that counterfactual claim
with respect to moral or aesthetic beliefs because we cannot imagine a world that is
exactly like this one except that in that world slavery is just or The Marriage of Figaro
is trash. (1996: 119)
He offers the same line of argument in Justice for Hedgehogs.9
Here the idea seems to be something like the following. The reason why
the best explanation of one’s belief that it rained earlier today might very well
invoke the fact that it rained earlier today is that we can both make sense of
and investigate the truth of the following counterfactual:
If it had not rained earlier today, then one would not now believe that
it rained earlier today.
In contrast, because we “cannot imagine a world that is exactly like this one
except that in that world slavery is just,” we cannot understand or investi-
gate the truth of the following counterfactual:
If slavery had not been unjust, then one would not believe that it is
unjust.
And according to Dworkin, our being able to understand and evaluate this
counterfactual is an (unfulfillable) necessary condition for the fact that slav-
ery is unjust to figure in the best explanation of one’s believing that slavery
is unjust.
This is a bad argument. In general, even if we cannot imagine a world
that is exactly like ours except that p is false, it does not follow that p can-
not be a part of the best explanation of why someone believes that p is true.
In order to see this, consider another case that Dworkin mentions in the
passage above: the case of mathematics. Imagine a mathematician who ini-
tially has no opinion about whether some mathematical conjecture is true
or false. Suppose that she subsequently succeeds in proving the conjecture
and thus comes to believe the relevant proposition on the basis of the proof.
In this case, the explanation of why the mathematician currently believes
the proposition is that she succeeded in proving the theorem. But of course,
that explanation entails that the relevant proposition is true. So it is essential
to the best explanation of why the mathematician believes the proposition
that the proposition is true; one could not offer an equally good explanation

9
  See his discussion of the “crucial counterfactual question” (2011: 73–4).
 Relax? Don’t Do It! 199

of why the mathematician holds the belief that is neutral with respect to the
truth of her belief.10
In short, (1) might be every bit as good of an explanation as (2):
(1) The mathematician believes that p because she proved that p is true.
(2) The mathematician believes that it rained earlier today because she
observed that it rained earlier today.
Of course, given that p is a mathematical truth, we will not be able to
imagine a world that is exactly like ours except for the fact that in that
world p is false. Because of this, the counterfactual “If p had been false,
then the mathematician would not have believed p” will strike us as unin-
telligible, in the way that counterfactuals whose antecedents consist in the
negation of mathematical truths generally strike us as unintelligible. (“If
2+2 had not equaled 4, then . . .”) But that has no tendency to cast doubt
on the truth-invoking explanation of the mathematician’s belief. Indeed,
our inability to evaluate or understand the relevant counterfactual does not
even mean that we cannot investigate or acquire evidence that bears on
the truth-invoking hypothesis as an explanation of why the mathematician
believes as she does. (For example, if we learn from the mathematician’s
diary that she first became convinced of the relevant proposition years ear-
lier, on the basis of a sheer hunch, and that she has believed it unwaver-
ingly ever since, that discovery might very well cast doubt on whether the
truth-invoking hypothesis is really the best explanation of her belief.)
The same point holds for another of Dworkin’s comparisons, one which
is in some respects more closely analogous to the moral case: the case of
the aesthetic. Given that aesthetic properties supervene on non-aesthetic
properties, Dworkin is surely right that we cannot imagine a world which
is exactly like our own except for the fact that The Marriage of Figaro is a
piece of trash. Still, given that we think that it is true that The Marriage of
Figaro is beautiful (as Dworkin does), it certainly seems as though we can

10
  Objection: But can’t we explain why the mathematician believes the proposition by
citing the fact that she takes herself to have proved it? One can take oneself to have proved
a proposition even if one has not actually proved it, and indeed, even if the relevant
proposition is false. So we can explain why the mathematician believes the proposition in
a way that is neutral with respect to whether the proposition is true after all, by citing the
fact that she takes herself to have proved it. Reply: It is a mistake to think that nothing has
been lost by substituting in this weaker, less committal explanation. For we can ask the
following question: what explains why the mathematician takes herself to have proved
the proposition? In a given case, the explanation for why she takes herself to have proved
the proposition might be the fact that she proved the proposition. (Given her mathemati-
cal competence, she would not take herself to have proved the theorem unless she had
actually done so.) So in a sufficiently comprehensive explanation of why she currently
believes the proposition, the fact that she successfully proved the theorem will still appear.
200 Sarah McGrath

ask whether the best explanation of why a particular person believes that it
is beautiful includes her appreciation of its beauty (a truth-invoking expla-
nation), or rather, because certain psychological mechanisms guarantee that
she will end up believing that something is beautiful just in case her friends
think that it is beautiful, regardless of its actual aesthetic merits. In any case,
whether it is ultimately defensible to think that someone’s belief could be
best explained in this way, no reason for doubting that it could is supplied
by the fact that “we cannot imagine a world that is exactly like this one
except that . . . The Marriage of Figaro is trash.”
More generally: nothing about whether some truth could be part of the
best explanation of someone’s believing that truth follows from its modal
status. But this observation shows that Dworkin’s counterfactual criterion
cannot be right, because the unintelligibility of the relevant counterfactu-
als arises from the fact that their antecedents consist in the negations of
necessary truths. In fact, notice that if the orthodox Kripkean view about
metaphysical necessity is correct, then Dworkin’s own account of what
distinguishes the types of beliefs that can be explained by truth-invoking
explanations, and the types of beliefs that cannot, does not even yield con-
sistent results. On the one hand, the belief that water is H2O is a belief about
the physical world, and thus seems eligible to be explained by citing the fact
that water is H2O, together with a story about how scientists recognized this
fact.11 On the other hand, given that water is H2O in our world, we cannot
coherently imagine a world that is exactly like this one except for the fact
that water is not H2O.
I conclude then, that if it is true that the injustice of slavery can play no
role in explaining why people think that it is unjust, this has nothing to do
with the fact that we cannot imagine a world exactly like ours except for the
fact that slavery is not unjust.

The Argument from Pointlessness

Dworkin has another, independent argument for the conclusion that the
truth of a moral view never plays a role in explaining why it is held. The key
idea here is that, when it comes to our moral views, it is inevitably “point-
less” or “otiose” (2011: 74) to appeal to their truth in explaining why we
hold them. He develops this line of thought as follows:
even if we assumed that moral truth does have mysterious causal potency, that
assumption could be of no help whatsoever in justifying our moral beliefs. We
would have to know, independently, whether those beliefs were true before we

11
  Dworkin himself explicitly cites our beliefs about chemistry as paradigms of beliefs
that can be explained by hypotheses that cite their truth (2011: 69).
 Relax? Don’t Do It! 201

could intelligibly cite truth as their parent. That requirement is particularly clear
when you offer to explain someone else’s moral opinions. You think that affirma-
tive action is unfair but your friend thinks it perfectly fair. You cannot think that
his belief is caused by the truth; if you want to explain his belief you must compose
a personal-history explanation. You find one that you think complete and persua-
sive: you cite his education in a knee-jerk liberal family. But now you change your
own mind: you are suddenly convinced by his arguments that affirmative action is
fair. You now think that what your friend believes is true, but you have discovered
nothing that could impeach your earlier explanation of why he believes it. If the
personal-history explanation was adequate before, it remains adequate now. You
may be tempted now to say that, after all, the truth did play a role in the causal story
of how he came to think what he does. But that shows only that [the appeal to truth]
is never more than a fifth, spinning wheel in any explanation. (2011: 74)
Consider first the claim that, in attempting to justify a moral belief, it is
unhelpful to appeal to its truth. There is an obvious sense in which this
is correct. If you and I  disagree about whether affirmative action is fair,
then it would of course be ridiculous for me to attempt to justify my belief
to you by citing its (putative) truth. However, there is no difference here
between our moral beliefs and our scientific beliefs. If the scientific com-
munity is divided about whether some chemical hypothesis is true, then it
would obviously be pointless for those who are already convinced of the
hypothesis to attempt to justify their belief by citing the (alleged) fact that
it is true. But this uncontroversial point about justification has no tendency
to show that (e.g.) the best explanation of why chemists believe that water
is H2O does not invoke the fact that water is H2O. The same point holds in
the moral domain. When William Wilberforce took to the floor of parlia-
ment to attempt to justify his belief that slavery is unjust, it would obvi-
ously have been pointless for him to cite the fact that slavery is unjust. But
this uncontroversial point about justification does nothing to show that
the injustice of slavery plays no role in the explanation of why some people
believe that it is unjust.
Consider next Dworkin’s discussion of your attempts to explain your
friend’s belief that affirmative action is fair. In Dworkin’s example, you orig-
inally accept a “personal-history explanation” of your friend’s belief, accord-
ing to which
(H1) My friend believes that affirmative action is fair only12 because
he grew up in a knee-jerk liberal family.

12
  Recall that in Dworkin’s example, you originally regard this as a “complete” expla-
nation of your friend’s belief.
202 Sarah McGrath

At a later time, you become convinced by your friend’s arguments that

affirmative action is fair. At this point, it’s natural for you to consider
another explanation of why your friend believes as he does:
(H2) My friend believes that affirmative action is fair because he pos-
sesses sound arguments for that conclusion.
Notice that H2, unlike H1, is a truth-invoking explanation: if your friend
possesses sound arguments for the conclusion that affirmative action is fair,
then it is true that affirmative action is fair. For this reason, the hypothesis
would not be considered a live option by anyone who (like your past self in
Dworkin’s example) is of the opinion that affirmative action is unfair.
Dworkin suggests that, when you change your mind about affirmative
action and the quality of your friend’s arguments, you should regard this as
irrelevant to the acceptability of the explanatory hypothesis that you have
accepted up until now, i.e. H1. (“You have discovered nothing that could
impeach your earlier explanation . . . if the personal-history explanation was
adequate before, it remains adequate now.”) But this claim neglects the way
in which the rational credibility of an explanatory hypothesis can depend on
which alternative hypotheses are taken to be live options. As philosophers
of science often note, an explanatory hypothesis can have its credibility dra-
matically boosted when a formidable competing hypothesis is eliminated
from consideration. In such cases, the credibility of the hypothesis increases
when the field of alternative hypotheses shrinks. Of course, the opposite
can also occur: the credibility of a currently accepted hypothesis might be
dramatically reduced when a new hypothesis is introduced, or when a previ-
ously rejected hypothesis is reinstated as a live option. Dworkin’s example
is a potential illustration of this last possibility. Back when you thought
that affirmative action is unfair and your friend’s arguments in its favor
were unsound, the truth-invoking explanation H2 was not among the live
options, because it is inconsistent with what you then took to be the case.
After you change your views in the relevant ways, H2 is no longer ruled out
by what you believe, so it can reenter the competition. Contrary to what
Dworkin suggests, this might very well affect whether it is rational for you
to believe that H1 is the actual explanation of why your friend believes as
he does.
What would it take to show that an appeal to truth is never more than
a “fifth, spinning wheel” in explaining our moral beliefs? Simply this: that
whenever we are tempted to accept a truth-invoking hypothesis of why
someone accepts a moral belief, there is always some superior “personal his-
tory explanation” that does not invoke truth. (Compare the task of showing
that there are no cases in which the best explanation of why someone holds
a mathematical belief invokes the truth of that belief.) But contrary to what
 Relax? Don’t Do It! 203

Dworkin seems to think, there is nothing in the passage quoted above that
bears on the possibility of successfully executing this project.
Appreciating the way in which the credibility of an explanatory hypothesis
can vary depending on which alternatives are considered live options puts us
in a position to answer a question that is made salient by Dworkin’s discus-
sion. The question is this: given that (as we have admitted) one could never
justify a moral belief by citing its truth, why should a moral realist care whether
the truth of a moral belief could ever play a role in explaining why it is held?
Here is a natural answer to this question. Presumably, something explains why
we hold the moral views that we do: the fact that one holds a moral belief is
never simply a brute fact. If there is some general reason for thinking that
the truth of a moral belief could never figure in the best explanation of why
it is held, then for any particular moral belief, any truth-invoking hypothesis
should not be considered a live option. And it might very well be that, once
all truth-invoking hypotheses are removed from the field, the most credible
remaining hypothesis about why one believes as one does is an explanation
which has the following property: if accepted, it would give one a reason to
abandon the moral belief that it explains.
Compare the empirical case:  if an oracle informed us that the correct
explanation of why many scientists believe that global warming is occurring
has nothing to do with the occurrence of global warming, thus eliminating
any truth-invoking explanation as a live option, it might be that the next
best explanation (the one that it would then be reasonable for us to accept)
would be one that undermines the rationality of continuing to believe in
global warming. Thus, the reason why a moral realist who acknowledges
that one could never justify a moral belief by citing its truth might never-
theless care about whether moral truth is the kind of thing that could play
an explanatory role is simply this: if all of the truth-invoking-hypotheses are
eliminated from consideration, it might very well be that the explanations
of her moral beliefs that it would then be reasonable for her to accept are
ones that undermine the rationality of her continuing to hold those beliefs.
More generally, the elimination of all-truth-invoking hypotheses might
leave debunking explanations as the most reasonable explanations of our
moral beliefs left standing.
I take it that this is a natural line of thought. Dworkin, however, would
regard it as resting on a fundamental confusion about the vulnerability of
our moral beliefs. For Dworkin holds that our moral views are not suscept­
ible to being undermined by purely psychological hypotheses about why we
hold them. The same view has been advanced by Thomas Nagel in The Last
Word.13 This is the final issue that I will take up.

13
  This is a central claim of ch. 6 of that work.
204 Sarah McGrath

Could Our Moral Views be Undermined by Non-moral

Considerations?
As noted, Dworkin holds that moral beliefs differ from ordinary empirical
beliefs in at least two significant respects. First, the fact that a given moral
belief is true never appears in the explanation of why it is held. Second,
the general lack of explanatory connection between moral truth and moral
belief does not cast any doubt on moral beliefs.14 Let’s turn to the second
of these two claims.
Given that (by Dworkin’s own lights) the absence of an explanatory con-
nection between a belief ’s truth and its being held tends to undermine the
credibility of an ordinary empirical belief, why wouldn’t the same hold for
our moral beliefs? For Dworkin, this difference between ordinary empiri-
cal beliefs and moral beliefs is a consequence of an even more fundamental
difference between them. In the case of an ordinary empirical belief, infor-
mation about its causal etiology can undermine its credibility. (When I dis-
cover that I only believe what I do about the weather because I have been
hypnotized to hold those beliefs, this tends to undermine their credibility.)
However, Dworkin holds that simply learning information about the causal
etiology of a moral belief is not sufficient to undermine its credibility in the
same way. This is because the undermining of a moral belief is something
that always requires moral considerations: merely learning purely descriptive
information about the causal history that accounts for why the belief is cur-
rently held is never enough to do the job on its own.15
This is a remarkable view. Imagine that the hypnotist has found altering
my beliefs about the weather so enjoyable that he decides to have some
additional fun with my moral beliefs. At some later time, I am presented
with a complete list of the things that I have been hypnotized to believe;
some of these beliefs concern the weather, others concern moral issues.
When I discover that one of my beliefs about the weather is on the list, this
undermines the credibility of that belief. But when I find that one of my

14
  Dworkin helpfully locates himself in dialectical space by noting that the first view
is one that he shares with a certain kind of “external skeptic” about morality, but that he
and the skeptic part ways with respect to the second view (2011: 70).
15
  “[A]‌ny argument that either supports or undermines a moral claim must include or
presuppose further moral claims or assumptions” (2011: 100). This is a major theme of
both Dworkin’s (1996) and part I of his (2011). But for applications particularly relevant
to the following discussion, see esp. 1996: 123–9 (arguing that even if one learnt that
one’s views about distributive justice are inevitably determined by one’s self-interest, that
would not give one a reason to doubt those views) and 2011: 77–9 (arguing that, even if
one learnt that one’s view about the fairness of affirmative action is due to the side effects
of a brain scan, that would not give one a reason to doubt that view). I discuss the latter
case below.
 Relax? Don’t Do It! 205

moral beliefs is on the list, this discovery does not similarly undermine its
credibility. For anything that could undermine the credibility of my moral
conviction would itself have to be or include a moral consideration, and the
fact that I hold this belief because I have been hypnotized is not a moral
consideration but rather a fact about my psychology.
Consider also traditional “debunking” explanations of our moral beliefs,
of the kind offered by Nietzsche, Marx, and Freud.16 These accounts of why
we hold the moral beliefs that we do are not themselves moral or normative
or evaluative claims; rather, they purport to be purely descriptive, natu-
ralistic, causal-explanatory accounts of why we end up holding the moral
convictions that we do. One might think that these causal-explanatory
hypotheses are implausible, or that we lack good evidence that they are
true. However, if it’s true that our moral views are not susceptible to being
undermined by the provision of purely non-moral information, then even
if we knew with certainty that one of these accounts was correct, this would
have no tendency to undermine the credibility of our moral convictions.
Perhaps there are other possible explanations of our moral convictions
that, if known to be true, would seem to cast those convictions in an even
worse light than the stories offered by Nietzsche, Marx, and Freud. But if
Dworkin is correct, then we have a kind of a priori guarantee that nothing
that we discover empirically could count as a good reason to lose confidence
in our moral convictions, so long as what is discovered is not itself a moral
consideration.
Why does Dworkin think that our moral convictions could not be
rationally undermined by empirical discoveries? Ironically, Dworkin thinks
that the relative immunity of our moral beliefs is underwritten by the truth
of (what he calls) Hume’s principle, according to which “no amount of
empirical discovery about the state of the world . . . can entail any conclu-
sion about what ought to be without a further premise about what ought to
be” (2011: 17). As Dworkin notes, Hume’s principle has traditionally been
thought to bolster the case for moral skepticism, but he contends that in
fact, it undermines the case for moral skepticism.17 His thought seems to be
something like this: rationally undermining a moral conviction requires an
argument that targets that conviction, but Hume’s principle guarantees that
the content of moral and non-moral claims is sufficiently distinct that any
such argument whose premises consist of non-moral claims will inevitably
fail to make contact with its intended target. Therefore, the discovery of

16
  For an overview, see Leiter (2004).
17
  On Hume’s principle and its anti-skeptical thrust, see esp. Dworkin (2011: 44–6). For
an excellent discussion of Dworkin’s use of Hume’s principle, see Shafer-Landau (2010).
Smith (2010) provides a useful critique of Dworkin’s conception of moral skepticism.
206 Sarah McGrath

purely non-moral information could never furnish the premises of a suc-

cessful skeptical argument. I will return to this use of Hume’s principle.
If Dworkin is correct in thinking that our moral beliefs are in principle
immune to having their credibility undermined by discoveries about their
causal etiology, then this would indeed be a significant respect in which such
beliefs differ from beliefs about the physical world. And it would certainly
provide a reason to relax about morality, or at least, to be less anxious about
its standing and status. I’ll argue, however, that Dworkin is not correct, and
that there is no significant difference here between beliefs about the physi-
cal world and moral beliefs: both are susceptible to being undermined by
purely descriptive information about their causal etiology.
In both Justice for Hedgehogs and “Objectivity and Truth,” Dworkin pro-
ceeds in the following way. He describes a hypothetical scenario in which
you learn surprising information about why you hold a certain moral belief.
He suggests (correctly, I think) that there is a strong temptation to think
that, in the relevant circumstances, you should lose confidence in the moral
belief, and that his cases are in relevant respects representative of other cases
that inspire the same intuitive reaction. But he argues that, even in these
cases, it is a mistake to think that the mere knowledge of the causal etiology
of your belief gives you a good reason to doubt it.
I concede to Dworkin that if his treatment of these cases is sound, then
his more general claims about the relative immunity of moral beliefs to
empirical undermining are plausible. I’ll argue, however, that his treatment
does not withstand scrutiny, and that in fact, the correct lessons to draw
from these cases contradict his more general claims. I’ll focus my attention
on his more recent discussion, although I believe that all of the same points
apply, mutatis mutandis, to his earlier one.
Here is Dworkin’s description of the case:

Until a year ago you thought affirmative action patently unfair. Then you had occa-
sion to think about the matter again and were convinced, by arguments that sud-
denly seemed compelling, that it is not unfair. One Tuesday morning you read, in
the Science section of your newspaper, of an astounding discovery. Everyone in the
world who has had a scalotopic brain scan (don’t ask) thinks that affirmative action
is fair, whatever opinion he held before the scan. The evidence is vast and conclu-
sive: there is no possibility of coincidence. You had a scalotopic scan shortly before
you rethought and changed your views, and you are left in no doubt that you would
not have changed them if you had not had the scan. (2011: 77)

In Dworkin’s continued telling of the story, you respond to this discovery

by subjecting your newfound belief that affirmative action is fair to an unu-
sual level of critical scrutiny. You reconsider the arguments that you take
to justify your belief and find that they still strike you as compelling. You
 Relax? Don’t Do It! 207

consider the belief in light of other things that you believe about related
matters. You expect to find conflict, but instead find that your other beliefs
support your newfound conviction. In fact, it is your old opinion, that
affirmative action is not fair, that now fits poorly with the rest of what you
believe. You attribute this to the following:
The effect of the scan, you now assume, was more general and pervasive than you
had thought: it affected widespread shifts throughout the full range of your moral
convictions so that all your convictions are now thoroughly integrated with your
new views about affirmative action. No matter how you test them, they all seem
right to you. (2011: 78)

Dworkin’s verdict about the case is uncompromising:

How should you react when you finish being bewildered? Surely your discovery
should have some impact on either your opinions or your confidence in those opin-
ions. But in fact it can have no impact whatsoever. (2011: 78)

That is, you should be no less confident of your belief that affirmative action
is fair after finding out about the brain scan and its effects on your thinking
than you were before. Central to Dworkin’s case for this uncompromising
verdict is his account of how you should view the effects of the brain scan
on your thinking, as someone who currently believes that affirmative action
is fair.
. . . you cannot regret having had the scan, at least not for this reason [i.e. a con-
cern that the process has replaced a true belief about affirmative action with a false
one]. You have no reason whatever to think you were right before . . . Before the scan
you would have had a very strong reason not to have the scan if its results could
have been predicted. But now you have the same reason for not regretting the scan;
indeed, for thinking yourself fortunate to have had it.
Do you have less reason to suppose your new views correct than you had to think
your old ones correct before the scan? No; on the contrary, you now think you
have more reason than you had then because you now think your earlier reasons
were unsound. Should you now doubt your ability to form any responsible judg-
ment at all on the question of affirmative action? No, because you cannot reject the
hypothesis that the brain scan improved your ability to reason about morality. On
the contrary, you have some evidence that it did so: you were in error about many
moral matters before the scan but are now reasoning better, or so you cannot help
but think. (2011: 78–9)

One thing that should make us immediately suspicious of this line of

argument is the following. If it really would be legitimate for you to rea-
son in the ways endorsed by Dworkin in the envisaged circumstances, then
it seems like parallel reasoning could render our beliefs about the physi-
cal world immune to being undermined by discoveries about their causal
208 Sarah McGrath

etiology—something that Dworkin himself agrees is perfectly possible.

Suppose that I discover that I currently believe that it rained earlier today
because I was hypnotized to think this. Of course, given that I currently
believe that it rained earlier today, I  cannot consistently view my being
hypnotized as a process that resulted in my acquiring a false belief. On the
contrary, to the extent that I make use of my belief that it rained earlier
today in evaluating that process, I will view it as one that resulted in my
acquiring a true belief. In Dworkin’s brain scan case, the scan results not
only in a reversal in my original opinion about affirmative action, but also
in widespread shifts in my views about related matters, so that my new view
about affirmative action is thoroughly integrated with the rest of my beliefs.
In order to make the cases parallel, we should imagine that the hypnotist
erases any beliefs I once held that suggest that it did not rain earlier today,
and replaces them with beliefs that cohere well with the belief that it did.
(For example, perhaps he induces in me the belief that there are puddles of
rainwater on the streets outside, puddles that would not exist in the absence
of earlier rain.) I will then be in a position to offer arguments in support of
my belief that it rained today, arguments that are valid and which proceed
from premises that I believe to be true. However, so long as I know that
I only believe the supporting premises because of the hypnotist, it would be
perverse for me to think that these arguments provide me with good reasons
to think that it rained earlier today, or with reasons for thinking that my
encounter with the hypnotist must have been a stroke of good epistemic
luck. Similarly, so long as you know that the other beliefs against which
you test your newfound belief about affirmative action are side effects of the
brain scan, it would be perverse for you to use this as evidence that the brain
scan must have been an epistemically beneficial event, on the grounds that
you are now thinking so much better about the topic of affirmative action
than you were before it occurred.
But perhaps there is some way of pulling apart the two cases. So let’s set
aside the comparison with the empirical case, and focus exclusively on the
brain scan case. Suppose that we extend Dworkin’s fiction in the following
way. As part of your continuing medical treatment, you must return to the
hospital for another, follow-up scan. This brain scan differs from the first,
although it too has been discovered to have an astonishing side effect: any-
one who undergoes it ends up believing that affirmative action is unfair
(regardless of what they thought about the issue previously). Moreover,
anyone who undergoes this second scan also ends up with beliefs about
surrounding topics that tend to support that belief. That is, this scan, like
the first, will bring about “widespread shifts throughout the full range of
your moral convictions so that all your convictions are now thoroughly
integrated with your new view about affirmative action.” Of course, if you
 Relax? Don’t Do It! 209

assess the costs and benefits of this second procedure from your current
perspective, as someone who firmly believes that affirmative action is fair,
you will view it as something that will leave you worse off epistemically. But
these unfortunate epistemic side effects seem like a small price to pay, given
the severity of the medical condition that makes this unusual line of treat-
ment necessary in the first place.
Of course, after you have had the second brain scan, everything will look
quite different, given what you will then believe about affirmative action
and related topics. Employing the kind of reasoning endorsed by Dworkin,
you will be in a position to conclude that it was actually the second scan that
was the epistemically fortuitous procedure, and the first that was epistemi-
cally harmful. You will be able to subject your recently recovered belief that
affirmative action is unfair to the same kinds of tests as those mentioned
above, tests which the belief will pass with flying colors. If Dworkin is right
to think that sufficiently extensive scrutiny of the relevant kind is sufficient
to justify maintaining a moral conviction with undiminished confidence,
then you will be justified in maintaining your new belief in this way, not-
withstanding your knowledge that you would be in exactly the opposite
position if you had not consented to the most recent scan.
We can even imagine a series of such scans: perhaps every morning you
have the scan which results in people’s believing that affirmative action is
unfair, and every afternoon you have the scan which leads people to believe
the opposite. At any point in time, you can subject what you then believe
about affirmative action to the kind of critical examination suggested by
Dworkin, thoroughly scrutinizing it against your other moral convictions
in order to make sure that it coheres well with them. (It always does.) Of
course, you know full well that a few hours ago your slightly earlier self
thought otherwise, and that a few hours from now your slightly later self
will also vehemently disagree with your current verdict. But why should you
be troubled by what either of them think, given how benighted their views
about affirmative action and related topics really are? Not only are your
slightly later and slightly earlier selves not your epistemic peers; the fact that
they think as they do gives you no reason at all to be any less confident of
what you now think. You thus retain your current view with undiminished
confidence, for the same reasons that Dworkin thinks that your knowledge
of the first scan should make no difference to your confidence that affirma-
tive action is fair.
But that would be absurd. If you actually found yourself holding a belief
about affirmative action in these circumstances, then the rational thing for
you to do would be to abandon that belief. If you were unable to abandon
it, then you should regard your inability to do so as a failure of rationality,
and not make use of the belief in your practical or theoretical reasoning.
210 Sarah McGrath

(For example, you should not use it to reason about which scans were epis-
temically fortuitous and which were epistemically beneficial.) Of course,
you should not regard this information about the causal etiology of your
current belief as a reason to adopt the opposite belief about affirmative
action. Rather, the rational stance is to hold no view about the fairness of
affirmative action, on the grounds that it is not something about which you
are in a position to have a reliable opinion.
The fact that your knowledge of the causal etiology of the belief does
not give you a reason to adopt the opposite belief is significant, for it sug-
gests that this knowledge serves as an undercutting as opposed to a rebutting
defeater for your current belief.18 Intuitively, a rebutting defeater for one’s
belief that p undermines one’s justification for believing p by providing a
reason to believe not-p. In contrast, an undercutting defeater undermines
one’s justification for believing p without providing a reason to believe
not-p.19 I believe that this observation generalizes: debunking explanations
of our moral beliefs are best interpreted as attempts to provide undercut-
ting, rather than rebutting defeaters. For example, someone who suggests
that once we fully understand the evolutionary explanation for why our
moral beliefs strike us as correct, we will have good reason to abandon those
beliefs, is not suggesting that we should come to hold the opposite moral
beliefs. Rather, she is suggesting that we should hold no moral beliefs at all.
Recall Dworkin’s idea that Hume’s principle ensures that our moral beliefs
could not be undermined by purely non-moral discoveries. Once we take
on board the distinction between rebutting and undercutting defeaters,
we are in a position to appreciate why Hume’s principle, even assuming
for the sake of argument that it is true, has at best limited anti-skeptical
import. For Hume’s principle is relevant only if we are concerned with chal-
lenges to our moral beliefs that purport to provide rebutting as opposed to
undercutting defeaters for those beliefs. If I initially believe that affirmative
action is unfair, then one way you can undermine my belief is by giving me
compelling reason to believe that affirmative action is fair; if you succeed
in doing this, then you have succeeded in providing me with a rebutting
defeater for my original belief. Plausibly, you can only provide me with
compelling reason to believe that affirmative action is fair by appealing to

18
  On undercutting vs. rebutting defeaters, see especially Pollock and Cruz (1999).
19
  A stock example that epistemologists use to illustrate the distinction:  you are in
a house that has some red walls and some white walls. You see a wall that looks red to
you, and the fact that it has this appearance justifies you in believing that it’s red. You
subsequently learn that the wall has a red light shining on it. This new information is
an undercutting defeater for your justified belief that the wall is red, since you are not
in a position to rule out that the wall is really white. But the new information is not a
rebutting defeater, since it does not give you a reason to believe that the wall is not red.
 Relax? Don’t Do It! 211

moral considerations. And this seems generally true: in order to provide a

rebutting defeater for a moral belief, one must appeal to moral considera-
tions. Providing purely non-moral information will not be sufficient, for the
non-moral information will not warrant taking up the opposite moral con-
clusion. Thus, it’s at least plausible that Hume’s principle does entail that
this kind of undermining cannot be a matter of purely empirical, non-moral
discoveries concerning (e.g.) the origins of our moral convictions, or why
those convictions currently strike us as correct.
But not all defeaters are rebutting defeaters. There is a second way in
which you might undermine my belief that affirmative action is unfair: by
providing me with information that suggests that I am not in a position to
form a reliable opinion about affirmative action one way or the other. To
the extent that you succeed in doing this, you have provided me with an
undercutting defeater for my belief. However, precisely because you do not
need to convince me of some moral conclusion in order to do this, there
is no requirement that the considerations that you offer have to include
any moral considerations. (For example, it would be enough to show that
I arrived at my opinion under the influence of a pill that severely impairs
my ability to think about complicated issues.) The point is a general one
about undercutting defeaters: a successful undercutting defeater need not
stand in any logical relation to the content of the belief that it undermines.20
That is why non-moral considerations can serve as undercutting defeaters
for moral beliefs, even if Hume’s principle is true.
In sum: debunking explanations of our moral beliefs are best understood
as attempts to provide undercutting as opposed to rebutting defeaters for
our moral beliefs, but Hume’s principle only poses an obstacle for rebut-
ting as opposed to undercutting defeaters of our moral beliefs. Therefore,
Hume’s principle is not relevant to traditional debunking explanations of
our moral beliefs.

20
  For example, suppose that I  perform some non-trivial mathematical calculation
and arrive at a particular number for an answer. Because I  know that I  am generally
competent when it comes to calculations of the relevant kind, I am rationally confident
in believing a certain mathematical proposition. I then learn that while performing the
calculation I was under the influence of a drug that interferes with my ability to think
coherently, albeit in imperceptible ways. The information that I performed the calculation
under the influence of a mind-altering drug undermines my belief in the relevant math-
ematical proposition, although it is not itself a mathematical proposition.
The possibility of such cases shows what is wrong with Nagel’s remark that “some-
one who abandons or qualifies his basic methods of moral reasoning on historical or
anthropological grounds alone is nearly as irrational as someone who abandons a math-
ematical belief on other than mathematical grounds” (1997: 105). One can have good
non-mathematical grounds for abandoning a mathematical belief, and the way in which
this is possible is suggestive of how historical or anthropological findings could in princi-
ple give one a reason to abandon or qualify one’s methods of moral reasoning.
212 Sarah McGrath

4. Conclusion

Relaxed realists hold that there are deep differences between moral truths
and the truths studied by the empirical sciences, but they deny that these
differences raise troubling metaphysical or epistemological questions about
moral truths. On this view, although features such as causal inefficacy,
perceptual inaccessibility, and failure to figure in any of the best explana-
tions of our empirical beliefs would raise pressing skeptical concerns were
they claimed to characterize some aspect of physical reality, the fact that
these features characterize the normative domain is not a good reason to
have skeptical doubts about it. To suppose otherwise is in effect to apply
standards and criteria of evaluation that are appropriate for one domain
to another, quite different domain. From this perspective, a great deal of
the metaphysical and epistemological theorizing that drives contemporary
metaethics—for example, the project of “locating” ethics with respect to a
scientific account of the world (cf. Jackson 1998)—can look like a kind of
category mistake, one that arises from looking upon the normative domain
as though it were a kind of extension of the physical world.21
My main goal in this chapter has been to put pressure on this picture.
Relaxed realists often compare morality and mathematics, but I have argued
that, when judged by the criteria of causal efficacy, perceptual accessibility,
and indispensability in explaining what we observe, mathematical proper-
ties as a class pass all three tests. To the extent that morality resembles math-
ematics in relevant respects, we should expect (some) moral properties to be
causally efficacious, perceptually accessible, and indispensable in explaining
what we observe as well. If it is claimed that no moral properties have such
features, then the analogy breaks down, and mathematics becomes unavail-
able as a useful example for fending off skeptical worries about the moral
realm. Moreover, the fact that some mathematical properties have these
features refutes the idea that questions about causal efficacy, explanatory
indispensability, and perceptual accessibility are only properly posed when
we are dealing with natural or broadly physical properties.
Consideration of Dworkin’s arguments revealed other ways in which it
is reasonable for us to expect moral truths to resemble ordinary empirical

21
  Cf. Scanlon’s critique (2014: 16–30) of the idea that Mackie-style concerns about
objective values as things “utterly different from anything else in the universe” succeed
in raising a genuine ontological issue about what there is. Dworkin dismisses both “the
project of reconciling the moral and the natural worlds” and “the project of aligning the
‘practical’ perspective we take when living our lives with the theoretical perspective from
which we study ourselves as part of nature” as “entirely bogus philosophical projects”
(2011: 9).
 Relax? Don’t Do It! 213

truths. In the case of an ordinary empirical belief, it is always a fair question

to ask whether part of the explanation of why that belief is held is that it is
true, and no compelling reason emerged to think that the same question is
suddenly out of place when it is asked of a moral belief. Our moral beliefs, like
ordinary empirical beliefs, are susceptible to being debunked if the right kind
of evidence emerges.
I will close with a suggestion about where I think the moral realist should go
from here. My own view is that realists should face challenges to our moral views
head on, whether those challenges consist of empirically motivated attempts
at debunking, or more abstract metaphysical and epistemological objections.
With respect to empirically motivated debunking arguments, realists should
agree that, if certain possible empirical discoveries were made about why our
moral convictions strike us as true, then we should lose confidence in those
convictions. But they should insist that such evidence actually be provided, as
opposed to merely gestured at. Similarly, realists should not concede that (e.g.)
the injustice of slavery played no role in the historical rise of the relevant belief,
in the absence of a full and compelling explanation that makes no reference to
its injustice.22 More generally, I believe that moral realists—who, after all, are
already committed to believing in objective moral facts—should be cautiously
optimistic that once we abandon overly restrictive models of causation, explana-
tion, and the contents of perception, it will be plausible to claim vindication for
morality on some or even all of these fronts. In addition to the significant efforts
that have already been made in this direction, I think that this is a juncture at
which the moral realist can find encouragement in the example of mathemat-
ics. The idea that we never perceive (instantiated) mathematical properties, or
that such properties are never causally efficacious, or play no role in explaining
why we hold the beliefs that we do, often depends, I think, on relatively crude
pictures of what would be involved in such transactions.23 Once such pictures
are abandoned, it becomes extremely natural to think that the best explana-
tion of why 2+2=4 is as widely believed as it is entails that 2+2=4, and that no
explanation that is neutral as to whether 2+2=4 could possibly rival it. Perhaps
something similar is true with respect to the belief that slavery is unjust.24

22
  Readiness to concede in advance that the best explanation will make no reference
to the injustice of slavery seems particularly inappropriate when there is as of yet no
informative, generally agreed upon account of the factors that make one explanation
“better” than another among theorists of explanation.
23
  For example, in an endnote (2011: 443), Dworkin rejects the claim that the correct
explanation of someone’s believing that 7+5=12 could invoke the truth of that belief on
the grounds that “seven and five do not cause people to think that together they make
twelve.”
24
 Earlier versions of this chapter were presented at Fordham University, SUNY
Buffalo, meetings of both the Young Ethicists’ Network and Paper Tigers at Princeton
214 Sarah McGrath

References
Baker, Alan. 2005. “Are there Genuine Mathematical Explanations of Physical
Phenomena?” Mind, 114(454): 223–38.
Dworkin, Ronald. 1996. “Objectivity and Truth: You’d Better Believe it,” Philosophy
and Public Affairs, 25(2): 87–139.
Dworkin, Ronald. 2011. Justice for Hedgehogs. Cambridge, MA:  Harvard
University Press.
Garfinkel, Alan. 1981. Forms of Explanation. New Haven: Yale University Press.
Harman, Gilbert. 1977. The Nature of Morality. Oxford: Oxford University Press.
Jackson, Frank. 1998. From Metaphysics to Ethics. Oxford: Oxford University Press.
Kim, Jaegwon. 1981. “The Role of Perception in A Priori Knowledge,” Philosophical
Studies, 40(3): 339–54.
Leiter, Brian. 2004. “The Hermeneutics of Suspicion:  Recovering Marx,
Nietzsche, and Freud,” in B.  Leiter (ed.), The Future for Philosophy, 74–105.
Oxford: Clarendon Press.
Lewis, David. 1986. “Causal Explanation,” Philosophical Papers, ii. 214–40. Oxford:
Oxford University Press.
Lipton, Peter. 1991. Inference to the Best Explanation. London: Routledge.
Nagel, Thomas. 1997. The Last Word. Oxford: Oxford University Press.
Parfit, Derek. 2011. On What Matters. Oxford: Oxford University Press.
Pollock, John, and Cruz, Joseph. 1999. Contemporary Theories of Knowledge.
Lanham, MD: Rowman & Littlefield.
Putnam, Hilary. 1975. “Philosophy and our Mental Life,” Mind, Language, and
Reality: Philosophical Papers, ii. 291–303. Cambridge: Cambridge University Press.
Scanlon, Thomas. 1998. What We Owe to Each Other. Cambridge, MA: Harvard
University Press.
Scanlon, Thomas. 2014. Being Realistic about Reasons. Oxford:  Oxford
University Press.
Shafer-Landau, Russ. 2010. “The Possibility of Metaethics,” Boston University Law
Review, 90: 101–17.
Smith, Michael. 2010. “Dworkin on External Skepticism,” Boston University Law
Review, 90: 509–20.
Sober, Elliott. 2000. “Quine’s Two Dogmas,” Proceedings of the Aristotelian Society,
74: 237–80.

University, and at the 9th Annual Wisconsin Metaethics Workshop; I am grateful to the
audiences present on those occasions for their feedback. Special thanks to Niko Kolodny
and Tamar Schapiro for serving as my commentators in Princeton, and to Jamie Dreier,
Billy Dunaway, Elizabeth Harman, Frank Jackson, Mark Johnston, Thomas Kelly, and
two anonymous referees for Oxford Studies in Metaethics.
 9
Wrong Kinds of Reason and the Opacity
of Normative Force
Justin D’Arms and Daniel Jacobson

1.  Right and Wrong Kinds of Reason

Roughly speaking:  A  reason is said to be “of the wrong kind” when,

although it counts as a consideration broadly in favor of (or against) having
an evaluative attitude, it does not bear on whether the object is valuable in
the respect relevant to that attitude.1 To say that something is a wrong kind
of reason (WKR), however, is not to say that it is a bad reason. Some WKRs
seem to provide excellent reasons to desire something—or to have an emo-
tion such as amusement, admiration, fear, and the like toward it.2 The
central case in the recent literature concerns what we will call a demonic
incentive: a demon credibly threatens to punish you unless you desire or
admire something that, quite obviously, isn’t desirable or admirable. When
the incentive is compelling enough, it apparently provides conclusive reason
to desire x or admire y, if you can, but not a reason that shows x to be good
or y admirable.
Just as WKRs are not always bad reasons, bad reasons are not always,
or even typically, WKRs. The distinction between reasons of the right and

1
  This is rough because a given consideration can have both kinds of normative force,
so strictly this is true of mere WKRs. We will hereafter drop the “or against” qualifier,
except when needed for clarity, but all our claims are meant to apply to both positive and
negative cases.
2
  As we’ll see, this claim proves contentious but not in a way that calls into question
the goodness of certain WKRs. Rather, what is contentious is whether wrong kinds of
reason are really reasons to desire or admire (etc.), or rather reasons to want or try to have
such states. Though we will continue to speak of WKRs, our argument is consistent with
such WKR skepticism, mutatis mutandis. We discuss this reason redescription program
in more detail in §3.
216 Justin D’Arms and Daniel Jacobson

wrong kind is orthogonal to the distinction between good and bad reasons.
If Giuseppe admired Mussolini for abolishing all the political parties in Italy
save the Fascists, then Giuseppe was mistaken about what makes people
admirable (according to us and, we presume, to you). But that is not to
say that he was taken in by a wrong kind of reason, by his own lights or by
ours. We can suppose that Mussolini’s decisiveness in pursuit of his political
aims was a right kind of reason (RKR) by Giuseppe’s lights, in that the role
it played in his psychology was to make Mussolini seem admirable to him.
To the rest of us, this is no reason at all for admiring Mussolini; it’s not a
wrong kind of reason, which would be a (putatively) good reason to admire
him that does not bear on his admirability.3
It should be obvious that demonic incentives can only function as WKRs,
and that Mussolini’s decisive action functions as an RKR for Giuseppe even
though it is a substantively bad reason. However, in other cases it is harder to
tell what kind of reason a consideration affords. Take the fact that some trait
of yours—your lame leg, for instance—is something that you are stuck with
through no fault of your own. This seems to count in some way against feel-
ing ashamed of it. But is this an RKR against shame, which demonstrates that
the trait is not really shameful? One might think so: that because you are not
responsible for the occurrence or persistence of your impairment, it cannot
reflect badly on you in the way it would need to in order to be shameful. But
most accounts of shame suggest otherwise: it is often claimed to be a crucial
difference between guilt and shame that only the former must impute respon-
sibility or blameworthiness.4 Certainly shame is very commonly felt at physi-
cal and mental incapacities, especially conspicuous ones, that make a person
unable to do what others can. This is some ground for thinking that shame-
fulness does indeed attach to such inabilities, and not only to traits for which
people are somehow responsible—even if decent people try to avoid saying so.5
Nevertheless, there are some good reasons not to be ashamed of traits
that you did not cause and cannot change, even if all this is true. Such
shame seems to be painful and useless. Perhaps the best way to respond to

3
  Some subjectivists may wish to say that it was a good reason for Giuseppe to admire
him. Some of what we say in what follows can be accepted by subjectivists, but perhaps
not all of it; the relationship between subjectivism and fitting attitude theories of value
is complicated but tangential to our main purpose, and we will not address it here. See
Blackman (MS).
4
  See e.g. Baumeister et al. (1994, 1995); Gibbard (1990); Taylor (1985): esp. pp. 61,
91), although we disagree with Taylor about the details of how responsibility figures in
guilt but not shame.
5
  Indeed, we expect some readers to balk at our mere use of the word lame—which,
they will feel, is insensitive even when literally accurate. We use it nonetheless, because
that understandable impulse illustrates one of the themes of this chapter: that some legiti-
mate moral qualms do not call into question the truth of a predication.
 Wrong Kinds of Reason 217

impairments over which we lack any control is to accept them with seren-
ity, and to focus our self-regarding emotions on things we either like or can
change about ourselves.6 This way of thinking treats the fact that you did not
cause and cannot alter your lameness as a WKR: a reason why it would be
better not to be ashamed, rather than a reason why the trait is not shameful.
These reflections are not intended to settle what sort of reason is provided
by the fact that your lame leg is no fault of your own. To the contrary, our
point is that the answer to that question is far from obvious, and the ques-
tion of what relevance this sort of consideration has to shame turns out
to be controversial. In certain cases, the very same consideration will be
thought a WKR by some and an RKR by others.7 But the claim that this is
a good reason not to be ashamed seems much less controversial. Someone
who advances such a consideration may be confident of its importance
while uncertain or even confused about just what kind of normative force to
assign it. We suspect that this is common. While it is easy to think that this
consideration counts against shame, few people will have thought carefully
about whether (and why) it counts against shamefulness as well.
The relevance of fault to shame is just one example of a widespread phe-
nomenon. It is one thing to recognize that some consideration bears on
the justification of an evaluative attitude. It is another thing to understand
how it so bears: just what kind of reason it is. This phenomenon we will
refer to as the opacity of normative force, the phrase we’ll use for a range
of cases where an agent takes some consideration to count in favor of an
evaluative attitude while being unsure, or somehow mistaken, about how
the consideration justifies the attitude. The central goal of this chapter is to
explicate some important problems arising from the opacity of normative
force, which are not considered in the WKR literature.
There is now an extensive philosophical literature on right and wrong
kinds of reason. This literature was originally focused on solving a tech-
nical problem for Fitting Attitude (FA) theories of value, which became
known as the wrong kind of reason problem.8 The problem, in short, is how

6
  Although it has become clichéd, the Serenity Prayer was written not by a Hallmark
copywriter but the theologian Reinhold Niebuhr; and its popularity speaks to its wide-
spread appeal. It begins: “God, give me grace to accept with serenity the things that can-
not be changed . . .” Surely shame is incompatible with such serenity.
7
  It is immaterial for present purposes what we think about this case, but our silence
on it can be misinterpreted as agnosticism or just coyness. We hold that the fact that
something is not your fault does not generically, or even typically, lessen its shamefulness.
But when you are ashamed of something you’ve done that you take to reveal something
about your will, then the fact that it was not your fault does count against its shamefulness.
8
 The problem for FA theories is mooted in D’Arms and Jacobson (2000b) and
Rabinowicz and Rønnow-Rasmussen (2004). Darwall (2006) finds precursors in
Strawson (1968). Attempts to solve this problem include Olson (2004), Stratton-Lake
218 Justin D’Arms and Daniel Jacobson

to distinguish reasons of the right kind from those of the wrong kind with-
out circularity. In one sense the distinction is easily drawn. A  WKR is a
consideration that does not bear on whether x is Φ (where Φ is some value
term such as admirable, shameful, or desirable) despite providing reason to
have some associated attitude F (respectively admiration, shame, or desire)
at x.9 Thus all and only those considerations in favor of F(x) that do bear on
whether x is Φ count as RKRs with respect to the Φ. But this is no real solu-
tion to the problem because it cannot be adopted by an FA theory, which
seeks to explicate value in terms of fitting attitudes. The terms “fittingness”
and “merit” are commonly used to describe the endorsement of an evaluative
attitude as correct, in contrast with its endorsement as prudent or virtuous.
Recent versions of FA theory typically interpret the claim that an attitude
is fitting in terms of reasons, so that for x to be admirable is for there to be
(sufficient) reason to admire x. The existence of WKRs shows that defend-
ers of FA theories need to refine their view. They must hold that for x to be
admirable, shameful, or desirable is for there to be sufficient reason of the
right kind for admiring, being ashamed of, or desiring x. But in that case,
FA theory cannot explicate the notion of a reason of the right kind, with
respect to some value Φ, as those considerations in favor of F(x) that bear
on whether x is Φ. It seems viciously circular for the theory to analyze the
shameful as whatever provides reasons of the right kind to be ashamed, and
then to say that reasons to be ashamed are of the right kind just in case they
bear on whether something is shameful.10 FA theories thus require a charac-
terization of the distinction between RKRs and WKRs that does not appeal
to whatever values the theory aspires to capture. This is the technical wrong
kind of reason problem that most of the WKR literature attempts to solve.
We will not be attempting to solve that problem here, though our dis-
cussion will identify some challenges for the extant proposals. Here we are
concerned with broader motivations for trying to distinguish right from
wrong kinds of reason, which are not specific to FA theory. Our motiva-
tions are broader because one need not hold any particular theory of value
in order to be committed to distinguishing fitting from unfitting emotions,
and right from wrong kinds of reason. Some things, such as rampaging
grizzly bears, merit fear. Other things, such as garter snakes, do not—even

(2005), Danielson and Olson (2007), and Skorupski (2007), among others. For some
other important approaches to WKRs, see n. 11. See also Jacobson (2011, 2013).
9
  More strictly speaking, this is how we characterize objective WKRs. We draw the
distinction between the objective and subjective aspects of WKRs in §2.
10
  In a subsequent paper, Rabinowicz and Rønnow-Rasmussen (2006) suggest that
the analysis can still illuminate something about the nature of value even if rendered
circular in this way, but most philosophers take FA theory to have greater ambitions.
 Wrong Kinds of Reason 219

though they cause unfitting fear in many people. And some good reasons
not to be afraid, such as that fear increases the chances of being attacked by
the grizzly, are not considerations of fittingness. These ordinary cases show
not only that all philosophers need to distinguish those reasons for fear that
bear on the dangerous, whether or not they hold an FA theory, but that
everyone draws such distinctions in ordinary thought.
The literature on wrong kinds of reasons for evaluative attitudes has
largely proceeded on the assumption that these reasons pose only a theoreti-
cal problem (for FA) rather than a practical problem, because it is obvious
which considerations are of the right kind and which of the wrong kind; the
only difficulty is to say what each have in common without circularity.11 We
argue that these are mistakes, albeit mistakes encouraged by the canonical
examples. No one would confuse a demonic incentive to admire for a rea-
son why something is admirable. (That is of course the point of such cases;
we are not disparaging them in the context of the technical problem for
which they were intended.) But the phenomenon of opacity of normative
force will demonstrate that the question of which considerations in favor of
F(x) bear on whether x is Φ proves difficult to answer once one considers
a richer array of cases. Moreover, WKRs give rise to errors and confusions
that have nothing to do with any specific theory of value.
In this chapter we will argue that, in addition to the obvious wrong kinds
of reason on which the literature focuses, there are also more interesting
wrong kinds of reason that do not in any straightforward way advert to
advantages of being in the state. Whereas the paradigm cases in the lit-
erature involve incentives that are impossible to mistake as bearing on the
value of the relevant object, our paradigms of interesting WKRs involve
considerations about the object that bear on the propriety of having vari-
ous evaluative attitudes toward it. We will begin by showing the extensive
breadth of WKRs:  they are common in realistic contexts, and they arise
across a broad range of evaluative attitudes and emotions. We then go on to
argue that the normative force of some such reasons can be opaque, as with
fault and shame. As a result, certain philosophical positions and debates can
be seen as misguided.

11
  There is another significant strand in the literature, however. Hieronymi (2005) and
Schroeder (2010, 2012) argue that the notions of right and wrong kinds of reason are
more general, applying not only to reasons for evaluative attitudes but also to reasons for
intention and belief, and in Schroeder’s case even further. And they conceive the WKR
problem as a matter of how to characterize these general notions in a way that explains
what they have in common across domains; and how, generically, to sort right from
wrong kinds of reason for any attitude (or activity). But Hieronymi’s and Schroeder’s
proposals would not solve the problem of the opacity of normative force that we develop
here either.
220 Justin D’Arms and Daniel Jacobson

Interesting wrong kinds of reason are important in several ways. First,

they show that even the most promising ideas in the recent WKR literature
do not allow one to discriminate RKRs from WKRs with full generality.
Perhaps more importantly, though, our cases show that the philosophical
problems generated by WKRs go well beyond the theoretical niche where
the literature has located them. Unlike the obvious cases, they are capable
of confusing people on questions of value. Specifically, they can be taken as
if they were RKRs, conflated with RKRs when they do not in fact compete
with them, and offered in support of evaluative judgments to which they
are irrelevant. We aim to illustrate how confusion over the normative force
of certain reasons has led to mistakes in various domains of philosophy,
including value theory and the philosophy of emotion. If we are right then
the problems caused by WKRs outstrip the wrong kind of reason problem
as it has ordinarily been conceived, and the extant solutions to the technical
problem offer no help with resolving the opacity of normative force.

2.  Obvious and Interesting WKRs

Due to the narrow focus of the WKR literature, not all of the different
ways in which reasons can be of the wrong kind have been widely appreci-
ated. In addition to exogenous incentives (like the demon’s threat), and the
inherent hedonic tone of an emotion (painful or pleasurable), they can also
appeal to ethical considerations (whether deontic or aretaic) about what to
feel.12 Consider three problematic examples of reasons to adopt or forgo
some evaluative attitude, all of which are drawn from recent philosophical
literature—though, tellingly, only the first comes from the WKR literature
itself. These are Crisp’s demon, Bittner’s argument against regret, and Gaut’s
immoral comedy.
Roger Crisp (2000) imagines a demon who threatens to punish you
unless you desire (to drink) a cup of mud. If the threat is credible and
the punishment severe, this demonic incentive seems to constitute
a conclusive reason to want the mud. Even so, the demonic incen-
tive clearly does not make this garden-variety mud desirable (that
is, good).

12
  One might balk at deontic assessments of emotional response on the grounds that
in order for it to be wrong to feel an emotion, you must be able to control what you feel,
and emotions are not voluntary. But even if emotions are thoroughly beyond volitional
control (which seems an overstatement), people are still subject to aretaic assessment of
their emotional dispositions as virtuous or vicious.
 Wrong Kinds of Reason 221

By contrast, were this peculiarly delicious mud, then that fact would count
toward its goodness; were it unhealthy, then that would count against it.
These conflicting reasons could be weighed against each other in determin-
ing the respects in which the mud was good or bad, much like we weigh the
pros and cons of a cup of coffee. But the demonic incentive is a different
kind of reason to desire mud than these, precisely because it does not bear
on the mud’s value. It is the quintessential WKR.
A similar phenomenon arises with emotions as with desire. The next case
does not require demons or other exogenous incentives, since it trades on
the inherent hedonic tone of regret: its painfulness.
Rüdiger Bittner (1992) argues that regret is always irrational. To regret
something involves thinking it a mistake, painfully, and it motivates
change of policy. For example, suppose that, after a close call in the
trees, one comes to regret skiing without a helmet. Then one will
be motivated—though perhaps not sufficiently motivated—to wear
a helmet when skiing in the future. Bittner’s argument for the irra-
tionality of regret does not deny that skiing without a helmet merits
regret: it is regrettable. Yet he claims that we can gain all regret’s advan-
tages without paying its hedonic cost. We do so simply by realizing
that skiing without a helmet is a mistake and deciding to correct our
behavior. Why then add the pain of regret to the error of recklessness?
If regret were indeed useless—which we very much doubt—then Bittner
would have a powerful argument that there is decisive reason not to regret
anything (insofar as you can avoid it).
Yet this reason not to regret our mistakes differs fundamentally from the
reasons that make them regrettable. They can both be decisive, albeit in
answering different questions: whether some action was regrettable, versus
whether to regret it. Were Bittner right, the fitting emotion would diverge
from what there is most reason to feel, all things considered. The pain of
regret and the pleasure of amusement do not bear on whether anything
is regrettable or funny, though they provide a sort of standing reason not
to regret anything and to be amused by everything. Hence these consid-
erations too are (obvious) WKRs, like demonic incentives but realistic and
ubiquitous.
Finally, consider an argument concerning the demands that morality
places on what sort of feelings to have.
Berys Gaut (2007: 241) writes, “Imagine a comedy full of hilarious
jokes, all of which were so vicious and cruel that audiences watched
in stony silence, without being amused at all, since they correctly
thought that it would be wrong to feel amusement.” Gaut’s imagined
audience takes moral considerations to determine what they ought
222 Justin D’Arms and Daniel Jacobson

to feel, even when these considerations are in tension with the comic
value of the jokes (which are stipulated to be hilarious).
Gaut’s immoral but hilarious comedy raises considerations that are neither
strategic nor hedonic: these are ethical reasons to have or not to have an
emotion. Whereas Crisp’s and Bittner’s WKRs both trade on the prospect
of pain, Gaut’s case instead trades on the claim that it would be wrong or
vicious to be amused by such cruel jokes.
Considerations about the morality of feeling some way, like considera-
tions about the painfulness of that feeling, are always WKRs:  they con-
cern what is good to feel, not whether the object of that feeling is funny,
shameful, and so on.13 What is less obvious, however, is what to say about
the cruelty of these jokes, and other such moral defects of the objects of
our sentiments. Gaut holds that such moral defects always make jokes
less funny: they always provide an RKR against amusement. We deny this
sweeping generalization. We claim that sometimes the cruelty of a joke is
merely a WKR:  a reason it would be wrong to be amused, but not one
that diminishes the funniness of the joke. The cruelty of jokes is our first
example of how the opacity of normative force can lead to philosophical
confusion and error.
Before continuing, we need to clarify a couple of tricky points about our
cases of putative WKRs whose normative force can be opaque. Suppose
someone asks, of some consideration c, whether it is an RKR or a WKR
(for having some response F to an object x, with respect to the question
of whether x has the evaluative property Φ). There are two different ques-
tions one might be asking in asking what kind of normative force c has
(and hence what kind of reason it provides): a subjective and an objective
question.
The subjective question is about how c functions in some agent’s psy-
chology: in a right or wrong-kinded way. Is the best understanding of the
agent—which may or may not be the way he understands himself—as tak-
ing c to bear on whether x is Φ, or on whether it is somehow good or right
to feel F(x)? Since some considerations can function differently for different
agents, in these cases there will be no answer to this question independent
of the details of a particular agent’s moral psychology. If c led him to F(x)

13
  This is the leading thought of D’Arms and Jacobson (2000a), which discusses sev-
eral cases that illustrate the point in detail. But this is not to deny that norms of virtu-
ous feeling can take into account considerations of fit. Sometimes what is vicious about
getting angry at someone is precisely that he is not responsible for the transgression that
angers you. Don’t blame the messenger, as it is said—because people have the tendency
to do just that. Even in such cases, however, the fact that it is wrong to be angry at the
messenger is not itself a reason why the anger is unfitting. Rather, the fact that the anger
is unfitting is (part of ) the reason why it is vicious.
 Wrong Kinds of Reason 223

by way of some internalized ethical norm about what sort of person to be,
or through some kind of appreciation of the advantages of feeling F(x),
then it functioned as a WKR for him on that occasion, regardless of his
self-understanding. If instead it led to F(x) because of his emotional sensi-
bility about the Φ—his sense of humor, in the case of the funny—then it
counts as an RKR by his lights, regardless of its substantive goodness as a
reason. The kind of positive light in which this consideration cast the object
was evaluative in the relevant way. These are not exclusive, as c might func-
tion in both ways for an agent.
The subjective question arises because, when it comes to the questions
raised by our cases, people’s evaluative sensibilities can differ in respects that
make a difference to the normative force of c. Someone who thinks that the
only way in which a trait can reflect badly on you is by reflecting badly on
your character may think that the only shameful traits are vices, and that
something counts as a vice—rather than a disease, perhaps—only if you are
responsible for its possession. Such a person will take the consideration that
his obesity is not his fault to undermine its shamefulness, whereas someone
who embraces the Serenity Prayer, and has come to the conclusion that she
cannot change being obese, simply thinks it better not to be ashamed. The
phenomenon of opacity arises in the subjective case because it is not always
obvious to others, or transparent to oneself, how a consideration figures in
someone’s moral psychology.
The objective question is whether c in fact supports the evaluative judg-
ment that x is Φ, or if it merely supports feeling F(x) on other grounds.
This is the question a person is normally asking when, for instance, he asks
whether his lack of responsibility for some trait is a reason of the right kind
not to be ashamed of it. He is not asking about his own perspective but
about the evaluative truth (though this talk of truth, fact, and objectivity
can be understood in familiar quasi-realist ways). When we suggest that
the fact that you are not responsible for your conspicuous impairment is
a WKR against being ashamed of it, or that the cruelty of a joke can be
merely a WKR against amusement, we are talking in the first place about
the objective question. This is what needs to be settled in order to sort out
what things are shameful, funny, and so forth. The opacity of normative
force arises in the objective case because one can think that some considera-
tion bears on whether to have some evaluative attitude without being sure
just how it bears.
These questions can hardly come up about demonic incentives and other
obvious WKRs because the nature of their force is very clear, and they do
not pose any problems due to differences between people’s evaluative sensi-
bilities. No one’s taste is such as to make the demon’s threat render the mud
more desirable to her. But our cases are different. Because the phenomenon
224 Justin D’Arms and Daniel Jacobson

of opacity has both subjective and objective aspects, we will be concerned

with both questions here.
Consider again Berys Gaut’s (2007) example of the hilarious joke so cruel
that it is wrong to be amused by it. We’ve noted that Gaut thinks that the
cruelty of jokes and their other moral flaws are always RKRs against amuse-
ment. He defends the philosophical thesis known as comic moralism:  the
view that the moral flaws in jokes and comedies are inevitably comic defects,
which diminish how funny they are.14 We have argued against comic mor-
alism elsewhere and will not recapitulate those arguments in detail; in our
view, the moral defects of jokes and comedies—their cruelty, offensiveness,
and so forth—do not always count as comic defects in them.15 Though we
grant that cruelty, offensiveness, and other vices sometimes mar jokes, in
other cases the very features that make jokes morally dubious are part of
what make them funny. In those cases, the fact that a joke is cruel is (at
most) a wrong kind of reason not to be amused, which does not diminish
how amusing the joke is. It is an interesting WKR, however, and we dis­
agree with Gaut about the (objective) normative force of this consideration.
One possibility is that this reflects a difference of sensibility; we just have
different senses of humor. Then the cruelty of a joke will sometimes be a
subjective WKR for us but always a subjective RKR for him.
However that may be, our disagreement concerns what we called the
objective question. He claims that to be amused by cruel and vicious jokes
is to take amusement in what is not amusing. In the course of an an­alogy
between the amusing and the pleasurable designed to illustrate just this
point, Gaut (2007: 239) claims: “it is not just that it would be wrong to
take this kind of pleasure, but that to do so would be to take pleasure in
what is not pleasurable, but is, rather, foul and disgusting.”16 Similarly,
to be amused by a foul and disgusting joke would be to take amusement
in what is not amusing but is, rather, foul and disgusting. But this claim
rests on the assumption that such properties as cruelty and disgustingness

14
  Although Gaut calls his view ethicism rather than moralism, we find this taxonomy
more perspicuous. The crucial point is that he claims that moral defects of jokes and
comedies are always pro tanto comic defects, which render them less funny.
15
  See D’Arms and Jacobson (2000a). See Jacobson (1997, 2008) for development of
the thesis known as immoralism: that moral defects sometimes count as comic or aesthetic
merits in jokes and artworks.
16
  Despite this claim, Gaut sometimes retreats to the weaker thesis that, although such
jokes are always rendered less funny by such flaws, they might still be funny. The question
then becomes: less funny than what? The obvious answer is: less funny than it would be
if it did not offend against morality or good taste. Gaut acknowledges that this claim is
highly implausible, to say the least, and rejects it. Nevertheless, he insists that even these
ineradicable flaws always make a comedy less funny. But what can this mean? This seems
like an ad hoc maneuver designed to save the theory from rampant counterexamples.
 Wrong Kinds of Reason 225

always undermine the humor of a comedy. The fundamental flaw in even

the weakest versions of comic moralism lies in this assumption that if two
values somehow conflict, as do the funny and the disgusting—in that fun-
niness is good but disgustingness bad—then anything that makes a joke dis-
gusting must thereby make it less funny. This claim is challenged by many
and disparate counterexamples. Although weak versions of comic moralism
can allow that some jokes are funny despite their cruelty or disgustingness,
the onus is on that theory to explain why humor trades so deeply and regu-
larly in these qualities.
The trouble for comic moralism is not merely that so much humor hap-
pens to transgress norms of good taste and morality, though this is quite
obviously true, but that the heart of much comedy lies in such transgres-
sion.17 Our point here is not just the psychological fact that transgressive
humor often amuses people, since we grant that people can be amused by
things that are not genuinely amusing (though the more prevalent is this ten-
dency, the less plausible the moralist claim becomes). It is rather the implau-
sibility of the thought that to be amused by a foul and disgusting joke is to
take amusement in what is not amusing but, rather, foul and disgusting. This
thought is challenged, not to say refuted, by such comedies as The Aristocrats,
in which many of the most celebrated and renowned comics delight at one
upping each other with displays of amusing disgustingness. That is the whole
joke: the disgustingness—or rather the panache with which the disgusting-
ness is embellished—is just what is funny about it. (It is also worth noting
that the movie revels not just in visceral disgustingness but specifically moral
disgustingness as well.) None of this is to deny that the disgustingness of
these jokes might count as a WKR against amusement:  one can think it
depraved to be amused by such humor without denying that it is funny, or
insisting that it would be more funny were it less disgusting. While this is not
our own view, our argument against moralism is entirely compatible with
the conviction that one ought never be amused by such jokes.
In fact, there is evidence that such WKR-based intuitions drive Gaut’s
view, even though they do not support his theory. After describing the com-
edy full of hilarious jokes that leaves its audience cold because of their moral
qualms, he asks rhetorically if such a play could be comically successful.18

17
  Even if one thinks that the use of “transgressive” as an honorific in art (and comedy)
is clichéd, and that contemporary art too often mistakes banal transgression for original-
ity and courage—as we are inclined to grant—this point is still deeply problematic for
aesthetic and comic moralism.
18
  Although Gaut puts this example into the mouth of what he calls the comic auton-
omist, he too is committed to the possibility of hilarious jokes at which it is wrong to be
amused. He takes on this commitment in order to avoid counterexamples from the many
cruel but funny jokes, by expressly denying that moral flawed jokes cannot be funny.
226 Justin D’Arms and Daniel Jacobson

Clearly not, he answers: “we value art in part because of the quality of the
responses it properly calls forth, and ex hypothesi the audience of this play
is morally cut off from responding to it. That looks like a serious defect
in the play” (2007:  241). Let us grant that the moral inaccessibility of a
comedy’s humor is some sort of defect in it. Nevertheless, it can hardly
be a comic defect—if that is understood as a defect in the funniness of the
play—since by stipulation its jokes are nonetheless hilarious. Indeed, even
obvious WKRs such as demonic incentives can make the comic value of
a joke or play morally inaccessible to a virtuous audience. By taking such
inaccessibility to be a comic defect, Gaut’s argument mistakes paradigmatic
WKRs about the propriety of amusement for RKRs. This is some evidence
that Gaut has been confused by opacity, in a way we think common.
The common mistake rests on the fact that people tend to accord a
defeasible presumption of warrant to their actual emotional response to an
object, such as not being amused by a joke. This presumption is defeasible,
and it is possible to judge contrary to one’s emotions; but when people
find themselves feeling some way and take a consideration to justify their
response (or lack of response), they often do not consider whether it justi-
fies in a right or wrong-kinded way. Even when they do consider the ques-
tion, they do not always answer it in the way that makes the best sense of
their evaluative perspectives. One way in which the normative force of an
interesting WKR can be opaque is that it can create a diversionary response.
It seems plausible that features of jokes can provide reasons to have one
or more negative responses, such as indignation or disgust, even though the
joke is also funny. In such cases, a morally sensitive person will sometimes
be indignant or disgusted. Perhaps he will also feel some amusement, or
perhaps not. If he is amused, he may reasonably be bothered by his own
response, finding it unseemly to be amused by cruelty. But there may be
no amusement because indignation about the cruelty of a joke, let alone
disgust at it, can prevent one from being amused by what is funny in the
joke.19 This does not show that the joke isn’t funny, however, even by the
morally sensitive person’s lights! For all we’ve said, it may be that the joke
would amuse him if it were not for the interference of his moral sensibility,
which prevents it, perhaps virtuously, from getting a hearing from his sense
of humor. Compare another case of emotional diversion. When a daunt-
ing bully insults you, you might not become angry simply because you are
terrified instead; your emotional response is to the threat rather than the

19
  Although people are capable of some degree of emotional ambivalence, it is a famil-
iar fact about the emotions that they can sometimes be mutually incompatible. These
incompatibilities may sometimes be due to different physiological responses that are part
of the syndrome of distinct sentiments—the bodily responses characteristic of anger or
disgust may simply inhibit amusement, for instance.
 Wrong Kinds of Reason 227

transgression. That does not show that there was no transgression, merely
that your fear diverted the insult from angering you.
The comic moralist may be in a similar position: indignant or disgusted
rather than amused. This is the most realistic way to understand Gaut’s
imagined audience, which is not amused by jokes stipulated to be hilarious,
because they think it would be wrong. While it is possible for someone
to be self-aware about this complex evaluative and psychological situation,
it is not obtuse to be led into some confused judgments under such cir-
cumstances. Most of us normally make our judgments about funniness on
the basis of what amuses us—at least under normal conditions, when we
have no apparent reason to mistrust our responses. So a person who is not
amused by a cruel joke, specifically because it is cruel, might easily be led to
suppose that the joke is not funny for that reason. Which is just to say that
he may be best understood as mistaking the normative force of the consid-
eration that the jokes are cruel, by taking it to be an RKR when it is really
a WKR. Whether or not this is true of Gaut, we hope to have made it per-
suasive that this sort of thing happens: the normative force of an interesting
WKR can be opaque even to the person whose reason it is.
If this argument is convincing, it goes a long way toward establishing
some of our central claims. But the argument against comic moralism is
only one example of what we claim to be a general phenomenon, and every
example can be questioned. We suspect that some readers will be skeptical
of whether good sense can be made over disagreements about what is funny,
and others may not be persuaded by the substance of our argument against
moralism. Hence in §4 we will offer an argument for opacity that does not
require us to land a substantive claim in the theory of value. There we con-
sider another debate drawn from outside the WKR literature, concerning
reasons for and against pity, where we argue that no matter which position
one finds most congenial, it proves compelling to understand the opposing
view as trading unwittingly in WKRs whose normative force is opaque.
The phenomena of opacity raise a novel and important problem posed by
certain WKRs, which is obscured by the focus of the literature on obvious
cases. Yet various solutions to the technical WKR problem have been offered
that might help with these issues, since they offer criteria for differentiating
WKRs from RKRs. Moreover, one of the most popular positions amounts
to skepticism about WKRs. If there are no real WKRs, only reasons of the
right kind for some other attitude, then there can be no interesting WKRs
whose normative force is opaque. In §3, we address this challenge and show
that the proposed solutions to the technical problem will not illuminate the
phenomenon of opacity.
Readers whose interests lie primarily in issues of moral psychology, and
who want more evidence of the pervasiveness of opacity and its application
228 Justin D’Arms and Daniel Jacobson

to live philosophical disputes, can skip §3 without much discontinuity. Readers

primarily interested in whether the extant literature on the WKR problem can
solve our issues of opacity may focus on the following discussion and skip §4.
Those with broader interests in moral philosophy, and the stamina to match
them, may find that the two sections together illuminate the interrelation of
an outstanding puzzle in metaethics with live disputes in moral psychology
and normative ethics.

3.  Opacity Meets the Traditional Solutions

to the WKR Problem

The recent literature canvasses a number of ideas about differences between

right and wrong kinds of reason that one might expect to allow us to determine,
of a given consideration, what sort of normative force it has (either in fact or by
the lights of some agent). We will now consider three of the most influential of
these and argue, to the contrary, that none of them solves the problem posed
by the opacity of normative force. The first suggestion is that WKRs are not
reasons for evaluative responses at all, but practical reasons for trying to respond;
call this the reason redescription program. The second suggestion is that RKRs
concern the object of the evaluation, whereas WKRs concern the goodness of
being in the evaluative state; this is the object/state distinction. The third is that
RKRs are unique in that they can be followed to the states for which they are
reasons; this claim, which we will explain further, is the followability thesis. Each
of these ideas contains some important insights and, for all we say here, they
may provide the foundation for a solution to the technical WKR problem. But
even if they succeed in identifying generic criteria of right and wrong kinds of
reason, they do not provide a way of determining, about all of the considera-
tions that rationally guide evaluative attitudes, which kind of reason they are. In
particular, we will argue that our examples of interesting WKRs present a class
of cases that the tools developed in the literature do not help to sort.
To this point we have been assuming that even obvious WKRs, such as
incentives, are authentic reasons for the attitudes they favor, since they certainly
seem like considerations that count in support of having those attitudes. But
it is sometimes suggested that putatively wrong kinds of reason for evaluative
responses are better described as reasons (of the right kind) to do something
else. This idea has been embraced by a number of contemporary philosophers
and is widespread in recent discussions of reasons and rationality.20 It is com-
monly used to claim that incentives to believe are not really reasons for belief,

20
  See Gibbard (1990), Hieronymi (2005), Parfit (2011), Skorupski (2007), and Way
(2012) who calls this view WKR skepticism and offers a novel defense of it.
 Wrong Kinds of Reason 229

incentives to desire are not reasons to want, and incentives for having an emo-
tion are not reasons to feel. According to this reason redescription program,
such considerations are better described as reasons to want to have these atti-
tudes, or to try to bring them about.
John Skorupski argues that reason redescription solves the technical
WKR problem. He claims that it provides a general method for deter-
mining the normative force of any given consideration: just ask what sort
of action or response the consideration is best understood to be a reason
for. Once you ask that question, you will see that WKRs are reasons to do
something: take steps to bring about the evaluative attitude. By contrast,
RKRs are what he calls reasons to feel—for instance to admire, desire, or
be ashamed. Concerning his variation on the demonic incentive, in which
the demon punishes you unless you admire a weak violin performance,
Skorupski (2007: 10–11) writes:
[T]‌he response [to the WKR problem] is an automatic consequence of identifying
the exact reason relation we are discussing. In the case of the violin performance, the
fact that the evil demon has his evil plans is a sufficient reason for me to do some-
thing—namely, bring it about that I admire the performance, if I can.
The trouble with Skorupski’s suggestion is that his general method, which
involves simply asking whether a consideration provides a reason to do
something or to feel something, becomes inadequate when one moves from
demonic incentives to more realistic and interesting cases. If one is uncer-
tain whether lack of responsibility constitutes an RKR not to be ashamed
or only a WKR (and similarly for our other examples), then no help is given
by reframing the question as suggested. Any confusion or uncertainty about
the normative force of the consideration will not be resolved, but merely
relocated, by this method.
The reason redescription program may ultimately be correct, but to settle
that question would require a general account of reasons. However that may
be, it offers no help with interesting WKRs like the ones we’ve been consid-
ering. In these hard cases where normative force is opaque, it is exactly as
difficult to determine whether a consideration is a reason to feel or a reason
to act, as to decide whether that consideration is an RKR or a WKR. We
think the same is true of the other solutions to the traditional problem: even
if true and important, they do not help determine normative force in dif-
ficult cases. Rather, they supply a potentially insightful description of these
considerations that can be utilized once their normative force has been set-
tled. Thus none of these ideas help answer the problems to which opacity
gives rise.
The second influential idea in the literature seizes upon an obvious dif-
ference between most WKRs, especially those on which the literature has
230 Justin D’Arms and Daniel Jacobson

focused, and RKRs. Roughly speaking, RKRs are about features of some
object held to be relevant to evaluative judgment, whereas WKRs are about
normative assessments of having an evaluative attitude toward that object.
This characterization seems to explain what is right about the RKRs and
wrong about the WKRs, specifically with respect to evaluative judgment.
The fact that the violinist’s performance is off-key is an RKR against admi-
ration because it concerns the object of the value judgment:  the perfor-
mance. The demonic incentive is a WKR because it concerns the cost or
benefit of the relevant evaluative attitude: admiration (of the performance).
The latter is what Derek Parfit (2011) calls a state-given reason, since it is a
respect in which it is good or bad to have the relevant state.
The idea that the distinction between right and wrong kinds of reason cor-
responds to the distinction between object-given and state-given reasons is
attractive, but it proves more difficult to formulate clearly than first appears.
The sharpest formulation comes from Jonas Olson (2004), who calls certain
reasons A-referential because they refer to the very attitude A that they are
held to be reasons for having. Olson proposes A-referentiality as the mark
of WKRs. Thus the facts that you will be punished unless you desire some
mud and that regret is painful are both determined to be WKRs by Olson’s
test, because they refer to the very attitudes that they purport to justify,
and some less obvious cases seem amenable to this treatment as well.21 But
Olson’s proposal has been shown to face difficulties from cases where rea-
sons that are clearly A-referential nonetheless seem capable of functioning
as RKRs.22
Our cases are different. They are apparently object-given, but their nor-
mative force may not be of the right kind. The consideration you fix on
(that your lame leg is not your fault, or that the joke is cruel), which both
explains your reaction and justifies it to you, is about the object of your
attitude (your lame leg, the joke), not about the attitude itself. Similarly, we

21
  Olson’s formulation is motivated by examples like our (2000a) case of the rich but
touchy friend who will cut off his largesse if he suspects that you envy him. Here his
touchiness consists precisely in an attitude toward your state of mind. Olson tries to rule
out such cases through a restriction against state-given considerations “in the guise” of a
property of the object, as he puts it. This seems right: in order to understand the reason as
any sort of consideration against envying, one needs to adduce the fact that he is touchy
specifically about being envied. So this consideration is covertly A-referential.
22
 In particular, the case given by Hieronymi (2005:  447)  and Rabinowicz and
Rønnow-Rasmussen (2006:  118)  of the reciprocal lover is problematic for this view.
This is someone who will respond to your loving her by loving you back, which is an
A-referential consideration that could coherently be thought to be an RKR that makes
her loveable. Note that this can be true even if the trait also justifies loving her in a
wrong-kinded way, by constituting an incentive for loving her. But whether one takes
reciprocality as a right or wrong kind of reason for love, or both, there will likely be no
confusion about either rationale.
 Wrong Kinds of Reason 231

suggest that such object-given considerations as the fact that your daughter
won runner-up in the elementary school spelling bee (an extremely modest
achievement), or the fact that the person who won the award you coveted is
your friend, are best understood as WKRs despite being object-given. That
is because they count as reasons for pride and against envy because of how
they bear on what kind of a person you would be to feel certain ways. They do
not show their objects to be respectively more prideworthy or not enviable;
rather, they make it good to be proud and bad to envious. All these con-
siderations can function as WKRs, and are typically better taken that way,
objectively speaking. Hence it does not suffice to establish that a considera-
tion is object-given, on its face, in order to establish its normative force.
It may be objected that, when we explain how these considerations can be
WKRs, we end up saying something that sounds A-referential: we mention
respects in which they make it in some way good or right (bad or vicious) to
have the attitude. This can be taken to suggest that, insofar as our cases are
understood to be WKRs, they are state-given. When the cruelty of a joke
is a WKR, for instance, the full description of the reason is something like
this: the joke is cruel so it would be wrong to be amused by it. In a sense,
we grant this point. The important insight behind the object/state distinc-
tion, we think, is just that wrong-kinded normative force has to do with
ways in which it would be in some way good or bad to have the attitude,
whereas right-kinded force has to do with ways in which the object is good
or bad. But this insight does not provide a way of assessing the normative
force of interesting WKRs. The considerations that agents fix on are often
simple truths about the object: the joke is cruel, the lame leg is not my fault.
Once we have decided that the reason provided is of the wrong kind (in
fact or by his lights), then we can say that its normative force is state-given
(or that it functions that way for the agent). That is, we can then interpret
him as having been moved by this feature of the object because of some
implication it has for why the attitude would be good or bad to feel. But the
state-givenness of the rationale need not be a feature of the consideration
that he found persuasive; it is rather an explanation of how that considera-
tion is best understood as functioning in his psychology. So the object/state
distinction does not settle the normative force of interesting WKRs; it offers
an illuminating description of that force, once it is identified.
The final suggestion we wish to consider dovetails nicely with the reason
redescription program. We have been working with a conception of reasons
that follows Scanlon’s (1998: 67) characterization of a reason as “a consid-
eration that counts in favor of some judgment-sensitive attitude”—which
seems inevitably to include WKRs. But Parfit (2011:  51)  adds another
requirement by stating that “[r]‌easons are things to which at least some
people might be able to respond.” This matters because various authors note
232 Justin D’Arms and Daniel Jacobson

that there is an asymmetry between RKRs and WKRs with respect to how
people respond to them. Consider the case of belief, where this is clearest.
It seems that non-epistemic considerations in favor of some belief can be
followed—that is, responded to directly as reasons—only by wanting or try-
ing to have the belief. One cannot adopt a belief because it would be good
to have it. If you are convinced by Pascal’s wager, for instance, then you will
find yourself wanting to believe in God—but not yet believing. It seems
that the only reasons for which one can form beliefs are considerations that
one takes to bear on the truth of the proposition believed. Similar claims
have been made about intentions and evaluative attitudes:  that the only
reasons for which one can intend are considerations that (one takes to) bear
on whether to perform the intended action, and the only reasons for which
one can admire are considerations that bear on admirability.
In this view, incentives can function as reasons to want or try to have
attitudes but not as reasons to have them. Although it is seldom noted,
this general point can be applied to ethical reasons concerning attitudes as
well. The consideration that one morally ought to feel some way or believe
something will typically be incapable of being an agent’s reason for those
attitudes, though it might be his reason for wanting to have them, and for
feeling guilty if he fails. If so then this asymmetry in the rational role that
various considerations can play lends principled support to the reason rede-
scription program. It also suggests a general test for whether reasons are of
the right kind, which one might try to use to determine the normative force
of a given consideration.
According to this proposal, the only considerations that can function as
an agent’s reason for having some evaluative attitude are considerations of
fittingness by his lights: those that he takes to bear on whether the object
of the attitude has the relevant evaluative property. More concisely, the fol-
lowability thesis states:
If an agent S can follow some consideration c directly to F(x) or to
not-F(x), then c is an RKR by S’s lights.
Although this is our own construction, one can find similar ideas through-
out the literature.23 Raz (2009:  40)  speaks of following a consideration
directly, which he explicates as a matter of coming to the attitude for some
reason c, where this does not require any effort or extra step in reasoning.
Other authors express the thought a little differently, but they are clearly
sympathetic to something like the followability thesis as characterized here.

23
 It is endorsed more or less explicitly by Hieronymi (2005), Raz (2009), and
Skorupski (2010). See also related claims in Parfit (2011) and Kolodny (2005).
 Wrong Kinds of Reason 233

But we contend that followability does not provide a device for deter-
mining the normative force of a given consideration for an agent, because
in the cases we have been considering, it is possible for an agent to follow
a consideration that is a WKR by his own lights to an evaluative attitude.
At least, that can be true for all that the agent himself or an astute observer
can tell, since it is possible to respond to interesting WKRs effortlessly and
without any extra step in reasoning. You can sometimes feel proud of your
child’s very modest success or not envy your friend’s triumph, and do this
directly and without effort, even though you do not take the crucial con-
siderations to merit your responses. Someone may think that it is bad for
him that his friend won a coveted award, in just the way that makes envy
fitting, but he might be a sufficiently good friend that he isn’t envious of
her—though he would envy anyone else who had won it. He has a concep-
tion of what kind of person it is best to be that calls for pleasure at a friend’s
deserved successes, and on this occasion at least he feels in accordance with
it. Then the fact that she is his friend is his reason not to be envious, but it
is a WKR by his own lights.
Similarly, a parent can be proud of his child for performances that do
not really merit pride, in his view. Because good parenting sometimes calls
for such responses, many parents muster them directly and unreflectively
without any conscious effort, at least after a while. They need not be deluded
about the relevant standards of performance in order to be proud of their
child; they can simply have inculcated a disposition to respond unreflectively
with pride to his “accomplishments,” in part because they emulate other sup-
portive parents. Moreover, the fact that one focused on a consideration, took
it to justify or undermine an attitude, and then acquired or shed the attitude
effortlessly, without any further step in reasoning, does not ensure that the
consideration was of the right kind. Hence the followability of a considera-
tion does not supply a test for determining its normative force for an agent.
These claims are contestable, and we only claim to have made a prima
facie case for them here. But it is important to note that, since the authors
who champion the followability thesis only consider obvious WKRs, not
the ethical considerations on which we’ve focused, they offer no argument
that interesting WKRs cannot be followed. Instead, they argue for a claim
that may well be true: that people are incapable of following incentives to
believe, desire, and feel. Once we expand our focus beyond incentives, how-
ever, the claim that a person can only follow considerations that bear on
the Φ by his lights is much less plausible. We grant that aretaic ideals about
what kind of person to be, and ethical prohibitions against certain sorts of
responses, are limited in their ability to regulate human psychology. But we
see no grounds for insisting that, contrary to appearances, they can never
be successfully internalized in ways like those we have been imagining here.
234 Justin D’Arms and Daniel Jacobson

A sophisticated line of objection begins by granting that, at least in the

domain of emotions, people may be able to fix on considerations that are
(subjective) WKRs and move effortlessly into the emotional state that
the reasons favor. Nevertheless, the objector insists that such cases do not
count as following reasons to the relevant evaluative attitudes, but rather
as following these reasons to the bringing about of the attitudes. The fact
that no effort is required to follow the practical reason merely makes these
cases unusual.24 While this claim saves the letter of the followability thesis,
it does so by restricting what counts as following a reason to an attitude
to only those cases where the consideration offers right-kinded support
for that attitude. One will therefore be able to say what kind of reason an
agent followed only once one knows what force the consideration had by
his lights; but, as we’ve argued, this can be hard to determine. Hence the
cost of this maneuver is to abandon followability as a test of normative
force. This objection therefore does not help solve the problems posed by
the opacity of normative force.
Nevertheless, certain lessons can be drawn from the literature on the
technical problem that can be applied to our cases. In the first place,
state-given reasons—that is, considerations about respects in which it
would be good or bad to have some attitude—are the paradigmatic WKRs.
Although most of the obvious cases fix on incentives, which are respects
in which it is prudentially good to have an attitude, the moral goodness
or badness of having an attitude is equally a state-given reason and, hence,
a WKR. But our examples of opacity show that reasons do not always
wear their state-givenness on their sleeve. Considerations about objects
that strike agents as compelling are sometimes best understood as having
unrecognized wrong-kinded rationales. Moreover, the fact that these con-
siderations can be followed, or at any rate that fixing on them can bring
one effortlessly to the relevant attitude, entails that the key psychological
truth about obvious WKRs—that awareness of incentives for an attitude
only brings about the desire to have it—cannot be operationalized as a test
of normative force in difficult cases. Hence none of the proposed solu-
tions to the technical WKR problem, even if correct, solve the problems
of opacity posed by interesting WKRs. These issues can be found at large
in philosophical debates outside of the technical problem, such as a recent
dispute over the role of fault and blame in pity and the pitiable.

24
 Pamela Hieronymi has suggested this line of reply to us in discussion. Parfit deploys
a parallel line of thought in the belief case, but he supposes that our psychologies would
have to be different in order for it even to be possible to believe effortlessly on the basis
of WKRs.
 Wrong Kinds of Reason 235

4. Pity, Fault, and Blame

Philosophers differ about the fittingness conditions for pity and, hence, about
what count as RKRs for pitying someone. In this section we consider two the-
ories of pity, contrasting a familiar Christian view suggested by Robert Roberts
with a motivated revision of Martha Nussbaum’s Aristotelian account. Our
aim is not to defend either of these views, both of which we find implausibly
simple and extreme, nor is it to put forward our own account of when pity is
fitting. We instead aim to show that, although considerations about fault and
blame clearly seem relevant to whether and how much to pity someone, their
normative force is opaque. Moreover, we will suggest that some of the strong-
est considerations about what to pity are better understood as reasons of the
wrong kind rather than as determining what is pitiable.25 These considerations
lead both Nussbaum and Roberts to commitments about the pitiable that we
find implausible. Although we may be unable to persuade partisans of either
account to give up their theories, we hope to persuade each side—as well as
philosophers with no theoretical stake in the matter—that the opposition is
best understood as trading illicitly in WKRs.
The discussion of pity will serve as a case study for our more general
thesis. The phenomenon of opacity matters because, in order to have views
about what is pitiable, shameful, funny, and so forth that reflect one’s evalu-
ative perspectives, and to engage in normative discourse about these claims,
one must distinguish RKRs from WKRs. But that task can be quite dif-
ficult, and the need for it too often goes unnoticed in realistic cases of the
sort we are discussing. This is especially true when one judges on the basis
of one’s actual emotional response (or lack thereof ), and the considerations
that one takes to justify them. If some of these responses are best under-
stood as being justified by WKRs, as we will suggest, then they do not sup-
port the evaluative judgments that they often entice people to make.
Consider the emotion sometimes referred to as compassion but more
commonly called pity.26 All sides agree that pity involves being pained, or
at any rate bothered, by another person’s suffering or misfortune.27 Robert

25
  By ‘pitiable’ we mean fittingly pitied, not able to be pitied or normally pitied.
26
  As both Roberts and Nussbaum note, “pity” sometimes has connotations of conde-
scension, which is why they opt for the term “compassion.” But all parties to this discus-
sion take themselves to be talking about a robust and familiar emotion kind, which is the
same emotion whether called pity or compassion.
27
  Nussbaum refers both to misfortune and suffering, and Roberts to distress; but one
can pity misfortune that is not painful. We will focus here on suffering, both for simplic-
ity and coherence with the literature.
236 Justin D’Arms and Daniel Jacobson

Roberts endorses the recognizably Christian view that all suffering merits
pity. “When compassion is an emotion rather than a character trait,” he
writes (2003:  295)—that is, when it is pity—“it is a construal of some
person or other sentient being as in distress.” This view seems to imply that
all suffering merits pity, even when the victim has brought it upon himself
through his malfeasance. But according to the Aristotelian view defended
by Martha Nussbaum (2001: 301; emphasis added), “compassion [i.e. pity]
is a painful emotion occasioned by the awareness of another person’s unde-
served misfortune.”
Although Nussbaum follows Aristotle in referring to deserved misfortune
and suffering, this seems a misleading expression of her view. She expressly
includes both prudential error (when someone brings misfortune on him-
self, for instance through foolishness) and moral error (when the misfortune
arises from the agent’s wrongdoing) as ways in which someone can deserve
misfortune, for the purpose of ruling out pity. But because it would take
an especially harsh retributivist to hold that the foolish deserve to suffer,
we think it more perspicuous to differentiate these errors by distinguishing
between suffering for which one is prudentially at fault and suffering for
which one is morally to blame. We will call both kinds of suffering criticiz-
able rather than deserved. Thus we stipulate that the reckless skier is at fault
for his self-inflicted injury, whereas the feckless criminal is to blame for her
incarceration. In each case the agent brought his suffering on himself in
some criticizable way, which according to Nussbaum (2001: 311) suffices
to ensure that one does not pity him:
Insofar as we believe that a person has come to grief through his or her own fault, we
will blame and reproach, rather than having [pity]. Insofar as we do feel [pity], it is
either because we believe the person to be without blame for her plight or because,
though there is an element of fault, we believe that her suffering is out of proportion
to the fault. [Pity] then addresses itself to the nonblameworthy increment.
Although this claim gestures at something true and important, it is much
too strong as it stands. Nussbaum’s official theory of emotion commits her
to holding that it is impossible to pity someone’s suffering insofar as one
judges it his own fault. But this cannot be right, unless it is a stipulation
about what she will count as pity—which is not what she intends. To the
contrary, Nussbaum (2001: 301) intends to address a “ubiquitous human
phenomenon” that plays a central role in other cultures and even other
primate species, and that is the same emotion discussed by Aristotle and
Rousseau.28 Surely people can and sometimes—we think often—do have
28
  Note that Rousseau (1987: 53) characterizes pity simply as “an innate repugnance
to see [one’s] fellow suffer.” One can agree that normal humans have such an innate
repugnance (as we do) without granting that all suffering merits pity (which we deny).
 Wrong Kinds of Reason 237

the same familiar emotion towards criticizable suffering as they do towards

faultless suffering. This seems especially clear when the sufferer is a loved
one and the error is merely prudential, but that is just the most obvious
counterexample. When considered as an empirical proposition, the claim
that such pity is impossible is untenable; in order to make it plausible, we
must revise it into either a weaker psychological claim or a normative claim.
The weaker but true psychological claim is that people tend not to pity
suffering they take the sufferer to have brought upon himself—or at least
they pity it less. Sometimes they do “blame and reproach” the sufferer, as
Nussbaum claims, but in other cases they laugh at him or are left indif-
ferent by his plight. We do not think simple generalizations can be made
about these circumstances, both because people differ in their emotional
tendencies and because the cases can be elaborated so differently. There is
an important difference between fault and blame; it is one thing to ski
without a helmet and get a head injury, quite another to attempt to injure
someone and wind up hurting oneself. It also matters who suffers, in that
the suffering of a loved one is far more likely to be pitied, even when he is
at fault or to blame, than that of a stranger. These are just two of the most
obvious distinctions, not the only ones, and how badly someone suffers
surely matters as well for any prediction of how much he will be pitied (and
by whom). So although Nussbaum is right that fault and blame are relevant
to the likelihood of someone being pitied, she exaggerates in suggesting it is
impossible to pity suffering unless one takes it as “falling on the person from
outside, so to speak” (2001: 313).
Nussbaum is committed to a relevant normative claim as well (although
she does not expressly discuss it) which seems to us more defensible than
her official theory of pity, though it is still too strong. The claim is that pity
can fittingly be felt toward suffering only when the sufferer is not criticizable
for bringing it about. Recall that the fittingness of an emotional response is
a matter of its correctness. Since Nussbaum identifies emotions with judg-
ments, and a judgment is correct just in case it is true, her view entails that
pity is fitting just when directed at uncriticizable suffering. This norma-
tive claim has the advantage of being compatible with the psychological
possibility of unfitting pity; hence it isn’t falsified, as is her official theory,
by the fact that people do sometimes pity suffering despite believing the
victim somehow responsible. We’ll call this view about the fittingness of
pity, without the psychological claim about its impossibility, the improved
Aristotelian view. While this gives up on some of Nussbaum’s claims, it is
a coherent and arguable position about the fittingness of pity, and one to
which she seems committed.
Although the improved Aristotelian view tracks certain emotional dis-
positions, it conflicts with others, and we do not endorse it as a theory of
238 Justin D’Arms and Daniel Jacobson

the pitiable. Michael Weber (2004) argues, to the contrary, that fitting pity
is not restricted to suffering that is faultless or even blameless.29 It is fitting
to pity someone who ends up in a wheelchair because he falls asleep at the
wheel of his car, Weber claims, even though his injury is largely his own
fault. Weber also argues, in our view persuasively, against the attempt to
understand this pity as being directed only at the “nonblameworthy incre-
ment” of the suffering. As Weber notes, it is hard to draw a general distinc-
tion between a blameworthy element and a non-blameworthy increment of
suffering. In the case of the sleepy driver, this requires arguing that his severe
injury is out of proportion to his mistake, even though it is just the sort of
thing that can be expected to happen when people fall asleep at the wheel,
which is precisely why the driver should have pulled over.
We find it telling that Nussbaum seems to have difficulty embrac-
ing the consequences of her own theory. She writes unsympathetically of
Americans who supposedly judge that a sexual assault is not pitiable because
“the woman ‘brought it on herself ’—by walking alone in a dangerous place,
for example” (2004: 213–14). But it seems as though even the improved
Aristotelian view must side with these Americans against Nussbaum here. If
the assault is a (foreseeable) consequence of the criticizable decision to walk
alone in the dangerous place, then the Aristotelian considers pity for the
victim unfitting. Since both the assault victim and the sleepy driver made
risky choices and got horribly but foreseeably unlucky, both are criticizable
in ways that supposedly vitiate the fittingness of pity.
Nevertheless, there may be good reasons to pity the victim, of which
Nussbaum could avail herself. It is widely held that there are bad social
consequences of “blaming the victim” in cases of sexual assault; indeed,
this was precisely what Nussbaum was complaining about when criticiz-
ing Americans for doing something very like what her theory forces upon
her. But while these moral considerations about what to feel may be good
reasons to pity the victim, they are WKRs. Considerations about the social
consequences of pitying are not about what merits pity but about a different
question, namely what good pitying can do—they are a form of incentive.
To be clear, in our opinion the imagined victim’s suffering really is piti-
able, in part because she is merely at (prudential) fault rather than (moral)
blame; but that is just to deny the central commitment of the Aristotelian
view. The best position available to Nussbaum, compatible both with her
theoretical commitments and her intuition that one should pity the assault
victim, is to hold that pity for the victim is defensible, indeed mandatory,
but for reasons that are of the wrong kind with respect to pity. When she

29
  The distinction between fault and blame is ours rather than Weber’s, though we
suspect that he would agree with it. We are translating some of his claims into our terms.
 Wrong Kinds of Reason 239

finds herself feeling pity toward the victim of sexual assault and angry at the
callousness of those who do not, she cannot see her pity as an expression
of her view about what it is correct to pity, without giving up her theory.30
But it is open to her to see it as a reflection of a moral sensibility that is
concerned with other values as well. In other words, her pity for the assault
victim might be responsive to a good WKR.
There is another way to resolve the tension between Nussbaum’s theory
and her intuition about the assault victim, which also appeals to opacity.
This is to jettison the least plausible aspect of the Aristotelian view:  its
claim that anyone who suffers through prudential error does not merit pity
(except for the uncriticizable increment of his suffering). What is the appeal
of that claim to the Aristotelian? Suppose that you find yourself criticizing
and reproaching the person who suffers due to his own mistake, as we some-
times do—and as Nussbaum claims is inevitable. Then it may be difficult
to pity him, especially if you feel a conflicting emotion such as irritation
or amusement (via ridicule). Moreover, there are considerations that you
might take to justify your lack of pity. In some cases one might justifiably
criticize the sufferer rather than pitying him in order to help him learn from
his mistake, since to mollycoddle someone who came to grief through his
own bad decisions could discourage learning from them. Furthermore, one
inevitably has limited emotional resources, and there is so much suffering
in the world that one cannot respond even to all the genuinely pitiable suf-
fering. Perhaps then one should save one’s pity for those who did nothing
to bring about their suffering. These are good reasons not to pity suffering
that was brought about by prudential error, but they concern issues about
respects in which it is for the best not to pity: they are WKRs.
Now consider the Christian view mooted by Roberts, on which all suf-
fering merits pity. We will suggest a similar conclusion about this claim: it
would be improved by acknowledging that in some cases the good reasons
for pity are of the wrong kind. Note first that this view too is in tension
with most people’s emotional dispositions, in that we all tend to pity some
suffering more than others, for various reasons not limited to its intensity.
Of course not all the ways people do feel need be ratified as fitting. Most
people endorse some differentiation in pity, though, or would do so if they
reflected on it. Take a case of clearly blameworthy suffering, such as that of

30
  In fact, Nussbaum’s overt view, in contrast to the improved Aristotelian view we are
attributing to her, cannot even allow that it is possible to pity the victim insofar as one
acknowledges the (stipulated) truth that he acted recklessly and is therefore criticizable
for his plight. While Nussbaum could say that it is not genuine pity but some other emo-
tion that one can feel toward the victim, this would be ad hoc, undermine her claim to
speak about the familiar emotion discussed by Aristotle and Rousseau, and commit the
No True Scotsman fallacy.
240 Justin D’Arms and Daniel Jacobson

a terrorist who places a nail-bomb in a crowd of innocents—indeed, delib-

erately sets it right next to a small boy whose body is torn apart by shrapnel
when the bomb explodes. When the teenage terrorist is wounded during
his capture, he is sent to the hospital where he is treated. The nurses find
themselves responding to him, habitually, as an ordinary patient. “You see
a hurt 19 year-old and you can’t help but feel sorry for him,” a nurse says,
though she makes a pact with another nurse not to express this pity as they
ordinarily would. They correctly deem the terrorist to be blameworthy for
his suffering, and therefore not to merit the pity they can’t help but feel.31
While the nurses feel pity, due to a non-rational tendency to empathic
response to suffering, they do not endorse it. So they do not go on to make
any (erroneous) judgment.
When the terrorist is convicted and sentenced to execution at dawn, he
will likely suffer greatly during his final night. What should one make of
the nun who sits vigil for him, praying for his soul and, as predictably as
the nurses, pitying him? Most plausibly and charitably, she does not har-
bor illusions about his blameworthiness. Rather she embraces a conception
of how to live and what to feel that she takes to be embodied by Christ,
who enjoined compassion even for the worst sinners. We want to suggest
that many of Christ’s teachings—like those of many spiritual figures—are
better understood as WKRs. They set an ethical standard for how to feel,
rather than expressing norms of fittingness.32 Consider the admonition to
respond to a transgression against oneself by turning the other cheek. That
is a call against feeling anger that would nonetheless be fitting, since such
transgressions are precisely what merit anger. Similarly, Christ’s example
can be understood as encouraging pity at all suffering, even suffering that
doesn’t merit pity. If this is what moves the nun, though, then her justifica-
tion for pity should not be expressed as the view that all suffering merits
pity. Instead, she holds that the best way to live involves pitying all suffer-
ing—perhaps especially the suffering that others do not pity—regardless of
considerations of blame and desert.
We suggest that those who are attracted to the Christian ideal, like
Roberts, should keep in mind that it is one thing to claim that some
emotional response is fitting, another to claim that it is what a virtuous

31
  This case is (obviously, for now) drawn from the recent bombing of the Boston
Marathon, but it is realistic enough that we expect it to resonate after the details of the
incident are forgotten. The description and quotation are drawn from newspaper reports.
What follows about the death sentence and the nun is wholly fictional.
32
  Buddhism and Stoicism are problematic in this regard, because their therapeutic
advice often conflates claims about fittingness (that your child’s death is a matter of indif-
ference) with clearly ethical, purely forward-looking considerations (that the water of
your tears won’t grow a new son).
 Wrong Kinds of Reason 241

person would feel. That is, they must differentiate right from wrong
kinds of reason. Having done so, they need not disagree with the com-
monsense position, grounded in normal patterns of emotional response
that may well include their own, which denies that all suffering equally
merits pity, without regard to blameworthiness. They can instead inter-
pret the pull of their distinctively Christian intuitions, whose normative
force is initially opaque, as expressing norms of virtue rather than stand-
ards of fittingness.
We expect some readers to demur from this suggestion, as perhaps Roberts
would, and insist that they take all suffering to merit pity. We need not
convince them in order to make our central point. The more you are con-
vinced of this view of when pity is fitting, the more bizarre the Aristotelian
account must seem, on which no suffering that can be aptly criticized merits
pity, whether it is blameworthy or merely due to fault. Rather than see-
ing Nussbaum and others who are attracted to the Aristotelian position as
being simply wrong, however, one can take them to be responsive to WKRs.
Perhaps the most compelling such consideration is that this world is filled
with suffering, and we all have limited emotional resources, so we should
try to direct our pity at those who are least responsible for their suffering
(except perhaps when a more pressing social norm overrides). Analogously,
those who favor the Aristotelian view can think Roberts and his follow-
ers not simply to be feeling pity groundlessly, but to be responding to the
Christian injunction, another WKR. Those who reject both theories, as we
do, may conclude that both should be seen as being sensitive to good rea-
sons, albeit reasons of the wrong kind.
The general point of this section is to illustrate a philosophical debate
from outside the WKR literature whose participants are arguably seizing on
considerations that are good reasons to pity, but better understood as func-
tioning as WKRs with respect to the merit of pity.33 The normative force
of these considerations can be opaque to those who are confident that they
provide good reasons to have some evaluative attitude. Of course it is open
to philosophers to insist that they are putting them forward as considera-
tions of fittingness—that is, as RKRs. The question then becomes how they
think it most charitable to understand their opponents: as putting forward
a deeply misguided view of what merits pity, or as mistaking a good WKR
for an RKR.

33
  In general, we suspect that if a consideration is a good WKR then those who are
moved by it are probably taking it in a wrong-kinded way. While this presumption is
defeasible, we doubt that it is defeated by the fact that the subject holds a philosophical
theory to the contrary.
242 Justin D’Arms and Daniel Jacobson

The problem of opacity holds not just for partisans of fitting attitude
theories of value, if our diagnosis is correct, but for all those who want to
distinguish fitting from unfitting emotions—which includes almost every-
one. Neither Roberts nor Nussbaum defends an FA theory, but until they
attend to the different kinds of reasons to favor or oppose pity in various
cases, it is not even clear that they really disagree about the pitiable, their
philosophical commitments to the contrary notwithstanding.
This chapter has tried to demonstrate that wrong kinds of reasons are an
unrecognized source of error in evaluative thinking. The argument to that
conclusion has been complex, and it is worth summarizing some of its main
elements. In the first place, we claim that WKRs are a broader category than
they are often thought to be, encompassing not only incentives but other
considerations in virtue of which it is good or right, or alternatively bad or
wrong, to have some response to an object. The fact that regret is painful
makes it (prudentially) bad to regret your foolish mistake, and the fact that
the person who won the award you wanted was your friend makes it (mor-
ally) bad to envy him. But the foolish action is regrettable and the award
enviable nonetheless.
Unlike incentives for being in a state, the normative force of some WKRs
is not obvious, partly because they refer to features of objects. It can be clear
that some seemingly object-given consideration provides good reason for
or against pity or shame, for instance, without it being at all clear whether
it supports the evaluative judgment that someone’s suffering merits pity, or
that someone’s social disability is not shameful. It may instead only support
the ethical judgments that it is better to pity those people and not to be
ashamed of such things. Moreover, sometimes a person can suppose that
the consideration supports an evaluative conclusion when his own sensibil-
ity is best understood to be taking it as a WKR. This sort of confusion is
especially likely because some WKRs can be followed; unlike incentives,
they are considerations that can sometimes enable people to have or with-
hold various responses effortlessly, though they do not support the evalua-
tive verdicts those responses are normally taken to justify. Hence the most
widespread and pressing problem with WKRs is that, in various realistic
scenarios, people can fix on what they take to be a good reason to pity, or
not to be amused or ashamed, without recognizing whether it bears on what
is pitiable, funny, or shameful.34

34
  We are indebted to audiences at the Wisconsin Metaethics Workshop 2012, the
OSU/Maribor/Rijeka Philosophy Conference 2011, University of Sydney 2009, and
SPAWN 2007; Geoff Sayre-McCord; and an anonymous referee for their helpful com-
ments. This chapter was supported by a grant from the John Templeton Foundation.
 Wrong Kinds of Reason 243

References
Baumeister, R. F., Stillwell, A. M., and Heatherton, T. F. 1994. “Guilt:  An
Interpersonal Approach,” Psychological Bulletin, 115: 243–67.
Baumeister, R. F., Stillwell, A. M., and Heatherton, T. F. 1995. “Interpersonal Aspects
of Guilt: Evidence from Narrative Studies,” in June Tangney and Kurt Fischer
(eds), Self-Conscious Emotions: The Psychology of Shame, Guilt, Embarrassment, and
Pride, 255–73. New York: Guilford Press.
Bittner, Rüdiger. 1992. “Is it Reasonable to Regret Things one Did?” Journal of
Philosophy, 89: 262–73.
Blackman, Reid. MS. “Pathetic Rationality and Practical Reason.”
Crisp, Roger, 2000. “Review of Joel Kupperman, Value . . . And What Follows,”
Philosophy, 75: 458–62.
Danielson, Sven, and Olson, Jonas. 2007. “Brentano and the Buck-Passers,” Mind,
115: 511–22.
D’Arms, Justin, and Jacobson, Daniel. 2000a. “The Moralistic Fallacy:  On the
‘Appropriateness’ of Emotion,” Philosophy and Phenomenological Research,
61: 65–90.
D’Arms, Justin, and Jacobson, Daniel. 2000b. “Sentiment and Value,” Ethics,
100: 722–48.
Darwall, Stephen. 2006. The Second-Person Standpoint. Cambridge, MA: Harvard
University Press.
Gaut, Berys. 2007. Art, Emotion and Ethics. Oxford: Oxford University Press.
Gibbard, Allan. 1990. Wise Choices, Apt Feelings. Cambridge, MA:  Harvard
University Press.
Hieronymi, Pamela. 2005. “The Wrong Kind of Reason,” Journal of Philosophy,
102: 437–57.
Jacobson, Daniel. 1997. “In Praise of Immoral Art,” Philosophical Topics, 25: 155–99.
Jacobson, Daniel. 2008. Review of Berys Gaut, Art, Emotion and Ethics. Notre Dame
Philosophical Reviews: <http://ndpr.nd.edu/news/23394-art-emotion-and-ethics>.
Jacobson, Daniel. 2011. “Fitting Attitude Theories of Value,” The Stanford Encyclo­
pedia of Philosophy: <http://plato.stanford.edu/entries/fitting-attitude-theories/>.
Jacobson, Daniel. 2013. “The Wrong Kind of Reason Problem,” International
Encyclopedia of Ethics.
Kolodny, Nico. 2005. “Why Be Rational?,” Mind, 114: 509–62.
Nussbaum, Martha. 2001. Upheavals of Thought. Cambridge:  Cambridge
University Press.
Olson, Jonas. 2004. “Buck-Passing and the Wrong Kind of Reasons,” Philosophical
Quarterly, 54: 295–300.
Parfit, Derek. 2011. On What Matters. New York: Oxford University Press.
Rabinowicz, Wlodek, and Rønnow-Rasmussen, Toni. 2004. “The Strike of the
Demon: On Fitting Pro-Attitudes and Value,” Ethics, 104: 391–423.
Rabinowicz, Wlodek, and Rønnow-Rasmussen, Toni. 2006. “Buck-Passing and the
Right Kind of Reasons,” Philosophical Quarterly, 56: 114–20.
244 Justin D’Arms and Daniel Jacobson

Raz, Joseph. 2009. “Reasons: Practical and Adaptive,” in David Sobel and Steven
Wall (eds), Reasons for Action, 37–57. Cambridge: Cambridge University Press.
Roberts, Robert. 2003. Emotions:  An Essay in Aid of Moral Psychology.
Cambridge: Cambridge University Press.
Rousseau, Jean Jacques. 1987. Discourse on the Origins of Inequality, in The Basic
Political Writings, ed. and tr. Donald Cress, i.  25–109. Indianapolis:  Hackett
Publishing Co.
Scanlon, T. M. 1998. What We Owe to Each Other. Cambridge, MA: Belknap Press
of Harvard University Press.
Schroeder, Mark. 2010. “Value and the Right Kind of Reason,” Oxford Studies in
Metaethics, 5: 25–55. Oxford: Oxford University Press.
Schroeder, Mark. 2012. “The Ubiquity of State-Given Reasons,” Ethics, 122:
457–88.
Skorupski, John. 2007. “Buckpassing about Goodness,” in T. Rønnow-Rasmussen,
B. Petersson, J. Josefsson, and D. Egonsson (eds), Homage à Wlodek: Philosophical
Papers Dedicated to Wlodek Rabinowicz: <http://www.fil.lu.se/hommageawlodek/
index.htm>.
Skorupski, John. 2010. The Domain of Reasons. Oxford: Oxford University Press.
Stratton-Lake, Philip. 2005. “How to Deal with Evil Demons:  Comments on
Rabinowicz and Rønnow-Rasmussen,” Ethics, 115: 778–98.
Strawson, P. F. 1968. “Freedom and Resentment,” in Gary Watson (ed.), Free Will,
59–80. Oxford: Oxford University Press.
Taylor, Gabrielle. 1985. Pride, Shame, and Guilt:  Emotions of Self-Assessment.
New York: Oxford University Press.
Way, Jonathan. 2012. “Transmission and the Wrong Kind of Reason,” Ethics,
122: 489–515.
Weber, Michael. 2004. “Compassion and Pity:  An Evaluation of Nussbaum’s
Analysis and Defense,” Ethical Theory and Moral Practice, 7: 487–511.
 Index
abortion 99, 134
Adams, F.
intentions  22 nn9, 10
Strong Belief Thesis  22 n10
Anderson, R. Lanier.  149 n27
Stanford school  126 n3
Anscombe, E.
natural expression of
intentions 25, 26 n16
Very Weak Belief Thesis  36–7
anthropology 211 n20
Aristotle
emotion  236, 239 n30
misfortune and suffering  236
moral philosophical tradition  140
moral philosophy  138
moral values  152
Nicomachean Ethics 152–3
Nietzsche, influence on  138 n15
practical syllogism  17
slavery, “natural”  135 n13
WKR problem  235–9, 241
Armstrong, D. M.
disjunctive properties  168
Audi, R.
intention 28–9, 31, 40
non-naturalism 154 n4
Strong Belief Thesis  22
Baker, A.
number theory  191 n5, 192 n6
Baumeister, R. F. et al.
responsibility 216 n4
Bedke, M.
Coincidence Argument  102–24
debunking argument  x. 90 n27
evolution 83 n14
moral realism  x
non-naturalism  78 n4, 122 n19, 123
non-reductivism 78 n4
normative systems  83 n15
process of being "off-track"  90 n27
Beethoven, L. van  127, 143
Bentham, J. 134
Berker, S.
evolution 94 n32
Berry, J.
Pyrrhonian argument  147–8 n26
Bertrand, M. and Mullainathan, S.
stereotyping 96 n35
Bittner, R.
regret 220–2
Blackman, R.
subjectivism 216 n3
Bloomfield, P.
companions in guilt  146
Boghossian, P.
blamelessness 52–6
concept-constitution 53, 57
concept-possession ix
epistemic merit  45
meaning-entitlement
connection 46–9, 52–6
pejoratives 51 n19
warrant-transfer  49 n12, 52–6
willingness, language of  45 n1
Bourget, D.
philosophical views, survey of  146 n23
Boyd, R.
moral disagreement  140
Bratman, M.
cognitivism, criticism of  19, 21 n7
deliberation 37 n33
intention and belief  27 n18, 38
n34, 40
means-end coherence  23–4,
26, 27 n18
reasoning 12
Strong Belief Thesis  22 n10
Brink, D.
moral disagreement  140
moral explanations  133
Broome, J.
means-end coherence  20 n4
rationality  50 nn14, 15
reasoning to an intention  13–15
wide-scope requirements  18 n1
Brunero, J.
cognitivism about practical
rationality  ix, 18–42
means-end coherence  ix, 21 n7
Buddhism 240 n32
246
Chalmers, D. J.
doxastic possibility  34 n29
philosophical views, survey of  146 n23
Christensen, D.  100 n36
disagreement 136 n13
evidence for p 85 n20
independence requirement  81 n13
Christianity 239–41
Clark, M.  149 n
Nietzsche, interpretation of  126
Stanford school  126 n3
Clarke-Doane, J.  102 n, 149 n27
disagreement in mathematics  132 n8,
144 n21
mathematical Platonism  113 n10,
124 n21
cognitivism
about practical rationality  18–42
belief consistency  18–19
explanatory claim, problems for  Crisp, R.
39–42
intention consistency  18, 25
Means-End Coherence (MEC)  ix,
18–27, 32–5, 38–42
Strong Belief Thesis and  18–32
arguments for  25–32
cognitivism with  20–5
definition 18–19
unknown failures to intend  38–9
Very Weak Belief Thesis  36–8
Cohen, J.
moral depravity  133
Coincidence Argument (CA)  102–24
allodoxic possibilities  119–20
coincidence arguments and  104–6
conditions 102–3
counterpossibilities 105
defeat 114
epistemic internalism  106
evolutionary debunking and  121–2
generic skepticism  110–11
happiness 118, 122
inexplicable alignment  112–14
justification of normative
beliefs 111–14
modal truths  120–1
Moorean reply to  121–3
non-coincidence 107–8
non-metaphysicalism 122 n18
non-naturalism and  105–6, 111, 114,
116 n13, 120, 122–4
normative facts  106–9
obliviousness 114–16
to necessary truths  116–20
Index
perception 108
posteriori knowable identities  118 n15
random or unreliable
analogies 109–10
reliability 113
sensitivity requirement  114–15,
118, 120–1
skepticism and  122
skeptics vs. realists  108
subjunctives 119, 123–4
supervenience 117
truth-tracking 105
Constantinescu, C.
moral vagueness  xi, 152–82
Copp, D.  102 n, 183 n34
moral language  172 n22
truth 105 n5
vague properties  170 n21
Cornell realism  169
demon’s threat  220, 222
non-naturalism 154 n4
Cuneo, T.  100 n36, 102 n, 183 n34
non-naturalism 154 n4
Cuneo, T., and Shafer-Landau, R.
moral claims  93 n28
Dancy , J.
action and practical reasoning  1–17
belief 3
deliberation 6
non-naturalism 154 n4
resultance 156 n6
Danielson, S. and Olson, J.
Fitting Attitude (FA) theories  218 n8
D'Arms, J. and Jacobson, D.
feeling 222 n13
Fitting Attitude (FA) theories  217 n8
immoralism 224 n15
WKR problem  xii, 215–42
Darwall, S.
Fitting Attitude (FA) theories  217 n8
Darwin, C.
Darwinian Dilemma  102 n2, 105
evolutionary dilemma  76–7, 86 n21
Davidson, D.
Strong Belief Thesis  22 n10
Davis, W.
Strong Belief Thesis  22 n10
death penalty  134
debunking see evolutionary debunking
Doris, J. M.  149 n27
Doris, J. M. and Plakias, A.
moral disagreement  139, 144
 Index
Dougherty, T.  100 n36, 173 n25,
182 n34
facts 177
Dworkin, R.
as moral realist  xii
Hume's principle, use of  205, 210
Justice for Hedgehogs  187, 196–8, 206
legal and ethical questions  156 n8
mathematics 198–9
moral belief  81, 189, 196–213
moral facts and properties  155
moral indeterminacy  175 n28
moral luck  81
moral skepticism  205 n17
moral vagueness  156–7
Elga, A.  100 n36
good 85 n20
independence requirement  81 n13
possibility of error  83 n17
Enoch, D.
Coincidence Argument  103, 107, 112
evolution  81 n11, 107
moral truth  143 n20
non-naturalism 154 n4
evolutionary debunking  76–100
abortion 99
altruism 94–5
challenge of  76–7
Coincidence Argument and  121–2
debunker's argument  77–8
deontology 93–5
evaluative judgments  79–82
evaluative realism  87–9
evidence of error  98
evolution as problematic  96–7
good reason, principle of  82–7, 98
independent ground  98–9
Inverse Rule of Debunking  98–9
mistaken-ness  80 n9, 90–1
moral realism  89–93
off-track processes  90–1
perception 82–3
possibility of error  83–4
racism 92
skepticism and 80
undermining 97–100
utilitarianism 94–5
violence 93–4
Field, H.
debunking evaluative realism  89 n25
mathematical Platonism  113 n10
247
Finnis, J. 147
Fitting Attitude (FA) Theory, see Wrong
Kind of Reason (WKR) Problem
FitzPatrick, W.
Coincidence Argument  103
evolutionary debunking  79 n7
non-naturalism 154 n4
Fodor, J.
autonomy of sciences  133 n10
Freud, S. 205
Garfinkel, A.
explanations 191 n4
Gaut, B.
immoral comedy  220–7
Gewirth, A.  143 n18
Gibbard, A.
Coincidence Argument  103, 113
concept-constitutingness 57 n35
moral vagueness  174
nonanalytical naturalism  155 n5, 170
normative facts  113, 155 n5
ought, concept of  46 n4, 73
responsibility 216 n4
WKR skepticism  228 n20
Gillett, C. and Rives, B.
determinates/determinables 173
Goethe, J. W. von  127, 142–3
Greene, J.
deontological judgments  93 n30
emotion 94
moral intuition  95
violence 93
Greene, J. D. et al.
deontological judgments  104, 108 n7
Grice, P.
conversational pragmatics  22 n9
Strong Belief Thesis  22 n10
Hampshire, S. and Hart, H. L. A.
Strong Belief Thesis  22 n10
Harman, G.
bragging 22 n9
cognitivist approach  19
evidence for p 81 n12
explanatory indispensability test  186,
188–92, 197–8
flaming pussycat case  133–4
Strong Belief Thesis  21, 22 n10,
23–4
Hieronymi, P.
reciprocal lover  230 n22
RKR problem  230 n22, 232 n23
248
Hieronymi, P. (Cont.)
WKR problem  219 n11, 228 n20,
234 n24
history (discipline)  211 n20
Hitler, A. 133
Holton, R.
intentions  29 n22, 36 n32
Strong Belief Thesis  29 n22
Very Weak Belief Thesis  36 n32
Horwich, P.
metasemantic strategies  48 n9
Huemer, M.
Coincidence Argument  103
non-naturalism 154 n4
Hughes, G. E. and Cresswell, M. J.
instrumental belief  34 n28
Hume, D.  140–1, 143
Hurley, S.
disagreement 153 n2
Hussain, N.
fictionalist reading by  126 n2
Stanford school  126 n3
Hutcheson, F. 138
Hyde, D.
semantic vagueness  161
vague properties  161
Intelligent Design  148
Jackson, F.  214 n
locating' ethics  212
moral properties  169
non-analytical naturalism  170
Jacobson, D.
immoralism 224 n15
WKR problem  218 n8
Janaway, C.
Nietzsche, interpretation of  126
Jech, T.
set theory  132 n8
Joyce, R.
belief pills  109–10, 115
Coincidence Argument  103, 111
counterpossibilities 105
real facts  105
Kahane, G.
Coincidence Argument  103
evolutionary debunking  79 n7,
90, 95
Kant, I.  134, 135 n13, 136, 138, 140–4,
146 n23, 158, 176–7
Kaplow, L. and Shavell, S.
Index
fairness considerations  142
Kavka, G.
toxin puzzle  48 n10
Keefe, R.
supervaluationism 173–4
Kelly, T.  214 n
disagreement 136 n13
good 85 n20
Kim, J.
perception 193, 195
Kitcher, P.
moral skepticism  135 n13
scientific realism  136 n13
Kolodny, N.  214 n
diachronic requirement  49 n13
Enkratic Requirement  41 n39
instrumental rationality  42 n40
normative reasons  50 n15
RKR problem  232 n23
Kolodny, N. and Brunero , J.
means-end coherence  20 n5
possibility 35 n30
Korsgaard, C. M.  143 n18
Kramer, M.
unknowable obligations  176
n29, 181–2
Kripke, S.
evidence for p 81 n12
metaphysical necessity  200
legal positivism  147
Leiter, B.
best explanation argument  130, 138
debunking explanations  205 n16
egalitarian premise  143 n19
moral disagreement  126–49
moral skepticism  126–49
Nietzsche, interpretation
of x–xi, 126–49
Lewis, D.
belief revision  89 n25
causal explanation  191 n6
conversational score  162
disjunctive properties  168
moral vagueness  161
libertarianism 146 n23
Lillehammer, H.  183 n
reason and deliberation  178 n32
Lipton, P.
geometrical facts  190–2
Loeb, D.  147 n27
moral disagreement  131, 139, 142,
144, 147
 Index
MacIntyre, A.
meta-disagreement 147 n25
Mackie, J.
moral disagreement  140
objective values  212 n21
Marx, K. 205
Marxism 133, 148
mathematics see relaxed realism
McGrath, S.  179 n33, 183 n
knowledge 59 n37
moral realism  xi–xii, 186–213
Means-End Coherence (MEC) see
cognitivism
Mele, A.
free throw shooter  29 n21
Strong Belief Thesis  22 n10
Mellor, D. H.
natural properties  169, 172
metasemantic strategies  45–74
basic metasemantic idea  46–52
cognitive 'power or ability'  60
enkratic permissions  61
entitlement 47–8
Meaning Entitlement Connection
(MEC) 46–52
blamelessness and  52–6
moderate internalism  73–4
MPP-ish inferences  47–8, 53, 55 n30,
63, 66 n46
narrow-scope rational require-
ment  26–7, 41 n39, 50, 60–1
objections and responses to  65–71
Opposing Rule requirement  61–2, 65
ought-judgments  46, 49–50, 52, 57,
58 n36, 60–6, 69–74
perfect rationality and
non-accidentality 60–4
perfectly rational being  56–60
practical rationality and  49–50
principles of rationality  56–60
rational requirements  71–3
reasons-transferring principles  48–9
warrant transfer  54–5
wide-scope rational requirement  18
n1, 20, 26–7, 41 n39, 49
Mill, J. S.  138, 140–1
Moore, G. E.
Moore's paradox  25–6
skeptical/non-skeptical
scenarios 111, 121–2
moral realism see relaxed realism
moral vagueness  152–82
action-guidingness of morals  176
249
ambivalence  153 n1, 174
atomism 155–6, 182
baldness  152, 157, 161, 163–5, 168,
170, 173
belief-degrees 163 n15
borderlineness 157–9, 164
classical treatment of  152
cognitivism 155–6, 182
colour 166–70
conceptual derivatives  168
constitution  163, 165 n16, 166
conversational scores  162–3
correspondence 155–6, 182
courage  157, 167, 172–3
determinables 167 n19
determinates 167 n19
dilemma 159–61
disagreement and  152–4
disjunctive properties  168
epistemic conception  160–1
constraint on normative
reasons 179
ethical commitments  181
ethical facts  176–7
Euthyphronic region  163–5
harmfulness 159 n9
ignorance 175 n28
impermissibility  169, 178, 180
imprecise gradability  157
judgment-dependent matters  163–4
kindness  167, 172, 180
legal questions  156 n8
margin-for-error principle  175, 178
material constitution  181
maximally-improved rationality  178–9
moral facts  176–7
moral facts and properties  155
moral grounds  180–1
motivating reasons  177–8
myth of independent existence  161
non-metaphysical cognitivism  155 n5
non-naturalism and  154–6, 175–82
non-reductivism and  155–6, 173, 182
normative reasons  177–80
objectivism  155–6, 163, 182
ontic conception/view  160,
162, 166–70
open-mindedness 174 n27
permissibility  160–5, 167, 174–6,
178, 180
point-properties 167
powers of perceptual
discrimination 180
250
moral vagueness (Cont.)
psychological accounts  163 n15
quasi-realism and  175–6
range properties  167
Rationalism and  155–6, 177–80, 182
reasons to believe in  156–9
resemblance 168, 172
responsiveness 163
resultance 156 n6
semantic account of  160–1, 162–6
sharp moral properties  167, 171–82
epistemicism and  175–82
pointilism 171–3
supervaluationism 173–5
soriticality 157–9
supervaluationism and  162
sharp moral properties and  173–5
supervenience and  155–6, 165, 181–2
proportionality constraint  181–2
tallness 166–7
tolerance 157–9
trope-theory 164–5, 171–3
uncertainty 175 n28
unknowable-truths objection  176, 181
vague moral predicates  156
vague moral properties  161–70
weight 166
Moran, R.
gaps 8
Mussolini, B.  134 n12, 216
Nagel, T.
moral reasoning  203, 211 n20
relaxed realism  xii, 187 nn1–2
natural law theory  147
Nazism 144
Nietzsche, F.
debunking explanations  205
moral skepticism and
disagreement  x–xi, 126–49
arguments for  132–5
astrology 130
attitude-dependent facts  128 n4
best explanation argument  130, 132
current relevance  148–9
epistemic peers  136 n13
epistemological systems  146
feelings 137 n14
great philosophies  146 n23
meta-disagreement 147–8
middle books  129
moral properties or facts  128
moral propositions  131, 134–8
Index
Nietzsche's version  136–9
objections to  139–44
objectivity of morality  126–8
objectivity of value  127, 129
optimistic induction  136 n13
Parfit, similarities with  131 n7
pejorative morality (MPS)  127–8
philosophical critique  145–8
Platonism 128–9
prudential value  127
readings 126–7
relationalism 127–8
religion and intuitions  141 n16,
143 n20, 145
scope and grounds  127–32
supervenience 132 n9
utilitarianism  134, 135 n13, 136,
140, 142
vengeful thinking  130
Norcross, A.  141 n16
Nussbaum, M.
Aristotelian view of
emotions 235–9, 241
Fitting Attitude (FA) theory  242
misfortune and suffering  235 n27
pity and compassion  235 n26
Okasha, S.
biological altruism  95 n33
Olson, J.
A-referential reasons  230
Fitting Attitude (FA)
theory 217–18 n8
Paakkunainen, H.
metasemantic strategies  ix, 45–74
Parfit, D.
evaluative beliefs  80–2, 91, 104 n4
evolutionary debunking  121–2
followability thesis  232 n23
indeterminate questions  154
mathematics 188 n3
moral disagreement  131 n7, 153–4
moral facts and properties  155 n5
moral theory, optimism for  140–1,
143 n20
Nietzsche's moral skepticism
and 131 n7
non-metaphysical view  155 n5
non-naturalism  154–5, 179, 188 n3
Rationalism, thesis of  177
relaxed realism  xii, 187
On What Matters 187
 Index
state-given reasons  48 n10, 230, 232
sub-planning 38
WKR problem  228 n20, 234 n24
Peacocke, C.  causal efficacy  195–6
concept-possession ix, 45
metasemantic.strategies ix, 45
truth-preservation 54 n28
willingness, language of  45 n1
Peirce, C. S.  179
Plato
mathematical Platonism  113 n10, 123
moral philosophy of  136–8
Platonic objects  124
Platonism about value  128–9
Poellner, P.
Nietzsche, interpretation of  126
Pollock, J. and Cruz, J.
undercutting vs. rebutting
defeaters 210 n18
practical reasoning  1–17
action and belief  3–8
irrelevance of differences  8–9
to an intention  13–17
theoretical reasoning and  1–3
practical conclusions  9–13
Pryor, J.
hapless subjects  54–5
warrant-transfer 54–5
Putnam, H.
explanations 191 n4
skeptical meta–induction  135 n13
Pythagoreans 152
Quine, W.
epistemology of
mathematics 189–90, 192
evolutionary debunking  79
Rabinowicz, W. and
Rønnow-Rasmussen, T.
Fitting Attitude (FA) theory  217 n8,
218 n10
reciprocal lover  230 n22
racism  92, 135 n13
Railton, P.
historical explanation  133 n11
moral disagreement  153 n2
moral realism  139
relationalism 127–9
Raz, J.
followability thesis  232
practical reasoning  9–12, 16–17
underdetermination 12–13
251
relaxed realism  186–213
affirmative action  206–10
anti-realism and  196 n8
‘companions in guilt’  195
counterfactual questions  197–8
counterfactuals 196–200
debunking  186, 204, 211, 213
Dworkinian perspective  196–200
evolutionary biology  191 n5
explanatory hypothesis  202–3, 205
geometrical facts  190–1
Harman's test  186, 188–92
human perception  195
Hume's principle  205–6, 210
irreducible normativity  186
mathematics 187–8, 212–13
mathematical properties  192
mathematical truth  190–2, 198–9,
211 n20
modal status  200
moral claims  186–7
moral properties  186
moral/non–moral
considerations 204–11
natural sciences  189
number theory  191 n5
objective values  212 n21
objectivity and truth  206
perceptual accessibility  192–5
pointlessness 200–3
psychological explanations  191,
196–200
set theory  188 n3
undercutting vs. rebutting
defeater 210–11
religion 239–41
and intuitions  141 n16, 143
n20, 145
Richardson, J.
Nietzsche, interpretation of  126
Roberts, R.
Christian view of pity  235–6, 239–42
Roosevelt, F. D.  134 n12
Rosen, G. and Smith, N. J.  J.
sharp properties  167, 171
vague moral properties  161, 166–7
Ross, J.
intention-belief connection  40, 55 n31
resultance, notion of  156 n6
Rousseau, J.–J.
emotions 236
pity  236 n28, 239 n30
252 Index

Ruse, M. epistemicism  175 n28, 176 n30

Coincidence Argument  103 moral vagueness  153, 156, 161 n12,
Russell, B. 164–5, 171, 175 n28
vague moral properties  161 non-naturalism  154–6, 171, 181
pencil example  181
Sainsbury, R. M. Rationalism, thesis of  177
moral vagueness  152 realism, definition of  76 n1, 179
vague moral properties  161 Shapiro, S.
Sanford, D. H. moral vagueness  161–6
determinates vs. Sider, T.
determinables 167 n19 unknowable obligations  176 n29
disjunctive properties  168 Silk, A.
vague moral properties  166 n17 attitude-dependent facts  128 n4
Sayre-McCord, G. Singer, P.
moral depravity  133 ethics and intuitions  94, 104, 108
Scanlon, T. M. n7, 165
Being Realistic about Reasons 187 Sinhababu, N.
mathematics 189 vengeful thinking  130
moral facts and properties  155 Sinnott-Armstrong, W.
objective values  212 n21 debunking arguments  104
relaxed realism  xii, 87, 189 Skarsaune, K. O.
set theory  188 n3 Coincidence Argument  103, 107
WKR problem  231 Skorupski, J.
Schafer, K. WKR problem  218 n8, 228 n20, 229,
Coincidence Argument  103, 112 n8 232 n23
Schechter, J. and Enoch, D. slavery  135 n13, 198, 200–1, 213
metasemantic strategies  48 n9 Sliwa, P.  100 n36
Schiffer, S. evolutionary debunking  79 n7
vague moral properties  161, 163 n15 Smith, M.
Schopenhauer, A.  137–8 brain surgery  15
Schroeder, M. differential responsiveness  72 n55
Enkratic Requirement  41 n39 Dworkin’s conception of moral
state-given v. object-given skepticism 205 n17
reasons 48 n10 Smith, R.
WKR problem  219 n11 Nietzsche, interpretation of  138 n15
Schroeter, F. Sober, E.
ought, concept of  46 n4 Quinean approach  190
Schwitzgebel, E. Socrates 152
psychology and mistaken Sophists 137
belief 38 n35 Sorensen, R.
sciences 133 n10 moral vagueness  159
Sedgwick, A. 141 unknowable obligations  176 n29
Setiya, K.  74 n59 Sosa, E.  149 n27
cognitivist approach  19, 20 nn5, 6, 21 disagreement 153 n2
n7, 39 n36, 55 n31 Spinoza, B. de  136, 141, 146 n23
sexism 135 n13 Stoicism  138, 240 n32
Shafer-Landau, R.  183 n Stratton-Lake, P.
Coincidence Argument  103 Fitting Attitude (FA)
companions in guilt  146 theory 217–18 n8
constitution, conception of  165 Strawson, P. F.
n16, 181 Fitting Attitude (FA) theory  217 n8
Dworkin’s use of Hume’s Street, S.
principle 205 n17 Coincidence Argument  103, 109
 Index 253
evolutionary debunking/Darwinian perfectly rational being  56–60
Dilemma  76 n1, 77–83, 86 rational disposition  56 n33
nn21–2, 87 n23, 88–91, 95, 102 reduction 57 n34
n2, 104–5, 111 supernatural deities  59 n37
Jupiter case  109–10, 115 White, A. R.
non-natural normative facts  105 belief 2
reliability (good/no good White, R.  100 n36
accounts) 108 n7 independence requirement  81 n13
Strong Belief Thesis see cognitivism Wielenberg, E. J.
supervaluationism see moral vagueness cognitive processes  107
supervenience, see Coincidence Coincidence Argument  103, 107
Argument (CA); moral vagueness; evolutionary debunking  81 n11
Nietzsche, F. Wilberforce, W. 201
Williams, B.
Taylor, G. moral luck  81
responsibility 216 n4 sub-planning 38
terrorism 239–40 unknowable obligations  176 n29
theism 146 n23 Williamson, T.
Thomism (St. Thomas Aquinas)  140 concept-possession 56 n32
Tye, M. conditionals 47 n8
vague moral properties  161 epistemicism  175 n28, 176
metasemantic arguments  45 n2
utilitarianism  94–5, 134, 135 n13, 136, permissibility 162 n14
140, 142 supervaluationism 173 n26
Wittgenstein, L.
vagueness see moral vagueness skepticism 84 n18
van Roojen, M. Wolf, S.
individual agents  74 n58 disagreement. 153 n2
Vasile, L. J.  182 n34 Wright, C.
moral disagreement  153 n2 skepticism 84 n18
Vavova, K. Wrong Kind of Reason (WKR)
evolutionary debunking  x, 76–100 Problem 215–42
Velleman, J. D. A-referentiality 230–1
cognitivist approach  19 blame 235–42
Strong Belief Thesis  22 n10, Christian suffering  239–41
24–31 compassion 235–42
Very Weak Belief Thesis see cognitivism demon's threat  220–1
desire 234
Wallace, R. J. emotion 235–42
cognitivist approach  19–20, fault 235–42
32–5, 40, 42 Fitting Attitude (FA) theory  216 n3,
Way, J. 217–19, 240 n32, 242
WKR skepticism  228 n20 followability thesis  228, 232–3
Weatherson, B. immoral comedy/comic
evidence for p 85 n20 moralism 220–7
Weber, M. immoralism 224 n15
compassion and pity  238 incentives 233, 242
Wedgwood, R.  149 n27 No True Scotsman Fallacy  239 n30
cognitive 'power or ability'  60 object/state distinction  228–31
concept-possession 52 obvious WKRs  220–8
enkratic disposition  58 n36 opacity of normative force  217–20,
non-naturalism 154 n4 227, 228–35
ought, concept of  46, 73 pity 235–42
254 Index

Wrong Kind of Reason (Cont.) subjectivism  216 n3, 218 n9

reason redescription program  228–9 suffering 235–42
regret, argument against  220–1 terrorism 239–40
Right Kind of Reasons (RKR) victimhood 238
and 215–20 WKR skepticism  228–30