Ethics and Morals Phylosophy

Ethics and Moral Philosophy
Studies in Moral Philosophy

Series Editor
Thom Brooks (Newcastle University)

Editorial Board
Chrisoula Andreou (University of Utah)

Mark Bevir (University of California, Berkeley)
Clare Chambers (University of Cambridge)
Fabian Freyenhagen (University of Essex)
Tim Mulgan (University of St Andrews)
Ian Shapiro (Yale University)
VOLUME 1
The titles published in this series are listed at brill.nl/simp

Edited by
Thom Brooks
LEIDEN BOSTON
LEIDEN BOSTON
2011
This book is printed on acid-free paper.

Library of Congress Cataloging-in-Publication Data
Ethics and moral philosophy / edited by Thom Brooks.
p. cm. -- (Studies in moral philosophy ; v. 1)
Includes bibliographical references and index.
ISBN 978-90-04-20342-6 (hardback : alk. paper) 1. Ethics. I. Brooks, Thom.
BJ21.E855 2011
170--dc23
2011034741
ISSN 2211-2014
ISBN 978 90 04 203426
Copyright 2011 by Koninklijke Brill NV, Leiden, The Netherlands.
Koninklijke Brill NV incorporates the imprints Brill, Global Oriental, Hotei Publishing,
IDC Publishers, Martinus Nijhoff Publishers and VSP.
All rights reserved. No part of this publication may be reproduced, translated, stored in
a retrieval system, or transmitted in any form or by any means, electronic, mechanical,
photocopying, recording or otherwise, without prior written permission from the publisher.
Authorization to photocopy items for internal or personal use is granted by KoninklijkeBrillNV
provided that the appropriate fees are paid directly to The Copyright Clearance Center,
222 Rosewood Drive, Suite 910, Danvers, MA 01923, USA.
Fees are subject to change.
CONTENTS
Notes on Previous Publicationvii
List of Contributors....................................................................................ix
Introduction 1
part I
Practical Reason
Standards, Advice, and Practical Reason11
Chrisoula Andreou
Does Rationality Consist in Responding Correctly to
Reasons?25
John Broome
Practical Reason, Value and Action57
Alison Hills
Normativity and Practical Judgement77
Onora ONeill
part II
Particularism
Ethics Without Reasons?95
Roger Crisp
Defending the Right107
Jonathan Dancy
part III
Moral Realism
Moral and Theological Realism: The Explanatory Argument125
Russ Shafer-Landau
Anti-Reductionism and Supervenience147
Michael Ridge
vi
contents
part IV
Virtue Ethics
Han Feizis Criticism of Confucianism and Its Implications
for Virtue Ethics173
Eric L. Hutton
Aristotelian Virtue and the Interpersonal Aspect of Ethical
Character207
Maria W. Merritt
Good but Not Required?Assessing the Demands
of Kantian Ethics237
Jens Timmermann
Virtue, Character and Situation259
Jonathan Webber
part V

Doing Harm, Allowing Harm, and Denying Resources287
Timothy Hall
Time-Relative Interests and Abortion317
S. Matthew Liao
The Basis of Human Moral Status335
S. Matthew Liao
The Mixed Solution to the Number Problem357
Martin Peterson
Trivial Sacrifices, Great Demands371
William Sin
Essentialism and Anti-Essentialism in Feminist Philosophy385
Alison Stone
Bibliography409
Index417
Notes on previous publication

Chrisoula Andreou, Standards, Advice, and Practical Reason, Journal
of Moral Philosophy 3(1) (2006): 5767.
John Broome, Does Rationality Consist in Responding Correctly to
Reasons? Journal of Moral Philosophy 4(3) (2007): 34974.
Alison Hills, Practical Reason, Value and Action, Journal of Moral Philosophy 4(3) (2007): 37592.
Onora ONeill, Normativity and Practical Judgement, Journal of Moral
Philosophy 4(3) (2007): 393405.
Roger Crisp, Ethics Without Reasons? Journal of Moral Philosophy
4(1) (2007): 40-49.
Jonathan Dancy, Defending the Right, Journal of Moral Philosophy
4(1) (2007): 8598.
Russ Shafer-Landau, Moral and Theological Realism: The Explanatory
Argument, Journal of Moral Philosophy 4(3) (2007): 31129.
Michael Ridge, Anti-Reductionism and Supervenience, Journal of
Moral Philosophy 4(3) (2007): 330-48.
Eric Hutton, Han Feizis Criticism of Confucianism and Its Implications for Virtue Ethics, Journal of Moral Philosophy 5(3) (2008):
42353.
Maria W. Merritt, Aristotelian Virtue and the Interpersonal Aspect of
Ethical Character, Journal of Moral Philosophy 6(1) (2009): 2349.
Jens Timmermann, Good but Not Required?Assessing the Demands
of Kantian Ethics, Journal of Moral Philosophy 2(1) (2005): 927.
Jonathan Webber, Virtue, Character and Situation, Journal of Moral
Philosophy 3(2) (2006): 193213.
viii
notes on previous publication
Timothy Hall, Doing Harm, Allowing Harm, and Denying Resources,

Journal of Moral Philosophy 5(1) (2008): 5076.
S. Matthew Liao, Time-Relative Interests and Abortion, Journal of
Moral Philosophy 4(2) (2007): 24256.
S. Matthew Liao, The Basis of Human Moral Status, Journal of Moral
Philosophy 7(2) (2010): 15979.
Martin Peterson, The Mixed Solution to the Number Problem, Journal
of Moral Philosophy 6(2) (2009): 16677.
William Sin, Trivial Sacrifices, Great Demands, Journal of Moral Philosophy 7(1) (2010): 315.
Alison Stone, Essentialism and Anti-Essentialism in Feminist Philosophy, Journal of Moral Philosophy 1(2) (2004): 13553.
List of Contributors
Chrisoula Andreou is Associate Professor in Philosophy at the
University of Utah.
Thom Brooks is Reader in Political and Legal Philosophy at Newcastle
University. He is editor and founder of the Journal of Moral Philosophy.
John Broome is Whites Professor of Moral Philosophy and Fellow of
Corpus Christi College at the University of Oxford.
Roger Crisp is Professor of Philosophy and Fellow of St Annes College
at the University of Oxford.
Jonathan Dancy is Professor of Philosophy at both the University of
Reading and the University of Texas-Austin.
Timothy Hall is Associate Professor in Philosophy at Oberlin College.
Alison Hills is Lecturer in Philosophy and Fellow of St Johns College at
the University of Oxford.
Eric Hutton is Associate Professor in Philosophy at the University of
Utah.
S. Matthew Liao is Associate Professor in the Center for Bioethics at
New York University.
Maria W. Merritt is Assistant Professor at the Johns Hopkins Berman
Institute of Bioethics.
Onora ONeill is Professor Emeritus in Philosophy and former Principle
of Newham College at the University of Cambridge. Baroness ONeill
of Bengarve is a Crossbench Peer in the House of Lords.
Martin Peterson is Associate Professor in Philosophy at Eindhoven
University of Technology.
list of contributors
Michael Ridge is Professor of Moral Philosophy at Edinburgh

University.
Russ Shafer-Landau is Professor of Philosophy at the University of
Wisconsin-Madison.
William Sin is an Assistant Professor at the Hong Kong Institute of
Education.
Alison Stone is Reader in European Philosophy at Lancaster University.
Jens Timmermann is Reader in Moral Philosophy at the University of
St Andrews.
Jonathan Webber is Senior Lecturer in Philosophy at Cardiff University.
Introduction
Thom Brooks
Ethics and moral philosophy is an area of particular interest today. This
book brings together some of the most important essays in this area.
The essays have all appeared recently in the Journal of Moral Philosophy,
an internationally recognized leading philosophy journal. This book is
divided into five parts: practical reason, particularism, moral realism,
virtue ethics, and ethics and moral philosophy more generally. My
introduction will give some background to these areas and introduce
the essays in this book.
Part I: Practical Reason
The first four essays focus on the topic of practical reason. Chrisoula
Andreou (2006) asks the question: is there a mode of sincere advice in
which the standards of the adviser are put aside in favor of the standards of the advisee? Andreou argues that the advisers standards should
not be set aside when offering sincere advice to others. Her claim is that
it is a mistake to hold that practical reason judgements evaluate actions
from within the agents system of standards as opposed to the system of
the judgers.
The second essay is by John Broome (2007).1 He asks whether rationality consists in responding correctly to reasons. Many philosophers
answer this question in the affirmative: rationality does consist in
responding correctly to either reasons or beliefs about reasons. Broome
argues that this widespread understanding is wrong. Rationality is not
about correctly responding to reasons or beliefs about reasons. Instead,
I am rational where I respond correctly to my beliefs about reasons.
1
This essay was originally published in a special issue on Metaethics arising from a
special conference at the University of Cambridge organized by Fabian Freyenhagen
and Hallvard Lillehammer. Essays included in this book include Broome (2007), Hills
(2007), ONeill (2007), Ridge (2007), and Shafer-Landau (2007).
thom brooks
Alison Hills (2007) considers how we might determine which theory

of practical reason is correct. Many philosophers have tried to link
theories of practical reason with success conditions of action. Hills
argues that this is link has many benefits, but it cannot be a substitute
for relating practical reason to value.
The final essay in this part concerns normativity and practical judgement. Onora ONeill (2007) considers the relationship between norms
and action guidance. Often norms are considered indeterminate which
acts are particular and determinate. This has led many to question
whether norms can helpfully specify particular actions to be done: how
may we link indeterminate norms with determinate acts? ONeill
argues that practical judgement is a part of practical reasoning as its
aim is to integrate (and not prioritize or trade-off) a plurality of norms.
This project of taking norms seriously is understood as a matter of
working towards practical judgement and ONeill offers several
insights on how this may be best realized.
Part II: Particularism
The second part includes a debate between Roger Crisp and Jonathan
Dancy. This debate took place within a special issue dedicatedtoDancys
work on particularlism.2 Crisp (2007) offers a critique of Dancys arguments for strong holism about reasons. This critique is based upon several considerations including the view such as the genuine distance
between this holism and generalism and the concern that Dancy rests
his case on common-sense morality without justification. Weak holism
has problems of its own, namely, that it offers a generalist conception of
a normative reason. Crisp concludes that generalism, then, is true.
Dancys chapter is a direct response to the chapter by Crisp, as well
as other philosophers. Dancy (2007) begins his chapter by addressing
Crisps many criticisms. While there is much the two may agree on,
disagreement arises on whether ultimate reasons may be invariant:
Dancy argues they belong in Wonderland. Dancy then proceeds to
address other criticisms of his particularist position with reference to
work by Sean McKeever and Michael Ridge on default reasons, as well
as Richard Norman on reasons for action. Through responding to
critique Dancy reveals some new understandings of the particularist
2
This issue was Journal of Moral Philosophy volume 4, number 1 published in 2007.
introduction3
project that move beyond his earlier work into further interesting
domains.
Part III: Moral Realism
Part three concerns moral realism and an indirect debate between Russ
Shafer-Landau and Michael Ridge. Both essays were originally presented together in a panel at Journal of Moral Philosophy co-sponsored
conference on Metaethics.3 Shafer-Landau addresses sceptical arguments against theological realism and moral realism. He turns his focus
to an explanatory argument: moral facts lack causal-explanatory power
and, thus, we have most reason to deny the existence of moral facts.
Shafer-Landau argues that there is no parity between these realisms.
The vulnerabilities of one are rather different from the other and the
theological realist has the greater vulnerability. Moral realism may be
in a more safe position than we initially thought.
Michael Ridge does not directly address this particular line of argument, but he does focus on Shafer-Landaus then current work on
moral realism. Ridge focusses on Shafer-Landaus attempts to explain
supervenience with reference to his mora realism. Ridge argues that
these attempts are unsuccessful. Moreover, there are several implausible conclusions that follow from it. Ridge argues that we must have
greater caution in applying explanations for supervenience in the area
of the philosophy of mind in other contexts, such as metaethics.
Part IV: Virtue Ethics
The fourth part includes three chapters on virtue ethics. The first is
very different from the other two. This part begins with a fascinating
and highly illuminating essay by Eric Hutton. Hutton (2008) explores
the implications for contemporary (and largely Western) work in the
area of virtue ethics arising from Han Feizis critique of Confucianism.
This chapter will be of great interest to those interested in virtue ethics,
comparative philosophy, Chinese philosophy, and the non-Western
thought more generally.4 Confucius has been seen to offer a theory of
See footnote 1.
It has long been a goal of the Journal of Moral Philosophy to offer a mainstream
venue for work in non-Western philosophy. We continue to work arduously to attract
3
4
thom brooks
virtue ethics. Hutton considers the work of Han Feizi, an early critic.
Huttons project is to investigate whether these criticisms of Confucius
and his theory of virtue ethics offer revealing criticisms of virtue ethics
as understood today. Hutton answers in the affirmative and he argues
that this work helps highlight a new problem overlooked by current
virtue ethicists.
The next two chapters focus on character and virtue ethics. Maria
Merritt (2009) examines whether the Aristotelian idea of character as
firm and unchanging relates well to the empirical reality. She argues
that we should focus our attention primarily on substantive ethical
considerations and not characterological assessments. JonathanWebber
(2006) also considers our character in light of new developments in
experimental psychology.5 He argues that the traditional account has
much to recommend it. Experimental psychologists assume a behavioralist construal of traits in terms of stimulus and response rather than
inclinations. The traditional account has greater predictive power and
rests on a more defensible foundation.
Part V: Ethics and Moral Philosophy
This final fifth part brings together seven chapters that stretch
across ethics and moral philosophy more broadly. Timothy Hall (2008)
examines the oft-claimed moral distinction made by many non-
consequentialists between doing and allowing harm. Hall introduces a
third distinction, namely, denying resources. He argues that this third
distinction is separate from the first two and an important feature that
non-consequentialists should take into account.
The next two chapters are by S. Matthew Liao. First, Liao (2997)
considers the concept of time-relative interest originally introduced by
Jeff McMahan. This concept is meant to help us solve certain puzzles
about the badness of death, but it has also been used to explain how
abortion is permissible. Liao challenges this finding on a number of
grounds and argues that the justification of abortion must be found on
different grounds. In his second chapter included here, Liao (2010)
turns our focus to the question of what it is about our moral status as
new submissions in this area and it is an area of great personal interest as well (see
Brooks 2002a).
5
Webbers essay was the winner of our only prize for best article published in the
Journal of Moral Philosophy for 2006.
introduction5
human beings that gives this status such importance. Is this merely
speciesism? Liao argues in favour of what he calls the genetic basis for
moral agency as an account of rightholding. This account has the
attraction of accounting for all human beings without advocating some
form of speciesism.
Martin Peterson (2009) examines the number problem (see Brooks
2002b). The number problem is how to best decide whether to save
group x or y where these groups lack no morally relevant differences
and you may only choose one or the other. Different options are considered: we should save the greatest number, we should toss a fair coin
or we should set up a weighted lottery. These are all jettisoned in favour
of a new mixed solution where both fairness and the number of persons to be potentially saved count.
The following chapter is about our duties to those in severe poverty.
William Sin (2010) considers what follows if persons in affluent countries could easily save those in severe poverty in other countries. Sin
argues that accepting a principle of easy rescue whereby it is wrong for
these people not to provide easy rescue may lead to a problem of
demandingness. In such cases, what may first appear to be trivial sacrifices may result in great demands.
Alison Stone (2004) examines the essentialism debates of feminists.
Are there properties that are essential to women which all women
share? Stone argues against essentialism and instead favours the view
that women share a genealogy: women acquire feminity by appropriating and reworking existing cultural interpretations of feminity. This
anti-essentialist position is not only more sound than essentialism, but
it may also be more easily reconciled with feminist politics.
The final essay is on the demands of Kantian ethics. Jens Timmermann
(2005) assesses the alleged demandingness of Kantian ethics concerning our duties in a highly illuminating essay that defends Kantian ethics from several criticisms made of it.
Conclusion
A final comment should be made about how these essays were originally selected for publication. All chapters originally appeared in recent
issues of the Journal of Moral Philosophy. I founded the journal with
Fabian Freyenhagen in 2003 and it launched the following year. The
JMP has been published quarterly since 2009. All submissions were
thom brooks
subjected to rigorous anonymous peer review by our international editorial board and referees. We publish a list of our referees in the final
issue of each annual volume. Our standards are high and our acceptance rate is less than 7%. We receive submissions from all over the
world with the great majority (about two-thirds) coming from the
United States or Great Britain. Furthermore, we have endeavoured to
ensure high quality review standards with swift turnarounds and we
are normally able to vet about 90% of all submissions in three months
with about 85% reviewed in two months or less.
The essays in this volume have satisfied our high standards and all
have appeared in recent issues of the Journal of Moral Philosophy ensuring their quality and their timeliness. Each was selected for this book
because of its real contribution to the general topic. Rather than offer a
single narrative, the essays instead may be read in any order and present a number of important perspectives and insights on the topic that
should help provide further clarifying illumination on central debates
and ideas in the field. Readers interested in learning more about what
other essays have been published in the JMP should consult this books
bibliography towards the end where the full publication details of all
articles is listed. Anyone interested in submitting new work for future
issues should submit through our online submission system found on
our website.6
This book is amongst the first in our new Studies in Moral Philosophy
book series to be published by Brill, the publishers of the JMP. The
book series aspires to fulfil the same high standards and quality of its
sister journal. We will aim to publish leading work in the areas of moral,
political, and legal philosophy. This book helps launch this new series.
I hope you will enjoy reading the essays in this book as much as
Ihave. Many thanks are due to Suzanne Mekking at Brill for her strong
support of this book and new book series. I am also most grateful to
Fabian Freyenhagen for his assistance over the years, as well as to our
advisory committee for this book series. Finally, my most sincere
thanks to the authors for choosing the Journal of Moral Philosophy for
their work. The journals success is primarily not through any particular effort of its editors, but rather a reflection of the high quality of our
authors contributions. We look forward to producing further volumes
on other topics.
6
The Journal of Moral Philosophy website may be found here: http://www.brill.nl/
jmp.
introduction7
Bibliography
Andreou, Chrisoula (2006). Standards, Advice, and Practical Reason, Journal of Moral
Philosophy 3(1): 5767.
Broome, John (2007). Does Rationality Consist in Responding Correctly to Reasons?
Journal of Moral Philosophy 4(3): 34974.
Brooks, Thom (2002a). In Search of Shiva: Mahadeviyakkas Virashaivism, Asian
Brooks, Thom (2002b). Saving the Greatest Number, Logique et Analyse 45(177/178):
5559.
Crisp, Roger (2007). Ethics Without Reasons? Journal of Moral Philosophy 4(1):
4049.
Dancy, Jonathan (2007). Defending the Right, Journal of Moral Philosophy 4(1): 8598.
Hall, Timothy (2008). Doing Harm, Allowing Harm, and Denying Resources, Journal
of Moral Philosophy 5(1): 5076.
Hills, Alison (2007). Practical Reason, Value and Action, Journal of Moral Philosophy
4(3): 37592.
Hutton, Eric (2008). Han Feizis Criticism of Confucianism and Its Implications for
Virtue Ethics, Journal of Moral Philosophy 5(3): 42353.
Liao, S. Matthew (2007). Time-Relative Interests and Abortion, Journal of Moral
Liao, S. Matthew (2010). The Basis of Human Moral Status, Journal of Moral Philosophy
7(2) (2010): 15979.
Merritt, Maria W. (2009). Aristotelian Virtue and the Interpersonal Aspect of Ethical
Character, Journal of Moral Philosophy 6(1): 2349.
ONeill, Onora (2007). Normativity and Practical Judgement, Journal of Moral
Peterson, Martin (2009). The Mixed Solution to the Number Problem, Journal of
Moral Philosophy 6(2): 16677.
Ridge, Michael (2007). Anti-Reductionism and Supervenience, Journal of Moral
Shafer-Landau, Russ (2007). Moral and Theological Realism: The Explanatory
Argument, Journal of Moral Philosophy 4(3): 31129.
Sin, William (2010). Trivial Sacrifices, Great Demands, Journal of Moral Philosophy
7(1): 315.
Stone, Alison (2004). Essentialism and Anti-Essentialism in Feminist Philosophy,
Journal of Moral Philosophy 1(2): 13553.
Timmermann, Jens (2005). Good but Not Required?Assessing the Demands of
Kantian Ethics, Journal of Moral Philosophy 2(1): 927.
Webber, Jonathan (2006). Virtue, Character and Situation, Journal of Moral Philosophy
3(2): 193213.
part I
Practical Reason
Standards, Advice, and Practical Reason

Chrisoula Andreou*
1.Introduction
It is typical for a human being that has reached maturity to have a
system of standards for evaluating actions that, taken as a whole, concerns not only her own actions, but also the actions of others. The
standards of one individual can conflict deeply with the standards of
another. Someone might, for example, judge that her spouse should
become a vegetarian, based on her conviction that, other things equal,
people should not eat animals if they can nourish themselves well without doing so. Her spouse, however, might judge that it is fine to eat
animals, so long as one monitors ones purchases in order to avoid supporting cruelty against animals. Or, a parent might judge that his child
should find a steady job rather than try to make it as a filmmaker, based
on the conviction that a life of security is preferable to a life of striving.
His child, however, might judge that theres no profit in trading a life of
striving for a life of security, just as, to quote Socrates, theres no profit
in someone selling his son or daughter into slaveryslavery under savage and evil menfor even a great deal of money.1
An individuals standards for evaluating actions can conflict not only
with the standards of others, but also with her own felt inclinations and
her own intentions. Someone might, for example, have a hankering for
veal, and form the intention to have some, even though it follows from
her standards that, given what the production of veal involves, she
should refrain from eating veal. Or, someone might desperately want
his wife to break ties with a certain acquaintance, and beg her to do so,
* My thanks to Noell Birondo, Sarah Buss, Earl Conee, John Dreher, Stephen Finlay,
Bruce Landesman, Elijah Millgram, Russ Shafer-Landau, Cynthia Stark, Mariam
Thalos, Mike White, and two anonymous referees for valuable comments on earlier
versions of this paper. Some of the ideas in this paper were presented at the Eastern
Division APA meeting in December 2002 and at the Pacific Division APA meeting in
March 2004. I appreciate the helpful feedback I received during these meetings.
1
Plato, Republic (trans. G.M.A. Grube; Indianapolis: Hackett Publishing Co., 1986),
589e.
12
chrisoula andreou
even though, according to his standards, his request is inappropriate

and her agreeing to it would be unduly indulgent.
Note that, as I am using the notion of a standard, standards go hand
in hand with attitudes of approval and disapproval. If an individual
prefers hiking to watching films, but does not think that hiking is more
worthy of approval than watching films, then, according to my terminology, his preference for hiking over watching films reflects his tastes
rather than his standards. If, on the other hand, an individual ranks
exercising above smoking because she thinks that exercising is more
worthy of approval than smoking, then, according to my terminology,
this ranking reflects her standards, which may or may not fit neatly
with her tastes. Note that insofar as one need not see all vices as moral
deficiencies, ones system of standards need not be exhausted by ones
moral standards. If, for example, one sees smoking as a vice but not as
immoral, then ones system of standards includes more than ones
moral standards.
In evaluating ones own actions and arriving at all-things-considered
judgments about what one should do, one is bound to see ones own
system of standards as necessarily relevant. Yet, in arriving at an allthings-considered judgment about what another should do, there are,
it seems, two approaches that one can take: one can evaluate things
from within ones own system of standards; or one can evaluate things
from within the others system of standards. This at least is what is suggested by Gilbert Harman, who distinguishes between evaluative judgments in which there is a relativity to the judgers values, where the
judger is the person doing the evaluating, and evaluative judgments in
which there is a relativity to the agents values, where the agent is the
person the evaluative judgment is about.2
The idea that one can evaluate an action from within the agents system of standards, even if the agents system of standards conflicts with
ones own, underlies the following thesis concerning advice:
Thesis 1: There is a mode of sincere advice which is such that, in cases of
conflict, the standards of the adviser are put aside and actions are evaluated from within the advisees system of standards.
Gilbert Harman, Moral Relativism, in G. Harman and J.J. Thomson, Moral

Relativism and Moral Objectivity (Oxford: Blackwell, 1996), p. 62. As the term values is
being used by Harman, ones values ground ones attitudes of approval and disapproval,
and so reflect (or coincide with) ones standards rather than ones tastes.
2
standards, advice, and practical reason13
This thesiswhich is taken for granted by many, including, for example, Bernard Williams, Gilbert Harman, and Donald Hubin3might
seem quite plausible. Indeed, it might seem as though the following
even stronger thesis is warranted:
Thesis 2: All genuine and sincere advice is such that, in cases of conflict,
the standards of the adviser are put aside and actions are evaluated from
within the advisees system of standards.
For it might seem as though there is something to the following line of

reasoning:
In cases of conflict, someone who offers evaluative judgments based on
her own system of standards rather than on the advisees system of standards can be interpreted as a campaigner or as a bluffer, but not as a sincere
adviser. This is because giving genuine and sincere advice involves seeing
things from the advisees perspective rather than ones own; and seeingthings from the advisees perspective involves, among other things,
putting aside ones own standards and reasoning on the basis of the advisees system of standards.
Though there is, I grant, some plausibility to the idea that giving
genuine and sincere advice involves seeing things from the advisees
perspective rather than ones own, thesis 2 is false. Indeed, even thesis
1which is significantly weaker than thesis 2, and which is taken for
granted by Williams, Harman, and Hubinis, I will argue, untenable.
I will first directly address two sorts of cases in which it might appear
as though an adviser is evaluating things from within the advisees system of standards even though this system conflicts with her own; such
cases are, I think, best interpreted in ways that dissolve this appearance.
I will then argue that the nature of sincere advice precludes an advisers
putting aside her own system of standards in favor of a competing system of standards.
After defending my position, I will consider its implications for a
popular Humean view of practical reason. According to this Humean
view, practical reason judgmentswhich are concerned with what an
agent has reason to do or not to do and which can figure as advice
evaluate actions from within the agents (as opposed to the judgers)
3
See Bernard Williams, Internal Reasons and the Obscurity of Blame, in
B. Williams, Making Sense of Humanity (Cambridge: Cambridge University Press,
1995), pp. 3545; Harman, Moral Relativism; and Donald Hubin, The Groundless
Normativity of Instrumental Rationality, Journal of Philosophy 98.9 (2001),
pp. 44568.
14
chrisoula andreou
system of standards. If, however, I am right, and the nature of sincere

advice precludes an advisers putting aside her own system of standards
in favor of a competing system of standards, then the popular Humean
view is untenable.
2. Intentions, Standards, and Advice
Let us first consider two sorts of cases in which it might seem as though
an adviser is evaluating things from within the advisees system of
standards even though this system conflicts with her own.
Intention-based Advice
Suppose Jen is watching a fictitious made-for-television movie. At one
point, the main character, Moe, decides to make his spouse jealous. The
plan he settles on involves flirting with his neighbor. When Moe tells
his friend what he is up to, his friend responds with disapproval, saying
You shouldnt do anything that will make your spouse jealous. How
childish. Although Jen agrees with the friends evaluation, her first
thought is actually The neighbor?! That wont work. You should flirt
with the video store manager. It might be supposed that this bit of
potential advice is based on Moes standards. Given that Jen agrees with
the evaluation of Moes friend, it is quite clear that when Jen judges,
You shouldnt flirt with the neighbor. You should flirt with the video
store manager, this evaluation is not based on her own standards. But
it does not follow from this that her evaluation is based on Moes standards. Jens judgment You shouldnt flirt with the neighbor. You should
flirt with the video store manager is based, I contend, neither on her
standards nor on Moes standards. Instead, it is based on Moes intention, which might conflict with both sets of standards. Jen takes it that
Moe intends to make his spouse jealous and that he is mistaken in supposing that flirting with his neighbor will do the trick. Jens judgment is
no less applicable if Moes intention to make his spouse jealous is
akratic, failing to fit with even his own standards.4 Either way (whether
4
Note that to say that a bit of advice is applicable is not to say that a potential adviser
would always readily provide the advice. If, for example, Moe were part of Jens life
rather than a character on TV, Jen might not be willing to give Moe advice regarding
an intention that conflicts with her standards. For such advice might aid Moe in carrying out his intention, and Jen might not want to provide such aid, regardless of how
Moes intention is related to his own standards.
Moes intention is akratic or not), Jen can affirm that if Moe intends to
make his spouse jealous, he should flirt with the video store manager.
Her judgment is thus based on Moes intention, rather than on either of
their standards, which are, at least for the moment, being put aside.
Intention-based judgmentslike Jens judgment that Moe should
flirt with the video store managerfigure prominently in our interactions with others. When I tell the operator that I am trying to call
Greece and ask her what I should do, she does not, before answering,
inquire into the reason for my call so as to determine whether or how
well my intention fits with my standards. She simply provides me with
the information I need to carry out my intention. Similarly, when I tell
my financial adviser that I want to liquidate some funds in order to buy
a Corvette, but also want, as much as possible, to avoid paying penalties, she does not, before advising me, inquire into the reason for my
proposed auto purchase so as to determine whether or how well my
intention fits with my standards. Again, she just proceeds to provide
me with the information I need to carry out my intention. (She also
provides me with information concerning how carrying out this intention will affect my prospects of carrying out other intentions I have
expressed to her; but this is still a far cry from making an evaluative
judgment from within my standards.) We have, so far, no support for
the thesis that there is a mode of sincere advice in which the evaluation
is made from within the advisees system of standards.
Advice where a Restricted Set of Options is under Consideration
Suppose a friend who is against abortion but is also not ready to become
a mother finds out she is pregnant and, fearful and distressed, comes to
you for advice. After reflecting on your friends casetaking into
account, among other things, how the various available alternatives
would affect those involvedyou come to the conclusion that your
friend should have an abortion. You know, however, that your friend
would never have an abortion. For her, the only conceivable options are
to have the baby and raise it herself or to have the baby and give it up
for adoption. So, when your friend asks you what she should do, you
tell her that she should give the baby up for adoption. Though sincere,
this advice seems to conflict with your standards, and so it might be
assumed that the advice is based on your friends standards. But there
is really no conflict between your advice and your standards. To the
contrary, your advice reflects your standards. More specifically, it
16
chrisoula andreou
reflects the fact that your standards favor the following ranking:
(1) your friend has an abortion; (2) your friend has the baby and then
gives it up for adoption; (3) your friend has the baby and then raises it
herself. Your advice implicitly involves a contrast between options (2)
and (3), which are the restricted set of options under consideration.
Given that your friend would never consider having an abortion but
has not yet formed any intention about what to do with the baby once
she has it, your advice focuses on damage control or on making the
best of things. Such advice is still derived from your standards and so
we do not have a case in which the standards of the adviser are put
aside in favor of the standards of the advisee.
My approach thus far has been defensive. In the next section, I will
offer a positive argument in favor of my contention that there is no
mode of sincere advice in which the adviser puts aside her own system
of standards in favor of a competing system of standards.
3. Why Advisers Must Favor Their Own Standards
It might be suggested that even if I could conclusively show that the
evaluative judgments we, as individuals, offer as advice are never based
on the standards of the advisee, this may simply reflect a defect on our
part, rather than something about the nature of advice. Maybe we are
substandard advisers, self-centered and resistant to taking the perspective of others even when this is called for. In this section, I will argue
that the nature of sincere advice precludes an advisers putting aside her
own system of standards in favor of a competing system of standards.
Suppose a potential adviser is being asked for some advice. Suppose
further that the request is not for intention-based advice, and that there
is a question as to whether the adviser is to use her own system of
standards or the advisees competing system of standards to arrive at an
all-things-considered evaluative judgment that can serve as sincere
advice. If, on the one hand, the adviser is a cognitivist about her standards, then she accepts her standards as true. It thus makes sense for the
adviser to rely on her own standards in making an evaluative judgment. This is because, in cases of conflict, a sincere adviser relies on her
own beliefs rather than on the beliefs of the advisee, as is illustrated by
the following example (in which the conflict is between what are
uncontrovertibly beliefs): Suppose that, according to both Nico and
Sarah (who may have shared standards for evaluating actions), Nico
should really relax and enjoy himself for a few hours. Suppose further
that, while Nico believes that going to a bar for a couple of drinks would
be both relaxing and enjoyable for him, Sarah thinks Nico is mistaken
about this. Her view is that if Nico goes to a bar for a couple of drinks,
he will actually end up stressed and depressed. Clearly, Sarah will, if she
is being sincere, advise Nico against going to a bar for a couple of
drinks. Though she can agree that, relative to Nicos beliefs, Nico should
go to a bar for a couple of drinks, she cannot sincerely provide Nico
with advice that conflicts with her own beliefs. (Of course, if she is
unable to convince Nico that he is mistaken, she can sincerely tell him
to do what he thinks he should do, on the basis of her view that Nico
should not just blindly follow her advice.5)
If, on the other hand, the adviser is a non-cognitivist about her
standards and thinks of judgments derived from an individuals standards as essentially expressive, then again the adviser will have to rely on
her own standards in arriving at a sincere all-things-considered evaluative judgment. For a sincere expressive judgment will aim to convey
the judgers own attitudes, and so will be derived from the judgers
standards, not from the standards of the advisee. So whether the adviser
is a cognitivist or a non-cognitivist about her standards, it makes sense
that her all-things-considered evaluative judgment will be derived
from her own standards.
Of course, an adviser can still make judgments like According to As
standards, A should (all-things-considered) X, which she can affirm
even if her evaluative judgment is that A should (all-things-considered)
not X. But, unlike the judgment A should (all-things-considered) X,
the judgment According to As standards, A should (all-thingsconsidered) X is about a system of standards, rather than being an
evaluation derived from within a system of standards. My conclusion is
that the nature of sincere advice precludes an advisers putting aside her
5
Do what you think you should do can be interpreted as derived from within
Sarahs system of standards for evaluating agents (rather than from within Nicos potentially competing system of standards for evaluating agents, or from within Sarahs system of standards for evaluating actions). My paper is concerned with judgments
derived from standards for evaluating actions. In any case, in saying Do what you
think you should do, Sarah is not putting aside her own system of standards for evaluating agents in favor of Nicos potentially competing system of standards for evaluating
agents. (Nicos system of standards for evaluating agents might require that, in cases of
conflict, an agent should faithfully follow the dictates of a certain authority rather than
do what he thinks he should do.)
18
chrisoula andreou
own standards and evaluating things from within the advisees system
of standards.
Clearly then, when it comes to seeking advice other than intentionbased advice, there is something to be said for seeking out someone not
only sincere, knowledgeable, and clear-headed, but also someone who
largely shares ones own standards, or at least the standards that are
likely to be relevant given the issue at hand. Such a search resembles the
search for a potential true friend, given Aristotles conception of friendship. For, the true friend is, according to Aristotle, another self .6 More
specifically, a true friend is not only someone whose intentions one
trusts, but also someone whose judgment one trusts, and so someone
who at least generally approve[s] of the same things.7 With such an
individual, one is, according to Aristotle, more able both to think and
to act, since one is better guarded against temptations and errors.8
Atrue friend is prone to encourage one to reason well and to be steadfast when one is weak-willed, since she will not want her friend to go
wrong; she might also, in some cases, be better positioned to reflect on
a particular situation.
One might, of course, be interested in having ones standards challenged. In such cases, it might also make sense to seek out someone
who, though she often approves of the same things, does not share all
of ones standards, and, in particular, does not share the standards that
are relevant given the issue at hand. Considering a defense of standards
that one does not share, especially if the defense appeals at least in part
to standards that one does share, is sometimes a sensible strategy for
improving ones ability to think and act. This strategy is, however, dangerous in cases where one is tempted to put aside ones standards
because they are proving to be quite demanding; for, in such cases, one
might be prone to act on rationalizations one is not really convinced by.
There is also the possibility that part of the reason one is asking a
particular person for advice is because one is interested in getting
aclearer picture of this persons standards, perhaps in order to get a
clearer picture of how this person would be likely to react to an action
one is contemplating. Though the point of this might be to open ones
standards up to challenges, it need not be. For the person one is asking
6
Aristotle, Nicomachean Ethics [cited hereafter as NE], trans. David Ross, rev. J.L.
Ackrill and J.O. Urmson (Oxford: Oxford University Press, 1998), IX.4, 1166a3132.
7
Aristotle, NE, IX.3, 1165b2729.
8
Aristotle, NE, VIII.1, 1155a1516.
for advice need not be someone whose judgment one trusts; she might,
for example, simply be someone with power over ones prospects.
If the nature of sincere advice precludes an advisers putting aside
her own standards in favor of the standards of the advisee, then in what
sense does being a good adviser require seeing things from the perspective of the advisee? My reasoning suggests that one respect in
which the good adviser will see things from the perspective of the advisee is that the good adviser will not ignore the decisions that the advisee has already made. For, a good adviser will, at least typically, focus
on providing advice that is not bound to be ignored. She will thus need
to know what options the advisee is willing to consider, and the decisions that an advisee has already made are relevant for determining
this. Also, an advisees decisions determine what intention-based
advice is applicable. There is arguably a further respect in which the
good adviser will see things from the perspective of the advisee. For
there is a great deal of plausibility to the idea that the good adviser will,
in some cases, base her advice on the agents likes and dislikes. But all
this is consistent with the advisers favoring her own standards.
Note that I am not committed to the view that a good adviser is willing to impose her standards on others. To the contrary, my view is consistent with the idea that a good adviser will never attempt to pressure,
let alone compel, others to act in accordance with her standards, even
if she is convinced that her standards are correct. For my view is consistent with the idea that a good adviser will recognize her fallibility
and respect the autonomy of others. As John Rawls stresses, even if
there is an objectively or intersubjectively valid set of standards of evaluation and one sees ones standards as capturing ones views concerning these valid standards, one can still, in light of the burdens of
judgment, acknowledge a reasonable pluralism of standards based on
reasonable disagreement.9
4. Standards, Advice, and Practical Reason Judgments
If my reasoning in this paper is correct, then there is no room for
thepopular Humean view of practical reason that is accepted by, for
example, Bernard Williams, Gilbert Harman, and Donald Hubin.10
9
John Rawls, Justice as Fairness: A Restatement (Cambridge, MA: Belknap Press,
2001), pp. 3537.
10
See Williams, Internal Reasons and the Obscurity of Blame; Harman, Moral
Relativism; and Hubin, The Groundless Normativity of Instrumental Rationality.
20
chrisoula andreou
According to the popular Humean view, practical reason judgments

which are concerned with what an agent has reason to do or not to do
and which can figure as sincere adviceevaluate actions from within
the agents (as opposed to the judgers) system of standards.11 So if, for
example, an agents standards permit her to eat veal and this is common knowledge, then, according to the popular Humean view, the
evaluative judgment that the agent should refrain from eating veal is,
assuming it is put forward as a practical reason judgment, nothing
more than a bluff (unless, of course, someone makes it the case that the
agent should refrain from eating veal by introducing threats or incentives into the situation).
Implicit in the popular Humean view is the following idea:
There is no question of ones standards being true or false. Ones standards, unlike ones beliefs, do not represent, either well or poorly, some
aspect of reality; instead, they are essentially expressive.
For those who are not opposed to this idea, the popular Humean view
might seem plausible for several reasons. First, the view allows that, as
is commonly taken for granted, the aptness of a practical reason judgment does not vary depending on who is making the judgment. This
follows from the fact that there is, according to the Humean view, a
single system of standards that is relevant for arriving at practical reason judgments, namely the agents system of standards. Second, the
In the words of proponents of the popular Humean view, practical reason judgments take the agents perspective on a situation (Williams, Internal Reasons and the
Obscurity of Blame, p. 36), illustratea relativity to the agents values (Harman,
Moral Relativism, p. 62), evaluate actions from within the agents system of norms
(Hubin, The Groundless Normativity of Instrumental Rationality, pp. 46566).
Note that the popular Humean view differs from simple instrumentalism, according
to which practical reasoning and practical reason judgments are not in order until the
agent has settled on some end. For a defense of this latter view, which is less popular
among Humeans, see Candace A. Voglers Reasonably Vicious (Cambridge, MA:
Harvard University Press, 2002), especially chapter 7. For a critique of the view, see
Christine Korsgaards The Normativity of Instrumental Reason, in Garrett Cullity and
Berys Gaut (eds.), Ethics and Practical Reason (Oxford: Clarendon Press, 1997),
pp. 21554.
Note also that it is common ground that the popular Humean view is not Humes
view. There is, nonetheless, a great deal of controversy concerning how Humes position, particularly in the Treatise, is best interpreted. For two very interesting discussions concerning this issue see Elijah Millgrams Was Hume a Humean?, Hume Studies
21.1 (1995), pp. 7593, and Nicholas Sturgeons Hume on Reason and Passion, in
Donald Ainslie (ed.), Humes Treatise: A Critical Guide (Cambridge: Cambridge
University Press, forthcoming).
11
popular Humean view can account for the connection between accepting a practical reason judgment and being at least somewhat motivated
to act accordingly. Since the standards that are relevant for arriving at
practical reason judgments are the agents, and since the agents standards are essentially expressive, it makes sense that, barring akrasia,
agents are motivated to act in accordance with practical reason judgments that they accept. Finally, unlike the view that an agents reasons
are determined by the agents felt inclinations, which is explicitly
rejected by prominent Humeans as overly-simple, the popular Humean
view recognizes that, when there is a conflict between our standardsand our felt inclinations, it is our standards, rather than our felt
inclinations, that we are bound to see as necessarily relevant in evaluating our own actions.
Despite these potential virtues, the popular Humean view is untenable because it conflicts with the fact that the nature of sincere advice
precludes an advisers putting aside her own system of standards in
favor of a competing system of standards. As evaluative judgments that
can figure as sincere advice (and, more specifically, that do figure as
sincere advice when sincerely expressed in the form You should X),
practical reason judgments cannot be derived from a system of standards that conflicts with the judgers system of standards. Given that,
when agent and judger are not one and the same, the agents system of
standards can conflict with the judgers system of standards, it follows
that the idea that practical reason judgments evaluate actions from
within the agents system of standards is unacceptable.
It might seem as though a small alteration can save the popular
Humean view. The Humean can abandon the idea that practical reason
judgments are evaluations derived from within a set of standards, and
maintain instead that practical reason judgments either amount to or
derive from judgments about the agents standards. The Humean can,
more specifically, maintain that the practical reason judgment A should
(all-things-considered) X amounts to or derives from the judgment
According to As standards, A should (all-things-considered) X. If
practical reason judgments amount to or derive from judgments
about the agents standards, then a judger could affirm the practical
reason judgment A should (all-things-considered) X even if her
competing standards support the evaluative judgment A should (allthings-considered) not X. A judger could, for example, affirm the practical reason judgment Al should (all-things-considered) eat the veal
that has been placed in front of him by his host even if the judgers
22
chrisoula andreou
competing standards support the evaluative judgment Al should (allthings considered) not eat the veal that has been placed in front of him
by his host.
The view that practical reason judgments amount to or derive from
judgments about the agents standards is, however, problematic. This
becomes apparent via consideration of first-person practical deliberation. When, in practical deliberation, one concludes that one should
(all-things-considered) X, this practical reason judgment figures as an
evaluative judgment derived from within ones system of standards
rather than as a judgment about ones system of standards, or as a judgment that derives from a judgment about ones system of standards.
Consider, for example, the following straightforward bit of practical
deliberation:
Premise: Animals have the right to a decent quality of life.
Conclusion:
I should boycott factory-farm-produced meat.
The conclusion certainly seems to be derived from within a system of
standards. The premise, which figures as providing direct support for
the conclusion, could not figure in this way if the argument could be
rewritten in either of the following ways:
Conclusion: According to my standards, I should boycott
factory-farm-produced meat.
Conclusion 1: According to my standards, I should boycott
factory-farm-produced meat.
Conclusion 2:
I should boycott factory-farm-produced meat.
Unlike the judgment According to my standards, I should boycott factory-farm-produced meat, the judgment Animals have the right to a
decent quality of life is not a judgment about the agents standards. As
such, the latter judgment cannot provide direct support for the
former.
It might be suggested that the premise in the original bit of practical
deliberation must not be taken at face value, but must instead be interpreted as elliptical for the judgment According to my standards, animals have the right to a decent quality of life. But why should we accept
this suggestion? It seems dogmatic to insist that the agent cannot
takeher conclusion to be directly supported by a judgment about the
rights of animals, but must instead take it to be directly supported by a
judgment about herself. If, however, we allow that the agent can take
her conclusion to be directly supported by a judgment about the rights
of animals, then we must also allow that the agents practical reason
judgment is an evaluation derived from within her system of standards,
rather than a judgment that amounts to or derives from a judgment
about her system of standards. But then we cannot accept the suggestion that practical reason judgments amount to or derive from judgments about the agents standards. So the slightly altered Humean view
fares no better than the popular Humean view it was meant to replace.
5.Conclusion
In the bulk of this paper, I argued against the thesis that there is a mode
of sincere advice in which the standards of the adviser are put aside in
favor of the standards of the advisee. Though plausible-sounding, this
view is untenable. There are, I grant, cases in which it might seem as
though an adviser is evaluating things from within the advisees system
of standards even though this system conflicts with her own; but such
cases can, and indeed must, be interpreted in ways that dissolve this
appearance. For the nature of sincere advice precludes an advisers putting aside her own system of standards in favor of a competing system
of standards. It follows that a popular Humean view of practical reason
is untenable.
Does Rationality Consist in Responding

Correctly to Reasons?
John Broome*
1. Rationality and Responding to Reasons
Some philosophers think that rationality consists in responding correctly to reasons. For a reason I shall explain in section 2, not many
philosophers accept this idea as it stands. But very many accept some
variant of it. This paper looks for what truth it contains. I shall pursue
the truth through many of the twists and turns the idea can take. I cannot go through every one; there are so many. But I hope nevertheless to
extract all the truth contained in the idea.
Variants will come later. This section clarifies what the idea means in
its original form. What does it mean to say that rationality consists in
responding correctly to reasons? Partly, it is to say that:
Equivalence. Necessarily, you are rational if and only if you respond correctly to reasons.
As I shall put it, rationality is equivalent to responding correctly to

reasons. To say that rationality consists in responding correctly to reasons goes further; it implies that rationality can be reduced to responding correctly to reasons. Some philosophers hope to reduce rationality
to reasons through this formula or some related one. One purpose of
this paper is to argue they cannot do so. Rationality must be recognized
as an independent source of requirements.
For this purpose, it turns out that I need only concentrate on versions of equivalence, rather than on putative reductions. I shall argue
that equivalence and its related formulae are false. A fortiori, it will
follow that no reduction of this sort is possible.
* Thanks to Jacob Busch, Olav Gjelsvik, Christine Korsgaard, Sven Nyholm, Derek
Parfit, Wlodek Rabinowicz and Andrew Reisner for very useful comments. Thanks
also to audiences at Cambridge, Harvard, Oslo, Oxford and St Andrews, who all contributed a great deal. My work on this paper was supported by a Major Research
Fellowship from the Lever-hulme Trust, to whom I am very grateful.
26
john broome
Another purpose of the paper is simply to explore a part of the relation between rationality and reasons. I shall look for what truth there is
in equivalence and its relatives. I shall give most attention to this one
side of equivalence, because it is nearest to the truth.
Entailment. Necessarily, if you are rational you respond correctly to
reasons.
As I shall put it, rationality entails responding correctly to reasons.

Responding Correctly to Reasons
I need to start by elucidating the notion of responding correctly to
reasons.
First, the reasons it refers to are normative reasons. Furthermore,
they are reasons that are owned by you, as I put it. There may be a reason for you to be punished, but this reason does not call for a response
from you. Your punishment is the responsibility of the authorities, not
you. You do not own this reason. On the other hand, if there is a reason
for you to get lunch, you probably own it. It calls for a response from
you. So reasons refers to your normative reasonsthe normative reasons that are owned by you.
Next, what does correctly mean? It is simply a placeholder.
Equivalence claims there is some way of responding to reasons such
that, necessarily, you are rational if and only if you respond to reasons
that way. Responding correctly means responding in that way, whatever it is. Correctly cannot be omitted from the formulae, because
there are ways of responding to reasons such that it would not be
rational to respond in one of those ways. Those ways do not count as
correct.
What way of responding to reasons is correct, then? How must you
respond if you are to be rational? Suppose you have a reason to F; to be
rational must you F? No. Suppose you also have a reason not to F, as
you certainly may. If you had to F in order to respond correctly to a
reason to F, to respond correctly to your two conflicting reasons, you
would have both to F and not F. You could not respond correctly to
both reasons, therefore. But we must not interpret respond correctly
in a way that makes it impossible for you to respond correctly to conflicting reasons. If we did, equivalence would entail that you cannot be
rational, since you inevitably encounter conflicting reasons. We cannot
have that.
does rationality consist in responding correctly? 27
Let us go more carefully. Suppose you ought to F. I take it for granted

that, if this is so, there is some explanation of why it is so. The explanation may be that the balance of reasons comes down on the side of
Fing. In that case, you ought to F because on balance your reasons
require you to F. More briefly: your reasons require you to F.
Possibly the explanation of why you ought to F is not given by the
balance of reasons. Perhaps, say, some rigid deontic rule determines
that you ought to F. Nevertheless, we count the explanation of why you
ought to F as a reason for you to F. So in this case too, we may say your
reasons require you to F. (Your reasons refers to one reason only.)
Taking all cases together, we may conclude that you ought to F if and
only if your reasons require you to F.
This allows us to say that, to respond correctly to reasons, you must
F whenever your reasons require you to F. This interpretation of
respond correctly to reasons makes it possible for you to respond
correctly when your reasons conflict. We could express the same interpretation in a different but equivalent way. Your reasons require you
to F if and only if you ought to F. So we could say that, to respond correctly to reasons, you must F whenever you ought to F.
That you F whenever your reasons require you to F is a necessary
condition for you to respond correctly to reasons. There is at least one
other necessary condition as well. Even if you F whenever your reasons
require you to F, you might not be responding correctly to reasons; it
might just be a coincidence. Some appropriate connection must hold
between your reasons and your Fing. It may need to be an explanatory
one. Alternatively, a mere counterfactual connection may be enough.
For instance, the necessary condition might be that you would not have
Fed had your reasons required you not to F.
I shall add a clause to my formulae requiring an appropriate explanatory or counterfactual connection. But since it does not matter for my
argument, I shall not try to say just what sort of connection would be
appropriate.
Perhaps more conditions are necessary. But I am going to assume we
have now arrived at necessary and sufficient conditions for you to
respond correctly to reasons. You respond correctly to reasons if and
only if, whenever your reasons require you to F, you F and there is an
appropriate explanatory or counterfactual connection between your
reasons and your Fing. This is an analysis of you respond correctly
to reasons.
28
john broome
It makes equivalence come down to:

Equivalence analysed. Necessarily, you are rational if and only if, whenever your reasons require you to F, you F and an appropriate explanatory
or counterfactual connection holds between your reasons and your Fing.
One side of it is:

Entailment analysed. Necessarily, if you are rational then, whenever your
reasons require you to F, you F and an appropriate explanatory or counterfactual connection holds between your reasons and your Fing.
Since I am not interested in the explanatory or counterfactual

connection, I shall generally concentrate on this central part of
entailment:
Core condition. Necessarily, if you are rational then, if your reasons
require you to F, you F.
This schema is supposed to hold for all substitutions of a verb phrase

for F. We can therefore insert a universal quantifier into the core condition when we choose. We get: necessarily, if you are rational, you F
whenever your reasons require you to F.
I now turn to evaluating equivalence.
2. The Quick Objection
I shall start with a quick objection to it. Suppose your reasons require
you to F but you are ignorant of those reasons. Suppose you are not at
fault in being ignorant; you have no evidence of the reasons. If you do
not F, you might nevertheless be rational. So the core condition is false.
Therefore, equivalence is false because it entails the core condition. So
far as your rationality is concerned, ignorance of your reasons constitutes an excuse for not Fing.
Here is an example. The fish on the plate in front of you contains
salmonella. This is a reason for you not to eat it, and let us assume all
your reasons together require you not to eat it. But you have no evidence that the fish contains salmonella. Then you might eat it even
though your reasons require you not to, and nevertheless you might be
rational.
In this example, you are ignorant of the non-normative fact that
the fish contains salmonella, which constitutes a reason for you not to
eat it. In another version of the example, you believe the fish contains
salmonella, but, without fault, you do not believe this constitutes a

reason for you not to eat it. Here you are ignorant of a normative
fact. Either sort of ignorance is an excuse for not responding correctly
to reasons.
Why is ignorance an excuse? Because of a fundamental feature of
rationality: that your rationality depends only on the properties of your
mind.1 Rationality supervenes on the mind, I shall say. If your mind in
one situation has the same properties as it has in another, then you are
rational in one if and only if you are rational in the other. This is a conceptual truth; it is part of the concept of rationality. True, we sometimes
say that a non-mental act of yours is irrational. If you believe your reasons require you not to eat the fish, and yet you eat it, we might say
your act is irrational. But that is because we normally assume that you
do a particular act only if you intend to do it. You are indeed irrational
if you believe your reasons require you not to eat the fish and yet you
intend to eat itthis is a consequence of the enkratic condition, which
I shall come to in section 4. Given our normal assumption, you are
therefore irrational if you believe your reasons require you not to eat
the fish and yet you eat it. Your act shows you are irrational, so we treat
the act itself as irrational. But in this case too, whether or not you are
irrational depends only on the state of your mind. If we learn that our
normal assumption is false, and you eat the fish without intending to
perhaps you eat it accidentally because it is concealed under some
mashed potatowe shall not think you irrational.
Go back now to the case where the fish contains salmonella but you
do not believe it does, and you eat it. Compare it with the case where
the fish does not contain salmonella and you do not believe it does and
you eat it. The properties of your mind might be the same in either
case. (An externalist about the mind can recognize this possibility as
well as an internalist.2) Clearly you could be rational in the second case.
Therefore, because rationality supervenes on the mind, you could be
rational in the first case too. So your ignorance is an excuse in that case.
A parallel argument shows that ignorance of a normative fact can also
be an excuse.
The quick objection is that ignorance is an excuse. Many philosophers find it convincing. As a result, few accept exactly equivalence.
1
Ralph Wedgwood argues this point convincingly in Internalism Explained,
Philosophy and Phenomenological Research 65 (2002), pp. 34969.
2
Thanks to Olav Gjelsvik for this point.
30
john broome
Instead, they assume that a rational persons response to reasons has to

be filtered through her beliefs in some way. They think rationality is
equivalent to responding correctly to beliefs about reasons, or to
believed reasons, or something of that sort. I shall come to thoughts
like this in section 4.
3. Attitudinal Reasons
Before that I need to investigate a possible reply to the quick objection.
It can be argued that your ignorance of a reason does not always excuse
you. Some reasons may impose strict liability on you, as I shall put it.
By this I mean that, necessarily, if you are rational you respond correctly to these reasons, whether or not you believe they exist, and
whether or not you believe they are reasons. If you do not respond correctly to them you are automatically irrational. (By irrational I mean
not rational.)
How could a reason impose strict liability, in view of the supervenience of rationality on the mind? I argued on grounds of supervenience
that a reason cannot impose strict liability, using the example of salmonella. But my argument would not work against a reason that is itself a
state of mind or a fact about a state of mind. And, at least at first sight,
there seem to be reasons of this sort.
For instance, here are some claims about reasons that are plausible at
first sight:
Attitudinal reasons
R1. If you believe p, your belief is a reason for you not to believe not p.
R2. If you intend to F, your intention is a reason for you not to intend
not to F.
R3. If you believe p and you believe that if p then q, your two beliefs are
together a reason for you to believe q.
R4. If you intend to F and you believe you cannot F unless you G, your
intention and belief are together a reason for you to intend to G.
In each of these claims, the reason is said to be an attitude of yours or a

combination of attitudes. But we might equally well take the reason to
be the fact that you have the attitude or combination of attitudes. It is a
harmless regimentation to count every reason as a fact. Either way,
I shall call these reasons (if they are indeed reasons) attitudinal
reasons.
Although these claims are plausible at first sight, I do not believe
attitudes or combinations of attitudes can be reasons in this way.
Here is why. Take R3 as an example. For p substitute Emissions of

carbon dioxide do not cause global warming and for q Emissions of
carbon dioxide are harmless. There are various pieces of evidence for
the proposition that emissions of carbon dioxide are harmless, and
other pieces of evidence against it. No doubt these pieces of evidence
constitute reasons for or against believing the proposition. But now
suppose you believe that emissions of carbon dioxide do not cause
global warming, and you also believe that, if emissions of carbon dioxide do not cause global warming, they are harmless. According to R3,
these beliefs of yours are a reason to believe that emissions of carbon
dioxide are harmless. But that is not credible. Your beliefs could not
stand as a further reason for believing emissions of carbon dioxide are
harmless, alongside the evidence. You cannot by means of your beliefs
bootstrap a new reason into existence, to add to the evidence.
I can reinforce the example. R3 entails that, if you believe p and you
believe that if p then p, these beliefs constitute a reason for you to
believe p. That cannot be so. We can take it for granted that you believe
the tautology that, if p then p. Given that, R3 entails that believing a
proposition gives you a reason to believe it. Any belief you have
gives you a reason to have it. That cannot be so; it would be absurd
bootstrapping.3
Similar bootstrapping arguments will work against the other putative attitudinal reasons. However, the conclusion that there are no attitudinal reasons is controversial, and I do not wish to rest the main
argument of this section on it. So, for the sake of argument in this section, I shall suppose that attitudinal reasons do exist. I shall show that,
even if they do, they do not constitute a basis for equivalence.
Each attitudinal reason is matched by a corresponding necessary
condition for rationality:
Conditions of rationality
C1. Necessarily, if you are rational, you do not believe p and believe
not p.
C2. Necessarily, if you are rational, you do not intend to F and intend
not to F.
3
Thomas Nagel pointed out to me that believing p at one time might constitute a
reason to believe p at a later time. The fact that you believe p one time constitutes, at a
later time, indirect evidence of p: you would probably not have believed p if you had
not had evidence for it. But in the claim I am objecting to, the beliefs are supposed to
be contemporaneous.
32
john broome
C3. Necessarily, if you are rational then, if you believe p and you believe
that if p then q, you believe q.
C4. Necessarily, if you are rational then, if you intend to F and you believe
you cannot F unless you G, you intend to G.
I have stated these conditions only approximately. For example, you

might be rational even if you do not believe everything that follows by
modus ponens from what you believe.4 But I shall not burden you with
the more complicated, accurate formulations, nor the corresponding
accurate formulations of attitudinal reasons. Since they are only illustrative examples, rough formulations are good enough in this paper.
I hope it is obvious that there are necessary conditions for rationality
of this sort.
If attitudinal reasons exist, they impose strict liability just because of
their correspondence with conditions of rationality. If you have an attitudinal reason to F but do not F, you fail to satisfy the corresponding
condition of rationality, so you are not rational. You have contradictory
intentions or contradictory beliefs, or you do not believe what follows
by modus ponens from the contents of your beliefs, or you are irrational in some other way. That is so whether or not you believe you
have a reason to F.
Ignorance is No Excuse
If there are indeed attitudinal reasons, they impose strict liability. They
do so without violating the supervenience of rationality on the mind,
because they are states of mind themselves. But you might still be
uneasy, and feel that ignorance of a reason must be an excuse. However,
we can see from the nature of attitudinal reasons themselves how they
can impose strict liability. You can respond to an attitudinal reason
without believing it exists or, if you do believe it exists, without believing it is a reason. Given that you can do so, neither the fact that you do
not believe it exists nor the fact that you do not believe it is a reason
constitutes an excuse for not responding to it.
How can you respond to an attitudinal reason without believing it
exists? You often respond to an attitude without believing it exists. You
may respond to a belief of yours that it is raining by putting up your
4
A point made by Gilbert Harman in Change in View (Cambridge, MA: MIT Press,
1986).
umbrella. To do so, you do not need to have a second-order belief that

you believe it is raining; you do not need to believe the attitude exists.
I am not suggesting your attitude in that case constitutes a reason.
But here is an example of responding to an attitude that (I am assuming
in this section) does constitute a reason. Suppose you believe it is raining, but then you look out of the window and see the rain has stopped.
You now believe it is not raining. We are assuming this belief constitutes a reason for you not to believe it is raining. Typically, you will
indeed no longer believe it is raining once you acquire the opposite
belief. Some unconscious process generally ensures that, when you
come to believe a proposition, you do not also believe its negation. The
operation of this unconscious process counts as responding to your
reason, which is your belief that it is not raining. For the process to
work, you do not have to believe the reason exists. In this case, you do
not have to believe you believe it is not raining.
You may respond to any attitudinal reason like this, through automatic processes. But sometimes your automatic processes let you
down; you do not respond automatically. Even then, you may respond
nonetheless, through a conscious process of reasoning. Suppose, as you
wake up one morning on a cruise, you hear gulls from your cabin, so
you believe there are gulls about. Suppose you also have a standing
belief that, if there are gulls about, land is nearby. According to my
example R3 of an attitudinal reason, these beliefs constitute a reason
for you to believe land is nearby. But if they have not come together in
your mind, you may not yet believe land is nearby. However, by a piece
of conscious reasoning, you can reason your way to believing there is
land nearby. This reasoning will involve calling to mind the proposition that there are gulls about, and the proposition that, if there are
gulls about, land is nearby. But it need not involve any second-order
belief that you believe these propositions.
At least, my own view about reasoning is that it need not involve
second-order beliefs. Others disagree: they think reasoning must
indeed involve second-order beliefs, at least if it is to be full-blown,
critical reasoning.5 It does not matter in any case. The question is
5
For instance, Tyler Burge in Reason and the First Person, in Crispin Wright, Barry
C. Smith and Cynthia MacDonald (eds.), Knowing Our Own Minds (Oxford: Oxford
University Press, 1998), pp. 24370. My own view is outlined in my The Unity of
Reasoning?, forthcoming in Simon Robertson, John Skorupski and Jens Timmerman
(eds.), Spheres of Reason (Oxford: Oxford University Press).
34
john broome
whether you can respond to an attitudinal reason without believing

it exists, and the answer is that you can. You can respond by automatic processes. It is also at least arguable that you can respond by
conscious reasoningeven if not the full-blown, critical sortwithout
believing the reason exists. We can add that, a fortiori, if you do believe
an attitudinal reason exists, you can respond to it without believing it
is a reason.
All this shows how attitudinal reasons may impose strict liability.
For them, ignorance is no excuse.
Responding Correctly to Attitudinal Reasons
Attitudinal reasons provide a reply to the quick objection. They are not
subject to this objection, since they may impose strict liability. So if
attitudinal reasons exist, they might provide support for equivalence.
They could not support equivalence exactly. Even if attitudinal reasons exist, they are not all the reasons there are. I take it for granted that
many reasons are not attitudes. If the fish contains salmonella, that is a
reason not to eat it, but it is not an attitude. The quick objection applies
to these other reasons. Consequently, the core condition set out in section 1 is not true in general; you may be rational even if you sometimes
do not F when your reasons require you to F.
Nevertheless, the core condition might be true if it was restricted to
attitudinal reasons only:
Necessarily, if you are rational, you F if your attitudinal reasons require
you to F.
Equivalence might be true if it was restricted in the same way:

Necessarily, you are rational if and only if you respond correctly to attitudinal reasons.
That is to say, rationality is equivalent to responding correctly to attitudinal reasons.

It might even be true that rationality consists in responding correctlyto attitudinal reasons. That is to say, rationality might be reducible to responding correctly to attitudinal reasons. You might think that
could not be true because we already need to know what rationality is
before we can identify attitudinal reasons. I explained that each attitudinal reason corresponds to a condition of rationality. You might think
we can identify attitudinal reasons only through their correspondingconditions of rationality. But even if that is so, it might nevertheless
be the case that the conditions of rationality hold only because of their
corresponding attitudinal reasons. There might simply be a class of
attitudinal reasons like the ones I described, and rationality might
consist in responding correctly to them. The reasons might explain
the conditions, even though we identify the reasons through the
conditions.
So attitudinal reasons provide a potential defence of equivalence and
the reductionist idea that rationality consists in responding correctly
to reasons. However, I shall now argue that this defence fails. Even
restricted to attitudinal reasons, equivalence is false.
Your attitudinal reasons may conflict with each other; you may have
an attitudinal reason to F and an attitudinal reason not to F. I shall give
an example soon. When your reasons conflict, you respond correctly
to them so long as you do not go against their balance. Inevitably, even
if you respond correctly, you will not satisfy one or other of your conflicting reasons. But I explained that these reasons impose strict liability. That is to say, in failing to satisfy one of them you are inevitably
irrational. So, although you respond correctly to reasons, you are irrational. Therefore rationality is not equivalent to responding correctly
to attitudinal reasons.
Here is the argument more briefly. To be rational, you have to satisfy
some particular conditions. But attitudinal reasons can be overridden,
so responding correctly to them cannot guarantee that you satisfy any
particular condition. Therefore rationality is not equivalent to responding correctly to attitudinal reasons.
Next an example. Suppose you intend to G, and you believe you cannot G unless you H. According to R4 among the examples above, this
intention and belief together constitute an attitudinal reason for you to
intend to H. But suppose also that you intend not to H. According to
R2, this intention constitutes an attitudinal reason for you not to intend
to H. So you have an attitudinal reason to intend to H and another not
to intend to H.
To respond correctly to your reasons in this case, you must go with
the balance of your reasons. Let us say that the balance is in favour of
your not intending to H. Then you respond correctly to your reasons if
you do not intend to H. But you intend to G and you believe you cannot
G unless you H. Given that, the condition of rationality C4 set out
above tells us you are not rational. So you respond correctly to your
reasons but you are not rational. Therefore, rationality is not equivalent
to responding correctly to reasons.
36
john broome
There is a lacuna in this argument. It shows that you can respond

correctly to your attitudinal reasons to F and not to F and yet be irrational. But the conclusion we need, to reject equivalence, is that you
can respond correctly to all your attitudinal reasons and yet be irrational. Implicitly I assumed you are responding correctly to all your
other attitudinal reasons, but that might be impossible. The argument
requires two attitudinal reasons to conflict, which the example shows is
possible. But suppose there cannot be a conflict between two of your
attitudinal reasons unless you fail to respond correctly to some other
attitudinal reason. Then, even if you respond correctly to the two particular conflicting reasons, you do not respond correctly to all your
reasons.
So a reply to my argument is to claim that your attitudinal reasons
are linked together in such a way that there cannot be a conflict between
any two of them unless you fail to respond correctly to some other one.
Call this the network theory.
Is it true? I think not. So far as I can see, every one of your attitudinal
reasons might be opposed by another. Then, when you respond correctly to all your attitudinal reasons, you will inevitably fail to satisfy
some of them. Because each one imposes strict liability, that means you
are irrational.
I can illustrate with the same example. The two attitudinal reasons
I mentioned conflict. According to the network theory, therefore, you
must have some other attitudinal reason that you do not respond correctly toone that I have not yet mentioned. An obvious candidate is
this: your intention to G and your belief that you cannot G unless you
H might together constitute a reason for you not to intend not to H.
Since you do intend not to H, you are apparently not responding correctly to this reason.
But actually you may be responding correctly. This reason not to
intend not to H may be opposed by another attitudinal reason. For
instance, suppose you believe you ought not to H, and this belief constitutes a reason for you to intend not to H. Then the balance of reasons
might favour your intending not to H, so you are responding correctly
to these reasons too.
According to the network theory, then, there must be yet another
attitudinal reason that you are not responding correctly to. Perhaps
it is this: your belief that you ought not to H and your belief that you
cannot G unless you H might together constitute a reason for you not
to intend to G. Since you do intend to G, you are apparently not
responding correctly to this reason.
But again, actually you may be responding correctly. This reason not
to intend to G may be opposed by another attitudinal reason. Perhaps
it is opposed by a belief that you ought to G. So far as I can see, this
could go on for ever.
I conclude that attitudinal reasons do not provide a successful
defence of equivalence. Entailment, restricted to attitudinal reasons,
may yet be true:
Necessarily, if you are rational, you respond correctly to attitudinal
reasons.
This seems plausible to me,6 if attitudinal reasons exist. But since

Ibelieve attitudinal reasons do not exist, and since I have achieved my
main aim of rejecting equivalence, I shall not pursue it further. From
here on I set attitudinal reasons aside.
4. The Enkratic Condition
Many philosophers accept the quick objectionevidently they are not
impressed by attitudinal reasons. It leads them to abandon equivalence.
They do not think that rationality is equivalent to responding correctly
to reasons. Instead they adopt some modified idea in which your reasons are filtered through your beliefs. They may think that, necessarily,
you are rational if and only if you respond correctly to beliefs about
reasons, or to believed reasons, or something of that sort. Can we find
a modified idea that is true?
In this section and the next two I shall investigate in turn three alternative modifications. Each can be understood to claim that:
Modified entailment. Necessarily, if you are rational, you respond correctly to beliefs about reasons.
That is to say, rationality entails responding correctly to beliefs about

reasons. Each alternative modification interprets the expression
respond correctly to beliefs about reasons differently. I shall accept
modified entailment under some interpretations. One of the modifications is also intended to support:
Modified equivalence. Necessarily, you are rational if and only if you
respond correctly to beliefs about reasons.
Thanks to Julia Markovits for pointing this out to me.
38
john broome
But I shall not accept modified equivalence under any interpretation.

The first modification starts with the core condition from section 1.
An easy modification of it takes us to:
Direct enkratic condition. Necessarily, if you are rational, then, if you
believe your reasons require you to F, you F.
This cannot be generally true, for the following reason. Suppose you
believe your reasons require you to F, and you intend to F, but you are
prevented from Fing by something outside your mind. You may be
rational. For instance, suppose you believe your reasons require you to
vote for Gore, and you intend to do so, and you do all the appropriate
things, but surprisingly the machinery works in such a way that you
vote for Bush. You fail to do what you believe your reasons require you
to do, but you might nevertheless be rational.
This is another application of the principle that rationality supervenes on the mind, which I introduced in section 2. If the machinery
had worked differently but your mind had been just the same, you
would have voted for Gore. In that case, we may presume you would
have been rational, since voting for Gore is what you believe your reasons require you to do. Given the way the machinery actually works,
you vote for Bush. But the operation of the machinery cannot affect
your rationality. Since your mind is the same in either case, you are
rational in this case too.
The consequence is that the direct enkratic condition cannot be true
unless whether or not you F is entirely determined by the properties of
your mind. The condition cannot be true if Fing is a bodily act, or a
state such as living in Australia.
It might nevertheless be true in important cases. I shall come back to
those in section 5. But for cases where your Fing is not determined by
the properties of your mind, we need to modify the core condition
some more. We can make it:
Enkratic condition. Necessarily, if you are rational, then, if you believe
your reasons require you to F, you intend to F.
This condition encounters no objection from the supervenience of

rationality on the mind. True, when you believe your reasons require
you to F, some external force may prevent you from intending to F, just
as an external force may prevent you from Fing. For instance, a psychologist might have wired your brain so as to prevent you from
intending to F. But then the external force has affected your rationality;
the scientist prevents you from being entirely rational. So this is not an
objection to the enkratic condition.
The enkratic condition says that akrasia is irrational. This is a common opinion, and one I shall accept without more ado. I treat this as a
concession, since I am about to show that it gives the best support I can
find to modified entailment. I do not think the condition as I have
stated it is exactly true. A more accurate version is:
Necessarily, if you are rational, then, if you believe your reasons require
you to F, and you believe you will F if and only if you intend to F, you
intend to F.
But for the purposes of this paper, I shall stick to the simplified version
I started with.
As a further concession, I am happy to strengthen the enkratic condition by adding an explanatory or counterfactual clause. We get (also
adding a universal quantifier):
Strengthened enkratic condition. Necessarily, if you are rational, then,
whenever you believe your reasons require you to F, you intend to F and
an appropriate explanatory or counterfactual connection holds between
your belief that your reasons require you to F and your intending to F.
This condition constitutes an interpretation of modified entailment.

We have only to interpret you respond correctly to beliefs about reasons to mean that, whenever you believe your reasons require you to F,
you intend to F and an appropriate explanatory or counterfactual connection holds between your belief that your reasons require you to F
and your intending to F.
In this paper, I am looking for the truth in the idea that rationality
consists in responding correctly to reasons. I have found that one part
of it, or more accurately a modification of one part of it, is true.
Specifically, I have found a way to interpret modified entailment that
makes it true. Rationality entails responding correctly to beliefs about
reasons.
But that is as far as I can go. Rationality is not equivalent to responding correctly to beliefs about reasons. Many other conditions are necessary for rationality. I listed a few in section 3. To be rational, you must
not have contradictory beliefs; you must intend what you believe is
necessary to an end that you intend, and so on. Responding correctly
to beliefs about reasons is only one part of rationality. We have fallen
far short of modified equivalence.
40
john broome
5. Meeting Your Own Standards
In section 4 I insisted that many other conditions are necessary for

rationality besides the enkratic condition. I earlier listed some in section 3. However, I have no basis for these conditions apart from intuition. I admitted that the formulation of some is only approximate.
Moreover, there is a general reason to doubt conditions of this sort.
Take as an example the claim that, necessarily, if you are rational
you do not have contradictory beliefs. A paraconsistent logician sees
nothing wrong with having some pairs of contradictory beliefs, because
she believes that some contradictions are true.7 If a paraconsistent
logician believes p and believes not p, is she necessarily irrational?
Plausibly not. Even if paraconsistent logic is false, we might not think it
irrational to believe it. And if you do believe it, plausibly you are not
irrational if you have patterns of belief that conform to it.
This suggests we might weaken the condition against contradictory
beliefs. For example, we might make it:
Necessarily, if you are rational, you do not believe p and believe not p,
unless you believe there are true contradictions.
I would be happy to accept some weakened formula such as this.

But the example of a paraconsistent logician may suggest a more
radical alteration to what I have been saying. Conditions of rationality
such as the ones I mentioned all impose strict liability in the sense
I adopted in section 3. Each asserts that you are irrational if you fail to
meet some condition, and this is so whatever your own beliefs are
about the matter. It makes no difference whether or not you believe
your reasons require you to meet the condition, or what you believe
about the rationality of meeting it. In view of the example of paraconsistent logic, this sort of strict liability may seem too unforgiving. We
might look for more liberal conditions of rationality.
As an alternative, we might think of rationality as meeting your
own standards, so that you are irrational only if you fail by your own
standards.8 The direct enkratic condition from section 4 offers a way to
capture this thought. I repeat it here:
7
See Graham Priest and Koji Tanaka, Paraconsistent Logic, in Stanford Encyclopedia
of Philosophy, http://plato.stanford.edu/.
8
This is the view of rationality presented by T.M. Scanlon in What We Owe to Each
Other (Cambridge, MA: Harvard University Press, 1998), pp. 1832.

Direct enkratic condition. Necessarily, if you are rational, then, if you
believe your reasons require you to F, you F.
I explained in section 4 that this condition is not true universally,

because it is inconsistent with the supervenience of rationality on the
mind. But for this section only, I shall restrict the range of verb phrases
that can be substituted for F to ones that describe being in a particular
state of mind. For instance, Fing might be intending to go to Venice, or
it might be not having contradictory beliefs. With this restriction, the
condition is consistent with the supervenience of rationality on the
mind.
One instance of the direct enkratic condition, restricted this way, is:
necessarily, if you are rational, you do not have contradictory beliefs
if you believe your reasons require you not to have contradictory
beliefs. Another is: neces-sarily, if you are rational, you intend what you
believe is necessary to an end that you intend, if you believe your reasons require you to intend what you believe is necessary to an end that
you intend. And so on. The direct enkratic condition gives us a liberal
version of each of the conditions of rationality that I have mentioned.
It tells us that to be rational you must satisfy each of these conditions,provided that you yourself believe your reasons require you to
satisfy it. For another example, it gives us this liberal version of the
enkratic condition: necessarily, if you are rational, you intend to do
what you believe your reasons require you to do, if you believe your
reasons require you to intend to do what you believe your reasons
require you to do.
Once we notice this consequence of the direct enkratic condition,
we might decide to give up all the separate conditions of rationality. We
might replace them with the direct enkratic condition on its own,
which gives us a liberal version of each.
If all the conditions of rationality could be replaced by just this one,
it would allow us to say that rationality is equivalent to responding correctly to beliefs about reasons. The following chain of argument would
get us to that conclusion.
First, we strengthen the direct enkratic condition by adding an
explanatory or counterfactual condition (and a universal quantifier)
to get:
Strengthened direct enkratic condition. Necessarily, if you are rational,
then, whenever you believe your reasons require you to F, you F and an
appropriate explanatory or counterfactual connection holds between
your belief that your reasons require you to F and your Fing.
42
john broome
Compare modified entailment from section 4. I repeat it here:

These two formulae are equivalent if we interpret you respond correctly to beliefs about reasons to mean that, whenever you believe your
reasons require you to F, you F and an appropriate explanatory or
counterfactual connection holds between your belief that your reasons
require you to F and your Fing. So the strengthened direct enkratic
condition, like the strengthened enkratic condition, constitutes an
interpretation of modified entailment. It says that rationality entails
responding correctly to beliefs about reasons.
We can go further. We are pursuing the thought that the direct enkratic condition replaces all the separate conditions of rationality. If it
does, that one condition encompasses the whole of rationality.
Consequently, we could convert modified entailment from a one-way
conditional statement to a biconditional. We would arrive at modified
equivalence. We would have shown that rationality is equivalent to
responding correctly to beliefs about reasons. This sections particular
interpretation of this idea is that rationality is equivalent to meeting
your own standards.
I have two objections to this conclusion. The first is to the last step of
the argument: modified entailment cannot be extended to modified
equivalence as I have just suggested it can. The second is an objection
to the direct enkratic condition, which means it is an objection to modified entailment itself, interpreted as the strengthened direct enkratic
condition.
Objection to Modified Equivalence
For the moment, let us suppose the strengthened direct enkratic condition is true. I shall argue it cannot possibly be the only condition of
rationality. So modified equivalence does not follow.
The direct enkratic condition is supposed to replace all the various
more particular conditions of rationality that I mentioned. For instance,
instead of the condition that you do not have contradictory beliefs, we
have a particular instance of the direct enkratic condition: that you do
not have contradictory beliefs if you believe your reasons require you
not to have contradictory beliefs.
The particular conditions of rationality assert strict liability: you are
not rational if you do not conform to them, whatever you believe about
the conditions themselves. Replacing them with the direct enkratic

condition removes strict liability. Whether or not you are rational
depends on your own beliefs about conditions of rationality. Your
rationality is judged by your own standards.
That is a nice liberal thought. But however liberal we wish to be, your
rationality cannot be judged entirely by your own standards. The direct
enkratic condition itself asserts strict liability. If you believe your reasons require you to F but you do not F, the direct enkratic condition
asserts you are irrational. It asserts that this is so, whether or not you
believe the reasons require of you, that, if you believe your reasons
require you to F, you F.
So some strict liability is inevitable. Given that, we have no strong
liberal motive to restrict it to the direct enkratic condition alone. And
there is an obvious reason why we cannot replace all particular conditions of rationality with the direct enkratic condition. The following
example illustrates it.
Compare two people. One believes her reasons require of her that,
whenever she intends to F, she does not intend not to F. The other
believes her reasons require of her that, whenever she intends to F, she
also intends not to F. Suppose that the first has no contradictory intentions. But suppose the second has only contradictory intentions, by
which I mean that each of her intentions is contradicted by another.
So far as the direct enkratic condition is concerned, both these people might be rational, since they meet their own standards. But they are
certainly not on a par so far as rationality is concerned. The second
cannot be entirely rational, since she has contradictory intentions.
There must therefore be some condition of rationality that she fails to
satisfy. She satisfies the direct enkratic condition, so there must be
some other one.
Rationality cannot be equivalent to meeting your own standards,
because some peoples standards are more in accordance with rationality than other peoples are.
Objection to the Direct Enkratic Condition
For this section, I restricted the direct enkratic condition in a way that
makes it consistent with the supervenience of rationality on the mind.
But the following example shows that even the restricted version is
false. Supervenience is not the only objection to the direct enkratic
condition. Consequently, modified entailment, as I interpreted it in
this section, is false.
44
john broome
Suppose you do not believe God exists, but you believe your reasons
require you to believe he does exist. According to the direct enkratic
condition, restricted to states of mind, you are irrational. But that is not
necessarily so.
Suppose you believe there is no evidence of Gods existence, and that
is why you do not believe he exists. Indeed you are unable to believe he
exists. Nothing you could dogoing regularly to church, or taking a
course of religious instruction, or anything elsewould bring you to
believe he exists. However, you believe your reasons require you to
believe God exists, on grounds of personal safety. You believe it is a
good idea to believe he exists, just in case he does. You may be perfectly
rational in having these beliefs.
To be sure, your belief that you ought to believe God exists may be
false. At least two arguments can be raised against it. One comes from
evidentialism. Evidentialists think personal safety is no reason to
believe God exists; only evidence can constitute a reason. So they think
you are wrong to believe your reasons require you to believe God exists.
Still, even if evidentialism is true, you need not be irrational in having
a different opinion. Your belief that your reasons require you to believe
God exists will be false, but it need not be irrational.
A second argument comes from the principle that ought implies can.
Since you cannot believe God exists, perhaps it follows that it is not the
case that your reasons require you to believe God exists. Still, even if it
is true that ought implies can in this context, you need not be irrational
in believing the opposite. Once again, even if your belief is false, it need
not be irrational.
That shows the restricted version of the directed enkratic principle is
false. On the other hand, what is its attraction? Why should we think
you are necessarily irrational if you are not in a mental state that you
believe your reasons require you to be in? You would indeed be irrational if your mental states were controlled by your intentionsif, necessarily, whenever you intended to be in a mental state, you were in it.
The enkratic condition tells us you would necessarily be irrational if
you did not intend to be in a mental state you believed your reasons
required you to be in. Consequently, you would necessarily be irrational if you were not in it. But actually many of your mental states,
including many of your intentional attitudes, are not controlled by
your intentions.
We might be inclined to believe the direct enkratic condition even
so. We might think a rational person simply has a psychological
disposition to be in a mental state whenever she believes her reasons
require her to be in it. Her disposition must be maintained by some

causal process, but the process need not involve her intending to be in
the mental state. It could be an automatic, subpersonal process.
But the example shows this is not credible. You believe there is no
evidence for Gods existence, but you believe your reasons require you
to believe God exists. No automatic process will cause you to believe
God exists. Indeed, an automatic process will probably cause you not to
believe God exists. Any automatic process will follow your belief about
evidence, rather than your belief about what your reasons require you
to believe.
So we have to reject the direct enkratic condition, even restricted to
mental states. There is no good reason to believe it, and the counterexample shows it is false.
Scanlons Condition
T.M. Scanlons paper Structural Irrationality9 suggests an alternative
condition that is not subject to this objection. Scanlon thinks that some
of our attitudes are judgment-sensitive, as he puts it. Beliefs and intentions are, for instance. Judgment-sensitive attitudes he earlier defined
as attitudes that an ideally rational person would come to have whenever that person judged there to be sufficient reason for them.10
The direct enkratic condition implies that a rational person has an
attitude whenever she believes her reasons require her to have it. So at
first Scanlons view seems to be just an application of the direct enkratic
condition. But it is not, because Scanlon intends sufficient reason to
refer only to reasons of a particular sort. He says:
We need to narrow the relevant reasons to what Parfit calls object-given
reasons: evidence for p, in the case of belief that p, and reasons for doing
A, in the case of intention to do A.11
Derek Parfit contrasts object-given reasons with state-given reasons.12

State-given reasons to be in a particular state are ones that have to do
with the state itself. If it would be safest to believe God exists, that is a
9
T.M. Scanlon, Structural Irrationality, in Robert Goodin, Frank Jackson, Michael
Smith and Geoffrey Brennan (eds.), Common Minds: Themes from the Philosophy of
Philip Pettit (Oxford: Oxford University Press, 2007).
10
Scanlon, What We Owe to Each Other, p. 20.
11
Scanlon, Structural Irrationality, op. cit.
12
Derek Parfit, Rationality and Reasons, in Dan Egonsson, Jonas Josefsson, Bjrn
Petersson and Toni Rnnow-Rasmussen (eds.), Exploring Practical Philosophy: From
Action to Values (Alderhot: Ashgate, 2001), pp. 1939.
46
john broome
merit of the state of belief, so it is a state-given reason (if it is a reason

at all) to believe God exists. Object-given reasons are ones that have to
do with the states object. Scanlon specifies what he takes to be objectgiven reasons for beliefs and intentions: for beliefs, evidence, and for
intentions, reasons for doing the thing intended.
Scanlons definition of a judgment-sensitive attitude amounts to this.
When Fing is a judgment-sensitive attitude,
Scanlons condition. Necessarily, if you are rational, then, if you believe
your object-given reasons require you to F, you F.
This is a variant on the direct enkratic condition. But the formula has
to be interpreted carefully if it is to catch Scanlons meaning correctly.
We must give a particular interpretation to the clause if you believe
your object-given reasons require you to F.
One way to satisfy the condition expressed in this clause would be to
have a belief that you could express using a sentence of the form My
object-given reasons require me to F. But hardly anyone ever has that
sort of a belief, because hardly anyone has the concept of an objectgiven reason. Another way to satisfy the condition is by having a belief
that you could express using a sentence of the form X requires me to F,
where X (whether you know it or not) constitutes your object-given
reasons. For instance, you satisfy the condition if you have a belief that
you could express by the sentence The evidence requires me to believe
the climate is changing. A technical way of putting this is that believe
in the conditional clause sets up a referentially transparent context.
We can make the transparency explicit when Fing is believing or
intending. For believing and intending, Scanlons condition comes
down to:
Necessarily, if you are rational, then, if you believe your evidence requires
you to believe p, you believe p.
Necessarily, if you are rational, then, if you believe your reasons require
you to F, you intend to F.
The first of these formulae is plausible; I do not deny it. The second is
simply the enkratic condition, which I accept.
So Scanlons condition may well be true of beliefs and intentions.
It makes better psychological sense than the direct enkratic condition.
We can expect automatic, subpersonal processes to follow your beliefs
about object-given reasons rather than your beliefs about state-given
reasons. In my example, you believe it would be safest to believe God
existsa state-given reason to believe he exists. But you believe the

evidence does not support Gods existencean object-given reason
not to believe he exists. Automatic processes, following the evidence,
are likely to stop you believing God exists. In general, automatic processes help to make beliefs judgment-sensitive in Scanlons sense. The
same is probably true for intentions.
It may be true for some other attitudes too. Scanlons condition may
be true of various attitudes. I have introduced it because it is a variant
of the direct enkratic condition that, unlike the direct enkratic condition itself, is true in at least some applications. It is not subject to the
objection I raised against that condition. We could substitute it for
thedirect enkratic condition in the argument of this section. If we did,
the argument would show that, strengthened with an explanatory or
counterfactual condition, Scanlons condition is a version of modified
entailment. So here is another interpretation of modified entailment
that may well be true.
I am pursuing whatever truth resides in the idea that rationality consists in responding correctly to reasons. In section 4 I identified one
piece of truth in it: the strengthened enkratic conditionone interpretation of modified entailmentis true. Scanlons condition may add
another bit of truth. Strengthened, it constitutes another interpretation
of modified entailment. Unfortunately, the truth it can add is not much.
Part of Scanlons condition turns out to be the enkratic condition itself,
whose truth we have already recognized.
Can this condition take us any further? Does it give any support to
modified equivalence? It does not. It is not subject to my objection to
the direct enkratic condition, but it offers no hope of overcoming my
objection to modified equivalence. It cannot possibly replace all the
other conditions of rationality, such as the ones listed in section 3.
6. Responding Correctly to Non-normative Beliefs
The quick objection shows that rationality is not equivalent to responding correctly to reasons. It led us to consider instead the modified idea
that rationality is equivalent to responding correctly to beliefs about
reasons, or at least that it entails responding correctly to beliefs
aboutreasons. The quick objection does not tell us just what the content of those beliefs must be. They might have a normative content;
sections 4 and 5 considered that possibility. I concluded in section 4
48
john broome
that rationality does entail responding correctly to any belief whose

content is that your reasons require you to F. That is the strengthened
enkratic condition, which I accepted.
It is also possible that rationality entails responding correctly to
some of your beliefs that have a non-normative content. Even some
beliefs with a non-normative content may fairly be called beliefs about
reasons. Suppose you believe there is a tiger-snake in the path ahead.
Your belief has a non-normative content. But suppose that, if there was
a tiger-snake on the path ahead, that fact would constitute a reason for
you to stop. Then what you believe would, if true, be a reason for you to
stop. So we may fairly call it a belief about a reason. It is a non-normative belief about a reason, or a non-normative belief in a reason. We
may also fairly say you believe a reason exists, though you may not
believe it is a reason.
You can respond to your belief by stopping. You may do so automatically: as soon as you acquire the belief that there is a tiger-snake in the
path head, you automatically stop. In doing this, you may not form any
normative belief, such as the belief that you have a reason to stop. You
might not even have the concept of a reason. Yet you could still respond
to your belief that there is a tiger-snake on the path ahead, by
stopping.
Since you can respond to non-normative beliefs of this sort, we may
ask whether rationality is equivalent to, or entails, responding correctly
to them. Parfit says:
We are rational insofar as we respond to reasons, or apparent reasons. We
have some apparent reason when we have some belief whose truth would
give us that reason.13
I think Parfit is saying that rationality is equivalent to responding correctly to what I have called non-normative beliefs about reasons. This
section considers whether that is so.
Suppose you believe there is a tiger-snake on the path ahead, and
you respond by stopping. Do you respond correctly to your belief?
Probably not. If there is a tiger-snake on the path ahead, that is one
reason for you to stop. But probably you have other reasons either to
stop or not to stop. Perhaps you need to get back home, or perhaps you
are enjoying the exercise. All these reasons together determine whether
Parfit, Rationality and Reasons, p. 25; original emphasis.
13
or not your reasons require you stop. There is therefore no such thing
as responding correctly to just one of them. You can respond correctly
only to all of them together; you do so by doing what all of them
together require. Consequently, we could not count you as responding
correctly to a non-normative belief in just one of the reasons.
Could you ever respond correctly to a non-normative belief about a
reason? The tiger-snake does not provide an example, but are there any
examples? Suppose you have a non-normative belief that a reason
exists, and suppose that reason, if it did exist, would be enough on its
own to make it the case that your reasons require you to F. I call a belief
of this sort a perfect reason. Then, if you F (and if an appropriate
explanatory or counterfactual connection holds between your belief
and your Fing), we could count you as responding correctly to your
belief.
An Example
Here is a plausible example of that sort. Suppose, gazing at the horizon
at night, you believe you see a red light. What you believe is that you see
a red light. Plausibly, if true, this would be a perfect reason to believe
you see a coloured light. Therefore, if you respond to your belief
thatyou see a red light by also believing you see a coloured light, plausibly you are responding correctly.
Moreover, it is also plausible that, unless you respond correctly in
this way, you are not rational. Necessarily, if you are rational you
respond correctly to this reason. Rationality entails responding correctly to it. I do not assert this is true; only that it is plausible. I cannot
think of any more convincing example. So for the sake of pursuing the
argument further, I shall assume it is true.
If it is, it illustrates another sort of strict liability. You are irrational if
you do not respond correctly to your non-normative belief about a reason, and this is so whatever your normative beliefs may be. Section 3
examined strict liability in responding to a reason. Here we have strict
liability in responding to a belief about a reason.
This raises a question. The claim is that you are strictly liable for
responding correctly to a non-normative belief of yours, whatever your
normative beliefs may be. But might you not have a normative belief
that imposes a conflicting liability? Is that a problem?
In the example, you have the non-normative belief that you see a red
light. I said that, plausibly, you are not rational unless you believe you
see a coloured light. But suppose you also have the normative belief
50
john broome
that your reasons require you not to believe you see a coloured light.
Perhaps you know you will be severely punished if you believe you see
a coloured light. According to the direct enkratic condition, you are
not rational if you believe you see a coloured light. So it seems you are
not rational whether or not you believe you see a coloured light. Is that
a problem?
It is not, because I have already rejected the direct enkratic condition
in section 5. This may be a case like the one where you believe your
reasons require you to believe in God. Suppose, given that you believe
you see a red light, nothing you can do would stop you from believing
you see a coloured light. Even if you believe your reasons require you
not to believe that, there need be nothing irrational about you if you do
believe you see a coloured light.
An alternative reaction to the difficulty would be say you must be
irrational anyway, because your normative belief is irrational. If you
can see a red light, you must be irrational in believing your reasons
require you not to believe you see a coloured light. But that is not credible. To be sure, according to evidentialism your normative belief is
false. According to evidentialism, your evidence provides all the reasons you can have for believing or not believing anything. In this case
you have conclusive evidence that you see a coloured light, so you have
a conclusive reason to believe you see a coloured light. The fact that you
will be punished for doing so provides no reason against it. But whether
or not evidentialism is true, it is debatable. We must not impugn a persons rationality just because she has a belief that conflicts with
evidentialism.
In any case, in one way or another, we can accept strict liability
for this example. Plausibly you can respond correctly to your nonnormative belief about a reason, and you are not rational unless you do.
Practical Reasons
The example of seeing a red light is contrived, and it is an example of a
reason to believe. Can we find an example among practical reasons
reasons to do something?
Before I try out an example, I need to make an adjustment to the
notion of responding correctly to a non-normative belief. Suppose you
have a non-normative belief whose content would, if true, be a reason
to do some act. Suppose the act is not a mental one. Then it could not
be the case that, necessarily, if you are rational you do this act. That
condition would violate the supervenience of rationality on the mind.

You might fail to do the act because of some obstruction outside your
mind, which has nothing to do with your rationality. So if we insist
that, to respond correctly to a non-normative belief in a reason to act,
you must do that act, modified entailment will inevitably be false
whenever the act is not a mental one.
I have already been through this argument in section 4. There it led
me to adjust the notion of responding correctly to a normative belief;
here I shall correspondingly adjust the notion of responding correctly
to a non-normative belief. To respond correctly to a non-normative
belief in a perfect reason to do something, you must intend to do that
thing. You do not have to do it.
We are looking for an example of a non-normative belief in a practical reason such that rationality entails responding correctly to this reason. Derek Parfit offers one.14 Suppose you are due to have an operation
next Tuesday. However, owing to a cancellation, you have the chance of
changing your appointment to Wednesday. Anaesthetics will be available on Wednesday but not Tuesday. So if you change your appointment, you will have only slight pain on Wednesday, but if you do not
change it, you will have agony on Tuesday.
To this non-normative fact, let us add others such as: the surgeon
will be just as careful on either day; you have nothing else to do on
Wednesday; Tuesday is equally as real as any other day of the week; if
the calendar had been designed differently, a different day would have
been Tuesday; and so on. Take the big conjunction of all these nonnormative facts. I shall assume that enough facts are included to ensure
that this conjunction constitutes a perfect reason for you to change
your appointment to Wednesday. It makes it the case that your reasons
require you to change your appointment.
Suppose you believe the big conjunction. What you believe would, if
true, be a perfect reason for you to change your appointment. To
respond correctly to this belief would be to intend to change your
appointment. Furthermore, according to Parfit, provided we have
included enough facts within the conjunction, you would necessarily
be irrational if you had this belief and did not intend to change your
appointment. So Parfit thinks this is an example of what we are lookingfor. You have a non-normative belief in a reason to do something,
14
Derek Parfit, Reasons and Persons (Oxford: Oxford University Press, 1984),
p. 124. Parfit further works out this example in Rationality and Reasons.
52
john broome
and necessarily, if you are rational and you have this belief, you intend
to do this thing. That is to say, rationality entails that you respond correctly to your belief.
This is another assertion of strict liability. You are not rational if you
fail to respond to your non-normative belief, whatever your own beliefs
may be about the normative situation. This time I am not convinced.
Suppose you have a normative belief that conflicts with your nonnormative one. Suppose you believe that your reasons require you not
to change your appointment. That might be because you have some
false, bizarre normative theory according to which you should avoid
pain on Wednesdays much more than on Tuesdays. Then the enkratic
condition entails that you are irrational unless you intend not to change
your appointment. But Parfit thinks you are irrational unless you
intend to change your appointment. We may take it for granted that
you are irrational if you have both intentions; since they contradict
each other, they violate C2 among the conditions of rationality listed in
section 3. So, whatever intentions you have, you are irrational.
What should we conclude? A similar problem came up with the red
light example. But there the conflict between your normative and your
nonnormative belief was generated by the direct enkratic condition.
Isolved it by rejecting that condition. In our present practical example,
the conflict is generated by the enkratic condition. The enkratic condition is on stronger ground, and I accept it.
We might instead claim you are irrational because of your bizarre
norma-tive theory. We might say it cannot be rational to hold such a
bizarre theory. But I see no grounds for that claim. There is no logical
inconsistency among your beliefs; your belief that your reasons require
you not to change your appointment is not logically inconsistent with
your non-normative belief in the big conjunction. Moreover, your
belief in a bizarre normative theory may be supported by all the evidence you have. Perhaps you acquired it during your upbringing by
testimony from people you had good reason to believe were reliable.
We are not entitled to impugn your rationality just because you hold a
false and bizarre normative theory.
I conclude that you may rationally intend not to change your
appointment, despite your non-normative belief whose content would,
if true, be a perfect reason to change it. The content would, if true,
make it the case that your reasons require you to change your appointment. But that is only because of the normative facts, and you do not
believe those normative facts. Instead, you have a normative belief that
implies your reasons require you not to change your appointment. You
may not be irrational in having this belief. Consequently, you may
intend not to change your appointment, and nevertheless be rational.
A parallel argument could be brought to bear on any non-normative
belief in a practical reason. You may be rational even if you fail to
respond correctly to it.
That need not be the end of the story. Derek Parfit has pointed out to
me a way of going further. The problem revealed in the example is that
your non-normative belief might conflict with a normative belief of
yours. If you respond correctly to your normative belief, you will not
respond correctly to your non-normative belief. That is why you may
be rational even if you do not respond correctly to your non-normative
belief. But suppose you have no contrary normative belief. Then you
have no grounds for failing to respond correctly to your nonnormative belief. In that case, we might conclude you are necessarily
irrational if you do not respond correctly to it.
That may be correct for Parfits particular example of pain on
Tuesday. In this example, it is very obvious that what you believe, if
true, would be a perfect reason for you to change your appointment.
Given that, perhaps it takes a normative belief (which would have to be
bizarre) to oppose its obviousness, and make it rational for you not to
change it. I have no firm view about this example.
However, the rule does not generalize. The question is this. Suppose
you have a non-normative belief in a reason. Suppose that what you
believe, if true, would be a perfect reason for you to F. Suppose you
have no opposing normative belief. Are you necessarily irrational if
you do not respond correctly to your belief? That cannot be so in general. It need not be obvious that what you believe constitutes a perfect
reason for you to F. It might be a perfect reason only because of some
fact that you do not believe and to which you have no easy access. In
that case, you may fail to F and yet be rational.15
For example, suppose you believe this liquid is orange juice. Suppose
that, if it is indeed orange juice, that is a perfect reason for you not to
drink it, because you have overnight become so allergic to orange juice
that drinking it will kill you. But suppose nothing has given you any
inkling of this allergys sudden onset. Then you might drink the liquid
and not be irrational in doing so.
15
This point is made by Andrew Reisner in Conflicts of Normativity (unpublished
DPhil thesis, Oxford University, 2004), ch. 2.
54
john broome
Conclusion
This section has investigated the idea that rationality is equivalent to
responding correctly to non-normative beliefs about reasons. These are
the conclusions that have emerged.
First, to many of your non-normative beliefs about reasons, there is
no such thing as responding correctly. You can respond correctly only
if what you believe would, if true, constitute a perfect reason. But many
of your nonnormative beliefs about reasons are about reasons that conflict with others. Your belief about a tiger-snake is an example.
Second, plausibly there are some non-normative beliefs about reasons such that, necessarily, if you are rational you respond correctly to
them. Your belief about seeing a red object is an example. This means
that, at a pinch, we may once again endorse modified entailment.
Ireproduce it here:
We can apply this idea to non-normative beliefs about reasons. But in

doing so, we have to give it an etiolated interpretation. In my previous
interpretations of this formula, I treated it as containing an implicit
universal quantifieras applying to all your beliefs about reasons. This
time we have to treat it as containing an existential quantifier. It says
that some of your nonnormative beliefs about reasons are such that,
necessarily, if you are rational, you respond correctly to them. Only of
these beliefs can we say that rationality entails responding correctly to
them.
Third, it is doubtful that any of your non-normative beliefs about
practical reasonsreasons to actare among the beliefs that satisfythis condition. The example of pain on Tuesdays turns out not to
satisfy it.
Finally, besides modified entailment applied to non-normative
beliefs, there are many other conditions of rationality. The enkratic
condition is one; this is a condition on normative beliefs. Another is
that, necessarily, you are not rational if you have contradictory intentions. Others are set out in section 3. So rationality is very far from
equivalent to responding correctly to non-normative beliefs about
reasons. Responding correctly to some of these beliefs may be a part of
rationality, but it is very far from the whole of it.

7.Conclusion
In this paper I have investigated the idea that rationality is equivalent

to responding correctly to reasons. I have followed the idea through
many variations. I hope I have done enough to scotch it.
In sections 2 and 3 I rejected equivalence as it stands. In the following three sections I moved on to modified equivalence, the idea that
rationality is equivalent to responding correctly to beliefs about reasons. I concluded there is some truth in one part of it: modified entailment, the idea that rationality entails responding correctly to beliefs
about reasons. The strengthened enkratic condition constitutes one
interpretation of modified entailment, and I accepted it as true. In sections 5 and 6 I found some truth in two other interpretations of modified entailment.
However, there are necessary conditions of rationality that are not
captured by modified entailment under any interpretation. For example, to be rational you must not have contradictory intentions. So
rationality is not equivalent to responding correctly to reasons, or to
responding correctly to beliefs about reasons.
A fortiori, rationality does not consist in responding correctly to reasons, or in responding correctly to beliefs about reasons. Rationality
must be an independent source of requirements in its own right.
Practical Reason, Value and Action

Alison Hills*
1.Introduction
Practical reasons are considerations that favour or disfavour action.
They have normative force. For example, eating a hearty breakfast
givesyou more energy throughout the day and makes you less likely to
eat chocolate before lunch. These considerations may favour eating
breakfast heartily, and so may be practical reasons for doing so. Eating
a hearty lunch, by contrast, sends you to sleep for the afternoon. This
may be a consideration against lunching too well, and so a reason
against it.
Though we have an intuitive grasp on what practical reasons are and
which considerations favour or disfavour action, there are many rival
theories of practical reason. Are all practical reasons based on the
agents desires? Are they all reasons to maximize some good? Or do
you have a reason not to kill, for example, even if you could thereby
minimize the number of killings?
How should we decide which theory of practical reason is correct?
One possibility is to link each conception of practical reason with a
theory of value, and to assess the first in combination with the second.
So, for example, one argument in favour of maximizing consequentialist theories of practical reason is that practical reasons are reasons to
promote the good, and more good is always better than less. Scanlon
has tried to argue against this view by criticizing the associated teleological theory of value, claiming that some values are to be responded
to in ways other than just promotion, and so theories of practical reason with non-maximizing elements are to be preferred.1
Recently some philosophers have taken a different approach. They
have tried to link theories of practical reason with theories of action
* Thanks to audiences at the University of Essex and the University of Cambridge

Metaethics Conference for helpful questions and comments.
1
T.M. Scanlon, What We Owe to Each Other (Cambridge, MA: Harvard University
Press, 1998).
58
alison hills
instead, drawing an analogy with recent work on the connection

between epistemic reasons and belief. The underlying idea is simple.
Epistemic reasons are reasons for belief. To know what will count as
good or bad epistemic reasons (and perhaps what will count as an epistemic reason at all) we need to know what a belief is, what its function
is and what counts as correct belief. Belief, they claim, has a constitutive aim, an aim in virtue of being a belief. Belief aims at the truth. This
constitutive aim grounds a clear standard of success or correctness for
belief.
If the standard of success for belief is truth, does it follow that a false
belief is not really a belief at all? Obviously not. True and false beliefs
are both kinds of belief, in virtue of their aiming at the truth (unlike
other kinds of propositional attitudes, such as fantasies or assumptions).
True beliefs achieve this aim; false beliefs do not. False beliefs are therefore genuine beliefs but they are failures, according to a standard of
evaluation that is internal to belief itself. Velleman, for example, has
defended this kind of view.2
It is not straightforward to move from this intuitive idea to work out
the details of a theory of reasons for belief from the fact that a successful or correct belief is a true belief. Reasons for belief must in some
sense be connected to the truth of the beliefbut in what way connected? Nevertheless, it seems that it might be possible to make some
progress on this front.
Since it is not unreasonable to expect some success with deriving
epistemic reasons from the standards of correctness for belief and perhaps from its constitutive aim, we might hope to do something similar
for action. Korsgaard and Velleman in particular have tried to give an
account of practical reason on the basis that actions too have a constitutive aim, and hence that there are standards of success intrinsic to
action. That it aims at meeting this standard can be used to distinguish
actions from non-actions, but the standards themselves can be used to
evaluate amongst actions, to assess actions as successful or failed.
Both Velleman and Korsgaard have extremely large ambitions for
this programme. Both hope to defend a substantive conception of
practical reason (in both cases broadly Kantian) by linking it to the
standards constitutive of successful action. Both also expect to make
2
J.D. Velleman, The Aim of Belief , in his The Possibility of Practical Reason (Oxford:
Oxford University Press, 2000).
practical reason, value and action59
progress in explaining the normativity of practical reason in the

same way.3
In this paper, I have a much more modest goal. I will try to show that
it can be illuminating to think of practical reason in terms of the success conditions of action, but ultimately this is in addition to, rather
than a substitute for, relating practical reason to value as well.4
I will make an assumption in this paper that practical reason is
related to action, in the following intuitive sense:
Practical reasons are reasons to perform actions that meet the standards
of success for action.5
In other words, practical reasons are reasons for action, but more precisely, they are reasons to perform actions that meet the standards of
success intrinsic to action. It is perhaps not obvious what it means to
have reasons to meet these standards, but I hope that this will become
clear through a number of examples. On the assumption that this intuitive connection holds, I will set out three different conceptions of
action and corresponding success conditions, and explain how each is
linked to a particular conception of practical reason and, in two cases,
to a theory of value too. My goal is to describe these different accounts,
rather than to defend any in particular, though I will suggest that some
are more satisfactory than the others.
3
See C.M. Korsgaard, The Locke Lectures (forthcoming, available online at:
http://www.-people.fas.harvard.edu/~korsgaar/) and J.D. Velleman, Practical Reflection
(Princeton: Princeton University Press, 1989), and his The Possibility of Practical
Reason. Korsgaard thinks that acknowledging these standards of success will leave us
in a position to answer the sceptic who challenges the authority of practical rationality.
Velleman hopes that this will turn out to solve the problem of how an agents practical
reasons are related to her desires.
4
In this paper, I discuss only practical reasons, rather than practical reasoning or
practical rationality. I do not expect the link between practical reasons and correct
practical reasoning or practical rationality to be straightforward, so the argument here
should not be taken to endorse any particular view about practical reasoning or practical rationality.
5
I am therefore assuming for the purposes of this paper that practical reasons are
essentially reasons to act, rather than reasons for emotions, desires or anything else.
We do often talk of reasons for wanting X or for feeling sad or angry. But here I will
assume that, insofar as these are practical reasons, they are reasons to take steps in
order to feel the relevant emotion or to acquire the pertinent desire: in other words,
they are reasons for action after all. If there are reasons simply to have a certain
emotion, for example, rather than to bring it about, they are of a very different kind
from the practical reasons described in this paper, and so, I will assume, deserve to be
treated differently and labelled differently too.
60
alison hills
2.Execution
Perhaps the simplest, most basic conception of action is that action

aims at the execution of the agents aim, goal, desire or intention (for
simplicity, I will not distinguish between these here).6 We can call this
action as execution.
According to action as execution, behaviour is classed as an action if
it aims at the execution of the agents goal, and this aim gives action
an internal standard of success.7 If a person successfully executes her
goal, her action is a success. If she does not, her action has failed. She
may still have performed a genuine action, but one that is a failure
according to the standards internal to action itself.
For example, suppose that I want to eat a hearty breakfast. I cook
myself a large bowl of porridge and eat it. My action is successful. But
if I burn the porridge and it turns out to be inedible, I dont get what
I want. My action obviously has failed.
If practical reasons are simply reasons to perform successful actions,
and successful actions are those in which the agent successfully executes her aim or desire, then practical reasons are reasons for the agent
to take the means to satisfy her aim or desire. That is, they are instrumental reasons to achieve some arbitrary end that the agent happens to
want. We can call this the simple conception of action as execution.
6
The successful execution of your aim involves more than that your aim causes you
to move in such a way that it is satisfied. This is for reasons familiar from Davidsons
example of the climber who is caused by his desire to be rid of the extra weight to let
the rope holding his companion slip from his grasp: the desire causes the desired outcome but not in the way appropriate for action (D. Davidson, Freedom to Act, in his
Essays on Action and Events [Oxford: Oxford University Press, 1980], p. 79). If you are
to act, executing your aim, your aim must cause you to move in the right way. It is very
difficult properly to characterize what the right way is, but it involves your having
control over what you do, in order to satisfy your desire. Davidsons climber does not
execute his aim because he does not control what he does; and so he does not count as
acting at all. This control may be conscious, but it may not be (this conception of action
includes both activity, in which the agent has control of what he does, but not consciously, and autonomous action, where the agent consciously controls what he does;
cf. Velleman, The Possibility of Practical Reason, pp. 19399).
7
What does it mean for an action (or a belief ) to aim at anything? Velleman, in The
Possibility of Practical Reason, understands this aim in terms of sub-personal motives
(e.g. pp. 2029, 17099). I do not intend to commit myself to any particular account of
this constitutive aim: all that this means is that it is appropriate to evaluate types of
behaviour (or propositional attitudes) by the relevant standards given by that aim. But
I do discuss in more detail whether the aim relevant to each conception of action can
reasonably be attributed to the agent herself. Since agents do aim to execute their goals,
the aim relevant to this conception of action can be attributed to the agent.
There is something quite attractive about the idea that the basic evaluation of an action as a success or failure is in terms of whether the
agent achieved his aim or not. And there is also something appealing
about tying reasons for action to the success conditions of action. But
unfortunately the associated theory of practical reason does not fit at
all well with our intuitive grasp of what practical reasons we have. It is
simply not plausible that agents have practical reasons to satisfy their
desires, no matter how ill considered, nor is it plausible that they have
no reason to do anything that they do not already want to do.
Consider a famous example from Williams.8 Suppose that I think
that a bottle of clear colourless liquid contains gin, whereas in fact it is
full of petrol. If I pour myself a glass and drink it, is my action s uccessful?
I intended to get myself a drink from that bottle and I did. According
to the simple theory, my action is successful because I got what I aimed
at, and I had reason to do what I did. It is not very plausible, however,
that any consideration favoured my drinking petrol, nor that I had any
reason to do so.
There is a worse problem for the simple account, however. One of
my aims in acting was to get a gin and tonic, and I did not get one, so
I failed to achieve my aim. It appears to follow from the simple theory
that my action is a failure and also a success, that I had no reason to do
it and also that I had some reason to do it.
The simple theory appears, therefore, to be incoherent. But it
seems that whilst keeping to the spirit of that theory, we can modify it
to avoid this problem. The success of my action should not be judged
on the basis of an aim of mine (e.g. to drink from that bottle), which
is based on a false belief. The satisfaction of my other goal (to get a
gin and tonic), by contrast, is relevant to an evaluation of my actions
success. This accords with my own judgment of my drinking the petrol,
namely that my action was disastrous, and that I had no reason to
do it.9
8
B.A.O. Williams, Internal and External Reasons, in his Moral Luck (Cambridge:
Cambridge University Press, 1981).
9
There are other ways of restoring the coherence of the simple account. We could
claim that I performed more than one action, one successful (drinking from the bottle), one not (drinking a gin and tonic). This seems to multiply the number of actions
I performed more than is necessary or plausible, however. Alternatively, we could
claim that my action was successful in one respect and not in another. But this conflicts
with my own, very plausible judgment, that my drinking petrol can in no way be
described as a success.
62
alison hills
A more sophisticated version of action as execution considers action

as not merely the carrying out of the agents immediate aim or desire,
but the execution of the aims or desires she would have under certain
conditions, for example, if she were well-informed and had no false
beliefs and perhaps if she had exercised imagination or thought carefully about what she really wanted.10
This theory of practical reason is more plausible than the simple version. For practical reasons are not now restricted to instrumental reasons to satisfy the desires I now happen to have. Sometimes I do not
have reason to satisfy a current desire (it is based on false belief orinsuf
ficient information) and sometimes I have additional reasons: reasons
to satisfy desires I would have had if I had had better information.
This sophisticated version of action as execution supports a conception of practical reasons that are similar to those which BernardWilliams
calls internal reasons. According to Williams, internal reasons are reasons that are grounded in the contents of the agents subjective motivational set (or S), where this includes the agents motives in a broadsense:
desires, emotions, values and so on. Internal reasons are reasons to satisfy the aims that the agent would have by a process of sound deliberation, starting from her S. Sound deliberation includes the removal of
false belief and the addition of relevant information, though in principle it can go beyond these. Practical reasons according to the sophisticated conception of action as execution are all internal reasons.11
For an example in which it is important that the agent is fully informed, consider
the following. Suppose that I want a nice drink and have the choice between a glass of
water and a gin and tonic. Not having experienced the latter, and lacking the capacity
to imagine how delicious it is, I drink the water. I have had a nice drink, but I could
have had a much nicer one. There is a sense in which my action is successful, but a
more important sense in which it is not (or at least, it is not as successful as it might
have been). I may have had some reason to drink the water, but I had more reason to
drink the gin.
11
One particularly interesting consequence of this is that it supports a different
argument from the one that Williams makes against the possibility of external reasons,
that is, reasons that are not based in the appropriate way on the agents subjective motivational set. Williams argues that external reasons cannot play an adequate role in an
explanation of why an agent acted. If the agent can act for the reason, it must be possible for him to be motivated by the reason, so the reason must be related in an appropriate way to his S. So it must be an internal reason.
A different line of argument against external reasons would be the following:
10
1.Reasons for action are essentially connected to the success conditions of action:
all reasons for action are reasons to perform successful actions.
2.An action is successful if it is the successful execution of the agents aims (broadly
conceived to include aims that are not based on false belief or ignorance).
Williams goes further than the sophisticated conception, by relating

reasons for action to those aims that the agent would have if she deliberated soundly. The sophisticated version of action as execution could
be modified further to include these reasons, in which case its associated theory of practical reason would be exactly equivalent to Williamss
conception of internal reasons. Williams is not very clear about what a
sound deliberative route is, and this has been a matter of intense
debate. So it is not completely clear what the resulting theory of practical reasons includes (does it, for example, include reasons to be prudent and reasons to be moral, on the basis that any rational agent would
have these motivations if she deliberated soundly?) Though the modified theory is not very clear, however, it is likely to be an improvement
on the preceding theories in terms of fitting our intuitive judgments
about our reasons for action.
But this extensional adequacy is at a price. For it implies that an
action should count as a failure if it does not satisfy some aim that the
agent would have had if she had soundly deliberated. For example, if
I underwent an extensive process of sound deliberation, I might be
brought to have all kinds of new desires. Perhaps I would have much
more altruistic desires if I deliberated soundly. But it seems very odd
to classify my action as a failure simply because it fails to satisfy a
desire that I would have had if I were in highly unrealistic idealized
circumstances.
3.Therefore all practical reasons are reasons to perform actions that successfully
execute the agents aims (broadly conceived).
4.External reasons for action are reasons to perform actions that are independent
of the agents aims.
5. Therefore no practical reasons are external reasons.
With this line of argument we can expand on a remark of Williamss that an agent who
acts morally badly can be criticized: he may be ungrateful, inconsiderate, hard, sexist,
nasty, selfish, brutal, and many other disadvantageous things (B.A.O. Williams,
Internal Reasons and the Obscurity of Blame, in his Making Sense of Humanity
[Cambridge: Cambridge University Press, 1995], p. 39). Actions can be assessed and
evaluated in many ways by standards that are not internal to action itself, in the sense
that they are not based on the nature of action and its conditions of success. Similarly,
you can criticize a car in ways related to its function (it wont steer round corners, the
brakes keep failing) and in ways that do not (its colour clashes with your dress).
Criticisms of action in moral terms, as sexist, selfish, brutal and so on may well be
legitimate (this argument is therefore independent of metaethical scepticism about
value) but these standards are not related to the nature of action, and what it is to be a
successful or failure as an action. Therefore they are independent of practical reason.
Of course we consistently say that everyone has reason to be kind, considerate and
gentlebut that is an attempt to persuade them to become more sympathetic and nice,
to change their motivationsor as Williams more bluntly puts it, it is bluff.
64
alison hills
The problem is that we began with a quite plausible claim, that success in action was related to achieving your aim. But this linked practical reason to your current desires or goals: a successful action, and
what you had reason to do, was to get what you wanted. It was reasonable to ignore those of your actual desires based on false belief or ignorance: for it made sense to describe your action as failed when you
drank petrol instead of a gin and tonic. You, here and now, would probably class those actions as failed. But it does not make sense to call your
action a failure simply because you have not satisfied some desire that
you do not yet have. This simply does not fit with the conceptionof
action as execution. The only reason to make this modification is
because without it the associated conception of practical reason is
unacceptable. The modification is ad hoc, and on that basis should be
rejected. But without it the theory of practical reason is not adequate.
In addition, if we do say instead that practical reasons are reasons to
satisfy the desires that you would have if you had no false beliefs, and
you had all relevant information, we run into a well-known problem.12
On this account, all reasons for action are instrumental reasons to satisfy your aims. But if you really have a reason to take the means to your
aim, then it seems that you must have a reason to pursue that aim in the
first place (or at the very least, no reason not to). Otherwise, you could
acquire reasons to do stupid, pointless and immoral actions just by
forming the appropriate desires. If you find yourself with a desire for
which you have no reason, for example, a desire to kill yourself and
others, then you have no reason to take the means to satisfy it. Rather,
you have reason to get rid of the desire, if you can. The claim that there
are only instrumental reasons for action seems to be false.13
Discussed, amongst many others, by C.M. Korsgaard, The Normativity of
Instrumental Reason, in G. Cullity and B. Gaut (eds.), Ethics and Practical Reason
(Oxford: Oxford University Press, 1997), J. Broome, Normative Requirements, Ratio
12 (1999), pp. 398419; J. Broome, Normative Practical Reasoning, Proceedings of the
Aristotelian Society, Supplementary Volume 75 (2001), pp. 17593; and D. Enoch,
Agency, Schmagency: Why Normativity Wont Come from What is Constitutive of
Agency, Philosophical Review 115 (2006), pp. 199241.
13
The best response to this objection may be to insist that the problem itself is incoherent. Practical reasons just are reasons related to success in action, where this is the
successful execution of the agents aims. Asking questions about whether or not we
have practical reason to have the aims that we in fact have is simply a mistake. Compare
the parallel argument about belief. Epistemic reasons relate to the correctness conditions of belief; in the crudest terms, they are reasons to have true beliefs rather than
false. Someone might well ask you whether you have reason to form propositional
attitudes that aim at the truth: what reason do you have to form beliefs? There may be
12
There must be something wrong with the combination of the claims

that practical reasons are dependent on the success conditions of action
and that successful action is the successful execution of aims or desires.
Therefore, we must drop either that conception of successful action, or
the link between success in action and practical reason. In this paper,
I am assuming that the latter link holds, so I will consider a different
conception of action to see if it can support the connection.
3.Production
There are problems with the idea of action as execution and the related
conception of practical reason as purely instrumental. But this is not
the only way to think of action. A very common and appealing conception is that action essentially aims at getting good results, or more formally, at the production of a state of affairs that is in some way good.
Successful action achieves this aim, that is, it actually produces goods.
Unsuccessful action does not. Practical reasons, on this view, are reasons to perform actions which will produce results that are as good as
possible. If a state of affairs would be better than the alternatives, the
agent has reason to bring it about; otherwise not. We can call this conception of action, action as production.14
a number of possible responses that you could make, but it would be a confusion
to think that you could answer in terms of epistemic reasons. This is the wrong sort
of reason. Epistemic reasons are reasons to have correct (true) beliefs rather than
incorrect (false) ones, as it were, rather than reasons to have beliefs at all. Similarly, it is
a confusion to ask if you have practical reason to act: practical reasons are reasons to
act successfully (satisfying your desire) rather than unsuccessfully (failing to achieve
your aim), not reasons for having some particular aims rather than others. But whilst
this response to the problem is coherent, I think, it is not convincing. For it still seems
that if you have a reason of kind K to F in order to G correctly or successfully rather
than unsuccessfully, then you must have a reason of some kind (not necessarily of
type K) to G. If you have epistemic reason to form true beliefs rather than false beliefs,
then you must have a reason of some kind (not necessarily an epistemic reason) to
form beliefs in the first place, rather than some other kind of propositional attitude
with different correctness conditions. One might insist that practical reasons just
are reasons to act successfully rather than unsuccessfully, not reasons to act at all, but
if so, there must in the first place be a reason of some kind to act. And if there is such a
reason, insisting on delimiting a subset of reasons connected with the success conditions of action and calling only those practical reasons is neither well-motivated nor
helpful.
14
One significant problem for this theory concerns the claim that actions aim at
the production of good consequences. Whilst it is plausible that any agent aims at executing her goals or intentions, clearly some agents aim at producing good consequences
but some instead aim at producing results they expect to be indifferent or perhaps even
66
alison hills
The appeal of action as production is perhaps the reason why utilitarianism (or more generally consequentialism) is perennially popular.15 If action aims at producing good results, then as an agent, the
fundamental way in which you should respond to value is by p
roducing
more of it (thereby acting successfully). So some maximizing version of
consequentialism seems to be inevitable, for example, utilitarianism
(everyone has reason to maximize happiness) or egoism (everyone has
reason to promote her own interests).16
There are some benefits to thinking about successful action as the
production of some good, and practical reasons as reasons to promote
the good. For example, it is not true that all practical reasons are instrumental reasons, so the problems that afflict theories of practical reason
that are purely instrumental do not apply. The question why be an
agent? can be answered. According to action as production, there are
certain standards that you must meet if your action is to be successful,
bad. Moreover, an agent can successfully produce good results without aiming at anything in particular. So the aim of producing good results (or even of producing
any results) cannot be attributed to the agent herself. This is a problem for the theory, since it is not obvious how else to understand the idea that actions aim at
production.
15
This might be a friendly amendment to Foots claim that the intuition behind
consequentialism is that it can never be right to prefer a worse state of affairs to a better. But whereas most people do not think at all in terms of states of affairs, it is more
plausible that they do think of actions as aiming to produce good results (P. Foot,
Utilitarianism and the Virtues, in her Moral Dilemmas [Oxford: Oxford University
Press, 2002]).
16
The picture is so seductive that it influences even Velleman, who initially seems to
be trying to give a rather different account, whereby the constitutive aim of action
is selfknowledge (Velleman, The Possibility of Practical Reason, pp. 2024; Practical
Reflection). Action, or perhaps autonomous action, is distinguished from mere behaviour, according to Velleman, in that an agent is in conscious control of his action
whereas one is not similarly self-aware in other forms of behaviour (e.g. in Freudian
slips). This sounds quite different from the idea of action as production. But in order to
derive a theory of practical reason from this constitutive aim, Velleman expands his
story so that you have most reason to do what it would make most sense for you to do.
For example, it might make most sense for you to cook dinner, since you are hungry
and reasonably enjoy cooking, so that is what you have most reason to do (Practical
Reflection, pp. 198ff.). In order to derive a full theory of practical reason from the minimal idea that action involves self-knowledge, he resorts to action as the production of
self-knowledge. And it is here that he goes wrong, because as Setiya points out, even if
self-knowledge is in some sense characteristic of autonomous action (or acting for reasons) it is not in the sense that action can produce more or less of it, and we have reason
to produce more. See K. Setiya, Explaining Action, Philosophical Review 112 (2003),
pp. 33993, especially pp. 37188.
namely your action must promote the good, and since you have reason
to meet those standards, because the good is to be promoted, you also
have reason to be an agent.
Making this response to the problem obviously means leaving aside
the possibility of an explanation of reasons for action in terms only
of the standards of action, without reference to value. Here instead the
reason to be an agent, that is, to perform actions that have certain
standards of success, is explained in terms of the reason to produce
certain values. It may seem now that characterizing action in terms of
production is making no contribution at all to the theory of practical reason. But this is not so. Claims about the nature of action are
no longer playing the kind of ambitious role that Korsgaard and
Velleman, for example, hoped that they would. They are not explaining
the whole of practical reason. But they still have a purpose. Instead of
deriving all practical reasons from the nature of action, we have an
interconnection between value, practical reason and the nature of
action. Values are the kind of thing that should be brought about as
much as possible. Action is essentially production; successful action
is the production of good states of affairs. So we have reason to act
successfully and to produce good states of affairs and these practical
reasons can be seen as connected both to the nature of value and to the
nature of action.
The difference between thinking of action as production and as execution is substantial and they tend to support different theories of practical reason. This is particularly noticeable if the agent does not intend
or aim at producing what is in fact a good state of affairs.
Suppose that it is good to maximize happiness but the agent instead
intends to make herself rich. She may successfully execute this intention without maximizing happiness. According to action as execution,
her action is successful. According to action as production, her action
is a failure, even though she did what she set out to do.
Now suppose that she does not execute this intention very well, and
she ends up poor, but at the same time making others happy. According
to the view that action is essentially production, her action is successful: she has produced a good state of affairs. According to the view that
action is essentially execution, her action has failed; she has not
achieved her aim. In these circumstances, it may be a matter of luck
that she produced a good outcome. But equally it may be predictable
that by forming apparently unrelated aims and intentions, she may
68
alison hills
achieve a good state of affairs (either by successfully or, as in this example unsuccessfully, pursuing the alternative ends).
It is for this reason that consequentialist theories of practical reason
may be self-effacing, that is, they may require an agent not to follow
that theory of practical reason. For example, a good utilitarian may aim
to look after herself and her friends and carry out that aim, rather than
aiming to maximize happiness, precisely because she has reason to
maximize happiness. This may be the best way for her to do so. This
aspect of consequentialist theories has been thought to be a problem by
some and defended by others.
But the real problem for this theory is much worse: there is something very peculiar about classing an action as a success when the agent
not merely has completely failed to do what she aimed to do, but may
not even have been carrying out an aim at all. She may have brought
about an event with good consequences, but this is not the same as acting successfully. In fact, she may be able to produce the good consequences without acting at all, that is, without carrying out any particular
goal or intention.
For example, suppose that tripping over the kerb I fall againstanother
pedestrian who was about to step into the road, not having seen the
lorry speeding towards him. My tripping has saved his life, bringing
about good consequences, and therefore is classified as successful
action. But obviously it is not: it is not an action at all.
The theory of action as production, as it stands, may classify as
examples of successful action kinds of behaviour or events that are not
actions.17 The combination of claims that successful action just is the
production of good results, and that practical reasons are reasons to
perform successful actions, is unacceptable, and we will have to reject
one of them.18 If we are to maintain the intuitive connection between
practical reason and successful action, we will have to find a different
way of relating action, value and practical reason.
17
The theory that successful action is the production of good results and the claim
that actions aim at producing therefore come into conflict, at least when the latter is
understood in terms of the agent herself aiming at bringing about some results. A similar criticism is made of these kinds of theory by T. Schapiro, Three Conceptions of
Action in Moral Theory, Nous 35 (2001), pp. 93117.
18
This is not an objection to maximizing consequentialist theories of practical reason, for it is obviously open to someone who wanted to maintain that practical reasons
are reasons to produce good outcomes to drop the claim that practical reasons are
reasons to perform successful actions. But this claim is intuitively plausible, and a theory that links practical reason, value and action is intuitively appealing.

4.Commitment
A common thought is that action can express value. This was taken
extremely literally by Wollaston, whose views are described by Schapiro
as follows:
The thoughts which Wollaston takes to be expressed in action are essentially propositional in nature. For example, by firing on an approaching army, a body of soldiers offers the salute of an enemy, a salute
whichdeclares that the two armies are indeed foes. If the actual relation
which holds between them is one of friendship, then the proposition
embodied in this act is false. Similarly, by using and disposing of an
object in his possession, a man declares it to be his own. If he does not
in fact stand to the object as an owner to his property, then the declaration is false Action, on this account, is a form of assertion, a mode of
expression in which we purport to represent things as they really are.
And this, Wollaston claims, is the sense in which actions can be or fail to
be in conformity with the nature of things, independently of their
consequences.19
Wollaston suggests that action is associated with an assertion, which

can be assessed for truth. Action is successful if the associated assertion
is true.
Hume famously is sceptical about this sort of theory on the basis that
there is no such thing as a true or false action. Schapiro thinks that
Humes worry is along the right lines: To say that an expressive action
is true or false is simply to say that something other than the action
itself, namely the belief it asserts, is true or false; it makes no sense to
say that the act of asserting that belief is itself either true or false.20 This
was supposed to be an account of successful action, but by focusing on
assertion it has missed out the component that is distinctive to action
and therefore fails.
But we should not dismiss expressivist views of action too quickly.
The problem is to find the right account of how action can express
value. We cannot appeal to assertion, but more promising is a kind of
propositional attitude that is distinctive to action but can express value:
intention.
I take intention to be a kind of commitment to do something, either
for its own sake or for the sake of some further end. When you intend
Schapiro, Three Conceptions, pp. 9697
Schapiro, Three Conceptions, p. 98.
19
20
70
alison hills
something, you commit to it in the following sense: you aim at it, try to
achieve it, choose what to do on the basis of their contribution to it,
and monitor your success in achieving it. When you intend to X for its
own sake, you aim at it and try to achieve it for its own sake.21
Intentions can express value in the following sense. If something
is valuable for its own sake, then it is worth pursuing for its own sake.
If something is worth pursuing for its own sake, then it is appropriate
to aim at it, to try to achieve it, to choose your actions on the basis of
their contribution to it, and to monitor your success at achieving it for
its own sake. So when you intend X as an end, you treat X in ways that
would be appropriate if X really were worth pursuing for its own sake.
Similarly, if you intend Y as a means to some other end Z, you treat Y
in ways that would be appropriate if it really were valuable as a means
to Z.
For example, if you intend to kill someone, your intention is inappropriate, because killing is not worth pursuing for its own sake. If you
intend to help Sarah because she is in need, your intention is appropriate. Intentions can also express non-moral value. It is appropriate for
you to commit yourself to having a hearty breakfast, but it is not appropriate to intend to eat a large lunch, because the former is good for you,
the latter is not.
Note that this is different from the traditional claim that action is
under the guise of the good, according to which, when an agent desires,
aims or intends something, she sees it as good in some respect. An
agent can intend to eat a large lunch without regarding eating a large
lunch as good or worth-while in any respect. But she will thereby treat
eating a large lunch in ways that would be appropriate only if it had
qualities that she herself does not think that it has.22
The simplest way of connecting action and commitment is to say
that actions aim at having appropriate commitments, and that an
21
This conception of intention as commitment is very similar to that defended by
M. Bratman, Intentions, Plans and Practical Reasoning (CSLI, 1999), except that he
does not discuss the connection between intention and value.
22
As a consequence, there is a problem in understanding in what sense actions aim
at appropriate commitment, since this is not something at which the agent herself
aims. The best way of understanding this claim is, I think, the following. Anyone who
can be considered an agent has intentions, and in appropriate circumstances, those
intentions are commitments that can be assessed as appropriate or inappropriate. Since
commitments can be evaluated in this way, and actions essentially involve commitments, it is reasonable to assess actions by this standard, even though agents themselves need not aim at having appropriate commitments.
action is successful if the intention on which the agent acts is appropriate. An intention is appropriate if the agent intends to X for its own
sake, only if it is valuable for its own sake; he treats Y as if it were valuable as a means only if it is valuable as a means.23 The action you have
most reason to do is the action whose commitment is most appropriate. We can call this action as commitment.
The relevant conception of practical reason, given this idea of
action as commitment, is that an agent has reason to perform some
action if the intention on which the agent acts is appropriate (that is,
appropriately responsive to value). As in the case of action as production, the goal here is not to explain fully practical reason in terms of the
nature of action, but rather to give an account of action, practical reason and value that fits together. The view of value that fits with action
as commitment is quite different from that underlying action as production, where value was thought of as that which should be brought
about. Instead, value is conceived of as that which should be responded
to in the right way, for example, by aiming at it, trying to bring it about
and so on. According to this view, what is important in terms of action
is not production of value but orientation with regard to value.24
Action as commitment might well be suited to a kind of nonconsequentialist moral theory, for what is important about action is
commitment (e.g. trying not to kill people) rather than production
(e.g. making sure as few people are killed as possible). For example, it
may support a version of the doctrine of double effect, for the distinction between what is intended (what one is committed to) and what is
not intended (whether it is foreseen or not) is very important on this
view of action.
The idea of action as commitment, and related theories of practical
reason and morality, are interestingly close to Kants conception of the
good will. According to Kant, a good will wills the right action for
the right reason (the relationship in Kant between practical reason
and value is too complicated and controversial to discuss here). But a
23
Intending to perform Y for the sake of X is an appropriate commitment only if X
is valuable and Y is a means to X, however. If X is not an appropriate end, the agents
overall commitment cannot be appropriate even if Y is a suitable means to the end she
in fact chose.
24
Correct orientation with regard to value requires at least that you intend to X for
its own sake only if it is valuable for its own sake. But it may also be necessary that you
formed appropriate intentions for the right reasons, that is, that you intend to X for its
own sake because it is valuable for its own sake.
72
alison hills
problem often raised for Kants good will is relevant here: surely it is
important not just what someone wills, but also what they actually do.
In our terms, we may have given an account of successful intention,
but we have missed out an aspect of action that is surely crucially
important: carrying out that intention. No one could call an action
successful however worthy the intention if the agent utterly failed to
execute it. Suppose that you form an appropriate intention: to work all
morning. But then you find yourself engrossed in a variety of entertaining but trivial distractions. This is not a paradigm of successful
action. Nor could it be said that you did what you had reason to do. You
had reason to work all morning, not just to form the intention to do so.
This suggests that commitment alone cannot be what is essential
to action. It must at least be combined with another feature. And the
obvious candidates are the two that have already been discussed: execution and production. This leaves us two possibilities: first, that successful action is an appropriate intention plus successful execution of
that intention;25 second, that successful action is the production of
good consequences together with an appropriate intention.
There is a problem with the latter theory as it stands. An agent may
have an appropriate intention and produce good consequences, but not
through successfully executing her intention. For example, you may
intend to help others and you may actually improve their welfare, but
precisely by failing to act on your intention to help and thus forcing
those others to fend for themselves and become self-reliant. This surely
should not count as successful action, however. So we need to amend
the second theory to: successful action is the production of good consequences through an appropriate intention. This rules out accidentally
contributing to good consequences whilst at the same time having a
good intention.
This brings the two accounts much closer together, for in both cases
successful execution of an appropriate intention is essential to successful action. But there may still be a difference between the two. The successful execution of an appropriate intention may not produce the best
consequences, for the long-term consequences may be much worse
25
As before, execution here must not just be that the intention is in fact satisfied, or
even that the intention is satisfied though some behaviour of the agent. The behaviour
must be caused in the right way by the agents aim in order for the agent to
execute her intention: the agent must control her behaviour in order to satisfy her
intention. It is very difficult to say what this control consists in, and unfortunately this
is a question I cannot pursue here.
than expected. This would count as successful action according to

action as commitment plus execution but not necessarily according
to action as production through commitment.
According to these conceptions of action, practical reasons are either
(a) reasons to have appropriate commitments and execute them successfully or (b) reasons to produce good consequences through appropriate commitments. Thus it follows that I did not after all do what
I had reason to do when I wasted time all morning when I intended to
work. I may have had an appropriate intention, but I did not execute
it successfully, nor did I produce good results through that intention.
My action was not successful, so I had no reason to do what I did;
instead I had reason to act successfully, that is, to carry out the intention (or to produce good results through that intention).
On the other hand, even if I had worked all morning and therefore
I had done what I intended, but there was something much more
important that I could have been doinglooking after my sick child,
for examplemy action would not have been successful and I would
not have done what I had most reason to do. Forming the intention to
look after my child and carrying it out (thereby producing good consequences) would have been a successful action. I might have had some
reason to work all morningsince working was valuable, that intention would not have been utterly inappropriate, as it would be if I had
intended to do something badbut working would not have been
what I had most reason to do. Looking after my child was more valuable, and so an intention to do so was more appropriate, and it is what
I had most reason to do.26
How should we decide between these two conceptions of action? We
can turn to commonsense judgments about what action is and what
counts as successful action. For example, if you aim to do something
good and you achieve your aim, but you nevertheless produce less
good than you could have done, does your action count as successful?
26
It does not follow from the fact that practical reasons are reasons to perform successful actions, where successful actions are those with appropriate commitments plus
execution, or production through commitment, that if the agent fails to act successfully, no practical reasons applied to her. Of course, she may have had practical reasons
to act, that is, to carry out her intention, or to produce good results through her intention, to which she failed to respond. The important point is that if she failed to act
successfully, she cannot have had practical reasons to do what she did (i.e. reasons to
fail to act). The practical reasons that applied to her, which she may have acknowledged
or ignored, were reasons precisely to carry out an appropriate intention or to produce
good results though an appropriate intention, rather than to fail to do either of these.
74
alison hills
If you are inclined to say no, you will be drawn to theories of practical
reason that emphasize reasons to produce good consequences; if you
would say yes, you will be supporting a theory of practical reason that
emphasizes reasons to respond appropriately to value.
Your view on value obviously may feed in and influence this decision
too. These two conceptions of action both support conceptions of value
as something to be produced and as something to be responded to in
other ways. How important is either of these two components? If you
think that it is fundamental to value that it is to be brought about as
much as possible, you should favour the conception of action that
emphasizes production. If, by contrast, you think that responding to
value in ways other than production is important, you should favour
action as commitment plus execution.
I cannot settle this question here, so I will leave these two accounts
of action still standing. Practical reason for each of these is a complex
matter, including reasons to respond to value appropriately, instrumental reasons (to carry out intentions successfully) and perhaps reasons to produce good consequences too. But we have been led to this
more complicated account of practical reason by an attempt to give an
adequate account of action (and of value) and simpler theories proved
to miss out one or more of the essential features of successful action.
5. Basic Action and Mature Action
There is still a serious question, however, over whether our final
accounts of action are too complicated. Can we consider small children
and animals as forming appropriate intentions and carrying them out?
If not, are we to conclude that they do not act after all?
This raises a significant question for the whole approach of this
paper. Can we in fact give one account of action, identifying a single
standard of success? Perhaps different kinds of action have different
standards. For example, perhaps animals and small children perform
basic actions which have as their standards of success: the agents
successfully executing her aim or intention. Perhaps we have been setting out a different kind of action, which we might call mature action,
whose standard involves an assessment of what the agent is doing
(through an assessment of her intention, and perhaps of what she produces) as well as whether she successfully achieved her aim.
I think that there is something to this complaint, but that there are
important connections between the different conceptions of action.
I suggest that basic action is a genuine kind of action. This allows us to

describe small children and animals as agents, at least in a basic sense,
provided that they can execute their own aims in the requisite manner.
These creatures are not (yet) capable of responding to value appropriately, so though we ourselves can judge that they have good or bad
aims, they themselves cannot, or at least, they cannot in the same way.
Some basic agents may not form intentions to act at all, but simply
carry out their aims without committing to them. But even if they do
form commitments to act, their intentions cannot properly be regarded
as expressing value, because they are not capable of understanding
value or responding to it appropriately: they cannot orient themselves
with regard to value. It is unreasonable to evaluate them by standards
that they themselves cannot understand, and with which they cannot
comply.
But some basic agents become capable of responding to value and
practical reasonthey become sufficiently mature to be treated as
morally responsible, for exampleand they therefore have reason to
respond to value. They can be expected not just to carry out their intentions, but also to have appropriate intentions in the first place (and perhaps to aim at producing good results).
The standard of success for evaluating their mature actions are obviously not completely independent from the old standards of basic
action, though they are more demanding. In both cases, the execution
of the agents aim is important. But whereas this is all that is needed for
successful basic action, a mature action can count as unsuccessful if
you fail to have appropriate intentions, even if you successfully achieve
your ends. This new component for the assessment of action is not
something artificially bolted on to the old standard, unconnected to
the nature of action itself, however. Successful action, in mature agents,
involves the execution of an intention. But intention is a commitment,
and in mature agents who can orient themselves with regard to value,
it is entirely appropriate to regard their commitments as expressing
value. So mature action involves expressing value, and it is entirely reasonable to evaluate mature actions as successful or not on that basis.
Those who think that there is something illegitimate about such
an evaluation, and complain about associated theories of practical reason, may be confusedly thinking that the basic conception of action is
the only one that there is. Or they may think that it is appropriate
to judge an action unsuccessful if and only if it fails to meet the aims
that the agent herself sets. But it may be appropriate to evaluate action
76
alison hills
with regard to standards that the agent was capable of responding

to, even if in fact she did not, because they are standards she ought to
have acknowledged. On this view, any successful action will involve
the agents successfully executing her aim. But an agent can satisfy her
aim without her action qualifying as successful, because her aim was
wrong.27
6.Conclusion
Many of the debates about the nature of practical reason discuss the
nature of value and the relationship between value and practical reason. I have suggested that in these discussions there may be a missing
third term: the nature of action. Since practical reasons are reasons for
action, it is natural to think that different theories of action will support
different theories of practical reason, and vice versa. Therefore we can
find a new way of looking at old disputes about practical reason, and
hopefully new insights too, by thinking about the related conception
of action.
I have suggested that some of the reasons that (a) practical reason as
instrumental reason and (b) consequentialist theories of practical reason persist is that the conceptions of action that are most obviously
related to them are initially very appealing. But in both cases, the conception of action and therefore of practical reason is either flawed or
limited. A better conception of action is of action as involving commitment. This supports a more complex theory of practical reason, which
may well be non-consequentialist.
Previous discussions of practical reason and the standard of success
for action have tended to be enormously ambitious, expecting to isolate a single standard of success for action and derive an entire substantive theory of reasons for action. It is hardly surprising that we cannot
do so much from so little. But we can still find illuminating the connections between value, practical reason and action, and reflecting on each
may affect the way we need to conceive of the others.
27
Of course, it is hardly going to worry an immoral person that, according to us, her
actions are not successful, even though she gets what she wants. That was not the purpose of the argument, however, which should not therefore be criticized for failing to
answer this kind of sceptic, a task which in any case may be impossible: there may
simply be no way to convince an immoral agent by philosophical argument that she
should be a better person.
Normativity and Practical Judgement

Onora ONeill
Discussions of normativity and practical reason often focus on reasons
for adopting, or for rejecting, specific norms. This focus is useful for
addressing questions about the nature and justification of norms and
their division into various modal types (requirements, prohibitions,
permissions, etc). In particular, it is the right focus for raising questions
about the nature and justification of ethical norms and their division
into various modal and other types (ethical obligations, prohibitions,
permissions, recommendations, etc). Yet a focus on reasons for adopting specific norms, including adopting specific ethical norms, does not
seem to be enough to guide action. Norms are always indeterminate;
acts are always particular and so determinate. A given norm can always
be satisfied by a plurality of possible acts. So it seems that there will
always be a gap between norm and act, and that while practical reasoning may be able to justify specific norms, it will not by doing so show
which particular acts are required. Moves from a specific norm to one
or another particular enactment of that norm are generally seen not as
instances of reasoning, but rather as applications of that norm. Yet the
use of the term application is more obscure than it seems: action that
changes the world to fit a norm is quite different from cognition that
fits or applies a concept to an aspect of the way the world is.
This gap between indeterminate norms and their particular enactments is not confined to the domain of morality. It arises for norms
that bear on many different aspects of action, including norms thought
of as social or legal, grammatical or technical, epistemic or cultural, as
well as ethical norms. It arises for norms that prescribe with varying
modalities, including those that we see as expressing requirements,
obligations, and permissions, and those that we see as formulating recommendations, advice or warnings. In all cases, reasons for adopting a
specific norm do not yield reasons for selecting one rather than another
act that instantiates that norm.
Can we then think of practical reasoning as justifying acts as well as
norms? In order not to prejudge this question I shall rely on an ample
I hope at least non-excludingview of practical reasoning, formulated
78
onora oneill
by Phillipa Foot, who suggested that practical reasoning is what you

have if you are good at making the kinds of practical choices that arise
for human beings, given the material circumstances of their existence.1
This characterization strikes me as sufficiently ample to avoid begging
questions. Being good at making choices is not just a matter of being
good at choosing norms. Some people who adopt, indeed seek to live
by, admirable norms make a poor fist of enacting them. These are the
people whom we think of as having poor judgement. Now it may be
that practical reason can offer no guidance to practical judgement, and
that reason can reach no further than the justification of norms. But we
should not reach this conclusion by adopting too limited a view of
practical reason. A fortiori, we should not restrict our account of practical reasoning to those norms that incorporate quantifiable requirements, such as those formalized in models of rational choice. Norms
may incorporate quantifiable requirements or considerations, but need
not do so. Whether or not they do so, it is still an open question whether
practical reasoning can tell us anything about the move from adopting
justifiable norms to enacting those norms.
That being said, we need to say something preliminary about the
way in which a focus on norms can help us to understand how practical
reasoning bears on norms, while leaving it open how it is to bear on
acts that instantiate those norms. There would be no point in speaking
of practical reasoning unless we can see why it offers a basis for saying
that some norm is or is not a reason for doing some action of type
A. Norms pick out types of action that fall under specific act descriptions: it is their propositional structure and content that make norms
apt for reasoning. However, by the same token the fact that a norm
picks out a type of action means that it can offer no more than a reason
for doing some action of a specified type, and not a reason for doing a
particular act of that type.
Since practical reason seeks to guide action it must be future oriented, rather than directed at particular acts that have been done and
can be individuated. We can, of course, formulate definite descriptions
of future acts. For example, I can use definite descriptions to specify the
types of acts I intend to do at some future time: the walk that I intend
to take after lunch, the drink that I hope to enjoy at a certain pub,
1
Phillipa Foot, Morality as a System of Hypothetical Imperatives, in her Virtues
and Vices (Oxford: Basil Blackwell, 1978), pp. 15773.
normativity and practical judgement79
the meeting with friends at the pub that I shall arrange. But this is not
to say anything about the particular acts that I will do, beyond stating
what types they will be with some generality. The definite descriptions
do no more than specify the types of act that are to be done. Although
they individuate the agent by whom and the occasion on which they
are to be done, they do not distinguish between the innumerable ways
in which the specification of action might be met. It may seem, then,
that the gap between norms (for which reasons may be given) and
enactments of those norms (for which no complete reasons can be
given) marks a limit of practical reasoning.
Taking a Relaxed View of Norms
So the optimistic thought that norms are apt for practical reason
because they combine propositional structure and content (making
them apt for reasoning) with a capacity to guide action (making them
practical) is too quick. This might suggest some problem with reliance
on the idea of a norm. The term has a chequered history, and may seem
in some ways to be too narrow and in others too broad to provide a
useful focus for practical reasoning. However, I do not think that the
difficulties lie specifically in the use of the term norm, rather than of
other terms for practical propositions that combine pro-positional
structure and content with claims to be action guiding (such as principle, rule, standard, law). We should not and we need not be the p
risoners
of past uses of the term norm (particularly now that the term normative
is used in a much broader way). As I see it, there are reasons for being
rather relaxed in our understanding of both terms, and rather self conscious about their relations to other terms that serve cognate purposes.
Indeed the only feature of norms that it seems to me essential to retain
is the thought that their use is to guide the way action changes the
world rather than to describe the way the world is. A focus on direction
of fit suggests four issues that might arise if we relied on traditional,
more restrictive understandings of norms. I shall discuss these in turn
and suggest that there is nothing particularly problematic about centring discussion of practical reasoning on norms.
1. Norms and Other Directives
One set of considerations that might lead us to wonder whether norms
offer too narrow a focus for practical reasoning is that there are other
80
onora oneill
practical uses of propositions, which we would not usually view as

norms, which also incorporate act descriptions to which the world is to
be fitted and which are taken to guide action. For example, there are
intentions and instructions, and other practical uses of propositions
that are sometimes termed directives, which we might be reluctant to
think of as norms, although they function like norms in that they are
intended to guide action rather than to fit the way things are. In each
case we think of ways in which action can be shaped to fit one proposition rather than another, and contrast this with uses of non-normative
principles that aim to fit a proposition to the way the world is or will
beincluding the way certain actions are or will probably be.
This wider range of practical uses of propositions is the focus of
Anscombes classic work Intention, and the action guiding aspect of the
practical uses of propositions she discusses are vividly encapsulated in
her example of the man whose wife complains of his shopping performance saying Look, it says butter and you have bought margarine,
who would get things very wrong if he replied What a mistake! We
must put that rightand altered the list to read margarine.2 The shopping list is intended to be normative for the shoppers performance,
rather than predictive of it: the mismatch calls for different shopping,
not different listing, i.e. for changes in action rather than for changes in
theory. In my view the fact that we would not traditionally speak of an
instruction or intention to buy butter as a norm is not a deep problem.
Instructions and intentions function normatively, and there is no reason to be squeamish about thinking of them as norms (in a moment
I shall allude to some historical reasons why we may [mistakenly] think
that we should be squeamish).
The same relaxed view of norms will allow us to think of other types
of directive, such as intentions, vows and promises, orders and requests
as normative, although they might not traditionally have been called
norms. Norms in a traditional, sociological sense of the term are perhaps best thought of as a subclass of directive, which not merely guide
action but are thought of as having an authority that is not traceable
either to individual decisions or commitments (as are intentions, vows
and promises), or to second and third person impositions (as are
orders, instructions and requests). To put matters vaguely, norms, in
the traditional sociological usage of the term, are impersonal directives,
2
G.E.M. Anscombe, Intention (Oxford: Basil Blackwell, 1957), p. 57.
often seen as backed by social, legal or other authority. Norms in this

narrower sense may be ethical, grammatical, social or legal and are
often domain specific (norms of international behaviour, norms of economic rationality, etc.). However, we do no harm in classifying a fuller
range of uses of practical propositions as norms, provided that we realize that we have broadened the traditional understanding. What is
common to all these, and many other, practical uses of propositions is
that they set out standards to which the world should be adjusted,
rather than aspects of the way the world is.
2. Norms and Modality
A second rather different caveat is that some propositions that function
normatively are not accurately characterized as directives, in that they
do not specify requirements for the aspects of action for which they are
normative. Propositions can have other practical uses. They can be
used to formulate advice or warnings, recommendations or guidance,
or Kantian counsels of prudence, none of which strictly speaking
require or direct action of a specific type. But once again it seems to me
that we can take this point lightly and extend the term norm to cover
this wider range of practical uses of propositions: like requirements
and prohibitions they function normatively to specify ways in which
the world should be changed. Examples of norms that do not require
but rather recommend or warn against specific types of action are very
common in self-help manuals, good practice guidance, proverbial wisdom and traditions of virtue.
It is, I think, useful to adopt a capacious use of the term norm, which
covers uses of practical propositions of all sorts, regardless of their
source (first, second or third personal, or impersonal), or the modality
of their prescription (requirement or prohibition, warning or advice).
Taking this elastic view of norms is not only acceptable but advantageous, because it reconnects norms with contemporary understandings of normativity. However, if we accept this broadened understanding
of norms two further points follow.
3. Socially Embedded Norms
A long tradition in sociology, often reflected in ethical writing, assumes
that the crucial feature and advantage of focusing on norms in discussions of practical reasoning is that (unlike abstract principles and
rules) they are socially embedded or entrenched. There is, of course, a
82
onora oneill
lot of history here. Some strands in that history urge us to think that
norms provide a uniquely suitable focus for practical reasoning precisely because they are socially embedded, because they are the norms
of this or that group at this or that time, and that this is what makes
them important, practical and motivating. Others have seen the understanding of norms as socially embedded as a great deficiency of a sociological understanding of norm-based practical reasoning. If norms
are seen as practical propositions that are entrenched in the lives of
some group, the objection runs, they offer a highly problematic focus
for practical reasoning. Some embedded practical propositions may
provide poor reasons for actionand the fact that they are embedded
adds only the questionable weight of an argument from authority.
(We need only think of appeals to entrenched norms of revenge, honour killing or female genital mutilation.) By contrast, abstract principles of action that at a given time are not well embedded may sometimes
offer significant reasons for action.
However, letting go of the traditional, narrower sociological understanding of norms as embedded is not problematic. In fact, I suspect
that current usage of the term has long since distanced norms and normativity in discussions of ethics and practical reason from the older
sociological conception of norms as socially embedded, although there
are pockets of lingering confusion.
4. Norms as Motivating
And there is a fourth point that follows from accepting a relaxed view
of norms. An explicit rejection of the older sociological conception of
norms as socially embedded undermines the psychological corollaries
of such views, and requires us to set aside the assumption that norms
are invariably motivating. There is no doubt a complex story to be told
about the ways in which norms come to be accepted, understood,
rejected and modified, and about the emergence of particular norms in
the lives of individuals and societies. There is also no doubt a story to
be told about what Sabina Lovibond has called the animal precursors
of practical reasoning.3 However, once we accept a broader understanding of norms as practical propositions that may or may not be
embedded in the life of this or that group, we will also have to accept
3
Sabina Lovibond, Practical Reason and its Animal Precursors, European Journal
of Philosophy 14.2 (2006), pp. 26273.
that they have no automatic psychological or motivational role. This

may be an advantage. It may be thatas with theoretical cognition
the best account of the development of the emergence and embedding
of specific norms in agents or in groups would not be given in terms
that stress those specific norms. The embedding of norms may be better explained in terms of the entrenching of lower level routines, habits,
inhibitions and tendencies. The most psychologically effective way of
living up to some norm may not make it an explicit focus of practical
reasoning. For example, if I seek to live up to a norm of law-abidingness,
I may do best if I seek to be conformist and inconspicuous, rather than
to check my compliance with the statute book with high frequency;
more generally morally responsible ways of living may not depend
heavily on explicitly moral reasoning or motivation, any more than
cognitively responsible ways of living need depend on explicit efforts to
conform to epistemic norms.4 Explicit thought about norms is needed
when we address questions of justification, but in acting on norms and
in thinking about motivation we may often do better to rely on shortcuts, habits and heuristics.
Indeterminate Norms, Determinate Acts
With these preliminaries on one side, we return with greater clarity
to consider how normsof any sortare supposed to guide action.
This is not the same as asking which norms can be justified, for which
contextsuncontroversially a central task of practical reasoning. Yet
the question of whether and how norms (justified or not) can guide
action seems to me a more fundamental question than those about the
justification of norms, or about the relation of justification to motivation. If we cannot understand how norms can shape action, an account
of their justification may be of little practical use.
4
This is not a new thought. Whitehead expressed it as follows: It is a profoundly
erroneous truism, repeated by all copy-books and by eminent people when they are
making speeches, that we should cultivate the habit of thinking of what we are doing.
The precise opposite is the case. Civilization advances by extending the number of
important operations which we can perform without thinking about them. Operations
of thought are like cavalry charges in a battlethey are strictly limited in number, they
require fresh horses, and must only be made at decisive moments (A.N. Whitehead,
An Introduction to Mathematics [New York: Henry Holt and Co., 1939], p. 61). For a
recent summary of some evidence see Malcolm Gladwell, Blink: The Power of Thinking
without Thinking (London: Penguin Books, 2006).
84
onora oneill
On the surface norms seem defective guides to action. Norms, taken

in the relaxed sense I have suggested, are abstract entities with propositional structure and content to which action is to conform. A norm
formulates some standard or requirement, some recommendation
or permission; action is then supposed to be guided or shaped by
that standard, requirement, recommendation or permission. However,
while norms are indeterminate, the acts done in living up to themor
in failing to live up to themhave to be determinate in all respects.
How, then, can norms guide action? Isnt the temptation to think of
practical matters, including morality, as a matter of being guided by
certain norms an illusion, because norms are never enough to shape
action?
There is some temptation to think that this issue could be avoided if
only norms were algorithms that provide wholly definite instructions
for each context and can specify exhaustively what must be done in living up to them. This thought seems to me highly implausibleand
particularly implausible for the case of ethically important norms:
An algorithm is a finite procedure, written in a fixed symbolic vocabulary, governed by precise instructions, moving in discrete stepswhose
execution requires no insight, cleverness, intuition, intelligence or perspicuity, and that sooner or later comes to an end.5
Strictly speaking, algorithms are therefore possible only within formal

systems, where contexts and moves can be exhaustively specified.
A system of arithmetic will provide an algorithm for multiplication; a
statement of the rules for playing noughts and crosses will allow us to
work out (although it wont state) an algorithm for avoiding defeat even
when the other player has the first move. But when we multiply or play
noughts and crosses in real life, we need more than algorithms: multiplication can be done in ones head or aloud, or in writing, and the
order of the multiplicands can be varied. Noughts and crosses can be
played on paper, on a blackboard or on the sand, using varied marks,
and various implements to make them and so on. There are no true
algorithms for action.
This may seem surprising. Does not Benthamite Utilitarianismosten
sibly aspire to provide an algorithm for morality, supposedly allowing
us to calculate which act is optimific? The formal structure of utilitarian
5
David Berlinski, The Advent of the Algorithm: The Idea that Rules the World
(New York: Harcourt Inc., 2000), p. xviii.
reasoning may look algorithmic. The instructions for utilitarian calculation tell us to list all options; to reckon the probable outcomes of each;
to calculate and sum the expected utilities of these outcomes for all
parties; and finally to maximize. In practice, as we all know, utilitarian
calculation cannot even approximate the underlying algorithm: we can
specify only selected options; we are often unsure about their probable
outcomes; our calculation of expected utility for anyone (let alone
everyone) is pretty gestural. Only the maximizing looks even close to
algorithmichowever, on reflection we see that it too is not wholly tied
down and could be done in various ways.
On second and more cheerful thoughts, it perhaps is not important
if norms are not algorithmic. For they can at least formulate constraintson or advice for action, and perhaps all that matters is that we
choose some act that meets the constraint or recommendation set by a
norm that we are seeking to respect. Yet how are we meant to do this?
A standard answer is that it is a matter of judgement. But this is not
reassuring. Invoking judgement without explaining how it is to work
seems to leave us no clearer about what we need to add to norms if they
are to offer practical guidance. Yet perhaps practical reasoning can take
us no further.
Writing of theoretical rather than practical judgement, Kant commented that reason could not take us all the waya point that has been
endorsed by many later writers:
[I]f it [the understanding] wanted to show generally how one ought to
subsume under these rules, i.e., distinguish whether something stands
under them or not, this could not happen except once again through a
rule. But just because this is a rule, it would demand another instruction
for the power of judgement, and so it becomes clear that although the
understanding is certainly capable of being instructed and equipped
through rules, the power of judgement is a special talent that cannot be
taught but only practiced.6
Does an analogous problem arise for the case of practical judgement?

Or can we say something about how practical judgement is to gobeyond
norms and select one rather than another enactment of a norm?
6
Immanuel Kant, Critique of Pure Reason (trans. and ed. Paul Guyer and Allen W.
Wood; Cambridge: Cambridge University Press, 1999), A133/B172. Here Kant is
writing specifically about subsumptive or determinant judgement, the most standard type of theoretical judgement. See below for some comments on reflective
judgement.
86
onora oneill
Picking and Choosing
One minimalist answer to this problem would be that practical judging

is no more than a matter of lighting on some act that fits the normunder
consideration. For example, if we are aiming to live up to a norm of
promise-keeping we need only find some act that meets the terms of
the promise previously made. Any such act will do, and there are no
significant differences between the various acts by which a given promise might be kept. Or if we seek to keep to a norm of parsimony, we
need only find some way of living that is (adequately) parsimonious.
And so on.
On such views, practical judging is simply a matter of picking some
act that satisfies the relevant norm, and it does not matter which of
many differing available congruent acts is chosen. The original of such
cases is illustrated by the predicament of Buridans ass, who could not
find any reason for preferring one over another bundle of equidistant
hay. Here, some writers want to say, choosing gives way to mere picking. As my hand hovers over an array of equally accessible tubs of margarine, do I really choose one over another? Or is this a case of mere
picking?7 This seems an apt enough characterization of the particular
case. There is no reason to think that the choice of one rather than
another tub of margarine from a number that are equally accessible,
equal in price, and indistinguishable in quality and appearance is more
than a matter of picking. It is not a matter of choice because there is
(ex hypothesi) no basis for choice and so no better reason for choosing
one rather than another such tub.
But this example of mere picking is, I think, a limit case which we
cannot take as a model for thinking about all cases in which a variety of
discernibly different acts would satisfy the constraints set by some
norm. Most choosing is not a matter of mere picking. Indeed, some
would argue that the Buridan case, and similar cases, are degenerate
examples of choosing, precisely for this reason. As Leibnitz thought,
where things are absolutely different there can be no choice, because
choice must have some reason or principle.8 Even if we regard mere
7
See Edna Ullmann-Margalit and Sydney Morgenbesser, Picking and Choosing,
Social Research (1977), pp. 75767, for the distinction. They illustrate picking with the
case of selecting among tins of soup rather than tubs of margarine. I have stuck with
margarine, in order to emphasize the links to Anscombes argument.
8
Leibnitz expressed the thought repeatedly in varying ways. For a typical passage
see G.W. Leibnitz, Discourse on Metaphysics (trans. Peter G. Lucas and Leslie Grint;
Manchester: University of Manchester Press, 1953), para. 3.
picking as a limit case of choice, it at least seems plausible that reasoned

choice must refer to some norm. Moreover, there is often reason to
think that the various ways of meeting the claims of a norm are not
equivalent, and that some are better than others. Here it seems that we
are not dealing with mere picking, but rather are deploying some form
of reasoned choice. If so, practical judgement must generally be more
than mere picking, and there may be grounds for thinking that in a
given situation with given norms, one judgement may be better or
worse than others, and more generally that the judgements made
by some persons might usually be better or worse than those made by
others. This thought returns us to the problem of understanding how
practical judgement works.
Practical Judgement and Multiple Norms
Is practical judgement an aspect of practical reasoning, or is it only a
pompous term for acts of picking, by which we instantiate an indeterminate norm with any determinate act that fits the norm? If so, what is
it that we admire in acts and persons that we think of as exhibiting
good judgement? What makes one way of instantiating a norm an exercise of good judgement and another an exercise of poor judgement?
Much of the literature that I have read on ethical judgement offers
astonishingly little help here. Often this is because work that purports
to be relevant to ethics and practical judgement is in fact about (one
type of) theoretical rather than about practical uses of judgement, and
in particular about the distinctive problems that arise for theoretical
judgement when it is unclear which concepts or standards should be
applied. Kant divided theoretical judgement into determinant and
reflective judgement, on the basis that:
If the universal (the rule, principle, or law) is given, then judgment which
subsumes the particular under it is determinate If, however, only the
particular is given and the universal has to be found for it, then the judgment is simply reflective.9
In both cases theoretical judgement presupposes that a p

articularaspect
of the world is there to be judged. Where the judgement is determinant
the task is to see whether a certain concept or description applies;
9
Immanuel Kant, Critique of Judgement (trans. J.C. Meredith; Oxford: Oxford
University Press, 1952), vol. 5, p. 179; original emphasis.
88
onora oneill
where it is reflective the task is to see which of many possible concepts

or descriptions is appropriate. None of this is relevant to practical
judgement, where the task is not to apply a concept or description to an
act (which does not yet exist) but to enact some norm or principle.
A large amount of writing on ethics (much of it Wittgensteinian in
flavour) assumes that practical judgement is form of reflective judgement, and so that it must address some version of the problem of relevant descriptions. For example, Peter Winch, John McDowell, David
Wiggins, and at times Bernard Williams, depict judgement as the crux
of the moral life, yet focus not on practical judgement but on judgement of the context or situation in which action is undertaken. They
often describe ethical judgement as a matter of appreciating or appraising or attending to what is salient about situations and cases of ethical
significance. A typical formulation is given by McDowell, when he
characterizes judgement or deliberation as a capacity to read the details
of situations or a capacity to read the details of situations in the light
of a way of valuing actions or a capacity to read predicaments correctly.10 Yet the analogy between practical judgement and reading texts
or appreciating situations is unconvincing. When we act we may as a
preliminary matter have to decide how to view the situation in which
we already find ourselves, and in which we seek to act: here reflective
judgement may indeed be needed. But even when reflective judging is
completed, and we have determined how to view the situation, we will
still need to decide what to do: and that is where practical judgement
does its work. A focus on reflective judging will not reveal whether or
how practical judging works.
If we think about action that conforms to a single norm there seems
to be little that we can say about practical judgement. Any act that
meets the relevant normthat satisfies the standard or constraint it
setsseems to be as good as any other, and there is no room for distinguishing better from worse judgement. In this case we are indeed
reduced to picking one of many possible acts that enact a practical
principle. However, the thought of seeking to meet the constraints of a
10
See John McDowell, Deliberation and Moral Development, in S. Engstrom and J.
Whiting (eds.), Aristotle, Kant and the Stoics (Cambridge: Cambridge University Press,
1996), pp. 1935; the passages cited are on pp. 23 and 26. This spectators view of moral
judgement is well entrenched: see also David Wiggins, Deliberation and Practical
Reason, in his Needs, Values, Truth: Essays in the Philosophy of Value (Oxford: Blackwell,
1987), pp. 21537; Bernard Williams, Persons, Character and Morality, in his Moral
Luck (Cambridge: Cambridge University Press, 1981), pp. 119.
single norm is highly artificial. We constantly need to act in ways that

meet multiple constraints and standards. In buying margarine I do not
merely need to fit the purchase to the shopping list, but to do so without breaching a large range of other norms that are also relevant to
shopping. I will probably take care to meet norms of not stealing, not
assaulting the shopkeeper, and not offering counterfeit coins, as well as
norms of domestic life, such as taking the purchased margarine home,
rather than (for example) feeding it to the pigeons or taking it back to
the shelf where it was displayed.
Action on multiple norms can be challenging. If I have to break
some bad news in a way that is honest, does not undermine the confidence of the person hearing it, and yet is not so shrouded in euphemism that the message does not get across, then the choice of words,
tone and way of communicating may begin to present a real challenge.
If I have to bring in a building project on time and on budget and to the
standards specified in the project documents, a lot will be demanded.
Practical judgement, I suggest, is not something different from acting
on norms: rather it comes into play where and because numerous norms
have to be simultaneously taken seriously and observed. Hence practical
judgement can indeed be reasoned, because it is norm guided, and
norms are apt for reasoning.
If this account of judgement is correct, it follows that in the end there
is always a point at which mere picking has to take place. When all
steps have been taken to conform to the full range of normsethical
and otherthat an agent sees as relevant, there will still be a range of
possible acts that fit all these requirements. Here reason giving has to
stop, and here it is not needed. Picking comes into the picture when the
differences between the possibilities from which the agent selects are
indeed a matter of indifference. It is typically below the level of reason
giving, and below the level at which any norms are relevant if I pick one
rather than the other of two coins of equal value in giving change, or if
I place my tub of margarine in one rather another equally suitable position in the shopping basket. Typically many of the aspects of action that
fall below the level of attention or of intention are picked rather than
chosen for reasons that are formulable in norms.
In offering this picture of practical judgement as reasoned, I have
said nothing about the justificatory arguments that might be given in
favour of some norms or against others. I have made no assumptions about the quality of arguments that we may be able to offer in
favour of specific norms that are important in specific aspects of life, or
90
onora oneill
about the possibility of offering arguments in favour of wholly general,

unrestricted norms of ethical importance. My question has been only
whether if we had such arguments we would still find that norms were
impotent to guide action, and I have argued that we would not provided we had reasons to think that a plurality of norms that are apt for
reasoning make their demands, but that we could not expect those
norms to guide us all the way down to one rather than another particular act.
Some Conclusions
Practical judgment, I have argued, is a matter of finding acts (or policies) that meet the constraints of a plurality of norms that specify
requirements and recommendations of a variety of types. However, the
way that we deploy a plurality of norms is not best thought of as a matter of balancing one against another. There is no metric for balancing
or trading off different types of norms, and there are no reasons for
thinking that high success in living up to one norm will generally compensate for failure to respect another. Great success in keeping a business afloat will not exactly compensate for handling stolen goods, or
failing to supply goods that are fit for purpose; great success in surgery
will not compensate for performing the wrong operation on a patient
(Ive done a very good tonsillectomy, although I didnt take out your
grumbling appendix). Practical judgement is an aspect of practical
reasoning because it aims to integrate rather than to prioritize or trade
off a plurality of norms. This task can be done better or less well, and
there are good reasons for thinking of some people as having good
and others poorer judgement. But even those with the most intelligent
and careful capacities for practical judgement will in the end have to
pick among possible acts between which there is, as we say, nothing to
choose. Provided picking is used only for this unavoidable task and not
prematurely where respect for a range of norms has to be integrated,
there will be no deficiency in practical judgement.
It follows that practical judgement will be at its most demanding
when agents seek to respect multiple norms whose requirements are in
tension, or even contingently11 incompatible. Two strategies may be
11
If two norms are intrinsically incompatible they cannot both offer reasons for
action. The defect then lies in claims that both are justified, and is not resolvable by
practical judgement about their enactment.
relevant here. The first is a matter of forward planning and avoidance.

We know a lot about the circumstances and actions that are likely to
create conflicts between the requirements of different norms, and can
try to avoid those situations. Those who make excessive or conflicting
undertakings (bigamists, fraudsters) will not be able to honour all their
commitments. Those who impose excessive or incompatible demands
on others are likely to face them with impossible demands, which no
exercise of practical judgement can integrate (the worst excesses of the
target culture). By contrast, foresight, care and good institutional structures can do a fair amount to avert such problems, by forestalling,
reducing and averting contingent conflicts between principles, so easing the tasks of practical judgement.12
But individual foresight and social reform have their limits. Often
there is no way of acting that satisfies all the norms than an agent would
wish to respect, and nothing that could have been foreseen or done to
avert the potential for conflict. The most extreme examples are dirty
hands problems, where institutions, practices and prior action make it
hard or impossible for those who have to act to meet all the norms to
which they are committed. Even when nothing is so deeply awry, we
may often find that the background of institutions and practices, of
habits and customs, of virtues and failings, of skills and incompetence,
of capabilities and vulnerabilities which shape action also hinder
attempts to live up to multiple principles. In making practical judgements it is pointless, indeed misleading, to assume these realities away.
It may be true that had institutions and practices been better, or had
agents made better decisions in the past or been more competent, less
conflict would have arisen. But in the world as it is, agents may be unable to avoid a degree of failure, including moral failure, because no
amount of thoughtful practical judgement enables them to integrate
and live up to all the norms to which they are committed in the situations that actually arise.
Where realities force hard choices it may simply be impossible
for agents to meet all the norms that they seek to respect. The most
that they can then do is to recognize the claims of unmet, contingently
unmeetable requirements and recommendations. But the fact that a
norm proves contingently unmeetable in some situations will not
12
See Ruth Barcan Marcus, Moral Dilemmas and Consistency, Journal of Philosophy
77.3 (1980), pp. 12136, here p. 121: as rational agents with some control of our lives
and institutions, we ought to conduct our lives and arrange our institutions so as to
minimize predicaments of moral conflict.
92
onora oneill
wholly cancel its claims. Unmeetable norms may leave remaindersthat

call for attitudinal responses such as expressions of regret or remorse,
or for more active responses such as apologies, commitments to reform,
to compensation, restitution and other forms of making good.13 The
importance of active foresight and institution building, and of an active
approach to dealing with remainders in the wake of practical conflict
between norms and failure (moral and other) does not show that norms
are redundant or useless. It shows that living up to them can be hard
and demanding. Taking norms seriously is a matter of working towards
practical judgement, expressed in action that seeks to enact requirements and standards and to make good where failure has not been
avoided.
13
See Barbara Herman, Performance and Obligation, in her The Practice of Moral
Judgment (Cambridge, MA: Harvard University Press, 1993), pp. 15983; and Onora
ONeill, Instituting Principles: Between Duty and Action, in Mark Timmons (ed.),
Kants Meta-physics of Morals: Interpretative Essays (Oxford: Oxford University Press,
2002), pp. 33147.
part II
Particularism
Ethics Without Reasons?*

Roger Crisp
Jonathan Dancys Ethics Without Principles is a deeply suggestive and
insightful discussion of some of the most fundamental issues in philosophical ethics.1 I cannot say I agree with Dancy even on the broad
lines of the positions that emerge in his book; but thinking about his
arguments has made me fully aware of their power, and helped me to
make a good deal of progress in clarifying my own views.
What is Dancys book really about? It is of course a defence of a particularist conception of morality. But what is that? On the first page we
are told:
A particularist conception [of ethics] is one which sees little if any role
for moral principles. Particularists think that moral judgement can get
along perfectly well without any appeal to principles, indeed that there is
no essential link between being a full moral agent and having principles.
This makes it sound as if Dancys concern is to answer the questions:

What is it to be a moral person? or: How ought one to make moral
decisions? And those are indeed two of the three questions he lists as
his concern at the bottom of that first page. But this cannot be Dancys
main concern. In a discussion of codifiability, Dancy describes utilitarianism (he calls it welfarism) as at the other end of the spectrum
from particularism. A utilitarian may well think that the moral person,
the person who makes moral decisions as they ought, will not make
any reference to principlesif that is the most effective way to maximize utility. In fact, all the positions on how reasons work that he discusses (including what he calls generalism) are consistent with the
view that human beings possess a remarkable capacity to understand
what to do in each individual case, requiring no reference to principles.
* For comments on an earlier draft, I am grateful to Nick Zangwill and participants
in the conference on moral particularism held at the University of Kent, Canterbury,
on 1 December 2004.
1
Jonathan Dancy, Ethics Without Principles (Oxford: Clarendon Press, 2004). All
unattributed page references in the text are to this work.
96
roger crisp
Dancys real concern is his third question: How is it possible for an

action to be right or wrong?
It would be no part of the kind of answer Dancy is looking for to his
question to say something like: Well, God created the world. Thats
what made it possible for an action to be right or wrong. Dancy is concerned with the very nature of rightness and wrongness themselves,
and in particular with the nature of moral reasons. Indeed, on the second page of his book he says that he has written a work which is not an
investigation of the subtleties of our moral life, but one about how to
understand the way in which reasons work. The view about reasons
that Dancy himself wishes to defend is holism:
H: A feature that is a reason in one case may be no reason at all, or an
opposite reason, in another. (p. 7)
Holism can be understood in two quite different ways. According to

what we might call weak holism:
WH: A feature that is a reason in one case may be no reason at all, or an
opposite reason, or always the same reason, in another.
WH can be understood as a view about the nature of reasons.

Essentially, the suggestion is that there is nothing in the idea of a reason
which means that any reason-giving feature must always function in
the same way across cases. But it may turn out, on investigation, that
there are in fact reasons which do always function in the same way
that is, that there are cases of what Dancy calls invariance (pp. 7778).
And because Dancy appears to allow for the possibility, if not the actuality, of invariance in Ethics Without Principles, it might seem that it is
WH that he has in mind.
A strong version of holism, however, will not allow invariance:
SH: Any feature that is a reason in one case may be no reason at all, or an
opposite reason, in another.
In his earlier work, Moral Reasons, it seems that Dancy may have
been aiming to defend SH. At one point, for example, he considers the
common charge against his particularistic view, grounded on holism
in the theory of reasons, that it is always bad knowingly to cause pain.2
Rather than allowing that a particularist could accept that indeed it is,
Dancy seeks to show that it is not.
In opposition to both versions of holism is atomism:
2
Jonathan Dancy, Moral Reasons (Oxford: Blackwell, 1993), pp. 6162.
ethics without reasons?97

A: A feature that is a reason in one case must remain a reason, and retain
the same polarity, in any other. (p. 7)
Holism and atomism concern the behaviour of reasons. But, as we

have already seen, Dancys book is also an attack on principles. What is
the relation, as he sees it, between views about reasons and views about
principles? Holism is said to line up with particularism:
P: The possibility of moral thought and judgement does not depend on
the provision of a suitable supply of moral principles. (p. 7)
And atomism with generalism:

G: The very possibility of moral thought and judgement depends on the
provision of a suitable supply of moral principles. (p. 7)
We have already seen that Dancys concern cannot be with what

moral thought and judgement, in practice, are like. So we need to interpret particularism and generalism in the light of his real concern,
and what he says about holism and atomism. I suggest that they are
best understood as adding clauses to holism and atomism, about
the relation between reasons and principles. So weak particularism
would be:
WP: A feature that is a reason in one case may be no reason at all, or an
opposite reason, or always the same reason, in another. If there turns out
to be no feature that is always the same, then there will be no general or
universal principle stating the reason-giving status of any such feature.
WP, then, is consistent with there being general or universal moral

principles. But there might still be truths about what one should do in
particular cases which could be stated only in the form of particular
non-universalizable propositions.
A strong form of particularism will claim:
SP: Any feature that is a reason in one case may be no reason at all, or an
opposite reason, in another. So there can be no general or universal principles stating the reason-giving status of any such feature.
And according to generalism, understood in the same way:

G*: A feature that is a reason in one case must remain a reason, and retain
the same polarity, in any other. There is at least one such feature, and it
can be described in the form of a universal principle.
According to Dancy, it is holism that provides the main argument

for particularism (p. 73). So let me first discuss SH. I find four main
problems with it, or with Dancys arguments for it.
98
roger crisp
First, even if SH is accepted, the difference between a particularism

based on it and generalism may be quite small. Particularism can be
more or less radical. The most radical form would say that, of any reason identified in any particular case, we could make absolutely no
sound predictions about whether it would have the same force or even
polarity in any other case, even one very similar to the existing one.
One obvious problem for radical particularism is that it looks as if
we can come up with some plausible generalizations about reasons,
such as:
T: In at least nearly every case, the fact that some action is one of causing
suffering to a non-rational sentient being counts against it.
This is consistent with SH, since it may be the case sometimes that
the property some action has of being a causing of suffering to a nonrational sentient being does not count against it, or even counts in its
favour.
Inconsistent with SH would be:
T*: In every possible case, the fact that some action is one of causing suffering to a non-rational sentient being counts against it.
Note that these principles are quite similar. Note further the implication of the distinction between the criterion of rightness and decisionprocedures which allowed us to focus on Dancys real concern.
Someone who accepts T may think that, in practice, it would be wiser
to teach children T*, on the ground that one maximizes the chances of
action in conformity with the real reasons.
But even if we assume that this is not the case and that we should
teach children T rather than T*, it seems because the principles are so
similar that the difference between particularism and generalism may
not be especially great. To be sure, an ideal moral agent, according to
the particularist, is likely to need pretty good judgement to tell when a
case is an exception to a principle. But a generalist like Ross is also
going to require judgement about the weight of different principles in
particular cases.
The second problem is Dancys unreflective reliance on commonsense views about which reasons we in fact have. One obviously plausible strategy for understanding how reasons work is first to offer a
(defeasible) account of certain reasons, and then assess how they
behave in different circumstances. But such an account should not be
taken directly from common sense, as it so often is in writing about
reasons. Consider the following example. In his first chapter Dancy

criticizes what he calls the subsumptive option, a form of generalism
according to which something like the following is true:
SO: There is at least one universal moral principle, and moral agents may
therefore proceed to make each moral decision by seeing what that principle tells them to do in each particular case.
SO is consistent with many different normative theories. One might

easily imagine a subsumptivist utilitarian, for example. But Dancy
claims that SO has been largely discredited, partly because it can make
no sense of the relevant notion of moral regret.
The argument about regret rests straightforwardly on an appeal to
common-sense phenomenology. Dancys example is the standard one
of someone who has to break an important promise to help someone in
the street. Now of course most of us understand fully what that feeling
of regret is like. But whether we should give it any weight depends on
whether it is a response to genuine reasons or not. A utilitarian defender
of SO will say that it is not, and that therefore we should not appeal to
it when assessing accounts of the nature of reasons.
A third, and related, problem with Dancys argument is more serious, since it occurs in his main argument for holism and hence particularism (pp. 7378). This argument rests on certain examples. Dancy
starts with theoretical reason, and claims that everyone will agree that
theoretical reasons function holistically. If that were right, then we
would have at least some reason to expect the same to be true of practical reasons. Dancys example here is a case in which it now seems to me
that something before me is red. Normally, that is a reason for me to
believe that there is something red in front of me. But in a case where
I also believe that I have taken a drug that makes blue things look red,
and red blue, the appearance of a red-looking thing before me is reason
for me to believe that there is a blue thing, not a red thing, before me.
Dancy is right that, if his alleged reason is a reason, we are all going
to accept that it functions holistically. But the example depends on a
failure to distinguish, in the theoretical sphere, between ultimate and
non-ultimate reasons. The principle at stake here is something like:
You have an ultimate epistemic reason to believe that some object is
the colour it appears to be, if you believe that you are viewing that
object in normal conditions. Normal conditions could be further specified, and would doubtless include the lack of influence of visually distorting drugs.
100
roger crisp
Dancys examples of practical reasons are faulty in the same way, as

were the examples he used in Moral Reasons. He offers two:
[T]hat there will be nobody much else around is sometimes a good reason for going there, and sometimes a very good reason for staying away.
That one of the candidates wants the job very much indeed is sometimes
a reason for giving it to her and sometimes a reason for doing the
opposite. (p. 74)
These reasons are non-ultimate or derivative, and depend for their

normative force on certain higher-order reasons which can be seen to
subsume them into a single generalist form. On Monday, I think I shall
enjoy some solitude; so the fact that there will be nobody else around
on the hills on that day is a good reason for going there. But on Tuesday
I have had enough, and want some company. So the fact that there is
nobody around counts against heading for the hills. But on both days
what I am after is enjoyment, something (perhaps) I always have ultimate reason to pursue for myself.
A similar story can be told about the second case. Imagine a scenario
similar to that of Bernard Williamss famous story of George.3 In the
morning, Georgina is interviewed for the job of primary school teacher.
She wants the job very much. Her wanting the job is a reason to give it
to her. In the afternoon, Georgina is interviewed for the job of chief
researcher in a chemical weapons plant, where weapons are being
developed for purposes of genocide. She wants the job very much, and
in this case her wanting the job is a reason not to give it to her. What
might be the higher-order principle here? One obvious one might be
the utilitarian principle that we should do what produces the most
good; but there are many other possibilities.
The final, and most weighty, difficulty for SH is the plausibility of
claims such as T*. Here the challenge to the holist is simpleto describe
some possible case in which the fact that some action causes suffering
to a nonrational sentient being does not count against it. Until that
challenge is adequately met, then SH must be rejected.
Weak holism, however, is not falsified by the availability of universal
moral principles such as that captured in T*, since WH allows for the
possibility of invariance. According to the advocate of WH, we should
3
Bernard Williams, A Critique of Utilitarianism, in J. Smart and B. Williams (eds.),
Utilitarianism For and Against (Cambridge: Cambridge University Press, 1973),
pp. 75155, at pp. 9798.
not begin enquiring into our reasons, or their nature, on the assumption that they must, to be reasons, function universally in the way that
atomism says. But reflection upon the nature of explanation suggests
analogies with the nature of justification which throw doubt upon WH.
Important here is the distinction Dancy draws between favouring
and enabling (pp. 3843). Here is one of Dancys own examples:
1. I promised to do it.
2.My promise was not given under duress.
3. I am able to do it.
4. There is no greater reason not to do it.
5.So: I do it.
Premise 1, Dancy suggests, is a favourer: the fact that I have promised counts in favour of my doing what I have promised. 3, however,
merely enables 1 to favour 5. It does not itself add to the reasons in
favour of acting. Dancy believes that 2 is also an enabler rather than a
favourer.
In an earlier article on particularism, I sought to draw an analogy
between explanation and justification, claiming that just as what
I called a full explanation of some event, such as an action, will include
everything, including background or enabling conditions, so the full
story about justification should not leave anything out.4 Dancy says
that, because I insist on bringing enabling conditions into the justification, in that earlier argument I do not respect the distinction between
right-making features and enabling conditions (p. 47). I did not, however, wish to claim that all enabling conditions must feature in a justification (though they should feature in a causal explanation). For
example, if I now have a strong justification for -ing, the fact that God
has not just struck me dead is one of the conditions that enables me
now to act. But it should not be part of my justification. Rather, what
should be taken up into the justification are all those allegedly merely
enabling conditions that are in fact justificatory. This is why I am what
Dancy calls an expander (p. 96), and why I seek what he calls guarantees (pp. 9597). As Dancy puts it: Every ultimate reason is a feature
whose presence guarantees that there is a reason (i.e. that there is
Roger Crisp, Particularizing Particularism, in B. Hooker and M. Little (eds.),
Moral Particularism (Oxford: Clarendon Press, 2000), pp. 2347, esp. at pp. 3436,
4244.
4
102
roger crisp
something to be said for doing the action), and any reason we can rest
satisfied with is an ultimate reason (p. 96; see also pp. 12527).
Dancys first objection to this idea is that a guarantee that there is a
reason to do the action need not itself be a reason in favour of doing it,
since the only way I could reach a guarantee would be by adding enablers and the absence of disablers to an incomplete reason. That would
mean that anything that could be a reason for doing the action has got
submerged in a host of considerations that are doing other jobs. The
crucial point here is that a combination of a favourer and an enabler
need not itself be a favourer at all (p. 96).
On that final point, as I have said, I think Dancy is right. But I think
also that the method of expansion he describes is better understood in
terms of finessing a favourer. To use the example of promising again,
what happens is that, once we learn that a promise has to be freely
given, we amend the claim that having promised to counts in favour
of -ing to the claim that having freely promised to counts in favour
of -ing. So I have little doubt that premise 2 should be incorporated
into the justification of my action.
In support of his claim that what is doing the normative work in this
case is not the combination of the facts that I have promised and that
I was not under duress when I promised, but rather just the fact that
I have promised, Dancy suggests that even those whose promise has
been deceitfully extracted often feel some compunction in not fulfilling
their promise, even though they realize that their promise does not
play its usual reason-giving role. I confess that this is a phenomenon
I fail to recognize. But even if it exists, it is best explained as a remnant
of our moral upbringing, in which we are taught the simple principle
that it is wrong to break promises. A second example of Dancys concerns the reason you have to help someone who asks you the time. If
their purpose were to distract you so that their accomplice could steal
your bag, that reason would not exist. But Dancy is unwilling to say
that the real reason must be construed as their asking you the time for
some appropriate purpose. Again, this seems to me a case where the
alleged enabler should be incorporated into the favourer, enabling us
better to understand the justification in question.
Here we should also note Dancys readiness to allow the possibility
of a principled but holistic ethic, which might include a principle
suchas:

Q: If you have promised, that is some reason to do the promised act,
unless your promise was given under duress. (pp. 81; cf. pp. 11217)
This principle is meant to be holistic in the sense that what is doing

the normative work when I have freely promised is the fact that I have
promised. And the polarity of that reason changes to neutral when the
promise is unfree. But the idea that what is really providing normative
weight in this case is merely the fact that you have promised does seem
to me misleading. It is the fact that you have freely promised that makes
the difference, and that can be turned into an atomistic generalist principle quite happily.
As for Dancys premise 3, I suggest that it could feature in a justification of my trying to , and indeed it is not implausible to claim that all
reasons for action are reasons for trying to act. So in many cases what
Dancy alleges to be a mere enabler is best understood as part of a
favourer. If some feature that is claimed to be an ultimate reason for
action in one case fails as a reason in some other case, then either the
two features must be allowed to differ or we have followed the white
rabbit into a world where anything can happen. Normative principles,
that is to say, are like typical natural laws. Each system helps us to
understandin one case, why something happened; in the other, why
someone should do something. Unexplained exceptions can only hinder our understanding, and spur us on to avoid them through further
specification. Dancy may claim that he is not leaving exceptions unexplained; but if that is the case then the ultimate reasons must be understood to differ in the two cases in question. If am justified in -ing but
not in -ing, that must be because my ultimate reasons differ. If those
reasons are the same, then there just cannot be any difference in justification; and this is indeed part of what we understand by a reason.
I have suggested an analogy with natural laws, and accept that some
natural laws are statistical. Much has been done to find an underlying
non-statistical account at the quantum level, and at present there is
general consensus that it is unavailable. The projection postulate is correct, and God really does play dice. In ethics, then, it might be argued
that whether we should be holists turns on the question of what our
ultimate normative reasons turn out to be, and how they behave from
case to case. They might turn out to be analogous to the projection
postulate and to behave in a particularist way, or, as the generalist
might claim, to be analogous to the Schrdinger equation, or perhaps
104
roger crisp
better, given their fundamental role in ethics, to the law of the conservation of energy. Something like this view that we should have an open
mind on how reasons behave is what lies behind WH.
But the particularist analogy fails. Justificatory reasons must function universally, like standard causes. The WH picture seems to be the
following. At the quantum level, we might have some set of initial conditions s which issues in outcome o. But we cannot strictly claim that
s causes o, because we know that some identical set of initial conditions
s* might issue in an out-come p which is non-identical with o. What we
have here is a case of genuine indeterminism, which is to that extent
inexplicable: it just happens that way. Similarly, it may be that some
feature of an action f functions as an ultimate justificatory reason for
performing some action, but in a similar situation fails so to function
or even has the opposite polarity.
Ultimate reasons for action cannot be like that, however, since they
help us to understand why someone has a justification for acting. If
I claim to be justified in -ing because of some feature of -ing, but
then refuse to allow that this same feature justifies me in -ing, then
I have further work to do in specifying exactly what my reasons are.
A staunch holist might refuse, but this is to follow the white rabbit.
What is more likelyas usually happens in Dancys caseis that the
justification will in fact be further specified, but holism retained
through the use of a distinction between enablers and favourers that,
respectable enough though that distinction is, is not operating here.
WH allows that, as things turn out, some reasons may in fact always
have the same normative force across different cases. So let me end by
saying a little about Dancys views on invariance:
It may be true that every reason may alter or lose its polarity from case to
case, even though there are some reasons that do not do this. If they dont
do it, this will be because of the particular reasons they are. Invariant
reasons, should there be any, will be invariant not because they are reasons but because of their specific content It is like the claims that a man
can run a mile in four minutes, that Sam Smith is a man, and that Sam
Smith cannot run a mile in four minutes. These claims are compatible,
and so are the claims that reasons are variable qua reasons though some
reasons are (necessarily, given their content) invariant. The invariance,
where it occurs, derives not from the fact that we are dealing here with a
reason, but from the particular content of that reason. (p. 77)
The thought here is that, just as some men turn out to be able to run
a mile in four minutes and some do not, so some reasons may turn out
to be invariant and others not. But Dancy would be unhappy with this
view of reasons:
It is, I suppose, conceivable that though the vast bulk of reasons function
according to a holistic logic, there are a few whose logic is atomistic. But
if this were true we would have a hybrid conception of rationality. There
would just be two sorts of reasons, each with their own logic, and moral
thought would be the uncomfortable attempt to rub such reasons
together. It is much more attractive, if at all possible, to think of our reasons as sharing a basic logic, so that all are atomistic, or all are holistic.
(p. 77)
Dancy is somewhat non-committal on whether he thinks that there

are any invariant reasons. But this passage strongly suggests that he
thinks that all reasons are in fact variant. If so, then he owes the generalist, or the holistic invariantist, an account of how an actions being a
causing of suffering to a non-rational, sentient being can fail to count
against it.
Dancy does discuss a case of invariance similar to that I have just
alleged: the fact that an action causes gratuitous pain to unwilling victims. He then allows that in the famous case of the fat man stuck in the
cave, the fact in question is some reason against lighting the dynamite
and blowing him out. He goes on:
The question I want to raise is whether the fact that this featureis functioning as the reason it here is, is in any way to be explained by appeal to
the (supposed) fact that it functions in the same way in every case in
which it occurs. It seems to me that this feature is the reason it is here
quite independently of how it functions elsewhere. (p. 78)
This passage seems to be in effect a restatement of the holistic conception of a reason, according to which how a reason functions in one
case is clearly going to be independent of how it functions in another,
because of the possibility of variance. But we have seen that the very
concept of a reason is atomistic and generalist. So in that sense whether
some feature is a reasonthat is, an ultimate reason (the kind that
really matters)or not will not be independent of how it functionselsewhere. If it functions differentlyif its polarity switches, for example
then that will be a clue that what we are dealing with is not a
counter-example to generalism but a non-ultimate reason.
In conclusion, let me restate my main claims. The central topic of
debate here is whether holism or atomism captures the truth about
normative reasons. Holism can be understood as a weak view about the
106
roger crisp
nature of reasons, which would allow for invariance, or as a strong view

which rules out invariance. There are four main problems with strong
holism, or Dancys arguments for it: a plausible particularism based on
it is likely to be in practice very close to generalism; Dancy assumes the
correctness of certain commonsense views about our reasons; his
examples are of non-ultimate reasons; and there are certain universal
principles it is hard to believe could be wrong, such as that the causing
of suffering to a non-rational sentient being always counts against an
action. The main difficulty with weak holism is that justification can be
understood as analogous to explanation, and this gives us an atomistic
and generalist conception of a normative reason. Here Dancys arguments frequently depend on identifying as an enabler what is more
plausibly seen as part of a favourer. Generalism, then, is true. We should
be relieved about this, since it leaves no mystery in ethics analogous to
that discovered by physicists at the quantum level. God may play dice
down there, but not up here in the world where we live and act. Indeed,
it is hard to see how he could.
Defending the Right

Jonathan Dancy
1. Ultimate and Non-ultimate Reasons
Roger Crisp distinguishes two forms of holism, weak and strong. The
main difference between these lies in the consequence they take to follow from a common holistic premise:
A feature that is a reason in one case may be no reason at all, or an opposite reason, in another.
Strong holism says, effectively, that if this is true there can be no general principles stating the reason-giving status of any such feature.
Weak holism allows that there will be such a principle for any invariably reason-giving featureand that there may be such features.
Crisp starts by considering the stronger form of holism. He makes
various points about the sort of argument that I have produced in its
favour. One way of arguing for holism is by appeal to examples. Crisp
repeats his complaint (in the Hooker/Little volume1) that the examples
I offer fail to distinguish between ultimate and non-ultimate reasons.
In his earlier version, he characterized an ultimate reason as a consideration that we can rest satisfied with as a reason. This made it easy for
me to say that I was perfectly satisfied with lots of reasons that Crisp
wanted to call non-ultimate. Now he has a different account of the ultimate: These reasons are non-ultimate or derivative, and depend for
their normative force on certain higher-order reasons which can be
seen to subsume them into a single generalist form.2
What is going on here? There are, I think, various possibilities. One
is that what this means is that the considerations we ordinarily think of
as reasons are not really reasons at all. Properly speaking, they are parts
of reasons. A part of a reason is not subsumed into something higherorder; it is an element in a complex which is itself a reason. No part of
B. Hooker and M. Little (eds.), Moral Particularism (Oxford: Oxford University
Press, 2000).
2
Roger Crisp, Ethics Without Reasons?, p. 44, this volume.
1
108
jonathan dancy
that complex is more a reason than any other part. If we look inside
that reason, dismantle it, as it were, what we find are things which do
not favour anything (at least not here). The only real reasons are the
ultimate ones.
A different interpretation allows that non-ultimate reasons are reasons all right, but claims that these reasons get their normative force
from being sub-sumed under certain higher-order reasons, which are
invariant. (I think this claim of invariance is what Crisp means by his
talk of their having a generalist form.) This interpretation would fit
something Crisp said in his earlier attempt; he held there that the virtues were ultimate reasons, suggesting that the ultimate reason for
doing many things was that doing them would make ones life go better.
It seems to me, however, that though one can generalize in this way, in
doing so one goes a step too far. The reason why things make ones life
go better is that they are fun, rewarding, engrossing and so on, worth
doing in their own right. It is not the other way around. Perhaps better:
that something is fun, or in some other way worth doing, is not a reason to do it because one has a general reason to make ones life go better, and having fun is one way of doing that (sometimes). That this
would be fun is its own reason to do it, and ones life goes better to the
extent that one is able to do a wide variety of the things that one has
good reason to do. The so-called ultimate reason is the derivative one,
I would sayif it is a further, or distinct reason at all. If there is normative force here, it comes up from below, not down from above.
We will return to these issues. But for the moment we are only considering the fortunes of strong holism, which goes further than anything I was prepared to defend in Ethics Without Principles. As Crisp
points out, strong holism is in danger of being refuted by the discovery
of a single invariably relevant consideration, complex or simple. (I say
in danger of because I remain worried that the complex consideration
at issue may not itself be a reason, but rather a mere guarantee that
there is a reason.) So we pass to weak holism, which allows for the possibility of invariance. Crisp opens his enquiry by noting that, according
to the weak holist, we should not start off with the assumption that our
reasons must somehow function universally, as the atomist claims they
must. But, he says, the relation between justification and explanation
suggests otherwise. I found it quite hard to work out why he says this.
There are two points on which Crisp and I agree. First, he accepts the
crucial distinction between favourers and enablers. Second, he allows
that a combination of favourers and enablers should not necessarily
defending the right109
itself be a favourer. With these things in hand, as I see it, Crisp is thinking that, though we start from some favourer (perhaps that I promised
to do it), and add an enabler to it (that my promise was freely given),
the demands of justification and explanation drive us onwards until we
reach a specification of some complex of features which is an invariant
reason, even if the things from which we started, and what we have at
each intermediate stage, are not. What we have at the end of this process, I urged, was not a reason of any sortnot, that is, some sort of
favourerbut a guarantee that something favours the action, a guarantee which will contain a mention of the favouring consideration, but
which will not give it any pride of place. This would not respect Crisps
earlier thought that the normative force of the reason derives, by subsumption, from that of the guarantee. The guarantee has normative
force merely because of the force of the reason whose presence it guarantees; again, things are the other way around.
Crisp has seen this all before, and does not think it is a correct
account of what is going on. One way in which things have got distorted is that, he says, he does not think that every enabler must feature
in the relevant justification. For instance, I may have a strong justification for doing something which I could not do if God were to strike me
dead, but that God has not struck me dead is no part of the justification
for my doing that thing. This, however, is not to the point. An enabler
is not a feature required for me to do the action; it is a feature required
for the reason to do it to be the reason it is. What is enabled is not the
action but the reason; it is not that I am enabled to do the action but
that the feature is enabled to favour the action. I am sure that Crisp
knows this crucial aspect of my notion of an enabler perfectly well, but
he seems to have forgotten it at this point. However, having forgotten it,
he is in a position where he can say that only those enabling conditions
need to be part of the justification that are themselves justificatory. But
this was just the position I was complaining about. I dont feel that enabling conditions justify anything. One might call the relevant reason
justificatory, or one might call the whole story justificatory, but a
required enabler is only justificatory in a weak sense. We see this sense
emerging when Crisp writes It is the fact that you have freely promisedthat makes the difference.3 The difference it makes, I would say,
is the difference between a promise that gives us a reason and one
Crisp, Ethics Without Reasons?, p. 47, this volume.
110
jonathan dancy
thatdoesnt. But this difference is not enough to call the enabling condition justificatory. The only thing it justifies is seeing a reason to keep
this promise when one sees no reason to keep another. If the question
raised by someone who seeks justification is why there was a reason to
keep the first promise and no reason to keep the second, then this is
what the fact that the enabler was present in the first case and not in the
second will provide. But in that case what it is justifying in the first case
is not doing the action, but rather doing it when one is not going to
keep ones promise in the second, similar case. What I would say, then,
is that the enabler only succeeds in justifying ones comparative choice.
But in that role, it is acting as a favourer. It favours this rather peculiar
object: keeping ones promise in the first case and not in the second.
But this does nothing to show that it is part of the justificatory story for
the first case alone.
As I see things, it is the comparative question that drives the demand
to include within the relevant reason, properly understood, everything
whose presence or absence can make a difference to the normative
score, whether reason, enabler, intensifier, disabler or attenuator.
There remains a worry about the White Rabbit. Crisp writes: If some
feature that is claimed to be an ultimate reason for action in one case
fails as a reason in some other case, then either the two features must
be allowed to differ or we have followed the White Rabbit into a world
where anything can happen.4 Now I would have thought that it is
straightforwardly impossible for something that is an ultimate reason
in one case to fail to be a reason, indeed an ultimate one, in another.
But suppose instead that something that is claimed to be an ultimate
reason in one case turns out not to be any such thing in another. Then,
I think, by the definition of ultimate reason we know straight off that
that feature is not an ultimate reason anywhere. The worry about the
White Rabbit only emerges if we have a feature which is a reason in one
case and not in another, though there is no explanation of this at all.
This is at least conceptually possible, I would claim, though it is not
possible if we are dealing with ultimate reasons, which are defined in
such a way as to have to be invariant.
So what I want to say about this is that if we are dealing with a feature
which is a reason in one case and not in another, there will be an explanation of this, which will consist in the presence of some disabler or the
4
Crisp, Ethics Without Reasons?, p. 47, this volume.
absence of some required enabler. There is no threat of the White

Rabbit here. But a similar question cannot be asked about ultimate reasons. If so, the fact that there is no answer to it when we do ask it (everything we can find to say having been used up already) does nothing
to banish us to the world of the White Rabbit. My own view is that it is
ultimate reasons that belong in Wonderland.
2. Default Reasons
What Sean McKeever and Michael Ridge offer here is not so much a
new criticism of particularism as a supposedly insuperable difficulty
for a new aspect of particularism, or rather, for an aspect of particularism that has only recently begun to receive the attention it deserves.
In my Moral Reasons I mooted the possibility of default reasons, but
only tentatively. I had two reasons for being tentative. The first was
that I was not sure that it was necessary to sully the pristine purity of
particularism with the complications attendant on the notion of
defaults. The second was that I was not quite sure that this notion
ofdefaults would not turn out in the end to be a poisoned chalice. Like
Simon Kirchin, McKeever and Ridge take the view that without
defaults, particularism is much less attractive than it otherwise would
be, much less consonant with our moral intuitionsby appealing to
which, after all, it gains much of its initial plausibility. Alan Thomas is
rather of the opposite opinion, but I side with McKeever and Ridge
on this. They also think, however, that no sense can be made of the
distinction between default and non-default reasons, once we come to
look at the matter carefully. The upshot is that particularism needs
something that it cannot get. I have certainly learnt a lot from their
discussion of the issue.
McKeever and Ridge distinguish three ways in which one might try
to understand the notion of a default reason. The first, mine, is metaphysical. The second, which they associate with Lance and Little, is
epistemological. The third is pragmatic. In what follows I will try to
defend, or perhaps rather to reconstruct, the metaphysical approach;
I think that the other approaches only look attractive if the metaphysical one cannot be got to run.
McKeever and Ridge find nothing especially wrong with the intuitive, though metaphysical, way in which I introduce the notion of
a default reason. Some reasons, I suggested, arrive switched on.
112
jonathan dancy
Others arrive switched off, but can be switched on by appropriate contexts. And the ones that arrive switched on can be likewise switched off,
if the situation so determines. So all this is formally compatible with
the constraints of holism, and it is therefore a picture that is available to
particularism.
As well as the metaphor of switching on and off, I offered a somewhat less metaphorical way of thinking about the default. If a default
reason is not switched off, but functions here at the default setting, as it
were, there is nothing to explain. It is only when things are not as they
are default-set to be that we begin to ask questions. If a default reason
is not here the reason it is antecedently set up to be, there must be an
explanation of the matter. The reverse of all this is true for the nondefault reason; here explanation is called for if these are reasons at all,
while no explanation is needed if they remain normatively inert in the
background.
But even this less metaphorical way of capturing the notion of a
default is not something of which one can be very proud, and it is easy
to see that there may be difficulties involved in trying to work the general idea out in more detail, as we would all wish to do. What I had not
foreseen is how easy it would be for the whole idea to be undermined,
as McKeever and Ridge have shown it to be. The first move in this process involves the link between defaults and enablers. A non-default reason is one that needs something to switch it on, and I have a term for
this sort of switcher-on: an enabler. So the default reasons must be ones
that need no enabler. If they are like the non-default reasons in needing
an enabler, it can hardly be true that they arrive switched on. They
would only be switched on if the enabler is there, which it might not be.
Of course one might try to say that what they need is not so much an
enabler as the absence of disablers. That sounds much better; there are
things that can switch them off, but so long as none of those disablers
is present they will stay switched on. The trouble is that there seems to
be no difference between the presence of an enabler and the absence of
a disabler, nor between the absence of an enabler and the presence of a
disabler. Our understanding of these terms is run entirely by appeal to
certain subjunctive conditionals, and subjunctive conditionals are not
capable of drawing distinctions fine enough to distinguish the presence
of something that would stop a default reason from being a reason
from the absence of something required for that reason to be a reason.So the absence of a disabler is an enabler, and so default reasons,
which certainly must have disablers if holism is to be true, must also
have enablers; and so the distinction between default and nondefault

reasons collapses.
This all seems to me to be very persuasive. It serves as further
evidence of something of which I was already convinced, namely, the
general futility of appeals to subjunctive conditionals. This is one of my
favourite themes. The trouble is that I seem to forget it when it suits
mesomething which I would be quick to point out in anyone else.
The question then seems to be whether some alternative way of characterizing the relevant difference can be found. What is the relevant difference? There seem to be two differences. The first is the difference
between the default and the non-default. The second is the differencebetween enabler and disabler.5 My mistake was in trying to rely
entirely on the second distinction to make sense of the first.
There does seem to be a distinction in role that is not a mere reversal
of some subjunctive conditional. When I wrote of switching on, I did
not have in mind anything as negative as merely failing to stand in the
way, which is all that is achieved by the absence of a disabler. Here is a
way to think about it. I can get over the stile by myself, thank you very
much, without your help, though if you get in the way I cant get over it
at all. Your not getting in the way, that is, your not preventing me from
getting over, is not at all the same as your helping me over. It enables me
to get over, but it does no more than that; when I do get over, I do so
under my own steam. Someone else might be unable to get over the
stile without help. I suggest that we try to think in the same sort of way
about the distinction between default and non-default reasons.
This seems a promising line, and the question is whether we canmake
adequate theoretical sense of it. There is a distinction that I appealed to
in Ethics Without Principles, which I owe to Marcus Giaquinto. He suggests that a belief that is caused by certain experiences may still be able
to count as a priori or conceptual knowledge, so long as the experiences that cause it do not act as grounds for it. There is of course the
classic Quinean objection to any such claim that a priori or conceptual
knowledge is possible, which is that any sentence might come to be
rationally rejected as false in the light of experience. Giaquinto admits
this, but claims that it is irrelevant. The truth of the relevant sentence is
It was an undergraduate at Georgetown, Adil Haque, who first suggested to me
that it was a mistake to conceive of the difference between enablers and disablers so
symmetrically if I wanted to use those terms to make room for defaults. But I didnt at
the time see what he was getting at.
5
114
jonathan dancy
indeed dependent on experience, in the way that Quine claims, but this
is negative dependence; a belief is negatively dependent on experience if
its retention is dependent on the absence of awkward experience.
A belief is positively dependent on experience if experience acts as its
ground, that is, if the acceptability of getting the belief depends on the
presence of supporting experience. Giaquintos point is that a belief
can be negatively dependent on experience without being positively
dependent on it. And to constitute a priori, or conceptual, knowledge,
the belief only needs positive independence. Quines point that it is
negatively dependent on experience is irrelevant.
This seems to be another instance of the distinction between the
absence of an impediment (a disabler) and the presence of something
more positive, in this case a ground. The absence of an impediment is
nothing like the presence of a ground, nor is the presence of the impediment anything like the absence of a ground. Could we therefore try
saying, as it were, that a default reason is negatively dependent on the
absence of a disabler, but not positively dependent on anything?6 If we
could make this out, something would have been salvaged from the
flames. What we are after is a distinction between something worth
calling support and the mere absence of an impediment. We find versions of this distinction in pairs of the following sorts:
I can bicycle from home to campus in 15 minutes if there isnt a strong
head-wind. (The absence of the headwind does not produce any of the
power required.)
I can bicycle from home to campus in 10 minutes with a strong following
wind. (The wind helps; it produces part of the power required.)
I can get over the stile if you dont get in my way, hinder me. He cant get
over the stile without your help.
I can get over the stile if I am not too tired. My not being tired doesnt
help me get over the stile; my being tired prevents me from doing what
otherwise is in my power, which amounts to saying that it deprives me of
the power.
In some cases, the agent can do it without help; in others, he cant.

But what sort of help do we have in mind here? It was not supposed to
be the sort of help provided by another reason. So when I do the journey in 15 minutes we want to say that the prime agent is me, not the
Or on anything specific/local, at least.
absence of wind. When I do the journey in 10 minutes, however, there

is a sense in which the wind and I do it together. That seems to be a
case, then, in which the eventual agency is joint. But this is not the only
possibility. It is not that together, you and he can get him over the stile.
That would be a special case. The more normal situation is that he can
get over the stile, though he needs your help to do it.
The absence of a disabler for a default, i.e. the presence of an enabler,
merely permits something. The presence of a disabler prevents it. What
is the relation between permitting and preventing? That which is permitted may, for all that, not happen. That which is prevented will
certainly not happen. The opposite of preventing is therefore not permitting but ensuring, or making happen. But an enabler does not
ensure. Perhaps, then, a default reason requires the absence of a disabler; it can only be the reason it is set up to be if nothing prevents it
from being so, that is, if all required enablers are present. But it does
not require anything to make it a reason.
It appears therefore that we need a distinction between not needing
help and having no enablers, or between allowing something to be a
reason and making it into a reason. For default reasons need no help
but, as all reasons may do, they have enablers. The question now seems
to be what to say about the non-default reasons, which we are now
understanding as needing help if they are to be reasons at all. What
account are we to give of the role played by a consideration whose presence converts what would otherwise not be a reason into a reason?
I used to speak of such things as enablers, but now it seems that I had
too few categories, and of the wrong sort. Enabling is too thin a category to make the required sense of default reasons. The enabling/
disabling distinction is a thin logical distinction, within which we must
observe certain thicker or meatier distinctions. In this it resembles
Giaquintos remarks about the dependence/independence distinction.
He suggests that we need a meatier form of dependence, dependence
on a ground, or groundedness; dependence on the absence of recalcitrant experience is merely a permitting condition, not an ensuring one.
Suppose that her asking him not to do it is a reason for him to do it.
Some form of positive intervention is required for this to be the case;
that she asked him not to do it is made into a reason for him to do it,
not merely permitted to be one, which is all that an enabler could do.
(Perhaps they are playing some peculiar form of parlour game.)
Theintervening consideration need not be a further reason. If I climb
over the stile with your help, it is still me that climbs over the style.
116
jonathan dancy
Suppose that you increase my confidence to such an extent that I

become able to do what previously I could not do. You help me do it,
but it is I who do it. There is a distinction between our doing something
together and my doing it with your help. (Think of supervising a graduate thesis, and of the cases where the one begins to slide imperceptibly
into the other.) In this way we may manage to retain the sense that a
feature that would not otherwise be a reason is made into a reason by
the presence of some further feature that is not itself a reason at all. But
some features do not need to be turned into reasons by the presence of
some other feature, and these are the default reasons.
3. Reasons for Action
Richard Norman suggests that there is potential trouble in the idea of a
reason for action. He thinks that this phrase encourages the idea of
reasons as pushes or pullsan idea that is apparent in talk of positive
and negative polarity. I dont agree with him on this point. First, I think
of polarity in terms of the positive and negative terminals of a battery,
and electricity does not work by pushes and pulls. Second, I do not
think of reasons for action as considerations somehow capable of getting an inert body to move, and I dont see why one should think that
the very phrase reason for action encourages one to do so. But Norman
makes a more positive suggestion:
When [reasons] are employed to reach decisions about what to do here
and nowthe question which they address is not Should I act? but
Would it be better for me to do this or that?favouring and disfavouringis more a matter of exhibiting a match or mismatch between the
character of the circumstances and the character of the action.7
In fact there are two suggestions here, the link between which is
obscure. The first is that a reason for action is a consideration that
makes it better to act in one way than in another. This seems to differ
from the standard account of reasons as favourers in two ways. First, it
thinks of favouring as making good; I take this to be a distinct shift,
and will say why below. Second, it thinks of reasons as always comparative in their focus; a reason does not just favour an action; it favours
doing this one rather than doing that one.
Richard Norman, Particularism and Reasons, p. 34, this volume.
Normans second suggestion is a modern version of the classical

intuitionist view, first seen in Clarke, that the right action is the
one that most fits the situation, broadly conceived. (The difference is
that Normans version speaks about reasons rather than about rightness.) As such, I have no quarrel with it. One of the attractions of the
notion of fittingness is that there is a comprehensible notion of partial
fittingness, in terms of which we can hope to explicate the notion of a
contributory reason. Modern intuitionism, which speaks as much of
reasons as of rightness, has cause to latch onto the notion of fittingness
with gratitude. It is with Normans first suggestion that I am here
concerned.
That first suggestion, as I said, had two parts. Its second part seems
to me to be innocuous. One could understand reasons as favourers
while still insisting that favouring is always comparative: a reason is a
consideration that favours doing one thing rather than doing another,
or than any other. What really concerns me is the suggestion that reasons should be thought of as considerations that make something better, whether this is better than it otherwise would be or than something
else is. This seems to me to be a positive mistake.
In Ethics Without Principles I drew a distinction between two normative relations, the relation of favouring, or being a reason for, and
the relation of making right, the right-making relation. Effectively the
same distinction can be drawn between the relation of favouring and
the good-making (or better-making) relation. These last two are distinct relationseven though there are obviously interesting connections between them. To favour is not the same as to make good. Why
not? First, there are reasons for belief, and it is not clear that they necessarily make it better to believe one thing rather than another. The question what should I believe here? does not seem to be the same as the
question what would it be better for me to believe here? Second, what
is favoured is an action. Favouring is a normative relation that holds
between the consideration that is the reason and an action of a certain
type. At this point we have to be careful. The favouring relation does
not stand between a consideration and a particular action. The consideration is particular enough, but what stands on the right-hand side of
the relation is doing an action of a more or less specific type. As far as
the reason is concerned, any action of that type is acceptable; the reason does not distinguish between them. But, further, the normative
relation can obtain even when no action of that type is in fact done. The
consideration that favours must be the case if it is to stand as a reason;
118
jonathan dancy
but the action for which it is a reason may not, for all that, be done. This
is just as well, because there are reasons against as well as reasons in
favour, and we would not want to understand all reasons against as
some-how ensuring that an action of the sort that they are against still
gets done.
There is a problem here, which was, as far as I know, first noticed by
H.A. Prichard, and which I call the Prichard problem. How exactly are
we to make sense of the idea of a relation that can be instantiated even
when only one of its relata in fact obtains (or occurs, or is the case)?
One way in which the Prichard point is important is that it reminds us
that choosing an action is not at all like choosing a chocolate. It is not
as if the possible alternative actions are laid out for us on a tray, in their
rich particularity. Only one of those possibilities is going to have a rich
particularity, and the others are not even going to exist.
It is important, of course, that in addressing the Prichard problem
we focus on the case before us. It might not be enough, for instance, to
appeal to intensional contexts, since it is not at all clear that reasonclaims are intensional. But my purpose here is not to solve the problem
but to use it to point out a difference between favouring and goodmaking. The Prichard problem does not really apply to the good-making relation, any more than it does to the right-making relation. One
can refrain from doing an action for the reason that it would be wrong.
This is properly expressed in terms of a subjunctive conditional. To do
it would have been wrong, and our reason for not doing it can be thatit
would have been wrong; we dont need to say that the reason was that
it was wrong. And we can take a similar approach to good-making.
It would have been better to have given the car keys to someone else,
given the amount of beer I consumed that evening. I didnt do so, and
so there was no action of the relevant type that was made better by the
amount of beer consumed. But that doesnt raise a Prichard problem,
because the subjunctive reading (it would have been better ) is perfectly satisfactory. Things are different with the favouring relation. It is
not that an action of the relevant type would be favoured if one were to
do it. Doing an action of that type is already favoured, even if one
doesnt do it. Otherwise we would be restricted to saying that there
would have been some reason to do it if we had done it. This is not the
same as saying that if we had done it, we would have done something
that there was reason to do.
There are other things to be said about the difference between
favouring and good-making, some of which I tried to say in Ethics
Without Principles. I was discussing there the relation between favouring and right-making (or ought-making)and in my discussion
I made some mistakes, which I hope here to correct. For a while, therefore, my discussion will focus on right-making rather than on goodmaking.8 What stands on the right-hand side of the favouring relation
is an action (or a belief, or a feeling), the action favoured by the reason.
What stands on the right-hand side of the right-making relation? In
Ethics Without Principles (pp. 2223) I suggested that it is a judgement,
the judgement that something is right. But that suggestion involves a
cross-categorization. The contrast I am after is between two substantial
relations, the relation of favouring and the relation of right-making.
Though there will be points to be made about judgement, because
the rightness made is a proper object of judgement while the action
favoured is not, those points are consequent on the distinction between
two relations, which itself needs to be made in other ways. So: what
stands on the right-hand side of the right-making relation? It looks as
if it is the rightness made (though of course it is an action, belief or feeling that has that rightness, in the sort of cases we are thinking about).
And if so, we do indeed have a contrast; one relation has an action as
one of its relata where the other has a property. But here again I am
uneasy. Surely what is favoured is an action, and what is made right is
an action; what is the difference here? We might say that what is made
is the rightness of an action. But couldnt we equally say that favouring
is making favoured, so that our distinction can only be between making
right and making favoured? That would not be much of a distinction.
I had wanted to say something like this, that the favouring relation
is intrinsically normative; it is the relation of calling for. The rightmaking relation is normative only because of the normativity of the
rightness made. Rightness-making, after all, is just one form of the relation of making-it-the-case-that, which is not itself normative at all; it is
the form in which what is made the case is that the action is right. But
couldnt one equally think of favouring as making it the case that the
action is favoured, so that neither relation is more intrinsically normative than the other?
Things do not seem to be going well for the distinction at the
moment. But there is something salvageable here. There is nothing
more to right-making than making it the case that the action is right.
I am grateful here to Douglas Farland for various salutary corrections.
120
jonathan dancy
That side of the story is sound, and so is the claim that any distinctive
normativity here derives from the normativity of what is made. But
though it is true that where there is favouring, what favours makes it
the case that what it favours is favoured, this does not yet establish a
perfect parallel. It is true that by loving Sarah, I make it the case that
Sarah is loved, and loved by me. But that does nothing at all to show
that loving is just one form of the general making-it-the-case-that relation. It is not. It is a relation whose right-hand side is a person (let us
assume). And I want to say something similar about favouring: it is the
relation of calling for, and calling for something is not the same relation
as that of making it the case that something is called for.
Further, the concept of right-making assumes as given the concept
of rightness, and that concept is not (in the relevant sense) relational;
that is to say, if rightness is a relational concept, this is not because
whatever is right is made right by something. We understand this form
of the making-it-the-case-that relation by appeal to an independent
understanding of what it is that is made. Things are the other way
round with the favouring relation. We understand what it is to make it
the case that something is favoured in terms of a prior and essentially
relational notion of favouring, just as we understand what it is to make
it the case that someone is loved in terms of a prior and essentially relational notion of loving.
Understood in this way, we can say that what lies on the right-hand
side of the right-making relation is a proper object of belief, something
capable of being the case, while what lies on the right-hand side of the
favouring relation is an action, which is not a proper object of belief but
something capable of being done. A second difference is that the righthand side of the favouring relation is something that is not capable of
being the conclusion of an inference. One can conclude an inference by
doing something, but one cannot infer an action from anything. The
right-hand side of the right-making relation seems, by contrast, well
set up to be the conclusion of an inference. In this sense the favouring
relation is more essentially practical than is the right-making relation.
There is a spin-off benefit of these thoughts. It seems to me that
Christine Korsgaards complaints about what she calls substantive realism derive largely from failing to see the difference between the two
relations. She writes, for instance:
According to substantive realism, then, ethics is really a theoretical subject. When we ask ethical questions, or practical normative questions
more generally, there is something about the world that we are trying to

find out. The world contains a realm of inherently normative entities or
truths, whose existence we have noticed, and the business of ethics is to
investigate them further.9
And later we read:

The substantive realist assumes that we have normative concepts because
we are aware that the world contains normative phenomena, or is characterized by normative facts, and we are inspired by that awareness to
construct theories about them.
But that is not why we have normative concepts. The very enterprise
we are engaged in right now shows why we have those: it is because we
have to figure out what to believe and what to do. Normative concepts
exist because human beings have normative problems.10
These complaints seem to me to be blunted by the fact that Korsgaards

substantive realist is perfectly capable of thinking that the basic normative relation is the favouring relation, and that the tracking of that is not
an interest in which normative facts are made the case, but in how to
act or in what to believe. The practical focus of that interest is not
undermined by supposing that there is the making-right relation in
addition to the favouring relation; adding this second relation does
nothing to turn ethics from a practical enquiry into a theoretical one.
Finally, what is true of the right-making relation seems equally true
of the good-making one. So my eventual conclusion is that Norman
has conflated two relations, the favouring relation and the goodmaking (or better-making) relation. These relations are indeed related,
but they are distinct.
Now even if there are these two normative relations, favouring and
good-making, it is quite possible for the very same thing, feature, or
consideration, to stand on the left-hand side of both of them at once. So
a favourer can be a good-maker, and often will be. But this does nothing to obscure the distinction between the two relations. And of course
it remains possible that something be favoured (a belief, say) without
being thereby, or therefore, rendered good.
Now consider deliberationof which the proper conclusion is an
action. To say that the favouring relation is the good-making relation is
to suppose, I think, that every deliberation must conclude in a thought
about somethings being good (or better, or best), and that our actual
9
C. Korsgaard, The Sources of Normativity (Cambridge: Cambridge University
Press, 1996), p. 44.
10
Korsgaard, The Sources of Normativity, p. 46.
122
jonathan dancy
decision is a further stage, not a proper part of the deliberation at all. It

is as if one cannot pass to action except via a thought about goodness.
But this seems to me unnecessary. Why cannot one simply say to oneself I promised to do it; so Ill do it? Why is one really to be thought of
as saying I promised to do it; so doing it would be good; so Ill do it?
In doing this, one is tracking the favouring relation by moving to the
thing that is favoured by what has gone beforewhich is not the same
as moving to the conclusion that the action is right or best. I think that
the insertion of the evaluative belief is a redundant product of bad
theory.
Suppose that this is right. There then arises an interesting question,
which concerns the relation between this conclusion and the buckpassing view that to be right is to be the action which one has most
moral reason to do.11 I am strongly attracted to this view. But if to be
right is to be most strongly favoured, what will happen to the distinction between two normative relations? Will it collapse? I am not sure
that it will. The right-hand side of the favouring relation is an action,
not favouredness. That is the right-hand side of a different relation, the
favoured-making relation. There is a close relation between these relations, as there is between the relation of loving and the loved-making
relation. But they are not the same. If the favoured-making relation and
the right-making relation are tightly interconnected, as they would if to
be right were to be most favoured, this would do nothing to unsettle
our original distinction between two normative relations.
11
I call this a buck-passing view by analogy with the now standard buck-passing
view that to be good is to have reason-giving features. The view in the main text passes
the deontic buck; the standard view passes the evaluative buck. For all this, see my
Should We Pass the Buck?, in A. OHear (ed.), The Good, the True and the Beautiful
(Cambridge: Cambridge University Press, 2000), pp. 15973.
part III
moral realism
Moral and Theological Realism: The Explanatory

Argument
Russ Shafer-Landau*
I am very interested in the possible connections between theism and
moral realism. Not because I am a theist, but because I am not.
As I will understand the view here, moral realism stands for the idea
that there are some moral claims that are true in a certain way. Their
truth does not depend on the attitudes that anyone takes towards their
content. Nor are they true, when they are, because of being endorsed,
implied or entailed by norms that are constructed from our evaluative
attitudes.
Theological realism, as I will understand the term here, is simply
theism: the view that God exists. The God I will be talking about is the
traditional God of Western monotheism: an omniscient, omnipotent,
and morally perfect agent.
Among philosophers, it is a common thought that moral and theological realism can easily be prised apart. Most analytical philosophers
these days are agnostics or atheists, and so reject theism. But even those
who are unkindly disposed towards moral realism do not think that its
vulnerability lies in a commitment to theism.
The story among non-philosophers is quite different. In that arena,
its a common thought that the status of morality and religion are very
closely connected. On a popular view, morality can be objective only if
God exists. Thats a central reason why atheism is taken, by so many, to
be such a threat to morality.
We are all familiar with the standard lines of popular thought that
seek to tie the fate of moral and theological realism very closely
together. Ill take the liberty of placing some fancy philosophical terms
* I appreciate the help I have received in working on this paper as it was given in the
form of talks at Oxford University, Cambridge University, the University of Colorado,
and the workshop on Realism and its Critics at Hamilton College. I would especially
like to thank Krister Bykvist, Geoff Ferrari, Alison Hills, Kinch Hoekstra, Brad Majors,
Graham Oddie, Jonas Olson, Charles Pigden, Carolina Sartorio, Elliott Sober and
Ralph Wedgwood for their generous comments in helping to improve earlier, rather
terrible, drafts of this paper.
126
russ shafer-landau
in the mouths of the men on the Clapham omnibus, so as to more precisely capture popular thinking in this area. In short order, the most
influential of these arguments are the following:
1.If morality is objective, then it cannot be the result of human
creation.
2.If morality cannot be the result of human creation, then it must be
authored by God.
3. Therefore if morality is objective, then it must be authored by God.
1.Genuine moral requirements entail categorical reasons of obedience.
2.Categorical reasons are possible only if underwritten by God.
3.Therefore genuine moral requirements presuppose Gods existence.
1.Genuine moral requirements must exert a reliable motivational
influence on those who are bound by them.
2.Moral requirements can exert such an influence only if they are
backed by divine sanction.
3.Therefore genuine moral requirements must be backed by divine
sanctions.
4. Therefore genuine moral requirements exist only if God exists.
Each of these arguments contains at least one premise that expresses a
philosophical claim accepted by most philosophers today. Still, I think
that each argument is unsound. And so do most other philosophers.
Given that, it may seem that theres little here worth discussing. The
received view nowadays is that the fate of moral realism is not tied to
that of theological realism, and so we can entirely leave aside matters in
the philosophy of religion when doing metaethics. I am not so sure.
Thats not because I am worried that objective morality may after all
require Gods existence. Rather, I am concerned about the pressure that
arguments for religious skepticism place on arguments favoring moral
realism. For the considerations in favor of such skepticism also seem to
cast strong doubt on the merits of moral realism.
There are threereally, fourarguments in this connection that
seem especially worrisome. I have space here to discuss only the last of
these, but I want to sketch them each so as to give a fair impression of
the scope of the relevant concerns.
The first argument is a form of genealogical critique. It does not seek
to vindicate atheism, or moral nihilism, but rather to show that any
positive theistic or moral belief is epistemically unjustified. On such a
moral and theological realism127
view the origins of our beliefs in a given area are directly relevant to
their epistemic merits. Most people hold the religious or moral beliefs
they do because of the way that they have been raised, or because of the
company they now keep. But there is no reason to suppose that such
influences track whatever truth there might be in these domains, especially since such influences have led people to contradictory beliefs. If
there are moral or religious facts, then given how our beliefs in these
areas have arisen, it would be sheer luck were they to land on the truth.1
But such luck vitiates the epistemic credentials of the associated beliefs.
Thus even if there is a God, and even if there is a set of strongly objective moral truths, our beliefs about them are epistemically unjustified.
The second argument is one from disagreement. The argument
really, a pair of argumentsbegins with the observation that the
breadth and depth of moral and religious disagreement is far greater
than that found within the natural and mathematical sciences. The first
version of this argument seeks to draw an ontological point from this
observation. On this line, the best explanation of such disagreement is
that there is no objective reality awaiting our discovery.2 Theists and
moralists disagree so often and so intently because they are projecting
their conflicting commitments or sentiments on a world that contains
no god, and no moral facts. The alternative accountthat half of those
involved in such disagreements have faulty belief-forming mechanisms
which stand in the way of their appreciating realityis both less parsimonious, and requires the presentation of a defensible religious and
moral epistemology, neither of which, it is claimed, is forthcoming.
The second version of the argument from disagreement is more
modest, and claims that even if there is a god, or objective moral facts,
the widespread disagreement on moral and religious matters undermines any justification we might have for our moral and religious
beliefs.
The last of the arguments says that we have most reason to deny
Gods existence, because God is explanatorily superfluous. God must
go the way of the ether, and the Aztec divinities. We havent any need of
1
The argument in the moral domain has been pressed in especially forceful ways by
both Sharon Street and Richard Joyce. See Street, A Darwinian Dilemma for Realist
Theories of Value, Philosophical Studies 127 (2006), pp. 10966, and Joyce, The Evolution
of Morality (Boston: MIT, 2006), ch. 6.
2
See J.L. Mackies so-called argument from relativity for a now-classic statement of
this worry, in his Ethics: Inventing Right and Wrong (New York: Penguin, 1977),
pp. 3840.
128
russ shafer-landau
these things to explain our encounters with the world. Ditto for God
all that he was once invoked to explain can nowadays be more simply
accounted for by reference to other, wholly natural phenomena. Belief
in God is therefore epistemically unjustified; further, belief that God
does not exist is highly justified.
The parallel with moral facts, realistically construed, is straightforward. Moral facts are unable to explain our observations of the way the
world works; an actions being morally required, or there being a moral
reason opposing some policy, does not explain why anything occurs as
it does. Moral facts are certainly unable to explain the goings-on of the
inanimate world. And even where they have most promise to explain
the animate worldnamely, in revealing why people do what they
dowe can more parsimoniously account for such goings-on by citing
non-moral facts about human psychology. Since moral facts are
explanatorily superfluous, therefore, we have most reason to believe
that they do not exist.
Lets begin with a more careful presentation of this last argument,
which will be the focus of the ensuing discussion. It is a variation on a
familiar argument, owing to Gilbert Harman,3 that the explanatory dispensability of moral facts entitles us to deny their existence.
Here is an initial go:
Something exists only if it is required in the best explanation of what we
believe, do or observe.
Neither religious nor moral facts are thus required.
Therefore there are no religious or moral facts.
We neednt consider the second premise just yet, for the argument is
unsound even if it is true. The first premise is false. The worlds contents
are not limited by what manages to satisfy our explanatory needs. What
serves such needs is a function of our interests and our ignorance. But
the true contents of our world are not fixed by such factors. Much of
what exists in the universe fails to figure in our best explanations, either
because we are ignorant of it, or ignorant of or uninterested in the
explananda that it would account for.
It is better, then, to conceive of the argument as having an epistemological, rather than an ontological, point. Thus conceived, it would go
something like this:
3
See his The Nature of Morality (Oxford: Oxford University Press, 1977), pp. 310,
and his Moral Explanation and Natural FactsCan Moral Claims be Tested against
Moral Reality?, Southern Journal of Philosophy, suppl. 24 (1985).
1.We have reason to believe that something exists only if it is required

in the best explanation of the events that we undertake or experience.
2.Neither religious nor moral facts are thus required.
3. Therefore we lack reason to believe that there are such facts.
This argument, if sound, would suffice to warrant a form of
Pyrrhonian skepticism about moral and religious facts. Such skepticism claims that our religious and moral beliefs lack justification, and
that agents are unjustified in holding such beliefs. But many critics seek
to vindicate a stronger conclusion, namely, that we not only lack justification for such beliefs, but possess justification for thinking them
false. Call this stronger view Academic skepticism.
Academic skepticism entails the Pyrrhonian variety, but not vice
versa. One may lack justification for a belief without also having justification for thinking it false. I am not justified in believing that it will
rain in Monte-video on 3 June 2020. But neither am I justified in thinking that it wont.
To vindicate the stronger, Academic form of skepticism about religious and moral beliefs, we would need a different argument from the
one weve just seen. Perhaps something like this:
4.If anything that x is invoked to explain can be better explained without positing xs existence, then we have reason to deny xs existence.
5.Anything that divine or moral facts are invoked to explain can be
better explained without positing their existence.
6.Therefore we have reason to deny the existence of divine and moral
facts.
This argument, even if sound, is not yet enough to vindicate
Academic skepticism about moral and religious facts. Look at its conclusion. It says only that there is reason to deny the existence of such
facts. It doesnt say that such a reason is indefeasible, or even the best
available reason.
To get Academic skepticism, we need both of the arguments just
considered. We can combine them to get the strong form of skepticism
about religious beliefs that is so widespread in todays philosophical
community, and the strong form of moral skepticism that is less pervasive, but no less threatening for that. The argument would take this
form:
7.If there is no reason to believe p, and some reason to deny p, then
there is most reason to deny p.
130
russ shafer-landau
8.There is no reason to believe in moral or divine facts, and some reason to deny their existence.
9.Therefore there is most reason to deny the existence of divine and
moral facts.
I hope it is clear by implication that the reasons throughout are epistemic reasons, and not other types, such as prudential ones. If it isnt,
let me make it explicit now.
My strategy here will be to present what I consider to be the best
replies on behalf of the moral realist, and then see whether these replies
allow the theist to resist this extended chain of reasoning. Call that
chain, from (1) to (9), the Explanatory Argument.
Ill spare you a detailed examination of each premise, and instead
sort all moral realist responses into two large groups: that of the
naturalist, and that of the nonnaturalist. Naturalists have a ready
reply to these arguments. They will accept the explanatory requirements asserted in premises (1) and (4), but claim that moral facts can
meet them. By the naturalists lights, moral facts are ordinary, run-ofthe-mill, natural facts, i.e., empirically discernible facts. Such factsas,
for instance, that a great deal of pleasure is generated by an
actionmay indeed play an explanatory role in accounting for our
beliefs and experiences. So, on the naturalist line, premises (2) and (5)
would be false. That would explain why we can resist these skeptical
arguments.
Thus if moral naturalism is true, then moral realists will have a satisfactory reply to the Explanatory Argument. Since the reply relies on
invoking moral naturalism in defense of moral realism, a parallel route
is unavailable to the theist. For that parallel route would be to enlist
theological naturalism in defense of theological realism, and that
ongoing, Western conceptions of Godjust doesnt make sense.
Theological naturalism would seek to vindicate the explanatory necessity of divine facts by revealing them to be a species of natural facts. But
that simply runs counter to the essence of the monotheistic tradition,
which depicts God and his attributes as supernatural.
Theists might nevertheless accept the explanatory requirements of
(1) and (4), and target the negative claims of (2) and (5). Theists cannot
implement this strategy by invoking naturalism. If theists accept the
explanatory requirements, they will have to show that natural explanations are incomplete, and that they require supplementation from the
supernatural for their success. I think that most of our naturalistic
explanations are in fact incomplete. But that gives no reason to suppose
that they require such supplementation to adequately round them out.

Making such a case is a very tall order.
In any event, I am no moral naturalist, and so even if what Ive just
argued is on the mark, the prospects for parity between my sort of
moral realism and theological realism seems much more likely.
The success of the naturalist reply is predicated on accepting the
explanatory requirements advanced in premises (1) and (4). I am
unsure of whether moral facts can pass such tests. If the relevant
explananda include such things as being a virtuous person, or being
non-institutionally deserving of praise or blame, then moral facts can
surely pass the tests. But as I understand their use, these tests are meant
to determine, from neutral starting points, the contents of our ontology.
If we just assume the existence of moral explananda, then it will come
as no surprise that moral facts are required to best explain them. But
this surely is a victory too easily gained. If explanatory tests are to be
useful in determining whether classes of entities or facts exist, then
itseems that we cannot assume the existence of that class in applying
the tests.
To reinforce these points, consider the relevant parallel with theism.
Suppose that the theist claimed that God was necessary to best explain
various religious factsthat the world is imbued with immaterial souls
that are innately directed to divine purposes, that there is a heaven and
a hell, that the Christian scriptures are inerrant. If we start with the
assumption that there are such facts, it will be unsurprising to find
divine explanations needed to account for them. But those who base
their religious doubts at least partly on a version of the Explanatory
Argument would be right to claim that such an assumption begs the
question.
Things are just the same when it comes to moral matters. A concession that Harman made in framing his discussionthat even if there
are moral facts, they are explanatorily superfluouscrucially undermines his case. Once we grant the existence of moral facts, we can
always ask how they are to be explained, and given the existence of an
is-ought gap of some kind, it wont come as a surprise to find other
moral facts needed to do the relevant explaining.
So suppose we begin our inquiry without the assumption that there
are moral facts. Can we show that they are needed to explain what we
believe or experience? Some have thought so. We might think, for
instance, that the fact that two and two are four best explains why we
believe that they are, and that things are no different when it comes to
very widely endorsed, practically unshakeable moral beliefs, of the
132
russ shafer-landau
sort, say, that condemn genocide and slavery. On this line, those who
are appropriately sensitive to mathematical or moral reality can explain
the event of their believing, as well as the content of their beliefs, by
citing the mathematical or moral facts that have caused these doxastic
events and contents.
This may be true. But to show it so, we would have to show that alternative explanations are not as good. And the problem, at least in the
moral case, is that there are quite plausible alternative explanations
(those offered by error theorists and expressivists), ones that do not
invoke a moral reality that agents can accurately appreciate. Why
require moral truths to best explain our moral views, if we can cite the
social, parental and psychological factors that appear to so heavily
influence their content? This is the essence of Harmans challenge.
There may be a good answer to it. After all, we do need to go beyond
such factors to best explain why physicists hold the physical views they
do, biologists the evolutionary views they do, etc. But it seems that we
are licensed to move beyond purely genetic accounts of doxastic practices in the natural sciences because the phenomena under study possess recognized causal powers. We cannot assume that moral facts
enjoy the same status without begging the question in this context.
So at this juncture we are not in a position to vindicate the causal
efficacy of moral facts by claiming that they play an ineliminable role in
our holding the moral beliefs we do. We can successfully defend such
an argument only by revealing the flaws in anti-realist accounts. I dont
think that there is any short way to do this.
Consider, then, a different line of argument, meant to demonstrate
the causal efficacy of moral facts, that in its essentials can be run in a
way that is compatible with either moral naturalism or nonnaturalism.
The basic idea is this.4 Many nonmoral facts are counterfactually
dependent on moral facts. If certain moral facts had not obtained,
other nonmoral facts would not have obtained. This dependence may,
with qualifications, suffice to establish genuine causal power. Since
moral facts can often pass such a test, moral facts may possess genuine
causal power.
4
See David Brink, Moral Realism and the Foundations of Ethics (Cambridge:
Cambridge University Press, 1989), pp. 19296; Nicholas Sturgeon, Moral
Explanations, in D. Copp and D. Zimmerman, Morality, Reason and Truth (Totowa,
NJ: Rowman and Allanheld, 1985), and Sturgeon, Contents and Causes: A Reply to
Blackburn, Philosophical Studies 61 (1991), pp. 1937; Brad Majors, Moral Explanation,
Philosophy Compass 1 (2006), pp. 910; Graham Oddie, Value, Reality and Desire
(Oxford: Clarendon Press 2005), pp. 198203.
As all proponents of this strategy have acknowledged, counterfactual dependence has its limitations as a sufficient condition of causal
efficacy. Relata with common causes, and instances of backtracking,
pass the test but rarely reveal genuine causal relations. Exceptions can
also arise when the test is used as a necessary condition for causal
power. It fails, for instance, in cases of preemption and overdetermination. But barring aberrant cases, a counterfactual testhad X not
occurred, Y would not have occurred as it didis a reliable measure of
somethings causal (and hence explanatory) power. Lets proceed on
this assumption and see where it gets us.
Suppose that we invoke a moral fact to explain a nonmoral fact. The
employee pension fund is now drained; what accounts for this? It can
be perfectly natural to cite the venality, greed and moral corruption of
the corporate executives who perpetrated the fraud. And we might say
that we should respect the appearances, until we have good reason to
doubt them.
Unsurprisingly, there is a reason to doubt the appearances here, and
to deny that the citation of moral facts in everyday causal talk reveals
anything about the real contents and structure of the world. The reason
is that moral facts, by all accounts, supervene on causally efficacious
nonmoral ones, and since we should be loath to countenance widespread causal overdetermination, it is (say the critics) really the subvening nonmoral facts that are doing the causing, rather than the moral
facts that supervene on them.
That might be so. But there is reason to resist the criticism just
described. We are assuming that in standard cases, X causes Y if and
only if Y would have been different had X been different.5 Not only do
moral facts often pass this test, but, importantly, the nonmoral facts on
which they supervene often fail it. If we kept everything else fixed, but
It is important to note that almost every standard case is one that involves
supervenient phenomena; the only ones that dont are cases of causation at the fundamental physical level (assuming there is such a thing). All of the facts we inquire
aboutunless we are physicistsare supervenient facts. So we shouldnt seek to discount the applicability of the counterfactual test in the moral case just because moral
facts are supervenient ones; motivational, intentional, social, economic, biological, historical, political (etc., etc.) facts are all supervenient, too. It might, of course, be that all
moral cases are non-standard, but it is hard to see how to defend such a claim in a
non-ad hoc way. Or it might be that no supervenient phenomena really existthat
independent causal power is a prerequisite for ontological credibility, that supervenient phenomena lack such power, and that they therefore ought to be expunged from
the ontology. But if that were so, then the absence of moral facts would be no surprise,
given the absence of any people, actions, intentions or motivesall supervenient phenomena themselves.
5
134
russ shafer-landau
assumed that the executives were not greedy, venal and corrupt, then
the outcome would have been quite differentthe pension fund would
presently be well-endowed. So these moral failings do seem to be causally relevant. But the pension fund would remain depleted in the nearest natural possible world, one in which the specific natural facts that
constituted the corruption were changed only very slightly. This argues
for the greater causal-explanatory power of moral facts over certain
nonmoral ones.
Since moral properties are multiply realizable, there is a good deal of
variation in the make-up of the natural facts that can instantiate a given
moral property. This variability is what explains why citation of exclusively natural features often fails to satisfy a counter-factual test of
causal efficacya small change to the subvening facts, one that brings
us to the closest natural possible world, would not alter the relevant
outcomes in any way. So if we have to choose between these options in
fixing causal power, then whenever (keeping all else the same) alteration of moral facts would, and change of subvening natural facts would
not, yield a different outcome, we should be willing to credit the moral
facts with enough causal power to allow them to retain a place in our
ontology. After all, we extend the same courtesy to many other types of
supervenient facts, on precisely the same grounds.
There is a worry, of course. It was expressed a good while back, by
David Zimmerman,6 and his critique remains, I think, the best expression of doubt about this sort of strategy. His criticism is that the greater
generality achievable by adverting to supervening facts in causalexplanatory contexts is a double-edged sword. Rather than explain
(say) Hitlers deeds by reference to his depravity or evil nature, we
might more informatively account for his actions by citing the specific
nonmoral intentions and motives that prompted them. There are many
ways of being evil and depraved, some of which Hitler did not exemplify, and so some of which are irrelevant to accounting for his actions.
We therefore gain greater understanding of his behavior if we cite the
more specific nonmoral facts that realized the evil and depravity at particular times and places. In this way we can always replace a moral
explanation with a nonmoral one in order to gain greater understanding of the phenomenon in question.
6
David Zimmerman, Moral Realism and Explanatory Necessity, in D. Copp and D.
Zimmerman (eds.), Morality, Reason and Truth (Totowa, NJ: Rowman and Littlefield,
1985), pp. 79103, at p. 87.
Suppose that we must concede Zimmermans claim that nonmoral

facts do more in the service of perspicuous explanation than moral
facts can do. Still, we might then claim that causal power is enough to
secure a place in the ontology, even if what is causally efficacious is not
maximally explanatorily efficacious. Zimmerman could be right that
mention of finer-grained non-moral facts will always be more illuminating than reference to coarser-grained moral facts. But even if he is,
that does not undermine the (sometime) greater ability of moral facts
to pass a counter-factual test of causal relevance. And passage of such a
test is what really counts, for causal power is thought by all parties to
this debate to signify real ontological credibility, whereas explanatory
illumination is always relative to our interests and ignorance. For this
reason, if causal and explanatory power come apart, there does seem to
be a good basis for sticking with causal power as a means of discerning
the proper contents of our ontology.
Yet there remains a serious problem for the moral realist, and the
problem is simply put: passing the causal test is insufficient for establishing realistic status. Legal facts, for instance, can pass the test: had
traveling at a certain speed not been against the law, the police officer
wouldnt have pulled the driver over; had the conduct not been criminal, the defendant wouldnt now be behind bars; had the contract not
been validly drawn, specific performance would not now be required.
The appeal to legal categories can do genuine explanatory work, and
often because legal facts causally account for why things happen in our
world. But legal facts are conventional,7 and arent apt for a realistic
construal.
There isnt anything special about the legal case. We could make the
same point about other kinds of conventional facts. Had the joke not
been funny, the audience wouldnt have laughed. Had the dress not
been unfashionable, its owner wouldnt have been ridiculed. Had the
speaker not been rude, he wouldnt have been jeered at. Being humorous, fashionable or rude are all conventional matters, depending essentially on our contingent attitudes to features of our world. Sartorial
realism is implausible; realism about humor or etiquette no less so.
7
For those natural lawyers out there, we can restrict the discussion to legal facts
(such as those that enact a speed limit) that everyone acknowledges to be conventional,
and leave aside the controversial matter of whether moral constraints (realistically construed) play an ineliminable role in fixing the identity conditions of either a legal system or specific legal standards.
136
russ shafer-landau
And yet there are facts of these kinds that cause and explain things in
our world. Thus reliance on the causal test is insufficient to vindicate a
realistic understanding of the facts or entities that pass it.
I dont think that this points to a special problem for the counterfactual test. It, or something very like it, seems essential to our understanding of causal efficacy. And I dont see any natural amendment to
the test that would enable us, through its use alone, to discriminate
between facts that are best construed realistically, and those that are
not. That is because many conventional facts seem, in the examples
above, and in so many others, to be causally relevant to outcomes in
our world.8
Where does this leave us? We reached this point by casting doubt on
the tenability of the strategy that had us first assuming the existence of
moral facts, and then asking after their explanatory efficacy. Such a
strategy, endorsed by Harman, makes it far too easy to vindicate the
explanatory necessity of moral facts, as becomes clear when we consider the analogous case that might be made for theistic facts. Once we
grant the existence of the disputed class of entities or facts (theistic,
moral, etc.), and make them the relevant explananda, it wont be difficult to establish the need to invoke other facts of the same kind in
accounting for them.
So perhaps we can, without assuming the existence of the disputed
class, defend its existence by means of a counterfactual test of causal
efficacy. Yet as we have seen, though we may be able to do this for moral
facts, that does not suffice to vindicate their realistic status. Those
moral realists who rely on the test to underwrite their position thus
need to look elsewhere.
This isnt to say that the counterfactual test, or a near-cousin, is irrelevant in these debates. It might be, for instance, that counterfactual
dependence ordinarily supplies a necessary, but not sufficient, condition for somethings earning its way into the ontology. If that were so,
then passage of the test would be insufficient to vindicate moral realism, but failure would suffice to refute it.
8
One could deny that conventional facts are causally efficacious, and so claim that
we have good reason to regard only nonconventional facts that pass the causal test as
best construed realistically. Since moral facts are nonconventional, and pass the causal
test, we therefore have good reason to think that moral realism is true. The problem
with this is that we cant simply assume that moral facts are nonconventional. Passage
of the causal test was supposed to show that they are. But it cant do that, as the examples from the law and etiquette have shown.
If we insist on causal relevance as a necessary condition of ontological credibility, then a good deal more must be done to more precisely
fix the content of a causal test, and to show that moral facts are (un)able
to pass it. Those who enjoy strenuous exercise are encouraged to take
this path. But I think there is good reason to sidestep it.
If we proceed on a neutral basis, and assume neither that moral facts
exist, nor that they dont, then it is as yet unclear whether they are
explanatorily indispensable. If they are, then moral realism is so far in
the clear. But I want to proceed here on the assumption that we can do
all of the relevant explaining without invoking moral facts. My goal is
to show that even on this worst case scenario, the moral realist has a
defense against the Explanatory Argument that is largely successful.
This defense, I think, is unavailable to the theist. And so the parity that
is asserted at the heart of the Explanatory Argument can be resisted.
If we grant, for purposes of argument, that moral facts will fail the
explanatory tests, then nonnaturalists are faced with a real worry. For
theres no doubt that we have used these explanatory tests to very good
effect in accounting for why were no longer warranted (if ever we were)
in believing in ghosts, UFOs, or snake gods. I think, in fact, that the
explanatory tests are excellent devices for assessing the epistemic merits of our ontological commitments in a very broad, but not unlimited,
way. Specifically, their good work is limited to the natural realm, the
one whose contents are (at least in principle) empirically discernible.
How convenient for me! But there is a case to be made for such a
limitation. For the explanatory requirements, if strong enough to eliminate moral facts, will almost certainly eliminate all normative facts. If
we assume that moral facts lack explanatory power, there is every reason to make a similar assumption regarding all normative facts. That
Iought to believe that So Paulo is in Brazil would not explain why I do.
Testimony, and my past acquaintance with maps of South America,
would explain my belief. That I would be irrational for failing to dress
warmly in the snowy season would not explain why I have done so.
Rather, my wanting to stay warm, and knowing how to do it, could do
the needed explaining. That I am morally required to support my children would not explain why I do. That I believe that I am thus required,
in addition to the fact of my direct love for my children, and my ability
to support them, would together explain my support. As a general matter, its quite implausible to suppose that normative facts are going to
directly account for non-human events. But even when it comes to
human agency, there are explanations in terms of our psychological
138
russ shafer-landau
profiles and opportunities for action that might be capable of doing all
the needed explaining.
Normative properties are individuated by the standards they set, the
reasons they represent. Predictive or causal-explanatory failure is no
strike against the credibility of a normative standardthat people dont
adhere to it is not, by itself, good evidence that its content is mistaken.
Now these considerations, even if correct, do not amount to an argument for jettisoning moral and other normative facts from our ontology. We get that result only if we supplement these considerations with
the explanatory requirements in (1) and (4). Since I want to argue for
our justified belief in moral facts, I have to showagain, on the assumption that moral facts dont pass the explanatory requirementsthat
these requirements dont apply to them. I think that there is a direct
way to do this:
10.If the explanatory requirements (i.e., premises (1) and (4)) are both
correct, and perfectly general in application, then we have most
reason to deny the existence of all normative facts.
11.We dont have most reason to deny the existence of all normative
facts.
12.Therefore the explanatory requirements are either false, or, if true,
are limited in their application.
Let us proceed by assuming the truth of (10), which represents a
generalization of the concession I have been making for purposes of
argument. Whatever grounds there are for thinking that moral facts are
explanatorily dispensable will, so far as I can tell, generalize to all normative facts. If (10) is false, then whatever explains its falsity will also,
presumably, justify the explanatory necessity of moral facts. And while
(per our earlier discussion) this is not enough to vindicate moral realism, it is enough to insulate moral facts from the Explanatory Argument.
So anti-realists can have no quarrel with the concessive assumption
that (10) is true.
In support of (11), consider first some hopefully uncontroversial
examples: it is a fact that it is sometimes appropriate or legitimate or
justified to pursue ones happiness, a fact that one ought to believe in
ones own existence, a fact that one often should pursue efficient means
to ones chosen ends. More importantly, we can show that (11) is plausible by showing how implausible its negation is. The negation of (11)
that we have most reason to deny the existence of all normative
factsis, if true, a normative fact. So, by its own lights, we would have
most reason to deny it. If we have most reason to deny the negation of
(11), then we have excellent reason to affirm (11).
In other words, the explanatory requirements, and the conclusions
they are meant to help establish, are all themselves normative claims.
They tell us what we have reason to do or believe. But if we deny the
existence of all normative facts, then the conclusions of the skeptical
arguments are ones that we have most reason to deny. Since I am
assuming that the proponents of those skeptical arguments would find
such a view unacceptable, even they should accept the truth of (11).
Here is where we stand. I have tried to vindicate (12), by granting to
anti-realists the truth of (10), and then arguing for (11). If I have succeeded, then we either have to say of the explanatory requirements that
they are false, or that they do not apply to the normative realm. I prefer
to say the latter, just because of the excellent work they have done in
justifying the rejection of beliefs in phlogiston, the Greek pantheon,
Easter bunnies, etc. In other words, I think we are right to apply the
explanatory requirements when seeking to determine the natural contents of our world.
And this provides a basis for drawing a disanalogy between the
moral and the divine. The job description of normative facts does not
include the possession of explanatory power. They may indeed possess
such power, though I am assuming here that they do not. But because
their functional role does not require that they explain nonnormative
phenomena, but rather that they specify ideals, requirements, or standards that in some way must be met, an explanatory failure does not
license their expulsion from the ontology. By contrast, the job description of God does include explanatory power. Indeed, such power is
causal-explanatory power, which, moreover, must sometimes be exercised. God must, at the least, get the universe going, and will further, to
all but deists, intervene in our affairs at least occasionally in such a way
as to vitiate the causal closure of the natural.
Yet it is far from clear that supernatural explanations are superior to
natural ones in accounting for the goings-on in our world. Supernatural
explanations are certainly less parsimonious than natural ones. When
a battle is won, for instance, we can say, if we are so inclined, that God
had a hand in the victory, but what we must say is that these particular
soldiers, acting in these particular ways, in these specific circumstances,
carried the day. Perhaps a divine hand was guiding the combat, but
there is no need to invoke such an explanation when the natural one is
apparently complete. The lesson is easily generalizable. Since theistic
140
russ shafer-landau
explanations are less parsimonious, if they are to be superior to natural

explanations they must be substantially better along some other explan
atory dimension. It isnt clear what that dimension would be.
The theist might at this point try adopting a companions-in-guilt
strategy, and argue that many kinds of things whose existence we
strongly affirm are such that their functional role includes causal efficacy, though they fail to be explanatorily indispensable. Mental states
are perhaps the most common example in this regard. There are mental
states. They do cause things. And yet they also seem to be replaceable
in any explanation they might appear in. We dont really need to cite
mental factsat least in principleto explain why we do what we do.
There is a complete neurophysical story that can do that, and we have
warranted confidence that such a story will some day, in all its details,
be available to us. Once the neurons do their job, there is nothing left
for a belief or a desire to do. So beliefs and desires, and other mental
states, are explanatorily superfluous. Therefore, by the lights of the
explanatory requirements, we have most reason to deny their existence. Since, I am assuming, this last conclusion is falsewe dont have
most reason to deny the existence of beliefs and desiressomething
must be wrong with the explanatory requirements. And so the pressure
they put on theism may be resisted.
This familiar line relies for its success on something like what Kim
called the causal inheritance principle.9 I wont pause here to comment
on the merits of the principle, except to say that if it can be vindicated,
then the moral realistwhether naturalist or nonnaturalistis very
likely on firm footing as regards the explanatory requirements. But the
theist is just as likely not. For the success of this theistic reply depends
on the plausibility of the causal inheritance principle, and that principle, if true, is restricted in its application to phenomena that supervene
on, or are identical to, naturalistic facts or states. But divine facts are
neither. They arent identical to mundane natural facts. Nor is God
nomologically, metaphysically or conceptually constrained by such
facts; the divine does not supervene on the natural. Divine explanations compete with natural ones. They do so because the divine is neither identical to, nor supervenient upon, the natural world.
This point is crucial in seeing why the moral realist and the theist are
not equally vulnerable to the Explanatory Argument. If moral facts are
9
Jaegwon Kim, Multiple Realization and the Metaphysics of Reduction, Philosophy
and Phenomenological Research 52 (1992), pp. 126, at p. 18.
identical to natural ones, then there is no competition between them in

accounting for the worlds events. If moral facts instead supervene on
natural ones, then if there is any competition, it is as benign as the competition between physical facts, and the mental, biological or chemical
facts that supervene on them. If we insist on the multiple realizability
of these latter kinds, regard them as real, and have a causal test of ontological credibility, then we will render one of two verdicts. Either
(i) there is causal competition between supervening and subvening
facts, and the supervening facts sometimes win out (as they do in a
straightforward application of the counterfactual test, e.g., in the pension fund case above), or (ii) there is no real competition between such
facts, and the genuine causal efficacy of one entails that of the other. So,
for instance, one might think that there is no real competition between
a causal explanation of a paper cut in half if we cite, on the one hand,
the scissors that did the cutting, and on the other, the concatenation of
particles that constitute the scissors.
But since divine facts are neither identical to natural ones, nor supervenient upon them, then there is real competition between explanations that invoke only natural facts, and those that invoke God and his
powers and attributes. If, to take an earlier example, a battles outcome
can be accounted for just in terms of natural facts about military
deployment, then not only is there no need to invoke divine explanation, but the completeness of the natural explanation entails the falsity
of the divine one. Thats not so in the case of supervenient phenomena;
explaining (say) someones death by citing a terminal disease does not
falsify an account that is couched at the molecular level. (For, as a general matter, an explanation given in terms of a supervenient domain
does not falsify one given at the subvening level, if the supervening
features are realized by the subvening features.)
What this shows is that theists cannot avail themselves of the companions in guilt strategy, and so must, after all, accept the plausibility of
the explanatory requirements, and proceed to show, if they can, that
divine facts manage to pass them. Whether they can succeed in this
task is, of course, too large an issue to be considered here.
As far as I can tell, there is just one way for the theist to resist this line
of reasoning, and that is to claim that knowledge of Gods existence can
be gained in a wholly a priori way. Thus even if our concept of God
included the realized ability to causally intervene in our affairs, we
would not have to be able to cite evidence of such interventions in
order to justify a belief in his existence. So, for instance, if a version of
142
russ shafer-landau
the ontological argument were sound, and we could know this, then
that would suffice to justify belief in Gods existence, even if we were
unable to show that divine intervention was the best explanation of
anything that we had experienced.
This strategy bears an obvious resemblance to the sorts of intuitionist epistemological strategies sometimes adopted by moral realists. The
crucial difference is that intuitionists are not claiming to gain knowledge of any existential truths in an a priori way. If there are any a priori
moral propositions, they are restricted to ones that express moral principles (which, as I understand them, are conditionals), rather than
moral facts (which, by my lights, are instantiations of moral properties). So, for instance, one might be able to know a priori that killing
solely for enjoyment is immoral, though one cannot know a priori
whether any given instance of killing is immoral. And that is because
one cannot know a priori whether moral properties have been instantiated. One must know something of the worlds contingencies in order
to have such knowledge. Thus moral facts are not knowable a priori,
even if some moral principles are.
So, although some moral realists are perfectly comfortable invoking
the possibility of conceptual truths and a priori knowledge in moral
matters, they deny that we can know a priori either (i) that there are
moral facts, or (ii) that the specific moral facts are as we take them to
be. We must have some experience of the world in order to know that
there are moral features in it. We must know, at the very least, that there
are beings with intentions and deliberative powers, vulnerable beings
who can suffer, etc. There can be no moral facts in the absence of moral
agents. And whether there are any moral agents about is not something
we can know a priori.
Compare the situation for the theist. It might be a conceptual truth,
and hence a priori knowable, that if there is a God, then God possesses
more knowledge than human beings do. But the theist needs more
than this, of course, since atheists can sign on to such principles. What
the theist needs is a justification for believing that God exists, and such
a justification must, if I am right, proceed in one of two ways. Either the
theist must show, a priori, that God exists, or the theist must show, a
posteriori, that God is explanatorily indispensable.
Both such argumentative burdens are different from those inherited
by the moral realist. Moral realists cannot show, a priori, that moral
facts exist. (Nor do they need to.) Nor does the realist have to show that
moral facts are essential in explaining why the nonmoral world is as it
is. Thus the argumentative burdens on the moral realist differ very
importantly from those of the theological realist.
This concludes my assessment of the explanatory argument that
seeks to establish parity between the vulnerability of moral and theological realism. The take-home message here is that there is an important disanalogy between the cases that can be made against moral facts
and theistic ones. The explanatory requirements are best construed as
limited in their application to entities whose job description includes
having a discernible impact on the natural world. Since moral facts do
not, but many divine facts do, have it as part of their job description to
causally intervene in that world, moral facts are excluded from the purview of the requirements. Divine facts are not. And therefore divine
facts must earn their way into our ontology by passing the explanatory
requirements; their failure to do so would (absent a sound a priori
argument for Gods existence) provide us most reason to deny their
existence. Moral facts, by contrast, are immune to the explanatory
strictures. They might nevertheless manage to satisfy such demands.
But even if they fail such tests, we do not, on that account, have most
reason to deny their existence.
Postscript
I havent yet said anything that would give us reason to think that there
are moral facts. Indeed, on the very generous assumption that everything Ive argued for thus far is correct, we are left with a very large
question: if passing an explanatory requirement is not the appropriate
path to normative vindication, then what is? Why are we licensed (if we
are) in believing that there are moral facts?
Lets distinguish two questions that might tend to get entangled here.
The first is the one just asked: what reason do we have for thinking that
there are moral facts? The other, narrower question is: what reason do
we have for thinking that the specific moral facts are as we take them to
be? The former question is the more fundamental. If we cant supply a
positive answer to it, then theres no point in considering the second
question.
The answer to the first question is, at a sufficiently general level, the
same answer that we return to the question of what licenses belief in
the existence of natural facts. We believe that there is an external world,
a world of natural facts, because such a belief unifies a great deal of our
144
russ shafer-landau
particular experiences and beliefs, and presents itself as so plainly true

as to be, for all practical purposes, irresistible. True, there are skeptics
about the external world, and some of them are brilliant and capable of
avoiding any logical traps we set. There is nothing we can do but beg a
fundamental question against such people, as their position, so far as I
can see, need not involve incoherence. We will never convince them, as
anything we can adduce on behalf of the external world will presuppose its existence.
Most of us think that this intractable disagreement is compatible
with our being justified in believing that there is an external world. We
think this because of the aforementioned unification and practical irresistibility of such a belief. I think a perfectly parallel case can be made
on behalf of our confidence in the existence of moral facts. The atrocious immorality of certain actions just impresses itself upon us in a
way that makes the abandonment of such a conviction completely
untenable. Nor can we check our admiration at fidelity under extreme
temptation, or suspend our judgment of a persons virtue when she
publicly exposes the cruelties of tyrant. There are a great many easy
cases in morality, though these rarely get the press devoted to the hard
ones. These cases are easy because (at least in part) we arrive at our
beliefs about them without deliberative effort, and we attach to such
beliefs a near unshakeable conviction in their truth. They are as easy as
we think they are only if there are right answers in ethics. This general
commitment to moral facts unifies our moral beliefs and experiences
in just the way that an underlying commitment to the existence of an
external world unifies our empirical beliefs.
That is why we are licensed, if we are, in believing that there are
moral facts. But what licenses us in thinking that certain actions are
right, and others wrong, certain motives good, and others bad? Things
are less tractable here, and the parallel between moral and empirical
beliefs breaks down at this point.
We are searching for a replacement for the explanatory test of justified belief about ontological matters, a test that works so well in the
empirical realm. What, precisely, are we testing for? In the natural
realm, we are testing to see what sorts of entities exist. In the moral
realm, we are not testing to determine the contents of our ontologyat
least, once we grant the existence of a class of moral facts. We allow that
there are people, actions, intentions, motives, etc. And we allow that
there are moral properties. We dont really need to do any further ontological sussing out. We dont need to determine which entities do, and
dont exist. What we need, instead, is a way to determine which specific

moral claims are true. And whats potentially puzzling about moral
claims isnt that some, but not others, entail ontological commitments
that might be suspect. Either all moral claims do that, or none. Rather,
whats puzzling about moral claims is that we arent sure how we might
vindicate the truth of some, but not others.
And here, I am afraid, there is almost nothing constructive to be
said. What we are looking for is something more specific than a general
theory of epistemic justification. We might be foundationalists, coherentists, contextualists or reliabilists in science, as in ethics. The explanatory requirement does not serve as a general criterion for positive
epistemic status, but rather as a more specific test that can be utilized
within the context of these general theories of epistemic justification
and warrantthe test, specifically, of what our ontology should include
and exclude. The correlative test in the normative realm would not, as
I say, focus on determining the contents of our ontology, but rather on
whats to count as correct normative standards or particular normative
truths. The explanatory test enjoys allegiance from all parties to ontological investigations in the natural realm. There is nothing similar in
the normative realm.
What this means is that we have no uncontroversial, neutral starting
place from which to begin normative inquiry. We have only the guidance available to us from the edicts set forth by our preferred general
epistemology. So the coherentists among us will be counseled to proceed by gaining the needed doxastic alignment. Some foundationalists
might hope to vindicate their normative commitments by identifying
at least some of them as self-evident. Etc.
This isnt very exciting, but theres nothing, so far as I can tell, to be
done that would both make things more exciting, and preserve the
truth about normative inquiry. An explanatory test is out of place in
the normative realm, and theres nothing suited to serve as its substitute. That we lack a general test for normative commitment is something we simply have to live with. We must muddle on as we can.
Lest this be thought especially damaging to the prospects of veridical moral inquiry, remember that epistemic assessment is also a normative endeavor, as is inquiry that seeks to determine the proper
demands of practical rationality. So long as we thinkas all of us do
that we can have good reason to support our beliefs in these other
areas, we should, at least provisionally, extend the same courtesy to our
moral beliefs.
Anti-Reductionism and Supervenience

Michael Ridge*
According to anti-reductionist forms of normative realism, normative
predicates purport to refer to normative properties, and these properties
are not properly understood in any non-normative or purely descriptive
vocabulary. On these views, any attempt to reduce the normative to the
non-normative is doomed to fail. Normative properties are sui generis,
and must be understood on their own terms. G.E. Moore famously
defended a non-naturalist version of anti-reductionism, according to
which normative properties (in Moores case, specifically moral ones,
and more specifically goodness) are not only irreducible, they are nonnatural. What being non-natural came to in Moores sense is slightly
obscure, though it is nowadays often understood in terms of whether
the property in question would figure in any ideal empirical science or
be reducible to properties which would so figure.1 By contrast, other
anti-reductionist views insist that though the normative is irreducible,
it is in some important sense natural all the same. Here I want to discuss a problem for all forms of anti-reductionism about normative
properties, regardless of whether the view in question takes those
properties to be natural or non-natural.
The problem will be familiar to those who know the existing l iterature
on non-naturalism in metaethics. The argument begins with the follow
ing premise. There can be no normative difference without some nonnormative or descriptive difference as well. The normative in this sense
supervenes on the non-normative and descriptive, or so I shall argue.
Indeed, the supervenience of the normative, when properly formulated,
is plausibly characterized as analytic. Someone who denied a suitably
formulated supervenience thesis would thereby give evidence that he is
not a fully competent user of normative terms. Given supervenience,
* Many thanks to Brad Majors, Sean McKeever, David Robb, Russ Shafer-Landau,
Nicholas Sturgeon, the participants of my presentation of this material at Davidson
College, and the participants of the Cambridge Metaethics Conference.
1
For a survey of the bewildering senses given to natural and non-natural in this
domain, see M. Ridge, Moral Non-Naturalism, Stanford Encyclopedia of Philosophy
(2005), http://www.seop.leeds.ac.uk/entries/moral-non-naturalism/.
148
michael ridge
the very idea of an irreducible normative property begins to look somewhat obscure. The challenge put to anti-reductionists is to explain how
there could be normative properties which both necessarily supervene
and are irreducible. In this paper I want to consider and rebut two
attempts to deal with the challenge.
The first strategy for dealing with this challenge has recently been
put forward (in unpublished work in progress) by Nicholas Sturgeon.
Sturgeon argues that while most metaethical views can accommodate
some form of supervenience, there is no single formulation of supervenience which is hospitable to all prima facie plausible metaethical
views. Just which form of supervenience one endorses will on this
account depend on ones background meta-ethical theory. If this argument succeeds then it puts pressure on the anti-realist who argues that
non-naturalists cannot explain supervenience. For perhaps the only
form of supervenience which non-naturalists can be expected to accept
without begging any central questions is indeed a form of that doctrine
which they can, after all, explain.
In section one, I respond to Sturgeons interesting line of argument
in a concessive way. I think we do indeed learn something important
from Sturgeons critique, but the lesson is not that there is no form of
supervenience which can be used as a dialectical weapon in these
debates. Instead, Sturgeons own argument actually suggests a formulation of supervenience which does not beg any central questions, and
which therefore could reasonably be used in a premise against one or
another metaethical view. In section one I try to show how a suitably
disjunctive characterization of the subvenient base avoids any reasonable controversy of the sort Sturgeon has in mind. In the remainder of
this paper I try to show why anti-reductionist views will have problems
explaining even this form of supervenience.
The second strategy for dealing with the challenge to explain
supervenience takes a more direct approach. Here I focus on Russ
Shafer-Landaus recent attempt to meet the challenge by explaining
supervenience in anti-reductionist terms. In the remainder of the
paper, I argue Shafer-Landaus strategy fails, and that the failure is
instructive. For Shafer-Landaus strategy seems like one of the most
promising ones available to the anti-reductionist. Shafer-Landau tries
to show how what he takes to be the most promising strategy for
explaining supervenience in the philosophy of mind can be carried
over successfully to the normative domain. In particular, he argues that
in both cases, we can adequately explain supervenience by appealing to
anti-reductionism and supervenience149
a thesis about how each instantiation of any of the supervening properties is fully constituted by some concatenation of the relevant subvening properties.
I argue here that even if this strategy works in the philosophy of
mind, it remains problematic in the normative case. The possibility of
such a divide between the normative realm and other cases of supervenience is itself very instructive. For anti-reductionists (Shafer-Landau
included) very often rely on companions in guilt style arguments.
These arguments rely on the idea that a supposed problem in the normative case is really just a variant on a problem encountered in other
areas of philosophy, and that whatever solution works in that other area
can be carried over to the normative case. The more general lesson of
my argument against Shafer-Landaus proposed explanation is that
such companions in guilt arguments should be viewed with considerable suspicion. Strategies which are perfectly plausible in other domains
may not carry over plausibly to the normative, or so I shall argue. First,
though, we need to clarify just how the supervenience of the normative
should itself be understood.
1. Supervenience: Formulating the Challenge
Supervenience theses come in all different shapes and sizes, though.
Debates in metaphysics and the philosophy of mind have led to the
development of a wide array of different supervenience theses.2 How
exactly should we formulate supervenience in the context of challenging the normative anti-reductionist to explain it?
Supervenience theses can be global or local. Global supervenience
theses claim that no two possible worlds can differ in the supervening
respect without also differing in the subvening respect. By contrast,
local supervenience claims hold that two numerically distinct objects
in one and the same world cannot differ in supervening respects without also differing in the subvening respects.
One problem for local supervenience claims is that if the subvening
base is widened far enoughsay, to include spatio-temporal location
then supervenience may be trivially true for the reason that no
2
For a useful survey, see K. Bennett and B. McLaughlin, Supervenience,
Stanford Encyclopedia of Philosophy (2005), http://www.seop.leeds.ac.uk/entries/
supervenience/#4.3.
150
michael ridge
umerically distinct things can be precisely the same in all of the subn
vening respects. In which case it becomes trivial that no two things can
differ in the supervening respects without also differing in subvening
respects as well. This will be trivially true for the simple reason that no
two genuinely distinct things can fail to differ in some of the subvening
respects. Global supervenience theses skirt this problem by being formulated in terms of whole possible worlds, rather than in terms of individuals within the same world. The issue then becomes which kinds of
worlds really are genuinely possible, and which are not. I propose to
understand supervenience here as a global thesis, so as to avoid these
sorts of triviality worries.
The other issue is how we should understand the subvening
base. Here I want to consider a worry recently discussed by Nicholas
Sturgeon.3 The subvening base is sometimes characterized in terms of
the natural properties, but the question of just which properties count
as natural is itself deeply vexed. Moreover, on many ways of understanding the natural it will not be true that the normative supervenes
on the natural. Plausibly, supernatural facts about the next life can give
us reasons for action, for example, in which case two worlds could differ normatively without differing naturallysupernatural differences
in ones fate in the afterlife, or differences with respect to whether one
even has an afterlife, could be enough to explain differences in ones
reasons for action.
A seemingly more promising approach, often taken in these debates,
is to maintain that the normative supervenes on the non-normative.
This is indeed very plausible, but it is not clear that it is analytic. Here
the issue seems to be too theory-dependent for analyticity. For some
forms of descriptivism are committed to denying this sort of supervenience thesis. Let me explain why this way of defining supervenience
may be inhospitable to certain forms of descriptivism.
Consider the descriptivist view that being worthwhile just is
being pleasant. This may not be a very plausible view, but it is also not
3
In what follows, my thinking has been influenced by an as-yet unpublished paper,
currently in draft, by Nicholas Sturgeon on supervenience. See Sturgeon (unpublished
manuscript), Doubts About the Supervenience of the Evaluative. To be clear, as of the
time of my writing this article, Sturgeons own views are still potentially in flux, as he
tells me that his paper is not in his view ready for publication. I therefore do not attribute any of the ideas discussed in the text to him in the sense of representing his considered view, but only in the sense of being his ideas, regardless of whether he still endorses
them as true or not.
uncontroversially analytic that it is false. Suppose that the descriptivist

we have in mind also holds a form of dualism according to which
pleasure is itself a basic and irreducible property. In that case, the view
seems consistent with denying the supervenience of the normative on
the non-normative. For on the descriptivist view on offer, being worthwhile just is being pleasant; that is, these are one and the same property.
In that case, though, being pleasant is a normative property, not a nonnormative one. Since being pleasant is, in turn, an irreducible property,
it seems to follow that two worlds could differ normatively without
differing non-normatively. For two worlds could differ simply in the
distribution of pleasantness, given that pleasantness is irreducible.There
seems no plausible case, given a thoroughgoing dualism about pleasure, for supposing that two worlds could not differ just with respect to
pleasantness. According to the descriptivist view on offer, though, this
difference between these two worlds (that is, their difference in terms
of the distribution of pleasure) is itself a normative difference, and so
not a non-normative one. So two worlds can, on this metanormative
view, differ in their normative properties without at the same time differing in their nonnormative properties.
Another approach to defining supervenience is to insist that the normative supervenes on the descriptive, where to be a descriptive property is just to be a property which is, or at least can be, picked out in
purely descriptive terms. This account relies on some way of sorting
descriptive from non-descriptive terms, of course, but we can perhaps
rely on an intuitive sense of how such terms are sorted to draw the distinction. This is roughly how Frank Jackson proceeds in his defense of
the supervenience of the ethical on the descriptive.4
This approach does not beg any questions against descriptivists,
unlike the formulation in terms of the non-normative. For the descriptivist will of course allow that the normative supervenes on the descriptive for the simple reason that the normative just is the descriptive, and
everything supervenes on itself.
However, those who are inclined to reject descriptivism may reasonably find the supervenience of the normative on the descriptive problematic. For one thing, some philosophers find the very idea of dividing
predicates into the category of the normative and the descriptive
4
See F. Jackson, From Metaphysics to Ethics (Oxford: Oxford University Press,
1998), p. 120.
152
michael ridge
problematic, in which case the category of descriptive properties

(defined here in terms of descriptive predicates) is problematic.
I take an ecumenical approach here. For what seems undeniable is
that the normative either supervenes on the non-normative or on the
descriptive. Those who deny that the normative supervenes on the nonnormative because they are descriptivists will allow that the normative
does supervene on the descriptive (trivially). While those who deny
that the normative supervenes on the descriptive because they reject
the normative/descriptive distinction will nonetheless allow that the
normative does supervene on the non-normative. For the only reason
to reject that thesis is apparently a certain form of descriptivism, and
anyone who denies the coherence of the normative/descriptive distinction is ipso facto not a descriptivist, and hence has no grounds for
rejecting the supervenience of the normative on the non-normative.
Let us therefore take as plausible, and indeed analytic, the following
very modest global supervenience thesis:
(S) Necessarily: Two entire possible worlds cannot differ in their normative properties without also differing either (a) in their non-normative
properties or (b) in their descriptive properties.
(S) is extremely plausible, to the point that someone who denied it

would thereby betray incompetence with normative concepts. To deny
(S) would be to allow, for all that has been said so far, that it could have
been the case that the world was exactly like the actual world in all of
its non-normative and descriptive features, yet Hitlers actions were not
wrong. Since all the non-normative and descriptive facts are the same
in this possible world it will still be true that Hitler killed the same people, had the same intentions, etc. Such bare normative differences seem
inconceivable. Very plausibly, the normative facts are in some way
entirely fixed by the non-normative and descriptive facts, and this is
something we can know a priori. Indeed, it seems to be built into the
very meaning of normative discourse, and is in that sense analytic.
I therefore conclude that supervenience formulated in this way need
not beg any central questions.
2. Explaining Supervenience: Companions in Guilt and Constitution
Descriptivist forms of realism have an easy time explaining how
there could be genuine normative properties which satisfy (S), since
the supervenience of the normative on the descriptive entails (S).
By contrast, anti-reductionism seems unable to accommodate any

plausible explanation of how there could be genuine normative properties which satisfy (S). For if normative properties, non-normative
properties, and descriptive properties really are, in Humes terms, distinct existences, then it is very hard to see why it should be impossible
for the former to differ with no difference in the latter. In one version
or another, this problem has been put to anti-reductionists many
times.5
Expressivism has a much easier time on this front. The crucial point
is that the expressivists explanatory task differs from the explanatory
task facing the cognitivist. Whereas the cognitivist must explain a metaphysical relationship between two potentially distinct sets of properties, the expressivist instead needs only to explain the sensibility of a
practice of normative judgment governed by a supervenience constraint. Moreover, it is not hard to see how such an account would go.
Part of the point of expressivism is that we use normative discourse to
recommend courses of action on the basis of their non-normative and
descriptive properties. We use normative judgments to decide what to
do in any given non-normatively or descriptively characterized circumstance. There could be no sense given to the idea that we were recommending actions, or deciding on actions, on the basis of their
non-normative or descriptive features if we did not obey a supervenience constraint in making these judgments.
How might anti-reductionists reply to the charge that they cannot
explain how normative properties supervene? One strategy has been to
pursue what are sometimes called companions in guilt, though those
who deploy this strategy might prefer to call it one of companions in
innocence. The idea is to argue that we have supervenience in other
areas where reductionism does not seem plausible, but in which expressivism also does not seem plausible. For example, the mental plausibly
supervenes on the physical even if the mental is not reducible to the
physical. Yet almost no one would be tempted to infer from this that we
should be expressivists about discourse employing mentalistic idioms.
5
See, e.g., J.L. Mackie, Ethics: Inventing Right and Wrong (New York: Penguin Books,
1977), pp. 3842; S. Blackburn, Spreading the Word (Oxford: Oxford University Press,
1984), pp. 18290; and S. Blackburn, Ruling Passions (Oxford: Oxford University Press,
1998), pp. 31517. Even critics of Blackburns argument often allow that it is effective
against anti-reductionist views like Moores; see, e.g., J. Dreier, The Supervenience
Argument Against Moral Realism, The Southern Journal of Philosophy 30.3 (1992),
pp. 1338, here 1718.
154
michael ridge
The argument against anti-reductionism seems to generalize too

quickly, or so this line of reply insists.
Sometimes, defenders of anti-reductionism supplement the companions in guilt move with a positive story about supervenience, which
is good. The companions in guilt approach by itself leaves a bad taste.
One always wonders whether there might not be some relevant difference between the companions in virtue of which they will at some
point part company.
What positive story do anti-reductionists offer about supervenience,
then? They typically appeal to a strategy for explaining supervenience
which is employed in other contexts (such as the philosophy of mind),
and argue that this strategy can be carried over and successfully
deployed in the normative case as well. The most promising version of
this strategy appeals to the idea that the supervening class of properties
(the normative or the mental, for example) are constituted by or r ealized
by the subvening class of properties (the non-normative/descriptive
or the non-mental, for example) This is exactly how Russ ShaferLandau defends his version of non-naturalism. Here is how ShaferLandau puts it:
According to the sort of ethical non-naturalism that I favour, a moral fact
supervenes on a particular concatenation of descriptive facts just because
these facts realize 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 pencils 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.6
The strategy combines the companions in guilt approach with an

appeal to the idea of constitution.7 As Shafer-Landau presents it here,
the idea is to explain the supervenience of the normative on the descriptive by telling a suitable story about the constitution of normative facts
(and properties) by concatenations of descriptive facts (and properties). On Shafer-Landaus ontology, facts just are instantiations of properties,8 and he formulates the strategy in terms of the constitution of
properties as well as in terms of constitution of facts.
R. Shafer-Landau, Moral Realism: A Defence (Oxford: Oxford University Press,
2003), p. 77; emphasis added.
7
Or, as seems equivalent in Shafer-Landaus version, realization.
8
As he reminds us at Shafer-Landau, Moral Realism, p. 83 n 2.
6
Shafer-Landau actually aims to explain a slightly different supervenience thesis than the one discussed here, though. He is interested in a
supervenience thesis which holds within a single possible world, and
maintains that no two things within that world have the same normative status without differing in some descriptive respects. It seems clear,
though, that he would extend this strategy to deal with the global
supervenience thesis advanced here.
Shafer-Landaus positive explanation of supervenience dovetails
very nicely with the companions in guilt strategy. For the appeal to
constitution is supposed to be borrowing the best strategy for the companions. Shafer-Landau gives the examples of a pencil and its material
constitution and that of a subjective feeling and its neurophysiological
(and perhaps intentional) constituents.
Crucial to the success of this strategy is the idea that constitution is
distinct from identity. Otherwise, the view on offer would collapse into
a form of reductionism, with normative properties being identical with
certain constellations of descriptive properties. In itself this is not a
problem, since constitution by is indeed distinct from identity with.
Constitution is asymmetric; identity is symmetric. A clay statue is
constituted by the lump of clay, but the lump of clay is not constituted
by the statue; the clay is instead constituted by certain molecules, which
themselves are not constituted by the lump of clay, but by various
other subatomic particles, and so on. Whereas if a = b then it follows
that b = a.
However, matters are not so straightforward. One obvious disanalogy between the normative case and the case of companions in guilt is
that the supervenience of the normative on the non-normative and
descriptive is plausibly a priori, and indeed analytic. By contrast the
supervenience of the mental on the non-mental is hardly a priori, much
less analytic. In fact, the majority of human beings at various points in
history may well have denied the supervenience of the mental on the
non-mental out of an antecedent commitment to a certain sort of dualism about the mental. For many religious traditions would naturally
lead one to such a view of the mental. The same point holds about the
material constitution of macroscopic objects by subatomic particles.
Perhaps even more obviously, this constitution relation is not a priori,
much less analytic.
Shafer-Landau does not explicitly address this disanalogy in his
discussion, but there is at least one interesting move that an antireductionist can make here. The anti-reductionist can argue that the
156
michael ridge
preceding objection simply assumes that the analyticity of the supervenience of the normative on the non-normative/descriptive must be
taken as a theorem, rather than an axiom in the following sense. The
argument assumes that whatever metaphysically explains the supervenience of the normative on the non-normative/descriptive must
itself be analytic. For otherwise the explanation of supervenience will
not explain the analyticity of the supervenience thesis.
The anti-reductionist can simply reject this premise, though. The
anti-reductionist can insist that the analyticity of supervenience is axiomatic, and not derived as a theorem. It is, the reply continues, simply
a brute fact about our normative concepts that something cannot count
as rightness, for example, unless it obeys a suitable supervenience constraint. Crucially, on this view the explanation of how any given property could actually satisfy the supervenience constraint need not be
analytic, or even a priori. Indeed, at some points, this seems to be
Shafer-Landaus own view of the issue of analyticity.9
Here is an analogy that might help clarify the defence I am proposing on behalf of the anti-reductionist. Suppose that I just stipulate that
by shmong I shall refer to that property which obeys such-and-such
supervenience constraint, and which is also such that someone who
fails to do that which is shmong is therefore blameworthy. There shall
now be no mystery about why facts about which actions are shmong
will conform to the relevant supervenience constraintthat follows
trivially from the stipulated meaning of shmong. How any given property can count as being the property picked out by shmong, on the
other hand, is another matter, and that will involve explaining how the
property in question satisfies the relevant supervenience constraint.
The explanation of how a given property satisfies the relevant supervenience constraint need not itself be a priori or analytic, though. For
we do not at this stage need an explanation of why it is analytic that the
shmong facts supervene in such-and-such way; that follows from a
semantic axiom about shmong. All we need is an explanation of how
shmong so defined could manage to refer to a given property, and that
explanation need not itself be a priori or analytic.
9
See Shafer-Landau, Moral Realism, p. 88. See also Shafer-Landaus reply to a critique by Robert Mabrito in R. Shafer-Landau, Replies to Critics, Philosophical Studies
126 (2005), pp. 31329. Mabritos critique of Shafer-Landau is also itself worth reading
for those interested in this debate (see R. Mabrito, Does Shafer-Landau Have a Problem
with Supervenience?, Philosophical Studies 126 [2005], pp. 297311).
The same sort of argument could, it seems, be given in the case of

normative concepts. The meanings of these concepts are not given by
stipulation, of course, so the analogy is imperfect. Nonetheless, linguistic facts are true in virtue of conventions, and the relevant conventions
could just make it axiomatic by convention that normative discourse is
governed by a suitable supervenience constraint. The fact that our
actual conventions grew up more organically, and were not the result of
an act of stipulation, does not preclude the relevant conventions having
this shape. In which case the fact that it is analytic that the normative
supervenes on the non-normative and descriptive is a disanalogy with
the other examples of supervenience (e.g. in the philosophy of mind)
that makes no difference.
There is a residual explanatory question waiting in the wings here,
though. For why do we have predicates for normative properties as
defined by the anti-reductionist? The expressivist has an answer to the
corresponding question for their view, as we have seen. For without
the incorporation of a supervenience constraint, normative discourse
could not do the job for which it is needed. Without such a constraint,
we could not make sense of the idea that normative discourse is used to
recommend actions on the basis of their non-normative or descriptive
properties, or that normative judgment allows us to decide what to do
in any given circumstance once the relevant non-normative and features are given to us.
The anti-reductionist needs either a different answer, or an account
of how reference to the irreducible property he has in mind is essential
to the job of recommending actions on the basis of their descriptive
properties and forming rational plans for contingencies given in purely
descriptive terms. It is not obvious that either of these strategies can be
made to work. Indeed, the mere internal consistency of an expressivist
discourse shows that reference to such an irreducible property is not
the only way to get these jobs done. In which case, it would be mysterious why human beings ubiquitously and crossculturally seem to have
normative discourse in the sense the anti-reductionist suggests (with
supervenience simply stipulated by linguistic convention, etc.), instead
of in the expressivist sense.
Here is another way of posing the challenge to the anti-reductionist
who takes superveniences analyticity as axiomatic. Suppose for the
sake of argument that this gets the semantics for good correct. Now
consider the invented predicate shmood. The meaning of shmood is
given as follows: it refers to a property just like the property of being
158
michael ridge
good save for this one differenceshmoodness does not supervene.

Why should we have predicates for goodness rather than shmoodness?
Perhaps the reply from the anti-reductionist will be that there could be
no property as shmoodness, but then the burden of the argument is to
explain why this is so. Explaining why there could not be such a property will in effect be much the same as explaining why goodness must
supervene. This is not progress.
By contrast, the expressivist explanation of why we would have normative discourse which includes a supervenience constraint has the
virtue of relying on the following fairly uncontroversial premise: as
rational agents who need to cooperate and coordinate with one another,
we need a discourse that can allow us to both settle on the thing to do
and recommend actions on the basis of their non-normative or descriptive properties. Moreover, this explanation does not seem to push the
question back a stage in the way that the more stipulative anti-reductionist appeal to brute analyticity might.
Most likely, an anti-reductionist explanation of the ubiquity of
normative discourse in their favored sense will invoke more controversial premises than these. I shall not speculate here on how an antireductionist might try to tell such an etiological story, but leave this as
a challenge for their view. For the remainder of this essay, though, I put
this particular challenge to one side, as I do not here have the space to
explore the anti-reductionists resources for meeting it in anything like
adequate detail.
3. Four Challenges
I now want to present four challenges for anyone defending ShaferLandaus strategy for explaining supervenience. These challenges do
not draw their strength from the analyticity of supervenience. The antireductionists proffered explanation is cast in terms of constitution. We
are most familiar in ordinary language with the idea of constitution as
a sort of mereological notiona whole is constituted by its parts. An
army is (partially) constituted by the soldiers who make it up; a chess
board is constituted by its 64 squares; a sentence is constituted by the
words that make it up; we the people constitute a nation, and so on.
The anti-reductionist strategy, however, is cast in terms of the constitution of facts, where facts, in turn, are understood as instantiations of
properties. A pencils length at a given time, we are told, is constituted
by its molecular composition at that time.10 An instance of pain, we are

told, is constituted by a set of neural (and perhaps intentional) events.
In like manner, the argument goes, the admirability of an action is constituted by some set of descriptive features.
At this point, we need to be very clear about just what we mean when
we say that an instantiation of a property is constituted by the instantiation of some other property or set of properties. Otherwise we will
have simply traded one mystery (supervenience) for another (constitution of property instantiations). This would not be progress.
Fortunately, some of Shafer-Landaus companions (whether in guilt
or innocence) have some helpful things to say on this front. First of all,
metaphysicians have a name for the instantiations of propertiesthey
call them tropes. Tropes are abstract particulars which are wholly present in the object which instantiate them, but are incapable of being
wholly present elsewhere at the same time.11
Tropes are distinguished from properties, which are universals.
We must distinguish the property of being tall (a universal) from my
tallness (a trope, or abstract particular). Trope theorists sometimes
claim that it is an advantage of their view that they do not need to posit
a separate realm of universals. Instead, they suggest that we take universals to be nothing more than sets of tropes which resemble one
another in relevant respects. Such trope theorists are sometimes called
trope nominalists. Other versions of trope theory posit universals as
existing in some sense above and beyond the set of all of their tropes.
The arguments developed here should be neutral on this front.
Tropes are also often taken to be fully determinate, whereas the
universals they instantiated are determinable. The property of being
colored is determinablethere are many ways of being colored.
A given instantiation of the property of being colored, though (one of
its tropes), must be fully determinate, such as a very specific shade and
hue of green.
10
Actually, Shafer-Landau lumps length together with weight and claims that both
are constituted by molecular composition; see Shafer-Landau, Moral Realism, p. 77.
Weight, though, is relative to gravitational field, and so is partly constituted by more
than the objects molecular composition. Mass would be a much better example for his
purposes. This in no way undermines the argument, but it is worth getting the examples just right.
11
I take this account from D. Robb, The Properties of Mental Causation,
Philosophical Quarterly 47 (1997), pp. 17894, though it seems to be fairly standard
fare amongst trope theorists.
160
michael ridge
Within this framework, we can reformulate Shafer-Landaus hypothesis. He suggests that we can best explain the supervenience of the
normative12 on the descriptive by invoking the claim that every normative property is fully constituted by the instantiation of some cluster
of descriptive properties. Within the framework of trope theory, this
turns out to be the claim that every normative trope is constituted by
a suitable cluster of descriptive tropes. Presumably we can here read
constituted by literally as meaning something like made up of , in
which case the idea of constitution is still understood mereologically.
So far, so good.
How does the thesis that every normative trope is constituted by a
suitable cluster of descriptive tropes explain supervenience? By itself, it
does not logically entail supervenience. The thesis is logically consistent
with a given set of descriptive features constituting the instantiation of
a normative property in our world, and yet failing to constitute a normative property in another world which is otherwise identical to ours.
In order to explain supervenience, we need some auxiliary premises.
The same sort of point in the philosophy of mind emerges nicely in
David Robbs discussion of how trope theory can be deployed to explain
the supervenience of the mental on the physical. Robb offers a thesis
he calls trope monism which is very close to the philosophy of mind
analogue of Shafer-Landaus thesis about the normative. In Robbs
sense, trope monism is the thesis that every mental trope just is a physical trope, though mental types are not identical with any physical types.
This is not precisely the analogue of ShaferLandaus thesis, which
claims only that every normative trope is constituted by a descriptive
trope.
The point is that on either way of proceeding, some further assumptions are needed to explain supervenience:
Trope monism, as I have formulated it, entails Supervenience (which,
I shall assume, concerns types) so long as we are allowed two plausible
assumptions: (a) if x = y, then necessarily x = y, and (b) if a trope is an
12
Here I take some liberties with Shafer-Landaus text. He focuses exclusively on the
moral, and not on the normative more generally. However, I strongly suspect that he
would want to generalize his account to include non-moral reasons for action too, and
to defend a kind of anti-reductionist realism there too. Rather, he would have wanted
to extend the strategy in question when he still believed in this whole approach. ShaferLandau now rejects the strategy discussed in the text on the strength of some of the
arguments in the textmost specifically, the argument that constitution alone does not
explain supervenience.

F-trope, then it necessarily is an F-trope. Whenever a mental type is present, this will be in virtue of the fact that a mental trope is instantiated;
but by trope monism, this will also be a physical trope, which means that
a physical type is present. By (a) this mental trope will necessarily be
accompanied by the physical one (they are just the same trope), and by
(b), this trope will necessarily be a trope of the same physical and mental
types. So whenever a mental type is present, a physical type that necessitates it will be as well; and if this is necessarily true, the result is
Supervenience.13
Robbs strategy can be adapted in service of Shafer-Landaus theory

as follows. Let us assume that (a*) if x necessarily constitutes y then
whenever x is present, y is present too, and (b*) if a trope constitutes an
F-trope, then it necessarily constitutes an F-trope. Shafer-Landaus
hypothesis entails that whenever a normative type is present this will
be in virtue of the fact that a cluster of descriptive tropes (the one which
constitutes the normative trope in this case) is instantiated. By (b*),
this cluster of descriptive tropes necessarily constitutes the normative
trope in question. By (a*), it follows that whenever the cluster of
descriptive tropes is present the normative one will be too. So whenever a normative property is instantiated, it will be instantiated by a
cluster of descriptive properties, and the presence of this cluster of
descriptive properties necessitates the presence of the normative property. If this is necessarily true (assuming [a*] and [b*] are true), then
the supervenience of the normative on the descriptive follows.
Before explaining why I find this strategy problematic, it is important to be clear about the dialectic. Although I have followed Blackburn
in posing the challenge as one of explaining supervenience, this is
potentially misleading. Supervenience is analytic, and so perhaps in a
sense does not stand in need of explanation. What does stand in need
of explanation is how there could be properties which are irreducible
yet supervene in this way. Supervenience tells only that if there are
moral properties then they must be like this: they must supervene in
such-and-such ways. That in itself is compatible with an error theory,
according to which there are no such properties and indeed with the
conclusion that there could not be such properties. Moreover, Mackie
famously argued that there could be no such properties precisely on the
strength of worries about supervenience, among other worries. This is
the real challenge: to explain how normative properties, understood in
Robb, The Properties of Mental Causation, p. 187.
13
162
michael ridge
anti-reductionist terms, are as much as possible, given the modal properties attributed to them by supervenience. The fact that supervenience
itself is a trifling and analytic truth does nothing to dispense with this
very real explanatory burden.14
In fairness to Shafer-Landau, I should note that he has indicated (in
correspondence) that he is not sure he wants to adopt the Robb strategy for explaining supervenience. In fairness to myself, I should note
that he has not suggested a particular alternative to Robbs approach.
Moreover, the general appeal to whatever works in the philosophy of
mind seems to make it fair game to argue that Robbs approach is what
works in the philosophy of mind, yet will not work in the case of normative properties. In which case, the appeal to companions in guilt
should not ease our worries about normative supervenience. Finally, of
course, apart from Shafer-Landaus own view, it is interesting in its own
right to see whether a variation on Robbs approach could be made to
work in the case of normative properties.
I now present four challenges for the defender of this approach. The
first challenge stems from the fact that the bulk of the explanatory work
here is done by the thesis that (b*) if a trope constitutes an F-trope,
then it necessarily constitutes an F-trope. Why, though, should we take
(b*) to be true? Constituition is, after all, not identity; that is part of
Shafer-Landaus point. Why, though, could we not have the same trope
elsewhere without its constituting the same F-trope it constitutes here?
Some further metaphysical work needs to be done to vindicate this
hypothesis, or as yet we have no explanation of how normative properties could supervene. Note that Robbs version of the argument has an
advantage on this score. For Robb can explain the analogous thesis that
(b) if a trope is an F-trope, then it necessarily is an F-trope. For this
14
Many thanks to Brad Majors for drawing me out on this aspect of the dialectic,
which is, in my view, all too often submerged as implicit in ways that are confusing
in the existing debates. Here I also find it instructive to note the analogy with the philosophy of mind. If dualism in the philosophy of mind were true then it would be
plausible to infer that the mental does not necessarily supervene on the physical, as
supervenience in this case is not analytic. This shows how we can have irreducible
properties without supervenience. This contrast with the normative suggests an
explanatory itch which deserves to be scratched. How could there be irreducible properties which nonetheless supervene in the relevant ways? Perhaps the challenge can be
met, but it cannot be dismissed as not posing a genuine challenge once we distinguish
the issue of supervenience itself qua trifling analytic truth from the metaphysical question of how there could be normative properties with the modal profile entailed by
supervenience.
only relies on the necessity of numerical identity, which is much more

obvious and plausible than the necessity of constitution.
The second challenge is unfortunately not one that I can develop in
sufficient detail here. The challenge reflects the fact that trope theory
itself is hardly platitudinous. There are serious philosophical worries
about the very idea of a trope, and indeed about the reification of property instantiations more generally, quite apart from the trope framework. It is true, of course, that properties are instantiated, but we must
not directly infer from this that there are property instantiations any
more than we should infer from the fact that it is true that people do
things for the sake of others that there are sakes.15 To be clear, the worry
here is about the very idea that we should reify property instantiations
at all, and not whether we should adopt this or that theory of them,
much less whether we should call them tropes. A traditional theory of
universals might well get by with just objects and their properties without bloating its ontology with property instantiations as a third sort of
fundamental entity.16
This is, of course, not the place for a general discussion of the plausibility of trope theory. Suffice it to say that the Shafer-Landau strategy
gives some serious hostages to metaphysical fortune. Indeed, the hostages to fortune are not just the fortunes of trope theory, but of metatrope theory. For Shafer-Landaus argument will also rely on the idea
that instantiations of properties can themselves have properties in
order to make sense of the idea that a fact (which just is a property
instantiation on Shafer-Landaus view) can have the property of being
a reason. Understanding meta-properties in the trope frame-work may
lead to more specific problems, and not all trope theorists may be
inclined to posit meta-tropes in their ontology. In any event, if expressivism allows us to explain supervenience without such hostages to
metaphysical fortune, then that makes expressivism ex ante a theoretically safer bet than anti-reductionist cognitivism.
The third challenge stems from the fact that the proffered explanation of supervenience potentially has the odd and somewhat implausible consequence of ruling out certain forms of monism as a first-order
view. Historically, anti-reductionist views have been understood as
neutral on such first-order questions, though. Indeed, this is in a way
This is an old point of Quines.

Thanks to Brad Majors for useful discussion here.
15
16
164
michael ridge
one of the main virtues of anti-reductionism. Because it is entirely neutral on the first-order questions of the day, it can much better accommodate the sorts of intuitions that underwrite the force of G.E. Moores
Open Question Argument.
Why do I say that Shafer-Landaus explanation potentially rules out
certain forms of monism? Because according to trope nominalism, a
property just is a set of resembling tropes. Whether a given property is
normative or descriptive depends entirely on whether the relevant
resemblances are normative or descriptive (the or is inclusive). Antireductionism just is the thesis that normative properties are not identical with descriptive properties, however. In the context of trope
nominalism, this means that the tropes that make up a given normative
property cannot resemble one another in ways that would be salient
from a purely descriptive perspective.
Somewhat surprisingly, this is enough to rule out certain forms of
normative monism. For normative monism will hold that just one kind
of fact can constitute a reason, and the tropes that make up this sort of
fact may well be ones that resemble each other in important ways from
a purely descriptive point of view. For example, take a simple form of
egoistic hedonism, according to which a fact is a reason for an agent to
perform an action just in case it is a fact to the effect that the action
would promote the agents pleasure. The property of promoting ones
own pleasure may, though, be highly salient from a purely descriptive
point of view; in which case the tropes which make up this property
will indeed resemble one another in ways that are salient from a purely
descriptive point of view. After all, empirical psychology may well
make heavy use of this property in explaining the actions of intentional
agents, for a start. This, in turn, would entail that the property of being
a reason is, after all, a descriptive property, which is inconsistent with
anti-reductionism.
The anti-reductionist has at least two replies. The first is just to bite
the bullet, and perhaps be so bold as to claim as a virtue for his theory
that it allows us to refute egoism! This, though, would be to bite the
dialectical hand that feeds you. For ruling out egoism on purely metaethical grounds seems incompatible with the spirit, if not the letter, of
G.E. Moores Open Question Argument, which is itself perhaps the
strongest argument for antireductionism in the first place.
The second anti-reductionist reply to the objection from monism is
to reject the trope nominalists account of what makes a property count
as being of one type rather than another (physical versus mental;
normative versus descriptive). Rejecting the trope nominalists attempt

to reduce universals to sets of tropes, the reply will insist that we must
posit universals as existing in their own right, above and beyond the set
of all of their instantiations. The anti-reductionist is on this horn of the
dilemma driven quite literally to something like Platos Form of the
Good, their frequent attempts to distance themselves from that model
notwithstanding.
This may seem more plausible than biting the bullet, with respect to
certain forms of monism. However, this move also may be seen as biting the dialectical hand that feeds you, albeit a different hand this time.
For many of those who argue for trope theory do so on the grounds
that trope nominalism allows us to avoid some of the metaphysically
excesses of a theory of universals. If we are driven to a theory that posits universals anyway, then we may have trouble mounting a good
argument for trope theory in the first place. The appeal to companions
in guilt may now begin to look more like an appeal for help from some
fair weather friends who have no interest in sharing your pain.
The fourth challenge for the defender of this strategy for explaining
supervenience is that it seems to force us to reject a doctrine Jonathan
Dancy has called holism about reasons.17 Holism about reasons maintains that a fact which is a reason in one situation may be no reason at
all, or a reason with the opposite valence, in another sort of situation.
For example, the fact that watching TV would be pleasant may be a
reason here, but that very same fact may not be a reason in an other
situation, as when the pleasure would be sadistic (in watching news
about some tragedy, say).
Holism about reasons is controversial in some quarters, but I think
it is very plausible. One way of seeing its plausibility is that we do not
typically think of all reasons as being partially constituted by the fact
that the agent is able to perform the action in question. My reason to
visit my mother is that it will make her happy, not that it will make
her happy and I can visit her. Yet, plausibly there is a reason entails
can in much the same way that ought implies can. This is already
enough to establish a modest form of holism. For this entails that what
is a reason here may not be a reason there if I can perform the action
here, but not there.
17
See, for example, J. Dancy, Ethics Without Principles (Oxford: Oxford University
Press, 2004).
166
michael ridge
Finally, of course, even if holism is false, it is hardly obviously false.

Once again, the anti-reductionist had better not bite the dialectical
hand that feeds it and rule out of court an interesting first-order view
like this on purely meta-normative grounds. For that would be deeply
alien to the intuitions which make the Open Question Argument as an
argument for anti-reductionism plausible in the first place.
Why, though, do I suggest that Shafer-Landaus strategy as I have
reconstructed it forces us to reject holism about reasons? Recall that in
order to explain supervenience, Shafer-Landaus strategy required the
following auxiliary thesis: (b*) if a trope constitutes an F-trope, then it
necessarily constitutes an F-trope. Suppose a given fact constitutes a
reason. From (b*), it follows that this fact necessarily constitutes a reason whenever it obtains, regardless of context. This, though, just is the
rejection of holism.
Dancy himself is aware that his theory threatens to make
supervenience mysterious, and offers the following explanation of
supervenience:
My final question about supervenience is how to explain it What then
explains the fact that moral properties supervene on the non-moral? My
own view is that this is a consequence of the fact that they result from the
non-moral. The properties from which the wrongness of this action
results are the reasons why it is wrong, the ground for its wrongness. Now
we know already that if we move to another case holding just these properties fixed, we may yet get an action that is not wrong; differences elsewhere may conspire to prevent the original wrong-making properties
from doing that job in the new case. But if we move to another case holding fixed all the non-moral properties of the case whatever, we know in
advance that no conspiracy of that sort can happen. Whatever was a reason in the first case must remain a reason in the second, and nothing that
was not a reason in the first can become a reason in the second. So necessarily the reasons must remain the same, and if so, their rational result
must remain the same.18
The key move for our purposes comes in the sentence I have italicized.
This is enough to entail supervenience, but it hardly looks like an explanation of supervenience. For the italicized sentence just asserts a form
of superveniencethat there can be no moral difference without some
non-moral difference conspiring to make the difference. You cannot
explain supervenience by stating it.
Dancy, Ethics Without Principles, p. 89; emphasis added.
18
This last objectionthat the explanation on offer is incompatible

with holismis forceful enough by itself, but it is even more forceful
when combined with the point about monism. For now we seem inexorably led to a very specific first-order normative theory simply in virtue of accepting anti-reductionism and its apparently best explanation
of supervenience. For on the one hand, we must reject monism, and
hence must accept a plurality of types of reasons. On the other hand,
though, we must also reject the view that reasons can function holistically. In effect, we seem driven to an atomistic version of something
like W.D. Rosss theory of prima facie duties.19
Such a Rossian view may well be a plausible view, though I doubt
that the atomistic part of the view is all that plausible (see above). Even
if the view itself is plausible, what is not plausible, given the Open
Question Argument from which non-naturalism draws its plausibility,
is that our metanormative theory alone could drive us to hold such a
specific sort of view about the structure of our reasons.
Conclusion
In this essay, I have tried to do three main things. First, I have tried to
show that so long as we are willing to characterize the subvenient base
disjunctively, it is after all possible to formulate supervenience in a way
that does not beg any central metaethical questions. Second, I have
tried to clarify the structure of the challenge to anti-reductionists to
explain supervenience, and in this context have discussed whether the
putative analyticity of supervenience poses any additional dialectical
burdens on the anti-reductionist. I there conclude that the anti-reductionist can accommodate the analyticity of supervenience but then a
new challenge emerges, namely to explain why we have predicates just
like our normative predicates but which do not take supervenience as
axiomatically analytic. To my knowledge, this particular gloss on the
issue of analyticity has not been discussed before. Third, most centrally, I have posed four related challenges for what I take to be the
most interesting attempt to date by normative anti-reductionists
19
See W.D. Ross, The Right and the Good (Oxford: Oxford University Press, 2003
[1930]).
168
michael ridge
(namely, Shafer-Landaus attempt) to explain supervenience in genuinely anti-reductionist terms. By way of summary, those four challenges are:
1.A dilemma: Either explain why we should reject trope nominalism
(and hence why the Robb strategy for explaining supervenience in
the philosophy of mind is not the best strategy here too) or
explainwhy trope nominalism plus the proffered explanation does
not implausibly entail the rejection of monism. Either of these
options threatens to bite the dialectical hand that feeds the antireductionist. On the one hand there is the spirit of the Moorean
intuitions behind the Open Question Argument, which would
counsel against a direct metaphysical argument against monism.
On the other hand, there is the fact that the main force of the original argument was an appeal to companions in guilt; once we are
more selective about which companions we invoke, though, we
need an argument that we are not hanging out with the wrong
crowd. In any event, note how little ice the this is just like in the
philosophy of mind explanation is cutting now. We need a lot more
argument about just which of the available stories in the philosophy
of mind is the right one for the anti-reductionist in metanormative
theoryand the anti-reductionist may be forced to say not the
trope nominalism theory and not a standard universals theory
either, but instead a universals plus tropes theory. Needless to say,
its not obvious that this is the right story in the philosophy of mind,
in which case companions arguments really do not advance the
anti-reductionist position very far.
2.Provide some explanation of why we should reify tropes. Why not
go for a more elegant metaphysic that posits only universals and
objects which instantiate them, and doesnt bloat its ontology with
property instantiations as well? This challenge may be especially
forceful if the reply to point (1) above is to take the horn of the
dilemma which rejects trope nominalism. For one of the main arguments for tropes is that given trope nominalism they allow us to
avoid the problems associated with a theory of universals.
3.Mount a defence of the assumption that if a property constitutes
another property then it necessarily does. This assumption seems to
do the lions share of the explanatory work, but is in Shafer-Landaus
discussion a suppressed premise which gets no defence.
4.Explain why this explanatory strategy does not lead all too quickly
to the rejection of holism, or why its so leading is not problematic.
I do not pretend to have demonstrated beyond any reasonable doubt

that these four challenges cannot be met. Taken together, though, they
strike me as posing a rather deep challenge for an attempt to explain
supervenience in terms of a constitution relation between property
instantiations. In any event, in my view, the ball is now in the antireductionists court.
part IV
virtue ethics
Han Feizis Criticism of Confucianism and its

Implications for Virtue Ethics*
Eric L. Hutton
Although Confucianism is now almost synonymous with Chinese
culture, over the course of history it has also attracted many critics
from among the Chinese themselves. Of these critics, one of the most
interesting is Han Feizi (ca. 280233 BCE). For according to an early
source, Han Feizi studied under one of the greatest Confucian masters
of his age, Xunzi, but later turned his back on Confucianism.1 We do
not know what led Han Feizi to reject Xunzis teachings, but he clearly
dislikes the Confucians a great deal, even going so far as to include
them in his famous list of five vermin ( wu du) that prey like
parasites upon the body politic. While Han Feizis animosity toward
the Confucians is thus very apparent, it is less obvious whether he actually has good arguments against them, and that is the issue I want to
explore here, especially since this question has received little in-depth
discussion in English literature on Han Feizi.2
* For their comments on an earlier draft of this essay, I want to thank Erin M. Cline,
Eirik Harris, P.J. Ivanhoe, Rachana Kamtekar, Elijah Millgram, Joel Sahleen, and the
two anonymous reviewers for the Journal of Moral Philosophy, though I have not been
able to incorporate all their excellent suggestions. Likewise, I want to thank audiences
at the University of Michigan and the University of Chicago, as well as the audience at
the 2005 Conference on Chinese Philosophy in Analytical Perspectives at National
Chengchi University in Taiwan, where I presented a Chinese version of this paper.
1
The source claiming that Han Feizi studied with Xunzi is the biography of Han
Feizi in Sima Qians Shi Ji. However, this account was written well after Han Feizis
death, and it is the only early source to make such a claim. Also, Xunzi is almost never
mentioned by name in the Han Feizi (for the one clearest mention of him, see note 23
below), and the text never calls him Han Feizis teacher. These facts have led some to
suspect that the Shi Ji account is unreliable. Even if that is so, and Han Feizi in fact
never studied with Xunzi, it has no substantive impact on the thesis of this paper,
because we can still ask how well Han Feizi understood those he was criticizing and to
what extent his criticisms have force. I regard the story of Han Feizi studying with
Xunzi as merely a useful tool for trying to imagine how Han Feizi might have grasped
Confucian views quite well and anticipated their responses to certain criticisms.
2
To my knowledge, the most extended discussion of this topic in English is to be
found in Chung-ying Cheng, Legalism versus Confucianism: A Philosophical
Appraisal, in C. Cheng, New Dimensions of Confucian and Neo-Confucian Philosophy
(Albany, NY: State University of New York Press, 1991), pp. 33138. Among Chinese
174
eric l. hutton
If Han Feizi really did study with Xunzi, then one might expect Han
Feizis criticisms of the Confucian tradition to be especially incisive,
because they would be coming from someone with inside knowledge
of Confucianism. Ofcourse, there is no guarantee that this is the case
upon investigation, we may ultimately decide that Han Feizi simply
misunderstood Confucianism, or that he simply did not present good
arguments against it, either unintentionally or on purpose. Indeed,
some have called his arguments question-begging and outright blind
to the values espoused by Confucians.3 However, here I will contend
that when we reconstruct the debate between Han Feizi and the
Confucians, trying to be sympathetic to both sides and trying not to
make astraw man out of either, we can see Han Feizi as a thoughtful
and powerful critic, one whose arguments are not based simply on misunderstanding, erroneous reasoning, or lack of effort, and who therefore poses a genuine difficulty for the Confucians that has no simple
answer. Let me stress that in making this claim I do not mean to show
that Han Feizi ultimately wins the debate, because that project is far
too long and complicated to undertake here.4 My aim is rather the
more modest one of demonstrating that Han Feizi has put his finger on
a serious problem for Confucian views and that he cannot be easily
dismissed.
In addition, a growing number of scholars have proposed that
Confucianism should be regarded as a form of virtue ethics,5 and this
view makes possible interesting new approaches to understanding not
and other non-English publications, there is more treatment of this matter, though still
less than one might perhaps expect for a topic of such prominence in the text. Cf.
Liangshu Zheng , Han Feizi Zhijian Shumu (Taipei: The
Commercial Press , 1993), pp. 12328, for a list of some of these writings.
3
Cf. Cheng, Legalism versus Confucianism, p. 332. As will become clear, I disagree
with Chengs assessment. However, in order to leave enough space to set out my proposed interpretation, I will not be arguing against him directly. I leave it to my readers
to judge whose reading of Han Feizi is more persuasive.
4
Accordingly, here I am also not going to discuss the plausibility of Han Feizis
advocacy of fa (, conventionally rendered in English as laws) as the crucial element
of government. Stress on fa is one of the most distinctive aspects of Han Feizis philosophy, and so any judgment of the ultimate value of Han Feizis views would have to
include treatment of that issue. As noted in the main text, my concern here is solely his
criticism of Confucianism, and the matter of whether Han Feizis own positive proposals are any better than those of his rivals is really a distinct issue that deserves separate
consideration.
5
Cf. Nicholas Gier, The Dancing Ru: A Confucian Aesthetics of Virtue, Philosophy
East and West 51.2 (2001), pp. 280-305; Edward Slingerland, Virtue Ethics, the
Analects, and the Problem of Commensurability, Journal of Religious Ethics 29.1 (2001),
pp. 97125; Bryan Van Norden, Virtue Ethics and Confucianism, in M. Bo (ed.),
han feizis criticism of confucianism
175
only Confucian thinkers, but also their critics. If Confucianism is a

form of virtue ethics, we can then ask to what extent Chinese criticisms
of it parallel criticisms that have been launched against virtue ethics,
and what lessons for virtue ethics in general might be gleaned from the
challenges raised to Confucianism in particular. Accordingly, in tracing out Han Feizis debate with the Confucians, I will be drawing from
writings by contemporary proponents of virtue ethics and their critics,
because I hope to show how comparison with this literature provides a
helpful tool for sharpening our analysis of the ancient Chinese texts,
and at the end I will suggest how the exercise of thinking through Han
Feizis criticisms and possible Confucian responses has a broader philosophical payoff, namely by highlighting a general problem for current
defenders of virtue ethics that has not been widely noticed, but deserves
attention.
Before proceeding to these tasks, though, let me clarify two aspects
of my approach here. First, it is important to note that Han Feizi criticizes many thinkers and philosophical movements, and his remarks
often target different groups simultaneously. Thus, some passages to be
discussed below are attacking not just Confucians, but others as well.
Iwill not be commenting on these latter aspects of the text, however,
because although it is certainly worthwhile to examine all of Han Feizis
criticisms of other thinkers, in this paper I want to concentrate solely
on the ways he strikes at Confucianism.
The second matter to be clarified is my use of Confucian and
Confucianism, since scholars have recently noted ways in which the
conventional translation of the Chinese term ru () as Confucian can
be misleading, and they have also highlighted problems with using
-isms in the analysis of early Chinese thought.6 While I acknowledge
Comparative Approaches to Chinese Philosophy (London: Ashgate Publishing, 2003),
pp. 99121; idem, Virtue Ethics and Consequentialism in Early Chinese Philosophy (New
York: Cambridge University Press, 2007); and Jiyuan Yu, Virtue: Confucius and
Aristotle, Philosophy East and West 48.2 (1998), pp. 32347. My own argument for
such a view is set out in Eric Hutton, Virtue and Reason in Xunzi (unpublished PhD
thesis, Stanford University, 2001). The interpretation of Confucianism as a form of
virtue ethics is not uncontroversial. It has been challenged, for example, by Yuli Liu,
The Unity of Rule and Virtue: A Critique of a Supposed Parallel Between Confucian
Ethics and Virtue Ethics (Singapore: Eastern Universities Press, 2004). However, it
would detract too much from my main topic to engage in debate over that matter here.
It suffices for my purposes here that the virtue ethics reading is a plausible and widely
supported interpretation of Confucianism.
6
Cf. Nathan Sivin, On the Word Taoist as a Source of Perplexity. With Special
Reference to the Relations of Science and Religion in Traditional China, History of
176
eric l. hutton
the worries raised by these scholars, using such terms need not always
be gravely distorting, either, and this is particularly the case as applied
to the Han Feizi. For at the beginning of chapter fifty of the text, we find
the following statement as part of a tirade against fools and
charlatans:
Among the most prominent kinds of learning in this age are the ru The
greatest of the ru was Confucius Since the death of Confucius, there
have been the ru of Zizhang, the ru of Zisi, the ru of the Yan family, the
ru of the Meng family [i.e. Mencius], the ru of the Qidiao family, the ru
of the Zhongliang family, the ru of the Sun family, and the ru of the
Yuezheng family. Thus, after Confuciusthe ru split into eight factions
andeach faction claims that they are the true representatives of [the
way of] Confucius.7
Here, not only does Han Feizi quite clearly use the term ru to refer to
Confucius and those who follow him, but moreover he labels it a kind
of learning ( xue), and thus insofar as we are trying to understand
Han Feizis views in this essay, it is appropriate to speak of Confucians
and Confucianism, since he himself identifies his targets (or at least
one set of them) in this way.8
Accordingly, I will use Confucian to refer to Confucius and his
immediate followers, as well as those who subsequently championed
Confucius ideas about how to live, and Confucianism will refer to the
views of these people, but with the following qualification. The passage
just cited shows that Han Feizi also recognizes disagreements among
these Confucians, so that Confucianism is not a highly unified body
of thought for him (nor for me). Given this diversity within
Confucianism, a study of Han Feizis criticisms of Confucians should
Religions 17.34 (1978), pp. 303330; Michael Nylan and Mark Csikszentmihalyi,
Constructing Lineages and Inventing Traditions through Exemplary Figures in Early
China, Toung Pao 84 (2003), pp. 59-99; and Kidder Smith, Sima Tan and the Invention
of Daoism, Legalism, et cetera, The Journal of Asian Studies 62.1 (2003), pp. 12956.
7
All references to Han Feizi are according to the numbering for the version that
appears in D.C. Lau and F.C. Chen (eds.), A Concordance to the Han Feizi
(Hong Kong: The Commercial Press , 2000). This volume is part of
the Chinese University of Hong Kong Institute of Chinese Studies Ancient Chinese
Texts Concordance Series, and hereafter, this book and all other works from this series
will be referred to as HKCS for convenience. This passage is from Han Feizi, HKCS
50/150/16-19. The translation is adapted from Philip J. Ivanhoe and Bryan Van Norden
(eds.), Readings in Classical Chinese Philosophy (Indianapolis: Hackett, 2nd edn, 2005),
pp. 35152. All further translations here are my own, unless noted otherwise.
8
The same applies, mutatis mutandis, to my use of the terms Mohists and Mohism
here.
177
ideally consider each of the different Confucian views in turn. However,

besides limitations of space, our materials themselves do not permit
such a thorough investigation, because Han Feizi does not describe the
views of all the Confucian factions in detail, and we currently lack solid
textual evidence to understand each of them on our own. Hence, my
argument here that Han Feizi has a strong criticism of the Confucians
is meant as a claim about the Confucians generally, based on the best
and most substantive accounts of their ideas, rather than being a claim
about every individual Confucian of the early period, since we simply
lack evidence of some of their views.9
More specifically, then, in speaking of Han Feizis criticisms of
Confucianism, I intend primarily the early Confucian views for which
we have good evidence and which Han Feizi likely intends to attack,
namely those represented in the Analects, Mencius, and Xunzi. I include
the Analects and the Mencius, since these texts are our most reliable
sources for the thought of Confucius and Mencius and their disciples,
and Han Feizi explicitly mentions these figures in the list of names
cited above.10 I also include the Xunzi, because it seems clearly to fall
under the conception of Confucian with which Han Feizi and we are
operating here.11 Han Feizis criticism as described in what follows
seems to me to apply about equally to all three of these texts, but for the
sake of simplicity I will not be tracing out in detail how each is subject
to the problems Han Feizi identifies, and instead I will be using representative quotes from one or another of them. Readers interested in
9
Archeological excavations may one day remedy our lack of evidence. Certainly,
the finds at Mawangdui and Guodian, along with the so-called Shanghai Museum
manuscripts, have shed new light on Chinese thought in the early period. However,
scholars are still vigorously debating to what extent these newly discovered materials
can even be classed as Confucian, and their relation to the Confucian factions listed by
Han Feizi is certainly unclear, so I have not included them in my considerations here.
10
There is considerable controversy concerning the extent to which the Analects
accurately reflects the views of the historical Confucius. However, even if a given passage does not accurately represent the ideas of Confucius, most scholars agree that the
bulk of the Analects was put together by later followers of Confucius, and hence it is
still possible to use it as evidence for the Confucian ideas that Han Feizi is attacking.
The same applies to my use of materials from the Mencius and the Xunzi. Accordingly,
none of my points here depend upon correctly identifying whether a given Confucian
really said what is attributed to him, so long as it is still granted that the text from which
the quote is drawn represents the views of some Confucian or other.
11
The text of the Xunzi proclaims at various points (most famously in chapter six)
that it follows the true way of Confucius, as opposed to Mencius and others who have
distorted it, so by its content it fits the model of the ru that Han Feizi attacks. Also, in
the passage cited in the main text where Han Feizi lists the various Confucians who
squabble over the legacy of Confucius, the reference to the Sun family faction has
178
eric l. hutton
seeing further e vidence concerning particular texts are invited to consult the notes.
With these clarifications out of the way, let us begin by examining a
passage from Han Feizi that is fairly typical of some of his complaints
against Confucianism:
When a sage governs a state, he does not wait for people to be good in
deference to him. Instead, he creates a situation in which people find it
impossible to do wrong. If you wait for people to be good in deference to
you, you will find that there are no more than ten good people within the
borders of your state. But if you create a situation in which people find it
impossible to do wrong, the entire state can be brought into compliance.
In governing, one must use what works in most cases and abandon what
works in only a few cases. Therefore, the sage does not work on his virtue,
he works on his laws.12
First, although this passage does not explicitly mention Confucianism

or name Confucian thinkers, it definitely contains an implied criticism
of Confucianism, since one prominent tenet of Confucian thought is
the idea that a sagely ruler relies on his virtue to inspire people to be
good and thereby bring order to the state. This view is famously
expressed in Analects 2.3, which depicts Confucius as saying, If you try
to guide the common people with coercive regulations and keep them
in line with punishments, the common people will become evasive and
will have no sense of shame. If, however, youguide them with virtue
and keep them in line by means of ritual, the people will have a sense of
shame and will rectify themselves.13 Indeed, the passage from Han
Feizi above almost seems intended as a direct rebuttal ofthisclaim.14
been regarded by some as a reference to Xunzi, since his family name, Xun (), often
appears as Sun () in early sources (including elsewhere in the Han Feizi, cf. note 23
below). If that is correct, then Han Feizi definitely means to attack Xunzi along with
Mencius and Confucius, and we are further justified in using the Xunzi as a source text
for considering the views of those criticized by Han Feizi.
12
HKCS 50/152/10-11. Translation adapted from Ivanhoe and Van Norden,
Readings in Classical Chinese Philosophy, p. 357.
13
Translation adapted from Edward Slingerland, Confucius: The Analects
(Indianapolis: Hackett, 2003). For similar sentiments in the Mencius, cf. 2A3, and in
the Xunzi see HKCS 10/44/5-8 (D.C. Lau and F.C. Chen (eds.), A Concordance to the
Xunzi (Hong Kong: The Commercial Press , 1996)) and
William Hung (ed.), A Concordance to Hsn Tzu (Beijing: Harvard-Yenching
Institute, 1950), 33/10/36-39. (This latter volume is from the Harvard-Yenching index
series, and hereafter this book and all others from that series will be referred to as
HYIS for convenience.)
14
Of course, the idea that a sage governs by means of de ( virtue) is not limited
to the Confucians aloneit also appears in the Daodejing, for exampleand in that
179
The fundamental point of Han Feizis attack seems to be that the

Confucian view is impossibly idealistic, because they hopelessly overestimate the number of people who can be transformed and made good
through the power of virtue. In making this criticism, Han Feizi is
attacking both the Confucians goal and the methods they propose to
achieve it. For as the quote from the Analects shows, the Confucians
want political order, but more importantly they want that order to
result from the fact that the people are themselves good and selfrestrained. Furthermore, on the Confucian view the way to achieve
that order is, in the first instance, by being good oneself, which they
think will in turn move others to goodness. Thus, when Han Feizi
denies the power of virtue to transform others, he is saying both that
the Confucian approach to achieving order will not succeed and that
one simply cannot achieve the Confucian state in which the majority of
people are good and orderly of themselves.15 Instead, Han Feizi thinks
that at most, one can get them to be law-abiding, but getting the majority of them to be truly good is simply out of the question. (Note that
Han Feizis argument above is thus aimed mainly at the idea of governing through the transformative effects of virtue, not at moral cultivation per se. He has little complaint against the common people trying to
cultivate themselves morally, so long as it neither results in their being
disobedient to the ruler nor interferes with their fundamental tasks of
farming and warfare. He is, though, still quite pessimistic about how
many people can or ever will become genuinely good, either on their
own or through the influence of others.)
In response to Han Feizi, the Confucians could simply retort that
they are not being overly idealistic. They might claim that since history
tells us that the ancient sages Yao, Shun, Yu, and Tang, along with King
Wen, King Wu, and the Duke of Zhou, brought about a golden age of
regard Han Feizis criticism is not limited to just the Confucians, either. However, it
seems fair to say that the Confucians are certainly at least one major intended target of
the passage from Han Feizi.
15
In saying this, I do not mean that there is no place at all for virtue ( de) in Han
Feizis thought. For example, the Jielao () chapter of the Han Feizi contains quite
a bit of positive discussion of de. However, some scholars have suspected that this and
similar chapters were not written by Han Feizi. Resolving that controversy is beyond
the scope of this essay, so for my evidence I have excluded those chapters that are not
clearly consistent with the ideas that run throughout the bulk of the text. In the overwhelming majority of chapters in the Han Feizi, de has little role to play in the way of
life and methods of government that the text proposes. I thank Eirik Harris for pressing me on this issue.
180
eric l. hutton
harmony and prosperity by relying on the power of virtue to transform

their subjects and make them good, this shows that the Confucian
vision is quite realistic. While Confucian texts do regularly appeal to
history, as a strategy for argument this response is not very good,
because it depends on the Confucians being right about the details of
events that supposedly happened several hundred to two thousand
years or more before the life of Confucius, as Han Feizi himself points
out.16 Such historical accuracy is difficult enough for us now, and only
more so for those in ancient China. Also, from early on, rival accounts
of the sages were circulating, so the Confucian version of history was
already contested. For these reasons, it would not be very persuasive
for the Confucians to claim that their views are realistic, at least not on
the basis of such archaic history.
However, the Confucians have available another strategy. Although
it is often a powerful criticism to say that a given view is too idealistic
to succeed in practice, such is not always the case. For people may put
forward ideals that they know can never be fully realized, because they
nonetheless value the way that the ideal can shape action.17 As an example, the ideal of world peace is a lofty goal, but one whose prospects
Cf. Han Feizi, HKCS 50/150/20-24.
For a statement of such a view in the Western tradition, see Josiah Royce, The
Philosophy of Loyalty (New York: Macmillan, 1908), pp. 284-85, who defends the pursuit of such lost causes as follows:
16
17
One begins, when one serves the lost cause, to discover that, in some sense, one
ought to devote ones highest loyalty to causes that are too good to be visibly realized at any one moment of the poor wretched fleeting time world Loyalty
seeks, therefore, something essentially superhuman In its highest reaches it
always is, therefore, the service of a cause that was just now lostand lost because
the mere now is too poor a vehicle for the presentation of that ideal unity of life of
which every form of loyalty is in quest.
I should note that, as a psychological matter, one might dispute the claim that people
can really pursue ideals that they consider unachievable. For instance, one might claim
that, even when such people say they know their ideals are unachievable, deep down
they really do not believe this, and it is only because they think there still might be
some glimmer of hope for realizing their ideals that they continue to pursue them.
William James seems to adopt this sort of position (cf. William James, The Will to
Believe, in J. McDermott (ed.) The Writings of William James: A Comprehensive Edition
(Chicago: University of Chicago Press, 1977 [1896])). Determining which psychological account is correct goes beyond the scope of this paper. However, my primary point
concerns the kind of justification that might be offered for pursuing an ostensibly unrealizable ideal, and as I go on to argue, the Confucians seem to adopt a position of the
kind I describe in the main text. Even if they happened to be deceived about their own
motivations, we can still consider the logical force of their position as a reply to Han
Feizi. I thank Erin M. Cline for discussion of these latter issues.
181
may seem vanishingly slim, especially in recent years, and human

nature being what it is, world peace may ultimately be unachievable.
Nevertheless, someone might insist on maintaining world peace as an
ideal, even while understanding that it cannot be attained, on the
grounds that such an ideal at least helps motivate people to strive for
peace, whereas any other ideal would be worse, and might encourage
people to tolerate or even to endorse violent conflict. (Note that I do
not mean to support this last claim myself, but offer it only as an example of how someone might reason.) When someone espouses an ideal
in this way, it will not be persuasive to object that his or her goal is
impossible. Moreover, in acknowledging that the goal is impossible to
achieve, the person is admitting that his or her methods are incapable
of bringing it about, and therefore it will likewise be ineffective to
object that those methods will not succeed in achieving the goal.
In the case of the Confucians, while they often speak as though they
intend their ideals to be fully achievable, at times they also sound as if
they are offering them more along the lines just described, namely as
ideals they realize cannot be truly attained, but which they consider
worth pursuing nonetheless. Perhaps the clearest instance of this is
Analects 18.7, where Confucius disciple Zilu says, When the gentleman takes office, it is in order to do what is yi (righteous). As for the
fact that the Way will not be put into practice, this healready knows.
Analects 14.38 might be taken in this vein also. There, when Zilu tells a
gatekeeper that he studies with Confucius, the gatekeeper replies, Isnt
he the one who knows that what he pursues is impossible and yet persists anyway?18 Since these words come from the gatekeeper, and not
Confucius or his disciples, it is difficult to know how best to take them.
On the one hand, the gatekeeper apparently intends to be critical,
which might incline one to think that Confucius would not agree with
his remark. On the other hand, it seems hard to understand why the
gatekeeper would say that Confucius himself knows he is trying to
achieve the impossible (as opposed to trying to achievethe impossible
without knowing it) unless that is a fact that Confucius has somehow
advertised about himself. Also, whereas many other passages that
depict people criticizing Confucius end with rebuttals from Confucius
or his d
isciples (e.g. Analects 14.39, 18.6), this passage stops with the
Translation modified from Slingerland, Confucius: The Analects.
18
182
eric l. hutton
gatekeepers comments. That makes it seem as though we are to accept

the gatekeepers
characterization of Confucius as accurate (while
rejecting the implied criticism in it), especially when we keep in mind
ers, who
that the Analects was put together by Confucius follow
intended it to present a positive view of Confucius. To that extent, this
passage would likewise support understanding the Confucians as putting forth an ideal that they themselves accepted as not completely
possible.19
If that is right, then in criticizing the Confucians for being too idealistic, Han Feizi would have simply missed the mark. However, I think
we can find in Han Feizi another criticism that strikes more directly
and powerfully at the Confucian position. The basic picture is as follows. On the reading just proposed, the Confucians advocate their particular ideal while knowing that it is not fully achievable, because they
think it is nevertheless still a good ideal to pursue and one that, as a
guide for action, makes the world a better place rather than a worse
one. In general, the way to attack someone holding such aposition is
not to focus on the feasibility of the ideal, but rather to deny that having
and acting on such an ideal tends to make things better. I want to claim
that Han Feizi presents an argument of this sort, but in order to establish this point, I want first to review some recent criticism against virtue ethics, because it seems to me that Han Feizis approach is
interestingly similar, and so c onsidering those discussions will help us
to think more clearly about Han Feizis views.
Now contemporary virtue ethics is not a highly unified philosophical movement with a well-defined and distinctive position, but for present purposes we need not worry about what is the most plausible
account of virtue ethics. Instead, the most relevant point for my concerns here is that a prominent strand of thinking in virtue ethics
emphasizes that the behavior of an ideally virtuous person sets the
standard foror at any rate indicateshow people ought to act. One of
the most developed attempts to work out this idea comes from Rosalind
Hursthouse, who proposes that An action is right iff it is what a virtuous agent would characteristically (i.e. acting in character) do in the
19
For similar sentiments in the Mencius, cf. 2B13, where Mencius seems to countenance the possibility that Heaven does not wish for order in the world, and so there is
nothing human beings can do to bring it about. Despite this possibility, Mencius himself is not dissuaded from pursuing the Way. Cf. also Xunzi HYIS 103/28/36-41, HKCS
25/141/3-8, where the idea is suggested that the times may simply be such as to make
order impossible, but people should still cultivate themselves.
183
circumstances.20 The practical effect of this view and others like it is to

encourage non-virtuous people to imitate the virtuous person and
hopefully in the process become virtuous themselves.
However, critics have pointed out that the virtuous person is not
always a proper role model for the non-virtuous person. While there
are a variety of arguments to this effect,21 I want to consider just one,
which was stated by Bernard Williams. The use of Williams as an example may surprise some, because Williams was generally sympathetic to
many aspects of virtue ethics, but in fact he could also be quite critical
of some ideas associated with it. Inparticular, while responding to an
essay by John McDowell, Williams expresses the following worry about
imitating the virtuous person:
Aristotles [ideally virtuous person]was, for instance, supposed to display temperance, a moderate equilibrium of the passions which did not
even require the emergency quasi-virtue of self-control. But, if I know
that I fall short of temperance and am unreliable with respect even to
some kinds of self-control, I shall have good reason not to do some things
that a temperate person could properly and safely do. The homiletic tradition, not only within Christianity, is full of sensible warnings against
moral weight-lifting.22
To flesh this out a bit more, imagine two people, a healthy woman who
possesses the virtue of temperance, and a man having serious weight
problems due to binge eating, who with great difficulty exercises only
tenuous control over his gluttonous appetites. Imagine further that
both are invited to a party and arrive to discover that the host has laid
out an extensive buffet in the d
ining room. For the woman with temperance, it is perfectly appropriate to wander in and out of the dining
room as she eats and chats with others, since there is no temptation to
over-indulge, but for the man with gluttonous appetites, it is better to
stay out of the dining room altogether, lest he revert to his old ways. In
such a case, it seems that the right thing for the man to do is not the
same as what the virtuous person would do.
20
Rosalind Hursthouse, On Virtue Ethics (New York: Oxford University Press,
1999), p. 28.
21
For an essay that raises a very serious challenge to Hursthouses view and others
like it, see Robert Johnson, Virtue and Right, Ethics 113.4 (2003), pp. 81034.
22
Bernard Williams, Replies, in J. Altham and R. Harrison (eds.), World, Mind, and
Ethics: Essays on the Ethical Philosophy of Bernard Williams (Cambridge: Cambridge
University Press, 1995), p. 190. One might label this the Kids, dont try this at home!
objection.
184
eric l. hutton
Such criticism calls into question the use of the virtuous person as an
ideal, by showing how that ideal provides the wrong kind of guidance.
That point in turn brings us back to the issue from which we began,
namely how to construe Han Feizis criticisms of Confucianism. As
mentioned earlier, since the Confucians may allow that their ideal is
not fully realizable, merely accusing them of being too idealistic is not
a good criticism, and instead, if Han Feizi is going to make his attack
stick, he must claim that their ideal itself is the wrong kind of ideal to
follow. In my view, Han Feizi does exactly this, arguing along lines similar to those just described. Let me now review the evidence for such a
reading.
Han Feizi makes copious use of examples from history in his arguments, but there is one incident to which he refers repeatedly, and
which obviously made a deep impression upon him.23 This is the story
of Lord Zikuai of Yan. As he is depicted in the text, Lord Zikuai seems
a decent man, in many ways better than most other rulers of his age:
Lord Zikuai of Yan was the descendent of Duke Shi of Shao. He possessed
land several thousand li square, with several hundred thousand spearmen. He did not indulge in the pleasures to be had with young boys and
girls. He did not listen to the sound of bells and stone chimes. Within the
palace grounds, he did not build pools and towers, and outside the palace
grounds, he did not engage in trapping and hunting. Moreover, he himself took up the plough and hoe to cultivate the fields. The way Zikuai
belabored his body so as to care for the people was thus so great that even
those whom the ancients called sage kings and enlightened rulers, in tiring their bodies and caring for the world, were not greater than this.24
The idea that the king worries about the peoples welfare andat least
occasionallytakes up the implements of farming to aid in working
the fields were traits commonly attributed to the ancient sage kings
23
The incident is mentioned several times over five different chapters: HKCS
7/10/11-13, 35/109/16 35/110/13, 38/123/9-10, 39/127/17-24, and 44/134/27-30. The
most extended discussion is in chapter 35. This story may have been especially memorable for Han Feizi, since according to one reference (ch. 38), the king of Yan turned
down Han Feizis own teacher, Xunzi, in favor of Zizhi, a fatal mistake.
24
HKCS 44/134/27-30. In conversation, Masayuki Sato has pointed out to me that
the description of the king eschewing music in this passage implies that Zikuai was
trying to be a good king on the model of the Mohists, rather than the Confucians. This
may be so, but it does not particularly affect my argument here, because the criticism
of trying to follow the ancient sage kings, and especially imitating their actions of
handing rule over to a sagely minister, will apply just the same to the Confucians as to
the Mohists.
185
(especially Yu)25 at the time. The fact that Zikuai does these things
seems to show that he was consciously trying to imitate them, and
apparently with some success, since his accomplishments are said to
rival those of the ancient sages.
Despite these qualities, Zikuai is not a sage in Han Feizis eyes, but
rather provides an example to be avoided, for he had a fatal flaw that
cost him both his state and his life. The passage just cited continues:
Nevertheless, Zikuai died and his state perished after being taken away
from him by Zizhi, and he was ridiculed by all under Heaven. What is the
reason for this? It is because he did not understand the way to employ
ministers.26
In typical fashion, Han Feizi does not explain here what passed between
Zikuai and Zizhi, but rather assumes that his readers will know the
story, because itwas a quite infamous affair in recent history.27 The particular events he references are as follows. Zizhi, who was one of
Zikuais ministers, managed to dupe Zikuai into handing over power to
him. Zizhi, however, turned out not to be a competent ruler. The state
grew more chaotic, and after a few years, the state of Qi invaded (around
314 BCE). During the conquest Zikuai was killed, and Zizhi met with
an especially grisly fate, being made into meat paste.28 Here, Zikuais
fault is said merely to be that he did not understand how to choose and
employ ministers, and insofar as Zizhi was neither as honest or capable
as Zikuai thought, it is clear that this is a fault in Zikuai.
25
Yu supposedly engaged in physical labor personally to help clear away floods and
prepare land for cultivation. Versions of the tale can be found at: Han Feizi, HKCS
49/145/29-30; Zhuangzi, HYIS 91/32/25-27 (William Hung (ed.), A Concordance to
Chuang Tzu (Beijing: Harvard-Yenching Institute, 1947)); Xunzi HKCS
25/121/12-13, HYIS 93/25/23-24; and Analects 14.5.
26
HKCS 44/134/30-31.
27
The affair was infamous enough to be mentioned in bronze inscriptions of the
time. Cf. the discussion of the Zhongshan Wang Cuo vessel by Gilbert Mattos, Eastern
Zhou Bronze Inscriptions, in E. Shaughnessy (ed.), New Sources of Early Chinese
History: An Introduction to the Reading of Inscriptions and Manuscripts (Berkeley: The
Society for the Study of Early China and The Institute of East Asian Studies, University
of California, 1997), pp. 104111, esp. p. 109. The event is also mentioned in the
Mencius explicitly at 2B8, and implicitly at 1B10, 1B11, and 2B9.
28
The story of Zikuai and Zizhi is recorded in various places, but a fairly standard
account is contained in the Shi Ji , in the biography of Su Qin (cf. Sima Qian,
Shiji (Beijing: Zhonghua Shuju, 1998), p. 2268). Zizhis gruesome end is reported
by the Zhushu Jinian book five , Annals of King Yin , year two (for
a translation, cf. James Legge, The Shoo King (Taiwan: SMC Publishing, Inc., [1893]
1991), prolegomena p. 175).
186
eric l. hutton
However, this is not the entire picture. For we should examine

exactly how Zizhi managed to get power from Zikuai. In fact, the text
records five detailed versions of the story (which attests to Han Feizis
acute interest in the matter). While these versions all differ, in their
broad outlines they are quite similar, so I will review just one as an
example:
Pan Shou said to the king of Yan, My king would do best to hand over the
state to Zizhi. The reason why people call Yao a worthy is that he tried to
hand over the state to Xu You, but Xu You was adamant in not accepting
it, and thus Yao was famous for trying to turn over the state to Xu You,
but in fact did not lose the Empire. Now if my king were to turn over the
state to Zizhi, Zizhi will be adamant in not accepting it, and thus my king
will have fame for trying to turn over the state to Zizhi, and will have the
same conduct as Yao. Thereupon the king of Yan accordingly took the
state and assigned it to Zizhi.29
Here Pan Shou refers to a common legend that, because the sage kings
were truly concerned for the well-being of their people and the state
rather than for keeping power to themselves, as they grew older they
sought out worthy ministers with whom to entrust the government.
These worthy men, though, were likewise not interested in power, so
they either refused or accepted only with much reluctance.30 Either
way, the state was well governed, and both sides were acclaimed for the
way they displayed great virtue. In the case of Yan, Zizhi initially made
a show of refusing the throne, butcontrary to poor king Zikuais
expectationshe eventually accepted it, and things only went downhill
from there.
In analyzing the kings downfall, we should see that part of the kings
mistake was his faith in Pan Shou, and in a sense it is the same mistake
as the one Han Feizi notes in the previous passage cited, namely that
Zikuai did not know how to choose and employ appropriately people
to serve him. For according to the stories, Pan Shou was secretly a
henchman of Zizhi, and the two conspired to bring about the kings
resignation.31 More generally, one broad theme that the various stories
of Zikuais downfall share in common and which Han Feizi emphasizes
is that a ruler must exercise great care when choosing whom to trust,
HKCS 35/109/28-31.
For a telling of the Xu You tale, cf. Zhuangzi HYIS 2/1/22-26. For a case of reluctant succession, cf. Mencius in 5A5.
31
HKCS 35/110/1-7.
29
30
187
lest his power be stolen away. However, another important element in

the story is the kings desire to imitate the sage Yao by giving up his
throne to a worthy minister, because it is this desire that makes him
vulnerable to manipulation by Zizhi.32 This desire to imitate great rulers of the past is something we already saw hinted at in the first passage
about Zikuai, and in all the other versions of the story, it is this kind of
desire that leads the king into disaster.
In the section of text containing these five stories, Han Feizi does not
particularly emphasize this desire as a fault,33 but elsewhere he criticizes anyone who is wedded to the past, with special reference to the
ideal of ceding rule to a worthy minister:
The ancients strove to the utmost after virtue. People of the middle ages
chased after clever stratagems. In todays world, they contest over who is
strongest. In ancient times, there were few problems to attend to, and the
preparations needed were simple. Things were basic, coarse, and not perfected. Thus, there were those who made farm implements from shells
and pushed carts by hand [to do their work]. In ancient times, people
were few, and so they were close to each other. Goods were plentiful, and
so people thought little of profit and gave way to each other easily, and as
a result there were those [e.g. the sages Yao and Shun] who, yielding
andgiving way, handed over the Empire to others. That being the case,
engaging in yielding and giving way, esteeming kindness and generosity,
and taking as ones way benevolence and munificencethese are all
32
Of course, it is not the desire alone that makes the king vulnerable in this way. The
kings susceptibility to being manipulated also stems from the fact that he makes this
desire known, and so lays open to his ministers an opportunity to gain influence over
him. Han Feizi recognizes this point and explicitly treats it as a factor leading to the
kings downfall at HKCS 7/10/8-16.
33
Actually, this depends on a textual problem. The five stories about the king of Yan
and Zizhi are intended as explanations for lines that originally read ,
, , . , ,
(HKCS 35/106/16-17). If we understand in the second sentence as referring
to , then that sentence would mean, The ruler of men perhaps may mirror himself
against antiquityand thus you have the example of Pan Shou discussing the facts
about Yu. In that case, Han Feizi would be explicitly using the story of Zikuai being
deceived by Pan Shou to illustrate the idea that the ruler should not try to imitate past
sages, which is the point I want to stress. However, some manuscripts have instead
of , and one of the stories contains the line ,
(HKCS 35/110/4-5). This argues in favor of reading instead of , so the sentence in
question would then mean, The ruler of men perhaps may take a mirror for himself
from his scholars, but those residing with him may be neither suitable nor forthright,
and thus you have the example of Pan Shou discussing the facts about Yu. While the
latter reading (using ) is less favorable to me, I believe it is possible to support my
point using other passages from the Han Feizi, which I do in the main text, so ultimately the resolution of this textual problem is not essential to my interpretation.
188
eric l. hutton
push-government [i.e. government methods as crude as pushing a cart
by hand]. When one dwells in a time of many problems, to use implements suited for few problems is not the kind of preparation made by a
wise person. To face an age of great contention, but follow in the tracks of
those who yielded and gave way, is not the way a sage brings about order.
And so, the wise person does not ride in a push-cart, and a sage does not
practice push-government.34
First, although Han Feizi does not mention the case of Zikuai yielding
the throne to Zizhi here, given that he shows strong interest in the story
in other places, it seems most likely that he has that case in mind when
attacking those who would follow ancient ways by yielding and giving
way. Second, note that Han Feizi does not criticize the action of giving
away the throne per se; here, at least, he seems to accept that it was a
sagely thing to do for the ancients.35 Rather, his point is that what was
appropriate for the early kings is not appropriate for modern rulers,
and if we think of this in terms of the lesson learned from the case of
Zikuai and Zizhi, then the thrust of his argument seems to be that
modern rulers should not imitate past sages, lest disaster follow.36
Now the idea that one should follow the ways of the ancients was a
central part of Confucianism, and the Confucians were fond of the stories about sagekings ceding their rule to a worthy minister as a manifestation of their virtue.37 Hence, in making such remarks Han Feizi is
HKCS 47/140/20-24. Commentators disagree about how to construe parts of this
passage. Here I follow the reading of Qiyou Chen , Han Feizi Jishi
(Taipei: Huazheng Shuju, 1987).
35
Han Feizis exact view on this is difficult to pin down. He nowhere denies that the
sage kings of old ceded the throne to others, and in one place he seems to describe it as
the natural thing to do, given the circumstances (e.g. HKCS 49/145/28 49/146/6).
However, elsewhere he seems to say that it set a bad precedent which has led to chaos
ever since (HKCS 51/153/11-21), but even there he does not seem to deny that it was
an effective decision for the ruler at the time. It is worth noting that Xunzi displays a
certain hostile attitude toward the legends about the sages abdicating, so perhaps Han
Feizi picked up part of Xunzis view here (cf. Xunzi HYIS 67/18/53 68/18/72; HKCS
18/86/6 18/87/5).
36
It is worth noting that in the Zhuangzi, at HYIS 43/17/34-35, there is a passage
that is very similar to the arguments we find in Han Feizi. However, given the many
problems in dating the later chapters of the Zhuangzi, it is difficult to tell whether Han
Feizi is influenced by the Zhuangzi author or vice versa.
37
Cf. e.g. Analects 20.1 and Mencius 5A1 (esp. ), 5A4 (esp.
), 5A5, 5A6. Although Mencius denies that the king can directly give the
empire to anyone, he accepts the basic idea that the king chose a worthy person to rule
in his place. I discuss Mencius further in the main text, p. 441. For a nice example of
how Mencius encourages imitating the sages, see 6B2. As mentioned above (cf. note
35), Xunzi seems suspicious of the legends about sages abdicating, but the basic point
of the objection, as I note in the main text, has to do with the idea of imitating the
34
189
criticizing the Confucians,38 and his criticism isat a general level

like that of Williams, in that he is pointing out how trying to live up to
the Confucian ideal may well make things worse, rather than better. To
that extent, Han Feizi is casting doubt on the worthiness of their ideal
as a guide for action in the first place, and as I noted earlier, this is precisely the kind of strategy needed to ensure that his attack against the
Confucians hits home, beyond merely complaining that their aim is
unachievable and that their methods will not work (which Han Feizi
does as well, of course).
Before considering how the Confucians might respond, however, it
is worth considering the degree of similarity between the arguments by
Han Feizi and Williams a little further. In Williamss example, the reason why the non-virtuous person should not imitate the virtuous person is that it can be dangerous for someone who is defective or has
lesser ability to attempt what a person who is perfected or who has
much greater ability does. The quotes from Han Feizi, though, suggest
a different concern, namely that conditions have changed so much that
the methods used by the ancients simply cannot be effective any longer,
regardless of whether one is a sage or not, and that is why it is dangerous
to use them. In this respect, Han Feizis point is not exactly the same as
the one that Williams is making.
However, although Han Feizi criticizes those who would follow the
ancients in the last passage cited, he himself occasionally seems to
encourage rulers to imitate past sages like Shun and Yu.39 In such
instances, the actions of thesages that he praises are more or less the
same policies that he usually recommends, namely heavy use of punishments and rewards, but such passages show that Han Feizis position
is not a straightforward rejection of the past.40 Rather, his view seems to
ancients, and the Xunzi frequently promotes that idea, e.g. HYIS 45/12/36 46/12/39,
63/17/24-26, 106/31/1-5, HKCS 12/59/1-3, 17/81/4-6, 31/145/1-5, so the Xunzi is still
subject to the problem that Han Feizi is raising.
38
As before (cf. note 14 above), the criticism here may not be limited to the
Confucians, since the Mohists also frequently defended their views by reference to the
ways of the ancient sages. However, the Confucians seem to have emphasized the legend of sages ceding the throne more than the Mohists did, which makes it more likely
that the Confucians are the primary target of such remarks by Han Feizi.
39
For example, HKCS 19/33/11-15 reads like a Confucian text in praising the former kings, but the government methods praised there are the same sort that Han Feizi
promotes throughout the rest of the text.
40
Ruying Liu , Le Lun Han Fei de Xian Wang Guan ,
Jiang Huai Lun Tan 1982(1), pp. 102-104, gives a nice and succinct discussion of this issue.
190
eric l. hutton
be that in some cases it might be alright for contemporary rulers to

imitate the sages (when the ancient techniques can still work), but in
other cases not. That in turn raises the question of how to determine
which of the ancient ways can be followed and which can not, and who
is able to make such judgments.
By and large, Han Feizi seems to think that those ancient methods
depending on the ruler to possess good character and act virtuously
toward his subjects will not work in current times. However, he does
not state outright that they can never work again, and a few passages
imply that they might still be usable. For instance, in the midst of a
discussion of the importance of the power of position ( shi), Han
Feizi comments:
The reason why I discuss the power of position is for the sake ofmediocre rulers. These mediocre rulers, at best they do not reach the level of
[the sages] Yao or Shun, and at worst they do not behave like [the archtyrants] Jie or Zhou. If they hold to the law and depend on the power of
their position, there will be order; but if they abandon the power of their
position and turn their backs on the law, there will be disorder. Now if
one abandons the power of position, turns ones back on the law, and
waits for a Yao or Shun, then when a Yao or a Shun arrives there will
indeed be order, but it will only be one generation of order in a thousand
generations of disorder.41
In contrasting Yao and Shun with Jie and Zhou, Han Feizi seems clearly
to be thinking in terms of the traditional view accepted by the
Confucians, on which Yao and Shun ruled with benevolence and righteousness, and Jie and Zhou ruled with cruelty and debauchery.
Therefore, when he then considers what would happen when another
Yao or Shun arrives (as opposed to just a nameless sage), I take him to
be referring to a sage who is like Yao and Shun precisely in the respect
of ruling with benevolence and righteousness. Since he says that when
such a Yao or Shun comes, there will indeed be order, Han Feizi thus
seems to acknowledge that someone might be able to rule effectively
using the virtue-based techniques of the past sages, though it would
require sagely character and intelligence to make this work (for another
passage of thissort, see p.442 below). Yet, as this same passage indicates, Han Feizi regards such people as exceedingly rare, and for that
41
HKCS 40/129/7-9. Translation from Ivanhoe and Van Norden, Readings in
Classical Chinese Philosophy, pp. 33031.
191
reason he proposes a different method that can work for the far more
common mediocre rulers.42
While ultimately Han Feizi would probably prefer not to rest the fate
of the government on any particular strengths of the ruler, but instead
depend on a system of laws and administrative techniques that will
work for all but the most kind or most cruel rulers, insofar as Han
Feizis claim is thus that the mediocre rulers incapable of governing
successfully by virtue should avoid trying to follow the past sages
approach, his thought does seem much closer to Williamss argument
that agents of lesser ability and virtue should not imitate those of
greater ability and virtue, because disaster will likely result. In this light,
the story of Zikuai and Zizhi serves as an especially powerful case for
Han Feizi against the Confucians, because on a very plausible reading
of the passage cited earlier, in yielding the throne Zikuai was motivated
more by concerns for enhancing his reputation and keeping his power
than he was by any genuine desire to find the most worthy person to
administer the state. To that extent, he was trying to make a show of
virtue that he did not really possess, which helps explain his failure
but such morally mediocre rulers as Zikuai are precisely the norm,
according to Han Feizis view, and hence one should avoid those ideals
(e.g. that of the Confucians) that will normally lead to ruin. While Han
Feizi is not so explicit as all this, one can perhaps see at least a hint of
this Williams-style concern about the abilities of agents in his treatment of the story of Zikuai and Zizhi in this vein, given how he sometimes stresses Zikuais lack of understanding.43
42
In conversation, Li Chenyang has suggested to me that in this regard Han Feizi is
suggesting yet another kind of criticism of the Confucians, namely that they have no
acceptable fallback plan for when it turns out that their ideals cannot be realized. This
is an interesting and plausible idea that deserves more consideration. However, I will
not be discussing it further here, in orderto be able to give sufficient attention to the
line of argument I am already tracing out in the maintext.
43
There is another passage that seems to support understanding Han Feizis worry
in this way. At HKCS 38/121/24 38/122/5, Han Feizi relates the tale of Duke Wen of
Jin and the eunuch Pi. When Duke Wen was early on exiled from his state, Pi was sent
to try to assassinate him twice, but failed both times. When Duke Wen eventually manages to return to Jin and become ruler, the eunuch Pi requests an audience with him.
Through an intermediary, the Duke asks why the eunuch acted with such alacrity in
trying to kill him previously, implying that the eunuch must harbor some hatred
against him. Pi answers that he was merely following orders and that he has no personal hatred toward the Duke. He also cites the story of Guan Zhong, who nearly shot
and killed Duke Huan of Qi, but was later employed by Duke Huan, who realized that
Guan Zhong would make an effective minister. Commenting on this story, Han Feizi
192
eric l. hutton
To sum up briefly where we have arrived, Han Feizi is complaining

that following the Confucian ideal and imitating the past sages will
likely make things worse, and there are two possible reasons for this
that are distinct, though not incompatible: simply because of changes
in the world, what worked in the past will likely fail miserably in the
present, even if practiced by equally capable people;44 or, even if the
world has not changed substantially, lords of inferior quality will likely
encounter calamity if they attempt what more capable lords were previously able to accomplish. Williamss point concerns primarily the latter
case, whereas Han Feizi tends to stress the former, but there are also
moments where Han Feizi seems a bit more like Williams. At any rate,
regardless of exactly how similar one takes Han Feizi to be to Williams,
to the extent that Han Feizis criticism stresses the disaster that will
likely result from imitating the sages, his basic point is the same as
Williams, and at this juncture let us turn to consider how the Confucians
might respond.
To begin with, the Confucians might dispute the example being
usedagainst them. Certainly, Mencius remarks at 2B8 suggest that not
all Confucians would endorse Zikuais deed:
Shen Tong [a minister of Qi] asked in a private capacity, May a punitive
attack be launched against Yan?
notes that rulers such as Duke Huan and Duke Wen were capable of making use of
such previously hostile men. He then remarks:
The lords of subsequent ages have not been as intelligent as these two dukes, and
the ministers of subsequent ages have not been as smart as those two men. When
a minister who is not loyal serves a lord who is not intelligent, then if the lord does
not realize his disloyalty, then one will have villainy such as that of Cao of Yan,
Zihan, or Tian Chang [who murdered their lords]. Or if the lord does realize the
ministers disloyalty, then the minister will use the examples of Guan Zhong and
the eunuch to explain himself. The lord is then surenot to execute him, because he
thinks himself to have the virtue of Duke Huan or Duke Wen. When a minister is
enemy to the lord, but the lords intelligence cannot get clear about this and instead
gives him more materials with which to work, while the lord thinks himself smart
and is not on guard, then if his ruling lineage is no more, is this not most
probable?
Here it seems pretty clear that the problem is that inferior lords are trying to imitate
superior rulers and getting themselves into trouble doing so.
44
It is worth noting that on this reading of Han Feizi, we can now see an additional
reason why it will not be persuasive for the Confucians to respond to Han Feizis charge
of being too idealistic by insisting that history shows their aims and methods to be
realistic. For even if the Confucians were right about the past, Han Feizi will respond
by arguing that what succeeded for the sages in the past simply may not succeed for
different rulers in the present or future, and thus the appeal to history proves nothing.
193
Mencius answered, It may. It was not allowable for Zikuai to give Yan
to another person, nor was it allowable for Zizhi to receive Yan from
Zikuai. Suppose there were a well-bred man here whom you liked, and to
whom you privately gaveyour official rank and salary without telling the
king, and suppose this well-bred man likewise received these things from
you without a mandate from the kingwould that be permissible? How
is that any different from this [i.e. the affairs in Yan]?
Mencius clearly disapproves of Zikuai and Zizhi, and even thinks that
they deserve punishment, though the remainder of the passage (not
translated here) shows that he thinks Qi was wrong to have undertaken
the task of punishing Yan. His reasons for disapproving of Zikuai and
Zizhi, which are merely hinted at above, are made more explicit at 5A5
and 5A6. According to those passages, the ancient sage kings did not
directly give the empire to their successors. Instead, they recommended
those men to Heaven, to whom the empire really belongs, and Heaven
accepted them as successors. Accordingly, Mencius view would seem
to be that the state of Yan, as a landholding originally granted by the
Zhou dynasty king, belongs to the Zhou king (and ultimately to
Heaven), not to Zikuai individually, so Zikuai had no authority to give
it away on his own, but at most could only recommend to the Zhou
king that Zizhi succeed him.45 Confucians might then follow out
Mencius thought by saying that Zikuai does not represent a case where
someone succeeded in imitating the sages but met with disaster.
Such a response to Han Feizis criticism, however, would not serve
the Confucians well. For one thing, Han Feizi has other examples that
seem less subject to the same complaint. Consider the following:
In ancient times, King Wen dwelled between Feng and Hao, with a territory of only a hundred li square. He practiced benevolence and righteousness and was friendly to the Western Rong [a barbarian tribe], and
subsequently came to be king over all under Heaven. King Yan of Xu
lived to the east of Han, with a territory of five hundred li square. He
practiced benevolence and righteousness, and thirty-six states cut off
pieces of their territory [i.e. to present as a tribute and symbol of submission] and came to pay him court. [The king of Chu] feared being harmed
by him, so he raised troops and attacked Xu and subsequently destroyed
it. Thus, King Wen practiced benevolence and righteousness and came to
be king over all under Heaven, but King Yan practiced benevolence and
For more discussion of Mencius views on this matter, cf. David Nivison, Mengzi
as a Philosopher of History, in A. Chan (ed.), Mencius: Contexts and Interpretations
(Honolulu: University of Hawaii Press, 2002), especially pp. 29698.
45
194
eric l. hutton
righteousness and lost his state. This shows that benevolence and righteousness were useful in ancient times, but are not useful in present
times.46
Here, the problem does not seem to be that King Yan did not succeed
in imitating the sage king Wen. On the contrary, the problem seems
rather to be that he was too successful. For as the Chinese scholar Lin
Weiyi has astutely noted, in this passage Han Feizi seems to admit that
virtue does have suasive power even in his own contentious age, since
he says that thirty-six states were won over by King Yans benevolence
and righteousness.47 King Yan was thus succeeding quite well in following the path of King Wen, but this very success made him a threat to
Chu and so brought about his destruction. Now we are not told the
details of King Yans rule, so it might perhaps still be open to the
Confucians to argue that King Yan was not really properly imitating
King Wen. Yet, for the Confucians to dispute in this fashion every such
example that Han Feizi raises does not seem a promising strategy,
because it will begin to look like special pleading if it turns out that
any time someone ostensibly acting as the Confucians direct meets
with disaster, it is because they were not really correctly imitating the
sages. Furthermore, such a tactic would threaten to undermine the
Confucians own views, since in some cases they seem to admit that
truly virtuous people may fall victim to disaster through their good
behavior.48
Apart from these worries, though, there is another, more fundamental problem with trying to escape Han Feizis criticism by arguing that
HKCS 49/146/8-11.
Cf. Weiyi Lin , Fa Ru Jian Rong: Han Feizi de Lishi Kaocha :
(Taipei: Wenjin Chubanshe , 2004), pp. 3133. It is an
interesting question to what extent this admission may undermine Han Feizis other
remarks about the importance of virtue for good government. I cannot pursue that line
of inquiry here, however. At theveryleast, this seems to show that when Han Feizi says
in the passage that benevolence and r ighteousness are not useful ( yong) for governing, what he means is that they are not sufficient for governing well, not that they are
completely ineffective.
48
For example, Xunzi seems to think of Bi Gan as a genuinely virtuous person, and
Bi Gan tried to save his lord, the tyrant Zhou, from impending doom by remonstrating
with him, but Zhou responded by having his heart cut out (cf. Xunzi, HYIS 50/13/1019, 66/18/24-29; HKCS 13/63/28 13/64/9, 18/84/14-18). In this case, to say that
whenever doom befalls a seemingly good person, it is because that person is not really
good, would threaten to undermine the belief in Bi Gans goodness, and likewise for
other figures whom the Confucians esteem. For similar cases in the Analects, cf. 18.1,
which mentions Bi Gan, and also 15.9, which notes that death may be the price of virtue. This last idea is also expressed in the Mencius in 3B1, for example.
46
47
195
Zikuai did not successfully imitate the sages. In particular, such a tactic
seems to miss the point of Han Feizis criticism, for (subject to certain
qualifications already noted above) Han Feizi is aiming to discourage
people from even trying to imitate the past sages, not just from successfully imitating the past sages. Indeed, since Han Feizi thinks we cannot
be sure about the distant past, he seems committed to a high degree of
pessimism about peoples ability to follow in detail the actual behavior
of sages like Yao and Shun.49 His repeated criticisms of imitating past
sages would then be pointless, if he thought almost no one could truly
imitate them anyway. Accordingly, the only charitable way to understand his criticisms of imitating the past is to construe them more
broadly, namely as applying to those attempting to imitate the past,
regardless of whether or not they actually do it correctly. On this understanding, the Confucian strategy considered in the previous few paragraphs will not make a difference even if successful, because Zikuai was
clearly trying to follow the model of the sages, even if he did not succeed perfectly, and to the extent that this attempt led him into disaster,
that provides reason enough against trying to imitate the past sage
kings.50 Thus, if the Confucians are going to defend themselves against
Han Feizis attack, they must adopt a different strategy.
Now there is an alternative available to the Confucians. Namely, they
can deflect the attack launched by Han Feizi against their ideals by
claiming that such criticism misconstrues the role that these ideals are
supposed to play in guiding peoples actions. Since I have been suggesting a parallel between Bernard Williams and Han Feizi, let me go back
to Williams for a moment in trying to explain this idea. His argument
gets its bite by taking the idea of what the virtuous person would do in
the circumstances to refer to fairly specific actions and then pointing
out the dangers of imitating those actionsin terms of my earlier
example, what the temperate person does at the party is to go in and
out of the dining room partaking of the buffet, but that is not what the
person fighting gluttonous appetites should do. Similarly, Han Feizi
Cf. note 16 above.
It is worth noting that although Williamss stand on this issue is unclear, a similar
idea seems to be suggested in his original critique, in particular in his analogy with
weight-lifting (cf. p. 433 above). Namely, I may hurt myself by being unable to stay
steady once I have the weights over my head, but if the weights are too heavy for me,
I may also tear muscles and injure my back just trying to lift them, while still not budging them an inch. Thus, if I am not up to the task, I have good reason not even to
attempt it.
49
50
196
eric l. hutton
takes it that the Confucians want rulers to govern as the ancient sages
would, but then points out how the specific actions taken by the past
sages, such as ceding the throne, result in ruin when copied by others.
In both cases, it is presumed that the point of upholding the virtuous
person or sage as an ideal is to encourage people to imitate the specific
things they did.
However, one might reject this understanding of the role of the ideal
figure. If so, there are at least three alternatives to consider. Before discussing those options, though, let me note what this approach means in
terms of the story of Zikuai. Specifically, it means that the Confucians
would be saying that Zikuais actions do not in any way represent what
they are advocating. For Zikuai was trying to imitate the actions of the
sages, but that is not what they are advising people to do. Rather, on the
alternative view, what the Confucians espouse is not the simple imitation of actionssomething which perhaps even they could admit51
would indeed be subject to the kind of critique that Han Feizi articulatesbut rather the imitation of something else, and since Zikuai was
not even trying to imitate (much less succeeding in imitating) this, his
downfall and other similar stories about the perils of imitating past
sages actions do nothing to impugn the Confucian ideal.
If one does not take what is most admirable and worthy of imitation
in the virtuous person or sage to be the particular things he or she does,
then what should one be trying to imitate instead? The first possibility
to consider is that it is the ideal figures good judgmentthe ability to
pick the course of action that is truly best. On this view, the advice to
imitate the sages amounts to the idea that one ought to do what they
did, but internally and mentally, rather than externally and physically,
or in other words, to think the same way they thought. For evidence
that Confucians might adopt this approach, we may look to the Xunzi,
which in counseling the novice states, When the teacher explains thusand-so, and you also explain thus-and-so, then this means your understanding is just like your teachers understanding.52 Since here the text
51
The discussion at Mencius 4A17 of whether or not to save ones drowning sisterin-law in violation of ritual propriety might serve as an example of how the Confucians
recognized this problem. Similarly, Analects 9.3 suggests an acknowledgment that
strict adherence to the dictates of ritual may not always be best, and the Xunzi repeatedly advocates an ability to adapt in response to changing circumstances, e.g. HYIS
7/3/15, 17/6/50, 23/8/87, 28/9/48, 45/12/25, 52/14/17, HKCS 3/10/2-4, 6/25/2, 8/32/11,
9/37/19, 12/58/3, 14/67/8.
52
Xunzi, HYIS 5/2/38, HKCS 2/8/2. Compare Analects 13.5, which makes it clear
that in learning the Odes one is supposed to go beyond merely being able to recite them
197
focuses on understanding ( zhi), it is clearly not advising one merely

to parrot ones teacher, but rather to come to have the same way of
thinking.
At first glance, this view may appear to provide a way out of the difficulties posed by Han Feizi and Bernard Williams. For one thing,
judgment is not the same as action, but rather good judgment may be
expressed in a variety of different behaviors. Moreover, good judgment
may include sensitivity to ones own limitations, whether actions used
in the past will still work, and so on. Accordingly, different people
applying the same kind of thinking in different situations will likely
arrive at different deeds, and hence one who is trying to imitate the
sages need not be committed to pursuing lines of action that, either
because of ones own deficiencies or because of the circumstances, tend
toward disaster.
There are, however, two ways of construing this idea of imitating the
ideal figures good judgment, and both are subject to problems. First, if
we conceive of the ideal figure as arriving at decisions about what to do
through some particular deliberative procedure, then imitating his or
her judgment will involve imitating that persons specific train of
thought. Such a construal, though, seems susceptible to the same criticism from Han Feizi and Williams, only this time directed at the agents
thought processes instead of his or her actions. In particular, they
might argue that just because a sage thinks in a certain way, this does
not entail that it is the appropriate way for not-yet-sagely intellects to
think. If I am not as smart as a sage, then using the same methods as a
sage to decide what to do may simply be too hard for me to do without
horrendous results,53 or even if I am capable of applying the same deliberative methods correctly, under sufficiently different circumstances
those methods might be precisely the wrong way to think.54
from memory, and Mencius 5B8, which suggests that the proper aim of studying
ancient works is ultimately to understand and befriend the people who composed
them.
53
For instance, even if the deliberative process includes a step where one considers
ones own weaknesses, if the person undertaking the deliberation has an overly optimistic view of his own abilities (as many novices do), then the deliberation will still go
awry.
54
For example, imagine general A who tries to think like great generals of the past.
Now imagine that he must fight against general B, who knows the approach of those
past generals, and knows as well that general A will try to think like them. Accordingly,
general A will be at a disadvantage, because general B will be able to anticipate how he
is likely to formulate his strategy and will be well positioned to neutralize it. In such
circumstances, it is better for general A not to strategize as previous generals have.
198
eric l. hutton
The other construal of imitating the ideal figures good judgment,

which can avoid this problem, is that one should imitate the person
simply in picking out the best course of action, where no particular
thought process is presupposed. If one treats this point as the basis for
modeling oneself after the virtuous person or sage, then the recommendation to imitate that person will be a recommendation to do as
the person would do, i.e. to take careful measure of the situation and
judge appropriately, rather than a recommendation to do what the
person would do, where this amounts to a prescription for particular
actions or even particular ways of thinking. In turn, since on this
understanding the object of emulation is now extremely general,
namely just getting it right, this conception of how to follow the ideal
will not fall prey to the kind of objection that Williams and Han Feizi
raise, because it is not committed to any fixed patterns of action or
thought that might be beyond the agents ability or that might lead to
disaster in different circumstances.
One might wonder, however, whether the Confucians could really
adopt the sort of strategy just outlined. In fact, a very similar view was
proposed by at least one Confucian, the eighteenth-century thinker
Zhang Xuecheng. On Zhangs view, the greatness of the sages rests in
their choosing the exact right response to their particular historical
context. For this reason, it would be a mistake to try to do the same
things that earlier sages did. This view is clearly expressed in Zhangs
Letter to Chen Jianting Discussing Learning, which criticizes what
Zhang regardsfairly or notas an excessive preoccupation among
his fellow Confucians with composing treatises (especially philological
studies of the Classics, as practiced by Dai Zhen).55 Zhang writes:
Learning has not yet advanced much since antiquity, because Confucians
of later times have taken the Six Classics as their model and Confucius as
their teacher in a mistaken fashion. Confucius put the Way into practice
but was unable to obtain an official position, so he [edited and] transmitted the Six Classics in order to hand down their teachings for myriad
future generations, yet this was something Confucius did because he had
no other choice. Even though later Confucians no longer live [as
Confucius did] in the time when the Zhou dynasty was declining and
55
For a more extended discussion of the relation between Zhang Xuecheng and Dai
Zhen than I can provide here, cf. Ying-shih Y, Zhang Xuecheng vs. Dai Zhen: A Study
in Intellectual Challenge and Response in Eighteenth Century China, in P.J. Ivanhoe
(ed.), Chinese Language, Thought, and Culture: Nivison and his Critics (Chicago and
LaSalle, IL: Open Court, 1996), pp.12154.
199
there was nothing to be done about it, they still insist that in order to
model oneself after Confucius and take him as ones teacher, one must
compose and transmit writings to pass on to future generations. But how
can one say that they have no other choice? Why do they so disregard the
people of their own time, while being so dedicated to future generations?
Thus, those who study Confucius should study what it was that Confucius
studied [i.e. how best to respond tohisown time], and should not study
what Confucius had no other choice but todo.56
Hence, for Zhang, the way to follow the sages is not to try to do what
they did, but rather to learn to appreciate how they made the right
decision for their situation, then look to ones own circumstances and
take appropriate action for ones own time.
Although Zhang lived many centuries after Han Feizi, the roots of
Zhangs view go way back in the Confucian tradition. In fact, one can
see hints of it in Han Feizis onetime teacher, Xunzi.57 Indeed, one might
even interpret the remarks from Xunzi about imitating the teachers
understanding cited earlier along such lines. Chapter twenty-one of
the Xunzi, which discusses the causes and cures for fixation or obscuration ( bi) that prevents people from truly grasping the Way, contains the following remarks:
Thus, among the cases of fixation, onecan be fixated on origins, or one
can be fixated on ends. One can be fixated on what is far away, or one can
be fixated on what is nearby One can be fixated on the ancient past, or
one can be fixated on the present. In whatever respect the myriad things
are different, they can become objects of fixation to the exclusion of each
other. This is the common problem in the ways of the heart The sage
knows the problems in the ways of the heart So, he is neitherfor the
origins, nor for the end results, is neither for what is near, nor for what is
far away,is neither for the ancient past, nor for the present.58
Here Xunzi seems to say explicitly that the sage is neither tied to past
ways, nor does he simply follow present customs. Instead, the sage
picks what is best based on the circumstances, and in some cases this
may mean following ancient ways, and in other cases departing from
them. Now this is not to say that Xunzi holds the exact same position
56
Cf. Xuecheng Zhang , Yu Chen Jianting Lun Xue (Letter
to Chen Jianting Discussing Learning), in X. Zhang, Zhang Xuecheng Yishu
(Beijing: Wenwu Chuban She , 1985), p. 86 (, ).
57
David S. Nivison, The Life and Thought of Chang Hseh-cheng (17381801)
(Stanford: Stanford University Press, 1966), p. 166, notes a similarity between Xunzi
and Zhang that is much like what I suggest here.
58
Xunzi, HYIS 78/21/6 79/21/29, HKCS 21/102/12 21/103/18.
200
eric l. hutton
as Zhang Xuechengand other parts of the Xunzi argue against any

easy attempt to equate their viewsbut rather to note that even at Han
Feizis time, there was a possibility that the Confucians could respond
along the lines contemplated earlier.59
There is, however, a price to be paid for adopting such a view, one
that perhaps explains why Xunzi and other early Confucians did not
adopt something like Zhangs view in the first place. For if one argues
that what is great and worthy of imitation in the sages is the general
way they responded correctly to their specific circumstances, rather
than the particular actions they took or the ideas they conceived, then
the model of the sages seems to provide very little substantive guidance
for those trying to follow in their footsteps. In other words, the
Confucian view would have escaped criticism, but only at the risk of
becoming vacuous. In turn, that would leave the Confucians with little
ground to stand upon in rejecting Han Feizis rival proposals for how to
live and run the government, and in fact, they would be hard pressed to
distinguish their own position from Han Feizis. For Han Feizi himself
says, [T]he sage does not expect to follow the ways of the ancients or
model his behavior on an unchanging standard of what is acceptable.
He examines the affairs of the age and then makes his preparations
accordingly.60 Perhaps theConfucians might insist that, to the extent
they could agree with this, the sage would still never disregard morality
in reacting to his circumstances, but given that the Confucians also
think of the sage as someone who bringsor can bringorder to the
59
For passages in other early Confucian texts that could be taken as suggesting the
idea that what is valuable and worthy of imitation in the sage is a kind of highly flexible
good judgment, cf. those mentioned in note 51 above. Also, there is another way in
which one might see this idea among the early Confucians. The Confucian notion of yi
(, usually translated as righteousness or appropriateness) is commonly interpreted
as being a virtue that consists in having correct judgment that is flexible and casespecific. I reject this interpretation, but will not dispute it here, because that discussion
would take us too far away from our main topic. However, it is important to see that
even if one accepts the common interpretation of yi, the Confucian view so construed
is subject to the same objections that I note in what follows in the main text, namely
that if yi consists simply in correct judgment, where this is highly flexible, then it lacks
content to guide the beginner or justify the specific approach to government favored by
the Confucians. Also, even on the common interpretation of yi, this virtue is something that must be slowly cultivated and is not immediately available to a beginner, so
it is likewise subject to the worries I consider on p. 449 of the main text.
60
HKCS 49/145/18-19. Translation from Ivanhoe and Van Norden, Readings in
Classical Chinese Philosophy, p. 340. Nivison, The Life and Thought of Chang Hsehcheng, pp. 12425 points out that Zhang Xuecheng has sympathies with Han Feizi and
that Zhangs own attempts to distance himself from Han Feis views are problematic.
201
whole world, the burden of proof seems to be on them to explain why

establishing order would never require disregarding virtue, as Han
Feizi suggests. (One wonders if Han Feizi has not, in the end, proved
himself too good a student of Xunzi, by taking the idea cited from
Xunzi earlier and running it to its logical conclusionHan Feizi might
say: morality itself can become an object of fixation, and so a true sage
will not be wedded even tomorality.)
There is one final possibility to consider, which is that the object of
imitation is the ideal figures good character, the tendency to be motivated by certain desires and feelings. On this construal, the advice to
imitate the sages amounts to the idea that one ought to be as they
were, i.e. virtuous, rather than to act or think as they did. This is more
substantive than just getting things right, and as before, such a view
looks like it might provide a way out of the difficulties posed by Han
Feizi. For insofar as proper feelings and desires may be compatible with
different behaviors and different ways of deliberating, it will again turn
out that one who is trying to imitate sagely character need not be committed to pursuing lines of action or thought that tend toward disaster.
Furthermore, this approach certainly seems to suit the Confucians
well, because they are indeed deeply concerned with how people are
motivated on the inside, and not just how they act on the outside.
The problem with this approach, however, is that according to the
Confucian view, as well as most contemporary proponents of virtue
ethics, one cannot simply will oneself into the proper character. Rather,
the cultivation of character takes time and is a process that works
through habituation.61 Furthermore, not just any kind of action will
serve for the proper habituation, but only certain kinds of action. Yet,
the idea that imitating the character of the ideal person involves a process of cultivation, which in turn requires certain kinds of actions, simply raises the specter of Williamss and Han Feizis criticism all over
again. For it may turn out that the kinds of actions necessary for
cultivating the appropriate character are precisely of the sort that tend
to ones ruin.
61
For example, consider Confucius statement in Analects 2.4 that it took him seventy years to get to the point where he could follow his desires without going astray,
Mencius warnings at Mencius 2A2 neither to neglect nor rush the process of moral
cultivation, and Xunzis remarks at HYIS 2/1/17-18, 89/23/68-69; HKCS 1/2/9-10,
23/116/13-15 about the need for slow accumulation ( ji) of efforts in order to
achieve sagehood.
202
eric l. hutton
To give an example of this worry, suppose I want to imitate a strongman. On the view under consideration, I want to imitate him in being
strong, rather than the particular things he does, such as lifting 300 lbs,
which would be extremely dangerous for me even to try, if I am a weakling. (Here I rely on Williamss formulation of the problem rather than
Han Feizis, because Williamss is more straightforward, but one could
easily imagine versions of the problem, mutatis mutandis, in terms of
Han Feizis worry.) Nevertheless, it may turn out that the only way for
me to become strong is by lifting heavy things, which is still quite
dangerous.
Now there is an objection that could be raised to this last point, in
order to save the Confucian position. Namely, just as in the way that, as
a matter of fact, many people do manage to become strong by lifting
heavy things without seriously injuring themselves, so by analogy it
would seem possible to imitate the virtuous persons virtue without a
high probability of falling into disaster. Note, however, that at this point
the relation between action and the ideal has now become rather tenuous; the actions required in such imitation may bear only a very vague
resemblance to the actions of the ideal figure (e.g. the strong person
pulls airplanes along the ground with his bare hands, but the beginner
lifts 5-lb dumbbells). If so, then it will be difficult to use the actions of
past sages to justify particular regimens of practice for current imitators, and that would again undercut the Confucians ability to justify
their preferred form of government against Han Feizis proposalsthe
problem of vacuity has thus returned. In terms of the analogy, if my
goal is to be strong, and I could accomplish this in more than one way,
say by lifting heavy things or by taking drugs, etc., then the goal of
being strong will not serve to justify one over the other. Similarly, Han
Feizi sometimes seems to suggest that if one really wants to be a benevolent ruler, as the Confucians espouse, one should follow his proposals
rather than theirs,62 and if the Confucians admit that there might be
many ways to be benevolent in order to avoid Han Feizis initial criticism, they will again have trouble ruling out this possibility.
The argument of the preceding three paragraphs still leaves open
one further move, the last I will consider. The previous argument works
by assuming that the beginner must imitate the sages actions in order
to become virtuous, but then notes how, in order to avoid Williamss
E.g. HKCS 42/130/24 42/131/1, 53/156/12-15, 54/156/19-27.
62
203
and Han Feizis critiques, that demand must be reformulated as the

idea that the beginner must in some sense imitate the sages actions in
order to become virtuous. Yet, once one has relaxed the sense in which
one will allow the beginner to imitate the sages actions, the sages
actions no longer provide a clear guide for how to behave. This argument, though, takes it for granted that the beginner should be striving
to imitate the sages manifestation of already-achieved virtue, and then
plays on the discrepancy between the actions allowed on the weak
sense of imitation and the sages heroic displays of virtue to undermine the role of the sage as a guide for action. Instead, one might propose that what the beginner needs to imitate is not already-achieved
virtue, but rather virtue-in-progress. That is to say, rather than doing
what the sages did in order to become virtuous, the novice should do
what the sages did in order to become virtuous. On such a view, the
actions of perfected sages can remain a model in the sense of an end
goal to aim at, but the beginner has a separate model for cultivation.63
This practice model would consist of less heroic and more homely
actions to imitate, actions that are accordingly more likely to be safe for
such a person to do in any circumstances, and hence this model would
be less susceptible to the worries raised by Williams and Han Feizi.
The Confucians do not explicitly distinguish between such models,
and if anything they tend to focus on heroic figures like Yao, Shun, and
Yu, which is perhaps what makes them a ripe target for Han Feizi.
However, such a distinction does not seem incompatible with their
views, either.64 Yet, while distinguishing a practice model from goal
model in this way would help, it would not necessarily solve the problems for the Confucians entirely. For the actions to be imitated on a
practice model, even if far less demanding than that of the goal model,
could still be subject to the problems pointed out by Williams and Han
Feizi, especially those of Han Feizi, if one grants that circumstances can
vary enough to make almost any given type of action likely to be ruinous. Much would depend on exactly how such a practice model is
formulated.
This idea was suggested to me by P.J. Ivanhoe.

More positively, one might see hints of such a distinction in passages such as
Analects 19.12, which seems to treat things like sweeping the floor, answering questions, and entering and withdrawing from a room as more basic practices in self cultivation that are to be followed by different and more advanced practices later.
63
64
204
eric l. hutton
In sum, Han Feizis challenge seems to raise a genuine difficulty for

the Confucians that cannot be answered in any simple and straightforward fashion.65 On the one hand, insofar as they cast their proposals for
how to live and how to run the state in terms of imitating the sages,
where the sages provide a fairly substantive model, they face the problem that following this ideal might actually make things worse. On the
other hand, if they take the sages as a model for imitation only in the
sense that they provide examples of right judgment or good character,
rather than specific guides for action and thought, then they undermine the distinctiveness of their own position and their ability to reject
alternative proposals like those of Han Feizi. For the Confucians to
escape this problem, they must treat the sages as a model for imitation
at some general level, general enough to avoid the kind of problem
raised by the case of Zikuai, but not so general that what is worthy of
imitation is merely the sages right judgment or good character (where
these have little connection to particular patterns of action and
thought). Articulating and defending that middle position, however, is
no easy task. Let me stress that in saying this I do not mean that the
Confucians have no answer to Han Feizis challenge. I think that they
may well have an answer in their conception of ritual and its relation to
virtue, especially because ritual may help provide something like the
kind of practice model discussed above.66 To examine that matter,
though, is too big a task to undertake here, so I will not attempt it now.
At minimum, what I do want to emphasize is that the kind of middle
position the Confucians must occupy to escape Han Feizis attack is,
65
I should add that I do not intend to claim that the line of argument I have been
tracing out is the only powerful criticism of the Confucians to be found in the Han
Feizi. I am perfectly willing to admit that there may be other such arguments in the text
(cf. note 42 above for a possible example), but I have not considered them here, so as to
be able to concentrate on the particular case where I think Han Feizi does make a
strong criticism.
66
In addition, the Confucians may have other resources to draw upon in answering
Han Feizis challenge. For example, early Confucians often stress the importance of
having a teacher to guide ones ritual practice (e.g. Xunzi HYIS 5/2/37-38, HKCS 2/8/
1-2), and presumably such a teacher would help steer students away from practices for
which they are not ready or which would have disastrous results in the current circumstances. Of course, these other Confucian ideas face challenges of their own, such as
the difficulty for a noviceespecially a rulerin identifying a good teacher (a point
that Han Feizi himself occasionally raises). As with the case of ritual, examining
whether and how these other notions can serve to defend the Confucians is too complicated a project to pursue in this paper, so I cannot do more than simply acknowledge
these possibilities here. I thank one of the referees for the Journal of Moral Philosophy
for pressing me on this point.
205
like most other middle positions in philosophical debates, a tricky spot

to occupy, and it is certainly not immediately clear that the Confucians
do successfully occupy it, even when their discussions of ritual and virtue are taken into account.
Since I have suggested a parallel between Han Feizis attack on
Confucianism and contemporary criticisms of virtue ethics, let me
close by trying to extract from the Chinese texts a lesson for contemporary philosophers. Two ideas are often considered to be part of virtue
ethics. One, as seen in Hursthouse, is the claim that the virtuous persons behavior indicates or sets the standard for how to act, and that in
this respect virtue ethics can provide substantive guidance for people.
The other is that the virtuous persons judgment cannot be reduced to
a grasp of rules, because no set of rules is sufficient to cover every possible situation of moral choiceor to put it another way, there are
unique s ituations calling for case-specific judgments that resist codification, and the virtuous person is one who makes the correct judgments in such cases.67 In thinking through the Chinese texts, though,
we discover that these two elements may be in tension with one another.
On the one hand, the first claim leads to the idea that one should imitate the virtuous persons actions. On the other hand, insofar as the
virtuous person is conceived as making correct judgments in unique
circumstances, then it will be a mistake to imitate those actions in any
other situation. At most, one can imitate the virtuous person in
responding properly to the particular salient features of the situation,
but since this is something that has no fixed form, the injunction to
imitate the virtuous person winds up yielding no specific guidance.68
The way to avoid this tension is for advocates of virtue ethics on the
one hand to recognize that not every action of the virtuous person is so
unique as to resist generalizationthat is what makes it possible for
their advice to imitate the virtuous person to be meaningful and helpful for those who are not virtuousand on the other hand to recognize
that not all actions of the virtuous person can serve as a model for
imitation. However, sorting out exactly which of the virtuous persons
67
For example, cf. John McDowell, Virtue and Reason, in idem, Mind, Value, and
Reality (Cambridge, MA: Harvard University Press, 1998), pp. 5073, and Jonathan
Dancy, Moral Reasons (Cambridge, MA: Blackwell, 1993).
68
To put this point in jargon that may be more familiar to contemporary philosophers, there seems to be a tension between using the good person criterion as a guide
for action and adopting a particularist stance toward moral judgments.
206
eric l. hutton
actions should be imitated, and which not, is a difficult task. In attempting it, the advocates of virtue ethics would perhaps also do well to distinguish something like a practice model from a goal model. Since the
Confucians (or at any rate, some of them) hope to occupy this kind of
middle ground, and since their notion of ritual may help in building an
appropriate practice model, this suggests that those interested in virtue ethics could perhaps learn something from how the Confucians
tried to do this, though that is, as I have said, a task I must leave for
another time.
Aristotelean Virtue and the Interpersonal

Aspect of Ethical Character
Maria W. Merritt
Aristotle holds that an agent acts virtuously only if his choices of
virtuous action proceed from a firm and unchangeable character.1
That is, the agent should be firmly established in his dispositions to
respond emotionally in the right way to the right objects of choice and
avoidance, to deliberate well, and to choose the right actions for the
right reasons. Aristotles normative ideal of firm and unchangeable
character is endorsed in the currently influential, broadly Aristotelean
school of thought known as virtue ethics.2
Drawing on central concepts of Aristotles Nicomachean Ethics,
I offer an account of how the normative ideal of firm and unchangeable
character is supposed to be realized psychologically (section II). I then
consider this aspect of Aristotelean virtue in light of present-day
empirical findings about relevant psychological processes (section III).
Since the Aristotelean conception of virtue emphasizes consistency
and integration within the agents set of motivations, an empirical
research topic of great interest is the psychological processes by which
individuals regulate themselvesor fail to regulate themselveswith
reference to their normative commitments.
The evidence suggests that these processes have an interpersonal
aspect. Typically, and often not consciously, their operation will be
mediated by our social bonds with other individuals whose values we
take ourselves to share. This proposition might seem to sit well with a
reasonably nuanced account of Aristotelean virtue. Aristotle allows
that among those who share the end of living a good life, social bonds
like friendship can help individuals to sustain virtuous character.
1
1105a30-1105b1. All quotations from the text of Aristotles Nicomachean Ethics
(hereafter, NE for short) will be in David Rosss translation, as revised by J.L. Ackrill
and J.O. Urmson, Nicomachean Ethics (Oxford: Oxford University Press, 1925).
2
A central instance is Rosalind Hursthouse, On Virtue Ethics (Oxford: Oxford
University Press, 1999), p. 136 and pp. 11319. On the centrality of practical reasoningchoosing theright actions for the right reasonsto the classical Western conception of virtuous character, see Julia Annas, Virtue Ethics, in D. Copp (ed.), The Oxford
Handbook of Ethical Theory (Oxford: Oxford University Press, 2006), pp. 51336.
208
maria w. merritt
But trouble emerges in the form of further empirical findings. The

same interpersonal processes that sometimes help to sustain character
may also disrupt it, even among agents who have the right values in
principle. The character-oriented self-understanding that would-be
virtuous agents share with their fellows can give rise to misplaced confidence, on the part of each individual, that his or her normative commitments will reliably issue in right action, even while an objective
viewpoint reveals that sometimes they do not. Moreover, ironically, the
more importance we place on seeing ourselves as aiming at the good,
the more vulnerable we may be to this insidious psychological dynamic.
To put it bluntly, the problem is not whether you have the right values but whether you act in accordance with the values you have, and
how interpersonal influences that are otherwise ethically constructive
may, over time, pull your actions apart from your professed values.
Fortunately, the empirical evidence also suggests some remedial measures, both practical and philosophical, to counteract the potentially
damaging effects of these influences. The practical measures include
a stance of due fallibilism and modesty about the state of ones
own character, caution against over-idealizing the character of close
associates, and a commitment to building mechanisms of objective
accountability into institutions and organizations. An important philosophical measure, I conclude (section IV), is for the advocates of virtue
ethics to address agents psychological need for a systematic decision
procedure that focuses attention primarily on substantive ethical considerations rather than characterological assessment. In order to put
virtue ethics effectively into practice, agents need a reliable way to track
and organize the many substantive ethical considerations that are
theoretically unified in the abstract conception of what the virtuous
person would do.
I. Moral Theory and Scientific Psychology
We need first to address a methodological concern: why and in what
ways does experimental scientific psychology matter for philosophical
moral psychology? Philosophers in growing numbers have taken up
the problem of compatibility between the two.3 This is not merely a
3
Jonathan Webber, Virtue, Character, and Situation, Journal of Moral Philosophy 3
(2006), pp.193213; John M. Doris and Stephen P. Stich, As a Matter of Fact: Empirical
aristotelian virtue and ethical character
209
matter of general-purpose intellectual responsibility, an interest in

keeping ones home discipline informed by relevant results from other
disciplines. The reason why moral philosophers should aim for compatibility with scientific psychology is s pecific to the purposes of normative theory. In general, any normative theory has to take account of
the empirical understandingif there is oneof what it applies to.
Where X is a kind of object (or state of affairs or event) in the natural
world, a theory intended as normative for the behavior (or occurrence)
of X calls for an empirically informed account of what X is.4
For example, since moral theories purport to articulate norms for
human action, they should be informed by an empirical understandingof what actions are, and of what people are doing when they engage
in behavior that folk psychology takes to be intentional. Similarly, as
I aim to illustrate in this article, moral theories that purport to articulate a psychology of virtuea set of norms for morally important dispositions of personal charactershould be informed by an empirical
Perspectives on Ethics, in F. Jackson and M. Smith (eds.), The Oxford Handbook of
Contemporary Philosophy (Oxford: Oxford University Press, 2005), pp. 11452; John
M. Doris, Precs and Replies: Evidence and Sensibility, Philosophy and
Phenomenological Research 73 (2005), pp. 63235, 65677, with commentaries by Julia
Annas (pp. 63642), Nomy Arpaly (pp. 64347) and Robert C. Solomon (pp. 64855);
John Sabini and Maury Silver, Lack of Character? Situationism Critiqued, Ethics 115
(2005), pp. 53562; Peter B. M. Vranas, The Indeterminacy Paradox: Character
Evaluations and Human Psychology, Nous 39 (2005), pp. 142; Rachana Kamtekar,
Situationism and Virtue Ethics on the Content of Our Character, Ethics 114 (2004),
pp. 45891; Peter Railton, Toward an Ethics that Inhabits the World, in Brian Leiter
(ed.), The Future for Philosophy (Oxford: Oxford University Press, 2004), pp. 26584;
Shaun Nichols, Sentimental Rules: On the Natural Foundations of Moral Judgment
(Oxford: Oxford University Press, 2004); Christian Miller, Social Psychology and
Virtue Ethics, Journal of Ethics 7 (2004), pp. 36592; John M. Doris, Lack of Character:
Personality and Moral Behavior (Cambridge: Cambridge University Press, 2002);
Gopal Sreenivasan, Errors about Errors: Virtue Theory and Trait Attribution, Mind
111 (2002), pp. 4768; Maria Merritt, Virtue Ethics and Situationist Personality
Psychology, Ethical Theory and Moral Practice 3 (2000), pp. 36583; Gilbert Harman,
Moral Philosophy Meets Social Psychology: Virtue Ethics and the Fundamental
Attribution Error, in G. Harman, Explaining Value and Other Essays in Moral
Philosophy (Oxford: Oxford University Press, 2000), pp. 16578; Owen Flanagan,
Varieties of Moral Personality: Ethics and Psychological Realism (Cambridge, MA:
Harvard University Press, 1991).
4
Thanks to Geoff Sayre-McCord for conversation on this point. For a similar point
based on common philosophical acceptance of the principle that moral judgments
supervene on non-moral facts, see Railton, Toward an Ethics that Inhabits the World,
pp. 27071): if we act rightly or wrongly, for the good or the bad, we do so within the
natural world we inhabit as empirical beingsany account of the domain of moral
thought and practice must be compatible with what we know of the domain of human
psychology, biology, and circumstance upon which it supervenes.
210
maria w. merritt
understanding of the psychological processes that affect the continuity

and sustainability of such dispositions over time.
Those who resist the intrusion of scientific psychology into moral
theory might couch their opposition as follows. First, since the primary
method of scientific psychology is controlled experimentation, whose
results are at best inductive generalizations, it cannot establish the
impossibility of exceptions. Second, it is perfectly in keeping with the
normative purposes of moral theory to exhort us towards an ideal that
is only ever realized in exceptional cases, cases of precisely the kind
that inductive generalizations leave room for.
This response contains a kernel of methodological truth. Since the
purposes of moral theory are primarily normative, not predictive or
explanatory, compatibility with scientific psychology is at most a constraint on moral theory, not a goal in its own right. Where to invoke the
constraint depends in part upon what empirical commitments a given
instance of moral theory incurs as it goes about its normative business.
A moral theory might tailor its psychology of virtue to normative purposes by articulating an ideal of practical rationality and knowledge of
the good, while recognizing that the ideal is rarely attained, and may
well be unattainable by the vast majority of men and women. Such a
moral theory might then incur no commitment to constrain its normative psychology of virtue by the inductive generalizations of scientific
psychology, although it could use them to diagnose and inventory the
ways in which people are liable to fall short of the ideal.
Rachana Kamtekar has argued that Plato, Aristotle, and the Stoics
take something like this approach to the psychology of virtue.5 By the
same reasoning, Kamtekar and other advocates of traditional virtue
ethics can claim that their approach is compatible with some of the
seemingly problematic findings of experimental psychology. The findings that have motivated virtue ethicists to rally most vigorously to this
defense are those of situationist social psychology, which suggest that
most peoples morally important behavior is often better explained
by chance situational factors than by dispositions of personal character. Even John Doris, who has presented at length the situationist
argument against routine attribution of virtuous (or any) character
traits, acknowledges that situationism leaves open the possibility of
5
Kamtekar, Situationism and Virtue Ethics on the Content of Our Character,
pp. 48285.
211
individual virtue in rare cases.6 Eric Hutton, in sympathy with traditional virtue ethics, has recently interpreted early Confucian thought
as compatible with situationist social psychology in just this way. While
the Confucians recognize that ordinary people are susceptible to the
vagaries of situational influence, at the same time they believe in rare
gentlemen and sages whose good behavior is utterly stable and
consistent.7
Yet the social function of moral norms is to regulate actual conduct
among ordinary men and women in human populations. And part of
the point of philosophical moral theory is to reflect critically on what
the content of moral norms should be, given their social function. If
traditional philosophical ideals of virtue are so lofty that even their
advocates expect them to be realized only rarely at best, how can they
be relevant to practical morality?
One possibility is that individuals might strive toward such ideals as
aspirational goals.8 Even when the goal is unattainable you can hope,
through the effort of trying to reach it, to become better than you might
otherwise have been. A second possibility is to adopt a division-oflabor model and hold out hope for a moral meritocracy to keep the rest
of us in line. Hutton suggests that in early Confucian thought, reflection on ideals of virtue could be crucial not despite the rarity of virtue,
but because of it. Given a traditional hierarchical society, it is ideally
the virtuous gentlemen and sages who will establish the social structures that ordinary (ex hypothesi, mostly non-virtuous) people inhabit:
if there are people who do have robust character traits and are resistant
to situational variation, they can design and reliably maintain the broad
range of institutions and situations that facilitate good behavior for
everyone else.9 Transposing the idea to modern liberal democracies,
we might hope that the individuals who lead particular communities,
Doris, Lack of Character, pp. 11012.

Eric L. Hutton, Character, Situationism, and Early Confucian Thought,
Philosophical Studies 127 (2006), pp. 3758, at p. 49.
8
A kind of informal, non-institutional moral education may occur through young
peoples attempts to imitate morally exceptional community members whom they
admire. But the uncritical acceptance of personal acquaintances as exemplars becomes
problematic to the extent that moral maturity demands independent thought. (See
section III.B below.)
9
Hutton, Character, Situationism, and Early Confucian Thought, p. 50. Hutton
emphasizes that he presents this solution as an exegesis of the early Confucian
thinkers, not as a positive practical proposal.
6
7
212
maria w. merritt
organizations and institutions will turn out to be the ones who prove
capable of virtue.10
These two possibilities are not mutually exclusive. Individuals could
be encouraged to aspire to virtue whether or not they hold positions of
social or political leadership. Those who do hold such positions could
be given to understand that their institutional roles carry an extra burden of responsibility to strive for virtue, commensurate with the importance of their influence on situational settings that guide the behavior
of men and women subject to their authority. But even if a few virtuous
super-individuals exist, there is no particular reason to be confident
that they will regularly occupy positions of power (not least because in
order to secure such positions they would have to compete against
individuals and organizations who value power above all). In general,
if virtue is so rare as traditional virtue ethicists are willing to maintain,
and if our behavior is influenced by morally arbitrary factors so extensively as they are willing to admit, we have good reason to doubt that
we should entrust the effective social functioning of moral norms to
the personal resolve and conscientious striving of individuals (ourselves or others, highly placed or not).
My concern in this article is with the psychology of interpersonal
processes as they affect morally important personal dispositions over
time. This issue is distinct from the problem highlighted by situationism, which is that morally irrelevant factors affect morally important
behavior at the time of action. Focusing on the influential Aristotelean
ideal of virtuous character as firm and unchangeable, I argue that the
evidence about interpersonal processes and their effect on personal
dispositions over time provides a distinct and additional reason, over
and above situationism, to predict that ordinary men and women will
run into trouble in the attempt to realize such an ideal.11 The safest bet
is to assume that however well-intentioned and conscientious we may
be, empirical generalizations about interpersonal processes apply to
10
Or third, as Hutton (Character, Situationism, and Early Confucian Thought,
p. 57, n. 41) suggests in passing, insofar as a society is democratic, and it is the ordinary people who direct the government and thereby their own lives, it may seem
equally imperative to make their behavior as little situationally-dependent as possible.
But to take this third option is to back off from the assumption that virtue is rare, thus
incurring greater exposure of moral norms to empirical generalizations about what
most people are like.
11
In principle, my point about interpersonal processes should stand on its own,
whatever ones view of the situationist problem.
213
each of usthat is, to each of us who take an interest in morality when

we make choices about how to act, how to live our lives, and how to
shape the social settings for which we bear some responsibility.
II. Firm and Unchangeable Character: Moral Psychology
and Development
How, in Aristotles ethics, is the normative ideal of firm and unchangeable character supposed to be psychologically realized? What is essential to Aristotelean virtue is the agents choice of some actions and
rejection of other actions, under descriptions that are sensitive to relevant ethical considerations.12 This is one of the points at which the current orthodox philosophical conception of virtue is directly indebted
to Aristotle, who defines it as a hexis prohairetike, a state of character
(hexis) concerned with choice (prohairetike).13 In Aristotles conception a vice or a virtue, as a hexis prohairetike, is a disposition to respond
in particular situations by choosing certain actions. Virtue is the disposition to respond by choosing the right actions.14 Virtue can be analyzed
into two basic components, practical wisdom (phronesis) and moral
virtue (ethike arete). As Aristotle conceives of them, each component
tends to perpetuate itself, and each tends to reinforce the other as their
activity continues over the course of the agents life. Once acquired, the
motivational structure of virtue thus tends to perpetuate itself, making
virtuous character firm and unchangeable. As Aristotle puts it, the virtues tend...by their own nature, to the doing of the acts by which they
are produced.15
Practical Wisdom
Practical wisdom is the ability to excel in deliberation about what is
truly valuable, from the standpoint of reflection on the shape of your
12
Cf. Bernard Williams, Ethics and the Limits of Philosophy (Cambridge, MA:
Harvard University Press, 1985), pp. 811.
13
NE 1106b36.
14
While choosing the right actions may often be effortless for the virtuous agent, it
is not mindless. As opposed to doing the same thing always and inattentively in the
same circumstances, like brushing your teeth every night before bed, virtuous choice
involves doing the appropriately different thing attentively in varying circumstances.
W.F.R. Hardie, Aristotles Ethical Theory (Oxford: Oxford University Press, 1968),
p. 104.
15
NE 1114b2729.
214
maria w. merritt
life as a whole. It involves an astute grasp, well-informed by experience,

of how such goods as friendship, honor, achievement, wealth, justice,
physical security, sensual gratification, intellectual activity, and so
forth, are in play under various kinds of circumstances. An agent who
has practical wisdom knows how to deliberate well and make good
choices of action where these goods are concerned, in all situations.
Aristotle holds that practical wisdom cannot be forgotten.16 Once
you have it you are not going to lose it, short of the most dire debilitation or destruction of your cognitive faculties. Suppose it is true that
certain goods in life are the best goods it is possible for human beings
to enjoy. Then once you have learned how to recognize them even in
their most subtle appearances, and how to deliberate so that through
your choices of action you do all you can to achieve them, not only in
familiar situations but in perplexing or novel situations as well, that
does not seem to be the kind of thing at which you can fall out of
practice.
The kinds of choices practical wisdom helps us to deliberate about,
choices about how to act with respect to the goods of life, are always
and everywhere with us. We have to make choices about these goods
constantly, although it may be possible to get out of practice at making
choices about some specific kind of good or other. For instance, if you
spent several years in seclusion engaged in excellent intellectual activity, you might get out of practice at the kind of high-pressure decisionmaking needed for excellent political activity. But even without
constant practice at deliberating about all the specific kinds of human
goods, you would not lose touch with more fundamental practical
insights, say, about the comparative value of various goods. You always
have the opportunity to practice using such insights, given that you
always have choices to make with respect to at least some of the goods
of life. At least, this is so for as long as you continue to possess moral
virtue.
Moral Virtue
Moral virtue is excellence in the activity of the emotions and appetites.
It keeps you in the attitude of caring appropriately about the goods of
life, so that your concern for getting your choices right about those
NE 1140b2930.
16
215
goods will guarantee you a succession of appropriate occasions for the

exercise of practical wisdom. In Aristotles view the life of virtue really
is the best life, and so long as the appetitive domain of your psyche is
in a condition to experience it as best, you can be counted on to stick
with that life.
More precisely, excellence in the appetitive domain is foundational
to Aristotles characterization of the virtuous man as one who is such as
to act for the sake of the noble. Aristotle claims that there are in general
three objects of choice and three objects of avoidance: the noble, the
advantageous, the pleasant, and their contraries, the base, the injurious, the painful.17 At the same time, pleasure accompanies all objects
of choice; for even the noble and the advantageous appear pleasant.18
Some objects of choice are merely pleasant: they appeal to any creature
subject to sensual appetites. But other objects of choice are pleasant
because and only so far as the agent is able to relate them to goods that
take a certain kind of rational awareness to appreciate, goods which are
for that reason superior. Rest following extreme physical exertion is
merely pleasant; running hard for miles to get an urgent message to the
commanding officer in a battle is pleasant, or rather is rightly to be
found pleasant, because the honor of serving well in battle is noble.
With respect to the objects of avoidance, running hard for miles is
merely painful, whereas it would be disgraceful to arrive too late with
the message because, feeling tired, one slowed down to a comfortable
jog. As a disgraceful action, slowing down should strike the virtuous
man as far more repellentand repellent in a more important way
than continuing to run as hard as he can over the necessary distance,
physical pain notwithstanding. In general, the virtuous man has accurate sensibilities and the right priorities when it comes to such distinctions among objects of choice and avoidance. He recognizes what is
noble and values it above all, and he recognizes what is base and reviles
it more strongly than he does anything else.
Moral virtue, then, is constituted by our values in the sense of our
attractions and aversions to the appropriate objects of choice, existing
in harmony with our appetites, emotions, and actions. Once the development of moral virtue initially establishes the attitudes of loving the
NE 1104b30-1105a1. For Aristotle, the noble is associated with the idea of a
given action as the done thing, or more precisely, what the man of practical wisdom
would do.
18
Ibid.
17
216
maria w. merritt
noble and despising the base, we will energetically seek to learn more
about how to make practical choices in accordance with those attitudes, progressing in the gradual acquisition of practical wisdom.
Abilities that we develop on the way to moral virtue, such as impulse
control and other forms of self-control necessary for keeping emotional and appetitive activity in line with our values, are in due course
supposed to reach beyond mere self-control to culminate in the highly
integrated self-mastery of virtuous character.
Meanwhile the process of acquiring practical wisdom teaches us
more about the nature of the nobler objects of choice, thereby adding
depth, intensity, and discerning precision to our love for the noble.
What constitutes practical wisdom is our commitment to our values in
reflective understanding, manifested in our intelligent recognition of
when and how they are in play on occasions of deliberation and choice,
including circumstances of adversity. The continued accumulation of
experience gradually reveals more and more of the ways in which the
nobler objects of choice really are superior, keeping us in the attitude of
caring about getting our choices right.
Interpersonal Aspect of Virtuous Character
Aristotle allows that even mature virtuous character needs at least
some s upport from social relationships. We can begin by remembering
that in his view the whole value of virtue lies in its active exercise, as
distinct from its mere possession.19 Now add the consideration that a
great many of the Aristotelean virtuescourage in battle, justice, liberality, magnificence, pride, due ambition, good temper, friendliness,
truthfulness, ready witrequire for their active exercise a social world
inhabited by a community of peers. This is so for reasons over and
above Aristotles argument that since human beings are social animals,
we can live the best human life only in society. More interestingly for
our purposes, it is so because Aristotles conception of virtue, founded
in love for the noble and contempt for the base, is highly sensitive to
standards of honor.
For Aristotle as for Homer, whom he quotes approvingly on this
point in his discussion of courage, honor is an essentially social phenomenon.20 In order for a man to be fully assured that in his conduct
NE 1098b33-1099a5; see also NE 1114a7-10.
NE 1116a20-26.
19
20
217
he meaningfully satisfies a standard of honor, he needs to win the recognition and respect of other men who are honorable themselves. And
for their part, each of them needs the same.21 The standard of honor
must be located in the judgment of fellow men of practical wisdom, as
opposed to the judgment of anyone at all, however wise or foolish. The
value of honor is in its guidance and confirmation of the honored mans
understanding of himself as virtuous; only honor bestowed by men of
practical wisdom can deliver this result.22
Similarly, in Aristotles account of friendship, friends are supposed
to love each other on the basis of virtue, each loving the other for his
good character. They choose to live together and they coordinate their
activities to pursue various specific ends that they share. The overall
end that unites them is the end of living a good life. One reason they
can help each other to pursue this larger end is that each admires and
tries to emulate the other. Given that Iregard my friend as someone
generally of good character, a person practiced in living well and better
at it in some respects than I am, I can learn from my friend how to
improve in those respects. And in the best kind of friendship there will
be reciprocity: my friend can learn from me in the same way, mutatis
mutandis.23
So, in Aristotles view, the virtuous man needs a social world for the
expression and exercise of the virtues. He needs not just any social
world, but one that prominently features social bonds with others who
share his evaluative sensibilities and priorities.
III. Empirical Psychology: Interpersonal Processes and
the Discontinuity of Character
It is a commonplace of Aristotelean moral psychology that in early life
the agent will, ideally, begin to adopt the values characteristic of virtue
by way of training and education in a supportive social community.
With increasing maturity and reflective understanding comes increasing independence in the agents commitment to the right values. The
goal of living up to the right values, initially a goal assigned by others,
NE 1095b26-29.
See also NE 1159a21-24. For a discussion of the general point see Bernard
Williams, Shame and Necessity (Berkeley: University of California Press, 1993).
23
NE 1172a11-14.
21
22
218
maria w. merritt
is transformed into a goal we set for ourselves. But even self-set goals
may continue to be influenced by the continuing normative expectations of selected others: those who share our values, and with whom we
choose to maintain interpersonal relationships. As noted above, this is
what we find in Aristotles ideal of the virtuous mans social world.
Using the resources of present-day scientific psychology, I shall suggest how interpersonal processes operate through social bonds with
like-minded others, with the effect of helping individuals to sustain virtuous normative commitments over time (section A); I shall then give
an empirically informed account of how certain aspects of the same
interpersonal processes also threaten to break the self-reinforcing
cycle between moral virtue and practical wisdom, prompting occasions of discontinuity that may over time become lasting disruptions
(section B). Rather than simply serving to strengthen character, as the
Aristotelean view would have it, social bonds with others who share
our evaluative sensibilities and prioritieseven assuming we have got
those right in principlemay also tend to undermine character.
A. Self-Evaluation and Self-Presentation
The agents values are the psychological ground of virtue. For
Aristotelean virtue, we may more accurately speak of positive and negative values, or values and anti-values, since virtue requires getting
it right not only about the noble as an object of choice, but also about
the base as an object of avoidance.24 This motivational structure brings
with it a self-evaluative concern to be, or to become, a certain kind
of person.
In the first place, the very activity or stance of valuing draws some
of ones attention to self-evaluation. As Agnieszka Jaworska has
pointedout, A person values herself in terms of how well she lives up
to her values.25 On top of that, the standard Aristotelean tradition
specifically encourages the use of broadly characterological attributions in evaluating ones own and others actions. Indeed, for Aristotle
the normative criterion of virtuous choice, the orthos logos, is set by the
idea of choosing as the man of practical wisdom would choose.26
NE 1104b30-1105a1.
Agnieszka Jaworska, Respecting the Margins of Agency: Alzheimers Patients
and the Capacity to Value, Philosophy and Public Affairs 28 (1999), pp. 105138,
at p. 115.
26
NE 1106b36-1107a2.
24
25
219
Similarly, contemporary neo-Aristoteleans like Hursthouse set out to

define right action in terms of what a virtuous agent would characteristically (i.e. acting in character) do in the circumstances.27 By implication, characterological self-evaluation is central to the first-person
practice of virtue ethics in the Aristotelean tradition.
Psychologists have studied a similar activity under the name of self-
discrepancy theory. This is a topic under the broader heading of
self-regulation, the study of behavior as a purposive, continually selfadjusting process of moving toward desired outcomes and away from
undesired outcomes.28 According to self-discrepancy theory, people
regulate their behavior in part by reference to self-guides, to which
they compare their perceptions of their actual selves in an ongoing
effort to become more like a desired self .29
Interpersonal processes that engage with self-regulation to influence
subsequent behavior constitute a major research topic in social psychology.30 Experiments reveal the cognitive, affective, and behavioral
manifestations of individuals interest in living up to the normative
expectations of selected others, whose responses can serve as benchmarks and provide external feedback to help guide the process of
becoming more like ones desired self.31 A pervasive type of behavioral
manifestation is self-presentation (also known as impression management), whereby each participant in a social encounter shapes the
impression he or she makes on others. Far from necessarily involving
Hursthouse, On Virtue Ethics, p. 28.
C.S. Carver, Self-Regulation, in A. Tesser and N. Schwarz (eds.), Blackwell
Handbook of Social Psychology: Intraindividual Processes (Oxford: Blackwell, 2001),
pp. 307328. See also C.S. Carver and M.F. Scheier, On the Self-Regulation of Behavior
(Cambridge: Cambridge University Press, 1998).
29
E.T. Higgins, Self-discrepancy Theory: A Theory Relating Self and Affect,
Psychological Review 94 (1987), pp. 31940; E.T. Higgins and O. Tykocinski, Selfdiscrepancies and Biographical Memory: Personality and Cognition at the Level of
Psychological Situation, Personality and Social Psychology Bulletin 18 (1992),
pp. 52735.
30
For reviews of the literature, see S. Chen, H.C. Boucher and M.P. Tapias, The
Relational Self Revealed: Integrative Conceptualization and Implications for
Interpersonal Life, Psychological Bulletin 132 (2006), pp. 15179; M.R. Banaji and D.A.
Prentice, The Self in Social Contexts, Annual Review of Psychology 45 (1994), pp. 297
332; H. Markus and S. Cross, The Interpersonal Self , in L.A. Pervin (ed.), Handbook of
Personality (New York: Guilford Press, 1990), pp. 576608.
31
M.R. Leary, The Self We Know and the Self We Show: Self-esteem, Selfpresentation, and the Maintenance of Interpersonal Relationships, in G.J.O. Fletcher
and M.S. Clark (eds.), Blackwell Handbook of Social Psychology: Interpersonal Processes
(Oxford: Blackwell, 2001), pp. 45777.
27
28
220
maria w. merritt
manipulation or any sort of conscious calculation, self-presentation

may often be neither accessible to conscious awareness nor controlled
by conscious effort. And far from being confined to the likes of shady
used-car salesmen, it is basic to successful communication and social
interaction, not only for first encounters but also over the long run in
relationships of collegiality, friendship, or love. Its interpersonal function is to allow us to select and streamline the information we convey
about ourselves, in a manner tailored to the relationship and circumstances at hand.32
In many interpersonal contexts, self-presentation can have the effect
of encouraging or confirming others beliefs that we possess dispositions of character that they (and we) consider desirable.33 Over time,
when we engage in repeated interactions with the same parties they
acquire expectations about how we will act not only during one
encounter, but also on future occasions. If we care about maintaining
our relationships, we should generally prefer to satisfy such expectations (or to renegotiate them so that its easier to satisfy them in the
future).34 It makes sense to act accordingly, at least when it is reasonable to assume that others stand in a basically cooperative attitude
toward us. The cumulative effect can be to establish and sustain dispositions of motivation and choice that tend to produce the desired
pattern of conduct.
For instance, studies of the Michelangelo phenomenon have found
that couples in close-partner relationships sculpt each others dispositions, values, and behavioral tendencies, through the expectations
that each forms of theother. In a stable, well-functioning relationship,
each partner endorses the others desired self, and communicates
expectations that tend to bring out the best in the other with reference
32
In the routines of everyday activity and interaction, there is simply too much
information about oneself to do anything but provide an edited, packaged version that
is relevant to the goals at hand. B.R. Schlenker and M.F. Weigold, Self-consciousness
and Self-presentation: Being Autonomous versus Appearing Autonomous, Journal of
Personality and Social Psychology 59 (1990), pp. 82028. See also B.R. Schlenker and
M.F. Weigold, Interpersonal Processes Involving Impression Regulation and
Management, Annual Review of Psychology 43 (1992), pp. 13368; M.R. Leary, Selfpresentation: Impression Management and Interpersonal Behavior (Boulder, CO:
Westview Press, 1996).
33
Leary, The Self We Know and the Self We Show, pp. 46263.
34
To be sure, even when there is some incentive to satisfy such expectations, there
might also be other incentives to opt out. The advisability of opting out will depend in
part on the cost of severing the relationships in play.
221
to this ideal. (In unhappy relationships the process backfires, each

partner pushing the other further away from his or her desired self.)35
Similarly, as we maintain interpersonal relationships and fulfill our
social roles over an extended period of time, the cumulative effect may
be to reinforce in each of us the personal dispositions that are most
closely bound up with the expectations we have a stake in fulfilling.
With regard to ethical practice, I suggest that whatever success we
may enjoy in meeting our own self-evaluative standards for personal
character, even in maturity and even assuming we have somehow really
got the right values, is typically subject to our ongoing engagement
with the expectations of particular other individuals with whom we
hold our values in common. What I mean to highlight here is nothing
as superficial as an immature, unreflective dependence upon others for
overt prompting about how to act, or a lack of motivation to act rightly
unless those others are currently present and looking on.36 The reinforced dispositions can issue in motivation and choice straightforwardly, and no less so in situations of which one believes not another
living soul will ever learn what happened. My point is that typically,
these deep springs of action may themselves be sensitive to ongoing
reinforcement from interpersonal processes.
B. Moral Misapprehension and Differential Accessibility
This picture has darker shadows. Empirical evidence suggests that the
very interpersonal processes which ideally support ethical character
also expose it to discontinuity and disruption, in ways that may be
extremely difficult for the agent to detect. To illustrate, consider a
hypothetical example featuring justice.
Angela works in the public defenders office of a large city, with colleagues she respects and likes. She went to work there in the first place,
and stayed, because she saw that her colleagues shared her values.
In her conduct, Angela consistently demonstrates a strong commitmentto the ethical principle that justice requires equal treatment under
the law. Where justice for indigent defendants is concerned, she dedicates herself to making sure that the criminal justice system delivers to
35
S.M. Drigotas, S.W. Whitton, C.E. Rusbult and J. Wieselquist, Close Partner as
Sculptor of the Ideal Self: Behavioral Affirmation and the Michelangelo Phenomenon,
Journal of Personality and Social Psychology 77 (1999), pp. 293323.
36
The more superficial form of reinforcement is emphasized by Kamtekar,
Situationism and Virtue Ethics on the Content of Our Character, pp. 48891.
222
maria w. merritt
them the treatment to which they are entitled by law. She does so in the
right spirit (caring that justice be served), and for the right reason (that
justice requires equal treatment). In short, Angelas v alues, choices, and
reasons for action are unimpeachable with respect to justice.
But over time, the groups caseload gradually increases. In their
efforts to keep up, they begin to cut corners, but in each case they manage to rationalize their actions as ethically unproblematic. Without
quite realizing it, Angela too makes more and more choices that are in
fact inconsistent with what justice requires, going in for her share of
well-meaning but unwittingly self-serving rationalizations. Angela
continues to understand herself as dedicated to justice, but in her
actual conduct she is no longer just. Indigent defendants suffer the
consequences.
If there is a single way to characterize the downward spiral of mistakes that Angela and her colleagues have fallen into, it is misapprehension of the objects of choice and avoidance. In terms of Aristotles
account of acting for the sake of the noble with respect to the objects
of choice and the objects of avoidance, they have become prone to
choose actions they ought to avoidactions that are advantageous
but wrong (base)under misapprehensions of them as advantageous
and right (noble).
A sadly compelling real-life example is the scandal that recently surrounded Patricia Dunn, former Chairman of the Board at the HewlettPackard Company (HP). Dunn is an expert in corporate governance.
When she joined the HP board, she brought with her an impeccable
record of ethical responsibility in powerful positions such as sole global
CEO of Wells Fargo Investment Advisors, where she had been the
principal fiduciary for more than a trillion dollars in assets.37 As board
chairman at HP, in a well-intentioned campaign to contain high-level
leaks to the press, Dunn trusted her in-house legal and ethics team to
determine the lengths to which hired private detectives could go in
their investigations of board members.38 Unfortunately, the detectives
went so far as to draw felony counts of fraud and identity theft, not only
against themselves but also against Dunn.39 At issue was the practice of
37
James B. Stewart, The Kona Files, The New Yorker, 19 February 2007,
pp. 15267.
38
Stewart, The Kona Files, p. 159.
39
A judge later dropped all charges against Dunn. But by then the episode had
imposed a heavy cost in organizational turmoil, as well as an enormous personal toll
223
pretexting: assuming a false identity to get confidential information

about targets of investigation. On a plausible interpretation, Dunn and
her closest confidantes were so preoccupied with their goal of containing leaksitself an entirely above-board objective, which Dunn was
specifically hired to pursuethat they all too readily overlooked the
base and shameful (not to say criminal) aspects of the actions taken
under their authority.
Research in social psychology strongly suggests that such misapprehensions are an ordinary and predictable result of human cognitive
functioning. They are ordinary in two senses: first, they are generated
by the routine operation of normal cognitive processes; second, far
from being peculiar to a few unusual individuals whose character is
rotten through and through, ethical failures of this kind are ubiquitous,
even among well-meaning, intelligent, sensitive people.40 Be our ethical
commitments ever so correct and thoughtfully adopted, we are (virtually) all susceptible to choosing seemingly (to us) innocuous actions
which objective observers would easily recognize as instances of ethical failure, such as conflict of interest, in-group favoritism, or stereotyping and prejudice.41 For example, several findings about physicians
prescribing practices are well-supported by empirical research: first, in
aggregate, physicians prescription practices are significantly influenced by even minor p
harmaceutical promotions (such as providing
pens and notepads emblazoned with the corporate logo); second, a
majority of physicians self-report that their p
ractices are not influencedby such promotions; and third, an even greater majority believe
that other physicians, unlike themselves, are indeed influenced by
promotions.42 Together, these findings suggest that individuals can
upon Dunn herself. Benjamin Pimentel, Court Dismisses Charges against Former HP
Chair, The San Francisco Chronicle, 15 March 2007, p. C1.
40
H.M. Bazerman and M.R. Banaji, The Social Psychology of Ordinary Ethical
Failures, Social Justice Research 17 (2004), pp. 11115; M.R. Banaji, Ordinary Prejudice,
Psychological Science Agenda 14 (2001), pp. 811.
41
D.A. Moore and G. Loewenstein, Self-Interest, Automaticity, and the Psychology
of Conflict of Interest, Social Justice Research 17 (2004), pp. 189202; N. Dasgupta,
Implicit Ingroup Favoritism, Outgroup Favoritism, and their Behavioral
Manifestations, Social Justice Research 17 (2004), pp. 14369; L.A. Rudman, Social
Justice in Our Minds, Homes, and Society: The Nature, Causes, and Consequences of
Implicit Bias, Social Justice Research 17 (2004), pp. 12942.
42
Jason Dana and George Loewenstein, A Social Science Perspective on Gifts
to Physicians from Industry, Journal of the American Medical Association 290 (2003),
pp. 25255.
224
maria w. merritt
easily recognize conflict of interest in their peers behavior while being

oblivious to it in their own.
The explanation for such obliviousness is that many less-than-ideal
choices of action issue from psychological processes that operate
largely outside conscious awareness, selectively determining what
comes most readily to mind and so directing behavior in ways that may
run counter to the subjects reflectively endorsed values and ethical
self-understanding.43 Part of the empirical and theoretical basis for this
explanation is found in the longstanding research program on intuition, judgment, and decision-making led by Daniel Kahneman and
the late Amos Tversky, now carried on by Kahneman.44 Kahnemans
explanatory framework emphasizes differential accessibilitythe
ease (or effort) with which particular mental contents come to mind
among various mental contents relevant to the judgment or decision at
hand.45 In the following account, I will first discuss patterns of differential accessibility and their consequences for ethically important behavior. Then I will turn to the causal contributions made by several factors,
especially interpersonal processes, which operate largely outside the
purview of conscious awareness to determine different degrees of
accessibility.
Consider the misapprehension of an advantageous action as morally
right when it is actually wrong. The mental contents that come easily to
mind track the features of the situation that recommend a certain
action as advantageous. The mental contents that do not come to mind
would (if they did come to mind) track the features of the situation that
might make that action wrong. These missing mental contents would
have prompted the agent to seek further information, entertain alternative courses of action, weigh the trade-offs that might arise between
important considerations, and generally to engage in critical, reflective
deliberation. But since they dont come to mind, the agent fails to consider them, so that her choice of action is uninformed by them. The
action then appears to be both advantageous and, at worst, morally
neutral. So, the agent chooses the advantageous action under an erroneous description of it as morally acceptable.46
Bazerman and Banaji, The Social Psychology of Ordinary Ethical Failures.
Daniel Kahneman, A Perspective on Judgment and Choice: Mapping Bounded
Rationality, American Psychologist 58 (2003), pp. 697720.
45
Kahneman, A Perspective on Judgment and Choice, p. 699.
46
Alongside Kahnemans research program, another school of thought prominent
in the empirical literature on intuition, judgment, and decision-making is led by Gerd
43
44
225
The initial error of choosing an advantageous action under a misapprehension of it as morally acceptable can be seriously compounded
by overconfident faith in ones own good intentions.47 Broad characterological self-attributions further increase the hazard. The belief
I am a good person, or even I am trying my best to be a good person,
encourages the expectation I would not deliberately do anything
wrong. Even if this expectation were accurate, its content is quite distinct from that of a further expectation that can too easily slip into its
place during the everyday flow of the first-person perspective on our
experience and conduct: I would not do anything wrong, deliberately
or not. Even though we might not openly, in full awareness, credit ourselves with moral infallibilitythat would be immodestwe are prone
to limit our field of vision unwittingly, so that the effective reach of
critical self-monitoring extends only to those occasions when we do
take time to think carefully. Left out of consideration are occasions
when everything seems fine, and we are not in fact doing wrong deliberately, but had we deliberated we would have noticed decisively
wrong-making features of our course of action. Instead, if and when we
consider the latter type of occasion in retrospect, the conviction of
being a good person tends to displace deliberative effort into after-thefact rationalization, a selectively confirmatory survey of all the reasons
that seem to support the past decision. In hindsight, we may even
enhance our description of the action as not only morally neutral, but
downright laudable.48
Gigerenzer. See, e.g., Gerd Gigerenzer, Peter M. Todd, and the ABC Research Group,
Simple Heuristics That Make Us Smart (New York: Oxford University Press, 1999).
Gigerenzer has identified a form of intuitive decision-making, one-reason decisionmaking, which is ubiquitous in everyday reasoning and may often be useful, but which
might occasionally lead to moral misapprehension.When there are too many relevant
considerations to take into account in a limited amount of time, we recruit a single type
of consideration to stand in for the harder-to-determine, all-things-considered best
choice. See also Walter Sinnott-Armstrong, Moral Intuitions as Heuristics, unpublished manuscript.
47
Returning to the real-life example mentioned above, Patricia Dunn has reportedly said, There has never been a whiff of scandal or taint related to my activities,
particularly any issues concerning my integrity or ethics. Indeed, in the roles I have
held, any such taint would be an instant career-ender, and for good reason. Quoted in
Stewart, The Kona Files, p. 154. My point in quoting Dunn here is not to accuse her of
moral conceit, but rather to note that even seemingly well-founded confidence in the
quality of ones character can sometimes be ethically counterproductive.
48
John Darley, The Cognitive and Social Psychology of Contagious Organizational
Corruption, Brooklyn Law Review 70 (2005), pp. 117794; Elliot Aronson, The Social
Animal (New York: Worth Publishers, 9th edn, 2003), pp. 14399; R.M. Jones, Law,
226
maria w. merritt
Through such unfortunate trains of cognitive processes, we can

i nadvertently acquire a new predisposition to continue misapprehending future situations and actions of a similar type. In Aristotelean
terms, this is the piecemeal corrosion of practical wisdom, or, hardly
less worrisome, the piecemeal corrosion of its building blocks (in the
case of agents who are only partly in possession of incipient practical
wisdom). The practically wise agent knows how to e stimate the comparative value of all relevant goods and evils as they come into play in
any particular situation. It is part of practical wisdom, and a necessary
step on the way to acquiring it, that we should learn to estimate the
value of specific goods and evils as they crop up, one by one or cluster
by cluster, in the local practical milieu we happen to inhabit. When,
instead, faulty estimations of some specific goods and evils become
habitual, we may continue to profess a generalized love of noble actions
and revulsion toward base actions, yet inadvertently decouple those
attitudes from some of their correct objects of choice and avoidance.
In this way, our aspirations toward moral virtue can pull apart from
true practical wisdom. The cycle of self-reinforcement between moral
virtue and practical wisdom is broken.
Moreover, the overconfident conviction of ones own moral probity
can be bolstered by interpersonal dynamicswhich may be otherwise
ethically salutary, as we saw in part (A) abovethat shape, encourage,
and support ones self-understanding as a person of good character.
Experimental studies have found that subjects accounts of their actions
are more likely to be distorted by self-enhancing cognitive bias when
their actions are known to others and they feel responsible for the outcome.49 In the words of one pair of psychologists, The presented self is
(usually) too good to be true; the (too) good self is often genuinely
believed.50
Not only do such interpersonal factors exert an unhelpful influence downstream of the faulty judgments that agents initially make
under conditions of differential accessibility. More importantly, they
Norms, and the Breakdown of the Board: Promoting Accountability in Corporate
Governance, Iowa Law Review 92 (2006), pp. 105158.
49
G. Weary, Self-presentation and the Moderation of Self-serving Attributional
Biases, Social Cognition 1 (1982), pp. 14059; Aronson, The Social Animal, p. 140;
Jones, Law, Norms, and the Breakdown of the Board.
50
A.G. Greenwald and S.J. Breckler, To Whom is the Self Presented?, in
B.R. Schlenker (ed.), The Self and Social Life (New York: McGraw-Hill, 1985), pp. 126
45, quoted in Aronson, The Social Animal, p. 139.
227
contribute directly to differential accessibility in the first place, by

influencing what kind of cognitive processes are likely to direct an
agents behavior.
Different degrees of accessibility between mental contents are determined jointly by two types of causal factors: the characteristics of the
cognitive processes that produce those mental contents in ones mind,
and the characteristics of the stimuli and events that one encounters in
the situation at hand.51 Let us look first at cognitive processes. In the
kinds of ethical failure that psychologists have studied most extensively
(such as conflict of interest and in-group favoritism), the cognitive processes that bring the actions advantageous features to mind are typically fast, automatic, effortless, associative, implicit (not available to
introspection), and often emotionally charged.52 The cognitive processes that would bring the wrong-making features to mind (but do
not, because they do not occur at the right moment) are slower, serial,
effortful, more likely to be consciously monitored and deliberately
controlled.53
What is missing in many cases of misapprehension is the engagement of the slower, deliberative type of process to examine the situation with some critical distance, search for further relevant information,
reason through important considerations, and in general to monitor or
check the choice of action prompted by the faster, effortless type of
process. The capacity for sound ethical deliberation is of little use if we
fail to employ it on occasions that call for it. As the psychologist John
Darley puts it, people are ethical, but only intermittently so.54 Enter
the second type of causal factor in the explanation of differential accessibility: characteristics of stimuli and events in the situation at hand.
Given that the slower, deliberative type of cognitive process is often
necessary to bring relevant moral considerations to awareness, what
conditions tend to stultify it? And what conditions tend to awaken it?
Kahneman, A Perspective on Judgment and Choice, p. 699.
Kahneman, A Perspective on Judgment and Choice, p. 698. See also Moore and
Loewenstein, Self-Interest, Automaticity, and the Psychology of Conflict of Interest.
53
Kahneman, A Perspective on Judgment and Choice, p. 698. It is important, however, not to over-generalize about the type of cognitive processing typically associated
with moral judgment. Even as many aspects of moral judgment draw on slow, deliberative cognitive processing, evidence suggests that other aspects are sub-served by fast,
automatic cognitive processing. See, e.g., Joshua D. Greene, et al., The Neural Bases of
Cognitive Conflict and Control in Moral Judgment, Neuron 44 (2004), pp. 389400.
54
Darley, The Cognitive and Social Psychology of Contagious Organizational
Corruption, p. 1185.
51
52
228
maria w. merritt
Psychological studies of accountability in judgment and decisionmaking open a revealing window onto differential accessibility.
Accountability is defined, for the purposes of assessing these studies,
as the implicit or explicit expectation that one may be called upon to
justify ones beliefs, feelings, and actions to others.55 It turns out that
accountability simpliciter may or may not awaken the effortful cognitive activity of deliberation. Whether it will do so depends, in part, on
the expected characteristics of the audience to whom one is accountable. Numerous experimental and field studies have shown that expecting to be accountable to an audience whose views one knows ahead of
time will stultify cognitive effort, as compared with expecting to be
accountable to an audience whose views one does not know.56
When experimental subjects are asked to make judgments or decisions, and they know the preferences of the audience who will evaluate
the result, their behavior conforms to the use of cognitive shortcuts
guided by their beliefs about what the audience will find acceptable,
bypassing the cognitive effort required to assess the object of judgment independently.57 Subjects who do not know the preferences
of their audience, by contrast, anticipate the need to justify their judgment or decision, prompting a critical, reflective internal dialogue that
awakens deliberation. These subjects more readily tolerate evaluative
complexity and exercise care in considering complex information,
weighing the reasons for and against alternative options, and recognizing trade-offs. Accountability to an audience with unknown views has
been shown significantly to reduce subjects tendency to proceed with
undue haste and substitute oversimplified, superficial, or misleading
decision procedures for effortful, systematic, self-critical deliberation.
It has also been shown to attenuate the effects of a wide variety of cognitive biases, such as oversensitivity to the order in which information
appears.58
55
J.S. Lerner and P.E. Tetlock, Accounting for the Effects of Accountability,
Psychological Bulletin 125 (1999), pp. 25575, at p. 255.
56
Lerner and Tetlock, in Accounting for the Effects of Accountability, provide a
comprehensive review and analysis of the extensive body of research on accountability
and its impact on judgment and decision-making.
57
Lerner and Tetlock, Accounting for the Effects of Accountability; J. Pennington and
B.R. Schlenker, Accountability for Consequential Decisions: Justifying Ethical Judgments
to Audiences, Personality and Social Psychology Bulletin 25 (1999), pp. 10671081.
58
Lerner and Tetlock, Accounting for the Effects of Accountability; see especially
Table 1, p.260.
229
But in practice, we cannot avoid making many of our judgments

and decisions under conditions in which we are thoroughly familiar
with the views of others to whom we are accountable. We often hold
important values, interests, and goals in common with them. Is there
any way to keep ourselves cognitively awake and alert under these conditions? Further studies of accountability have found that the potentially stultifying effect of having an audience with known views can be
mitigated by inducing a motivational orientation toward accuracy in
the decision-maker, as contrasted with a motivational orientation
toward social compatibility.59
For instance, one study manipulated the conditions under which
subjects had to make a business decision between two options for allocating a $10 million marketing budget: stay with a past decision that
had yielded disappointing results, or initiate an alternative likely to
improve the return on investment.60 Between two groups of subjects,
one group induced to be m
otivated by accuracy and the other induced
to be motivated by social compatibility, some subjects in each group
were told that they would have to justify their recommendations in a
face-to-face meeting with another person who endorsed staying with
the past, disappointing decision. Other subjects were told about this
other persons views, but were assured they would never have to meet
the person, nor would they have to explain and justify the present decision. Thus there were four combinations of experimental conditions:
accuracy motivation plus accountability to an audience with known
views, compatibility motivation plus accountability to an audience
with known views, accuracy motivation plus no accountability, and
compatibility motivation plus no accountability.
The results were that in all but one of the experimental combinations, subjects tended to conform to the known views of the other
person, to the detriment of return on investment. Only under conditions of accuracy motivation plus accountability did subjects escape
the influence of the other persons known endorsement of a plan
to throw good money after bad. These were the only subjects whose
59
S. Chen, D. Schechter and S. Chaiken, Getting at the Truth or Getting Along:
Accuracy-Versus Impression-Motivated Heuristic and Systematic Processing, Journal
of Personality and Social Psychology 71 (1996), pp. 26275; A. Quinn and B.R. Schlenker,
Can Accountability Produce Independence? Goals as Determinants of the Impact of
Accountability on Conformity, Personality and Social Psychology Bulletin 28 (2002),
pp. 47283. See also Jones, Law, Norms, and the Breakdown of the Board.
60
Quinn and Schlenker, Can Accountability Produce Independence?
230
maria w. merritt
decision-making was comparable to that of control subjects who had

no exposure whatsoever to anyone elses recommendations. An explanation for such results in terms of differential accessibility is suggested
by other studies showing that subjects with a social compatibility motivation are more likely to rest content with the low-effort cognitive
shortcut of mirroring a discussion partners opinions, while subjects
with an accuracy motivation are more likely to engage in effortful,
systematic reasoning, arriving at opinions unbiased by those of their
discussion partner.61
It is important to distinguish between the motivation to arrive at
an accurate judgment and the concern to be a certain kind of person,
the kind who is an independent thinker. While the accuracy motivation demonstrably improves decision-making performance, the concern to be an independent kind of person, with its characterological
slant, can distort behavior through the influence of self-presentation.
In one series of studies a sub-group of experimental subjects, whose
desired selves were distinguished by an emphasis on forming opinions
independently of what others think, were indeed more likely not to
conform under conditions that elicited conformity from other, less
independent-minded subjects. But the self-professed independents
took it too far, going out of their way not to conformeven to the point
of publicly reporting attitudes that differed from their private beliefs,
if such public reports would promote the impression that they were
independent.62
Accountability to independent, external review is the institutional
counterpart of the individual accuracy motivation. Well-governed
institutions should structure incentives for individuals accordingly.63
For example, in the wake of twentieth-century scandals involving
medical research with human subjects, the protection of subjects rights
and welfare is no longer left entirely to the personal character of
scientific investigators, however well-intentioned; under international
Chen, Schechter and Chaiken, Getting at the Truth or Getting Along.
Schlenker and Weigold, Self-consciousness and Self-presentation, pp. 82028.
63
However, for some domains, the empirical literature arguably suggests that selfinterested biases operate so far below the radar of conscious, deliberate cognition that
it may be more effective simply to remove the distorting influences altogether. For
instance, if the evidence shows that neither the disclosure of physicians involvement
with pharmaceutical companies nor limits on the size of promotional gifts to physicians will work to correct bias in prescribing practices, a policy of outright prohibition
might be preferable. Dana and Loewenstein, A Social Science Perspective on Gifts to
Physicians from Industry, p. 254.
61
62
231
guidelines and under the law of many countries, it is subject to prospective and ongoing independent review, prompting investigators to
think more sensitively and systematically about the ethical aspects of
their research.64 Similarly, the empirical literature on decision-making
can inform the design of better mechanisms to promote accountability
on the part of corporate leaders.65
IV. Conclusion: Virtue Reconsidered
For people who care about living up to their own values, the activity
known to psychologists as self-regulationmonitoring ones movement toward or away from the desired self is practically inescapable.
The psychological evidence suggests that self-regulation influences
conduct partly through its engagement with the ongoing social
demands of self-presentation, often through cognitive processes that
are not accessible to first-person introspection or control.
The influence of these social demands may edify personal character,but may also disrupt iteven when exerted through relationships
with the same people on both counts, and even if both you and those
other people (let us assume for the sake of argument) have the right
values in principle. When your internal sense of ethical accountability is pervasively influenced by your relationships with close associates whose values you know, admire, and sharealong the lines of
the ideal Aristotelean social worldyou run the risk of living in an
incubator for your own cognitive complacency.66 Even if the shared
values are actually right, what threatens to dull your deliberative edge
is your comfortable familiarity with the attitudes held in common.
Familiarity breeds a pattern of overconfident, low-effort, intuitive
decision-making, which may sometimes fail to track important moral
64
Henry K. Beecher, Ethics and Clinical Research, New England Journal of Medicine
274 (1966), pp. 1354360; Allan M. Brandt, Racism and Research: The Case of the
Tuskegee Syphilis Study, Hastings Center Report 8 (1978), pp. 2129; National
Commission for the Protection of Human Subjects of Biomedical and Behavioral
Research, The Belmont Report, U.S. Government Printing Office, 18 April 1979; U.S.
Code of Federal Regulations Title 45, part 46, Protection of Human Subjects; Council
for International Organizations of Medical Sciences (CIOMS), International Ethical
Guidelines for Biomedical Research Involving Human Subjects (Geneva: World Health
Organization, 2002).
65
Jones, Law, Norms, and the Breakdown of the Board.
66
Contrast the contemporary ideal of social diversity in educational environments.
232
maria w. merritt
considerations. Being such as to make the right choices when you do

think through your actions does not guarantee that you will in fact
think through all of the actions you ought to think through.
Add to this an explicitly characterological outlook on self-regulation
and self-presentation, manifested in ordinary circumstances of judgment and decision-making. Such an outlook can exacerbate a common
cognitive pre-commitment to credit yourself, in broadly characterological terms, with doing your best to be a good person, thus biasing
retrospective self-assessment in your favor, and preparing the ground
for future repetitions of every unwitting moral misapprehension.
But there is room for hope. The evidence suggests that cognitive
complacency is remediable to a significant extent by the motivation to
reach judgments and decisions that are well-founded and will stand up
to critical scrutiny. And it may be possible to modulate ones emotional
investment in the wish to be agood person, at least enough to reduce
the risk of overlooking ones own fallibility.
To revisit our pivotal question about moral theory and scientific psychology (section I above), what import do these inductive generalizations have for virtue ethics? Do they merely lengthen the (already long)
diagnostic list of ways in which people are liable to fall short of the
ideal, or do they indicate some constraint on viable content for the normative psychology of virtue?
The problem made salient by the empirical generalizations discussed
above is a common, well-intentioned inflection of the first-person perspective on ethical life: framing the ethical commitments that anchor
self-regulation primarily as ideals of personal character. To be sure,
ordinary notions of personal character and being a good person are
certainly not the equivalent of p
hilosophical virtue ethics. Nonetheless,
what philosophical virtue ethics recommends to the reflective agent is
recognizable as an intellectually disciplined cousin (and in some cases
a cultural progenitor) of the ordinary notions whose routine involvement in self-regulation invites the troubling outcomes discussed in
section III.B above. Orthodox virtue ethics, unlike moral theories in
the consequentialist or Kantian families, grounds the criterion of right
action in some conception of what an idealized yet fully human agent
would do. This builds into first-person ethical practice, as fundamental
and indispensable, a characterological orientation that is entirely discretionary for consequentialist or Kantian ethical practice.
Moreover, given that a motivation for accuracy can remedy cognitive complacency, consequentialism and Kantianism each provide a
233
sharply defined target for that motivation to home in on. Of course,

this in itself does not vindicate either type of theory as correct. And
obviously, no credible moral theory can interpret itself with respect to
real situations and problems. The work of deliberation remains for the
agent to do. The point is that virtue ethics makes extremely ambitious
cognitive demands on the aspiring agent, in that it burdens the accuracy motivation with so many disparate considerations to track and
organize under the abstract idea of what the virtuous person would do.
In order to ease the cognitive burden to the point of offering a reliable
guide to first-person ethical practice, advocates of virtue ethics may
be well-advised to make the way more navigable between that abstraction and the everyday arrival at choices of action.
With its ideal of personally embodied practical wisdom, virtue ethics offers a criterion of right action intended to register the importance
of multiple goods and evils in a unified rational perspective. Although,
strictly speaking, the theory does not prescribe cognitive shortcuts,
recommending instead that agents engage in deliberation to whatever
degree is necessary to track all the goods and evils in play, the imagined
exemplary figure of the virtuous person lends itself to oversimplified,
self-enhancing cognition. All the more so when the exemplary figure
you imagine is a composite drawn partly from (aspects of) real people,
with whom you have close emotional or collegial bonds, and who regularly affirm your own understanding of yourself as a good person. The
cognitive shortcut of doing what you expect your admired associates
to approve of, using them as stand-ins for the virtuous person or the
person of practical wisdom, becomes particularly tempting. But that
shortcut, simply because it is a shortcut substituted for systematic
deliberation, will not reliably lead to the right choice of action.
In order for virtue ethics to offer reliable guidance for first-person
ethical practice, it may need to introduce an explicit shift of emphasis
between the criterion of right action and the mode of deliberation
recommended for everyday life. The criterion of right action is defined
primarily in terms of what the virtuous person would do, and specified
in detail this will encompass and unify a variety of substantive moral
considerations. But how the aspiring agent ought to deliberate may
be better articulated by some sort of explicit procedure for systematically identifying and thinking through those substantive considerations, as right-making and wrong-making features of particular
prospective actions. The mental contents that ought to come to mind
those that enable us to home in on the target of the motivation for
234
maria w. merritt
moral accuracyare not adequately captured in the summary notion

of what the virtuous person would do, which well-intentioned agents
all too readily bring to mind and use without bothering to include the
specification necessary to flesh out the summary. Rather, the appropriate mental contents should, as experienced, approximate as closely as
possible how the virtuous person would think, and thisrequires that the
agent be attuned primarily and systematically to the substantive goods
and evils themselves.
The idea of what the virtuous person would do does not in itself
suffice to guide the agent in tracking and organizing substantive moral
considerations as the virtuous person would. Indeed, neither does that
idea in combination with a list (also abstract) of the considerations to
which, in general, the virtuous person would be sensitive. What the
aspiring agent needs, inaddition, is more explicit guidance for charting
particular courses of action by reference to those considerations.
Whereas for Aristotle himself, such guidance was to be provided by
mentors who would take the aspiring agent under their wing and by
friends who would form an audience of like-minded peers, the empirical evidence surveyed above suggests that the agent will need to refer as
well to some guiding procedure that remains independent of his or her
interpersonal relationships. To formulate such guidance is an especially demanding challenge for virtue ethicists, so far as they hold (for
good reason, in my opinion, although I will not argue for it here) that
the plural nature ofimportant ethical considerations makes decisionmaking algorithms impossible.67
In conclusion, what the evidence about interpersonal processes
reveals is a psychological need for systematic decision-making guidance, if aspiring agents are to adopt virtue ethics as a reliable guide to
first-person ethical practice. This psychological need does not necessarily constrain the normative content of the criterion for right action,
but it does place a serious and challenging demand on virtue ethics to
articulate some form of stand-alone decision procedure, comparable to
the procedures that consequentialist and Kantian moral theories can
more straightforwardly provide.
67
Cf. Hursthouse, On Virtue Ethics; Christine Swanton, Virtue Ethics: A Pluralistic
View (Oxford: Oxford University Press, 2003).
235
Aknowledgments
Credit for significant financial, institutional, and collegial support in
the production of this article is due to the Edmond J. Safra Foundation
Center for Ethics at Harvard University, where I held a Faculty
Fellowship in 2005-06. For thoughtful reading and discussion of an
earlier draft, I thank Dennis F. Thompson and the Faculty Fellows
seminar, and especially Renee M. Jones for extensive discussion,
exchange of ideas, and fresh insights into the psychological literature.
Precious Eboigbe provided excellent research assistance at Harvard.
Grateful acknowledgment for discussion and comments at various
stages of writing is also due to the following individuals and groups:
Neera Badhwar, Lorraine Besser-Jones, Alan Fuchs, Maximilian de
Gaynesford, Eric Hutton, Mark Leary, and Gopal Sreenivasan; the
Moral Psychology Research Group, especially John Doris, Gilbert
Harman, and Walter Sinnott-Armstrong; the Department of Philosophy
at Wake Forest University; and the audience at the Rocky Mountain
Virtue Ethics Summit, especially Rosalind Hursthouse, Karen Stohr,
Christine Swanton, and Kit Wellman. Some financial support for work
at earlier stages was provided by a Summer Research Grant at the
College of William and Mary.
Good but Not Required?Assessing the Demands

of Kantian Ethics*
Jens Timmermann
There is a strong sentiment in pre-philosophical moral thought that
actions can be morally valuable without at the same time being morallyrequired. For example, it is said to be admirable if a soldier throws
himself on a live hand-grenade to save his comrades, but it is arguably something we cannot expect him to do. A doctors volunteering
to give medical aid in a plagueridden town similarly seems to be a fine
thing, yet most of us would be unwilling to demand this degree of selflessness of her, or to blame her if she falls short of this ideal. Such
actions are widely regarded as heroic or saintly; and not everybody, so
the argument goes, need be a saint or a hero. Moreover, there might be
smaller things that are morally commendable while at the same time
going beyond the call of duty, such as buying a toy for the child of a
friend.1
Yet Kant, who takes great pride in developing an ethical system
firmly grounded in common moral thought, makes no provision for
any such extra-ordinary acts of virtue. The category of supererogatory
action is conspicuously absent from the table of moral categories of
freedom in his Critique of Practical Reason (V, p. 66); and he proposes
a classification of actions as either obligatory, permissible or prohibited
that in the eyes of one of his distinguished critics is totally inadequate
to the facts of morality.2 In the few passages in which the notion of
supererogation is actually mentioned Kant dismisses the idea of grand
* I should like to thank audiences in Stirling, Cracow and St Andrews, as well as
Rowan Crufts, Robert Louden and Thomas Hill, for comments on earlier drafts of this
paper.
1
Cf. J.O. Urmson, Saints and Heroes, in Joel Feinberg (ed.), Moral Concepts
(Oxford: Oxford University Press, 1969), pp. 6073 (first published in 1958). It should
be noted that Urmson (and others) seem to consider an extraordinary amount of selflessness the mark of supererogatory action. This assumption can by no means be taken
for granted. For those who believe in supererogation there can presumably be quite
ordinary acts that are still beyond the call of duty, as the last example shows.
2
Urmson, Saints and Heroes, p. 60. Cf. Elizabeth M. Pybuss defence of the tripartite division: Saints and Heroes, Philosophy 57 (1982), pp. 19399.
238
jens timmermann
and noble deeds as high-flown emotional nonsense.3 It is indeed difficult to see how within Kants general doctrine of duty any act could be
morally admirable or commendable but not at the same time required.
Acts of wide or imperfect dutya sub-class of which are seenastheclos
est thing to supererogation by one prominent modern interpreter4
are no exception. Kant makes no provision for going beyond the call of
duty because duty leaves the agent little room for choice:
[W]here the moral law speaks, there is, objectively, no further room for
free choice with regard to what there is to be done (Critique of Judgement,
V, p. 210).
Such considerations give rise to the fear that actions intuitively classed
as morally commendable but not required must be re-classified as
commands of duty by Kant, making his ethical theory as unbearably
demanding as certain strict variants of direct utilitarianism.5
3
Cf. Critique of Practical Reason, V, pp. 8485: By exhortation to actions as noble,
sublime and magnanimous, minds are attuned to nothing but moral enthusiasm and
exaggerated self-conceit; by such exhortations they are led into the delusion that it is
not dutythat is, respect for the law whose yoke (though it is a gentle one because
reason itself imposes it on us) they must bear, even if reluctantlywhich constitutes
the determining ground of their actions, and which always humbles them inasmuch as
they observe the law (obey it), but that it is as if those actions are expected from them,
not from duty as bare merit; and V, p. 157. Kants warnings are mainly psychological
and do not as such settle the question of how much duty demands. Preaching romantic
ideals distracts from plain and simple duty and encourages a view according to which
some things in morality are less stringent than others, which directly conflicts with the
categorical nature of all moral commands.
4
Thomas Hill in his seminal article Kant on Imperfect Duty and Supererogation, in
idem, Dignity and Practical Reason (Ithaca: Cornell University Press, 1992), pp. 14575
(first published in Kant-Studien 61 [1971]).
5
I.e. variants of direct utilitarianism that do not admit of supererogation.
It should be noted that even in pre-philosophical moral thought the ascription of
an act as supererogatory is one often not shared by the agent, especially in cases of
so-called saintly or heroic acts. The doctor referred to in the opening paragraph
is unlikely to say that practising in a plague-ridden town is a particularly fine thing
to do for a doctor who goes beyond the call of duty. Rather, she would probably refer
to the importance of saving peoples lives, or that she could not just leave them to
die, etc. Similarly, the soldier would probably say that he could not let his comrades
down, and that any of them would have done the same thing in his situation. Such
actions are regarded as heroic or saintly only from the third-person perspective. This
seems at least some indication that supererogation is not as firmly grounded in
common sense as some people think. The more ordinary case of buying a toy for
the child of a friend seems to allow of more room for judgement. It is also worth notingthat the phrase that a certain act cannot be demanded of a person is ambiguous.
Those who deny the possibility in supererogation are not committed to saying that
individuals therefore have a right to request the performance of acts of duty from
others.
good but not required?239
Kantian moral philosophers today are well aware of these problems.

Thomas Hill, Marcia Baron and David Cummiskey have vigorously
debated the difficulties surrounding the concept of supererogation in
the course of the past 15 years, but there is as yet no consensus.6 In this
paper, I wish to resume and, if possible, conclude the debate. There is
ample textual and philosophical evidence for the thesis that, within a
recognisably Kantian framework, moral choice is rather limited, and
that there can be no actions that are morally good but not required.
The paper divides into three sections. We shall first examine the
nature of moral goodness from a meta-ethical angle, introducing
some passages from Kants writings that present strong theoretical evidence against the case for supererogatory action. Secondly, we shall
pursue Thomas Hills suggestion that within the category of wide duty
we can accommodate some of the main features of actions classified as
supererogatory in other ethical systems: that specific acts falling under
some wide duties can be morally good but not required. We shall
conclude that, contra Hill, there are no actions of wide duty that can be
so characterized in any significant sense. On the basis of a clearer
understanding of why Kant was so reluctant to take the notion of supererogation on board, we shall finally, in the third section, address the
problem of how demanding the requirements of Kantian ethical theory
really are.
1. The Implications of Moral Goodness
Kants philosophy of value provides a strong argument against the idea
that heroic, saintly or indeed any other kind of action should be classified as morally good but not at the same time required. To begin with,
the famous opening passage of the Groundwork for the Metaphysics of
Morals declares moral goodness to be the only variety of goodness that
is unconditionally good (IV, p. 393). The good will possesses the kind of
The sum of recent scholarship is Thomas Hills comprehensive study Meeting
Needs and Doing Favors, in idem, Human Welfare and Moral Worth (Oxford: Oxford
University Press, 2002), pp. 20143. He discusses the criticism of Cummiskey and
Baron, whose contributions are in turn replies to his Kant on Imperfect Duty
and Supererogation. Chapter 1 of Marcia Barons Kantian Ethics Almost without
Apology (Ithaca: Cornell University Press, 1995) was first published as Kantian Ethics
and Supererogation in The Journal of Philosophy 94 (1987). For Cummiskeys contribution, cf. chapter 6, The (Not So) Imperfect Duty of Beneficence, in Kantian
Consequentialism (Oxford and New York: Oxford University Press, 1996).
6
240
jens timmermann
authority we can never take lightly in practical deliberation. It is even

the very condition of all other kinds of value. So how could unconditionally good action be optional?
A review of some passages in which Kant explains his conception of
moral goodness confirms this impressionthe claims of morality
appear to be so weighty that there can be no good reason ever to reject
a proposed moral action. In a passage from the Common Saying Kant
distinguishes prudential goodness and moral goodness. Whereas prudence depends on a variety of given ends and therefore admits of a
more and a less, a prudent action is good only on condition of my
moral worth as an agent. My ends must not conflict with moral goodness. As soon as they collide with the moral law of duty, I rationally
prefer the state of recognising goodness; but it is
not merely a better state [as his critic Christian Garve alleges, JT] but the
only one that is good in itself; it is a good from another sphere altogether,
where ends that may present themselves to me (and so too their sum,
happiness) are not taken into consideration at all, and where it is not the
matter of choice (a pre-given object) but the mere form of the universal
lawfulness of its maxims that constitutes its determining ground (The
Common Saying, VIII, p. 283).
Moral goodness always takes precedence over any other kind of value in
action, and not just that. Both subjectively and objectively, it silences or
eclipses any other claim to value, i.e. in a given situation nothing conflicting with moral value can have any value at all. The morally good
action is not just better, it is the only good option available to the agent.
There is nothing to be said for the prudent course of action. Kant
states that in deliberation I must first be certain that I am not acting
against my dutyonly afterwards am I permitted to look around for
happiness (ibid.).
In his Critique of Practical Reason Kant makes a similar point, this
time put forward in terms of the rationality of actions. This is his evocative chemical example of calcareous earth. Like a chemist, the philosopher can conduct an experiment with any mans practical reason
in order to distinguish the pure determining grounds of morality from
the empirical ones of inclination by adding the moral lawto the
7
What is the person motivated by who, contrary to both duty and momentary
desire, acts in accordance with his or her long-term interest? Even if prudence generates a motive analogous to respect in the moral case, as Kant sometimes seems to
empirically affected will.7 As usual, his example is that of the person

who would gladly lie because he can gain something by it. Kant
continues:
When an analyst adds alkali to a solution of calcareous earth in hydrochloric acid, the acid at once releases the lime and unites with the alkali,
and the lime is precipitated. In just the same way, if a man who is otherwise honest (or who just this once puts himself only in thought in place
of an honest man) is confronted with the moral law in which he cognises
the worthlessness of a liar, his practical reason (in its judgement of what
he ought to do) at once abandons the advantage, unites with what maintains in him respect for his own person (truthfulness), and the advantage,
after it has been separated and washed from every particle of reason
(which is now altogether8 on the side of duty), is weighed by everyone, so
that it can enter into combination with reason in other cases, only not
where it could be opposed to the moral law, which reason never abandons but unites with most intimately (Critique of Practical Reason, V,
pp. 9293).
The details of the chemical analogy need not concern us.9 The general
idea is sufficiently clear. Rationality determines the value of any kind of
action. As long as an agent considers possible courses of action from a
purely prudential point of view, reason points in the direction of his or
her advantage, of what is good for him or her. Moreover, he or she can
weigh and compare the various options: there can be a more and a less
in prudential matters.10 But moral reasons are different. They do not
acknowledge (cf. e.g. Refl. 1028), it cannot be effective in this case. Kants theory of
happiness-based motivation remains obscure.
8
Reading nun gnzlich for nur gnzlich.
9
Most of us will have a rough impression of what happens in experiments like
thatcountless chemistry lessons at school prove useful at last.
10
As one of the sentences leading up to the passage quoted indicates, the reason
why prudential judgements, unlike moral judgements, admit of degrees is to be found
in the empirical aspects involved in the pursuit of happiness: The distinction of the
doctrine of happiness from the doctrine of morals, in the first of which empirical principles constitute the whole foundation whereas in the second they do not make even the
smallest addition to it, is the first and most important task incumbent upon the
Analytic of pure practical reason, in which it must proceed as precisely and, so to speak,
as scrupulously as ever the geometer in his work [als je der Geometer in seinem Geschfte.
M. Gregor erroneously seems to read als jeder Geometer, as any geometer, and mistranslates the phrase accordingly, JT] (V, p. 92). The realization of our desires depends
on our knowing technical laws, which feed into hypothetical imperatives. They tell us
how to realize ends we happen to have. This is a highly complex empirical task.
Knowing what will make you happy in the long run is a matter of experience, and so is
the search for suitable means to that end. These means can be more or less suitable to
satisfy desires that can be more or less urgent. Kant optimistically contends that there
242
jens timmermann
just outweigh other, non-moral considerations. As soon as conflicting moral reasons enter the stage, the scene changes dramatically.
Moral considerations are accompanied by the characteristic rational
necessity of categorical judgements and consequently do not leave even
a single particle or appendage (Anhngsel) of rationality with what
previously appeared to be the prudent course of action favoured by
reason.11 The moral action alone is rational and good; all other value
depends on the unconditional value of moral goodness. Moral goodness does not just trump any other kind of practical value but completely annihilates it. (Note that makes Kants thesis of the necessity of
moral commands much stronger than standard modern overridingness.)12 If so, things are not looking good for supererogation. An action
correctly judged to be morally good cannot rationally be rejected in
favour of one that is not.
This may seem to leave open the possibility of rejecting a morally good
action (one that is not required) in favour of another action that is less
worthy but still morally good. But we have already encountered some
evidence that for Kant this is not a viable option either. The above
quotations from the Common Saying and the second Critique also
express the thought that the moral goodness defined by the categorical
imperativeas opposed to the prudential and presumably technical
goodness of hypothetical imperativesdoes not admit of degrees. It is
unconditional and makes an action necessary. There cannot be a lesser
kind of goodness that those engage in who do not opt for the extravagance of supererogatory acts.
There is further confirmation of this in a well-known passage from
the Metaphysics of Morals discussing the possibility of conflict amongst
different morally good courses of action. On strictly conceptual
can be no such vagaries in the moral sphere if moral judgement rests on pure a priori
principles.
11
The option that maximizes satisfaction of the agents desires in the long term presumably does not even deserve to be called the prudent choice any more (if the term
is supposed to imply normative approval).
12
This point is often overlooked, and rarely endorsed, by todays Kantians. A good
example is David Brinks Kantian Rationalism: Inescapability, Authority, and Suprem
acy, in Garrett Cullity and Berys Gaut (eds.), Ethics and Practical Reason (Oxford:
Clarendon Press, 1997), pp. 25591. Brink carefully discusses Kantian ethics in terms
of the three characteristics of moral judgement mentioned in the title of his paper but
stops short of the strong supremacy of moral value that renders all conflicting claims to
value null and void.
grounds, Kant rules out that there can be a conflict of duties. As duty
and obligation generally are concepts which express the objective practical necessity of certain actions, and two rules opposing each other
cannot both at the same time be necessary,13 it follows that a collision
of duties and obligations is unthinkable. However, there may be a conflict of grounds of obligation, one of which is not sufficient for obligation. If two such grounds conflict, the stronger groundnot the
stronger obligation!wins out to constitute a duty all things considered: fortior obligandi ratio vincit (Metaphysics of Morals, VI, p. 224).
The judgement of reason is unequivocal, even in the case of practical
conflict.14 It is true that other kinds of duty may follow from the vanquished ground. If I am unable to help a friend because my duty lies
elsewhere I may well incur a duty to explain the situation to her. But
when one ground of obligation loses out against another stronger
ground the latter does not constitute a lesser duty, but rather no duty
at all. Strictly speaking, the term duty should thus be used in the singular only, for all-things-considered token duties. This is the sense in
which a case of supererogatory action would have to go beyond the call
of duty. But there is no such case in Kantian ethics.15
In sum, the passages discussed present us with the following picture:
By virtue of the necessity of categorical judgement, moral value, all
things considered, (i) is absolute and does not admit of degrees; (ii) it
takes precedence over other kinds of practical value to the extent
that (iii) it even renders all other claims to practical value null and void;
and (iv) duty can be unequivocally determined even in cases of apparent moral conflict. We must conclude that there are no token actions
within Kants moral theory which are morally good but not required.
Good but not enforceable?Yes, that is just what duties of wide obligation are. Good prima facie or for the most part, but not all things
considered?Certainly, that will frequently be the case. Morally good
but not required?No, that would be a contradiction in terms.
13
In parentheses he adds: rather, if it is a duty to act in accordance with the one it is
not only not a duty to act in accordance with the opposed one, but even contrary to
duty.
14
It is, of course, conceivable that two actions are equally permitted. Two (or more)
actions can be permissible while it is obligatory to do either.
15
Morality may leave things undecided. In that case, two or more actions can be
permissible, but there would still be no opportunity for the agent to distinguish himself
as a saint or a hero. He would still have to do the one action or the other.
244
jens timmermann
2. Strict Duty, Wide Duty, Supererogation
Applying the commands of what is variously called strict, perfect or

unremitting duty is comparatively straightforward. They are negative
duties commanding specific acts, and they do not admit of any exceptions (or so it would seem). By contrast, wide duties are said to allow
for some discretionary judgement on the part of the agent.16 This is the
reason why Thomas Hill argues that the wider types of imperfect duty
share some essential features of supererogatory action and can help to
accommodate our intuitions about actions that are morally good but
not required.17 Let us therefore have a closer look at the kinds of duties
there are, and at their relation to supererogation.
As so often, understanding an important aspect of Kants ethics
involves interpreting one of his footnotes:
It must be noted here that I reserve the division of duties entirely for a
future Metaphysics of Morals, so that the division here stands only as one
adopted as convenient [beliebig]18 (for the sake of arranging my examples). For the rest, I understand here by a perfect duty one that admits of
no exception in favour of inclination (Groundwork, IV, p. 421 n.).
This has been taken as proof for the thesis that imperfect duties, as
opposed to perfect duties, do admit of exceptions in favour of inclination.19 If so, maybe imperfect duties tell us to do actions that somehow
are morally good but not required. However, Kant does not say that
imperfect duties may sometimes be set aside in the interest of inclination. Rather, he denies that perfect duties allow of any such exceptions.
Saying the latter could admittedly be taken to imply the former; but
this conclusion is by no means necessary.
16
Also, wide duties are said to be meritorious, whereas strict duties are owed.
In the case of wide duties, the object of my duty (e.g. a person in need of help) does
not have a right to my action (my helping him) even though I do have a duty to help
him.
17
Cf. Hill, Meeting Needs and Doing Favors, pp. 203ff.; Imperfect Duty and
Supererogation, pp. 149ff.
18
Here the word beliebig does not mean arbitrary (Paton), as it does in standard
German today. Kant does not wish to say that the present classification is without foundation. That would be highly surprising. Rather, the present classification is provisional
and inadequate for the purpose of a metaphysics of morals because it does not contain
the distinction between duties of right and duties of virtue.
19
Cf. Hill, Imperfect Duty and Supererogation, p. 148; idem, Meeting Needs and
Doing Favors, p. 214 and n. 34; and Mary Gregor, Laws of Freedom (Oxford: Basil
Blackwell, 1963), p. 96.
An interpretation along the following lines seems much more likely.

The above characterization of strict duties is trying to make the general
point that strict duty cannot be outweighed by any apparently morally
worthy end whatsoever. If you commit yourself to an end by incorporating it into your maxim, this will occasionally produce an incentive
for action when presented with cases that fall under that maxim. As
maxims of wide obligation aim at furthering an end, they temporarily
effect incentives similar to direct inclinations, e.g. the inclination to
help somebody; and the footnote referred to asserts that strict duties do
not admit of exceptions, not in the interest of wider obligation nor in
the interest of any other such inclination. Kant should have used the
term incentive (Triebfeder) rather than inclination (Neigung), a word
he usually reserves for a particular non-moral type of incentive. He did
not. Rather, he is lumping together everything that may seem to interfere with strict duty.20 Also, as we shall soon see, inclinations may play
a legitimate, if limited, role in making wider duties more specific.
Nevertheless, there cannot be exceptions in favour of inclination, i.e.
suspensions of duty in favour of an inclination.
Incidentally, there is an intriguing philosophical reason for the primacy of duties of strict obligation over duties of wide obligation.
Kants Groundwork tells us that not being able to act on maxims which
could not be thought a universal law without contradiction constitutes
a strict duty not to act on them. Deceitful promising, for instance,
could not even be thought a universal law because the universality of
such a maxim would thwart the very end of ones action. One has to
rely on the maxims not being universally valid to be able to act on it at
all; one has to make an exception for oneself, and that is why acting on
such a maxim is strictly immoral.21 By means of contrast, a world in
which everyone acts on the maxim of never helping anyone is not
inconceivable. It is nonetheless undesirable because if you abstract
from your own fortunate position, which is just a matter of good luck
and thus lacks moral weight, you see that you would yourself necessarily wish for other peoples help in the same way that they now wish
for yours (which following your selfish inclination you are tempted to
20
This imprecision is not too uncommon. Cf. sinnenfreie Neigung (propensio
intellectualis), Metaphysics of Morals, V, p. 213, which would normally seem to be a
contradiction in terms.
21
It is in this that Kant goes beyond rule utilitarianism, consequentialist universalisation, etc.
246
jens timmermann
deny them).22 Again, you would like to employ double standards, making an exception for yourself without any good, impartial justification.
The reason why duties of the wider variety are subordinate to duties
of strict obligation is disclosed in the passage immediately following
the four illustrations. Kant there states the distinction between the two
types of duty in the following way:
Some actions [actions contrary to strict/perfect duty, JT] are so constituted that their maxims cannot even be thought without contradiction as
a universal law of nature, far less could one will that they should become
such. In the case of others [actions contrary to wide/imperfect duty, JT]
that inner impossibility is indeed not to be found, but it is still impossible
to will that their maxim be elevated to the universality of a law of nature
because such a will would contradict itself (IV, p. 424).
Far less could one will that they should become such! Kant reveals the
reason for the primacy of strict duty in this sentence: all actions contrary to duty cannot possibly be willed to be universal laws, they all
generate what has been called a contradiction in the will; some actions,
in addition, also generate a contradiction in conception because their
maxim cannot even be conceived as universal if we want it to be successful. Maxims that, if universalized in the Kantian manner, generate
a contradiction in conception also generate a contradiction in the will,
whereas the reverse is not true. Actions contrary to wide duty only generate a contradiction in the will if we try to make them universal in
accordance with the test of the categorical imperative. They are therefore restricted by duties of strict obligation, the violation of which creates the additional difficulty of a contradiction in conception. In fact,
the categorical imperative in its first and primary formulation addresses
the question whether we can will the maxim to become a universal law,
covering all four possible cases, not whether we can think it as such.
Act only on a maxim that you can think a universal law! would be the
categorical imperative of strict duty.23
Similarly, the second variant formulation of humanity as an end
in itself provides us with two partially overlapping criteria: treating
persons as mere means to ones own ends (the strict criterion) and not
22
There is arguably even a particularly strong rational commitment for you to
accept other peoples help should you require it because you endorse your own
selfishness.
23
The details of this need to be spelt out. Not everything that seems to be a strict
duty turns out to be one. There may be cases of suicide or self-killing which should not
be subsumed under morally bad self-murder. Judgement must decide between maxims and imperatives that conflict.
treating them as ends in themselves (the general criterion). The former

is the stronger, exemplified only in actions contrary to perfect duty,
whereas the latter always applies whenever a duty, perfect or imperfect,
is violated.
Wide or imperfect duties are restricted by the demands of perfect
duty. Yet within these constraints not all imperfect duties are equally
wide. In Kants Metaphysics of Morals, we learn more about the manner
in which this type of duty can be made more precise. Ethical duties,
Kant says, are of wide obligation, whereas duties of right are of narrow
obligation (VI, p. 390). The former prescribe only the maxims of
actions, not actions themselves,24 which is taken to be a sign that it
leaves some leeway (latitudo) for free choice in following (complying
with) the law. However, he adds the following warning:
But a wide duty is not to be taken as permission to make exceptions to
the maxim of actions [Maxime der Handlungen] but only as a permission
to limit a maxim of duty [Pflichtmaxime] by another (e.g. love of ones
24
One has to be careful to distinguish between duties of right as such and perfect
duties to others. Hill seems at times to overlook this distinction (cf. Meeting Needs
and Doing Favors, p. 203, and Imperfect Duty and Supererogation, p. 149). Only the
former can be complied with by merely performing the action in question, the agents
character be what it may. The latter, strictly speaking, command the adoption of ends
in ones maxim, as do (wider) duties of virtue. The difference is that in the case of perfect duties to others, viewed from an ethical perspective, the realization of the end in
question is comparatively uncomplicated, whereas the ends specified by the latter need
to be actively pursued. It is the end of the honest shopkeeper to be honest in his dealings with his customers, no matter how young and inexperienced they are. This end is
part, or a direct consequence of, a general principle of honesty, adopted on recommendation of the categorical imperative. The prudent shopkeeper, by contrast, has a
maxim to do everything that is good for his business, and his treating customers honestly, his lack of true honesty notwithstanding, is a consequence of his commitment to
doing well financially because he perceives that treating customers decently, no matter
how young and inexperienced they may be, is on the whole the best policy if he wants
to achieve this end. Both shopkeepers comply with their strict duty of right to be honest in their dealings with customers, but from an ethical point of view only the truly
honest shopkeeper is committed to the morally obligatory end. When on p. 203 of
Meeting Needs and Doing Favors Hill says: Perfect duties directly prescribe or prohibit actions rather than the sort of indefinite maxims (to promote ends) prescribed
(directly) by imperfect duties and: We never satisfy the requirements of imperfectduty
simply by doing or refraining from external acts from nonmoral motives, such asgiving money to charity to impress people, one should therefore read duties of right for
perfect duties, more or less indefinite for indefinite and duties of virtue for imperfect duties in the first quotation and add nor do we satisfy the requirements of perfect
ethical duties if we perform external acts from nonmoral motives. On this reading of
perfect ethical duties, there is no difficulty in fitting in perfect duties to the self (cf. Hill,
Meeting Needs and Doing Favors, p. 205). They too should be construed as duties to
adopt ends that directly translate into action. Refraining from the acts mentioned on
the basis of any old end does not satisfy an ethical duty, not even a strict one.
248
jens timmermann
neighbour in general by love of ones parents), by which in fact the field
for the practice of virtue is widened (VI, p. 390).
The interpretation of this passage is disputed. Thomas Hill, for instance,

detects a terminological distinction between maxims of actions and
maxims of duty, which he then uses to defend his thesis that some
particularly wide duties allow for actions that are morally good but not
required. However, I rather doubt that any such distinction is intended.25
The passage rather confirms the strict interpretation of perfect as well
as imperfect duty offered above. It seems quite possible that Kant wrote
these lines in the later Metaphysics of Morals to correct the impression
his audience got from a glance at the misleading footnote in Groundwork
II discussed above. Maxims need to be applied or enacted; and if they
concern promoting an end rather than refraining from certain actions,
this can happen in various different ways.
Yet the role of inclination in the course of translating maxims into
action must be quite limited. The end is specified by the maxim.
A related inclination may therefore at most determine the means,
i.e. decide the question of how to comply with wide obligation in the
absence of decisive moral considerations regarding means. Consider
the following example. If the world is in a pretty sad stateas in fact it
is and has been at most points in human historywe could not, other
things being morally equal, choose not to help the needy to the best of
our abilities.26 We could not choose to be exempt even from wide duty.
Cf. Hill, Meeting Needs and Doing Favors, p. 221. As to Hills criticism of
Cummiskey, if there is no distinction intended, there is no distinction for Cummiskey
to overlook in his strict interpretation of the passage on pp. 11011 of Kantian
Consequentialism; and Hill cannot draw on it to show that the wider kind of maxim
admits of the kind of looseness required by his thesis that wide duty can (partially) take
over the role of supererogation. To make a terminological distinction, Kant would
have used parallel terms such as Maxime der Pflicht and Maxime der Handlungen, and
probably emphasized them typo-graphically (larger type, in effect todays italics or
bold type). The difficulty for Hill is that the sentence as quoted places no emphasis on
the two expressions at all. No contrast is intended. (This is admittedly obscured by the
veil of translation.) Maxim of actions (Maxime der Handlungen) and maxim of duty
(Pflichtmaxime) are mere stylistic variations. In particular, the maxims of actions in
question are dutiful maxims too, maxims in accordance with wide duty. Kant merely
wishes to point out that good maxims can be checked and made more specific by other
good maxims, but there must be no exceptions. Cummiskeys reading is essentially
correct.
26
On this reading, Kants Groundworks command in the context of the formula of
humanity that everyone further the ends of others so far as he can (soviel an ihm ist,
IV, p. 430) is to be understood as a very strict command (contra Hill, Meeting Needs
and Doing Favors, pp. 22728). It does not follow that we need to devote all our spare
25
However, there is still some leeway or playroom (Spielraum) because

the imperfect duty of beneficence only tells us to take the needs of others seriously. It does not tell us how precisely to go about realizing this
morally worthy end.
Maxims that pay respect to strict duty can be acted on more or less
straight-forwardly,27 whereas maxims corresponding to wide duties
cannot. We must often acquire a fair amount of knowledge about suitable means, in Kantian terms: we must familiarize ourselves with rules
of skill, the technical kind of hypothetical imperative, telling us how
best to put the ends we pursue into action;28 and quite often there is
more than one way to pursue ones ends skilfully. At this point choice
and latitude may enter deliberation.
Yet there cannot be exceptions in favour of inclination, or indeed
anything else. If Hill were right and those firmly committed to an
endas prescribed by wide dutycould face situations in which the
realization of the end was optional, i.e. pursuing the end was good but
not required, the question of what counts against performing the good
action arises. This cannot be a moral consideration, for in that case, as
we saw above, there would be a duty to act differently, and pursuing the
end in question would not be good. Yet if the consideration counting
against doing what is said to be morally good but not required is nonmoral, it must be based on inclination; and it is difficult to see how an
agent could be justified in preferring an action and an end based on
inclination to a moral one.29
time and resources to further other peoples ends, as Cummiskey argues, but all restrictions of our efforts must be on moral grounds (cf. section 3 below).
27
We may have our doubts about that
28
As I have argued elsewhere, assertoric imperatives and the categorical imperatives
are responsible for the choice of maxims (which contain ends), whereas problematic
imperatives, or rules of prudence, concern the realization of ends selected in accordance with the other two types of imperatives. Wide duties especially require a certain
amount of skill because they command the pursuit of sometimes complicated ends. It
is therefore likely that Kant has to posit an indirect duty to acquire the skills one needs
to act on the ends set by wide duty. If, for example, you live in a house by the sea where
shipwrecks are common, you may well have an indirect duty to perfect your swimming
skills. From this point of view, there is nothing morally good about perfecting ones
swimming skills as such, but it is the one means requisite for doing your duty of saving
people from drowning. At least the application of what you perceive to be your duty
can be messier and more empirical than a cursory glance at Kants ethical writings
reveals.
29
Note that I am not arguing that the categorical imperative can resolve the issue as
to which morally worthy endsay perfecting ones own talents vs. helping those in
need should be given priority in all cases. There is good reason to believe that this is
250
jens timmermann
To understand the role of inclination in the choice of certain actions

in accordance with wide duty, consider the following example. Listening
to the radio news you hear about a particularly pressing emergency,
such as a tidal wave. There can be little doubt that the people affected
require your help. In fact, if you were in their situation you would
expect the citizens of richer nations to come to your aid, and with good
reason. You have 20 to spare; there is no equally or more worthy purpose for you to spend this money on. This is a case of your maxim to be
beneficent. You decide that you ought to help.
However, your decision to help is not as easily realized as a decision
not to lie. You must know, for instance, that there are charitable organizations such as Oxfam and Brot fr die Welt which will put your money
to the best possible use (as well as other charities with perhaps somewhat less convincing credentials). Given that they are both charities
with an excellent record of helping the poor, you can give your money
to one or the otherbut part with your money you must. You cannot
refuse to do your duty this time simply because you have been charitable on past occasions, or because you would rather spend your money
on something else. Kant can thus suggest that our dutiful actions be
directed by inclination, other things being morally equal. You can
discharge of your duty by giving your money to Brot fr die Welt on a
visit to Germany or by sending a cheque to Oxfam. These actions are
morally equivalent. They are equally good means or ways of doing ones
duty. Morality tells you: do whatever you likebut do it.
Two things should be noted. First, the sense in which giving money
to Oxfam can be said to be good but not required crucially depends on
the description of the action. There is as such no duty to give money to
Oxfam; but that is not what you want to achieve anyway. Giving money
to Oxfam is not the end of your action as prescribed by the formal law
of morality. You wish to help the needy, which is indeed a duty, and
giving money to Oxfam is an excellent means of doing just that. It is
only in a somewhat Pickwickian sense that giving money to Oxfam is
good (a good means to a morally necessary end) but not requiredas
is giving money to Brot fr die Welt. Strictly speaking there is no
impossible, even in principle. According to the argument in the preface to The Common
Saying, no rule can govern conduct all the way down. You always need an act of judgement to decide that a given situation falls under the rule, no matter how specific it is;
cf. VIII, p. 275. Judgement must therefore decide matters of conflicting grounds of
obligation. This is, needless to say, really rather optimistic. Cf. also the closing remarks
of this paper.
dutyno duty sui generisto give money to either Oxfam or Brot fr

die Welt. Doing so is at best an indirect duty. As such, this choice is
morally neutral. The only thing that matters morally is your duty to
help, and the decision to fulfil your wide duty to be charitable alone is
morally good. As always, the correct action description is specified by
the maxim acted on, i.e. by the end specified in the maxim. When acting on your maxim of beneficence you fill in your cheque to Oxfam you
are doing your duty: that is the only morally relevant description.
Secondly, the room for exercising our free choice depends on the
equivalence of means to a morally necessary end. This case is different
from certain dilemmatic situations in which, for example, you are able
to save only one of two persons from a shipwreck. If there are no morally relevant considerations to help you decide whom to save, it does
not follow that you are permitted to save either (which then would, in
a sense, be good but not required). This case is different from the
Oxfam case because saving a single person is not a means to doing ones
duty to others. It is part of doing ones duty. If there are no moral criteria (such as debts of gratitude or prior arrangements) both persons
should be given the same chance of survival; you should toss a coin.
Only this fair procedurerather than plumping for the one person or
the other, which is bound to rely on prejudiceis compatible with
treating both of them as ends in themselves.30
A lecture on moral philosophy recorded by C.C. Mrognovius in the
winter semester of 178485 confirms the strictness of so-called wide
or imperfect duty.31 In a highly revealing passagecuriously not
included in the Cambridge translation of selections from the lecture32
Kant proposes the following shift of perspective:
All duties are owed duties with respect to God.33 With respect to human
beings, there are meritorious actions too, for with respect to them I can
30
I argue for thison broadly Kantian groundsin The Individualist Lottery:
How People Count, But Not their Numbers, Analysis 64 (April 2004), pp. 10612.
31
There are several ways in which wide duties towards others are still quite different from strict duties. Like Humes cautious and jealous virtue of justice, they depend
on the needs of others and generally on the right occasion. There is not always occasion
to help, but there is always occasion not to lie. Thats why they are called contingent
duties (zufllige Pflichten) in the course of discussing the formula of humanity. Once
they apply, of course, they are just as necessary as any duty.
32
As the square brackets on p. 246 indicate, there is a gap in Heaths translation; cf.
Immanuel Kant, Lectures on Ethics (Cambridge: Cambridge University Press, 1997).
33
Note that all duties are duties with respect to God. There are no duties to God,
strictly speaking, cf. Metaphysics of Morals, VI, pp. 44344.
252
jens timmermann
do more than I owe them: these are duties of virtue. In their case, I impose
a new duty on the others, rather than merely doing my duty as is the case
with duties of right. The other person does not have a right to duties of
virtue (Moral Mrongovius II, XXIX, p. 631).
It is common Kantian doctrine that in matters of wide duties to

others subsumed under duties of virtue in this passagethe other
person does not have a right to the agents performing this meritorious dutyas opposed to owed or strict duties, for which the agent
deserves no credit.34 In other words, there are acts that are demanded,
despite the fact that the person affected cannot demand the act of the
agent.
In his lectures, Kant seems to have been quite explicit about the
implications of this thesis: some actions are meritorious, and the other
person ought to be grateful, precisely because he is not entitled to my
performing the action in question. Such duties cannot be exacted. But
merit exists only with respect to the other person. Sub specie aeternitatis
there is no merit but only plain and simple duty; and God is under no
obligation to be thankful to you. Even meritorious actions are strict,
owed duties once one shifts focus from the goings-on between us mortals to the point of view of the divine being who unrelentingly demands
that we strive for perfection. Kants Metaphysics of Morals similarly
insists on moral perfection:
The wider the duty, the more imperfect therefore a mans obligation to
action, as he, nevertheless, brings closer to narrow duty (duties of right)
the maxim of complying with wide duty (in his disposition), so much the
more perfect is his virtuous action (Metaphysics of Morals, VI, p. 390).
Again, the conclusion must surely be that whenever an action presents

itself to us as morally good we are morally required to do it. Wide
or imperfect duties are less binding than perfect duties only in the
sense that they are restricted by the former, and that they crucially
depend on the appearance of the right kind of situation to applybut
if and when they apply they are just as binding as the other kind of
duty. We must strive to be perfect. The spectre of supererogation can at
last be dispelled.35
34
Kant switches to talking about necessary/owed (rather than strict or perfect)
duty vs. contingent/meritorious (rather than wide or imperfect) duty in discussing
the second variant of the categorical imperative in the Groundwork; cf. IV, pp. 42931.
35
Marcia Baron makes the excellent point that an ethical system that allows
for supererogation may be seen as encouraging stocking up ones account of moral

3. The Scope of Morality
If moral goodness is so infinitely precious; if even wide duty gives you

at most a choice of how, but not whether, you obey it; if in short there
is no place for supererogatory action in Kants ethicsdoes he want all
of us to become moral saints and heroes? In one sense the answer must
be yes. Kant clearly asks us to get our priorities right, as the abovementioned passages from his Common Saying and Critique of Practical
Reason suggest.36 Human agents must, as a matter of principle, first
check whether a proposed action is in line with the moral law, and pursue their own personal ends only within these limits. However, ideal
Kantian moral agents face the prospect of leading lives less dreary than
their unfortunate utilitarian colleagues. We are reminded of the indirect duty not to make oneself unhappy (Groundwork, IV, p. 399), which
is grounded in the moral dangers inherent in misery; and there are
three further mitigating reasons why the lives of Kantians are not as
unpalatable as it might at first seem.
First, even though we have an obligation to assist others in their projects and, generally, to make them happy, there are moral dangers in
helping others that we need to avoid. There is the familiar Kantian idea
that not everybody always deserves our help. Morally bad people may
notyou ought not to assist them in their morally flawed projects, no
matter how happy it would make them (cf. Metaphysics of Morals, VI,
pp. 48081). Apart from that, to help may be your duty, but it is worth
bearing in mind that acts of assistance can be demeaning for the recipient. This will restrain the Kantians altruistic efforts to some extent. In
the lectures on ethics, Kant cautions us against accepting, and being
bound by, the help of others, especially accepting benefitswhich anyone should do
only under the following conditions: first, out of dire necessity, and then
with complete confidence in the benefactor (he is no longer a friend, but
goodness by means of the occasional supererogatory act while frequently neglecting
the everyday responsibilities of life (Kantian Ethics Almost without Apology, pp. 3839).
There is no such danger if there is an upper limit of duty that everyone always has to
meet. Also, you cannot bargain with God (or with yourself) as there is nothing beyond
duty that you can do to make up for the failings of the past. As Baron also notes, there
is the additional advantage that Kantian ethics does not encourage us to rank virtuous
people or virtuous lives (Kantian Ethics Almost without Apology, p. 54 n. 64).
36
Cf. also Kants theory in chapter 1 of Religion within the Limits of Mere Reason
(cf. V, pp. 4653) that there are only two fundamental maxims: pursuing ones own
ends within the limits of morality and vice versa.
254
jens timmermann
a patron) that he will not count it as an obligation towards him (Moral
Mrongovius, XXVII, p. 1559; cf. the briefer account in Moralphilosophie
Collins, XXVII, p. 442).
Accepting favours can destroy the equality required for friendship; and
the fact that an act of assistance would be a burden is a very good
moralreason against it. We must similarly avoid being intrusive, patronising or paternalistic. Other-regarding action may be morally bad
even if the recipient happens to benefit. Partly with the intent of meeting these difficulties, Kant suggests earlier in the lectures that some
basic acts of assistance be re-classified as acts of strict duty:
One may have a share in the general injustice, even though one does
nobody any wrong by civil laws and institutions. So if we now do a kindness to an unfortunate, we have not made a free gift to him, but repaid
him what we were helping to take away through a general injustice. For if
no-one might appropriate more of this worlds goods than his neighbour,
there would be no rich people, but also no poor. Thus even acts of kindness are acts of duty and indebtedness, arising from the rights of others
(Moral Mrongovius, XXVII, p. 1539; Moralphilosophie Collins, XXVII,
p. 416).37
These considerations overall seem to point in the direction of a moral

theory that is demanding as far as the basic needs of others are concerned. Some unfortunate fellow human beings have to be raised to the
level of free and equal agents. Over and above that, the delicate give and
take of personal relations such as family and friendship would seem to
determine the nature and extent of ones furthering the ends of others.
Secondly, most philosophers concerned about the adverse effects of
moral heroism or sainthood think of morality as relating to our attitudes and actions towards other people; but the sphere of the moral
was much broader for Kant. As other-regarding actions are not always
good in Kantian ethics, self-regarding actions need not be morally neutral or bad. For Kant, there are duties to the self. Some self-regarding
actions may even be morally required. Although non-moral considerations such as a mere selfish desire or inclination can never outweigh
moral ones, the duties towards ones own person can surely morally
37
Cf. Even the civil constitution is pre-arranged so that we participate in public and
general oppressions, and thus we have to regard an act we perform for another, not as
an act of kindness and generosity, but as a small return for what we have taken from
him in virtue of the general arrangement (Moral Mrongovius, XXVII, p. 1552;
Moralphilosophie Collins, XXVII, p. 432).
outweigh grounds of duties towards others. We have a duty to develop

our talents, to stay in good health, and even an indirect duty to promote our own well-being. Without a doubt, this means that untrammelled altruism is not at all morally good. This does not, of course,
necessarily entail that a recognizably Kantian ethical theory would be
much less demanding subjectively, and some of the worries of those
who fear the strain of a command to strive for moral perfection still
apply, no matter whether morality is directed at the agent himself or at
others. Kant also seems overly optimistic about the powers of human
reason to decide apparent conflicts between morally worthy ends. Yet
the overall picture would be more balanced.
Thirdly, although it remains unlikely that an agent can never sit back
and say I have done enough it is worth remembering that the burden
of wide duties towards others does not just rest on a single agents
shoulders. How much exactly needs to be done, especially to meet the
basic needs of all human beings, obviously depends on many factors
that are not within our control. Yet it is essential to Kants universalist
ethics that the burden of helping others be shared equally. Everyone
must help to the best of his or her abilities to make the world a better
place; and to be sure if everybody did, saints and heroes would rarely
be needed. It may even seem morally objectionable if someone as a
matter of principle gives all her available income to charitable causes
because this attitude enables others with similar earnings to free-ride
on her beneficence. The even more extreme maxim of always giving
more than ones fellow moral agents clearly does not qualify as a universal law.
In non-consequentialist theories such as Kantian ethics it does not
just matter that the good deed is done but also who does it; and duties
of assistance to others have the interesting feature that if someone helps
a person in need someone else does no longer have to. Going beyond
the call of ones personal remit in the case of duty may have the consequence that others get away with doing less. Yet there is a delicate line
to be trodden. If it is clear that others will not contribute their fair
share, the Kantian can hardly walk away when she has done what would
otherwise have been her share of duty. It will presumably be her duty to
make up for the moral deficiency of others. The point worth stressing,
however, is that a Kantian ethical theory assigns duties to individual
agents, and it is their responsibility to live up to them. If every-one did
just that, we should not have to fear that the Kantian call for altruism
might be too demanding. This is the first important point to note.
256
jens timmermann
The second is that in an imperfect world with imperfect agents it may

be our duty de facto to do more than would be our fair share under
ideal conditions. The ideal Kantian agent may feel almost as great a
strain as his or her utility-maximizing colleague. However, I do not
think this counts against Kantian ethics. Rather, it counts against agents
who by neglecting their duties impose additional burdens on those
who care about the fate of their fellow human beings.
A strict interpretation of wide duty that depends on the individual
agents good judgement about what he or she ought to do in any specific situation has the additional advantage of solving what we may call
the problem of urgency. If, like Hill, we distinguish stricter and wider
imperfect duties not just for duty types but also for token dutiesif,
in other words, even at the level of deliberation about what to do in
a given morally relevant situation the categorical imperative tells us
sometimes to promote a certain endwe encounter the problem that
we do not really know when we should act on our maxim and when we
may legitimately make an exception. Hill admits that in some particularly extreme circumstances there may be no significant latitude for
choice.38 If, however, as argued above, the latitude even of imperfect or
wide duty is limited to choosing amongst equally appropriate means to
doing ones duty, and in a relevant case39 we must always do something
to fulfil our duty, the question when to act on our maxim never occurs.
There will still be particularly pressing cases in which we have even less
of a choice, because it is a clear case to fall under our maxim of beneficence, we are the only people who can do the task in question, and
there is only a single means through which to realize our end. But at the
token level, duty, in the singular, always commands with the force of
necessity.
In conclusion, Kant does indeed ask us to give morality first priority,
but his conception of morality is much more complex than that of most
contemporary ethics. Indeed, the thesis that morality is overriding
originated at a time when morality could not be reduced to altruism or
universal happiness. It depends on such a rich conception of morality
not to be completely ludicrous. At the same time, Kant was extraordinarily optimistic about the powers of moral judgement of any sincere
Hill, Meeting Needs and Doing Favors, p. 208.
I insert this caveat to indicate that there are probably many situations still in
which the moral law does not command any positive duty. Agents can then, of course,
act as they wish within the limits of the morally permissible (perfect duty).
38
39
human being: as morality is supposed to be synthetic a priori, making

up ones mind about what one morally ought to do is easy.40 However,
even though morality is not exclusively other-regarding, altruism is to
be taken very seriously.41 We can expect Kantian ethics to be somewhat
more demanding than ordinarily conceived. There is no suggestion
that, in a world like ours, being moral should be easy. In concrete practical terms, I should be surprised if Kants recommendations were very
different from the tithe suggested by Peter Singer in his Practical Ethics
chapter on Rich and Poor, however much he would be appalled by
Singers ethical theory.
Lastly, a word of caution. The argument of this paper purports to
show only that there are conceptual and philosophical points which
count against allowing for the possibility of actions that are morally
good but not required in any but the Pickwickian senses mentioned
above, and that within a recognizably Kantian theory the room of
choice amongst morally worthy options is severely limited. The main
reasons are Kants characterization of moral value as absolutely good,
and of moral action as necessary. Apparent conflicts of grounds of
duties would have to be decided by the moral judgement of each individual agent.
Kants optimism about the powers of human moral cognition is certainly quite extraordinary. Just consider the familiar recommendation
of the categorical imperative as the compass of moral judgement with
which the agent knows very well how to distinguish in every case as it
occurs what is good and what is evil, what is in conformity with duty or
contrary to it if only we point reason to its own principle (IV, p. 404).
Yet, even if we find it difficult to share this optimism, the Kantian lesson remains that moral decisions rest with the autonomous agent. It is
not the task of the moral philosopher to decide moral dilemmas, armchair style, but that of every human agent him or herself.
40
The empirical elements of prudential deliberation make it much more complicated. Cf. Groundwork, IV, pp. 401402.
41
There seems to be no exact decision procedure: How far should one expend ones
resources in practicing beneficence? Surely not to the extent that one would finally
come oneself to need the beneficence of others (Metaphysics of Morals, VI, p. 454). If
we go to extremes, rational inconsistency looms, as this quotations clearly shows. Still,
apparent moral value can only be outweighed by moral value, and what remains is
duty full stop yet making other people happy is not an unrestricted or unconditional duty.
Virtue, Character and Situation

Jonathan Webber*
1.Introduction
What is character? What do the distinctive patterns we discern in our
own and one anothers behaviour consist in? Exactly what does it mean
to call somebody honest, compassionate or courageous, and how are
such epithets earned? Answering these questions is central to assessing
the various ethical theories that enjoin us to develop morally sound
character traits, since we need to know in some detail what character
traits really are before we can discern whether and how they can be
developed. John Doris has argued that empirical evidence indicates
that we do not have characters as these are generally understood in
ethical discourse. There are no such traits as prudence, temperance,
courage, or fairness, he argues. There are only such traits as sailingin-rough-weather-with-friends-courage and office-party-temperance,
which are perfectly compatible with acting in a cowardly fashion when
sailing in rough weather with new acquaintances and drinking excessively at family gatherings. Each trait is to be specified with reference to
a range of features of the situations in which they are manifested, he
claims, with the result that each person has a wide range of traits each
with very restricted situational application. Following his terminology,
we can refer to this as the fragmentation theory of character.1
Doris goes on to argue that adopting this fragmentation theory
should lead us to a certain redirection of our ethical attention: instead
of attempting to improve our characters, we should invest more of our
energies in attending to the features of our environment that influence
behavioural outcome; we should try, so far as we are able, to avoid
*For comments on earlier drafts, I am very grateful to Chris Bennett, George
Botterill, Paul Faulkner, Chris Hookway, Sarah Parker, Tom Simpson, James Tartaglia,
Tom Walker, Suzi Wells, an audience at Keele University, and two anonymous referees
for this journal.
1
John Doris, Lack of Character: Personality and Moral Behaviour (Cambridge:
Cambridge University Press, 2002), chs. 15; Peter Goldie, On Personality (London:
Routledge, 2004), chs. 3 and 4; the sailing example is taken from p. 115.
260
jonathan webber
ethically dangerous circumstances and seek out situations conducive

to ethically desirable conduct. The ideal of character-development recommended by many ethicists should be abandoned in favour of an
ethic of situation-management. Maria Merritt has formulated a
variant of the fragmentation theory, according to which the aspects of
situations that need to be built into the concept of each trait are
always social aspects. She agrees that we should abandon characterdevelopment in favour of situation-management, adding that we
should focus on manipulating our social settings and relationships.2
But not all adherents of the fragmentation theory agree. Peter Goldie
advocates the aim of harmonizing ones fragmentary traits into semblances or simulacra of traits as traditionally conceived. If I know that
I am only sociable in certain situations, for example, and I approve of
that sociability, then it seems that I would prefer to be sociable generally. With the help of the twin executive virtues of circumspection and
strength of will, I can manipulate my surroundings so that they elicit
sociable behaviour, and continually doing so will gradually alter my
situation-specific dispositions, so that I am disposed to be sociable
across a wide variety of situations.3
These recommendations might be criticized for a variety of reasons.
Ones power to manipulate ones situations will always be limited, so it
might seem that the advice of Doris and Merritt does not leave us with
enough control over our behaviour. If ones traits are indeed fragmentary, then they will also be vastly more numerous than has traditionally
been thought, so Goldies advice might also seem impractical. But these
lines of thought will not be pursued here.4
The aim of this paper is rather to show that we should not accept the
fragmentation theory of character that these recommendations are
based on. The next two sections divide the evidence cited in favour of
the fragmentation theory between data that should be set aside and
data that needs to be considered. The following section shows that the
reliable data provides evidence in favour of the fragmentation theory
only if it is interpreted in the light of a behaviourist construal of character traits as dispositions to respond in certain ways to certain kinds
2
Doris, Lack of Character, pp. 14647; Maria Merritt, Virtue Ethics and Situationist
Personality Psychology, Ethical Theory and Moral Practice 2 (2000), pp. 36583.
3
Goldie, On Personality, chs. 3 and 4.
4
For development of the idea that the ethical advice issued by Doris and Goldie is
impractical, see my Character, Global and Local (forthcoming in Utilitas).
virtue, character and situation261
of stimuli, and that it is compatible with the traditional theory of character if traits are conceived as dispositions to be inclined with a certain
strength to behave in a certain way in response to a certain kind of
stimulus. We will see in section five that this latter conception is prominent in Aristotles Nicomachean Ethics and in recent characterological
ethical discourse influenced by it. Section six presents a pragmatic reason to prefer this conception of traits to the behaviourist one favoured
by proponents of the fragmentation theory, and the final section considers some implications of this discussion for future philosophical
investigations of character and virtue.
2. Specious Evidence
There are five kinds of experiment cited in favour of the fragmentation
theory, all of which are cited by Doris, and two of which are cited by
Goldie and Merritt.5 Of these five experiments, as we will see,twoshould
be discounted. There is also a sixth experiment discussed by Doris, but
although he has been criticized for taking this experiment as evidence
for his theory, he does not actually use it in this way, as we shall see.
One of the experiments to be discounted seems to indicate that people are more likely to exhibit helping behaviour if they have recently
had good luck. The subjects of this experiment were unaware that the
experiment was taking place. Having made calls from a telephone
booth, some found a dime in the coin return slot and others looked but
found nothing. Each subject was soon walking behind a woman who
dropped a folder full of loose papers and began to gather them up. Of
those who had just found a dime, 96 per cent helped her. Of those who
had not, just 13 per cent helped.6 All three advocates of the fragmentation theory cite this experiment.7 The results are taken to challenge the
5
Gilbert Harman presents a similar but distinct argument in his paper Moral
Philosophy Meets Social Psychology: Virtue Ethics and the Fundamental Attribution
Error, Proceedings of the Aristotelian Society 99.3 (1999), pp. 31531. He employs some
of the data employed by Doris, and argues that character is illusory and so characterological discourse should be abandoned altogether. I cannot do justice to his argument
here, but see my Character, Common-Sense, and Expertise (forthcoming).
6
Summary of Alice M. Isen and Paula F. Levin, Effect of Feeling Good on Helping:
Cookies and Kindness, Journal of Personality and Social Psychology 21.3 (1972),
pp. 38488, Study II.
7
Doris, Lack of Character, pp. 3032; Merritt, Virtue Ethics and Situationist
Personality Psychology, p. 366; Goldie, On Personality, pp. 3032.
262
jonathan webber
regularity theory: if behaviour could be influenced so strongly by such

a minor event, the thought runs, then the patterns of ones behaviour
would seem to reflect the vicissitudes of ones daily life rather than ones
character.
It has been pointed out that repetitions of that experiment have
yielded wildly differing results.8 But similar experiments have shown
that increased sunshine brings with it larger tips for waiters and an
increase in the number people willing to answer questions for a survey,
and that having recently been given a cookie increases the likelihood
that someone will agree to help another person with a task but not the
likelihood that someone will agree to hinder someone else where this
would be part of an experiment.9 Minor situational variations do seem
to have an impact on helping behaviour. But as critics have recently
argued, these experiments have involved only relatively trivial helping
scenarios. Nobody has shown that such minor situational variations
affect the likelihood of responding to someone seeming to be in serious
distress.10 We should therefore take these experiments to show only
that we should not judge character on the basis of such minor acts of
help as gathering up someone elses dropped papers.
The other experiment to be discounted is the Stanford Prison
Experiment, cited by Doris. Eighteen males who had volunteered to
partake in a study of prison life and were judged to be stable and mature
were randomly assigned roles of either prisoner or guard. Early one
Sunday morning, each prisoner was arrested by real police officers
with no warning, as family and neighbours looked on, and was charged
with public order offences, taken to a real police station, warned of his
rights, fingerprinted, and then transferred blindfold to what he was
told was the Stanford County Jail. He was searched, stripped, sprayed
with a delousing agent, dressed in a rough smock, and assigned a number to be used in place of his name. A chain was padlocked to his ankle,
to be worn at all times, even while asleep. The guards were not trained,
and were free to do just about anything they thought necessary to
maintain law and order. At 2.30 am on the first night, they woke the
prisoners and made them line up and give their identification n
umbers.
8
Christian Miller, Social Psychology and Virtue Ethics, The Journal of Ethics 7.4
(2003), pp. 36592, Appendix.
9
See, for example, John Sabini, Social Psychology (London and New York: Norton,
2nd edn, 1995), pp. 30710.
10
John Sabini and Maury Silver, Lack of Character? Situationism Critiqued, Ethics
115.3 (2005), pp. 53562 (53940).
Dissension was met with forced push-ups, sometimes with someone

pressing down on the prisoners back. The prisoners responded by barricading themselves inside their cells. The guards used fire extinguishers to fight through the barricades and remove the supposed ringleaders
to solitary confinement. From here the situation continued to worsen
dramatically until the experimenters called a halt after six days of the
scheduled fourteen.11
Whatever this might tell us about the conduct of experiments, it cannot provide strong evidence in favour of the fragmentation of character. Doris implies that the behaviour of these volunteers did not
conform to their own general behaviour patterns, and that most people
would behave the same way in the same circumstances. But the experiment has never been repeated and there were no control groups: the
behaviour of these few subjects simply cannot warrant such claims.
Although the experiment involved eighteen subjects, moreover,
these did not enter the experimental situation individually but together,
split into two rival groups. The implication that all the subjects playing
the guard role were themselves over-zealous to the point of violence is
therefore unwarranted. It is just as likely that this is true of only one or
two of the guards, and that the others found it too difficult to challenge
him or them. Similarly, there may have been one or two prisoners persuasive enough to convince the others to rebel. With no repetition and
no control groups, we cannot draw conclusions about the motivations
of the behaviour of individuals in this situation.
Finally, even if it is true that most people will behave unusually in
such a situation, or even that most people will behave in the same ways
in such a situation, the extremely disorienting nature of the opening
stages of the experiment for the prisoners and the sheer strangeness
and threatening instability of the situation faced by the guards make it
difficult to be confident of any extrapolation to less extreme situations.
Any evidence this experiment might provide for the fragmentation of
character, therefore, would be defeated by this concern, and so the data
should be set aside.
Doris also draws attention to studies of honesty and of extraversion
and introversion among schoolchildren. In one study, more than eight
thousand subjects between the ages of eight and sixteen were observed
in situations where they had the opportunity to cheat in tests, lie about
11
Summarized from the excellent website www.prisonexp.org, created and maintained by one of the organizers of the experiment, Philip Zimbardo. See also Doris,
Lack of Character, pp. 5153.
264
jonathan webber
whether they had cheated in those tests, fake a record of their athletic
performance, steal a small amount of unattended money, and so on.
The study found a high correlation between the proportion of a group
of children behaving dishonestly in one situation and the proportion of
the same group behaving dishonestly in another situation of the same
type. Children who cheated in an exam, it seems, are likely to cheat in
the next exam. But it also found very little consistency across types of
situation. A child who cheats in an exam seems no more likely than any
other child to steal unattended money or fake a record of athletic performance. So among children at least, there seem to be no unified traits
of honesty or dishonesty, only narrower traits such as being an examcheat or an unattended-money-thief.
Rachana Kamtekar raises a methodological concern about the use of
this data in the debate over character, asking whether it is legitimate to
make inferences about adult character traits from observation of children who might be more impressionable, less committed to particular
ideals of conduct, or less integrated than adults.12 But this is simply
unfair to Doris, who does not use the data in this way. He is aware that
children with developing personalities are plausibly thought to exhibit
less behavioural consistency than fully formed adults, and discusses
these experiments not so much for their evidential role as for the interpretive perspective they provide.13 In this, he is right. This data might
illuminate the development or structure of the trait of honesty, but cannot provide evidence against the idea that adult behaviour is regulated
by such general character traits as honesty, generosity and courage.14
We will therefore set it aside.
3. Significant Evidence
The three remaining kinds of experiment do provide reliable evidence
concerning adult character traits. Probably the best known, and
12
Rachana Kamtekar, Situationism and Virtue Ethics on the Content of Our
Character, Ethics 114.3 (2004), pp. 45891 (466 n. 30).
13
Doris, Lack of Character, p. 63.
14
Other objections to the use of this data are raised by Gopal Sreenivasan in Errors
About Errors: Virtue Theory and Trait Attribution, Mind 111.441 (2002), pp. 4768,
and by John Sabini and Maury Silver on pp. 54044 of their Lack of Character?
Situationism Critiqued, Ethics 115 (2005), pp. 53562. These criticisms are misplaced,
as Doris rightly does not use this data as evidence. See also note 23 below, and my
Character, Consistency, and Classification (forthcoming).
certainly the most dramatic, of these is the Obedience Experiment, cited

by Doris. Subjects were assigned the role of teacher through a rigged
draw and an actor posing as a learner was strapped into a chair with
an electrode attached to his wrist. The subject was seated before a
shock generator consisting of a line of switches ranging from 15 to 450
volts in 15-volt increments. Some of these were labelled: Slight Shock
at 15 volts, Moderate Shock at 75, Strong Shock at 135, and so on up
to Danger: Severe Shock at 360, and XXX at 435. The subject was told
to ask set questions in sequence. If the learner gave the wrong answer
or no answer, the subject was to issue a shock, starting at 15 volts and
increasing by 15 volts each time. The shocks were fake, of course.
But the shocks did not seem fake: the subject was given a genuine
shock of 45 volts as a sample, and the learner grunted at 75 volts, complained that the shocks were becoming painful at 120, refused to go on
with the experiment at 150, cried I cant stand the pain! at 180,
screamed at 270, refused to answer any more questions at 300 and 315,
screamed again at 330, and gave no response at all to any of the final
eight shocks. If the subject queried the procedure or asked to stop, the
experimenter replied politely but firmly: Please continue, then The
experiment requires that you continue, then It is absolutely essential
that you continue, and finally You have no other choice, you must go
on. The experiment ended either after the subject had reached the maximum shock or the subject refused to continue after the fourth reply.15
This experiment was repeated with thousands of subjects in various
countries across three decades, and consistently around 65 per cent of
subjects continued to administer the shocks all the way up to the maximum of 450 volts. Almost all went at least as far as 150 volts. These
results remained the same in versions in which the learner complained
of the effects the shocks had on his heart.16
This is not evidence of widespread sadism. The subjects often displayed striking reactions of emotional strain and afterwards often
reported significant levels of stress and nervous tension. They acted
against their compassionate inclinations, it seems, out of obedience to
the experimenter: in a variation where the experimenter was called
away and somebody posing as another volunteer took his place, the
15
Summarized from Stanley Milgram, Obedience to Authority: An Experimental
View (New York: Harper & Row, 1974), ch. 2.
16
See Milgram, Obedience to Authority, ch. 4 experiments 1 and 2, and ch. 6 experiments 5 and 8.
266
jonathan webber
proportion of subjects continuing to the maximum shock level was

reduced to 20 per cent.17
Not just any kind of authority figure will do, as Sabini and Silver
point out. The western societies in which these experiments were conducted were not particularly authoritarian: students often disobeyed
their teachers instructions, for example, and crime was not uncommon; members of the same society asked to predict how the subjects
would behave or asked how they themselves would behave generally
predict mass disobedience, and the actual results surprise and dismay
members of that society.18
But this does not mitigate against the idea that the subjects were
willing to obey authority of a particular kind. The experimenter was
not simply an authority in the sense of being in charge, but also in
the sense that he seemed well-versed in the ways of psychological
experiments. This impression was reinforced by the language he used
in response to protests: not I would like you to continue, but the
experiment requires that you continue, and so on. His expertise
also conferred a certain moral authority on him: his instructions indicated what an intelligent and well-informed person thinks morally
appropriate in such a situation. This point is supported by the version
of the experiment in which there were two experimenters who disagreed at 150 volts over whether to continue: all the subjects stopped at
that point.19 Although this is usually referred to as an experiment concerned with obedience, therefore, what it uncovered might be better
described as deference. In the version where the experimenter is
replaced by someone who appears to be another volunteer, this person
inherits the experimental set-up and so inherits some but not all of this
deference.
The other experiments are less complicated. One of these is the
Bystander Experiment, cited by Doris to show that behaviour is significantly affected by the presence of passive bystanders. One version of
this experiment is concerned with the likelihood of a particular individual helping someone apparently in distress. Students were solicited
to participate in a study of games and puzzles. Each subject was met by
See Milgram, Obedience to Authority, ch. 12 and ch. 8 experiment 13.
Sabini and Silver, Lack of Character?, pp. 54647, esp n. 33. See Milgram,
Obedience to Authority, ch. 3.
19
Sabini and Silver, Lack of Character?, pp. 54551, esp. 547 n. 33 and 55051. See
Milgram, Obedience to Authority, ch. 8 experiment 15.
17
18
a woman posing as a market research representative, and shown into a

room either alone or in the company of an experimental confederate
posing as another solicited student. From the room, the subjects could
see in an adjoining office a ramshackle bookcase precariously stacked
with papers and equipment.
Each subject was given questionnaires to complete, as was the confederate if there was one, while the woman went into the office, and
closed the door. For a few minutes, the woman could be heard moving
things around in the office, until there was a loud crash and a womans
scream. Oh my God, my foot, she cried. I Icant moveit. Oh,
my ankle. I cantcant cantget this thing off me. And so on, for
about a minute, after which she could be heard muttering, then leaving
the office by another door. All of this was a recording playing in the
office, but only 6 per cent of subjects later reported suspecting that this
was so. Of the subjects left in the testing room alone, 70 per cent offered
to help the woman, whether by entering the office or by calling out.
Experimental confederates posing as solicited students merely looked
up on hearing the scream, then returned to the questionnaires. Of the
subjects left with a confederate, only 7 per cent offered help.20
The final set of data is from the Samaritan Experiment, cited by all
three proponents of the fragmentation theory. Students at Princeton
Theological Seminary were asked to complete questionnaires in one
building, then walk to another building and give a short talk. When
leaving the first building, some were told that they were running late
and should hurry, some that they would reach the next location on
time, and some that they were running ahead of time and would arrive
early. On the way to the second building, subjects encountered someone slumped in a doorway, apparently in need of help. The differing
religious and moral outlooks avowed on the questionnaires made
no statistically relevant difference to whether the subjects stopped to
help. Neither did it matter whether a subject was among those asked to
talk about the Parable of the Good Samaritan. What was relevant was
the degree of hurry the subjects were in. Of those who had been told
that they needed to hurry, only 10 per cent stopped to help; of those
who had been told that they would be on time, 45 per cent stopped;
20
Summary of Bibb Latan and John M. Darley, The Unresponsive Bystander: Why
Doesnt He Help? (New York: Meredith, 1970), pp. 5760. See also Doris, Lack of
Character, pp. 3233.
268
jonathan webber
and of those who had been told that they would be early, 63 per cent
stopped.21
4. Interpreting the Data
It is unclear exactly why proponents of the fragmentation theory take
this data to be evidence in their favour. Part of the problem is that all
three give ambiguous definitions of the position they are opposed to,
the understanding of character purportedly prevalent in ethical discourse. Doris writes that a trait as traditionally construed would be
reliably manifested in trait-relevant behaviour across a diversity of
trait-relevant eliciting conditions that may vary widely in their conduciveness to the manifestation of the trait in question. Merritt writes that
a trait as traditionally construed will reliably give rise to the relevant
kind of behaviour, across the full range of situations in which the
behaviour would be appropriate, including situations that exert contrary pressures. Goldie writes that we tend to think of character traits
asstable and consistent in their manifestation in thought, feeling and
action across a wide range of different situationshonest people can
be relied on to act honestly wherever honesty is appropriate.22 What is
meant by a situation being conducive to a certain kind of behaviour, or
that kind of behaviour being appropriate in that situation?
These could be read as moral terms: the proponents of the fragmentation theory could be taken to be arguing that the data indicates that
whether or not one will respond correctly to a moral demand for compassionate behaviour depends not on whether one is disposed to do so,
but on further situational features such as instructions from an authority figure, ones degree of hurry, and the passivity of bystanders. Some
of their critics certainly take this to be their argument. Michael DePaul
and Christian Miller, for example, have both claimed that the data is
21
Summary of J.M. Darley and C. Daniel Batson, From Jerusalem to Jericho:
A Study of Situational and Dispositional Variables in Helping Behaviour, Journal of
Personality and Social Psychology 27.1 (1973), pp. 100108. See also Doris, Lack of
Character, pp. 3334; Merritt, Virtue Ethics and Situationist Personality Psychology,
p. 366; and Goldie, On Personality, pp. 6263.
22
Doris, Lack of Character, p. 22; Merritt, Virtue Ethics and Situationist Personality
Psychology, pp. 36566; Goldie, On Personality, p. 50. Doris provides much the same
understanding of traits in his earlier paper, where he first puts forward his argument
against the traditional conception of character: Persons, Situations, and Virtue Ethics,
Nos 32.4 (1998), pp. 50430 (506507).
irrelevant to characterological ethical discourse, since that discourse

does not hold virtues such as compassion to be widespread, but to be
ideals for which we should aim.23
Some of the language Doris employs suggests that this is indeed his
argument. In his discussion of the Samaritan Experiment, he anticipates the objection that the subjects who did not stop to help were
indeed compassionate but also had a strong concern for punctuality,
unctuality
and argues that this cannot be the case since the demands of p
seem rather slight compared with the ethical demand to at least check
on the condition of the confederate. This seems to assume that the
point at issue is whether the subjects in the experiment had the virtue
of compassion, whether they habitually responded compassionately to
the apparent distress of others when doing so was morally appropriate.
Similarly, he describes subjects in the Obedience Experiment as engaging in destructive behaviour, and summarizes the results of the Bystander Experiment as showing that [m]ild social pressures can result
in neglect of apparently serious ethical demands.24 It seems that this
language has led critics to take his central claim to be that the experiments show that people do not possess virtues, that they do not habitually respond in the right way to particular morally relevant situational
features.
But this is not the point of his argument. In the article in which he
first presented his view, he is careful to distinguish the descriptive psychology of traits from the normative theory of virtue in the opening
pages before presenting his central argument in terms of traits rather
than virtues, and then considering whether one might argue that such
an attack on the traditional notion of traits fails to impugn the notion
of virtue employed in ethical discourse.25 His book is also generally
clear that his argument aims to show not that people do not have
Michael DePaul, Character Traits, Virtues, and Vices: Are There None?, in
Bernard Elevitch (ed.), Proceedings of the Twentieth World Congress of Philosophy. IX:
Philosophy of Mind (Bowling Green, OH: Philosophy Documentation Center, 2000),
pp. 15053; Miller, Social Psychology and Virtue Ethics, IV. See also Sreenivasan,
Errors About Errors, p. 57. Similarly, when Sabini and Silver argue that the low correlations reported in the honesty experiments are perfectly compatible with the idea
that some people are consistently honest while non-honest people are inconsistent,
they are mistakenly taking Doris to be arguing that there is no consistent virtue of
honesty rather than that people generally do not act consistently, whether this consistency be virtuous or otherwise. See their Lack of Character?, pp. 54243.
24
Doris, Lack of Character, pp. 34, 39, 33.
25
Doris, Persons, Situations, and Virtue Ethics, pp. 505, 506507, 51113.
23
270
jonathan webber
virtues, but that they do not have any character traits as traditionally
conceived. This is also the claim made by Merritt and Goldie.
That character traits and virtues are indeed conceptually distinct will
be illustrated in more detail in the next section. Here it is sufficient to
note that the idea that one can be habitually too courageous or inappropriately honest is a philosophical commonplace. We should therefore take the terms conducive and appropriate in these definitions of
character traits as non-moral, descriptive terms: the proponents of the
fragmentation theory take the data to indicate that whether or not one
will respond to the apparent distress of another person by trying to
alleviate it depends not on whether one is disposed to do so, but on
situational features such as instructions from an authority figure, ones
degree of hurry, and the passivity of bystanders.
It is unclear just why they understand the data this way. Goldie does
not formulate an argument. Merritt does so, but only briefly. She claims
that the data shows behavioural dispositions to be related to such ethically arbitrary situational factors as ones degree of hurry or the presence of passive by-standers, and that the traditional understanding of
character should lead us to expect that this is not the case.26 But without further elaboration, it is difficult to see why the traditional theory
of character should lead us to expect that people do not have commitments to punctuality or to peer-approval strong enough to preclude
their offering help in certain types of situation.
Doris is more explicit. He hinges his argument on the claim that: If
a person possesses a trait, that person will engage in trait-relevant
behaviours in trait-relevant eliciting conditions with markedly above
chance probability p.27 We can make sense of this notion of chance
behaviour only as a comparison between the agent and the population
at large: if a person possesses a certain trait, then the probability of that
person behaving in a way that manifests that trait in a given situation is
significantly greater than the probability of a randomly chosen member
of the population doing so. But this seems to imply that traits are comparative notions, and that it is incoherent to hold that a whole population could possess any given trait. This notion is therefore distinct from
the notion traditionally employed by theories that encourage us all to
develop certain traits. But Doris does not want to provide such a comparative theory of traits: he agrees that in principle, every individual in
Merritt, Virtue Ethics and Situationist Personality Psychology, p. 366.
26
27
a population could possess a [particular] trait.28 It seems, therefore,

that his reference to probability should not be taken as a constitutive
claim, but as an evidential one: an abnormally high probability that a
given person will seek to alleviate the apparent distress of another person indicates the presence of the corresponding character trait, but the
trait could be possessed without that high probability being abnormal.
This reading fits well with his remark about the Obedience Experi
ment that personality research has failed to find a convincing explanation of [these] results that references individual differences.29 The
traditional understanding of traits should lead us to expect to find the
subjects of these experiments behaving in different ways according to
their differing traits. Instead, he claims, we find near-uniformity in the
Obedience Experiment, since even those who rebelled at some point
had seemingly high levels of shock at the experimenters instruction,
and we find that the likelihood that one will help a distressed stranger
varies with ones degree of hurry and with the presence or absence of a
passive bystander rather than varying from individual to individual.
It seems that Doris interprets the data this way: evidence for the regularity theory would be provided if there were a significant diversity of
behaviour among the subjects of these experiments, but there is not;
therefore the experiments that we would expect to provide evidence for
the regularity theory fail to provide that evidence. His argument is
therefore the following modus tollens: if we possess character traits
astraditionally construed, then these experiments should yield a diversity of behaviour; but they do not, so we do not possess character traits
as traditionally construed.30

Doris, Lack of Character, p. 39. He defends this claim at length on pp. 4651.
30
See also his statement of this argument in Persons, Situations, and Virtue Ethics:
we are justified in inferring the existence of an Aristotelian personality structure when
a persons behaviour reliably conforms to the patterns expected on postulation of that
structure. In the psychological lexicon, we can say that trait attribution requires substantial cross-situational consistency If I am right about the experimental data, systematic observation typically reveals failures of cross-situational consistency (p. 507,
italics original). When Sreenivasan objects that Doris mistakes evidence of flaws in our
everyday trait-attributions for evidence that there are no traits as ordinarily construed
(Errors About Errors, p. 54), he therefore fails to engage with the central argument
Doris provides (but see also p. 56). When Kamtekar claims that such empirical
arguments against virtue ethics wrongly suppose that traits must be distinctive to the
individuals possessing them (Situationism and Virtue Ethics on the Content of Our
Character, pp. 46768), she similarly fails to engage with this argument.
28
29
272
jonathan webber
It might be objected that at least some of the data does indicate a

diversity of behaviour. The Bystander Experiment suggests that
63 per cent of the population will offer help to a distressed stranger in
another room in the absence of a passive bystander but not in the presence of one; 7 per cent of the population will offer help in both situations, and 30 per cent will offer help in neither. The Samaritan
Experiment similarly suggests that 10 per cent of the population will
help a distressed stranger even when late for an appointment, 35 per
cent will not do so when late for an appointment but will do so when
on schedule for an appointment or when having time to spare, 18 per
cent will not do so when late or on schedule but will do so when having
time to spare, and 37 per cent will not help at all. But this objection
does not challenge the claim that the Obedience Experiment shows a
remarkable uniformity among subjects.
A stronger response is to query the notion of character traits
employed in the argument. This construes a trait in purely behaviouristterms, as the disposition to respond in a certain kind of way to a
certain kind of stimulus. But a trait might rather be construed as a disposition towards certain behavioural inclinations in response to a particular kind of stimulus. Situations can present an array of features
eliciting a variety of inclinations that cannot all be acted upon. Subjects
in the Obedience Experiment, for example, are presented with the
competing demands of concern for the wellbeing of the learner and
obedience or deference to the authority of the experimenter, and
so may have inclinations against administering the shocks, but also
stronger inclinations towards obedience or deference. The emotional
strain manifested by some of the subjects sweating, shaking, stammering, and even crying as they obeyed the experimenter is evidence of
their inclinations against administering the shocks. Ones overt behaviour is the result of the relative strengths of ones competing inclinations, on this picture, and the differences between the levels at which
different subjects ended the experiment reflect differences in the relative strengths of their competing inclinations just as do their differing
manifestations of stress.
We might term this understanding of character the regularity theory, since it claims behaviour to be regulated by long-term dispositions
to have inclinations of certain strengths to behave in certain ways in
response to certain kinds of stimuli, and the patterns we discern in the
behaviour of individuals over time to reflect these dispositions. Inorder
to count as a character trait, such a disposition must yield an inclination
of about the same strength whenever the subject is confronted with the
relevant kind of situational feature where this is described in nonnormative terms, though as we shall see in the next two sections this
inclination need not necessarily be consciously felt in order to play the
explanatory role required of it.
If we employ this regularity theory when interpreting the data, we
find the positive evidence in favour of differing character traits that
Doris urges us to look for: as we have seen, the subjects of the Samaritan
Experiment differ in their inclinations towards helping a distressed
stranger and towards punctuality; the subjects in the Bystander
Experiment differ in their inclinations towards helping an apparently
distressed stranger and towards winning or maintaining peer-approval;
and the subjects in the Obedience Experiment differ in their inclinations towards obedience to authority and their inclinations against
inflicting pain.31
As we will see in the next section, moreover, the regularity theory
underlies key aspects of Aristotles ethics and of some prominent recent
works influenced by it. The argument against characterological ethical
discourse mounted by Doris therefore misses its target: so long as that
discourse retains the regularity theory, the data Doris cites will present
no threat to it. But this is not yet to show that there is any positive reason to prefer the regularity theory to the fragmentation theory, which
will be shown in section six.
5. Character in Ethical Discourse
Various key aspects of characterological ethical discourse involve this
notion of a trait as a relatively stable disposition to be inclined with a
certain strength towards a certain kind of behaviour in response to
a certain kind of situational feature. The wellspring of the tradition,Aristotles Nicomachean Ethics, contains a number of prominent
31
As we saw in section 3, the authority involved in the Obedience Experiment is
that of an institutional expert and moral guide, not simply that of the person in charge.
Sabini and Silver question whether the behaviour of the subjects in the Samaritan
Experiment should be described in terms of punctuality, preferring instead to invoke
potential embarrassment at arriving late (Lack of Character? Situationism Critiqued,
p. 558). But these explanations seem complementary rather than contrary. Potential
embarrassment could enforce ones acting on ones commitment to punctuality, and
without that commitment it is difficult to see why somebody would be embarrassed by
being late rather than proud of having stopped to help.
274
jonathan webber
e xamples, and descendents of these ideas prominent in recent literature retain the conception of trait they involve, as we shall see.
Aristotle makes it abundantly clear that the kinds of traits that he
recommends as virtuous are certainly not dispositions to behave in a
certain kind of way whenever or almost whenever a certain kind of
situational feature is present. He considers such traits to be vices
of excess: giving money away at every or almost every opportunity is
not generosity, but profligacy; never or hardly ever recoiling from danger is not courage, but foolhardiness; and so on. The vice consists not
just in behaving in a certain way too often, but in behaving in that way
on inappropriate as well as appropriate occasions. Possessing a specific
virtue, on the other hand, means that one will act in a certain way
when one should, towards the things one should, in relation to the
people one should, for the reasons one should, and in the way one
should.32 Someone might commit murder in the face of real danger, but
would not thereby display the virtue of courage, since the intrepid act
was inappropriate for other reasons. Someone who tells the truth even
in situations where the demands of compassion make it inappropriate
does not fully possess the virtue of honesty.
Underlying this theory is the idea that each trait leads one to be
inclined with a certain degree of strength towards certain kinds of
actions in certain kinds of situations, and that such traits are virtues
only if these inclinations are tempered by other inclinations that constrain the range of occasions on which they will result in action. This
explains why Aristotle takes the full possession of one virtue to require
full possession of all virtues, and to hold that virtue in the primary
sense consists in this possession of full virtue. There is, for Aristotle, a
single web of interdependent virtues: full possession of any one virtue
means habitually being inclined to behave in a certain way with the
right degree of strength in the presence of a certain situational feature,
where what is right is relative to the strength of ones other habitual
inclinations in response to other possible situational features.33
John McDowell has argued for a similar position on the grounds that
virtue issues in nothing but right conduct.34 Philippa Foot, on the
32
Aristotle, Nicomachean Ethics, translated by Christopher Rowe, with introduction
and commentary by Sarah Broadie (Oxford: Oxford University Press, 2002). Quotation
is from 1106b2123; see also 1115b1520.
33
Nicomachean Ethics, 1144b321145a2.
34
John McDowell, Virtue and Reason, Monist 62 (1979), pp. 33150. Reprinted in
his Mind, Value, and Reality (Cambridge, MA and London: Harvard University Press,
1998), 2.
other hand, has disagreed with such a stringent criterion for virtue,
arguing that a trait counts as virtuous so long as it usually issues in right
conduct. On the occasions when it does not, it is not operating as a
virtue because its manifestation ought to have been precluded by
another trait, which the agent unfortunately lacks. But one can, according to Foot, possess a virtue that does not always operate as a virtue.35
McDowell and Foot nevertheless agree that the traits in question are
dispositions to be inclined with a certain degree of strength to a certain
kind of action in response to a certain kind of situational feature; they
disagree only over the criteria for such traits to be counted as virtuous.
Where McDowell holds that the murderers trait of courage cannot be
virtuous because it is not accompanied by the right level of aversion to
killing, Foot holds that it is virtuous so long as it is usually manifested
in good behaviour.
Aristotles discussion of self-control and weakness of will also seems
to involve this notion of character trait. The difference between g enuine
virtue and mere self-control, he claims, is that only the latter involves a
struggle to overcome contrary inclinations; the difference between vice
and mere weakness of will is that only the latter involves such a struggle; and the difference between self-control and weakness of will is that
the inclination one believes one ought to act on wins the struggle in the
former case, loses it in the latter. The picture is complicated, of course,
by the fact that one might have an additional disposition towards being
inclined to do whatever one believes one should do. But even when
such a trait is in play, it seems that weakness of will results from the
inclination acted upon being stronger than the combination of the
inclination one believes one should act on and the inclination to do
whatever one believes one should do. Self-control results from this
combination being stronger than the inclination to behave otherwise.
McDowell has recently echoed this view, claiming that the virtuous
person acts for a reason that is apprehended, not as outweighing or
overriding any reasons for acting in other waysbut as silencing
them.36 The virtuous person pursues the right course of action without
consciously considering any other. One might disagree and argue that
an action is virtuous even when the right reason is apprehended
as more important than reasons for contrary actions, whereas merely
35
Philippa Foot, Virtues and Vices, in her Virtues and Vices and Other Essays in
Moral Philosophy (Oxford: Blackwell, 1978), III.
36
Aristotle, Nicomachean Ethics, book 7 chapters 110, esp. 1146b315; McDowell,
Virtue and Reason, 3.
276
jonathan webber
self-controlled action involves a genuine struggle to act on the right

reason. But it would remain that virtue and vice involve a greater
difference between the strengths of competing inclinations than do
strength and weakness of will.
This difference between acting as a result of conscious considerationand acting without such conscious consideration reflects the difference between the two ways in which Aristotle and his followers
think that traits lead to action. When faced with a situation that seems
to call for mutually exclusive responses, these responses reflect different if overlapping sets of ones values and attachments. One will therefore need to deliberate, to consider which of the conflicting values one
attaches most weight to, which one is most strongly inclined to act
upon.37 But sometimes it just seems to one without conscious deliberation that a certain course of action is the right one to pursue. Aristotle
considered character to be reflected in such instances no less than in
the outcomes of deliberation, and this idea is emphasized by McDowell
along with Iris Murdoch, David Wiggins, and Martha Nussbaum,
among others.38
An inclination to behave in a certain way need not result directly
from a disposition to have that kind of inclination in response to a certain kind of feature, therefore, since it could result from the combination of a number of dispositions elicited by different features of the
situation. One may be less strongly inclined to tell the truth to someone
who might be upset by it than one would be otherwise, but be inclined
to do so nonetheless; ones inclination would therefore result from ones
disposition towards a strong inclination to tell the truth in conjunction with ones disposition towards a less strong inclination not to
upset people. This resulting inclination could result from conscious
37
See Bernard Williams, Internal and External Reasons, in his Moral Luck:
Philosophical Papers 19731980 (Cambridge: Cambridge University Press, 1981),
p. 104. Aristotles lengthy and scattered thoughts on deliberation are discussed in:
Richard Sorabji, Aristotle on the Role of Intellect in Virtue, pp. 205207; David
Wiggins, Deliberation and Practical Reason, pp. 23137; both in Essays on Aristotles
Ethics, ed. Amlie Oksenberg Rorty (Berkeley, CA: University of California Press,
1980).
38
Nicomachean Ethics, 1143b1114; see also 1109b1823, 1126b24 and 1142a27
30; Iris Murdoch, The Sovereignty of Good Over Other Concepts (Cambridge: Cambridge
University Press, 1967), p. 20; McDowell, Virtue and Reason, 2; Wiggins, Deliberation
and Practical Reason, pp. 23233; Martha C. Nussbaum, The Discernment of
Perception: An Aristotelian Conception of Private and Public Rationality, in her Loves
Knowledge: Essays on Philosophy and Literature (Oxford: Oxford University Press,
1990), esp. pp. 74 and 79.
deliberation, or from the processing underlying ones perception of the

situation. In the latter case, one need not even consciously feel the
inclination not to upset people.
The ideas surveyed in this section, drawn from Aristotle and prominent recent discussions of virtue involve the same idea of a character
trait, one that stands in contrast to the conception employed by the
proponents of the fragmentation theory of character. We have already
seen that the claim that people differ in character is positively confirmed by the data cited against it so long as we interpret that data using
this notion of traits commonly employed in characterological ethical
discourse, rather than the behaviourist notion employed by the critics
of that discourse. But this leaves open the question of which of these
conceptions we should prefer. The next section presents an argument
in favour of retaining the traditional notion employed by Aristotle and
his followers, rather than replacing it with the behaviourist notion
employed by Doris, Merritt, and Goldie.
6. Explanation and Prediction
The traditional conception of a character trait as a disposition towards
a certain inclination in the presence of a certain situational feature is
preferable to its behaviourist rival for the pragmatic reason that it provides a deeper level of explanation, and so has greater predictive power.
The behaviourist notion explains an action by the trait of responding in
a certain way to a certain combination of environmental stimuli, such
as sailing in rough weather with friends or passing a distressed stranger
while in a hurry. The traditional notion, however, can allow us to
explain why that combination of stimuli leads to that behaviour from
that person. It does so by referring to a number of inclinations and
their relative strengths elicited in that person by that situation. We can
therefore predict that persons behaviour in a novel situation that combines only features already observed, by considering the outcome of
the combination of the traits known to be elicited by those features.
The behaviourist notion, on the other hand, requires us to consider the
novel situation as eliciting a trait not previously observed, and therefore unknown.
Advocates of the behaviourist notion might argue that this deeper
level of explanation is simply not available, that we can do no morethan
observe overt action in response to environmental stimuli. They might
278
jonathan webber
further claim that belief in this deeper level of explanation is the reason
why predictions of behaviour often fail. One way to decide whether this
is true would be to undertake longitudinal studies. If consistent observation and classification of the actions of a few individuals over significant stretches of time were to yield reliable predictions of their
behaviour in novel situations, then the traditional notion would be
shown to be preferable to the behaviourist notion. If it were to yield no
such reliability, then the traditional notion would have no such
advantage.
Such longitudinal studies would require long-term detailed surveillance of the public and private lives of people unaware of being under
surveillance and are therefore unavailable to professional psychologists
for both logistical and ethical reasons.39 But these kinds of studies are
not necessary for settling this issue. Latitudinal studies, in which different subjects are tested in the same situation, can provide evidence of
this deeper level of explanation, and hence also evidence of character
traits as traditionally construed.
One study that provides such evidence was originally designed to
explain the following difference in homicide rates between the n
orthern
and southern regions of the USA: white males in large southern cities
are no more likely to commit homicide than are their counterparts in
large northern cities, but those who do commit homicide are significantly more likely to do so as a result of an argument; and white males
in rural areas and small urban areas in the southern states are no more
likely to commit homicide as a result of a crime than are their northern
counterparts, but are twice as likely to commit homicide as a result of
an argument.40
The experimenters tested the hypothesis that these differences were
due to a southern culture of honour, more pronounced in rural and
small urban areas than in larger, more cosmopolitan cities. Their subjects were white, non-Jewish, non-Hispanic, male citizens of the USA
studying at the University of Michigan. They were divided between
39
Some philosophers argue that longitudinal studies are available in the form of
novels and the writings of other acute observers of human behaviour. I defend this
view in Character, Common-Sense, and Expertise. But since the traditional notion of
a character trait can be shown to be preferable to its behaviourist rival by reference to
recent experimental data, there is no need to make a more controversial appeal to nonexperimental literature here.
40
See Richard E. Nisbett and Dov Cohen, Culture of Honor: The Psychology of
Violence in the South (Oxford and Boulder, CO: Westview Press, 1996), p. 21.
southerners, defined as having lived in southern states for at least six

years, and northerners, who had lived in southern states for less than
six years if at all. On average, the southerners had lived in southern
states for 80 per cent of their lives, the northerners for 5 per cent.41
Evidence of a southern culture of honour would be provided if southerners were shown to be significantly more likely than northerners to
react violently to insults.
Each subject was asked to complete a questionnaire, deliver it to a
room at the end of a long corridor, and return. Some subjects passed
someone in the corridor who had to close a filing cabinet to let them
pass, then had to do so again on their return, and so bumped them with
his shoulder and insulted them as they passed. For control subjects, the
corridor was empty. There are three variations of the experiment. In
one, subjects were then asked to complete a story, which began at a
party with Jill telling her fianc Steve that a mutual friend Larry, who
knew that Jill and Steve were engaged, had been making passes at her,
and stopped with Steve seeing Larry trying to kiss Jill. 75 per cent of
insulted southerners ended the story with Steve injuring or threatening
to injure Larry, whereas only 20 per cent of control southerners did so.
Having recently been insulted made no statistically relevant difference
to how northerners ended the story.
In the second variation, saliva tests were taken before the questionnaires were filled in and after the subject returned from dropping off
the questionnaire. Differences in levels of two hormones were measured: cortisol, associated with high levels of stress, anxiety and arousal;
and testosterone, associated with aggression and dominance behaviour. The cortisol levels of insulted southerners rose by an average of
79 per cent during the experiment, whereas those of control southerners rose by an average of 42 per cent. The cortisol levels of insulted
northerners rose by an average of 33 per cent, compared with the control northerners average of 39 per cent. Similarly, testosterone levels
rose 12 per cent for insulted southerners and only 4 per cent for control
southerners, where these rose 6 per cent for insulted northerners and
4 per cent for control northerners. So there is a significantly larger
average increase in both cortisol and testosterone levels during the
experiment for insulted southerners than for control southerners or
any northerners.
41
For more on the definitions of northern and southern used here, see Nisbett and
Cohen, Culture of Honor, Appendix C.
280
jonathan webber
In the third, chicken variation, the subject returning along the corridor was faced with somebody coming the other way, who was six feet
and three inches (one metre and 90 centimetres) tall, weighed 250
pounds (17 stone and 12 pounds; 114 kilograms), and played college
football. All subjects gave way. They differed in the distance at which
they did so. Insulted southerners yielded at an average of around three
feet (90 centimetres), control southerners at about nine feet. Insulted
northerners yielded at around six feet, control northerners at around
five and a half. The insults therefore made a significant difference in the
case of southerners, but not in the case of northerners.42
If we adopt the behaviourist notion of traits, these experiments
suggest only that none of the subjects have the trait of responding
violently to insults in experimental situations, that southerners are
more likely than northerners to have the trait of completing the story
violently after being insulted in an experimental situation, and that
southerners are more likely than northerners to have the trait of displaying increased bravado shortly after being insulted in experimental
conditions. These are distinct traits. The data concerning cortisol and
testosterone levels does not report any behaviour, and so does not indicate any difference in traits. The traits that are indicated, moreover,
cannot be linked to any data concerning homicide occurring outside
experimental conditions, since the traits manifested are relative to
those conditions. Indeed, since homicide is committed in a wide range
of situations, the behaviourist notion of traits precludes any explanation of the homicide data in terms of trait differences between southerners and northerners.
The traditional notion of traits, on the other hand, allows us to
explain the homicide data and the data from the story and chicken
variations of the experiment in terms of a trait prevalent among southerners but not northerners: a disposition to be strongly inclined to
respond violently when insulted. Evidence of the inclination elicited
by insults is provided by the saliva tests. The fact that nobody actually
did respond violently can be explained by the other relevant inclinations in play in the experimental situation. This information allows
predictions of results of future experiments that also test reactions to
insults or slights among northerners and southerners, but test these in
42
Nisbett and Cohen, Culture of Honor, ch. 4. For further data in support of the
theory of a southern culture of honour, see chs. 3 and 5; for more on what this culture
involves, see chs. 1 and 6.
different ways. The traditional notion of traits therefore genuinely does

have an explanatory and predictive power that the behaviourist notion
lacks. We should therefore choose to interpret data using this traditional notion.
7. Conclusions and Implications
At issue between the regularity and fragmentary theories of character
is a conceptual question: should we prefer the traditional notion of
traits that leads to the regularity theory, or the behaviourist one that
leads to the fragmentary theory? We have seen that the traditional
notion is preferable because it allows for a deeper level of explanation,
and therefore has greater explanatory and predictive power.
Evidence of this deeper level of explanation is provided by a set of
experiments that measure not just overt behavioural responses to environmental stimuli, but also inner events mediating between stimulus
and response: the saliva tests measure physiological priming for aggression and dominance behaviour, and the story experiment measures
cognitive priming for such behaviour; neither measures such behaviour itself. The chicken variation measures subtle dominance behaviour indicative of violent inclinations, but not violent behaviour. The
Samaritan and Bystander Experiments, on the other hand, report
only stimuli and overt behavioural responses, and the Obedience
Experiment measures only these things. There is a methodological lesson in this: if we are to retain the regularity theory of character and the
related notion of traits as dispositions to behavioural inclinations, then
empirical data employed in philosophical discussions of character
should be drawn from experiments reporting inner mediating events,
not those reporting only behavioural responses to stimuli, since inclinations are not always manifest in behaviour. In addition to this, there
are further implications for the philosophical discussion of character
and virtue to be drawn from the foregoing discussion.
First, the data concerning responses to insults also indicates that the
relevant dispositions vary with cultural background, which shows that
we could in principle inculcate a particular kind of habitual response to
insults in our children, at least by manipulating their cultural background. Whether an adults habitual response to insults can be altered
is another issue. Perhaps it is impossible, perhaps it requires relocation to another culture, or perhaps it is possible without relocation.
282
jonathan webber
Perhaps it is possible but extremely difficult, either with or without

relocation. This is an empirical issue. Whether adults can be enjoined
to inculcate in themselves a particular kind of habitual response to
insults, and if so how this might most expediently be done, are therefore matters requiring further empirical investigation.
It may be, of course, that some traits do not vary with cultural background, while others do. Perhaps our levels of obedience or deference
to certain types of authority, for example, are hard-wired. If so, ethical
theories enjoining inculcation of any contrary inclination would be
unacceptable, though ones urging us to resist inclinations issuing from
such hard-wired traits would not. Ethical theories of characterdevelopment would also have to consider the behaviour likely to result
from the combination of each recommended trait with this set level of
obedience or deference to certain types of authority. Further empirical
research is required, therefore, to discern whether any traits are inherent and unchangeable.
Where the recent philosophical literature on ethics and empirical
psychology has discussed character in general, moreover, it seems necessary to treat traits severally. Perhaps concern for peer approval is
unalterably high in humans, for example, where concern for the welfare of others is culturally variable. Perhaps concern for punctuality is
very difficult to change in adult life, but concern for keeping ones word
is not. Ethical injunctions and their implementation will need to take
account of such differences between traits. But this is not so straightforward as it might seem, since there seems no obvious reason to
believe that the folk psychological terms employed in ordinary English
individuate traits correctly. If traits are dispositions towards certain
inclinations of certain strengths in response to certain environmental
stimuli, their individuation requires reference to the range of stimuli
that elicit them. Discussion of the individuation of traits should also
draw on empirical investigation.
Philosophical discussions of the ethics of character therefore require
roponents
a considerable amount of empirical information. Two of the p
of the fragmentation theory have also argued, as we have seen, that
the project of character-development should be replaced with one of
situation-management. It might turn out that this advice is right, even
though the fragmentation theory of character need not be embraced.
It might simply be that character-development is not feasible, either
generally or with respect to certain traits. Whether or not situationmanagement is preferable to character-development, therefore, is an
issue to be settled by the same kinds of empirical investigation involved

in the discussion of the individuation and development of behavioural
inclinations.
This is not to deny, of course, that ethics is an irreducibly normative
discipline. It is not to suggest that psychology is a guide to moral value.
It is just to point out that since a practical ethical recommendation is
acceptable only if it is reasonable to expect people to follow it, theories
that recommend character-development should be formulated and
assessed with reference to empirical studies of the nature of behavioural inclinations, studies that are becoming increasingly available.
part V
ethics and moral philosophy
Doing Harm, Allowing Harm, and

Denying Resources
Timothy Hall*
I
Many non-consequentialists assert a moral difference between doing
and allowing harm. This view is attractive because of its prominence in
ordinary morality and its substantial intuitive purchase.
Most philosophers insist on a justification for the claimed moral difference between doing and allowing harm beyond appeals to intuition,
however. Proponents of the moral difference between doing and allowing have therefore undertaken two tasks. The first is to characterize the
distinction between doing and allowing harm in a way that classifies
all, or at least the largest part of, candidate behaviors as doings or allowings of harm. The second is to defend a moral difference between doing
and allowing harm so characterized.
There is a large and rich literature devoted to these tasks, accompanied by objections from a relentless and resourceful group of critics.
Within this literature there is a sizeable discussion of a group of behaviors that have come to be called the withdrawal of aid. For example:
The Withdrawal of Life Support: Smith, a terminally ill patient, is in great
pain and believes his life no longer worth living. He requests of Jones, his
doctor, that Jones remove the life support machine he connected to
Smith. Jones removes the life support from Smith, and Smith dies.
Withdrawals of aid have been of interest because they seem to many

commen-tators to be counterexamples to characterizations of doing
or allowing harm that are otherwise promising. A natural first attempt
to characterize the doing of harm, for example, is in terms of an agent
*I owe thanks to Barbara Herman, Seana Shiffrin, Michael Otsuka, Frances Kamm,
Jeff McMahan, Derk Pereboom, David Christensen, Donald Loeb, Shelly Kagan, Kai
Draper, and my colleagues at the Philosophy Department at Oberlin for comments on
this manuscript: Al MacKay, Peter McInerney, Martin Thomson-Jones, Todd Ganson,
and Dorit Ganson. Thanks, too, to anonymous referees and to Thom Brooks for constructive criticisms and editing suggestions.
288
timothy hall
acting. Yet the example above involves action, and many respondents
will claim that Jones lets Smith die.1
Matters are further complicated by examples such as the following:
The Enemy: Smith, a terminally ill patient, is in great pain, but wishes to
continue to be kept alive by life support devices. Smiths enemy, Jones,
learns of Smiths incapacity and sneaks into the hospital. Jones disconnects Smith from the life support device, and Smith dies.2
Many respondents will describe this example as a killing. Exactly

how this example differs from The Withdrawal of Life Support in a way
that explains a difference in terms of killing and letting die is not obvious, however.3
Moreover, many proponents of the moral difference between doing
and allowing harm are wont to explain many moral differences among
withdrawal cases in terms of the difference between doing and allowing harm. This makes the metaphysical question of their difference
more pressing.
Many commentators, such as Foot, Davis, Kamm, Bennett, Boorse
and Sorensen, Kagan, Hanser, Draper, and McMahan have addressed
the w
ithdrawal of aid.4 Several of them have offered positive accounts
1
It may not be obvious that the distinction between killing and letting die is an
instance of the distinction between doing and allowing lethal harm. One possibility is
that an agent could act and do enough damage to do lethal harm, but nonetheless not
kill because the relation between the action and the death of the victim is not sufficiently direct or immediate. Nevertheless, I shall, in what follows, use killing as if it
were synonymous with doing lethal harm, for simplicity of exposition.
2
See comparable examples discussed in Shelly Kagan, The Limits of Morality
(New York: Oxford University Press, 1991), pp. 101103; and Jeff McMahan, The Ethics
of Killing (New York: Oxford University Press, 2002), p. 379.
3
One might point out that the fact that Smith wants to die in the first example but
does not want to die in the second can explain how Smith is harmed in the one case but
not the other. What still remains to be explained, however, is the difference in agency
that many assert between the examples. Why does Jones allow death in the first case,
harmful or not, and kill in the second?
4
Philippa Foot, Killing and Letting Die, in B. Steinbock and A. Norcross (eds.),
Killing and Letting Die (New York: Fordham University Press, 2nd edn, 1994), pp. 280
89; Philippa Foot, Euthanasia, Philosophy and Public Affairs 6 (1977), pp. 85112;
Christopher Boorse and Roy Sorensen, Ducking Harm, The Journal of Philosophy 85
(1988), pp. 11534; Nancy Davis, The Priority of Avoiding Harm, in Steinbock and
Norcross (eds.), Killing and Letting Die, pp. 298354; Kai Draper, Rights and the
Doctrine of Doing and Allowing, Philosophy and Public Affairs 33 (2005), pp. 25380;
McMahan, The Ethics of Killing, pp. 37992; Jeff McMahan, Killing, Letting Die,
and Withdrawing Aid, Ethics 103 (1993), pp. 25079; idem, A Challenge to Common
Sense Morality, Ethics 108 (1998), pp. 394418; Matthew Hanser, Killing, Letting
Die, and Preventing People from Being Saved, Utilitas 11 (1999), pp. 27795; Kagan,
The Limits of Morality, pp. 10111; Frances Kamm, Morality, Mortality, vol. II
doing harm, allowing harm, and denying resources 289
of the m
etaphysics of withdrawals. These accountswith the exception
of Hansers, and Boorse and Sorensenscharacterize withdrawals as
the doing or allowing of harm, with the aim of employing that metaphysicalcharacterization in subsequent moral analysis of withdrawal
behaviors.
I wish to offer an alternative view. Withdrawals of aid have a greater
significance than as a proving ground for broader theories of doing
and allowing harm. I will argue that withdrawals of aid reveal that nonconsequentialism is best understood as consisting in three broad, morally distinct categories of behavior where commentators have ordinarily
identified two. I will call these categories standard doings of harm,
standard allowings of harm, and denials of resources.
I will argue that, with one exception, denials of resources are morally
distinct from standard doings or allowings of harm. They are morally
distinct because their moral status depends to a large extent on a right
to the resource denied. The moral statuses of standard doings and
allowings of harm, on the other hand, are not influenced to a great
degree by such rights. The exception is cases in which a person is herself the resource that she denies. Those cases are morally on a par with
standard allowings of harm.
A distinctive feature of my view is that the claim of moral distinction
is independent of whether denials of resources are the doing or allowing of harm. So the metaphysical question of whether denials of
resources are doings or allowings of harm can be left aside in understanding their moral status, I shall argue. I qualify standard doings and
allowings of harm as standard precisely to leave open the possibility
that denials of resources are also doings or allowings of harm.
I will proceed by considering Jeff McMahans analysis of the withdrawal of aid in section II. McMahans is the most sophisticated attempt
to analyze the moral status of withdrawals of aid in terms of doing and
allowing harm, but my comments will be applicable to other views of
this kind. In section III, I offer an argument for the moral distinctness
of denials of resources based on a comparison of well-matched examples. In section IV, I argue that the rights of persons conjoined with any
of a range of widely held justifications for the distribution of resources
in political society imply the moral distinctness of denial behaviors.
(New York: Oxford University Press, 1996), p. 28; Jonathan Bennett, The Act Itself
(New York: Clarendon Press, 1995).
290
timothy hall
In section V, I consider several objections to the two arguments. In

section VI, I conclude.
II
I will follow McMahan in his characterization of the withdrawal of aid.
In the paradigm cases of withdrawing aid, there is some threat to a
person, aid or protection against that threat, and a party other than the
threatened person who acts to withdraw the aid exposing the first party
to the threat.5 So, for example, in The Withdrawal of Life Support, there
is a patient who is threatened by organ failure, disease, or something
similar. The life support provides protection of a kind against this
threat. A doctor withdraws the life support, thus exposing the patient
to the preexisting threat.
McMahan offers an account of what makes withdrawals of aid doings
or allowings of harm. McMahan follows Foot in claiming that when aid
or support is initially provided by the withdrawing agent the withdrawal is an allowing of harm. When withdrawn aid is not initially
provided, the withdrawal is a doing of harm.6 More specifically,
McMahan claims that:
Iff:
(a)the withdrawing agent did not create or cause the threat to the threatened
party, and
(b) the aid or protection is provided by the withdrawing agent,
then the withdrawal of aid is an allowing of harm, and
iff a withdrawal is not an allowing of harm, it is a doing of harm.7
Though McMahan has been concerned largely with lethal harms, and thus the narrow distinction between killing and letting die, the full body of his work makes clear
that he means to characterize a broader distinction between doing and allowing harm.
See McMahan, The Ethics of Killing, p. 381; Killing, Letting Die, and Withdrawing Aid;
A Challenge to Common Sense Morality.
6
See Foot, Euthanasia, pp. 85112.
7
I have simplified McMahans account here somewhat, in part for expository
purposes, but in part for other reasons. In Killing, Letting Die, and the Withdrawal of
Aid, and in The Ethics of Killing, McMahan offers a further condition of a withdrawals
being the allowing of harm: that the aid be in progress or continuing, and not selfsustaining. As far as I can tell, however, this condition is redundant. It is equivalent
to requiring that the harmed party must have been threatened independently of the
5
Implicit in McMahans account are two further claims: the distinction

between doing and allowing harm exhausts withdrawal of aid cases,
and doing and allowing harm are mutually exclusive.
Some examples illustrate McMahans account. Consider again The
Withdrawal of Life Support. Here, McMahan identifies Joness action as
a letting die.8 The fact that Jones provided the aid to Smith initially is
the reason that, according to both Foot and McMahan, Joness withdrawal is a letting die.
For Foot and McMahan both, the provision of aid also includes
being part of a team, another member of which physically provided
aid, and whose collective goals are connected with the provision and
withdrawal of aid. So, the doctor in the following example also lets the
patient die:
A Teams Withdrawal of Life Support: Smith, a terminally ill patient, is in
great pain and believes his life no longer worth living. He requests of
Jones, a doctor, that Jones remove the life support machine sustaining
him. The support was originally connected by Black, another member of
the same medical team handling Smiths case. Jones removes the life support from Smith, and Smith dies.9
McMahan also claims that ones being useful, without performing

any aiding action, is a way of providing aid. In the following case, for
example, McMahan claims that because one is plugged into the violinist, one provides aid to him. So, the benefactor allows the violinist to
die by unplugging:
Thomsons Violinist: You are kidnapped and connected by a tube to an
innocent, unconscious violinist. As your organs are his only match, if he
is to live you must remain connected to him for nine months. At the end
of this time, the violinist will have made a full recovery. Supporting him
will keep you confined near him for nine months, with all of the associated burdens. If you are to free yourself, you must disconnect from the
withdrawing agent, rather than the agent creating a threat against a person. For more,
see McMahan, Reply to On Harming and Killing: Replies to Hanser, Persson and
Savulescu, and Wasserman, forthcoming.
8
There is the related question of whether the withdrawal here is passive, rather than
active, euthanasia. As it is beyond my purposes here, and as most commentators take
the active/passive distinction to be an instance of the distinction between doing and
allowing of harm, I leave this question aside.
9
McMahan also means provision to include a persons having a role by which he
might appropriately control the aid withdrawn, even if the person did not originally
provide the aid. I suspect that what McMahan means by a role reduces to the collective
action or ownership possibilities, however. See McMahan, The Ethics of Killing, p. 382;
and Killing, Letting Die, and Withdrawing Aid, pp. 25079.
292
timothy hall
violinist, which will result in his death. You disconnect yourself, and the
violinist dies.10
McMahan goes further than Foot in claiming that provision includes

cases in which an agent owns, or has some similar strong moral claim
to, the aid in question. In a case like the following, McMahan would
claim that Jones lets die rather than kills:
Joness Antidote: Smith and Jones live in a remote area somewhat near one
another, but are otherwise distant acquaintances. It happens, rarely, that
poison gas is released from a volcano in the area. To guard against this
possibility, Jones and Smith each have on hand an antidote to the poison,
which must be taken right before exposure. Before the danger, Smith
traded his antidote to a third party for something of value. Now the volcano releases poison gas, threatening them both. Smith, not wanting to
die, takes Joness antidote. Before Smith can use it, Jones takes it back.
Smith dies.
Many people would agree that Jones lets Smith die. Jones takes back
what is his, what he would have been permitted to refuse to provide
initially. So, Jones allows a threat to run its course, rather than doing
harm to Smith.
McMahans analysis of the withdrawal of aid has two problems. The
first is that not all careful readers will agree with McMahans intuitions
about doing and allowing harm in the examples above. Some respondents will be unwilling to say that the behavior in question is either
doing or allowing harm. For example, I can attest that some readers
will claim that Jones neither kills nor lets die in the examples above.
They will limit themselves instead to the claim that Jones withdraws aid
from Smith and Smith dies as a result.11
Some readers will, of course, share McMahans intuitions. Rather
than attempt to resolve this issue, however, I will leave it aside. For I am
interested in an independent problem of moral explanation.
McMahans account is, strictly speaking, a metaphysical one.
However, it is clear that a large part of his motivation is normative.
10
Judith Thomson, A Defence of Abortion, Philosophy and Public Affairs 1 (1971),
p. 5.
11
On Hansers view, Jones prevents Smith from being saved, rather than killing
Smith or letting Smith die. Jones acts to deny to Smith what Smith would otherwise
save himself with. So, Hanser claims, Jones does not let Smith die. On the other hand,
Jones does not cause the threat to Smith that kills him. So, Jones does not kill Smith,
either, according to Hanser. See Hanser, Killing, Letting Die, and Preventing People
from Being Saved, pp. 27795.
McMahans account carries with it the promise of generality in moral

explanation. For example, being able to determine of a specific withdrawal of aid, like the withdrawal of life support, that it is a doing of
harm would presumably reveal a significant similarity between the
withdrawal and a paradigm doing of harm, like shooting or stabbing.
This similarity, in turn, would presumably reveal some significant
moral similarity between the withdrawal and non-withdrawal behaviors, allowing for general moral explanation across the behaviors.12
However, McMahans analysis does not permit a satisfying generalityof moral explanation. For there to be a satisfying generality of explanation across a range of phenomena, there must be a close similarity
among the explanantia in the specific instances of the phenomena. For
a difference in voltage potential to provide a general explanation of
various phenomena like a working light switch, lightning, and dust collecting on a television screen, the specific instances of voltage potential
in each of these cases must be closely alike. For there to be a general
explanation of the moral statuses of various standard doings and allowings of harm and of withdrawals of aid, there must be a close similarity
in the specific instances of doing or allowing harm across all of these
cases. On McMahans account, various cases of doing and allowing
harm are not at all closely alike.
Consider the three important ways in which a withdrawal might be
an allowing of harm for McMahan: physically providing the aid that
one withdraws, being oneself the aid one withdraws, and owning or
having some claim to the aid withdrawn. The three salient facts here
are very different: physically providing a resource, being a resource,
and owning a resource (or having some other determinate right to it).
There does not appear to be any clear similarity on the basis of which
to provide a general moral explanation in terms of allowing harm
across all of these withdrawal-allowings.
McMahan, of course, attempts to unify these cases under the heading of provision. The use of this term only serves to forestall the
12
This characterization of McMahans ambition is not meant to imply a commitment to the claim that doing and allowing harm are always morally significant. Nor is
it meant to imply that there are no other morally important factors other than doing
and allowing harm in withdrawal cases or standard doings or allowings of harm. An
insightful generality of explanation is consistent with both of these concessions.
For a thorough discussion of generality in moral explanation, see Don Loeb,
Generality and Moral Justification, Philosophy and Phenomenological Research 56
(1996), pp. 7996.
294
timothy hall
question, however. Indeed, one must stretch to explain how the latter
two possibilities are cases of providing at all. Merely having a right to a
resource used by someone else is not obviously accurately described as
a case of providing. Nor is having ones presence be useful to someone
else obviously a case of providing.
Perhaps, though, one might claim that there is an attenuated sense of
providing that requires only that resources used by someone be owned
by a person to be provided by that personin something like the sense
it might seem correct, or at least not clearly wrong, to say that a victim
provides money to a thief. We might, then, notice that in the case in
which one is a resource one also has a right to the resource. For one has
a right to ones own body. Thus having ones body be useful to another
might be another way of providing.
However, the fact that an ownership right to a non-person resource
might well haveand on many views does havea very different
source than a right to oneself weakens the similarity between these two
ways of providing, even if they are really ways of providing. Furthermore,
having a right to a resource is still a very different matter than physically providing a resource. Physically providing a resource is a physical
act. Ownership is a moral claim.
One might claim that, in the appropriate conditions, physically providing a resource also generates a right to that resource. Then one
might claim that being a resource, owning a resource, and physically
providing a resource are all alike: they are ways one might have a right
to a resource.
However, physically providing a resource does not, in normal circumstances, give any right to the provider. The right to the resource is
usually predetermined by ownership or some other claim. I do not
come to have a right to Smiths life-support machine, for example,
merely by connecting Smith to it.
Furthermore, the resulting interpretation leaves aside McMahans
notion of provision altogether. Rather than there being three ways of
providing a resource, these three possibilitiesphysically providing a
resource, owning a resource, and being a resourcewould simply be
three ways of having a right to resource.
Finally, and most importantly, there would still be a gross dissimilarity between standard cases of allowing and withdrawal-allowings.
Withdrawal-allowings would be those allowings in which an agent has
some right to the resource he withdraws, whether through ownership,
being the resource, or having physically provided the resource.
A paradigm, non-withdrawal case of allowing harm would be an

agent not acting to prevent some independent threat from running its
course against someone. These two kinds of behaviors are still very
different. While both behaviors require some independent threat, that
fact is not sufficient to qualify them as allowings, as doing-withdrawals
would have this feature, too. Apart from there being an independent
threat, the two kinds of allowing behavior are markedly dissimilar from
one another. In the case of standard allowing, the feature other than
there being an independent threat is an agents refraining from acting.
In the case of a withdrawal-allowing, the feature other than there being
an independent threat is the action of withdrawing aid to which one
has a right. So, even if withdrawal-allowings were all alike in virtue of
an agents having a right to the aid withdrawn, it is hard to see how
allowing in the withdrawal-allowing cases is similar enough to the fact
of allowing in the standard allowing cases for there to be a general
moral explanation in terms of allowing across the two sorts of
allowings.
One might reply by offering a moralized account of allowing harm
generally. One might suggest that the distinctive feature of allowing
harm is an agent with a right to the aid withheld or withdrawn. This
fact is an obviously morally salient feature that might figure in general
moral explanations, too.
This view would still have the aforementioned problem of the dissimilarities in the origins of rights over oneself and over external
resources. More importantly, this reply would have the defect of leaving aside a large part of the explanatory aims of a project like McMahans.
Part of the point of McMahans analysis is explaining why a right one
has over oneself justifies some allowings of harm but does not justify
some comparable doings of harm. For example, why a right to oneself
permits one to refuse to provide costly aid, but does not justify doing
harm to an innocent person to prevent equally costly harm to oneself,
is precisely what an account of doing and allowing is supposed to
explain. The moralized analysis of allowing harm just suggested would
leave this fact to be explained by something other than the facts of
doing and allowing.13
McMahans characterization of doing-allowings and withdrawalallowings, then, makes these behaviors too varied among themselves
13
Precisely this complaint can be made against Foots moralized account of doing
and allowing harm. See Foot, Killing and Letting Die.
296
timothy hall
and too different from the paradigm cases of doing or allowing harm to
permit a satisfying generality of moral explanation across all doings
and allowings. This problem is not just an idiosyncracy of his account,
however. As I shall now argue, there is a significant moral difference
between withdrawal behaviors and paradigm examples of doing and
allowing harm.
III
In this section, I will argue that a category of behaviors called the denial
of resources is morally distinct from paradigm doings or allowings of
harm. I argue that this moral difference is independent of whether or
not the denials are doings or allowings of harm.
Before going on, I need to make some phrases clear. A standard
doing of harm is an agent acting and creating a threat that harms someone. Examples include paradigmatic cases of harming others, such as
shooting, stabbing, beating, or poisoning. Standard allowings of harm
are cases in which there is some threat of harm to another person
uncaused by, and otherwise independent of, the agent. Though the
agent could act to eliminate, forestall, or otherwise protect against the
threat, the agent does not, and harm befalls the threatened person.
A paradigm example would be an agent capable of aiding standing by
while another person drowns. The third category is what I will call the
denial of resources. By a resource, I mean something apart from a person herself that offers the prospect of being useful in furthering or protecting her interests. Sometimes, other persons can be a resource,
though most often I will be considering cases of non-person resources.
If there is a person to whom there is a realistic prospect of some
resources being useful, and another person, an agent, who acts to make
the resource unavailable to the first person, the agent denies the
resource to the first person. Withdrawals of aid just are one way of
denying resources: others include consumption of resources, the imposition of barriers to resources, or destroying resources.
My first argument for the moral distinctness of denials of resources
begins with a comparison of matched examples. My method differs
from a common one in normative ethics. Typically, a writer interested
in arguing for the moral significance of some factor of interest will
compare behaviors in circumstances in which all other things are equal,
except for the factor of interest. If there is some all things considered
moral difference between the behaviors, it is presumed that the factor

of interest is making that moral difference.14
My method will be to compare pairs of examples. I will consider one
pair taken from each of three behavior types: denying resources, standard doing of harm, and standard allowing of harm. The pairs of examples will be closely alike, except for the differences in the categories to
which they belong. First, I will show that a right to a resource makes a
significant moral difference between the two denial behaviors. Then,
I will show that the same right does not make a significant moral difference between the standard doings of harm or the standard allowings of
harm. Because the right to a resource makes a moral difference in the
denial pair that it does not in the other pairs, I will conclude that there
is a moral difference between denials and each of the other behavior
types. The argument to come will illustrate the method.
I begin with two denial cases. The first is Joness Antidote. Note two
things about this example. First, it is permissible for Jones to take the
antidote from Smith. Jones would have been permitted to refuse to give
the antidote to Smith in the first place. Since Smiths possession lacks
Joness permission, Jones is permitted to take the antidote back.
Second, Smith is not permitted to use violence to resist Jones taking
back the antidote. Again, Jones is taking back what he has a right to
take, the antidote he would have been permitted to refuse Smith in the
first place. Smith cannot use violence to prevent Jones from taking
what he is entitled to take and what Smith has no right to keep.15
Consider a second example:
Smiths Antidote: Smith and Jones live in a remote area somewhat near
one another, but are otherwise unrelated. It happens, rarely, that poison gas is released from a volcano in the area. To guard against this
possibility, Jones and Smith each had on hand an antidote to the poison.
Before the danger, Jones, desiring some luxury item and discounting the
risk, sold his antidote to a third party. The volcano releases poison gas,
threatening them both. Jones, not wanting to die, takes Smiths antidote.
Smith dies.
14
See the qualifications in the use of this method discussed in Kagan, The Limits of
Morality, pp. 1115; Kagan, The Additive Fallacy, Ethics 99 (1988), pp. 531; Kamm,
Morality, Mortality, vol. II, pp. 1786; Peter Unger, Living High and Letting Die
(New York: Oxford University Press, 1996), pp. 1023.
15
At least, Smith is not permitted to use violence if he is aware of the provenance of
the antidote.
298
timothy hall
In this example, Jones acts impermissibly in taking the antidote from

Smith. He has no right to the antidote, and Smith has a right to the
antidote. Smith would have been permitted to use violence to resist
Jones, had he been able. He was not required to stand by and let Jones
take what he has a right to use to save his own life.
The difference between the two cases is obvious. In the first case,
Jones has a right to the resource and Smith does not. In the second
case, it is Smith who has the right.
By a right, I mean a moral, and not a legal, right. That the rights are
as I stipulate is plausible enough. The ordinary legal right conjoined
with the other details of the case will suffice for the rights to be as I
assert, in the minds of most readers. The most obvious factors that
might tell against the legal property right determining the moral right
are not present. Both Smith and Jones had ample, and equal, opportunity to acquire the antidote outright. Neither one was forced by extraordinary pressures, including unjust threats of v iolence, to relinquish his
property after acquiring it.16
The claim that a moral right of ownership determines the permission to withdraw the antidote, and the permissibility of violent resistance, does not risk vacuity. A moral right of ownership is not just an
all things considered permission to make exclusive use of a resource.
It is a stand-in for a specific set of reasons for that permission, reasons
provided by a theory of justice in distribution. The details of that theory will determine the details of the explanation of the permission
to withdraw. So it is possible for a right to a resource to provide an
informative explanation of the all things considered moral status of
withdrawing a resource.
Furthermore, in these examples, concern for the right to life is
replaced by, or subordinate to, concern for a right to the resource. In
Joness Antidote, for example, Joness right to the antidote and Smiths
lack of such a right are, in conjunction with the other circumstances,
sufficient to justify the action that results in Smiths death. There is no
further complaint one might make on Smiths behalf that Jones violates
Smiths right to life. In Smiths Antidote, by contrast, Joness behavior is
a serious wrong, and Smith would be permitted to respond to it as he
would a standard killing that violated his right to life.
16
Some might claim that the moral rights to the antidote cannot be satisfactorily
evaluated without a wider view of the biographies of these two men, or of the society
as a whole, than provided here. The reader is invited to fill out the story as required.
The subordination of the right to life is noteworthy because the

reasons that typically determine whether a person has a right to a
resource are very different than those that determine whether a being
has a right to life. Whether a person has a right to a resource is likely to
be determined by a legal or conventional system, which in turn is subject to principles of justice. These principles might be egalitarian principles, or principles guaranteeing a minimum welfare, or principles
that require that the state not interfere with voluntary transactions
among individuals, to provide some examples. Whether a being has a
right to life will be determined by such factors as whether the being has
a sophisticated psychology of a certain kind, including, perhaps, such
characteristics as being capable of shaping her actions in accordance
with some conception of the good.
That a right not to be harmed is replaced by, or subordinate to, a
right to a resource is a distinctive feature of denial behaviors generally.
The right not to be harmed does not do the moral work it does in a
standard doing or allowing of harm case.17
We can make this last observation more precise. When the denying
agent has the right to the resource, and the victim does not, the denial
does not infringe on the victims right not to be harmed. The denial is
in that respect on a moral par with a standard allowing of harm. When
the denying agent does not have the right, and the victim does, the
denying infringes the victims right not to be harmed. (Alternatively,
the violation of the right to a resource is on a par with an infringement
of the right not to be harmed.) The denial is thus on a moral par with a
standard doing of harm. Rights to resources make no comparable difference in standard doing or allowing harm cases.
To see all of this, we can contrast the denial behaviors with standard
doings of harm:
Drive, Joness Antidote: Jones is in a public park, and accidentally touches
a poisonous plant. Though this sort of poisoning is rare, Jones made
some provision for the possibility, and his antidote is not far away. To
reach the antidote and save his life, however, Jones will have to drive his
car over Smith, who is napping in the park. A sound sleeper, Smith cannot be awakened in time. Jones drives ahead, killing Smith.
17
Boorse and Sorensen call this effect the multiplier effect, as does Kamm in the
context of abortion. See Boorse and Sorensen, Ducking Harm; Frances Kamm,
Creation and Abortion (New York: Oxford University Press, 1992), pp. 2041.
300
timothy hall
Joness action in this case is wrong, most people would agree. Smith,
furthermore, would be permitted to use even lethal force to save his
own life, if he were able and if that were necessary for him to save his
life.
A second case:
Drive, Smiths Antidote: Jones is in a public park, and accidentally touches
a poisonous plant. Jones had made some provision for the possibility, but
sold his antidote some time ago. It happens that Smith is napping nearby,
and Smiths antidote to this poison is just a further, short distance away.
Because Smith is a sound sleeper, for Jones to get to Smiths antidote in
time, Jones would have to drive over Smith. Jones drives ahead, killing
Smith.
In this case too, most people would agree, Joness action is wrong and
Smith would be permitted to use even lethal force to save his life, were
he able and were it necessary.
In each of these standard doings of harm, Joness action is impermissible. The fact that Jones has a right to the antidote in one case and not
in the other does not make a difference in the permissibility of his
behavior. Yet these cases are comparable to the denial cases discussed
above. In both the denial and standard killing cases, Jones either acts
with the result that Smith dies or Jones himself dies. Joness action in all
cases is necessary for his using the antidote. In none of the cases need
Jones aim at the death of Smith to acquire the antidote.18
One might object that Drive, Joness Antidote is importantly dissimilar to Jones
Antidote. In the latter case, Smith is vulnerable because of his need for the antidote. Yet
Smith has no right or claim to the antidote. In Drive, however, Smith is vulnerable
because of his position in the park. Yet, Smith has a right to sleep in the park. So, in
Joness Antidote, Smith is denied only that to which he has no right, but he is killed in
Drive, Joness Antidote where he has a right to be.
We might imagine that there is a rule against sleeping in the park, however. Even
with Drive, Joness Antidote so modified, it is not permissible for Jones to kill Smith.
We might also offer more indirect considerations. Consider the following example:
18
Joness Ambulance: Jones is rushing to the hospital to save the lives of five, whom
he came across poisoned near his property. A person has innocently fallen unconscious across Joness private roadway, however. The roadway is the only path to the
hospital. So, should Jones stop, the five in the rear of the ambulance will die.
Should Jones proceed, the one lying unconscious in the road will be killed.
Most people will think here that Jones may not drive over Smith. Jones must stop, even
at the cost of the five lives in the rear of the ambulance.
If Jones may not drive over an innocent Smith, laid across Joness private roadway,
even to save the lives of five people, then surely Jones may not drive over Smith in
a comparable case in which one life is at stake, even if it is Joness own life at stake.
The right to the resource is not important in determining the moral

status of Joness killing of Smith. In these standard killing cases, Joness
behavior would be condemned on the ground that it violates Smiths
right to life. In these cases, Smiths right to life, and not the right to the
resource, dominates the analysis of Joness behavior.
There is a similar contrast between the denial cases and two standard
allowings of harm:
Shelter, Joness Antidote: A poison gas has been released from a volcanic
vent on the valley floor. Jones has found some natural shelter, where he
is safe from the poison. From his shelter, he can see Smith some distanceaway. Smith, weak from poison, is struggling to reach Joness antidote, which sits on the ground nearby. Jones can help Smith use his
antidote only by leaving his shelter. Smith and Jones differ in their sensitivity to the poison in the following way, however. If Jones leaves his shelter now, his exposure would be too serious for the antidote to save him.
He must wait for the poison to dissipate a bit before the antidote will be
effective, whereas Smith could use the antidote now. So, Jones does not
leave the shelter until the poison dissipates enough for it to be safe, and
Smith dies.
Most people would judge that Jones acts permissibly in this case. Most
people would agree that Smith would not be permitted to kill Jones, if
that were necessary to save his life, either. Imagine, for example, that
Smith and Jones were arranged such that Smiths shooting Jones would
cause Jones to fall out of his shelter, knocking the nearby antidote to
Smith. Smith would not be permitted to shoot.
Consider a second example:
Shelter, Smiths Antidote: A poison gas has been released from a volcanic
vent on the valley floor. Jones has found some natural shelter, where he is
safe from the poison. From his shelter, he can see Smith some distance
away. Smith, weak from poison, is struggling to reach his own antidote,which sits on the ground nearby. Jones can help Smith reach his
antidote only by leaving the shelter. Smith and Jones differ in their sensitivity to the poison in the following way, however. If Jones leaves his shelter now, his exposure would be too serious for the antidote to save him.
He must wait for the poison to dissipate a bit before the antidote will be
So, even if we assume Smith is innocently in a place he has no right to be in Drive, the
moral differences between Drive and Joness Antidote remain.
There may be an important qualification to be made here. If Smith knowingly places
himself in a position he has no right to be, many people will be sympathetic to the
claim that Jones may then drive over Smith to get to the life-saving antidote.
302
timothy hall
effective, whereas Smith could use the antidote now. So, Jones does not
leave the shelter until the poison dissipates enough for it to be safe, and
Smith dies.
Again, in this case, it is permissible for Jones to remain where he is,

most p
eople would agree. Again, we might imagine that Smiths shooting of Jones would cause both Jones and the antidote to tumble toward
Smith. Again, most people would agree that Smith would not be permitted to kill Jones, if that were necessary to save his own life.
These cases are comparable to the denial cases discussed above. In
each of these cases, Jones must allow Smith to die or he will die himself.
In each case, Jones determines whether Smith or Jones will be able to
use the resource in question.19 Yet there is no moral difference made by
the right to the resource between the standard allowing of harm cases.
As I claimed at the outset of this section, because a right to a resource
makes a moral difference in the denial cases, but not in the standard
doing or allowing cases, denials of resources are morally distinct from
standard doings or allowings.
A further clarification of the sense of moral distinction here is in
order. Typically, when one claims that doing harm is morally distinct
from allowing harm, one claims the difference between doing harm
and allowing harm sometimes makes an all things considered difference in permissibility. I am not c laiming that there ever are cases of
denials of resources that, other things equal, differ with respect to permissibility from comparable cases of both standard doings or standard
allowings of harm. Indeed, I claim that a denial of a resource will always
be on a par with either a standard doing or a standard allowing of
harm depending on the disposition of the right to the resource. What
I claim is that a denial of resources is permissible or not, or licenses
violent defense or not, for very different reasons than standard doings
19
There is a complication in Shelter, Smiths Antidote. It seems that Jones requires
that Smith be dead for the following reason. If Smith were alive, Jones would have no
right to use the antidote in preference to Smith. So, because Smiths being dead is a
necessary element in Joness plan, Jones must intend that Smith die. This intention
marks a difference between Shelter, Smiths Antidote and Smiths Antidote.
Even with the intention asserted by this objection, the argument for the moral
difference between denials and standard allowings of harm goes through. For the
intention that Smith die should, if anything, count against the permissibility of letting
Smith die. It is still permissible for Jones to let Smith die in Shelter, Smiths Antidote,
however. Jones need not surrender his own life to save Smiths. So, the difference in
permissibility between Smiths Antidote and Shelter, Smiths Antidote remains.
or allowings of harm.20 The reasons in denial cases have to do with the

distribution of resources. The reasons in standard doing or allowing
cases do not, or at least do not to nearly the same degree.
There is, nevertheless, a sense in which denials of resources occupy
an intermediate place between standard doings and standard allowings
of harm with respect to permissibility.21 Denying a resource is on a par
with a standard allowing of harm when the agent has a right to the
resource denied, and the victim does not have a right. Denying a
resource is on a par with a standard doing of harm when an agent does
not have a right to the resource and the victim has a right. In the ordinary course of things, it is easier for a person to have an exclusive right
to resource as Jones does in Joness Antidote than it is for a person to
have permission to undertake a standard killing, like a shooting or a
stabbing. Indeed, commonsense morality has it that most of us have
rights to some resources even if these resources are needed to save the
lives of others. We have permission to kill others, through behaviors
like stabbings or shootings, only under very unusual circumstances.
On the other hand, a standard letting die might be permissible even
if the agent has no right to a relevant resource other than herself.
(Consider the Shelter, Antidote examples.) A denial of resources in
which the agent had no right to the resource would be justified only
under unusual conditions, in the way a standard doing of harm is justified only under unusual conditions. So, in these respects, a denial of
resources is easier to justify than a standard doing of harm, but more
difficult to justify than a standard allowing of harm.
In the argument in this section, the denial of resources case I discussed were withdrawals of aid. However, nothing turned on the
examples being w
ithdrawals as opposed to other forms of denials of
resources,such as consumption or destruction of resources. Hence, the
considerations above can be generalized to other forms of the denial of
resources.
20
In some cases, the reasons in common between a denial case and a standard
doing or allowing might dominate. For example, an impermissible allowing of harm
might be impermissible because it is an easy rescue case. Denying resources in a comparable case might be impermissible, and largely for the same reason: aid would be low
cost to the benefactor but prevent great harm to a beneficiary.
21
The remarks below can be repeated for other elements of the moral status of a
behavior, such as whether the behavior licenses violent defense from a victim, whether
an agent is liable to compensate afterward, and the degree of wrongfulness or blame.
304
timothy hall
IV
In this section, I offer a second argument for the moral distinctness of

denials of resources. This second argument moves from two wellaccepted features of contemporary non-consequentialism to the
moral distinctness of denials of resources.
The first of these features is a commitment to personal rights. These
rights are the standard non-consequentialist rights against being
harmed by others, as well as permissions to refrain from making oneself or ones efforts available to others. I mean to consider among the
former group relatively uncontroversial rights such as the rights against
being seriously injured or killed. Though many non-consequentialists
take seriously other rights against non-injurious bodily invasion, subtle emotional harms, or disrespectful treatment, for simplicity I will
consider only rights against being seriously injured or killed.22 The permissions to refrain from making oneself or ones efforts available will
likewise include permissions to refrain from being seriously injured or
killed in the service of others.
Non-consequentialists hold, of course, that both the rights and
the permissions will sometimes stand against such things as maxi
mizing the good or realizing greater interpersonal equality. Nonconsequentialists typically believe, for example, that one may not maim
an innocent bystander, even to save ones own life. The maiming is
ruled out despite the fact that it would maximize utility, in that two
persons would live and no one would die if the maiming occurs, and
despite the fact that it would be less unequal for there to be one
maimedand one whole person than one whole person and one dead
person. Similarly, non-consequentialists hold that one is sometimes
not required to submit to maiming that would save the life of another
person, despite the prospect of maximization of utility or greater
equality.
These rights are usually taken to be prepolitical, as well. By that,
I mean that the rights do not depend on a political society for their
existence, and they persist even within a political society.23 These rights
22
For an extended discussion of emotional harms and any rights we may have
against the infliction of such. See Judith Thomson, The Realm of Rights (Cambridge,
MA: Harvard University Press, 1990), pp. 24971.
23
Insofar as I say that the personal rights are prepolitical, I share with natural rights
theorists the claim that people enter political society with rights the state may not deny.
can form the basis of complaints of injustice against a government,

even against a government otherwise justly constituted and conducted. For example, that an otherwise just government kidnapped
and maimed its citizens for the benefit of others would give rise to
powerful complaints of rights violations, according to most nonconsequentialists.
The second feature is the distribution of non-person resources. By
this, I mean that, within political society (at least), resources other than
persons themselves are subject to distribution. A distribution here
means the allotment of rights to the use of resources. Rights here is
meant to encompass a variety of forms of claims and permissions to
use resources, not just private property in the sense meant in a contemporary Western legal system. The specifics of the distribution will
depend on the specification of the principles of just distribution.
For most non-consequentialists the principles of justice in distribution are much more outcome-oriented than are principles informed by
the personal rights. For example, some non-consequentialists will
endorse a requirement of equality of access to welfare (or some component of welfare) derived from material resources. In considering
whether various possible outcomes meet the requirement, the calculations of interpersonal equality will not, typically, place any special
weight on how the equality is arrived at. That is, the fact that resources
would have to be appropriated from some and distributed to others, as
opposed to left in the possession of some, is not itself taken to have
moral significance. This will be true even if the outcomes of the denials
involve serious bad effects for some. It is worth considering an example
illustrating this point:
Two Groups of Potentially Blind People (Denial): An initial distribution
of non-person resources is taking place at the founding of a society.
The society has two small subgroups, one whose members have an inherited disposition to blindness in one eye, and a larger class disposed to
blindness in both eyes. Without treatment, nearly every member of each
group will develop the blindness to which he is disposed. Natural medicine is available to treat either sort of blindness. The medicine cures the
one-eye blindness, and weakens two-eye blindness to one-eye blindness.
I do not, however, necessarily assert the range of natural rights to property that philosophers like Locke or Nozick assert. See Robert Nozick, Anarchy, State, and Utopia
(New York: Basic Books, 1974), pp. 17482; John Locke, The Second Treatise of
Government, II:v:2551.
306
timothy hall
This medicine is the only way either sort of blindness can be treated.
The medicine is scarce, however, and can be employed to prevent blindness in the one group or the other, but not both.
Non-consequentialists inclined toward an egalitarian distribution of

resources would be willing to endorse the distribution that provides
the medicine to those susceptible to blindness in both eyes. The outcome that results in this case is less unequal than that which would
result otherwise. Non-consequentialists would be willing to endorse
this outcome despite the fact that denial of the medicine to those susceptible to blindness in one eye results in a very bad outcome for those
people.
Furthermore, the fact that the distributive principles tell in favor of
distributing the medicine to the completely unsighted is decisive in this
case. There is no further complaint to be made on behalf of the completely sighted that the denying of the medicine to them violates their
rights against being harmed.
By contrast, most non-consequentialists would not endorse violating a personal right even if so doing would realize equality. For
example:
Blinding Action (Standard Doing): An initial distribution of non-person
resources is taking place at the founding of a society. The society has
two small subclasses. One subclass has an inherited disposition to total
blindness. This subclass can prevent blindness to itself through preparing
a medicine from an otherwise useless plant, and this is the only way to
prevent the blindness. In preparing the medicine, however, they will
release fumes that will cause another, smaller, and otherwise healthy subclass of citizens to go blind in one eye.
Most non-consequentialists would forbid this way of realizing equality,as it would involve the violation of the rights of the healthy subclasswho would be made blind in one eye. Importantly, the fact that
the case involves a standard doing of harm, as opposed to a denial of
resources, is part of the explanation for the rights violation that occurs
in this case.
Similar points might be made in regard to an alternative, and widely
endorsed, standard of distributive justice: the assurance of a social
minimum. Should a standard doing of harm be necessary to realize a
minimal level of welfare, again as in a case like Blinding Action, that fact
would tell against realizing the minimum. No such constraints would
be recognized if resources had to be denied to some in order for a social
minimum to be realized.
So, the two features of non-consequentialism taken together providea strong case for a moral difference between standard doings and
allowings of harm on the one hand and denying resources on the other.
The fact that the personal rights tell against standard doings of harm
undertaken in service of equality, or some social minimum, but not
against the denial behaviors involved in a distribution of non-person
resources, indicates a moral difference between standard doings of
harm and denials of resources. Similarly, the fact that a person is permitted to refuse sacrifices imposed by standard doings of harm, and
thus allow another person to be harmed, even at the cost of equality or
some social minimum, indicates a moral difference between standard
allowings of harm and denials of resources.
V
I turn now to six objections to the two arguments just presented.
1. One might object that the second argument does not show that
denying resources is morally distinct from standard doings of harm,
despite differing assessments of cases like Two Groups of Potentially
Blind People and Blinding Action. In Two Groups of Potentially Blind
People, the better placed will appropriate the medicine, or the worse
placed will do so, or neither group will appropriate. If neither group
appropriates, then each group will suffer the blindness to which they
are susceptible. So, if either party appropriates, the other party will be
no worse off than if no one had appropriated. In these conditions, the
objection runs, it is permissible for the group whose denial would be
less bad than to deny the resource. It is permissible, the objection
claims, even if denials are morally comparable to standard doings of
harm.
To see the point, we might consider an example of a standard doing
of harm:
Falling People: Two people are falling toward one another, with a net
below them both. If each person does nothing, then they will collide and
both will be killed. Each person has a ray gun with him, with which he
might disintegrate the other person.24
24
Michael Otsuka, Killing the Innocent in Self-Defense, Philosophy and Public
Affairs 23 (1994), pp. 7494.
308
timothy hall
In this example, many people will agree that it would be permissible for
each person to try to disintegrate the other. Disintegration is plainly a
standard doing of harm, however. Since the only alternative to one person d
isintegrating the other is both people dying anyway, if one man
kills the other the dead man is no worse off than if neither had acted.
It is therefore permissible for each to try to disintegrate the other.
In response, we might note the moral differences between the appropriation of resources and a Falling People-type example. In Falling
People it seems as if the people still retain their rights against being
attacked, despite the permission of each to try to disintegrate the other.
It seems best to say that the person who is disintegrated has her right
infringed by the other person, rather than that she has no right at all.25
In the case of justified appropriation of resources, however, as in Two
Groups of Potentially Blind People, the people to whom resources are
not justly distributed are not naturally described as having their right
to the resource infringed by its denial. The principles of distribution
to which one appeals settle precisely the question of whether they have
a right.
2. One might also complain that the case made above by the second
argument is insufficiently general. I considered only two sorts of principles by which a non-consequentialist might decide the allotment
of non-person resources: equality, and the guarantee of a minimum
level of welfare. There have been other principles defended by other
authors, principles that are not so teleologically oriented toward welfare goals. For example, the principles of justice in distribution advocated by Robert Nozick are famously unconcerned with welfare
targets.26 Perhaps the case advanced above would not show that nonconsequentialists like Nozick are committed to denying resources
being morally distinct from standard doings of harm or standard
allowings of harm.
25
I understand by a justifiable infringement of a right a situation in which one
might permissibly proceed with a behavior that a right ordinary forbids, but in which
some moral residue of the right remains. In some cases, as in taking the stores someone
has left in an empty cabin in an emergency, the residue of the right might entail compensation be paid to the right-holder. In Falling People, the residue of the right might
tell in that respectperhaps the survivor should try to make good to some degree to
the surviving dependents of the other person. Perhaps the right tells in a more subtle
way, via agent-regret not appropriate in a denial of resources case.
26
Nozicks view can serve as a stand in for certain interpretations of Locke, as well.
Nozick, Anarchy, State, and Utopia, pp. 17482.
The case I offered above depends only on the underdetermination of

the distribution of non-person resources by the personal rights. That
the distribution of resources is an open question given the personal
rights is enough to show that there is a morally significant difference
between denials and standard doings and allowings. Even on Nozicks
view these conditions obtain.
On Nozicks view of justice in initial acquisition, resources may be
claimed from an unowned state, through some unspecified act of
appropriation, as long as the subsequent denial leaves no one worse off
than he is with the item in an unowned state.27 So, even on Nozicks
view, the constraints imposed by the personal rights underdetermine
the just allotment of resources. Furthermore, denying a person the
opportunity to improve his condition in a state of nature does not,
famously, tell against appropriation of resources on Nozicks view.
The analogous result from creating a threat against someone else
preventing his condition from improvingwould presumably be ruled
out by the personal rights, even on Nozicks view. If one were to injure
a person so that improvement in his defective eyesight which would
have otherwise occurred through natural bodily processes does not
occur, for example, then ones action would violate that persons rights.
So, even on Nozicks view, the personal rights rule out a standard doing
of harm, but not an otherwise comparable denying of resources.
3. One might also offer an objection that would bear on each of the
two arguments I have presented for the moral distinctness of denials of
resources. That objection claims that there are cases of the denial of
resources that are morally on a par with standard doings of harm, irrespective of the rights to the resources involved. One such case is a close
variant of one discussed by McMahan, for a somewhat different
purpose:
Pipe Sealer: A pipe carrying poison springs a leak, and Jones places a
patch of his own over the pipe, sealing it. A year goes by, with the patch
intact. Jones needs the patch for some other important but not vital purpose, and removes it from the pipe. Some people die when the poison is
released.28
27
Nozick himself uses appropriation to mean both the establishment of a moral
claim, and the denial of resources. I here use denial where it is appropriate. See Nozick,
Anarchy, State, and Utopia, pp. 17482.
28
McMahan offers a close variant of this example to point out the importance of aid
not being self-sustaining, as he calls it, if withdrawing it is to be an allowing of harm.
See McMahan, The Ethics of Killing, p. 381.
310
timothy hall
Many people will respond to this example by claiming that what Jones
did seems impermissible, and seems impermissible, furthermore,
despite his right to the resource involved, his patch. The prospective
victims of this behavior would have been able to use violence to prevent Jones from removing the patch, most people will agree. The case
thus seems like it is morally on a par with a standard doing of harm.
In response, one might deny that Jones has a right to the patch. Even
if Joness placing of the patch is consistent with his retaining legal ownership over it, legal ownership is not the last word on the moral rights
to the resources involved. One can argue that the vital importance of
the patch in place implies that others protected by it have a right that it
remain. One can argue that they come to have such a right through the
moral equivalent of eminent domain appropriation of legal property
a case made more plausible in this case by Joness initial provision of
the patch.
Evidence for this view can be found in considering a case in which it
was clearly understood by Jones and by the protected parties that Jones
retained a right to the patch. Suppose, for example, that Jones provided
the patch only on the condition that, should he need it for an emergency himself, he would be permitted to remove the patch at a moments
notice. In such a case, Jones removing the patch does not seem on a par
with a standard killing. It should not be thought that this non-parity is
owing to the agreement alone. It does not seem to be true that this contract is on a par with a contract that Jones be permitted to shoot other
parties if that is necessary for him. That latter contract would be much
more controversial among non-consequentialists, perhaps because of
the belief that the rights to life of the other parties are inalienable. There
will not be such a principled objection to other parties recognizing
Joness right to a patch, even when the patch provides vital assistance.
On the other hand, one might argue that, owing to the longstanding
presence of the patch, Jones creates a threat, rather than removing protection against one. Intuitions about this suggestion will vary. The metaphysics of threats is a notoriously difficult subject. A case like this
might be a hard case for a view on offer. The fact that there are hard
cases for the determination of a threat, however, is not itself an objection to the claim that the presence of an independent threat is morally
significant. That there is, in any event, a metaphysical distinction
between creating and not creating threats seems no less plausible than
the much-attacked claim that there is a distinction between doing and
allowing harm that animates the received view.
A related complaint is that there are cases that appear to fall under
my characterization of denials of resource, but seem (also) to be paradigmatic cases of doing harm. For example, if one removes a manhole
cover from the patch of road over which Smith will walk, one might
claim, one harms Smith when he falls down the hole. For purposes of
moral explanation, Smiths right against being harmed is in question
here, not his right to the use of the manhole.
I do not see it as an embarrassment, however, that denials of resources
are much more prevalent behaviors than one would expect from considering only the withdrawal of life support or other unusual behaviors. The strong intuition that removing a manhole is on a par with
striking someone with a club or otherwise causing some threat to harm
the person may be a function of the strength with which we think a
person is entitled to use resources like manholes in public areas like
streets to protect himself from an injurious fall. Note that our intuitions would be different if we were considering two shipwrecked men
who contend over one life preserver, and we were also sure that one of
the two men had a right to the life preserver. There, as the discussion in
section III made clear, we would not view the rightful owners denying
the life preserver as on a par with initiating a threat against the other
person. Were we to imagine, then, some extremely unusual situation in
which the manhole were owned by one person and not another, and in
which there was a conflict of interest like the life preserver example, it
is not at all clear that removing the manhole would be on a par with a
standard doing of harm.
4. One might also object that both arguments above do not address
denials of resources in which the resource denied is the agent herself.
For example, consider again Thomsons Violinist. In this case, the denial
seems permissible, and it seems permissible in large part because of
the right Jones has to his body. This is not a right that most nonconsequentialists will recognize as distributed, however, that is
bestowed in accord with principles of just distribution of resources
within a society. It is a right Jones has in virtue of being the person
occupying his body.
That is the same right to which it would seem appropriate to appeal
in cases of standard allowings of harm. If, for example, Jones were
called by the hospital and informed of the violinists need, rather than
being kidnapped, Jones would be permitted to refuse to connect himself, and would be so permitted because the right he has to his own
body entails a permission to refrain from making bodily sacrifices on
312
timothy hall
behalf of others. So, denials of resources in which the agent herself is

the resource denied are not morally distinct from standard allowings of
harm. Insofar as the right an agent has to her own body, and a subsequent right to refrain from sacrificing on behalf of others, is the operative factor in both denials of resources and in standard allowings of
harm, these behaviors are morally alike.
In response to this objection, I offer a concession. Denials of
resources in which an agent herself is the resource denied are not morally distinct from standard allowings of harm. They are on a par with
standard allowings of harm.29 Thus the distinction between denials and
each category of standard doings and standard allowings is not a neat
one; denials of non-person resources are morally distinct from standard doings of harm and standard allowings of harm, and denials of
person-resources are morally on a par with standard allowings of harm.
5. One might also object that the permissibility of denials of resources
is not always determined in largest part by the right a person has to the
resource withdrawn. One might consider an easy rescue version of a
denial case, a case in which the prospective withdrawer unquestionably
has a right to the resource withdrawn, but in which the cost to the prospective withdrawer of not withdrawing is low, and the benefit to someone else of making use of the resource is very high. We might imagine,
for example, a version of Joness Antidote in which only Smith is threatened, and the cost to Jones of letting Smith use his antidote is trivial.
Since Joness withdrawing the antidote would be impermissible because
it would be grossly uncharitable, the objection goes, the importance of
a right to a resource in determining permissibility does not distinguish
all denials from standard allowings of harm. Some denials are like
standard allowings of harm in that reasons of relative costs and benefits
to providing aid are the most important factors in determining
permissibility.
It should be emphasized, in response, that my view is that denials of
resources are on a par with otherwise comparable standard allowings
of harm when the denying agent has an exclusive right to the resource
denied. The parity claim holds even if a denial and a comparable allowing of harm are both impermissible. There is still a point to distinguishing allowings of harm from doings of harm even if two instances are
29
An exception to this concession should be made for cases in which one voluntarily provided aid initially and thereby prevented other forms of aid from being provided
to the beneficiary. In such cases, a beneficiary may have a right to continued aid.
impermissible. The impermissibility of each behavior might be owing

to different reasons. On my view, the reasons that make easy-rescue
allowings of harm impermissible will apply to denials of resources
when the agent has, and the victim does not have, a right to the resource
denied. If the victim has a right and the agent does not, the denial is
wrong for reasons applying to standard doings of harm. Furthermore,
what a prospective victim might permissibly do in response to either
an attempted impermissible doing or an attempted impermissible
allowing of harm varies. The victim might permissibly respond with
violence in the one case, and not the other.
6. Finally, one might object that the account of denials I have offered
has assumed a clear, exclusive right to a resource. One should not
expect there will always be such a right. There might be cases, one
might claim, in which there are no determinate rights to a resource. Or,
a right might be possessed by a third party, not by either the agent or
the victim in a denial case. For example, consider variants of Joness
Antidote in which neither Smith nor Jones has a preexisting right to the
antidote. In one case, we might imagine the antidote is simply unowned.
In another, we might imagine a third party, himself unthreatened, owns
the antidote. My account would, at least, be silent about such cases.
For my purposes, though, a right to a resource need not be determined by anything so formal as a legal or conventional system of property, itself governed by principles of justice in distribution. A right
might be determined by simpler, ready-to-hand factors. Consider a
variant of Joness Antidote in which neither Smith nor Jones has his own
supply of antidote. Exactly in between them, suppose, is one dose of an
unowned antidote. It seems to me there are at least three possible
responses to this situation:
1.The situation is first come, first served. Whoever possesses the antidote first has a right to it.
2.The situation calls for equal chances. Smith and Jones should flip a
coin to determine the right to the antidote.
3.There are no ways to establish a right. Should one man take possession, the other can justifiably wrest possession away, if he is able.
ossibility might be seen as an alternative characterization of
(This p
an equal right.)
Opinion will vary among these three possibilities, with the first two
being more commonly held. For my purposes, I need not choose
among them. My argument for the moral status of denials does not
314
timothy hall
require me to choose among these views of emergency-rights, so to

speak, anymore than it requires that I choose among principles regulating a more systematic distribution of resources within a society.30 It is
enough for me to note that deciding among these views is a distributive
matter. It is not a matter of the denying agent s having to provide reasons that overcome the moral presumption against standard killings.
Hence, the moral distinctness of denials is preserved even in cases in
which neither the agent nor the victim has a preexisting right to the
resource.
VI
Denials of resources are morally distinct from standard doings or
allowings of harm because moral rights to resources are much more
important in determining the moral status of denial cases than they are
in determining the moral status of standard doings or allowings of
harm. By contrast, the personal rights that are of great importance in
standard cases of doing or allowing harm are not very important in
denial of resources cases, or, if they are important, are subordinate to
rights to resources.
As I emphasized at the outset, the preceding arguments for the moral
distinctness of denial behaviors are consistent with denials being
doings or allowings of harm. Even if one accepts a metaphysical view
like McMahans, the fact that rights to resources play a much more
important role in denials than in standard doings or allowings of harm
remains. So, one would have to recognize moral differences between
denials and standard doings and allowings by recognizing moral differences among different kinds of doings and allowings of harm, on a
view like McMahans.
Indeed, the moral distinctness of denials of resources is consistent
with doing and allowing of harm being themselves intrinsically morally significant in denial cases. Consider a variant on McMahans view,
for example. On this variant, an agent who denies a resource to which
30
One might worry that taking an unowned resource from someone who already
possesses it is on a moral par with a standard doing of harm, not because of view (1),
first come, first served, but because the denial itself has a default status on a par with a
standard doing of harm. One reason to reject this view is precisely because different
assumptions as to rights to the resource will affect the permissibility of denial behaviors in ways that standard doings of harm are not affected.
he has a right and to which the victim has no right allows harm; an
agent who denies a resource to which he has no right and to which the
victim has a right does harm. Consider a denial-doing of harm. On this
view, the denial is a doing of harm because the agent lacks a right to the
resource denied and the victim has a right. On such a view, the fact that
the denial is a doing of harm might itself be a morally significant fact.
One can readily accept that a doing of harm that is determined as a
doing of harm by a lack of a right to a resource could itself be salient in
the explanation of the moral status of the behavior.
Nevertheless, as I argued in section II, there would not be a close
similarity between the sort of doing and allowing that occur in denial
cases and the sort that occur in standard cases. So, generality of moral
explanation across all doings and allowings in terms of doing and
allowing harm will be lost.
The preceding discussion has taken place at some remove from practical moral problems. Nevertheless, the distinctness of denying
resources bears in significant ways on practical moral questions. I will
give two examples.
The first is the withdrawal of life support in medical contexts. One
much-discussed issue is the agency involved in the withdrawal of
life support, in particular whether it is active or passive, doing or allowing. On the view I have laid out, at least some moral questions that
were thought to be questions of the significance of agency are questions
of just claims to property. For example, on my view, a doctor will be
permitted, absent reasons of charity, to remove life support when she
has a claim to this support. Establishing such a claim will involve
consideration of a range of factorssuch as her relationship to the hospital in which the equipment resides, her contractual obligations to the
patient, the role of financially interested third parties like insurers or
the statethat play a much more prominent role than would be true of
determining whether a standard act of killing violated a patients right
to life.
Among other things, this result makes the path clearer for the involuntary withdrawal of life support. It is easier, other things equal, to
show that a person in need lacks a claim on a resource because of more
pressing demands of others than it is to show that that person lacks his
right of life because of those demands.
Another issue of practical significance is abortion. On a prominent
line of argument, abortion is analogous to the withdrawal of aid. The
example Thomsons Violinist above is, of course, famously advanced to
316
timothy hall
argue that even if a fetus had a right to life many abortions would be
permissible.
The moral distinctness of denying resources casts doubt on any
direct inference from the permissibility of withdrawal behavior in an
example like Thomsons Violinist to the permissibility of initiating
threats, such as the mechanical or chemical attacks of most abortions.
While withdrawal behaviors can be justified by an appeal to the rights
the benefactor has to the resource withdrawnher bodyjustifying
the initiation of threats against the fetus requires overcoming the fetus
assumed right to life. The burden involved in the latter might be significantly greater than in the former.
Time-Relative Interests and Abortion

S. Matthew Liao
Time-Relative Interests and the Badness of Death
Many people believe that the badness of death is measured in terms of
its effect on the overall value of life as a whole.1 In particular, most people believe that other things being equal, it is worse for a twenty-yearold than for an eighty-year-old to die, because death would deprive the
twenty-year-old of a more valuable future than it would the eightyyear-old. However, this whole-lifetime view faces a puzzle: It would
also be worse for a late foetus to die than a twenty-year-old, since death
would deprive the late foetus of a more valuable future than it would
the twenty-year-old. Yet, most people believe that death is worse for a
twenty-year-old than for a late foetus.
The concept of a time-relative interest is introduced by Jeff McMahan
to solve this puzzle.2 To have time-relative interests is to be able to stand
in some psychological relations to ones future and past selves. The
strength of ones present time-relative interests depends on how
strongly one is psychologically connected to or invested in those
future and past selves. For example, an infant will typically have a
weaker time-relative interest in continuing to live than a grown adult,
since an infant has little or no awareness of his or her future self. Or, a
patient in the middle stage of Alzheimers disease will have weaker
time-relative interests than a normal ten-year-old child, because the
Alzheimers patient will be less psychologically connected to or invested
in her future and past selves than a normal ten-year-old child.
The concept of a time-relative interest can explain why death is
worse for a twenty-year-old than for a late foetus, because while the late
foetus has more valuable future than the twenty-year-old, it is also less
See, e.g., J. Rachels, The End of Life: Euthanasia and Morality (Oxford: Oxford University Press, 1986), p. 6; D. Marquis, Why Abortion is Immoral, Journal of Philosophy
86 (1989), pp. 183203, at p. 189; J. Broome, Weighing Lives (Oxford: Oxford University
Press, 2004).
2
J. McMahan, The Ethics of Killing: Problems at the Margins of Life (Oxford: Oxford
University Press, 2002), pp. 80, 17074, 18388.
1
318
s. matthew liao
psychologically invested in its future than the twenty-year-old. In other

words, on a time-relative interest understanding of the badness of
death, one should discount the late foetuss weak time-relative interests
from its valuable future measured from a whole-lifetime perspective.
Once one does this, one can explain why death can be worse for the
twenty-year-old than for the late foetus.
Because the concept of a time-relative interest seems to provide a
good explanation to the above puzzle, there is a reason to believe in its
validity. A number of writers believe that the concept of a timerelative interest can also explain why abortion is permissible.3 Indeed,
McMahan can be interpreted making this claim in his book, The Ethics
of Killing, and David DeGrazia has also put forward arguments along
this line in the case of early abortion. For example, DeGrazia says that
the proper basis for assessing the harm of the death to the presentient
fetus is its time-relative interest in remaining alive Since the vast
majority of abortions involve presentient fetuses, this application of the
[Time-Relative Interest Account]is enormously important.4
Or, McMahan writes regarding late abortion,

These two claimsthat the central moral objection to killing a developed
fetus is that this would frustrate its time-relative interest in continuing to
live and that this time-relative interest is comparatively weaktogether
imply that the killing of a developed fetus is substantially less seriously
objectionable than the killing of a person, and perhaps not seriouslyobjectionable at all, if other thingssuch as the effects on the biological parentsare equal.5
Indeed, McMahan speaks of the Time-Relative Interest Account as

providing a basis for an argument for the permissibility of abortion.6
I say that McMahan can be interpreted as making this claim,
because elsewhere in his writings, it is not as clear that he makes this
claim. In particular, McMahan holds the view that the Time-Relative
Interest Account of the wrongness of killing is just one component of a
more comprehensive Two-Tiered Account of killing, according to
McMahan, Ethics of Killing, pp. 267362; D. DeGrazia, Human Identity and
Bioethics (Cambridge: Cambridge University Press, 2005), pp. 27994; D. DeGrazia,
Identity, Killing, and the Boundaries of Our Existence, Philosophy and Public Affairs
31 (2003), pp. 41342.
4
DeGrazia, Human Identity and Bioethics, p. 288.
5
McMahan, Ethics of Killing, p. 339.
6
3
time-relative interests and abortion319
which the wrongness of killing beings who are above a certain threshold of respect does not vary according to their strengths of timerelative interests, other things being equal.7 McMahan proposes that to
be above this threshold of respect, one must have a certain minimum
level of psychological capacities such as autonomy, which presupposes
self-consciousness and some degree of rationality.8 But he does not say
whether having this minimum level of psychological capacities is
equivalent to having a certain minimum level or strength of timerelative interests. The assumption that having a minimum level of psychological capacities is equivalent to having a minimum level of timerelative interests implies that an indi-vidual with a minimum level of
psychological capacities will always have stronger time-relative interests than a being who has less than the minimum level of psychological
capacities. This means, for example, that an elderly human person who
has the minimum level of psychological capacities but who only has a
few minutes to live will still have stronger time-relative interests than a
normal healthy animal, e.g. a dog, who has less than the minimum
level of psychological capacities. Independent of our discussion, this
assumption and its implication seem to me to be correct because
although the dogs future life will contain vastly more good, the elderly
human person who will die in a few minutes still has a very strong
time-relative interest in avoiding death in a few minutes, that is, he has
a very strong time-relative interest in living for many more years. It is
true that his interest in continuing to live will be frustrated but it is still
an interest that he has and it seems that it would be stronger than any
interest the dog might have.
In any case, this assumption is important for our purpose because if
the two thresholdsthe minimum level of psychological capacities
and the minimum level of time-relative interestswere not equivalent,
then the Time-Relative Interest Account of the wrongness of killing
would not be providing the justification for the permissibility of abortion. Abortion would be permissible because embryos and foetuses
have less than the minimum level of psychological capacities, and not
on account of the strengths of their time-relative interests.
If the two thresholds were equivalent, then the Time-Relative Interest
of the wrongness of killing could provide a justification for the permissibility of abortion, because the embryos and foetuses having less than
7
8
320
s. matthew liao
the minimum level of psychological capacities is equivalent to their

having less than the minimum level of time-relative interests, and on
the Time-Relative Interest Account of the wrongness of killing, killing
embryos and foetuses via abortion is therefore less wrong given that
they have less than the minimum level of time-relative interests, other
things being equal.
Since we are interested in knowing whether the concept of a timerelative interest can be used to justify the permissibility of abortion,
since some writers such as DeGrazia have explicitly used this concept
to justify some forms of abortion, and since, as I have said, the assumption that the two thresholds are equivalent has independent plausibility, we should assume, for the purpose of this paper, that the two
thresholds are indeed equivalent, that is, having the minimum level of
psychological capacities is equivalent to having the minimum level of
time-relative interests.9 Given this assumption, a case for the TimeRelative Account of abortion can be presented as follows:
1.Killing is more wrong and less permissible when it is an offence
against persons. Call this the morality of respect for persons or
morality of respect, for short.
2.Outside of the morality of respect, killing is less wrong and may be
more permissible. That is, such an act may be weighed and traded
off in a manner approved by consequentialists10 and such an act is
not directly subject to or opposed by a deontological constraint.11
3.To be within the morality of respect, a being must have a certain
minimum level of time-relative interests.12
4.Embryos/foetuses have less than the minimum level of time-relative
interests.
5.Therefore, killing embryos/foetuses is not governed by the morality
of respect.
6.Therefore, killing embryos/foetuses via abortion is less wrong and
may be more permissible.
For this argument to be complete, it is important to define what
counts as having the minimum level of time-relative interests.
9
In personal correspondence, McMahan accepts that this assumption could be
consistent with his view.
10
11
12
McMahan, Ethics of Killing, pp. 26061.
McMahan suggests the following rule of thumb.13 While most people

believe that it is seriously wrong to kill persons like you and me, they
do not believe that it is as seriously wrong to kill some animals like
dogs. This suggests that these animals have less than the minimum
level of time-relative interests.14 If this is right, it follows that embryos/
foetuses also have less than the minimum level of time-relative interests. For, compared to a normal dog, a foetus and certainly an embryo
have much weaker time-relative interests. As McMahan describes a
foetus,
The developed fetus cannot envisage or contemplate its future and hence
cannot have future-directed psychological states, such as intentions; it
would, if it were to become a person, be unable to recall its life as a fetus;
and it now has no psychological architectureno beliefs, desires, or dispositions of characterto carry forward into the future. It is, in short,
psychologically cut off or severed or isolated from itself in the future. Its
future is, figuratively speaking, relevantly like someone elses future. It is
for this reason that despite the great good in prospect for it, the developed fetus has only a comparatively weak time-relative interest in continuing to live.15
And he writes that if the possession of certain psychological capacities is what relevantly distinguishes us from animals and so is the basis
of respect, the developed fetus must fall outside the scope of the morality of respect, for it clearly does not possess the capacities that distinguish persons from animals.16
Given this, if the killing of animals like dogs is not governed by the
morality of respect, the killing of embryos and foetuses should also not
be. If so, the Time-Relative Interest Account seems to entail that killing
embryos and foetuses via abortion is not as wrong and may be more
permissible.
In this paper, I first argue that if the Time-Relative Interest Account
permits abortion, then it would also permit infanticide. Given the
impermissibility of infanticide, this is a strong reason to reject the
Time-Relative Interest Account. I next consider and reject the proposal
that the Time-Relative Interest Account can at least explain the permissibility of early abortion, even if it may not be able to explain the

15
16
13
14
322
s. matthew liao
permissibility of late abortion. This does not mean there is no way to

justify the permissibility of abortion. My purpose here is only to show
that the Time-Relative Interest Account does not by itself entail the
permissibility of abortion.
The Time-Relative Interest Account and Infanticide
If the Time-Relative Interest Account permits abortion, then it would
also permit infanticide. The reason should be obvious. Newborn babies
also have weaker time-relative interests than normal dogs. Indeed, a
newborn baby hardly can contemplate its future and have futuredirected psychological states, such as intentions any more than a developed foetus approaching birth would be able to. Like a developed
foetus, if a newborn were to develop to maturity, it would be unable to
recall its life as a newborn. It is, similar to a developed foetus, psychologically cut off or severed or isolated from itself in the future. Hence,
on the Time-Relative Interest Account, a newborn also has less than
the minimum level of time-relative interests. If so, it should follow that
killing a newborn is also not governed by the morality of respect. If so,
infanticide should be permissible on this account, that is, infanticide
should not be directly subject to or opposed by a deontological
constraint.
McMahan acknowledges that this is an implication of the TimeRelative Interest Account, and he tries to defuse the objection in two
ways.17 First, he suggests that infanticide might not be so wrong. In this
regard, he points to historical precedents and cultural practices that
have permitted infanticide, including in the United States and Britain
where certain infants with Down syndrome have been allowed to die.
This strategy amounts to biting the bullet and is, I believe, not promising. Despite the cultural practices and historical precedents that
McMahan cites, including the practice regarding infants with Down
syndrome, the Time-Relative Interest Account seems to imply that all
forms of infanticidewhether of unhealthy infants or of healthy
infantsare permissible, that is, they are not subject to deontological
constraints. Yet, most people, as McMahan acknowledges, simply do
not believe that infanticide is generally permissible. I am going to
assume that they are right. This means that if the Time-Relative Interest
17
Account has this implication, then this is a serious problem for this
account.
McMahans second strategy is to argue that there are differences in
the strength of time-relative interests between a developed foetus and a
newborn that might account for the distinction between abortion and
infanticide. It might be worth noting that I do not think that McMahan
believes in this strategy. He is inclined, I think, towards the first strategy. As he says at one point, I believe, therefore, that common sense
beliefs about the morality of infanticide have to be revised. Because the
newborn differs only slightly in nature and status from the developed
foetus, there is no basis for a radical moral distinction between abortion and infanticide.18 Nevertheless, it is instructive to consider what
McMahan says here or what someone else who also holds this view
could say to see if the Time-Relative Interest Account can avoid this
implication.
McMahan offers three suggestions. First, he points out that a newborn infant is exposed at birth to a large amount of external stimuli,
which impels its mind to operate at a higher level, thereby accelerating
its psychological development.19 As a result, the infants time-relative
interest in continuing to live will be stronger than when it was a
foetus.
However, that a newborn infant is exposed at birth to a large amount
of external stimuli may just be a contingent fact. For, we can imagine
that after birth, an infant is placed and raised in a dark room in
which external stimuli are minimized. Infanticide would not be more
acceptable just because the newborn infant is now less stimulated
psychologically.
But let us suppose that being born inevitably causes the newborn
infants mind to be additionally stimulated. The issue here is whether
the newborn will achieve a level of time-relative interests that is higher
than that of normal animals such as dogs. It should be obvious that
immediately after birth, it is doubtful that a newborn infant would
achieve a level of time-relative interests comparable to that of a normal
animal, let alone a higher level of time-relative interests, even taken
into account the external stimuli that it may receive. If this is right, then
even granting that there may be a slight difference between the infant
and the foetus in their strength of time-relative interests, this would
18
19
324
s. matthew liao
hardly explain how the Time-Relative Interest Account need not entail
the permissibility of infanticide.
McMahans second suggestion is that at birth a newborn infant
begins actively to form special relations with the parents and others in
ways in which it is not possible for a foetus. According to McMahan,
the special relations may magnify the reason these people have not to
frustrate its time-relative interest in continuing to live.20
First, it is worth mentioning that developing foetuses can also
actively form some special relations with the parents and others. At late
stages of a pregnancy, a foetus may kick back when its foot is touched
and may respond to music and other kinds of sounds from parents and
others. Secondly, when someone is contemplating infanticide, it is
unlikely that they would try to form special bonds with the newborn
infant. If so, in such a case, it is unlikely that there would be much
stronger special relations to the infant after it was born than before it
was born. This means that while special relations may explain why
some people do not kill their newborns, for people who are thinking
about killing their newborn, it does not seem to provide any reason
why they should not do so. Finally, it is hard to see how special relations
at an early stage of infancy could drastically change a newborns timerelative interests. If it does not, then it remains the case that the newborn does not have significantly stronger time-relative interests than
the foetus. If so, the Time-Relative Interest Account still seems to entail
the permissibility of infanticide.
McMahans third suggestion is that the reasons favouring killing a
foetus are often stronger than in the case of infanticide. He points to
Judith Jarvis Thomsons argument that the foetus inside the pregnant
womans body is dependent on the continued use of her body for survival, which can be quite burdensome and invasive to the pregnant
woman.21 Newborns on the other hand exist independently and the
sacrifices they may require from others for their survival may be of a
fundamentally different kind and possibly less burdensome. According
to McMahan, this means that possible justifications for infanticide may
be more limited and substantially weaker than possible justifications
for abortion. If so, according to McMahan, this may be the main reason
why infanticide is generally more objectionable than abortion.
J.J. Thomson, A Defense of Abortion, Philosophy and Public Affairs 1 (1971),
pp. 4766, at pp. 4849; McMahan, Ethics of Killing, p. 344.
20
21
This strategy also does not work because we are not just interested in
any distinction between abortion and infanticide. Rather, we are interested in a distinction that would enable one to say that on the TimeRelative Interest Account, while abortion is permissible, infanticide is
not. While McMahans point above offers an explanation why infanticide may be generally more objectionable than abortion, it does not
change the fact that on the Time-Relative Interest Account, infanticide
is still permissible; that is, it is not subject to deontological constraints.
Again, the permissibility of infanticide on the Time-Relative Interest
Account derives from the fact that newborn babies have weaker timerelative interests than normal dogs. Given that it is permissible to kill
dogs because they have less than the minimum level of time-relative
interests, it should be even more permissible to kill infants given that
they have even weaker time-relative interests. The fact that there may
be additional reasons for favouring abortion does not change the fact
that from the perspective of time-relative interests, an early foetus and
an infant are intrinsically similar. So, if it is permissible from the perspective of time-relative interests to have an abortion, then from the
same perspective infanticide should also be permissible.
Hence, McMahan has not defused the charge that the Time-Relative
Interest Account entails that infanticide is permissible. Given the serious implausibility of this implication, one might also question its
implication about the permissibility of abortion.
The Time-Relative Interest Account and Early Abortion
At this point, some might draw a distinction between early and late
abortion, and some might argue that, irrespective of the Time-Relative
Interest Accounts implication for late abortion, it can at least explain
the permissibility of early abortion.22 One way to define early abortion
is as an abortion prior to the fetal brains acquiring the capacity to support consciousness, which occurs some time between the twentieth
and the twenty-eighth week of gestation.23 Since exactly when this
event takes place is indeterminate given our present understanding of
the nature of consciousness, McMahan proposes that we adopt a conservative estimate and stipulate that an early abortion takes place prior
See, e.g., DeGrazia, Identity, Killing, and the Boundaries of Our Existence.
22
23
326
s. matthew liao
to twenty weeks, or roughly about five months after conception.24

Accordingly, a late abortion would take place after the foetus has developed the capacity to generate consciousness.
It should be noted that McMahan himself does not try to show that
the Time-Relative Interest Account can explain the permissibility of
early abortion. The reason he believes that early abortion is permissible
is because he holds a certain view of identity, what he calls the Embodied
Mind Account of Identity, that says that we begin to exist only when
the capacity to generate consciousness is there.25 If the Embodied Mind
Account is correct, then an early abortion merely prevents someone
from existing.26 To be sure, as McMahan acknowledges, an early abortion does kill a developing human organism. However, McMahan
argues that if we are not identical to this organism, it is hard to see how
the organism can have a special moral status sufficient to make it seriously wrong to kill it.27 In contrast, late abortion does kill someone like
you and me. Let us call this the identity argument.
There are two problems with the identity argument. First, the
Embodied Mind Account is a controversial thesis. Even defenders of
liberal abortion policies such as Peter Singer and David DeGrazia do
not accept it.28 Many people accept instead what might be called the
Organism View of Identity, which says that we begin to exist when our
organisms begin to exist, which is sometime before our capacity for
consciousness begins to exist.29 If the Embodied Mind Account is false,
then McMahans identity argument already fails.
This said, it is worth pointing out that some people who reject
McMahans Embodied Mind Account may still employ a version of the
identity argument to defend early abortion. In particular, some people
believe that we begin to exist only when twinning is no longer possible.30 Since the possibility of twinning ends sometime after fertilization, it follows that early abortion, as defined by the period during
McMahan, Ethics of Killing, pp. 6694, 26769.
26
27
28
P. Singer, Rethinking Life and Death: The Collapse of Our Traditional Ethics (New
York: St. Martins, 1994); DeGrazia, Identity, Killing, and the Boundaries of Our
Existence.
29
A defence of this view can be found in E. Olson, The Human Animal: Personal
Identity without Psychology (New York: Oxford University Press, 1997).
30
See, e.g., Olson, The Human Animal; DeGrazia, Human Identity and Bioethics;
Singer, Rethinking Life and Death, p. 94.
24
25
which twinning is still possible, would be permissible, using the identity argument.
Elsewhere I have argued that the Embodied Mind Account is false,
that we should accept the Organism View, and that on the most plausible Organism View, we exist before the possibility of twinning ends,
namely, at the moment of fertilization.31 I shall not repeat those arguments here, and I shall simply say that on this particular Organism
View, an early abortion would still be killing one of us.
The second, more serious problem with the identity argument is that
if it were valid, it seems that this argument would be sufficient to establish the permissibility of early abortion. If so, it would render the TimeRelative Interest Account unnecessary. Indeed, as I said earlier,
McMahan himself does not defend early abortion using the TimeRelative Interest Account. This is a problem if the aim is to show that
the Time-Relative Interest Account, and not some other theory, can
explain the permissibility of early abortion.
Is there then a way to use the Time-Relative Interest Account to
demonstrate the permissibility of early abortion without appealing to
the identity argument and while avoiding the implication of infanticide? Here is a possibility: It might be pointed out that how exclusive or
inclusive the Time-Relative Interest Account is depends on where one
sets the minimum level of time-relative interests required to be within
the morality of respect. McMahan, as we have seen, sets it at the level at
which animals such as dogs would not be part of the morality of respect.
But one could lower the requirement in order to avoid the implication
that infanticide is permissible. Admittedly, once the minimum level is
lowered so that infanticide is excluded, this revised Time-Relative
Interest Account would no longer imply that late abortion is permissible, since the difference between a newborn and a late term foetus is, as
we have seen, not significant. However, it might nevertheless enable
one to argue that early abortion is permissible, for the following reason:
Even after drastically lowering the level required to be within the
morality of respect, early foetusesdefined as foetuses that have not
yet developed the capacity for consciousnesswill still not meet it,
even if they are one of us, so that an appeal to the identity argument is
not necessary. This is because early foetuses utterly lack any psychological connection with their later selves, and therefore utterly lack any
See S. Matthew Liao, The Organism View Defended, The Monist 89.3 (2006).
31
328
s. matthew liao
psychological capacity. Given this, on this revised Time-Relative

Interest Account, the killing of early foetuses would therefore not be
governed by the morality of respect. If so, the killing of early foetuses
via early abortion should be permissible. Indeed, David DeGrazia says
that the utter lack of psychological unity between the presentient foetus and later minded being it could become justifies a radical discounting of the harm of the foetuss death.32 However, it should be noted that
DeGrazia has not explicitly proposed that one should lower the minimum level of time-relative interests. Yet, this lowering seems necessary
if a Time-Relative Interest Account is to avoid implying that infanticide
is permissible, while it demonstrates that it can explain the permissibility of early abortion without relying on the identity argument.
So, is this proposal successful? I believe that it is not. For one thing,
once one lowers the minimum level of time-relative interests required
so that infanticide is excluded, this revised Time-Relative Interest
Account would imply that the killing of animals such as dogs should
also be governed by the morality of respect. After all, if infants are governed by the morality of respect in terms of their time-relative interests, then normal animals such as dogs, given that they have much
stronger time-relative interests than infants, should also be governed
by the morality of respect. Now some people might welcome this implication. But consider its consequences. Many people believe that once a
being meets the minimum requirement of the morality of respect for
persons, then that being has equal worth to all the other beings who
also meet this minimum requirement. Call this the Equal Worth View.33
On this view, even if other beings have comparatively stronger timerelative interests than this being does, all beings that meet the minimum requirement still have equal worth. So, for example, suppose
Bright, as a result of his intelligence, has comparatively stronger timerelative interests than Dull, who is constitutionally dim-witted.34 Since
both are above the threshold of the morality of respect, the Equal
Worth View implies that they both have equal worth, even though they
have different strengths of time-relative interests. One implication of
their having equal worth may be that other things being equal, e.g. if
32
DeGrazia, Human Identity and Bioethics, p. 288; DeGrazia, Identity, Killing, and
the Boundaries of Our Existence, p. 433.
33
McMahan discusses a specific version of the Equal Worth View, which he calls the
Equal Wrongness View (Ethics of Killing, pp. 23540).
34
The example of Bright and Dull comes from McMahan, Ethics of Killing, p. 234.
neither is culpable for the predicament he or she is in or if there are no

considerations of special relations, then killing either would be equally
wrong. Another implication may be that other things being equal, in a
situation in which one could save one or the other but not both, as both
have equal worth, one might for example have to toss a coin to decide
whom to save.35
If animals such as dogs are above the threshold of the morality of
respect, then the Equal Worth View implies that they have equal worth
like you and me. So, killing a dog would be just as wrong as killing, for
example, Bright. Or, if Bright and the dog are both drowning, as both
have equal worth, one might have to toss a coin to decide whom one
should save. These implications, however, seem counterintuitive.
Alternatively, one might reject the Equal Worth View and accept
instead the Differential Worth View which says that for two beings who
are above the threshold of morality of respect, the one with the comparatively stronger time-relative interests has the greater worth. The
Differential Worth View can avoid the implausible implications that
killing a dog is just as wrong as killing Bright and that one should toss
a coin when deciding whether to save Bright or a dog, because arguably, Bright has greater worth than the dog, since the former typically
has stronger time-relative interests. However, the Differential Worth
View now implies that Bright has greater moral worth than Dull, given
that Bright has stronger time-relative interests than Dull. Among other
things, this means that killing Dull is not as bad as killing Bright; or
when faced with a choice of saving Bright or Dull, this view implies
that one should save Bright instead of tossing a coin. Neither implication seems plausible.
Moreover, suppose the choice now is between a normal adult dog
and a newborn human infant. As we said earlier, normal adult dogs
have much stronger time-relative interests than newborn human
infants. On the Differential Worth View, it seems therefore that the dog
has greater worth than the newborn infant, other things being equal. If
so, killing the newborn would not be as bad as killing the dog; or if
confronted with the choice of saving the dog or the newborn, other
things being equal, it seems that one should save the dog instead of the
infant. Again, these implications seem counterintuitive.
35
330
s. matthew liao
Some might be tempted to resist this last example by arguing that

even though the newborn has weaker present time-relative interests
than the dog, the newborns total valuable future will be greater than
the dogs. The unstated conclusion here is that after taking into account
this factor, the newborn will have a greater worth than the dog, thereby
diffusing the coun-terintuitiveness of the Differential Worth View.36
First, whether the newborn will have greater valuable future than the
dog is a contingent matter. We can stipulate that the newborn has a
serious genetic disease and that it will die before its total valuable future
or the strength of its time-relative interests is greater than that of the
dog. Even so, it still seems absurd to think that killing the dog is worse
than killing this newborn. Secondly, even supposing that we are concerned with the typical case in which the newborn will have greater
valuable future than the dog, this argument involves a measure of confusion. The concept of a time-relative interest is a theory about how to
aggregate well-being over time.37 In particular, it takes an individuals
total valuable future and discounts them for the strength of the psychological connectedness to ones past and future selves. Since the total
valuable future has already been factored in and discounted, one cannot appeal to it again, as the above argument does. Otherwise, the
notion of a time-relative interest would become incoherent. Indeed,
proponents of the Time-Relative Interest Account of abortion were
able to argue that the late foetus is below the threshold of respect on the
ground that animals such as dogs are below the threshold of respect
and the late foetus has weaker time-relative interests than these animals. If one now claims that a newborn has greater worth than a dog,
even though the newborns time-relative interests are weaker than
those of the dog, then by the same logic, the late foetus should also have
greater worth than the dog, since like the newborn it also has greater
total valuable future than the dog. But if this is right, then it is no longer
clear that the late foetus is below the threshold of morality of respect, as
proponents of the Time-Relative Interest Account have claimed. If so,
the case for a Time-Relative Interest Account of abortion would falter.
Hence, while lowering the minimum level does enable a TimeRelative Interest Account to say that early abortion is permissible, the
consequences of the idea that killing animals like dogs are governed by
the morality of respect for persons seem unacceptable.
36
I passed over the possibility that the newborn and the dog may have equal worth,
since that possibility would still be counterintuitive.
37
See Broome, Weighing Lives, pp. 24951 for this point.
Let me suggest another argument against this proposal, which could

be controversial.38 While it is true that early foetuses lack time-relative
interests altogether, it is also the case that late foetuses really do not
have much of them either. Indeed, the degree to which a late foetus,
which has just (a few hours ago) acquired the capacity for generating
consciousness, can be psychologically invested in its future seems
hardly significant when compared to an early foetus. Given this, and
given, as has been granted earlier, that both are one of us, it seems
highly questionable that the early foetus utter lack of any time-relative
interest should matter so much such that the killing of it would be permissible while the killing of the late foetus would not be. Indeed, it
seems that the treatment of either should be the same. So, either it
should be permissible to kill both or it should not be. Since we have
argued that it is not permissible to kill late foetuses on the revised
Time-Relative Interest Account, it seems that it should also not be permissible to kill early foetuses on this account.
To develop this point further, consider an analogy. Suppose there is
a human being in a persistent vegetative state (PVS) and another in a
deep coma.39 In the case of the PVS patient, the higher cerebral brain
is completely damaged. This means that he utterly lacks the biological
structures necessary for having consciousness, which therefore means
that he utterly lacks any time-relative interest. In the case of the deep
coma patient, the reticular formation, which is needed for controlling
arousal in the cerebral hemispheres, is permanently damaged, but the
cerebrum is on the whole intact. This means that, arguably, the deep
coma patient still has some time-relative interests, albeit very weak
ones, since the physical substrate of the mind is preserved and potentially functional.40 Let us suppose that the orga-nisms of both continue
to function on their own without respirators. Moreover, let us suppose,
quite plausibly in my view, that both the PVS patient and the deep
coma patient are one of us.41 (This assumption is necessary because it is
necessary to ensure that the Time-Relative Interest Account is what
will be doing the work and not some version of the identity argument.
If this assumption is not granted, then this example is not analogous to
38
If someone finds this argument unpersuasive, there is still the first argument
above.
39
For a good discussion of these two cases along the lines I am suggesting, see
40
See also McMahan, Ethics of Killing, pp. 42650.
41
I defend this in Liao, The Organism View Defended.
332
s. matthew liao
the early/late foetus case I have been discussing.42 Also, given this
assumption, the not uncontroversial view that we die when our higher
brain, i.e. our cerebrum, dies would be ruled out.)43 The important
question for us here is that given that the PVS and the deep coma
patient are both one of us, but given that the PVS patient utterly lacks
any time-relative interest, while the deep coma patient has just a bit of
time-relative interests left, does this difference justify a very different
treatment of the two such that, for example, it would be permissible to
kill the PVS patient while it would not be permissible to kill the deep
coma patient?
My conjecture is that there should be no difference morally in the
way one treats the two cases. Indeed, some people believe that it would
be wrong to kill the deep coma patient, and that it might also be wrong
to allow the deep coma patient to die, unless the patient has previously
expressed an autonomous preference to be allowed to die in such circumstances.44 If they believe this, and if they believe that both the PVS
and the deep coma patient are one of us, then they should also believe
that it is wrong to kill the PVS patient, and that it might also be wrong
to allow the PVS patient to die, unless the patient has previously
expressed an autonomous preference to be allowed to die in such
circumstances.
Others such as Peter Singer and James Rachels have argued that the
fact that the PVS patient and the deep coma patient are alive does not
in itself tell us whether it is wrong to take their lives.45 Since they believe
that it is permissible to kill the PVS patient in certain circumstances,
they should also believe that it is permissible to kill a deep coma patient
in similar circumstances. In fact, Singer has argued that there is no
It should be noted that the difference between the PVS and the deep coma patient
may be somewhat greater than the difference between the early fetus and the late fetus.
This is because while both the PVS patient and the early fetus utterly lack any timerelative interests, the cerebrum of the deep coma patient is typically developed while
the cerebrum of the late fetus is typically not as developed. This suggests that the difference between the early and late fetus is even less than the difference between the PVS
and the deep coma patient.
43
J. Korein, Ontogenesis of the Brain in the Human Organism: Definitions of Life
and Death of the Human Being and Person, Advances in Bioethics 2 (1997), pp. 174.
Defenders of the higher-brain criterion include M.B. Green and D. Wikler, Brain
Death and Personal Identity, Philosophy and Public Affairs 9 (1980), pp. 10533; R.M.
Veatch, Whole-brain, Neocortical, and Higher Brain Related Concepts, in R. Zaner
(ed.), Death: Beyond Whole-Brain Criteria (Dordrecht: Kluwer, 1988), pp. 17186.
44
45
Singer, Rethinking Life and Death, p. 207; Rachels, The End of Life, pp. 4243.
42
principled reason why it would be wrong to take the lives of either,

given that neither has the capacity for consciousness.46 So, for Singer,
the treatment of either would be the same, even though one utterly
lacks any time-relative interest while the other one has some measure
of time-relative interests. Also, McMahan, who sees a metaphysical difference between the two cases in terms of time-relative interests, nevertheless says that Although deep coma is fundamentally different
from PVS in metaphysical terms, it may not be much different prudentially or morallypatients in a deep coma should be treated in much
the same way that we ought to treat patients in a PVS.47 This suggests
that the difference between the PVS and the deep coma patient in terms
of time-relative interests does not justify a very different treatment of
the two such that, for example, it would be permissible to kill the PVS
patient while it would not be permissible to kill the deep coma patient.
If I am right, given that this case parallels the early/late foetus case, it
seems that the difference between the early and late foetus in terms of
time-relative interests also should not justify a very different treatment
of the two. If so, and given that the revised Time-Relative Interest
Account does not permit late abortion, then it should also not permit
early abortion.
Conclusion
The concept of a time-relative interest is useful for solving certain puzzles about the badness of death. Some people believe that the concept
can also be used to show that abortion is permissible. In this paper,
I first argued that if abortion were permissible on the basis of the TimeRelative Interest Account, then infanticide would also be permissible
on the same basis. The implausibility of infanticide suggests that we
should also question the plausibility of the Time-Relative Interest
Account. I next considered and rejected the suggestion that the TimeRelative Interest Account can at least explain the permissibility of early
abortion, even if it has implausible implications when applied to late
abortion. If I am right, abortion, whether early or late, has to be justified on grounds other than time-relative interests.
Singer, Rethinking Life and Death, p. 105.

46
47
The Basis of Human Moral Status*

S. Matthew Liao
1. The Speciesism Challenge
Many people believe that all human beings have the same basic,
moral status, that is, that they are all rightholders. For example, this
belief is encapsulated in the Universal Declaration of Human Rights
(1948), which states that all human beings are born free and equal in
dignity and rights. A number of prominent philosophers have also
expressed this belief. For example, T.M.Scanlon says that the mere fact
that a being is of human born provides a strong reason for according
it the same status as other humans.1 Or, as the late Bernard Williams
said, there are certain respects in which creatures are treated in one
way rather than another simply because they belong to a certain category, the human species.2
However, the claim that all human beings are rightholders is in fact
surprisingly difficult to defend. When philosophers assess this claim,
they tend to find themselves either agreeing that not all human beings
are rightholders or adopting what Peter Singer and others have called a
speciesist position, where speciesism is defined as morally favoring a
particular speciesin this case, human beingsover others without
sufficient justification.3 Why is this?
* I would like to thank Jeff McMahan, Peter Singer, Julian Savulescu, James Griffin,
John Broome, Roger Crisp, David Archard, Nick Bunnin, the late Geoffrey Marshall,
Wibke Gruetjen, David Wasserman, Nick Shea, David DeGrazia, Joseph Shaw, Nick
Bostrom, Dan Robinson, Nic Southwood, Agnieszka Jaworska, Nathan Nobis, Michael
Walzer, Clifford Geertz, Chris Grau, Anders Sandberg, Steve Clarke, and audiences at
the Institute for Advanced Studies at Princeton, City University of Hong Kong,
Princeton University, University of Northern Florida, Joint Session of the Mind
Association and the Aristotelian Society (2004) at University of Kent, Georgetown
University Work in Progress Bioethics Seminar, City University of New York, University
of South Carolina, and the Society for Applied Philosophy Congress at St. Annes
College, Oxford, for their comments on earlier versions of this paper.
1
T. Scanlon, What We Owe to Each Other (Cambridge, MA: Belknap Press, 1998),
p. 185.
2
B. Williams, The Human Prejudice, in A. W. Moore (ed.), Philosophy as a
Humanistic Discipline (Princeton: Princeton University Press, 2008), pp. 13554 (142).
3
See e.g., P. Singer, Practical Ethics (Cambridge: Cambridge University Press, 2nd
edn, 1993). There is conceptual space for a position, according to which all human
336
s. matthew liao
Many philosophers believe that when we assess who can be a rightholder we must meet what might be called the Species Neutrality
Requirement.4 The Species Neutrality Requirement says that an adequate account of rightholding must provide some criterion for rightholding that in principle does not exclude any species and where the
criterion can be assessed through some objective, empirical method.
The reason for the latter condition is that there does not seem to be an
a priori way of knowing who the rightholders are. For example, if we
meet some alien beings and we want to know whether they are rightholders, it seems that we would not be able to know a priori whether
they are rightholders. If so, the only way to find out is to investigate
empirically what attributes they have and to consider if these attributes
are relevant enough to make them rightholders.
However, supposing that the Species Neutrality Requirement is correct, there does not seem to be a relevant empirical attribute that would
apply to all human beings. The most plausible attributes such as actual
sentience and actual agency do not apply to all human beings. For
example, some human beings such as anencephalic children and comatose persons lack actual sentience, and many human beings including
newborn infants lack actual agency. These human beings would not be
rightholders on these accounts.
Given this, those who wish to defend the claim that all human beings
are rightholders typically reject the Species Neutrality Requirement.
For example, some argue that all human beings are rightholders because
they have intrinsic worth or because they have dignity.5 Neither notion
is an attribute that one can empirically identify and assess. Others
assert that it is just self-evident that all human beings are rightholders,
and that it is not necessary to find out the particular attributes that
make them so.6 Although justification for any moral principle must end
beings as well as some other animals are rightholders. Such a position would not necessarily be speciesist. In fact, the account I shall develop may be an example of such a
position. As far as I am aware, such a position has not been advanced in the literature.
4
J. Feinberg and B. B. Levenbook, Abortion, in T. Regan (ed.), Matters of Life and
Death (New York: McGraw-Hill, 1993), pp. 195234.
5
G. Vlastos, Justice and Equality, in Jeremy Waldron (ed.), Theories of Rights
(Oxford: Oxford University Press, 1984), pp. 4176; R. Dworkin, Lifes Dominion (New
York: Alfred A.Knopf, 1993).
6
H. J. McCloskey, Respect for Human Moral Rights Versus Maximizing Good,
in R. G. Frey (ed.), Utility and Rights (Oxford: Basil Blackwell, 1985), pp. 12136;
R. Nozick, About Mammals and People, New York Times Book Review, 27 November,
II, pp. 2930; Williams, The Human Prejudice.
the basis of human moral status
337
at some point, if being human is a sufficient but not necessary condition for rightholding, it seems problematic that this approach is not
able to help us determine whether the beings with which we presently
live and the beings we might encounter in the future are rightholders.
If being human is a necessary condition for rightholding, it seems that
this approach would have the seemingly counterintuitive implication
that all non-human beings that we presently know as well as those we
might meet in the future cannot be rightholders.
Still others have argued that all human beings have rights, not in
virtue of the actual attributes they possess, but in virtue of belonging to
the kind of beings that typically have the relevant attributes for rightholding. For example, John Finnis says that to be a person is to belong
to a kind of being characterized by rational (self-conscious, intelligent)
nature.7 Scanlon says that the class of beings whom it is possible to
wrong will include at least all those beings who are of a kind that is
normally capable of judgment-sensitive attitudes.8 Indeed, Scanlon
believes that severely mentally handicapped human beings can be
wronged, even though they themselves do not and will not have the
capacity to understand or weigh justifications.9 For conveniences sake,
we can call this account the species norm account.10
The species norm account also fails the Species Neutrality Require
ment. To be sure, some members of a given species would have the
7
J. Finnis, A Philosophical Case against Euthanasia, in J. Keown (ed.), Euthanasia
Examined: Ethical, Clinical, and Legal Perspectives (Cambridge: Cambridge University
Press, 1995), p. 48. Readers will notice that Finnis is talking about persons, whereas
I have been discussing rightholders. In this paper, I understand the terms rightholding and personhood interchangeably. Some people who equate rightholding with
personhood also hold a particular conception of personhood, namely, all and only
those who are actual moral agents are persons. Consequently, they believe that all and
only those who are actual moral agents are rightholders. I do not hold this view. Others
equate personhood with actual moral agency, but not with rightholding. As many of
them also believe that some non-moral agents, e.g. animals, can also be rightholders,
these people argue that personhood is only a sufficient but not a necessary condition
for rightholding. Since my equating rightholding with personhood is a stipulation,
nothing substantive follows from it. If animals are rightholders, then, as I understand
these terms, they would be persons, even if they are not actual moral agents.
8
Scanlon, What We Owe to Each Other, p. 186.
9
Scanlon, What We Owe To Each Other, p. 185.
10
McMahan used the term species norm account to describe something different,namely, a view about the nature of fortune and misfortune. This term is useful,
though, for capturing the ideas advanced by writers such as Finnis and Scanlon. See
J. McMahan, The Ethics of Killing: Problems at the Margins of Life (Oxford: Oxford
University Press, 2002), pp. 14649, 209217.
338
s. matthew liao
required attribute for rightholding. But, as Scanlon admits, it can be the

case that other members would not have the required attribute. Also,
this account faces the following kind of objection advanced by Jeff
McMahan. McMahan describes a Superchimp who has rational capacities as a result of gene therapy, but is a chimpanzee.11 Since the
Superchimp does not belong to a kind that is characterized by having
rational capacities, it seems to follow, from the species norm account,
that the Superchimp is not a person. According to McMahan, this seems
absurd if the Superchimp indeed has rational capacities. Moreover, suppose Superchimps come to outnumber the normal chimpanzees, the
norm for the species would have changed and it would follow, on the
species norm account, that we would now need to treat the normal
chimpanzees as persons. As McMahan argues, this also seems absurd if
the normal chimpanzees do not have rational capacities at all.
In this paper, I shall assume that there is some merit to the belief
that all human beings are rightholders. I shall also assume that when
trying to justify this belief, one should try to meet the Species
Neutrality Requirement so as to not be speciesist. I shall further
assume that the methodology employed by adherents of the Species
Neutrality Requirement is valid, namely, it is possible to identify an
empirical attribute relevant for rightholding, and, pace Moore, it is
possible to derive moral claims such as rightholding claims from
empirical attributes.12 In light of these assumptions, I shall propose
an account that appears to meet the Species Neutrality Requirement
11
McMahan, The Ethics of Killing, pp. 14649. Some might question whether the
Superchimp would still be a chimpanzee if in fact it has rational capacities. This is a
difficult issue as it involves knowing the nature of being a chimpanzee. I do not know
of a good answer to this question at present. Hence, at least for now, I think we should
give McMahan the benefit of the doubt and allow for the disjunction between having
rational capacities and being a chimpanzee.
12
G.E. Moore, Principia Ethica (Cambridge: Cambridge University Press, 1959), has
cautioned that we should not try to derive an Ought from an Is. It is not possible to do
justice to this important question here, but arguably, empirical attributes in the world
play some role ingiving moral agents certain moral reasons for action (whether these
empirical attributes do sodirectly, as, e.g., when one derives an Ought directly from an
Is, or indirectly via some othermeans, seems to be another matter). For example, it
seems that an empirical phenomenon such as pain does play some role in giving moral
agents moral reasons to avoid causing it and toalleviate it if and when they can. If the
empirical phenomenon of pain did not exist, there would not be a moral reason to
alleviate it. If this is right, we can begin to understand howinternal, empirical attributes can play some role in determining what we ought to do. See, e.g., J. Griffin, Value
Judgement: Improving Our Ethical Beliefs (Oxford: Clarendon Press, 1996), for this
point.
339
and that at the same time appears to allow all human beings to be
rightholders. I call this the genetic basis for moral agency account of
rightholding.
Before we begin, it is worth noting that every plausible account of
human moral status comes with a certain theoretical baggage. That
is, each account will face difficult issues particular to it that it must
address. For example, on a sentience account of moral status, according to which beings that have the same amount of sentience have the
same moral status, it seems that one would have to address the issue
of how to treat a human being and a non-human animal, when both
have the same amount of sentience.13 Similarly, an account of moral
status, according to which all human beings are rightholders, will
imply that fetuses and embryos are rightholders since they are human
beings, and this will require discussions about issues such as abortion and embryonic stem cell research. Typically, plausible accounts
of moral status will have responses to these issues. For example, suppose one believes that all human beings are rightholders and that
abortion is permissible. One can follow Judith Jarvis Thomson and
hold the view that even if fetuses were rightholders, abortion would
still be permissible.14 It is beyond the scope of this paper to discuss all
the theoretical issues that can arise from the belief that all human
beings are rightholders. My aim here is therefore to set out clearly an
account that supports this belief, but not to defend this account
against all possible objections. Nevertheless, I shall explain how the
genetic basis for moral agency account is different from a potentiality
account and how it is preferable to an actual moral agency account of
human moral status.
2. The Genetic Basis for Moral Agency Account of Rightholding
My proposal is as follows: all human beings are rightholders because
they all have the genetic basis for moral agency; and it seems that having this genetic basis is sufficient for one to be a rightholder.
See, e.g., Singer, Practical Ethics.
J.J. Thomson, A Defense of Abortion, Philosophy and Public Affairs 1 (1971),
pp. 4766 (4849). In S. Matthew Liao, Rescuing Human Embryonic Stem Cell
Research: The Blastocyst Transfer Method, American Journal of Bioethics 5.6 (2005),
pp. 816, I argue that even if embryos were rightholders, there are still permissible
ways of pursuing embryonic stem cell research.
13
14
340
s. matthew liao
Following Carl Wellman and others, I take moral agency to be the

capacity to act in light of moral reasons.15 Moral agency can be contrasted with rational agency, which involves the capacity to know
something about causality such as if one does x, then y would happen,
and the capacity to bring about something intentionally. It can also be
contrasted with autonomous agency, which involves the capacity to
determine ones life course (autonomy) and the capacities to pursue
these courses (liberty). A moral agent need not act morally all the time
or at all.
The genetic basis for moral agency is the set of physical codes that
generate moral agency. In human beings, this set of codes is located in
their genome. We know this because a lot of complexity is needed as
the developmental basis for a complex adaptive phenotype like moral
agency, and the genome contains a significant proportion of this complexity. Also, the capacity for moral agency is grounded in psychological capacities such as rationality and empathy that uncontroversially
have a genetic basis. Indeed, rationality and empathy, two essential
components of the capacity for moral agency, develop in all normal
human beings according to a fairly predictable schedule.16 If the capacity for moral agency did not have a genetic basis, the development of its
essential components would not be so regular.
At present, we do not know exactly which set of genes is necessary
and sufficient for the genetic basis for moral agency (though rapid
advances in genomic technologies might mean that we could have this
knowledge sooner than we think). Also, it seems that some genes may
be necessary not only for the genetic basis for moral agency but also for
some other general capacities. But we can talk about a genetic basis for
moral agency as long as there are genes that definitely play no role in
forming the genetic basis for moral agency. For example, the genes for
my toenails or a gene whose expression serves only to produce pigment
in the eyes probably play no role in the formation of the genetic basis
for moral agency. Moreover, I shall shortly draw a distinction between
See, e.g., C. Wellman, Real Rights (Oxford: Oxford University Press, 1995), p. 113.
Piagets work in human cognitive development, for example, supports the idea
that rationality develops according to a fairly predictable schedule. For the claim that
the development of empathy in human beings also follows certain prescribed progression, see M.L. Hoffman, Moral Development, in M.H. Bornstein and M.E. Lamb
(eds.), Developmental Psychology: An Advanced Textbook (Hillsdale NJ: Erlbaum,
1988), pp. 497548; G.B. Martin and R.D. Clark, III, Distress Crying in Neonates:
Species and Peer Specificity, Developmental Psychology 18 (1982), pp. 39.
15
16
341
genes that make up an attribute and genes that undermine the development of an attribute, a distinction that could help us further to narrow
the set of genes that is necessary and sufficient for the genetic basis for
moral agency.
It should be mentioned that some people are hostile to the idea of a
genetic basis for human behavioral traits. A reason for this hostility is
due in part to the fact that, historically, there have been racist and sexist
attempts to show that human beings of a certain race or gender have
the genetic basis for higher intelligence than human beings of another
race or gender; and these attempts typically fail to consider seriously
the role that non-genetic factors play in the development of intelligence. Hence, by association, these people might also be hostile to the
idea of a genetic basis for moral agency.
However, the genetic basis for moral agency, as I understand it, gives
rise to a capacity and not a behavioral trait. This means that the fact
that human beings have the genetic basis for moral agency does not
mean that they will act morally. Also, nothing I have said precludes the
idea that non-genetic factors are also necessary for an adequate development of moral agency. Indeed, the idea of a genetic basis for moral
agency is compatible with the idea that much of the complexity needed
for the development of moral agency is located in the developmental
environment, a developmental resource that has been specifically
adapted to afford the development of moral agency (in a niche
construction-type way).
Here it is worth making two further points regarding the genetic
basis for moral agency. First, the idea of a genetic basis does not mean
that there is only one way this genetic basis can be sequenced or realized. In fact, this genetic basis could be multiply realizable. That is, it
could be the case that in certain environments, certain genes, A, B, C,
might be the genetic basis for moral agency for a particular being,
while in certain other environments, genes D,E,F would be the genetic
basis for moral agency in the same being.17
Secondly, to have the genetic basis for a certain attribute, the genes
that make up that attribute must be activated and be coordinating
with each other in an appropriate way. A being does not have the
genetic basis for a certain attribute if a being just possesses somewhere in its genome the genes that could make up the attribute, but
I thank David Wasserman for this point.
17
342
s. matthew liao
these genes are either not activated or are scrambled in such a way
that they do not coordinate with each other in an appropriate way.
To support the point about coordination, consider the following:
Suppose there is a book containing many random words, which if put
together in the right way, would result in a Shakespeare book. That
book would not be a Shakespeare book just because it contains the
correct words; those words must be organized in the right way.
How do we know that all human beings have the genetic basis for
moral agency? We know that all normal functioning human beings and
those at the beginning of life (e.g. infants, young children, fetuses) who
will develop normally have this genetic basis, since they exercise moral
agency or will exercise it; and after conception, the genetic codes of a
human being do not change very much, if at all. We also know that
most comatose human beings have this genetic basis, because they
have exercised moral agency. Most of the causes of anencephaly are
not genetic but are caused by environmental factors such as folic acid
deficiency.18 So we can assume that many of those with anencephaly
also have this genetic basis.19 Moreover, those with mild mental retardation, such as children with Down syndrome, typically exhibit some
moral agency. This suggests that they also have the genetic basis for
moral agency.20
Finally, to see how those severely defective human beings whose
conditions are the result of genetic defects rather than environmental
factors would also have this genetic basis, it is useful to distinguish
18
Other environmental factors include undiagnosed diabetes; hypervitaminosis A;
high temperatures of 102 degrees or higher for more than five hours; anticonvulsant
medication, especially valporic acid (valporate); or environmental/chemical exposure.
See, e.g., Center for Disease Control and Prevention Effectiveness in Disease and
Injury Prevention: Use of Folic Acid for Prevention of Spina Bifida and Other Neural
Tube Defects, 19831991, Morbidity and Mortality Weekly Report 40 (1991), 513516;
MRC Vitamin Study Research Group, Prevention of Neural Tube Defects: Results of
the Medical Research Council Vitamin Study, Lancet 338 (1991), pp. 13137.
19
At the moment, it is an open question whether anencephaly even has a genetic
cause. Scientists know that some genetic disorders such as Waardenburg syndrome
seem to correlate with a higher incidence of anencephaly. But there is no reason at the
moment to believe that anencephaly is the result of a defect of the genes that make up
moral agency.
20
About 34 percent of the mentally retarded population is severely retarded, and
only 12 percent of the mentally retarded population is classified as profoundly
retarded. See, e.g., American Psychiatric Association, Diagnostic and Statistical Manual
of Mental Disorders (Washington, DC: American Psychiatric Press, Inc., 4th edn,
1994); M.L. Batshaw and B.K. Shapiro, Mental Retardation, in M.L. Batshaw (ed.),
Children with Disabilities (Baltimore: Paul H. Brookes, 4th edn, 1997).
343
between genetic defects of the genes that make up an attribute and

genetic defects that undermine the development of an attribute. For
example, consider a human being born without a hand. This may be
because this human being lacks the genes to form the hand, or it may
be that certain conditions needed for the genes to form the hand,
e.g., prenatal nutrition, were blocked or lacking. In the former, this
human being would not have the genetic basis for having a hand, since
the human being lacks the genes that make up the hand. In the latter,
the human being would still have the genetic basis for having a hand,
because the genes that make up the hand are there and active, but they
were blocked from developing because of certain conditions.
On this distinction, the genetic defects that we are likely to encounter in these severely defective human beings are not defects in the
genetic basis for moral agency but at best defects that undermine the
development for moral agency. For example, consider Phenylketonuria
(PKU), Tay-Sachs, Sandhoff Disease and a whole cluster of about 7000
other kinds of genetic disorders, which are caused by the mutation of a
gene.21 The gene is typically necessary for producing a certain protein or
enzyme, which is then needed to change certain chemicals to other
chemicals or to carry substances from one place to another. Mental
retardation and other defects are typically caused by abnormal buildups of certain amino acids that become toxic to the brain and other
tissues, because the cell is unable to process these amino acids owing to
the mutation. But with treatment of a low enzyme diet as soon as possible in the neonatal age, normal growth and cognitive development
can be expected in many cases. For our purpose, this shows that the
brain tissue has initially developed normally and would have continued
to do so except for the abnormal build-up of the amino acids. Therefore,
following the distinction between genetic defects of the genes that make
up an attribute and genetic defects that undermine the development of
the attribute, single gene defects seem to be cases of the latter rather
than the former. Given this, one can say that human beings who have
these kinds of genetic defects most likely have the genetic basis for
moral agency.22
21
V.A. McKusick, Mendelian Inheritance in Man. Catalogs of Human Genes and
Genetic Disorders (Baltimore: Johns Hopkins University Press, 11th edn, 1994).
22
There are other kinds of genetic defects that involve more than a single gene. As
far as I know, either those defects are so severe that the fetuses typically die before
birth, or individuals with these defects typically only have mild mental retardation.
344
s. matthew liao
It might be necessary to concede that there is a theoretical possibilitythat ahuman being could lack the genetic basis for moral agency,
even if all present cases are not ones in which human beings lack this
basis.23 In particular, owingto advances in genetic engineering, someone might be able to create such an individual artificially. Of course, it
could be questioned whether such an individual would be a human
being at all. This raises the difficult issue of what makes a being human,
and whether the genetic basis for moral agency is an essential
human property. At present, I do not know of a good answer to this
question. Consequently, it should be admitted at least for now that it is
theoretically possible that a human being could lack the genetic basis for
moral agency. Still, we can conclude that for practical purposes, virtually
all living human beings we are likely to encounter will have the genetic
basis for moral agency.
The claim that the genetic basis for moral agency is sufficient for
rightholding is attractive for several reasons. First, as this account practically supports the widely held intuition that all human beings are
rightholders, for many, this would be a reason in favor of it.
Secondly, the genetic basis for moral agency is an identifiable, actual,
physical attribute. This means that this account meets one of the primary conditions of the Species Neutrality Requirement. It also means
that this account avoids speciesism. Indeed, if we were to learn that
chimpanzees or some other animals have the genetic basis for moral
agency, then they would be rightholders.
Thirdly, this account captures what is intuitively appealing (at least
to some) about the species norm account, namely, it too is motivated
by the thought that the kind of being that is typically characterized by
moral agency should be a rightholder. However, the genetic basis for
the moral agency account offers a more adequate interpretation of this
claim, because on this account, one actually possesses an identifiable,
actual, physical attribute and is not just a member of a group that possesses this attribute. In fact, this account can handle the kind of objection posed by McMahan against the species norm account. If a
Superchimp has a rational nature as a result of gene therapy, this
account can accommodate this case by locating this nature in the
See, e.g., American College of Obstetricians and Gynecologists, Prenatal Diagnosis of

Fetal Chromosomal Abnormalities, ACOG Practice Bulletin 27 (May 2001).
23
I thank Jeff McMahan for this point.
345
genetic makeup of the Superchimp. Or, if Superchimps come to outnumber the normal chimpanzees, this account can explain how the
two groups can be treated differently, if, for example, the Superchimps
have the genetic basis for rational nature while the normal chimpanzees do not.
Finally, since the genetic basis for moral agency is only a sufficient
condition for rightholding, it avoids the intuitive cost of denying the
status of rightholding to those non-human animals or other beings
who may plausibly qualify as rightholders but who may not have the
genetic basis for moral agency. For example, even if a dog does not have
the genetic basis for moral agency, the dog could still be a rightholder
on other grounds. Or, suppose it were possible to genetically engineer
a being to lack just one gene for the genetic basis for moral agency and
still to be human; although this human being would not have the
genetic basis for moral agency, this human being could still be a rightholder on other grounds.
Here it is worth mentioning that it is possible that some beings,
e.g. some alien being or some super artificial intelligent being, could be
made up of non-genetic, that is, non-carbon-based, isomorphic material, and still possess something functionally similar to the genetic
basis for moral agency. In my view, they would also be rightholders,
given that they have the physical basis for the development of moral
agency even though they do not have the genetic basis for moral agency.
Hence, a more precise name for this account should be the physical
basis for the development of moral agency account of rightholding.
However, since most of the living beings we know are genetic carbonbased life forms, to keep things simple, I shall continue to refer to this
account as the genetic basis for moral agency account.
The intuitive appeal of the genetic basis for moral agency account of
rightholding should be easy enough to grasp. Some people might of
course not share these intuitions and they might demand that one produces an independent argument as to why one should believe this
account. This demand may not be entirely fair though, since, as far as
I am aware, those who advance other criteria for rightholding typically
offer no non-circular, independent arguments for why their preferred
criterion is relevant for rightholding. To give just one example, suppose
one holds the view that if X has actual sentience, then X is a rightholder. It might be asked, why is this so? Asserting that pain is bad
does not seem to be providing an independent argument for this
account. Arguably, this is just a circular way of restating that actual
346
s. matthew liao
sentience is relevant. In any case, since my aim here to set out clearly
the genetic basis for the moral agency account and not to argue that it
is the correct account of moral status, I shall leave this matter here.
What I shall try to do in the remaining sections of the paper is to consider some possible objections to this account and to respond to these
objections on its behalf.
3. Just a Potentiality Account?
To start, some might wonder if the genetic basis for moral agency
account is just a potentiality account in disguise. That is, it might be
thought that having the genetic basis for moral agency is just having
the potential for moral agency. As such, so the argument goes, the
genetic basis for the moral agency account contributes nothing new to
the debate and inherits all the problems that have been attributed to the
potentiality account.
First, even if the genetic basis for the moral agency account were just
a potentiality account, this may not be a problem, because the arguments against the potentiality account are not conclusive. I do not want
to defend the potentiality account here, but let me present two oftrehearsed arguments against the potential account. One goes as
follows:
1. According to the potentiality account, if X is a potential F, then X
has the same rights and interests as an actual F, where F could be a
rightholder, human being, person, and so on.
2.However, potential Fs typically do not have the same rights and
interests as actual Fs. For example, a potential president does not
have the same rights and privileges as an actual president.
3.Therefore, the potentiality account is mistaken.
Clearly, proponents of the potentiality account would deny premise
1. No such proponent would hold the view that a potential rightholder has the same rights and interests as an actual rightholder.
A more plausible interpretation of the potentiality account is something like If X has the potential forV, where V denotes attributes such
as moral agency, sentience, and so on, then X is an F, where F could be
a rightholder, human being, person, and so on. For example, A.I.
Melden has proposed that if X has the potential for moral agency, then
347
X is a rightholder.24 If so, a proponent of the potentiality account need

not be making the kind of mistake that this argument attributes to her.
The other oft-rehearsed argument against the potentiality account is
that if the potentiality account were correct, then sperm and ova would
be rightholders, because on some notion of potential, they would be
regarded as having the potential to be rightholders.25
However, the term potential has different meanings. It can mean
possible. For example, flour, water, eggs and baking powder have the
potential to become a cake, which means that it is possible for these
ingredients to become a cake.26 Potential may sometimes also mean
probable. For example, one may say that the Lakers have the potential
to win the NBA Championship. By this, one may mean that there is
some probability that the Lakers will win the NBA Championship.
Finally, potential may mean that an entity, which has a certain nature,
has an inherent capacity to realize its particular nature. For example,
one may say that an acorn has a potential to become an oak. This may
mean that the acorn has the inherent capacity to realize its nature of
being an oak.27
On the most plausible reading of potential, namely, in the third
sense, sperm and ova do not have the inherent capacity to realize the
nature of being an agent. At best, they have the inherent capacity to
realize their nature of being functioning sperm and ova. Indeed, as Joel
Feinberg has pointed out, critics who continue to insist that the potentiality of the sperm and ova is identical to the potentiality of the zygote
A.I. Melden, Do Infants Have Moral Rights, in William Aiken and Hugh
LaFollette (eds.), Whose Child? (Totowa, NJ: Littlefied, Adams & Co., 1980), pp. 199
220 (21011).
25
See, e.g., J. Harris, The Value of Life (London: Routledge & Kegan Paul, 1985), pp.
1112; L.W. Sumner, Abortion, in D. Van DeVeer and T. Regan (eds.), Health Care
Ethics: An Introduction (Philadelphia: Temple University Press, 1987), pp. 16283
(174); M. Warnock, Do Human Cells Have Rights? Bioethics 1.1 (1987), pp. 114 (12).
26
See, e.g., J. Glover, Causing Death and Saving Lives (Harmondsworth: Penguin,
1977).
27
See, e.g., M. Reichlin, The Argument from Potential: A Reappraisal, Bioethics
11.1 (1997), pp. 123; S. Buckle, Arguing from Potential, Bioethics 2 (1988), pp. 227
53; R. Hursthouse, Beginning Lives (London: Blackwell, 1987); K. Young, The Zygote,
the Embryo, and Personhood: An Attempt at Conceptual Clarification, Ethics &
Medicine 10.1 (1994), pp. 27 (4); F. Wade, Potentiality in the Abortion Discussion,
Review of Metaphysics 29.2 (1975), pp. 21655 (245); G. Gillett, Women and Children
First, in K. Fulford, G. Gillett and J. Soskice (eds.), Medicine and Moral Reasoning
(Cambridge: Cambridge University Press, 1994), pp. 13143 (132).
24
348
s. matthew liao
are vulnerable to a reductio ad absurdum, namely, [a]t the end of that

road is the proposition that everything is potentially everything else,
and thus the destruction of all utility in the concept of potentiality. It is
better to hold this particular line at the zygote.28 Hence, the argument
that sperm and ova could be rightholders is also not a problem for the
most plausible potentiality account.
In any case, having the genetic basis for the development of a
particular attribute is not the same thing as having the potential for
a particular attribute. Consider a human being, Joe, with hands.
We would say that Joe actually has hands and that Joe had the potential to have hands. But suppose Joes hands were accidentally sawed off
in a wood factory. We would say that Joe no longer actually has hands
and that Joe does not have the potential to have hands. However, we
can still say that Joe has the genetic basis for the development of
hands, because Joe still has the genes for the development of hands.
Indeed, if Joe has a child, the child will most likely develop hands.
Consider another example. Friends of the potentiality account typically would accept that anencephalic infants do not have, and individuals in irreversible coma no longer have, the potential for moral
agency.29 As I have argued earlier though, anencephalic infants and
individuals in irreversible coma both still have the genetic basis for
moral agency. If so, having the genetic basis for the development of a
particular attribute such as moral agency is not the same thing as having the potential for a particular attribute such as moral agency.
4. Only Actual Agency Matters
As another possible objection, other people might argue that surely
moral agency matters only if one can actually exercise it. On this view,
the possession of the genetic basis for moral agency does not matter.
What matters is that one actually has the capacity to act in light of
moral reasons. Indeed, some people will say that the value of the genetic
basis for moral agency is entirely derived from the value of actual moral
agency.
Feinberg and Levenbook, Abortion.
See, e.g. Sumner, Abortion, p. 173; M. Freeden, Rights (Milton Keynes: Open
University Press, 1991), p. 59, who argues that the potentiality account has this implication. See also D. Marquis, Why Abortion is Immoral, Journal of Philosophy 86
(1989), pp. 183203.
28
29
349
But actual moral agency cannot be the sole ground for rightholding.
The reason is that if rightholding served any function at all, one would
be the following:
If and when the rightholders interest is in conflict with the same kind of
interest, that is, with the comparable interest, of a non-rightholder, the
rightholders interest should prevail.
A corollary of this is that if one were to give the interest of a nonrightholder priority over the comparable interest of a rightholder, then
one would be acting wrongly.
For example, normal adult human beings are typically regarded as
rightholders, whereas normal adult turtles are not. (Those who
believe that all animals have rights may be unhappy with this
example. If so, I suggest that they substitute the turtle with whatever
they would regard as a non-rightholder. The general point would
remain valid.) Suppose this is correct, and suppose rightholding has
the function I suggested, this means that if and when the interest of a
normal adult human being conflicts with the comparable interest of
a normal adult turtle, one ought to give the interest of the human
being more weight. Hence, if a normal adult turtle and a normal adult
human being both require rescue (suppose both are crossing the
street and are in danger of being hit by oncoming traffic), and suppose that one can only save one of them, then one ought to save the
human being, because the turtle is not a rightholder while the human
being is. If one did not do this, then one would be acting wrongly.
This does not mean that one does not have any duties at all regarding
non-rightholders. For example, if a turtle requires rescue and it
would cost one little effort to save it, it seems that one would have a
duty to save the turtle. It also does not mean that any interest of a
rightholder would be more important than any interest of a nonrightholder. If the turtle requires rescue and it would cost one little
effort to save it, and suppose that I, a rightholder, am having a cup of
tea at the moment, it seems that I may have a duty to save the turtle
even if my tea may become cold.
If actual moral agency is the sole ground for rightholding, though,
adult human beings would be rightholders while human infants would
not be, as the latter do not have actual moral agency. If rightholding has
the function I suggested, it would mean that an adult human beings
interest should be given priority over the comparable interest of an
infant. And, if one were to give an infants interest priority over the
350
s. matthew liao
comparable interest of an adult human being, then one would be acting

wrongly. Yet it is often permissible to give the interests of infants priority over the comparable interest of adult human beings. For example,
suppose a human baby and a human adult who has only a short period
to live are drowning, and one can only save one of them. It seems permissible to save the baby rather than the adult. Or, in a high sea when
a ship is sinking, it seems permissible to give infants the priority to be
in the lifeboat instead of the adults.30 These examples suggest that either
we are wrong to think that it is sometimes permissible to give infants
preference over adults, or actual moral agency is not the sole basis for
rightholding. Our intuition is that it is at least permissible and not
morally wrong to give infants preference over adults. This suggests that
actual moral agency cannot be the sole ground for rightholding.
Some might say that the human baby versus human adult case is
not a case in which comparable interests are at stake. In particular, it
might be said that from the perspective of a whole lifetime, the baby
has more at stake, because all things being equal, it has more years of
good life to lose.31 However, suppose the choice was between the same
adult (a rightholder with only a short time to live) and a long-living
turtle (a non-rightholder, which would lose many years of good life as
a result of death). It would not be permissible to save the turtle just
because it has more years to lose. So the fact that the baby may have
more years of good life to lose cannot be the reason why it is permissible and not morally wrong to save the baby.
But even supposing that the baby and the human adult both have
the same number of years to live, it still seems permissible to save the
baby at least sometimes. For example, it seems permissible for a mother
to save her baby instead of a stranger. One might add that this is so
not just because it is her baby. Indeed, suppose the mother was instead
faced with the choice of saving her turtle or a stranger, both having
comparable interests at stake. Even though it is her turtle, it would
be wrong for her to save the turtle, given that the turtle is not a
rightholder.
30
Some might suggest that perhaps infants are given preference because they are
smaller. However, this cannot be the explanation, since if someone were to have a small
turtle, and assuming that turtles are not rightholders, we would not give the turtle
preference over an adult male just because it is smaller.
31
See, e.g., J. Rachels, The End of Life: Euthanasia and Morality (Oxford: Oxford
University Press, 1986), p. 6; Marquis, Why Abortion is Immoral, p. 189.
351
Others might argue that the problem is that the status of rightholding does not have the function I discussed above. They might point out
that sometimes non-rightholders interests are given priority over the
comparable interest of arightholder. For example, suppose to rescue
an adult human being who onlyhas a short period to live, one must
blow up a gigantic planet that contains many wonderful life forms,
none of which are rightholders. One might decide against blowing up
the planet. If so, this seems as if non-rightholders interests can sometimes be given priority over the comparable interest of a rightholder.
However, the argument that the status of rightholding has the function I discussed above need not claim that the interest of a rightholder
can never be outweighed by any amount of non-rightholders interests.
In fact, the rule implies that the comparison should be one to one, that
is, one rightholders interest versus a comparable interest of a nonrightholder; not one to a larger number. On a one-to-one basis, I cannot think of an example where it would be permissible to give preference
to a non-rightholders interest over a comparable interest of a rightholder. At the same time, there are many instances where it seems permissible to save an infants life over that of an adult human being, even
if the infant does not have more to lose.
Still others might suggest that perhaps we save the infants not
because the infants are rightholders but because in doing so we would
be promoting the interest of rightholders, namely, the interest of the
parents. This argument may have some plausibility if the choice is
between saving an infant and the infants parent, and the parent has
given indications that he would want the infant rather than himself to
be saved. But even when the choice is between saving an infant and an
unrelated adult human being, it often still seems permissible to give
preference to the infant. If this is right, this is one reason why actual
moral agency cannot be the sole ground for rightholding.
5. Actual Agency Matters More
Still other people might accept that the genetic basis for moral agency
matters somewhat for moral status, but they might insist that actual
moral agency always matters more, and therefore, only those with
actual moral agency should be rightholders. For example, suppose an
IVF clinic is burning and one could either save an embryo, which has
the genetic basis for moral agency, or a five-year-old child, who has
352
s. matthew liao
actual moral agency.32 It seems that one should save the five-year-old.
If so, this seems to show that actual moral agency matters more.
First, this case does not show that actual moral agency always
matters more. Imagine a variant of this case in which another IVF
clinic is burning, and there is another embryo and another five-yearold child. This time, though, the five-year-old child has inhaled too
much smoke, has temporarily passed out, and will die soon (in a few
days time) and the embryo happens to belong to you. You have been
desperately trying to have a child for years and the doctors tell you that
this time there is a very good chance you will succeed, but that this is
your last chance to have your own biological child. Is it so intuitively
clear that you should save the five-year-old? My intuition is not as clear
in this case. Indeed, I think you may be permitted to save your last
embryo. Note that suppose the alternative is between saving the fiveyear-old and saving, for example, your beloved Picasso painting.
Clearly, you should save the five-year-old. This suggests that the reason
why you may be permitted to save the embryo in this case is not just
because the embryo is your embryo, since the Picasso is also your
Picasso. If this is right, there may be certain circumstances in which
saving someone with the genetic basis for moral agency matters more
than saving someone with actual moral agency.
In any case, it is not necessary to deny that often an actual moral
agent will matter more than a being with just the genetic basis for moral
agency. But this would not show that the two beings have different
moral status. Indeed, when a stranger and ones partner are both
drowning, other things being equal, ones partner typically matters
more than the stranger; but this would not show that the two have different moral status.
Because an embryo typically has a greater future potential than
a grown child, some people might worry instead that the genetic basis
for moral agency account implies that we should always save the
embryo rather than the child. This implication need not follow. First,
just as there could be agent-relative reasons for choosing ones
embryo,there could also be agent-relative reasons for choosing ones
grown child, even if the embryo indeed has a greater future potential.
Secondly, there could also be agent-neutral considerations why one
32
See, e.g., S. Matthew Liao, The Embryo Rescue Case, Theoretical Medicine and
Bioethics 27.2 (2006), pp. 14147.
353
would choose the grown child over the embryo. In particular, it could
be argued that the embryo will have little or no time-relative interests
while the grown child may have very strong time-relative interests.33
To have time-relative interests is to be able to stand in some psychological relations to ones future and past selves. The strength of ones
present time-relative interests depends on how strongly one is psychologically connected to those future and past selves. For example, an
infant will typically have weaker time-relative interests in continuing
to live than a grown adult, since an infant has little or no awareness of
his or her future self. Similarly, the embryo will have little or no timerelative interests, since the embryo will not have the required capacities
to be able to stand in some psychological relations to the embryos
future and past selves. The grown child, on the other hand, may have
very strong time-relative interests, since the child could by then have
fully developed p
sychological capacities. If so, this could provide an
agent-neutral reason to choose the grown child over the embryo,
because although the grown child may have less future potential, the
grown child may have stronger time-relative interests. Still, the fact
that there is a difference in time-relative interests between the embryo
and the grown child need not be taken to mean that there is also a
difference in moral status.34
6. Anencephalic Infants and Embryos Cannot Have Interests
Some people might ask, in what sense can anencephalic infants and
embryos have interests that can ground rights? Following Feinberg,
many people hold that to have an interest is to be interested in something, which requires that one has the capacity to desire and want
things and to be self-aware.35 Given that anencephalic infants and
embryos lack brains, they lack the capacity to desire and want things.
Consequently, on this view of interest, they cannot have interests that
can ground rights.
33
The concept of time-relative interest comes from McMahan, The Ethics of Killing,
pp. 80, 17074, 18388. See also S. Matthew Liao, Time-Relative Interests and
Abortion, Journal of Moral Philosophy 4.2 (2007), pp. 24256.
34
I develop this point in greater detail in Liao, The Embryo Rescue Case.
35
J. Feinberg, The Rights of Animals and Unborn Generations, in Rights, Justice,
and Bounds of Liberty (Princeton, Princeton University Press, 1980), pp. 15984.
354
s. matthew liao
First, it is worth pointing out that healthy newborn infants would

also not be able to have this kind of interest. Indeed, newborn infants
are hardly self-aware. In fact, it has been argued that infanticide is permissible given that newborn infants cannot have this kind of interest.36
Surely, this is a bullet one would not want to bite too quickly.
Secondly, unless proponents of this point wish to win the argument
by stipulation, there is an alternative view of interest that is equally
common and plausible. That is, one can have an interest if something is
in ones interest, namely, if something can affect ones flourishing, even
if one is not able to be interested in it.37 For example, as a number of
writers have argued, it is in the interest of a plant to be watered, as the
plant will die if it does not receive adequate water, even though the
plant is not able to be interested in being watered.38 Or, an individual in
a deep coma can have an interest in being fed, even if the individual
cannot desire being fed. If plants or individuals in deep coma can have
this kind of interest, then anencephalic infants and embryos can also
have this kind of interest. For example, it may be in their interest not to
be destroyed or mutilated, not to be used as food or objects of experimentation or play, and possibly not to be used as sources of organs.
Some people might insist that plants cannot have interests because,
according to them, plants are merely valuable in so far as they are valuable to us, and not for their own sake. But from a biological perspective, a plant can certainly need water for its own sake. In any case, there
is no reason to think that anencephalic infants and embryos cannot be
valuable for their own sake and cannot have interests in this sense.
Thirdly, Feinberg defends the more restrictive notion of interest
because he holds the view that being able to have an interest is a sufficient condition for rightholding, and the view that entities such as
plants cannot be rightholders.39 However, one can deny that being
able to have an interest is a sufficient condition for rightholding. If so,
there would be less pressure to have this restrictive notion of interest.
See, e.g., Singer, Practical Ethics, chs. 2 and 4.

For philosophers who have argued that Feinbergs notion of interest is too narrow
and that some nonsentient entities such as plants can also have interests, see,
e.g., G. Varner, In Natures Interests? Interests, Animal Rights, and Environmental Ethics
(New York: Oxford University Press, 1998); K. Goodpaster, On Being Morally
Considerable, Journal of Philosophy 75 (1978), pp.308325.
38
See, e.g., Varner, In Natures Interests?; Goodpaster, On Being Morally
Considerable.
39
Feinberg, Rights of Animals and Unborn Generations.
36
37
355
On a view according to which having an interest and having rights are

two different matters, the fact that a plant can have interests would not
mean that it is a rightholder. Indeed, anencephalic infants and embryos
are rightholders because they have the genetic basis for moral agency,
and not just because they can have interests.
7. A Reductio?
The last objection I will consider here is that some might think that
the genetic basis for moral agency account faces a possible reductio
ad absurdum. According to this line of thought, it seems that every
human cell has the genetic basis for moral agency, since theoretically
it is possible to produce a human being from any human cell. But if
every human cell has the genetic basis for moral agency, would it not
follow that every human cell is a rightholder? Clearly, human cells are
not rightholders. So there must be something wrong with this account.
This argument fails, however, because human cells are not beings;
they are parts of a being. It is not so easy to distinguish a being from
parts of a being, but it seems that the kind of being at issue is some sort
of a self-contained organism that is ready to develop into an individual
or that is developing as an individual as long as nutrients and normal
developing conditions are provided. On this view, a human cell is not a
being, because it is not a self-contained organism ready to develop into
an individual or that is developing as an individual as long as nutrients
and normal developing conditions are provided.
Moreover, for a human cell to become a human being, its nature
must be significantly changedto the extent that it would stop being
a human cell. Specifically, given present technologies, one must
de-differentiate the genetic material of a human cell, that is, make the
cell unspecialized, e.g. make a skin cell into an unspecialized cell; and
transfer the genetic material to a viable egg. As I noted before, for a
being to qualify as having the genetic basis for moral agency, it is not
enough just to have the genes for the genetic basis for moral agency
located somewhere in the being; these genes must be activated and be
coordinating with each other in the appropriate manner, which is not
the case with normal human cells.
This last point allows us to respond to some hypothetical examples
involving injecting or integrating the genes for the genetic basis for
moral agency into non-human entities. For example, suppose we
356
s. matthew liao
injected these genes into a cabbage, would the cabbage be a rightholder?40 To answer this question, we should inquire whether these
genes are integrated into the cabbage or not. Ifthey are, then the cabbage may no longer be a cabbage, since its nature may have changed.
In such a case, the resulting being may be a rightholder, given that it
has the genetic basis for moral agency. A real-life parallel may be transplanting human genes into a rabbit egg and creating an embryo with
human characteristics.41 Arguably, such an entity may be a rightholder.
On the other hand, suppose that the genes for the genetic basis for
moral agency are not integrated into the cabbage, but instead these
genes just sit inside the cabbage. In such a case, the cabbage may not
have the kind of moral status that a cabbage with integrated genetic
basis for moral agency may have.
8.Conclusion
In this paper, I proposed that the genetic basis for a moral agency
account of rightholding can allow all human beings to be rightholders
without being speciesist. Although I did not try to argue that this
account is the correct account of human moral status, I did argue that
this account is different from a potentiality account and that it is preferable to an actual moral agency account of human moral status. While
it was beyond the scope of this paper to consider all issues arising from
this account, I have laid out some of its particular strengths. By being
able to incorporate the intuition that an appropriate account of rightholding should be species neutral and by being able to explain how all
human beings are rightholders, the genetic basis for the moral agency
account appears to provide a real and coherent alternative to present
accounts of rightholding and therefore deserves further study and to be
part of the debate about the foundations of human moral status.
I thank Agnieszka Jaworska for this point.

Ying Chen, Zhi Xu He, et al., Embryonic Stem Cells Generated by Nuclear
Transfer of Human Somatic Nuclei into Rabbit Oocytes, Cell Research 13.4 (2003),
pp. 251-63.
40
41
The Mixed Solution to the Number Problem

Martin Peterson
1.Introduction
Suppose you face a choice between two alternatives. You must either
save two people and let one person die, or let two people die and save
one. There are no morally relevant differences among the people. What
should you do? This example has been discussed extensively in the literature in recent years and is commonly referred to as the number
problem.1
Unsurprisingly, several views have been proposed. Utilitarians maintain that one should always save the greatest number, because this
would bring about the greatest amount of wellbeing. However, as
famously objected to by Taurek and others, the utilitarian solution
seems to be unfair.2 Imagine, for instance, that two cruise ships have
struck two separate icebergs and are sinking rapidly. Now, the captain
of a nearby rescue boat has to decide which ship to assist. The rescue
boat can only assist one of the two ships. The mere fact that there are
more people aboard one of the ships is not sufficient reason for totally
ignoring those aboard the other ship. Each person should be granted
some possibility of being rescued; all of us have morally significant
interests that ought to be given at least some weight. In order to account
for this intuition, it has been suggested that the decision should be
taken by tossing a fair coin.3 Such a lottery would give each individual
an equal chance of being rescued. If fairness is all that matters, then this
proposal certainly makes sense.
1
The locus classicus is John Taurek, Should the Numbers Count?, Philosophy and
Public Affairs 6 (1977), pp. 293316. See also John Broome, Fairness, Proceedings of the
Aristotelian Society 91 (1990), pp. 87102, and his Kamm on Fairness, Philosophy and
Phenomenological Research 58 (1998), pp. 95561. More recent contributions to the
debate include Iwao Hirose, Saving the Greater Number without Combining Claims,
Analysis 61 (2001), pp. 34142, and Hirose, Aggregation and Numbers, Utilitas 16
(2004), pp. 6279.
2
In addition to the references given in the first footnote, see also Michael Otsuka,
Saving Lives, Moral Theory, and the Claims of Individuals, Philosophy & Public
Affairs34 (2006), pp. 109135.
3
See e.g. Taurek, Should the Numbers Count? and Broome, Kamm on Fairness.
358
martin peterson
However, according to some authors, a lottery in which each individual has an equal chance of being rescued is unjustified in many
cases, e.g. in a choice between saving either a million people or ten
people.4 A fifty percent risk of not saving a million people is simply
morally unacceptable. In response to this objection, Broome has proposed that the number problem could be resolved by setting up a
weighted lottery, in which the probability that the first group (of m people) is saved is m/m+n, and the probability that the second group (ofn
people) is saved is n/m + n.5
That said, arranging a lottery seems to be entirely out of the question
in a choice between saving a million or ten people, no matter the probabilities. In such an extreme situation, it can be argued that the positive
moral value gained by arranging a lottery is entirely outweighed by the
number of people saved when saving the greatest number. The present
article aims at exploring this claim further. The hypothesis is that
arranging a lottery contributes towards making an act fairer, while
other features, such as non-optimal consequences, contribute towards
making the act worse in other morally relevant respects. Doing the
right thing ultimately amounts to balancing different values against
each other. In what follows, this view will be referred to as the mixed
solution.
The mixed solution has been briefly touched upon by John Broome
and Iwao Hirose, but it has never been extensively researched.6 Contrary
to Broome and Hirose, I argue that the mixed solution can be defended
without a ssuming the possibility of interpersonal value comparisons.
That is, the mixed solution can be accepted without claiming that one
ought to save two lives rather than one because two counts for more
than one. This might look like a rather technical and modest point, but
closer examination reveals that it is not. A p
rominent reason for taking
the number problem seriously is the widespread belief that moral value
cannot be aggregated across different people, not even in principle.
Anyone wishing to combine the claims of different people into a unified, stronger moral claim has to explain how this can be done in a
non-arbitrary way. We are all individuals, and this fact is surely of some
moral importance.
See e.g. Broome, Kamm on Fairness, and Hirose, Aggregation and Numbers.
Broome, Kamm on Fairness.
6
See e.g. Broome, Kamm on Fairness, and Hirose, Aggregation and Numbers.
4
5
the mixed solution to the number problem
359
Before proceeding, an important restriction has to be introduced.

In what follows, the mixed solution will only be analysed from the
consequentialists point of view. Other kinds of positions, based on
notions such as rights, virtues and fair contracts, will not be touched
upon.7 This is because the consequentialists ethical framework is the
only one in which the concept of aggregation plays a prominent role.
Even though e.g. the advocate of a rights-based theory could claim that
it is worse to violate two rights rather than one, it is generally agreed
that he or she is in no way committed to this view. The violation of the
rights of one person might be equally as bad as, or even incomparable
to, the violation of the rights of two people. However, nearly all consequentialists believe that consequences can somehow be aggregated
interpersonally, so articulating a view not relying on that assumption
seems to be worthwhile.
The structure of this contribution is as follows. In section 2, the utilitarian solution to the number problem is scrutinized in more detail, as
is a defence for a similar view put forward by Hirose. In section 3, an
account of the mixed solution proposed by Hirose is analysed and criticized. Thereafter, in section 4, I develop my own account of the mixed
solution, and explain why it is more attractive than previous versions.
2. Utilitarianism and Beyond
Is there anything wrong with utilitarianism? As pointed out above,
utilitarianism provides an easy solution to the number problem. Fullfledged utilitarians could simply point out that one should save the
greatest number of people, because that will typically produce the
greatest amount of wellbeing. However, as briefly indicated above, utilitarianism fails to acknowledge the importance of fairness. Utilitarians
assign no moral importance what so ever tofairness, and this contradicts our considered moral intuitions. Of course, some utilitarians
might response to this argument by biting the bullet and saying that
fairness is a morally irrelevant concept. However, let us for the sakeof
the argument assume that there is something wrong with the utilitarian
analysis, in order to find out what we might learn from the number
problem.
7
For a contractualist approach to the number problem, see Tim Scanlon, What We
Owe to Each Other (Cambridge, MA: Harvard University Press, 1998).
360
martin peterson
Needless to say, it should be acknowledged that even if one finds

utilitarianism unacceptable, it does not follow that it would be wrong
to save the greatest number, i.e. that the Greatest Number view is false.
Hirose proposes a more sophisticated, non-utilitarian defence of the
Greatest Number view. His main point is that non-utilitarian consequentialists can justify the case for saving the greater number without
combining the goods of separate people,8 that is, without carrying out
any kind of interpersonal aggregation of moral value.
Briefly put, Hirose attempts to show that the following two principles, both of which are weaker than the utilitarian view, support the
Greatest Number view.
Symmetry: Two alternatives are equally good if they differ only with
regard to the identities of the people.
Pareto: If one state of affairs is better for some persons than another state,
and if it is worse for no persons, then it is better than the other.
Symmetry is a criterion of impartiality, which is accepted by most (but

not all) consequentialists. The Pareto condition has been imported
from economics. It is, of course, a substantial ethical principle. Most
notably, the Pareto condition is inconsistent with most principles of
equality, because if one improves the situation for one person while
leaving it unchanged for all others, this change will lead to increased
inequality. Arguably, this should be a matter of concern for most people. However, note that neither Symmetry nor Pareto presupposes any
kind of interpersonal aggregation of goods.
In order to see how Hirose applies Symmetry and Pareto for justifying the Greatest Number view, consider a choice among the alternatives listed in Table 1. If Alternative 1 is chosen, person A will be saved,
but not B and C, and so on. Now, Symmetry implies that Alternatives
Table 1
Alternative 1:
Alternative 2:
Alternative 3:
ABC
(saved, dead, dead)
(dead, saved, dead)
(dead, saved, saved)
Hirose, Aggregation and Numbers, p. 68.
361
1 and 2 are equally good, and Pareto implies that Alternative 3 is better
than Alternative 2. Therefore, given that the relation better than is
transitive, it follows that Alternative 3 is better than Alternative 1, i.e.,
we should save the greatest number.9
Unfortunately, Hiroses argument only works as long as all alternatives are non-probabilistic. Even though it is true that Alternative 3 is
indeed better than Alternative 1, nothing more can be concluded. In
particular, it does not follow that Alternative 3 is better than any of the
two major alternative positions in the debate over the number problem, the Equal-Chances view and the Proportional-Chances view,
mentioned in section 1. By letting (50% saved; 50% (saved, saved))
denote a lottery in which the probability is 50% that thefirst individual
is saved and 50% that the two others are saved, the Equal-Chances view
and the Proportional-Chances view can be described as in Table 2.
Evidently, Hiroses two principles, Symmetry and Pareto, are not sufficient for deciding which alternative to choose. Since Pareto is only
applicable to deterministic cases, it cannot be used for adjudicating
among probabilistic alternatives. Therefore, Hiroses argument fails to
show that the Greatest Number view (Alternative 3 above) is morally
better than the Equal-Chances view and the Proportional-Chances
view. Consequently, since the latter views are among the most wellestablished ones in the literature, the reasoning suggested by Hirose is
not sufficient for giving a comprehensive analysis of the number problem. It is simply not true that Symmetry and Pareto imply that consequentialists can justify the case for saving the greater number without
combining the goods of separate people.10
Table 2
The Equal-Chances view:

The Proportional-Chances view:
AB
C
(50% saved; 50% (saved, saved))
(33% saved; 66% (saved, saved))
9
This conclusions has been questioned in the literature; see e.g. Thom Brooks,
Saving the Greatest Number, Logique et Analyse 17778 (2002), pp. 5559.
10
362
martin peterson
3. The Mixed Solution
Consider two very different instances of the number problem. Firstly,

suppose that you could save either one or two people. Secondly, suppose that you could save either a group of 1000 people, or a group of
1001 people. Some authors have claimed that these two examples ought
to be treated differently.11 In the latter example, fairness seems to outweigh the opportunity to save one extra person; therefore, the decision
ought to be made by tossing a fair coin. However, when facing a choice
between saving either one or two people, the opportunity to save one
extra person is more important than fairness, since only one person
will be treated unfairly. Hence, in that case, one ought to save two people for sure, instead of arranging a lottery. This is the mixed solution.
It simultaneously takes into account two apparently conflicting considerations, viz. fairness and the number of people saved.
According to Hirose, an unfair distribution of chances is morally
bad, and this badness can be aggregated interpersonally.12 In the choice
between saving 1000 or 1001 people, coin-tossing amounts to saving an
expected number of 1000.5 lives, whereas if one decides to save the
greatest number, 1001 people will be saved. However, each of the 1000
people not saved will suffer from some unfairness u. The total amount
of unfairness created by saving the greatest number of people is u
1000. See Table 3.
For present purposes, it is helpful to assume that moral value is additive. Thismeans that in Table 3, tossing a fair coin is morally better than
saving thegreatest number, unless u is very close to 0. This is because
1000.5 + 0 > 1001 u 1,000, for every sufficiently large u. For the sake
of the argument, we assume that u is in fact sufficiently large, i.e., that
one ought indeed to toss a fair coin rather than saving the greatest
number in the choice between saving 1001 or 1000 people.
Table 3: Save 1000 or 1001?
Number of people saved Aggregated unfairness
Toss a coin
Save the greatest number
1000.5
1001
0
-u 1000
11
The most extensive discussion of this view can be found in Hirose, Aggregation
and Numbers. See also Broome, Kamm on Fairness.
12
Hirose, Aggregation and Numbers.
363
The point of this example is that the analogous conclusion does

nothold true when choosing between saving one or two people; see
Table 4. Given that u is of the same magnitude as before, it follows that
it is better to save the greatest number in this case, instead of tossing a
fair coin. Fundamentally, this is because there is only one person who
is treated unfairly, whereas the difference in the number of people
saved is the same as in the previous example, i.e., on average 0.5 extra
persons.
Note that the present line of thought, which I take to be a reasonable
explication of Hiroses view, relies heavily on a certain non-trivial
assumption. The assumption in question is that there is a larger total
amount of unfairness in the first example (Table 3) as compared to the
second (Table 4). Is this a plausible assumption? Well, it seems obvious
that more people will in fact be deprived of the chance of being saved if
the Greatest Number view is adopted in the example illustrated in Table
3, as compared to the example illustrated in Table 4. However, it is not
obvious that this is in itself a morally relevant fact. Consider a choice
between saving either 1000 people or 2000 people. Compare this with
a choice between saving either one or two people. It is not obvious that
the total aggregated amount of unfairness produced by saving the
greatest number in these two examples differs. For each person who is
left behind, there are two others who are saved; the proportions are the
same. Hence, it is not evident that Hirose is right when he claims that,
The bad of unfairness, done to each person, is constant, regardless of
the number of people concerned.13
Arguably, advocates of the mixed solution need a more precise
account ofunfairness. Isabel Burnham has proposed such an account.14
Table 4: Save one or two?
Number of people saved
Aggregated unfairness
Toss a coin
Save the greatest number
1.5
2
0
-u 1

Isabel Burnham, Pandemic Influenza: Whom Should we Treat? A Real-life
Number Problem (Cambridge: Department of History and Philosophy of Science,
2006).
13
14
364
martin peterson
Briefly put, her proposal is that unfairness should be measured by

applying the Gini-index to each individuals chances of being saved.
The Gini-index assigns a number between 0 (perfect equality) and one
(perfect inequality) to every distribution of numbers. (Usually, these
numbers measure each individuals income or wellbeing, but in the
present context, it is reasonable to suppose that they reflect each persons probability of being saved.) The formula for calculating the Giniindex is complex.15 However, in order to visualize the Gini-index
graphically, suppose that all persons are ordered along a horizontal axis
according to their probability of being saved, from the worst-off to the
best-off. The Lorenz curve is then the cumulative sum of numbers
obtained by adding the first persons probability of being saved to that
of the next person, etc. Let these cumulative numbers be represented
on the vertical axis. For example, if everyone enjoys equal chances of
being saved, the Lorenz curveis a diagonal line going from the lower
left corner to the upper right one,and if a single person will be saved
with probability one and all others leftbehind, the Lorenz curve is a
straight horizontal line that turns up at the right end of the vertical axis.
The Gini-index corresponds to the area trapped by the hypothetical
straight line denoting perfect equality and the actual Lorenz curve for
the distribution of probabilities, divided by the area of the entire triangle. Let us assume that this is a plausible way of measuring degrees of
fairness.
Now, it can be easily checked (and intuitively visualized) that it is, in
fact more fair to save two and let one die, compared to saving 1001 and
letting 1000 people die. This means that the value of u is not, contrary
to what Hirose assumes, the same in the examples illustrated in Tables
3 and 4. The value of u, i.e., the amount of individual unfairness, varies
from case to case. The correct value is obtained by calculating the Giniindex for each distribution of probabilities.
However, Hirose may nevertheless be right that the aggregated
amount of unfairness, if such an entity exists, is higher when 1000 people are left behind, compared to the case in which only one person is
left behind, because it may very well hold true that u1 1,000 > u2 1.
15
Let p be a vector consisting of each persons probability of being saved, and let p be
the average probability in p. The Gini-index G(p) is then defined as follows:
G(p) =1
pi p j .
2n2 p i =1
j =1
365
I shall refrain from discussing whether or not this is the case. My point
is simply that Hiroses analysis relies on an assumption that Burnham
has shown to be doubtful: a single, fixed value of u cannot be used for
analysing cases involving very different numbers of people.
John Broomes view is to a large extent similar to Hiroses. Broome
argues that, fairness is not everything. Fairness requires tossing a coin,
but just as Ithink the fairness of saving no one is outweighed by the
badness of the result, so I think the fairness of tossing a coin is outweighed by the expected badness of the result.16 I take this to mean that
Broome also accepts that unfairness can be aggregated into a whole.
Or, put slightly differently, that Broome agrees that it makes sense to
multiply u by 1000 when a thousand people are left behind, and so on
and so forth.
That said, the idea that fairness can be aggregated into a whole is
presumably a bit questionable. How can there be such a thing as an
aggregated amount of unfairness? Why should we believe that unfairness can be aggregated interpersonally, i.e. that it makes sense to say
that persons A and B together are treated more or less unfairly than a
single person C? Arguably, the available evidence about fairness, which
mainly consists of vague moral intuitions, only supports the idea that a
single person can be treated more or less unfairly in relation to other
possible ways of treating that particular person. To say that person A
and B are facing more unfairness than C is a question-begging claim;
anyone wishing to make that claim first has to tell us more about the
concept of fairness. Of course, a mere reference to the Gini-concept
would not do in the present context. Firstly, it is far from clear that it
makes sense (in a substantial way, not just in a mere technical way) to
compare the Gini-index for populations of different sizes. Secondly, a
more fundamental problem is whether there is any reasonable way at
all in which fairness can be interpersonally measured. The Giniformula in itself does not answer that question. It is a mere mathematical abstraction, which may or may not correspond to a phenomenon in
the world.
Surely, if 1000 people are treated unfairly, that is morally worse than
if one person is treated unfairly, all else being equal. However, this conclusion does not presuppose that there exists some collective, interpersonal entity corresponding to an aggregated amount of unfairness.
Broome, Kamm on Fairness, pp. 95657.
16
366
martin peterson
It is sufficient to point out that 1000 people are being treated badly; the
reason need not be that the Gini-index of the corresponding distribution of probabilities is morally unfavourable.
4. An Alternative Account of the Mixed Solution
This section proposes an alternative account of the mixed solution,
which avoids at least two of the problems associated with the BroomeHirose analysis. Firstly, the new account does not assume that unfairness can be aggregated interpersonally. As argued above, there is no
reason to believe that there is such thing as an accumulated sum of
unfairness for each group of people treated badly. Secondly, the new
account does not assume that moral value can be compared interpersonally; this point is analogous to the first one.
The main feature of the new account is that fairness is construed as a
moral value that is good for the individual. There are no collective
moral values. All moral values, including fairness, can at most be aggregated on a personal level. Consider the following, general formulation
of this idea:
Individualism: All moral values, including fairness, are individual values
which cannot be interpersonally aggregated.
This principle is best explained by way of analogy: Suppose that it

would be unfair to use illegal steroids while participating in the Olympic
Games. Then, the more clean athletes who face unfair competition, the
worse it would be from a moral point of view. However, according to
Individualism, unfairness is something that affects the individual only.
There is no such thing as an aggregated amount of unfairness materialized in the Olympic Games. The badness resulting from unfair competition is something that makes each individual worse off that he or she
would have been otherwise. The claim that it is worse from a moral
point of view if a large number of people are affected is just a claim
about the ranking of alternative acts. It is not a claim about some
amount of aggregated unfairness in the world.
Further to this point, even a winner who has benefited from illegal
steroids would be worse off than he or she would have been otherwise.
All things being equal, it is better to win in the Olympics without using
illegal steroids, because that would be a fair victory. Furthermore, all
things being equal, it is better to be defeated in the Olympics in a fair
competition than in an unfair one. However, nothing could be said
367
about whether the winner suffers more or less unfairness than a group
of other athletes do together. Such claims would invariably require
interpersonal aggregation of individual moral values.
Hirose defines the concept of aggregation as follows:17 a vector of
normatively relevant features (a1, a2,, ai)is at least as F as (a'1, a'2,,
a'j)if and only if f (a1, a2,, ai) f (a'1, a'2,, a'j)where f( )is a strictly
increasing function that is not bound above. Suppose that F is fairness.
Then, the principle of Individualism should be taken to mean that there
is no function f( ) that orders different distributions according to
degrees of fairness. However, it might still be possible to rank the set of
vectors according to some other feature F', such as choice-worthiness.
Individualism is a moral principle on the same level of abstraction as
the two principles suggested by Hirose, i.e. Symmetry and Pareto. This
means that all three principles are fundamental moral principles, from
which everyday moral rules are to be derived. Taken together, these
principles can explain the moral difference between the choice of saving 1000 or 1001 people, r espectively the choice of saving either one or
two people. First consider the choice between saving 1000 people or
1001. In Table 5, one could either save a group, A, of 1000 people,
oranother group B of 1001 people. By applying Symmetry and Pareto
to the latter alternative, it follows that this alternative is morally better
than saving the smaller group A. However, the third alternative, tossing
a fair coin, will be morally better than the second alternative in case
u > 0.5. Let us, for the sake of the argument, suppose that this condition
is fulfilled. It then follows that tossing a fair coin is the morally right
Table 5: Save 1000 or 1001?
A1
A2
A999
A1000
B1
B2
B1000
B1001
Save A
(1000)
Save B
(1001)
Toss a
coin
1u
1u
0u
0u
0u
0u
1u
1u
0.5
0.5
0.5
0.5
17
368
martin peterson
Table 6: Save one or two?
A1
B1
B2
Save one
Save two
Toss a coin
0 u'
1 u'
0.5
0 u'
1 u'
0.5
1 u'
0 u'
0.5
thing to do in the choice between saving A or B. Note that this conclusion has been reached without assuming that moral values, such as
fairness or utility, can be interpersonally aggregated.
Now consider the choice between saving either one or two people;
see Table 6. Since each person treated unfairly in this example faces less
unfairness than in the previous example, it may safely be assumed that
u' < 0.5. In this example, the moral principles considered so far are
insufficient for deciding what to do. This is because no alternative is
Pareto-better than the others. All that can be concluded is that saving
two people is morally better than saving one. In order to conclude that
saving two is better than tossing a coin, one has to assume that u is low
and that moral values are not individual, i.e., that they can be aggregated interpersonally by, e.g., adding each individuals value to that of
others. As argued above, this is a questionable assumption. Before one
decides whether to save one, or two, or to toss a fair coin, some additional moral assumptions are required.
However, there is an easy way out of this predicament. The key idea
is to make a sharp distinction between evaluative and normative claims.
In Table 6, each number represents each individuals moral value under
different alternative acts. The numbers can thus be seen as a summary
of a set of evaluative statements. This set of evaluative statements is, of
course, consistent with several different normative (action-guiding)
principles. For instance, it could be argued that if an alternative is not
Pareto-better than a set of other alternatives, then all those alternatives
are morally permissible. In the example illustrated in Table 4, this liberal normative principle would imply that saving two p
eople and tossing a fair coin are permissible alternatives, whereas saving one person
is not. That is, an act is morally permissible as long as it is not defeated
by another act, in terms of Pareto-betterness. Typically, this means
that a rather large subset of a set of alternative acts will be morally
permissible.
369
To sum up, the main argument of this article looks as simple as it can
r easonably be. Both fairness and the total number of people saved matter, but neither value can be aggregated interpersonally. Therefore, the
liberal normative principle outlined here ought to be adopted when
facing the number problem.18
18
I would like to thank Iwao Hirose and Barbro Bjorkman for helpful comments on
an earlier draft. My work on this paper has been supported in part by a grant from the
Swedish Emergency Management Agency.
Trivial Sacrifices, Great Demands*

William Sin
This is the Trilemma of Easy Rescue: Most of us accept the following
three ideas. The first is the Principle of Easy Rescue, namely, if an agent
can relieve some victims dire plight by making a small sacrifice, it is
wrong for her not to do so. The second is the Inexhaustible Nature of
Positive Duties. We believe that there is no ultimate amount of sacrifice
that an agent can make after which she will no longer be responsible for
helping the needy. The third is the Non-Extreme Nature of Morality,
that is, an acceptable moral theory will not require an agent to sacrifice
everything she has to help the needy.
Each of the above three ideas seem true when considered independently, but they also seem inconsistent, forcing us to jettison some of
them when we consider them together. To get out of the trilemma,
I suggest we give up the Non-Extreme Nature of Morality. If we accept
this contention, we will find that we are committed not only to
Extremism, but Super Extremism, according to which morality may
require agents to give everything to the needy, instead of giving away a
certain greater-than-moderate amount.
1. The Trilemma
Consider the case of famine relief. Numerous victims in poor countries
are starving and suffering. Agents in affluent countries have readily
available information about this situation. They can greatly help individual victims each time they contribute a small donation to them.
Suppose relief organizations are willing to save lives via private funding. There is a question concerning whether or how frequently agents
* I owe particular thanks to Brad Hooker, Jeff McMahan, Daniel Wikler, and Melissa
Lane, who have provided helpful criticisms and encouragement. Wong Wai-Hung and
Luke Mulhall have helped polish the English of this article.
372
william sin
should contribute such small donations to relieve the plight of the

starving victims.
There can be three responses to the case of famine relief. The first
response is from theorists of moderate morality. Moderates hold that,
in the case of famine relief, although morality imposes demands on
agents, these demands are not extreme. Moderates subscribe to the
thesis of demandingness: If a moral theory required agents to perform
actions which would undermine their lives, the theory would be inherently problematic.1 (In what follows, I will call this thesis the problem of
demandingness.) This view of moderate morality gains some support
from our intuition. Many people believe that agents in general can lead
a morally decent life without lowering the quality of their own lives.
However, if we accept the argument I propose in this paper, we will see
that, in regard to famine relief, Moderates view is not the most
plausible.
Second, there is Moral Extremism. Extremists believe that morality
may legitimately require agents to make extreme sacrifices.2 Thus, a
moral theory can impose extreme demands on agents without being
objectionable. The claims of Moral Extremism may conflict with peoples commonsense. It is not clear whether we should take this as a serious problem of Extremism. In any case, in this paper, I will argue for
not only an Extremist View, but a Super-Extremist View. The SuperExtremists hold that morality may demand agents to sacrifice everything to help the needy. Such a view is in greater conflict with
commonsense morality than Extremism is.
The third position in response to the demands of famine relief is
Moral Minimalism. Minimalists believe that agents have full prerogatives to decide whether and how much they would donate to save the
needy in poor countries. On their view, it is not wrong if agents choose
not to make any sacrifices at all. Minimalism seems implausible, for if
agents could save someones life by making a trivial sacrifice, it would
1
Bernard Williams, Moral Luck (Cambridge: Cambridge University Press, 1981),
p. 18. See also James Griffin, Value Judgement (Oxford: Oxford University Press, 1996),
pp. 9091; Samuel Scheffler, Human Morality (New York: Oxford University Press,
1992), p. 102; Barbara Herman, Morality and Everyday Life, Proceedings and Addresses
of the American Philosophical Association 74 (2000), pp. 2945.
2
See e.g. Peter Singer, Famine, Affluence and Morality, Philosophy & Public Affairs
1 (1972), pp. 22943; Peter Unger, Living High and Letting Die: Our Illusion of Innocence
(New York: Oxford University Press, 1996); Shelley Kagan, The Limits of Morality
(Oxford: Oxford University Press, 1989).
trivial sacrifices, great demands
373
be obviously wrong for her not to do so.3 However, there may be some
Minimalists who hold that morality may demand agents to make some
small, but less-than-moderate, amount of sacrifice for the needy. The
closer Moral Minimalism gets to the position of the Moral Moderate,
the harder Minimalism is to reject.
As a response to famine relief, Minimalism deservedly receives the
least attention from philosophers in the debate. But note that the
Minimalist View still dominates many peoples intuition about the case
of famine relief. People in the street will not condemn someone, who
has an average income in an affluent country, for choosing not to make
any donations to save the needy in poor countries. In the minds of
these people, even if it takes only trivial sacrifices to save the needys
lives, it will still be a charity, instead of a moral requirement, for agents
to make the donations.4
Extremism, Minimalism and Moderate Morality divide the conceptual space under which a theorist may take sides in the case of famine
relief.5 The soundness of these positions is dependent on a number of
considerations. One such consideration is the Principle of Easy Rescue.
That is, if an agent can prevent something very bad from happening, or
alleviate someones dire plight, by making a trivial to moderate sacrifice, then it will be wrong not to do so.6 In the case of famine relief,
itseems that agents have some obligation to help the needy. However,
note that complying with the Principle of Easy Rescue in such a
3
This is the Principle of Easy Rescue, which we will introduce shortly. We will consider its counterarguments in section two of the paper.
4
Singer, Famine, Affluence and Morality, p. 235.
5
There are different ways to assess the weight of a loss. We may assess the percentage of a persons total wealth or some absolute level of sacrifice. For someone as rich as
Bill Gates, an amount as much as one billion dollars is only a tiny percentage of his
wealth, yet a great sum in absolute terms. In this paper, while I use the idea of percentage to understand the concepts such as extreme sacrifice and moderate sacrifice, I have
in mind people with an average income in the affluent world. For a sharper understanding, we may need a progressive standard, or even a standard which is defined
with reference to an agents level of well-being after she has made the donation. See
Liam Murphy, Moral Demands in Nonideal Theory (New York: Oxford University
Press, 2000), pp. 2021.
6
T. M. Scanlon, What We Owe to Each Other (Cambridge, MA: Harvard University
Press, 1998), p. 224. Also see Singers Famine, Affluence and Morality, pp. 22943. On
another occasion, Singer expresses his argument in a form that is similar to the
Principle of Easy Rescue: [N]ot to relieve the suffering of the worlds poor, when we
can do so at little cost to ourselves, is a serious moral failing. Peter Singer, A Response,
in D. Jamieson (ed.), Singer and His Critics (Oxford: Blackwell, 1999), p. 302.
374
william sin
circumstance can become demanding.7 For the conditions which give

rise to the compelling demands may persist even after the first few
rounds of agents compliance, and the agents may have to respond to
the principles demands time and again.
Thus, as agents comply with the Principle, there may involve an iterative effect. That is, even though, in each instance of compliance, the
amount of demanded sacrifice is trivial, the Principle of Easy Rescue
can end up being very burdensome for agents to follow.8 Put the point
in a different way, the Principle of Easy Rescue prima facie prescribes
innocuous and harmless actions; however, when agents try to comply
with it under certain circumstances, their subsequent compliance will
affect their own state of well-being significantly.
Regarding the moral demands posed by the Principle of Easy Rescue,
our question is how many small donations agents are required to give.
In Figure 1, I use a dot to represent each of the trivial sacrifices that an
agent can make to save a life in a poor country.
These dots go downward from the left to the right. If the agents
keep contributing their resources, the dot enclosed in the square of
Super Extremism at the bottom right-hand corner represents the last
possible sacrifice that agents can make. The grey area on the right
represents the territory of Extremists, who hold that agents should
sacrifice at least up to the squared dot of Extremism at the top of the
grey area on the right. The white area in the middle represents
Moderates territory. Moderates hold that morality does not require
agents to give anything beyond the squared dot at the bottom right
corner of the white area, which I call the Moderates Maximum
Point. The grey area on the left may be taken as Minimalists territory.
Even Moderates believe that agents should sacrifice beyond this
region. (The boundary between the extreme area and the moderate
area is vague, as is the boundary between the minimalist area and the
moderate area.)
Note a common feature of Extremism and Moderate Morality. Both
Extremists and Moderates may endorse a Threshold View. Although
Extremists are not indifferent to how many sacrifices agents should
7
Elizabeth Ashford mentions the possibility for the Principle of Easy Rescue to be
extremely demanding in her paper The Demandingness of Scanlons Contractualism,
Ethics 113 (2003), pp. 273302.
8
See James S. Fishkin, The Limits of Obligation (New Haven: Yale University Press,
1982); Garrett Cullity, Moral Character and the Iteration Problem, Utilitas 7 (1995),
pp. 28999.
375
Extremism
The Point of Super

Extremism
Moderates Maximum Point
Figure1. The Boundaries of the Minimalists, the Moderates and the

Extremists
make in response to the needs of victims in poor countries, yet they
may not require agents to contribute up to their last penny; insofar as
agents have donated a significant sum, they may have fulfilled the
demand of Extremism.9 Similarly, according to the Moderate View,
although agents should make a substantial amount of sacrifice in
response to the needs of the victims, once their aggregate sacrifices
have reached a certain moderate level, agents have a margin of discretion to decide whether they want to contribute more or not.
Holders of the Threshold View reject Moral Minimalism and Super
Extremism. They think that, first, there is some substantial amount
9
One can be an Extremist without endorsing Super Extremism. This is a unique
view which I recommend in this paper to understand the idea of Moral Extremism.
Generally speaking, philosophers may define Extremism with reference to a Super
Extremist View, which may not represent adequately the attributes of Extremism.
Cf. Tim Mulgan, Demands of Consequentialism (New York: Oxford University Press,
2002) p. 25; Kagan, Limits of Morality, pp. 1 and67.
376
william sin
that agents should sacrifice in the case of famine relief. (As this amount
may be great, moderate, or even small, even some Minimalists may
support, in principle, a kind of Threshold View.10) Second, agents are
morally free to decide when they will stop giving once they have made
a certain sum of aggregate sacrifice to the needy. Of course, since holders of the Threshold View suggest different thresholds, each of them
has to explain why it is the threshold she suggests that is the right
threshold. But this is not the most important issue here. I will now turn
to the issue concerning iterative effects that give rise to burdensome
demands. In the following discussion, I will not differentiate between
different Threshold Views.
When trivial moral demands iterate and become burdensome to
agents, a typical response is to emphasize the distinction between the
Iterative View and the Aggregate View. The Iterative View focuses on
the marginal cost that agents bear if they are to save an additional victim. As long as each of the victims can be saved at trivial costs, the
Principle of Easy Rescue applies. Then, agents compliance with the
Principle may generate great demands in aggregate.
Critics of the Iterative View may, however, hold that we do not have
to understand agents obligation in terms of their capability to save the
next victim. We may understand agents obligation in light of a cumulative approach. That is, if an agent has already contributed a great deal
to rescuing victims in poor countries, then the total sacrifices will
exceed a trivial (or even moderate) amount and she is not required to
donate further. So, on the Aggregate View, the Principle of Easy Rescue
does not apply to those agents who have accumulated a great sum of
total contribution, even though they can save the next victim by making an additional sacrifice that is trivial on absolute terms.11
Generally speaking, the Aggregate View is employed by Moderates
to reject Extremism. However, note that, it is possible for one to uphold
the Aggregate View without specifying which aggregate sum agents
should sacrifice. As holders of the Aggregate View believe in the existence of a threshold dividing the sphere of obligation and that of supererogation, we may now take the Aggregate View and the Threshold
10
This will be an exceptional case of Minimalism. With regard to the mainstream
Minimalists, their position will involve a denial of the Threshold View.
11
On the distinction between the Iterative View and the Aggregate View, see also
Garrett Cullitys The Moral Demands of Affluence (Oxford: Oxford University Press,
2004), pp. 8284.
377
View to be substantially the same. If we do so, Moderates and Extremists

can be taken to be allies against the Iterative View.
For holders of Aggregate Views, a commonsense justification for the
existence of a threshold is that other things being equal, agents responsibility to help the needy may become proportionally weaker as their
sacrifices in different times are aggregated. So as agents keep contributing, they may reach the point where their obligation towards the needy
is completely fulfilled. According to holders of the Aggregate View, even
if we cannot point out exactly where this threshold lies over a range of
possible sacrifices, we may still confidently refer to a certain area where
agents can permissibly reject the call to provide further assistance.
However, the Aggregate View is problematic in the way it depicts the
moral relation between capable agents and the potential needy. Holders
of this view tend to believe that agents positive duty to assist the needy
is exhaustible. That is, it is possible for agents, after they have made the
required sum of sacrifice, to terminate for good their responsibility to
assist others. I do not agree with this view; I do not think agents can
save moral credits in the way that they save pension funds. If the
Aggregate View were true, then people would be allowed to retire from
the moral world of assisting others simply by making a lot of sacrifices
at some stage of their lives. This is odd and counterintuitive.12
I suggest a counter principle, namely, the Inexhaustible Nature of
Positive Duties, or, for short, the Inexhaustibility Principle. This principle states that there is no ultimate amount of sacrifice that agents could
make after which they would no longer be responsible for helping others. If we accept this view, we believe that agents may never complete
their moral tasks in regard to the assistance of the needy. Rather, agents
may have to continue to help the needy as long as they are in a position
to do so.
Of course, agents responsibility for assisting the needy might be
completely fulfilled when all the needy around were being helped and
there would not be more people in need in the future. But this possibility does not match the situation of our world. The application of the
Inexhaustibility Principle thus depends on contingent conditions,
including the fact that there are always victims waiting to be saved in
poor countries; agents in affluent countries have information about
12
In describing the various interpretations of the Aggregate View, Cullity has noted
in passing the oddness of this implication. See Moral Demands of Affluence, p. 84.
378
william sin
them; humanitarian agencies are willing to save lives via private funding, etc.
The Inexhaustible Nature of Positive Duties is especially hard to
reject when agents can save the next victim at a very small cost to
themselves. Suppose Joe has an average income in an affluent country
and he has aggregately contributed 5000 to Oxfam, which is indeed a
non-trivial sum. Now, if Joe can save the next victim only by giving
another 500, he may have reason not to help. However, if Joe can save
an additional victim by sacrificing one more penny, it seems wrong if he
does not do it, and this is so regardless of how many starving children
Joe has saved through his prior contribution to humanitarian agencies.13 In this case, the great aggregate sacrifice that Joe has made does
not provide him with any special prerogative against the claims of the
needy.
Consider this case. Suppose that Joe has discovered a magic button.
Every time he presses the button, he saves one life from starvation.
Suppose further that this button is portable and it will be effective only
if Joe does not press it again within every minute. So Joe can carry the
button around when he is doing other things, too.
In this case, no matter how many times Joe has pressed the button
before, the level of his aggregate sacrifice still provides no reason for him
to stop pressing the button for good. Or, no matter whether Jack has
donated 30 per cent or 80 per cent of his income to relief organizations
to save lives, he still has the same reason to press the button to save one
more life. What it shows, I think, is that when the Principle of Easy
Rescue is coupled with the Inexhaustible Nature of Positive Duties,
Super Extremism follows. When agents can save victims lives at no cost
to themselves, they may have to do it indefinitely. In the case of Joe he
may have to keep pressing the button until he physically cannot do so.
We are thus faced with a trilemma. First, the Principle of Easy Rescue
is compelling. It is hard to imagine that agents may not be responsible
for saving someones life if they can do it at trivial costs to themselves.
Second, positive duty seems to have an inexhaustible nature. It is not
something that can be fully completed. Agents will remain connected
with those who are in need insofar as they are in a position to provide
help. Third, it is hard to abandon the non-extreme nature of morality.
In the situation of the magic button, Joe faces a super-extreme demand.
Cf. Unger, Living High and Letting Die, pp. 5961.
13
379
He does not have the prerogative to stop helping others even after he
has contributed a great deal. This goes way beyond what people generally think morality may permissibly impose on agents.
2. The Arguments
Super Extremism is hard to accept. However, compared with the other
alternatives that are open to us, it seems that Super Extremism may be
the least implausible. My view can be expressed in this argument (SE):
1. We cannot reject the Principle of Easy Rescue.
2. We cannot reject the Inexhaustible Nature of Positive Duties.
3. The Synthesis: Taken together, the Principle of Easy Rescue and the
Inexhaustible Nature of Positive Duties imply Super Extremism.
Therefore, we cannot reject Super Extremism.
Some may try to resolve the Trilemma by rejecting the Inexhaustible

Nature of Positive Duties. Holders of different versions of the Aggregate
View might take this line of argument. They think that when agents
aggregate sacrifices have reached a certain limit, agents are no longer
responsible for rescuing the next victim. People who hold this view do
not reject the Principle of Easy Rescue; they only believe that there are
circumstances under which this principle ceases to apply. For, when
the costs of saving lives have become significant, it is no longer a case
of easy rescue.
I explained in the last section why this position is not plausible. My
worry is that the Aggregate View may give agents a prerogative to
refuse all future claims of assistance from the potential needy. This is
counterintuitive, especially when what it takes for agents to save the
next victim can be very trivial, like the effort involved in the pressing of
a button.
A second way of trying to resolve the Trilemma comes also from
holders of the Aggregate View. They may have doubts about the soundness of the Synthesis: Agents might accept the Principle of Easy Rescue
and the Inexhaustibility Principle without accepting Super Extremism.
But how could this be possible?
Note that the Inexhaustibility Principle imposes no requirement on
agents frequency of compliance. So, agents may comply with the
Inexhaustibility Principle by making periodic sacrifices to save lives
with larger intervals of time; for example, agents, like Joe, may press the
380
william sin
magic button once a week, or even just once a month. These agents have
not terminated their mission of saving lives, but their selected scheme
of contribution does allow sufficient room for them to live their own
lives. Since such a reaction to the case of famine relief does not involve
great sacrifices, the Synthesis in argument SE can be rejected.
This response from holders of the Aggregate View contends the
point that agents may follow the Inexhaustibility Principle without
having to meet burdensome demands. Their case might be supported
by the fact that agents in general comply with certain other persistent
demands too, such as the duty not to lie, not to steal and the duty not
to perform murder, etc., To fulfill these demands, agents may also need
to make an indefinite amount of sacrifices, big or small, from time to
time, but agents compliance in this aspect will not amount to SuperExtremism at all. However, for two reasons, the comparison is misleading. First, the persistent demands of the negative duties we
mentioned above have been accommodated into agents lives. Most
agents have accepted them as the background circumstance under
which they develop their life plans. So the demands and agents life
plans will not systematically conflict with one another. But with regard
to the case of magic button we discuss here, the demand emerges only
later on in agents lives. It is hard for agents to find room to accommodate the demands within the established order of their lives.
The second reason why the comparison is misleading is that in the
case of the magic button, if agents are allowed to comply with the
Inexhaustibility Principle only periodically, these agents can be acting
in contradiction to the prescription of the principle itself. Suppose Joe
chooses to press the magic button once a week. On the Aggregate View,
if Joe has just pressed the button, he has thereby obtained free time of
a whole week to deal with his own business without having to save
additional people he might save at trivial costs to himself. Most of us do
not think, and would not be convinced, that there can be such moral
free time (with respect to saving lives). Such a scheme of contribution
by periodic compliance does not differ in substance from the original
proposal of the holders of the Aggregate View; that is, agents may make
a great aggregate sacrifice and then become exempted from all future
obligations to help needy, including that of easy rescue. In both cases,
we allow agents to exit, temporally or permanently, from the obligation
to honour their positive duties. This is a permission that lets victims die
too easily.
381
The third way to try to resolve the Trilemma is by objecting to the

Principle of Easy Rescue. The following may be an argument for that
purpose:
1. Super Extremism is objectionable.
2.The Synthesis: The Principle of Easy Rescue and the Inexhaustible
Nature of Positive Duties imply Super Extremism.
3. We cannot reject the Inexhaustible Nature of Positive Duties.
Therefore, the Principle of Easy Rescue is doubtful.
What could be the problem with the Principle of Easy Rescue? Perhaps
we could doubt whether human actions should always be guided by
moral principles. For instance, features of the situation can vary insofar
as an agent has saved a lot of lives. There may be increasingly strong pro
tanto reasons for the agent to care for her own well-being once she has,
say, pressed the magic button a great number of times. This may justify
a different response when the agent comes upon the next victim.
My objection is that relevant features of the situation may not, however, change drastically even if agents have performed a great number
of easy rescues. As long as it takes very little effort to press the button,
and the performance of this action saves some starving victims life, the
reason for the agent to continue pressing the button overwhelms any
reason she might have for leaving the button for any substantial period
of time.14 (She can permissibly take breaks as long as these will enhance
her ability to keep on pressing the button in the future.) So, Moderates
cannot reasonably discard the Principle of Easy Rescue.
Here is another objection to the Principle of Easy Rescue: It is not
clear that even if agents can save peoples lives or alleviate someones
dire plight by easy rescue, they should always do it. For, if the good that
is generated is not great and persistent enough, it may not be worthwhile for agents to devote their efforts to the mission time and again.
Imagine a case in which an agent is left on a deserted island with a
dying patient, whose life can only be maintained by the operation of a
respirator. And it happens that this agent can make the respirator work
by peddling a bicycle every hour for ten to fifteen minutes to recharge
14
Brad Hooker, Intuitions and Moral Theorizing, in P. Stratton-Lake (ed.), Ethical
Intuitionism (Oxford: Clarendon Press, 2002), pp. 16183.
382
william sin
the machines battery.15 Now, if we focus on what the agent can do to

lengthen the patients life in the next hour, it involves only a small effort.
But in the long run, even if the supply of food is not a problem and even
if this respirator can work continuously in the middle of the night, as
the agent has to peddle intermittently every hour throughout the day, it
seems morally permissible for the agent to let the patient die. The
Principle of Easy Rescue does not provide the right guidance in this
case.
I think this objection forces us to clarify the Principle of Easy Rescue.
For it does not seem to be a substantial good if our effort only prolongs
someones life for a short while, with an unchangeably grim prospect.
Perhaps the Principle of Easy Rescue should be stated as if an agent
can relieve some victims dire plight greatly by making a small sacrifice,
it is wrong for her not to do so. Sometimes there is a vague boundary
between a great contribution and a small contribution to someones
well-being. But whereas to prolong a dying patients life for another
hour may not be a substantial good, it will be a substantial good if the
life can be prolonged for a week or so.
Here, the argument concerning the potential ineffectiveness of easy
rescue may be used by people who want to excuse themselves from
performing even acts of easy rescue. They might argue that the scale of
the problem of famine in poor countries in Africa is so great and complicated that small contributions made by individual agents are not
going to bring any perceptible difference to the overall situation. In
short, if a whole nation is starving, then no contribution from one
average person can change this fact.16
One problem with this argument is that, even if the argument shows
the insufficiency of small donations for solving the problem of famine,
it does not show that agents are therefore exempted from making contributions at all. On the contrary, it may show that agents have an obligation to make a greater amount of sacrifice, or that they have to
contribute to the best long-term solutions to the victims plight.
Another problem with the argument is that the projects aiming at promoting the long-term development of the victims communities are
15
This case was suggested to me by Daniel Wikler at the 2005 Philosophy Summer
School in WuHan, China.
16
Various writers have mentioned this view. See G.A. Cohen, If Youre an
Egalitarian, How Come Youre so Rich? Journal of Ethics 4 (2000), pp. 1516; Cullity,
Moral Demands of Affluence, p. 86; Unger, Living High and Letting Die, pp. 4041.
383
different from the projects of providing emergency relief to individual

sufferers. The focus of our debate has been on how small contributions
may do great help to individual sufferers, not on resolving the background problem which causes famine.
In this paper, I considered the positions of the Minimalists, the
Moderates, the Extremists, and the Super-Extremists. Most people
think that Super-Extremism is the least plausible of the positions on
this spectrum. However, if we accept the arguments I offer in this paper,
we will see that, in fact, Super-Extremism is the least implausible one.
Perhaps there really are situations in which morality demands agents in
general to make the most extreme sacrifices. Given the difficulties in
rejecting the Principle of Easy Rescue and in rejecting the Inexhaustible
Nature of Positive Duties, Super-Extremism is a view that we should
take seriously.
Essentialism and Anti-Essentialism in Feminist

Philosophy
Alison Stone*
The heated feminist debates over essentialism of the 1980s and early
1990s have largely died away, yet they raised fundamental questions for
feminist moral and political philosophy which have still to be fully
explored. The central issue in feminist controversies over essentialism
was whether there are any shared characteristics common to all women,
which unify them as a group. Many leading feminist thinkers of the
1970s and 1980s rejected essentialism, particularly on the grounds that
universal claims about women are invariably false and effectively
normalize and privilege specific forms of femininity. However, by the
1990s it had become apparent that the rejection of essentialism problematically undercut feminist politics, by denying that women have any
shared characteristics that could motivate them to act together as a collectivity. An anti-anti-essentialist current therefore crystallized, which
sought to resuscitate some form of essentialism as a political necessity
for feminism.1 One particularly influential strand within this current
has been strategic essentialism, which defends essentialist claims just
because they are politically useful. In this article, I aim to challenge
strategic essentialism, arguing that feminist philosophy cannot avoid
enquiring into whether essentialism is true as a descriptive claim about
social reality. I will argue that, in fact, essentialism is descriptively false,
but that this need not undermine the possibility of feminist activism.
This is because we can derive an alternative basis for feminist politics
from the concept of genealogy which features importantly within
* This article is a substantially revised version of my essay, On the Genealogy of

Women: Against Essentialism in Feminist Philosophy, in Stacy Gillis, Gillian Howie
and Becky Munford (eds.), Third Wave Feminism: A Critical Assessment (London:
Palgrave, 2004). I thank Gillian Howie, Vrinda Dalmiya, and the anonymous referee
for the Journal of Moral Philosophy for their helpful comments on earlier versions of
this article.
1
I take the locution anti-anti-essentialism from Cressida Heyes, Line Drawings:
Defining Women Through Feminist Practice (Ithaca, NY: Cornell University Press,
2000), pp. 6772.
386
alison stone
some recent theoretical understandings of gender, most notably Judith

Butlers performative theory of gender.
To anticipate, I will develop my argument for a genealogical and
anti-essentialist recasting of feminist politics in the following stages.
I begin by reviewing the history of feminist debates surrounding
essentialism, identifying in these apparently highly disparate debates a
coherent history of engagement with an essentialism that carries a
relatively unified sense. My overview of these debates will trace how
anti-essentialism came to threaten feminism both as a critique of
existing society and as a politics of change. I shall then assess two
attempts by feminist thinkers to surmount the problems posed by anti-
essentialism without reverting to the idea that all women share a common social position and form of experience. These attempts are, first,
strategic essentialism and, second, Iris Marion Youngs idea that women
comprise not a unified group but an internally diverse series. Both
these attempts, I shall argue, are unsatisfactory, because they continue
tacitly to rely on a descriptive form of essentialism, even as they explicitly repudiate it. Nonetheless, Youngs rethinking of women as a series
is important in indicating that we need to overcome the problems generated by anti-essentialism by reconceiving women as a specifically
non-unified type of social group. Building on this point, I shall argue
that feminists could fruitfully reconceive women as a particular type of
non-unified group: a group that exists in virtue of having a genealogy.
The concept of genealogy, as I understand it, provides a way to reject
essentialism (and so to deny that women have any necessary or common characteristics), while preserving the idea that women form a distinctive social group.
My project of reconceiving women as having a genealogy is loosely
derived from Judith Butler, whose declared aim in Gender Trouble is to
outline a feminist genealogy of the category of women.2 By briefly tracing out the Nietzschean background to recent feminist appropriations
of the concept of genealogy, I will suggest that women always become
women by reworking pre-established cultural interpretations of femininity, so that they become locatedtogether with all other women
within a history of overlapping chains of interpretation.Althoughwomen
do not share any common understanding or experience of femininity,
they nevertheless belong to a distinctive social group in virtue of being
2
Judith Butler, Gender Trouble (London: Routledge, 1990), p. 5; italics original.
essentialism and anti-essentialism in feminist philosophy 387

situated within this complex history. This rethinking of women as having a genealogy entails a concomitant rethinking of feminist politics as
coalitional rather than unified. According to this rethinking, collective
feminist activities need not be predicated on any shared set of feminine
concerns; rather, they may arise from overlaps and indirect connections between womens diverse historical and cultural situations. I hope
that my exploration will begin to show how a genealogical rethinking
of women could enable feminists to oppose (descriptive) essentialism
while retaining belief in women as a group with a distinctive, and distinctively oppressive, historyan ongoing history which is an appropriate target of social critique and political transformation.
Feminist Debates around Essentialism
The first step towards any defence of an anti-essentialist, genealogical
perspective within feminist philosophy is to recall what was centrally at
issue in the controversies over essentialism that dominated much 1980s
and 1990s feminist writing. Identifying any central themes within feminist discussion of essentialism is complicated, however, as this discussion contains a bewildering variety of strands. Given this variety, the
notion of essentialism itself has taken on a correspondingly wide range
of meanings for feminists, leading some commentators, such as Gayatri
Spivak, to conclude that essentialism is a loose tongue.3 Reviewing the
huge body of literature on this question, Cressida Heyes has highlighted
four different senses of essentialism, all regularly criticized within feminist discussion: (1) metaphysical essentialism, the belief in realessences
(of the sexes) which exist independently of social construction; (2) biological essentialism, the belief in real essences which are biological in
character; (3) linguistic essentialism, the belief that the term woman
has a fixed and invariant meaning; and (4) methodological essentialism, which encompasses approaches to studying womens (or mens)
lives which presuppose the applicability of gender as a general category
of social analysis. Heyes suggests that the first two forms of essentialism
[which are] premised on metaphysical realist claims about pre-social
truths have been marginalized within the typology of essentialism,
3
Gayatri Spivak, In a Word: Interview with Ellen Rooney, in N. Schor and E. Weed
(eds.), The Essential Difference (Bloomington, IN: Indiana University Press, 1994),
p. 159.
388
alison stone
and that feminists have most regularly addressed and opposed methodological essentialism.4
Heyess typology helpfully clarifies the range of possible varieties of
essentialism, but, because she introduces precise distinctions into the
essentialism which feminists have generally discussed and criticized as
a relatively unified phenomenon, her account obscures how feminist
debates around essentialism have actually developed. Despite the variety of strands within these debates, retrospectively they can be seen to
be engaged with an essentialism that has a relatively unitary meaning,
deriving from the traditional philosophical understanding of essentialism. This relatively unitary sense of essentialism gives feminist debates
a coherent history, within which different contributions can be recognized to interweave with and build upon one another. To support this
assertion, I shall briefly reconstruct this history, starting from the philosophical sense of essentialism which forms the point of departure for
feminist explorations.
Philosophically, essentialism is the belief that things have essential
properties, properties that are necessary to those things being what
they are. Recontextualized within feminism, essentialism becomes the
view that there are properties essential to women, in that any woman
must necessarily have those properties to be a woman at all. So defined,
essentialism entails a closely related view, universalism: that there are
some properties shared by, or common to, all womensince without
those properties they could not be women in the first place. Essential
properties, then, are also universal. Essentialism as generally debated
in feminist circles embraces this composite view: that there are properties essential to women and which all women (therefore) share.5
4
Heyes, Line Drawings, p. 37. Charlotte Witt argues that feminist critiques of essentialism have four similar targets: (1) the metaphysical belief that gender and sex
are core attributes of the self; (2) biological determinism; (3) the belief that the word
feminine has a fixed meaning; (4) the practice of making false generalizations about
women; see Witt, Anti-Essentialism in Feminist Theory, Philosophical Topics 23.2
(1995), pp. 32144.
5
Naomi Schor stresses that essentialism and universalism differ in This Essentialism
Which Is Not One: Coming to Grips with Irigaray, in N. Schor and E. Weed (eds.), The
Essential Difference (Bloomington, IN: Indiana University Press, 1994), p. 24. One
might, for instance, argue that there are certain characteristics which all women share,
but which are accidental rather than essential (that is, these features could be changed
without women thereby ceasing to be women)such as, perhaps, the feature of being
disempowered relative to men. Yet, though universal features need not be essential,
essential properties are necessarily universal, hence essentialism and universalism are
generally assimilated in feminist discussion.

It is notable that, on this definition of the essentialism with which
feminists have been concerned, the properties that are essential and
universal to all women can be either natural or socially constructed.
This is reflected in that critics of essentialism from the later 1980s and
1990s typically attack any view that ascribes necessary and common
characteristics to all women, even if that view identifies those characteristics as culturally constructed. Equally, though, it must be acknowledged that feminist thinkers often use essentialism and biological
essentialism as interchangeable terms (apparently precluding the possibility that essential characteristics of women could also be cultural).
There is an obvious reason for this elision: if there are properties necessary to and shared by all women, these properties, qua necessary, can
most readily be identified as natural. Thus, essentialism easily slides
into biological essentialism because womens necessary properties are
most readily identified as biological.
Such simple, biological essentialism was commonly held prior to
secondwave feminism, typically as the view that all women are constituted as women by their possession of wombs, breasts and child-
bearing capacity. Arguably, this view played a crucial ideological role in
justifying womens confinement to the domestic sphere as natural and
necessary. Second-wave feminists therefore opposed essentialism in its
pre-feminist, biological incarnation. However, feminist antipathy to
essentialism rapidly extended to elements of biological essentialism
perceived to persist within feminism. In the 1970s, socialist feminists
criticized the essentialism they detected in the work of some radical
feminists who urged revaluation of womens allegedly natural features,
such as their child-bearing capacity.6 Within these socialist feminist
critiques, (biological) essentialism was typically contrasted to social
constructionism, which relies on the distinction between biological sex
and social gender. On the social constructionist view, sexed biology is
both different from, and causally inert with respect to, genderan
individuals socially acquired role and sense of identity. So, while being
female may require certain anatomical features, being a woman is
something different, dependent on identification with the feminine
genderthe social traits, activities and roles that make up femininity.
Following this recognition of the gap between gender and sex, social
6
For a classic socialist feminist critique of radical feminism, see Lynne Segal, Is the
Future Female? (London: Virago, 1987).
390
alison stone
constructionists could reject biological essentialism for confusing these

two levels of analysis and consequently making a fallacious and ideologically motivatedattempt to read off the contingencies of social
arrangements from the necessities of biology.
Despite repudiating biological essentialism, many influential feminist theorists of the 1970s and early 1980s went on to endorse nonbiological forms of essentialism. Having identified femininity as
socially constructed, these theorists sought to identify an invariant set
of social characteristics that constitute femininity and that all women,
qua women, share. Possibilities included womens special responsibility
for domestic, affective or nurturant labour (as Nancy Hartsock argued),
their construction as sexual objects rather than sexual subjects (according to Catherine MacKinnon) or their relational, contextual and particularist style of ethical and practical reasoning (Carol Gilligan).7 My
claim that theorists such as MacKinnon are essentialists might sound
odd, given the frequent contrast between essentialism and social constructionism. Yet social constructionists can readily be essentialists if
they believeas do these influential feminist theoriststhat a particular pattern of social construction is essential and universal to all women.
Moreover, in the later 1980s, a large number of feminist thinkers
began to attack the positions of Gilligan, MacKinnon and others as
preciselyessentialist. These critics argued, in considerable detail, that
universal claims about womens social position or identity are i nvariably
false. It cannot plausibly be maintained that womens experiences have
any common character, or that women share any common location in
social and cultural relations, or sense of psychic identity. Essentialism,
then, is simply false as a description of social reality. Moreover, critics
pointed out that the descriptive falsity of essentialism renders it politically oppressive as well. The (false) universalization of claims about
women in effect casts particular forms of feminine experience as the
norm, and, typically, it is historically and culturally privileged forms of
femininity that become normalized in this way. Essentialist theoretical
7
See, respectively, Nancy Hartsock, Money, Sex, and Power: Toward a Feminist
Historical Materialism (New York: Longman, 1983); Catherine MacKinnon, Feminism,
Marxism, Method and the State: An Agenda for Theory, Signs 7 (1982), pp. 51544;
Carol Gilligan, In a Different Voice (Cambridge, MA: Harvard University Press, 1982).
Gilligan has subsequently revised her ethical theory to mitigate the exclusive tendencies that critics detected in it: see J.M. Taylor, Carol Gilligan and A.M. Sullivan, Between
Voice and Silence: Women and Girls, Race and Relationship (Cambridge, MA: Harvard
University Press, 1996).

moves thereby end up replicating between women the very patterns of
oppression and exclusion that feminism should contest. This point has
been pressed particularly forcefully by Elizabeth Spelman, whose classic critique Inessential Woman castigates recurring tendencies within
feminism to take certain privileged womens experiences or situations
as the norm.8
One might wonder whether we could defend essentialism without
postulating any social or cultural characteristics common to all women
if we instead identified womens essential properties with their biologically female characteristics. This need not entail returning to the traditional, misleadingly anatomical, definition of womanhood: one might
hold that femininity is socially constructed in diverse ways, but that all
these constructions are united in that they build upon and interact
with individuals biologically female characteristics. This option was
foreclosed by feminist philosophies of embodiment which developed
in the 1990s. Judith Butler, Moira Gatens and Elizabeth Grosz, in particular, argued that bodies are thoroughly acculturated, and therefore
participate in the same diversity as the social field that they reflect.9
These thinkers argued that our bodies are first and foremost the bodies
that we live, phenomenologically, and the way we live our bodies is
culturally informed and constrained at every point. Sexed embodiment is therefore not external but internal to the gendered realm of
social practices and meanings. Consequently, one cannot appeal to any
unity amongst female bodies to fix the definition of women, since the
meaning of bodies will vary indefinitely according to their sociocultural location.
Following this recognition of the cultural character of bodies, a
growing number of theorists in the 1990s rejected the previously popular essentialism/ constructionism antithesis. They argued that constructionism remains unduly close to essentialism, since it accepts the
existence of natural bodily properties but simply denies them any role
in constituting the essence of woman. According to these critics, constructionism remains problematic because, in retaining the belief in
natural properties of female bodies, it leaves permanently open the
8
Elizabeth Spelman, Inessential Woman: Problems of Exclusion in Feminist Thought
(London: The Womens Press, 1988).
9
See Butler, Gender Trouble; Moira Gatens, Imaginary Bodies (London: Routledge,
1996); Elizabeth Grosz, Volatile Bodies: Toward a Corporeal Feminism (Bloomington:
Indiana University Press, 1994).
392
alison stone
possibility of making a (spurious) appeal to these properties in the

attempt to ground unity amongst women. The most consistent form of
anti-essentialism, then, which developed in the 1990s, denies that any
featuresnatural or socialare common to all women, who are fully
socially and corporeally diverse.
The increasingly radical rejection of essentialism prompted a countertendency within feminist thought, however, emphasizing the
neglected importance and political potential of essentialism. Feminists
became increasingly concerned that accusations of essentialism often
silenced thinkers, condemning their arguments out of hand. Naomi
Schor, famously, complained that these accusations had become the
prime idiom of intellectual terrorism and the privileged instrument of
political orthodoxyendowedwith the power to reduce to silence,
to excommunicate, to consign to oblivion.10 This prompted a reconsideration of whether essentialism might be philosophically or politically
fruitful. Notably, feminist rejection of essentialism had posed several
interwoven problems. First, it had cast doubt on the project of conceptualizing women as a group.11 By denying women any shared features,
anti-essentialism seemed to imply that there is nothing in virtue of
which women could rightly be identified as forming a distinct social
group. This undermined feminism as a critique of existing society,
insofar as this critique is premised on the claim that women constitute
a distinctly disadvantaged or oppressed social group. Anti-essentialism
appeared also to have undermined feminist politics: if women do not
share any common social location, then they cannot be expected to
mobilize around any concern at their common situation, or around any
shared political identity or allegiance. Thus, anti-essentialism seemed
to undermine feminism both as social critique and as a political movement for social change.
Faced with these problems, anti-anti-essentialists reconsidered how
far some form of essentialism might be necessary for feminist social
criticism and political activism. One of the most important strands in
this reconsideration has been strategic essentialism: the defence of
essentialism not as a descriptive claim about social reality, but merely
as a political strategy. In the next section, I will argue that strategicessen
tialism is unstable: although it attempts to avoid endorsing essentialism
Schor, This Essentialism Which Is Not One, p. 42.
Iris Marion Young, Gender as Seriality: Thinking about Women as a Social
Collective, Signs 19.3 (1994), p. 713.
10
11

as a description of social reality, it ultimately remains forced to rely on
descriptive essentialism to support its claim to political efficacy. I will
then assess how a similar instability infiltrates Iris Youngs suggestive
attempt to reconceive women as a series, an attempt which nonetheless
paves the way for my subsequent argument that women might productively be reconceived as having a genealogy.
Strategic and Descriptive Essentialism
Despite their concern to reappraise the political fertility of essential
ism, few anti-anti-essentialists have sought to reinstate the belief in
shared social characteristics common to all women. Rather, anti-antiessentialists have tended to defend essentialism by arguing that it can
take multiple forms, some more complex and subtleand defensible
than its familiar ones.12 In particular, it has been argued that essentialism need not take the form of a descriptive claim about social reality.
According to strategic essentialism, which became increasingly popular in the later 1980s and 1990s, feminists should acknowledge that
essentialism is descriptively false in that it denies the real diversity of
womens lives and social situations.13 Nonetheless, in delimited contexts,
feminists should continue to act as if essentialism were true, so as to
encourage a shared identification among women that enables them to
engage in collective action. To take a controversial example, many of
the bold statements in Luce Irigarays later work have often been construed as strategically essentialist. In Thinking the Difference, she claims
that women share certain bodily rhythms which give them a deep
attunement to nature, and which mean that women are particularly
adversely affected by ecological disasters such as the Chernobyl accident.14 It seems plausible to think that, rather than attempting to
describe women as they really are, Irigaray is encouraging women
to think that they suffer particularly from environmental problems,
See, especially, Schor, This Essentialism Which Is Not One, p. 43.

Strategic essentialism is, of course, primarily associated with Gayatri Spivak,
who coined the term in Feminism, Criticism and the Institution, Thesis Eleven 10/11
(198485), pp. 17587. Spivak, however, introduced strategic essentialism in relation
to subaltern studies: her term was taken up within feminist contexts in ways she did
not intend.
14
Luce Irigaray, Thinking the Difference: For a Peaceful Revolution (trans. K. Montin;
London: Athlone Press, 1994), pp. 2426.
12
13
394
alison stone
as a strategic identification that will galvanize them to collectively resist

ecological degradation.
An objection immediately arises to this strategic essentialist position. Any political strategy is effective only inasmuch as it allows agents
to recognize and intervene into the real social events, processes and
forces which make up the social field. But it seems reasonable to think
that a strategy can be effective, in this sense, only insofar as it embodies
an accurate understanding of the character of social processes. This
implies that a strategy of affirming fictitious commonalities among
women will fail to facilitate effective action given a world where women
do not really have any common social characteristics or locations.
Rather, such a strategy appears destined to mislead women into fighting against difficulties which are either non-existent or more likely
really affect only some privileged subgroup of women.
This objection can be resisted, however, as it (implicitly) is by Denise
Riley in Am I That Name?. Riley claims that it is compatible to suggest
that women dont existwhile maintaining a politics of as if they
existed since the world behaves as if they unambiguously did.15 In
other words, for Riley, the fiction that women share a common social
experience is politically effective because the social world actually does
treat women as if they comprise a unitary group. Riley accepts that
women are not a unitary group and that the socially prevalent idea that
they are unified is false. Nevertheless, this false idea informs and organizes the practices and institutions that shape womens experiences, so
that thosevery differentexperiences become structured by essentialist assumptions. A strategy of affirming fictitious commonalities
therefore will be effective given this world in which (false) descriptive
essentialist assumptions undergird womens social existence.
Rileys argument has a problem, though: she cannot consistently
maintain both that womens social experience is fully diverse and that
this experience is uniformly structured by essentialist assumptions. If
essentialism informs and organizes the structures that shape womens
social experience, then this experience will be organized according to
certain shared models and will acquire certain common patterns and
features. More concretely, the idea that women are a homogeneous
group will structure social institutions so that they position all women
15
Denise Riley, Am I That Name?: Feminism and the Category of Women in History
(London: Macmillan, 1982), p. 112.

homogeneously, leading to (at least considerable areas of) shared experience. Thus, Riley (and other strategic essentialists) may be right that
essentialist constructions are socially influential, but they cannot, consistently with this, also maintain that descriptive essentialism is false.
Furthermore, it is not obviously true that any uniform set of essentialist
constructions informs all social experience. These constructions may
all identify women as a homogeneous group, but they vary widely in
their account of the context of womens homogenous features. Conse
quently, these constructions will influence social structures in correspondingly varying directions, against which no counter-affirmation of
common experience can be expected to be effective.
Strategic essentialists, then, have attempted to resuscitate essentialism by arguing that it can take a merely political and non-descriptive
form. But this attempt proves unsuccessful, because one cannot defend
essentialism on strategic grounds without first showing that there is a
homogeneous set of essentialist assumptions that exerts a coherent
influence on womens social experiencewhich amounts to defending
essentialism on descriptive grounds (as well). Advocates of essentialism therefore need to show that it accurately describes social reality.
Here, though, critics can retort that essentialism is descriptively false,
since women do not even share any common mode of construction by
essentialist discourses. Yet this retort reinstates the problem of antiessentialism: its paralysing effect on social criticism and political activism. Strategic essentialism has not resolved this problem, for it has not
stably demarcated any merely political form of essentialism from the
descriptive essentialism which critics have plausibly condemned as
false and oppressive.
Women as a Series
To resolve the problems posed by anti-essentialism, feminist philosophers need to oppose essentialism as a descriptive claimthat is, to
recognize the diversity of womens lives and social characteristicsand
yet to continue to identify women as a distinctive (and distinctly disadvantaged) social group. This conjunction of anti-essentialism with
feminist social ontology appears difficult to achieve, but could be
accomplished if we reconceive women as a social group of some specifically non-unified type. Iris Marion Young takes this step in her paper
Gender as Seriality.
396
alison stone
Young urges us to reconceive women as a series, where a series is a

kind of group that is non-unified: vast, multifaceted, layered, complex
and overlapping.16 Employing the terminology of Jean-Paul Sartres
Critique of Dialectical Reason, Young distinguishes series from groups
in the strict sense: the latter are collections of individuals who mutually
recognize significant areas of shared experience and orientation to
common goals. In contrast, membership in a series does not require
members to share any attributes, goals or experience. Instead, the
members of a series are unified, passively, through their actions being
constrained and organized by particular structures and constellations
of material objects. Women, for example, are passively positioned in a
series by the particular cluster of gender rules and codes which infuse
every-day representations, artefacts and spaces. Youngs understanding
of women as assembled into a series allows her to deny women any
common identity or characteristics, by arguing that they take up the
constraints of gender structures in variable ways, within the contexts of
entirely different projects and experiences. At the same time, Young
can consistently claim that women retain the broad group status
of a series insofar as the same set of feminizing structures remains a
background constraint operative upon them all. Having secured
women the status of a determinate social groupin this broad, nonunified senseYoung concludes that it is possible for women to
become conscious of their group status and so to become motivated
into co-operating together politically.
Unfortunately, Youngs approach has a drawback which is structurally similar to that of strategic essentialism: her defence of womens
group status tacitly reinscribes the descriptive essentialism from which
she explicitly distances herself. Although she denies that women share
a common experience or identity, she does maintain that all womens
activities and lives are oriented around the same or similarly structured
objects [and]realities.17 From Youngs perspective, there must be
some features that unify these social structures and realities such that
they can be said to co-operate in constituting women as a single,
distinct gender. As Young says, it is from the sameness of the objects
structuring womens activities that the loose unity of the series
derives.18 Although, as she admits, the content of these objects and
Young, Gender as Seriality, p. 728.
Young, Gender as Seriality, p. 728; emphasis added.
18
16
17

realities varies contextually, she still maintains that, despite their
diversity, these realities share certain unifying characteristics. In particular, for Young, they all embody the expectation, first, of normative
heterosexuality which, she claims, constitute[s] women crossculturally19and, secondly, of a sexual division of labourthe content
of which, Young observes, varies with each social system, [although in
each case] a division of at least some tasks and activities by sex appears
as a felt necessity.20 Of course, she also insists that women experience
these expectations differently given their diverse situations and activities. Yet the claim that these expectations organize the realities of all
womens lives is itself ambiguous. Insofar as all womens lives are organized by a sexual division of labour, the content of this division varies
widely, as Young herself admits. Likewise, the meaning of heterosexuality is highly varying (although, as a minimum, normative heterosexuality expects women to desire men, the meaning of this expectation
will vary greatly relative to changing conceptions of masculinity and of
desire, sexuality, and their social significance). So, Youngs claim is
plausible only if it acknowledges that the expectations that ultimately
organize all womens lives are themselves varied; but, consequently,
these expectations cannot be said to unify the structures by which
women are serially positioned.
Young, however, needs to identify this unity because she can only
retain a coherent feminine gender by arguing that, although women
have no common features, there are common featurescommon
expectationsorganizing all the social realities that constrain womens
lives. Hence she has to maintain, for example, that normative heterosexuality has a universal, cross-cultural meaning. This brings Young
into a difficulty, parallel to that of the strategic essentialistsshe continues, ultimately, to rely on a descriptive form of essentialism, insofar
as she has to affirm that there are certain universal norms that constitute all women as women (even though women do not share a common
experience of those norms). Youngs residual essentialism can be traced
back to the Sartrean framework from which she derives her concept of
a series. Sartre insists that series should not be equated with groups in
the strict, unified sense. Groups in the strict sense involve shared goals
and experience, so that series count as groups only in a broad or, as

19
20
398
alison stone
Sartre puts it, neutral sense.21 Yet the reason why Sartre continues to
identify series as groups in this broad sense is because he regards series
as self-alienated versions of strict, unified groups. For Sartre, series are
self-alienated groups in that their unity is located outside them, in the
objective artefacts or structures by which their members are constrained and organized.22 Thus, from Sartres perspective, series only
count as groups at all to the extent that they remain unified, albeit in a
relatively indirect, exteriorized way.
Youngs Sartrean concept of the series proves inadequate to the task
which she wishes it to play: that of enabling her to reconceive women
as a specifically non-unified type of social group. Despite this failing,
the importance of Youngs argument lies in her more basic insight that
reconceiving women as a non-unified type of social group could surmount the problems generated by feminist critiques of essentialism.
I therefore propose to develop Youngs insight, by jettisoning the idea
that women constitute a series, and instead rethinking women as having a genealogythat is, as constituting a group that is internally
diverse, and yet remains a group in virtue of having a complex history
composed of multiple, overlapping threads of interpretation.
Women as Having a Genealogy
In this section, I will argue that an appropriation of the concept
of genealogy can provide a way to reinstate the idea that women comprise a distinct social group even in the absence of any common properties that constitute them all as women. Several prominent feminist
thinkers have already drawn on the concept of genealogy: for example,
in Gender Trouble, Judith Butler proposes to outline a genealogical
understanding of what it means to be a woman.23 Similarly, Moira
Gatens praises the project of charting a genealogy of the category
woman or women. On this approach women itself is under
stoodto have a history, a genealogy, a line of descenta genealogical
approach asks: how has woman/women functioned as a discur
sive category throughout history?.24 These approving references to
21
Sartre, Critique of Dialectical Reason. I. Theory of Practical Ensembles (trans.
A. SheridanSmith; London: New Left Books, 1976), p. 256.
22
Sartre, Critique of Dialectical Reason, pp. 25859.
23
Butler, Gender Trouble, p. 5.
24
Gatens, Imaginary Bodies, p. 76.

genealogy by Butler and Gatens imply that ideas of femininity should
be understood as historically constructed in multiple, shifting ways,
their fluctuations in meaning registering changes in social relations of
power. In this, though, Butler and Gatens appear concerned to trace
the history of the concept woman rather than the history of women
themselves. Yet the concept of genealogy can only provide a way of
grasping women as a distinct (albeit non-unified) social groupifwomen
themselves have a genealogythat is, to anticipate, if their experiences
and psychologies are shaped in overlapping and historically interconnected ways. Thus, any anti-essentialist appropriation of the concept of
genealogy must clarify the relationship between the genealogy of ideas
concerning femininity and the genealogy of women themselves.
To do this we can return to Nietzsche, from whom both Butler and
Gatens derive their concepts of genealogy. Gatens draws deliberately
upon Nietzsche,25 and, although Butler draws more explicitly upon
Foucault than Nietzsche, she herself repeatedly stresses that Foucaults
practice of genealogical enquiry is deeply indebted to his interpretation
of the concept of genealogy in Nietzsche.26 In On the Genealogy of
Morality (1887),27 Nietzsche sketches a distinctive form of historical
enquiry which traces how historically changing conceptssuch as
guilt, duty, community, good and evil shape lived social experience. He also traces how the power relations that are at work within
peoples social experiences lead them to reshape those concepts in turn.
In explicating Nietzsches approach, I shall draw on Foucaults presentation of it in his important methodological essay Nietzsche, Genealogy,
History (1971).28
In On the Genealogy of Morality, Nietzsche denies that any common
characteristics unite all the institutions, practices and beliefs normally
classified under the rubric of morality. Nietzsche thus adopts an antiessentialist approach to morality, taking its constituent practices and
Gatens, Imaginary Bodies, pp. 7677.
See, especially, Judith Butler, Subjects of Desire: Hegelian Reflections in TwentiethCentury France (New York: Columbia University Press, 1987), pp. 22628; Gender
Trouble, pp. 2021, 150, 161.
27
Nietzsche, On the Genealogy of Morality (trans. Carol Diethe; Cambridge:
Cambridge University Press).
28
Foucault, Nietzsche, Genealogy, History, reprinted in J. Richardson and B. Leiter
(eds.), Nietzsche (Oxford: Oxford University Press, 2001), pp. 34159. On the importance of this paper for Foucaults subsequent work, see, inter alia, Michael Mahon,
Foucaults Nietzschean Genealogy: Truth, Power, and the Subject (Albany: SUNY Press,
1992).
25
26
400
alison stone
beliefs to be highly diverse,29 and to fall under the single rubric of

morality solely in virtue of belonging within a distinctive history.
Foucault stresses that, for Nietzsche, this history is not to be studied
through traditional modes of historical enquiry, which misleadingly
presuppose an underlying unity among moral phenomena. Instead, the
history of morality should be approached through a novel mode of
enquirygenealogywhich attends to the fluctuating and internally
heterogeneous character of its object of study.30 In particular, the genealogist traces how some contemporary practice (for example, punishment) or experience (for example, guilt) has arisen from an indefinitely
extended process whereby earlier forms of that practice or experience
have become reinterpreted by later ones. Thus, the genealogist treatsany
historically arisen phenomenon as the reinterpretation of a pre-existing
phenomenon, upon which the new interpretation has impressed
its own idea of a use function.31 A genealogy takes shape as a practice
or experience becomes subjected to repeated reinterpretations which
impact upon its meaning [Sinn], purpose and expectation.32
According to Nietzsche, any reinterpretation must install itself by
accommodating, as far as possible, the meanings embedded in the preexisting phenomenon,33 though necessarily it sheds the elements of
those meanings that remain incompatible with its own agenda. This
makes reinterpretation a conflictual, agonistic process, in which present forces strive actively to take over recalcitrant elements of the past.
The outcomes of these conflictual activities of reinterpretation are
always variable and contingent, as Foucault particularly stresses.34
A key point, though, is that any practice or experience that succumbs
to reinterpretation has itself already taken shape as the sedimentation
of earlier layers of interpretation. No common core of significance
29
See Raymond Geuss, Nietzsche and Morality, in Raymond Geuss (ed.), Morality,
Culture and History: Essays on German Philosophy (Cambridge: Cambridge University
Press, 1999), pp. 16769.
30
Foucault, Nietzsche, Genealogy, History, p. 346.
31
Nietzsche, On the Genealogy of Morality, p. 55.
32
33
As Geuss explains, Nietzsches conflictual model implies that no reinterpretation
will ever encounterjust a tabula rasa, but a set of actively structured forces, practices
etc. which will be capable of active resistance (Geuss, Nietzsche and Genealogy, in
Morality, Culture, and History, p. 13). This resistance ensures that new interpretations
will not in general be so fully successful that nothingremains of the pre-existing
meanings (p. 11).
34
Foucault, Nietzsche, Genealogy, History, pp. 34748.

persists through all these layers of interpretation. Because i ncompatible
elements of meaning become shed upon each occasion of reinterpretation, a process of attrition takes place through which earlier layers of
meaning gradually get eroded away altogether. For example, Nietzsche
traces how the experience of bad conscience becomes reinterpreted by
Christianityand consequently re-experienced, re-livedas sinfulness.35 In similar fashion, the earlier meanings of all the constituents of
the institution of morality are gradually and continuously being etched
out by new acts of reinterpretation.36
Nietzschean genealogy, then, traces how concepts such as guilt and
evil undergo varying interpretations, where these interpretations
and the concepts which they organizecontinually reshape our
experience and practices. To study these experiences and practices
genealogically is to situate them within a particular groupfor example, the group moralitynot because of any essential characteristics
that they share with all the groups other members, but just because
each member in the group stands in the appropriate historical relationship to (one or more of) the others. More specifically, a set of such
items is grouped together only in virtue of the fact that each takes shape
through the reinterpretation of one or more of the others. The items in
this group need not have anything in common, but need only be
connected together through a complex process of historical drift in
meaning. From a Nietzschean perspective, any set of concepts, experiences or practices that become related in this overlapping way has a
genealogy.
Nietzsches idea that any chain of historically overlapping phenomena has a genealogy makes it possible to reconceive women as a determinate social group without reverting to the descriptive essentialist
claim that all women share a common social position or mode of experience. Any such genealogical analysis of women must start by recognizing that concepts of femininity change radically over time, and that
these changing concepts affect womens social position and lived experience. In particular, a genealogical analysis of women is premised
This insight underlies Nietzsches claim that genealogists trace the descent
(Herkunft) of phenomena rather than their putative origins (Ursprnge). Because any
historical phenomenon arises gradually through the concatenation of an indefinite
multiplicity of events, there is no definitive point at which that phenomenon can be
said to have originated (see Nietzsche, On the Genealogy of Morality, pp. 47; Foucault,
Nietzsche, Genealogy, History, pp. 34243).
35
36
402
alison stone
on the viewarticulated in Judith Butlers workthat women only

become women, or acquire femininity, by taking up existing interpretations and concepts of femininity. As Butler puts it, taking on a gender
involves finding a contemporary way of organizing past and future cultural norms, a way of situating oneself in and through those norms, an
active style of living ones body in the world.37 As this remark suggests,
the taking on of femininity consists not merely of a process of mental
identification with existing concepts, butmore fundamentallyof a
process of acquiring a feminine way of living ones body, of inhabiting
ones physiology. Moreover, this does not entail being passively moulded
by exterior cultural forces; rather, women become women through
active appropriation and personalizing of inherited cultural standards.
Butler also stresses, however, that each appropriation of existing
standards concerning femininity effects a more or less subtle alteration
of their meaning: individuals always interpret received gender norms
in a way that organises them anew.38 In actively appropriating existing
standards, individuals necessarily adapt them with reference to the varied contexts, power relationships and personal histories within which
they are located. Received meanings regarding gender continuously
become subjected to practical reinterpretation, which individuals
undertake with tacit reference to their differing personal and cultural
experiences. That the meaning of femininity undergoes incessant modification implies that it is considerably less unified than one might, at
first glance, assume. There is no unitary meaning of femininity on
which all women agree: for, even although all women may identify with
femininity, they will always understand and live their femininity in
different ways. Nonetheless, according to a genealogical approach, all
women remain identifiable as women. Although they do not share any
characteristics simply qua women, in each case they become feminine
by reworking pre-established interpretations of femininity with reference to their specific situations. In virtue of carrying out this reworking, each woman becomes located within a historical chain of women,
a chain composed of all those who have successively engaged in reinterpreting the meaning of femininity. All women thus become located
within this ongoing chain of practice and reinterpretation, which
37
Judith Butler, Variations on Sex and Gender: Beauvoir, Wittig and Foucault, in
S. Benhabib and D. Cornell (eds.), Feminism as Critique (Minneapolis: University of
Minnesota Press, 1987), p. 131.
38
Butler, Variations, p. 131.

brings them into complex filiations with one another. The genealogy
that arises through this process is not only a history of concepts of femininity, but is also, simultaneously, a history of women themselves, as
individuals who become women by taking on and adapting existing
concepts of femininity.
Since genealogical reinterpretation is an agonistic process, each
reinterpretation of femininity must overlap in content with the interpretation that it modifies, preserving some elements of that pre-existing
interpretation while abandoning others. Through the resulting process
of attrition of meanings, each woman will find herself in a series of
gradually diminishing connections with women of previous generations. Moreover, within a single generation, each womans reinterpretation of femininity will overlap in content, to varying degrees, with
other womens reinterpretations; these overlaps must arise, insofar as
all these women are engaged in reworking the same set(s) of preexisting meanings. An understanding of women as having a genealogy
thus entails that, instead of forming a unitary group, they are connected
together in complex ways and to varying degrees, and, in particular,
that they are linked by their partially and multiply overlapping interpretations of femininity.
This genealogical conception of women might be criticized on
two, closely interrelated, grounds. First, one might object that it is not
ultimately very different from Youngs idea that women form a series.
Just as Young preserves womens status as a group by arguing that their
lives are shaped by a universal set of normative expectations, likewise
the genealogical view argues that women participate in a unitary
history. This history is unitary in virtue of the continuity between its
constituent phases, a continuity that arises insofar as every reinter
pretation of femininityand every re-experiencing of womanhoodbuilds upon, and so retains a partial overlap with, pre-existing
interpretations and experiences. But, just as the expectations that
Young believes to organize womens lives are more plausibly seen as
diverse than singular in content, likewise the history of femininity and
women might be most plausibly seen as discontinuous. Arguablyand
this is the second objectionthe genealogical view over-emphasizes
overlaps and continuities between different interpretations and expe
riences of femininity, ignoring the deep chasms that regularly open
up between understandings of femininitythe breaks in the chain of
(re)interpretation. This objection takes on special importance because
such discontinuities will typically reflect exactly those asymmetries of
404
alison stone
power that appear to impede the possibility of women achieving any

solidarity as a group.
Against these objections, I suggest that the strength of the genealogical approach is that it can accommodate the reality of historical discontinuity alongside that of continuity. According to this approach,
successive modifications in the meaning of femininity necessarily build
upon one another, leading to the formation of distinct historical patterns of interpretation of femininity, which branch apart from one
another in particular directions. This branching, moreover, will typically follow along differentials in power (which lead women to modify
the meaning of femininity in particular ways). As the branching occurs,
the process of attrition whereby earlier elements of meaning get worn
away will ensure that quite separate cultures of femininity emerge.
Women located within these separate culturescultures that occupy
different positions within relations of powerwill have ceased to share
any experience as women, even though they all identify themselves as
feminine. In such cases, women remain connected together only indirectlyvia the long chains of overlapping meaning and practice that
span the gulf between them. Thus, a genealogical approach itself
implies the inescapability of both continuities and discontinuities
within the history of femininity.
As a consequence of these discontinuities in the meaning of femininity, women must be considered not merely as an internally diverse
group but also, more strongly, as a group fractured and torn apart by
divisions in power. Recognizing this, for instance, Butler refers to the
incessant occurrence of rifts among women over the content of the
term (that is, woman).39 Yet, however severe this rift, it remains compatible with womens existing as a distinctive social group. This is
because women remain defined by a single history, even though this
history is extremely complex and follows multiple branches. Womens
history should be understood on the model of a tree, which remains
singular even as it continually ramifies into innumerable (multiply
interwoven) branches. Properly understood, then, the genealogical
view avoids reinstating descriptive essentialism because it holds that
what unites womentheir historyis internally complex. At the same
time, this view insists that this complex history remains singular, and
39
Judith Butler, Contingent Foundations: Feminism and the Question of Post
modernism, in S. Benhabib et al. (eds.), Feminist Contentions: A PhilosophicalExchange
(London: Routledge, 1995), p. 50.

hence does constitute women as a group whose history, and consequent social positions, can be identified as distinctively oppressive.
This rethinking of women and femininity as having a genealogy
opens up the possibility of an anti-essentialism that supports, rather
than paralyses, feminist politics. To the extent that women remain a
social group (united in their participation in a single history), they can
mobilize together in pursuit of distinctive concerns. Nonetheless, since
womens history is internally complex, and womens concerns are correspondingly diverse, the only mode of collective activity appropriate
for women must be similarly diversified allowing women to pursue
concerns that are specific to them as women, yet which differ from
one another as well. This mode of political activity must, in addition, be
capable of accommodating deep rifts and divisions among women, by
providing them with the space for difficult negotiation over their divisions. These two conditions suggest that a coalitional politics is uniquely
appropriate for women. This is, indeed, the political practice generally
endorsed by advocates of a genealogical approach.40 But what is a coalitional feminist politics?
Coalitions may be said to arise when different women, or sets of
women, decide to act together to achieve some determinate objective,
while yet acknowledging the irreducible differences between them and
the often highly divergent concerns that motivate them to pursue this
objective. On the basis of the idea that women have a genealogy, we can
explain why women might, despite these irreducible differences, reasonably seek to mobilize together on such a coalitional footing. First,
each womans historically shaped experience inevitably overlaps in
content with that of at least some other women, which gives them areas
of commonality that they might reasonably seek to transform together,
despite being very different in other respects (and so approaching these
objectives from quite disparate perspectives). Second, in each womans
case, there will be many other women with whose experience her own
has no direct overlap, and with whom she is only indirectly connected
(through the whole web of overlapping relations between women).
These might typically be women to whom she stands in a deeply asymmetrical power relationship. Nonetheless, there might be many cases in
which she could reasonably seek to act in concert with those women,
40
See, for example, Nancy Fraser and Linda Nicholson, Social Criticism Without
Philosophy: An Encounter Between Feminism and Postmodernism, in L. Nicholson
(ed.), Feminism/Postmodernism (London: Routledge, 1990), p. 35.
406
alison stone
because she could expect improvements in either of their situations to

indirectly have positive repercussions for the other. Since women
remain connected indirectly by long chains of reinterpretation of femininity, an improvement in the situation of any of the women should
impact positively upon those women whose interpretations of femininity overlap, so that, through a kind of wave effect, even the women
at furthest remove could anticipate some indirect benefit. Those women
might, at least, be benefited in the sense that any change exposes the
meaning of femininity as contingent and malleable, making it easier
to undertake transformative reinterpretations of the meanings of
femininity that have become sedimented within their own cultures.
Certainly, such potential gains would be obstructed if privileged groups
of women respond to improvements in their situations in ways that
reinforce hierarchies between women. Nonetheless, a genealogical
conception of women and femininity at least suggests ways to reflect
upon the spectrum of motivations that might lead women to enter coalitions, different motivations that correspond to womens varying
degrees of cultural overlap and connectedness.
Conclusion
I have attempted to reassess feminist debates around essentialism in a
way that brings out the philosophical, ethical and political significance
of the questions they have raised. As I have traced, these debates pose a
central, and widely acknowledged, dilemma: essentialism is plausibly
seen as false as a descriptive claim about the social reality of womens
lives, yet appears necessary to feminist politics and social criticism.
I have argued that this dilemma cannot be solved by endorsing essentialism merely as a political strategy, since essentialism can only be
defended on strategic grounds if it is held to be descriptively true as
well. Instead, the dilemma should be solved by accepting that essentialism is descriptively false, but reconceiving women as a specifically nonunified sort of social group. From this perspective, I have suggested
that we might rethink women, and femininity, as having a genealogy.
This provides a way to identify women as a definite social group without falsely attributing to them any common characteristics that constitute them all as women.
According to my argument, every woman becomes a woman by
taking over and reinterpreting pre-existing cultural constructions of

femininity, constructions which in turn exist as a result of preceding activities of reinterpretation, so that all these interpretations of
femininityand all the women who produce and experience them
come to belong within overlapping chains. These chains make up a
uniquealbeit complex and multiply branching history within
which all women are situated. Thus, although women do not share any
common characteristics, they are defined as a group by their participation in this history. This opens up various ways in which women might
become motivated to engage in collective action organized coalitionally.
I suggest, then, that the idea that women have a genealogy overcomes
the dilemma posed by feminist critiques of essentialism, explaining
howdespite their lack of common characteristicswomen can still
exist as a determinate group, susceptible to collective mobilization.
Bibliography
The Journal of Moral Philosophy originally published the essays in this
collection. This bibliography lists the full contents of all articles published since the JMP was launched in April 2004. The JMP acceptance
rate is about 7%.
Allen, Robert Francis, Robust Alternatives and Responsibility, Journal of Moral
Philosophy 1(1) (2004), pp. 2129.
Alm, David, Deontological Restrictions and the Good/Bad Asymmetry, Journal of
Moral Philosophy 6(4) (2009), pp. 46481.
Anderson, Scott A., Of Theories of Coercion, Two Axes, and the Importance of the
Coercer, Journal of Moral Philosophy 5(3) (2008), pp. 394422.
Andreou, Chrisoula, Standards, Advice, and Practical Reason, Journal of Moral
Philosophy 3(1) (2006), pp. 5767.
Bedi, Sonu, Expressive Exclusion: A Defense, Journal of Moral Philosophy 7(4) (2010),
pp. 42740.
Bennett, Christopher, Forgiveness and the Claims of Retribution, Journal of Moral
Philosophy 1(1) (2004), pp. 89101.
Bennett, Christopher, State Denunciation of Crime, Journal of Moral Philosophy 3(3)
(2006), pp. 288304.
Besser-Jones, Lorraine, Personal Integrity, Morality and Psychological Well-Being:
Justifying the Demands of Morality, Journal of Moral Philosophy 5(3) (2008),
pp. 36183.
Besson, Samantha, Democracy, Law, and Authority, Journal of Moral Philosophy 2(1)
(2005), pp. 8999.
Bittner, Rdiger, A Horse in the Basement: Nietzschean Reflections on Political
Philosophy, Journal of Moral Philosophy 7(2) (2010), pp. 32133.
Brake, Elizabeth, Rawls and Feminism: What Should Feminists Make of Liberal
Neutrality? Journal of Moral Philosophy 1(3) (2004), pp. 293309.
Reprinted in Thom Brooks and Fabian Freyenhagen (eds), The Legacy of John
Rawls. New York: Continuum, 2005.
Brock, Gillian, The Difference Principle, Equality of Opportunity, and Cosmopolitan
Justice, Journal of Moral Philosophy 2(3) (2005), pp. 33351.
Broome, John, Does Rationality Consist in Responding Correctly to Reasons? Journal
of Moral Philosophy 4(3) (2007), pp. 34974.
Brown, Stephen, Naturalized Virtue Ethics and the Epistemological Gap, Journal of
Butler, Brian E., Rorty, the First Amendment and Antirealism: Is Reliance upon Truth
Viewpoint-Based Speech Regulation? Journal of Moral Philosophy 1(1) (2004),
pp. 6988.
Carens, Joseph, The Integration of Immigrants, Journal of Moral Philosophy 2(1)
(2005), pp. 2946.
Carter, Alan, The Evolution of Rawlss Justification of Political Compliance: Part 1 of
The Problem of Political Compliance in Rawlss Theories of Justice, Journal of Moral
Philosophy 3(1) (2006), pp. 721.
Carter, Alan, Political Liberalism and Political Compliance: Part 2 of The Problem of
Political Compliance in Rawlss Theories of Justice, Journal of Moral Philosophy 3(2)
(2006), pp. 13557.
410
bibliography
Clark, Michael, Retribution and Organic Unities, Journal of Moral Philosophy 3(3)
(2006), pp. 35158.
Choi, Yoon, Revisiting Kants Ethics: Two Challenges to the Status Quo, Journal of
Cholbi, Michael, A Kantian Defense of Prudential Suicide, Journal of Moral Philosophy
7(4) (2010), pp. 489515.
Christie, Tim W., Natural Separateness: Why Parfits Reductionist Account of Persons
Fails to Support Consequentialism, Journal of Moral Philosophy 6(2) (2009),
pp. 17895.
Coyle, Sean, The Ideality of Law, Journal of Moral Philosophy 6(4) (2009),
pp. 52134.
Crisp, Roger, Ethics Without Reasons? Journal of Moral Philosophy 4(1) (2007),
pp. 4049.
Cruft, Rowan, On the Non-instrumental Value of Basic Rights, Journal of Moral
Philosophy 7(4) (2010), pp. 44161.
DAgostino, Fred, The Legacies of John Rawls, Journal of Moral Philosophy 1(3) (2004),
pp. 34965.
Dancy, Jonathan, Defending the Right, Journal of Moral Philosophy 4(1) (2007),
pp. 8598.
DeGrazia, David, Moral Vegetarianism from a Very Broad Basis, Journal of Moral
Philosophy 6(2) (2009), pp. 14365.
de Muijnk, Wim, Thinking about Normativity: Ralph Wedgwood on Ought, Journal
Deonna, Julien A., The Structure of Empathy, Journal of Moral Philosophy 4(1) (2007),
pp. 99116.
Dietsch, Peter, Distributive Lessons from Division of Labour, Journal of Moral
Philosophy 5(1) (2008), pp. 96117.
Enoch, David and Ehud Guttel, Cognitive Biases and Moral Luck, Journal of Moral
Philosophy 7(2) (2010), pp. 37286.
Eylon, Yuval, Just Threats, Journal of Moral Philosophy 6(1) (2009), pp. 94108.
Flikschuh, Katrin, Duty, Nature, Right: Kants Response to Mendelssohn in Theory and
Practice III, Journal of Moral Philosophy 4(2) (2007), pp. 22341.
Friedman, Alex, Intransitive Ethics, Journal of Moral Philosophy 6(3) (2009),
pp. 27797.
Fuller, Lisa F., Poverty Relief, Global Institutions, and the Problem of Compliance,
Journal of Moral Philosophy 2(3) (2005), pp. 28597.
Reprinted in Thom Brooks (ed.), The Global Justice Reader. Oxford: Blackwell,
2008.
Goddu, G. C., More on Blameworthiness and Alternative Possibilities, Journal of
Grau, Christopher, Moral Status, Speciesism, and Liaos Genetic Account, Journal of
Green, Michael, Social Justice, Voluntarism, and Liberal Nationalism, Journal of Moral
Philosophy 2(3) (2005), pp. 26583.
Hall, Timothy, Doing Harm, Allowing Harm, and Denying Resources, Journal of
Harcourt, Edward, Crisps Ethics Without Reasons?: A Note on Invariance, Journal
Hayward, Tim, Thomas Pogges Global Resources Dividend: A Critique and an
Alternative, Journal of Moral Philosophy 2(3) (2005), pp. 31732.
Hendrix, Burke A., Authenticity and Cultural Rights, Journal of Moral Philosophy 5(2)
(2008), pp. 181203.
bibliography411
Hills, Alison, Practical Reason, Value and Action, Journal of Moral Philosophy 4(3)
(2007), pp. 37592.
Hirose, Iwao, Aggregation and Non-Utilitarian Moral Theories, Journal of Moral
Philosophy 4(2) (2007), pp. 27384.
Hofmeyr, Benda, The Power Not to Be (What We Are): The Politics and Ethics of Selfcreation in Foucault, Journal of Moral Philosophy 3(2) (2006), pp. 21530.
Holroyd, Jules, Substantively Constrained Choice and Deference, Journal of Moral
Philosophy 7(2) (2010), pp. 18099.
Hookway, Christopher, Ethics and the Pragmatist Enlightenment, Journal of Moral
Philosophy 3(2) (2006), pp. 23136.
Hutton, Eric, Han Feizis Criticism of Confucianism and Its Implications for Virtue
Ethics, Journal of Moral Philosophy 5(3) (2008), pp. 42353.
James, Aaron, Rights and Circularity in Scanlons Contractualism, Journal of Moral
Philosophy 1(3) (2004), pp. 36774.
Kaufman, Alexander, Rawlss Practical Conception of Justice: Opinion, Tradition and
Objectivity in Political Liberalism, Journal of Moral Philosophy 3(1) (2006), pp. 2343.
Kirchin, Simon, Moral Particularism: An Introduction, Journal of Moral Philosophy
4(1) (2007), pp. 815.
Kirchin, Simon, Particularism and Default Valency, Journal of Moral Philosophy 4(1)
(2007), pp. 1632.
Knight, Carl, Egalitarian Justice and Valuational Judgment, Journal of Moral Philosophy
6(4) (2009), pp. 48298.
Knight, Kelvin, MacIntyres Progress, Journal of Moral Philosophy 6(1) (2009),
pp. 11526.
Kristjnsson, Kristjn, A Utilitarian Justification of Desert in Distributive Justice,
Laden, Anthony Simon, Taking the Distinction between Persons Seriously, Journal of
Landemore, Hlne, Politics and the Economist-King: Is Rational Choice Theory the
Science of Choice? Journal of Moral Philosophy 1(2) (2004), pp. 17796.
Landrum, Ty, Persons as Objects of Love, Journal of Moral Philosophy 6(4) (2009),
pp. 41739.
Lang, Gerald, Luck Egalitarianism, Permissible Inequalities, and Moral Hazard,
Lefkowitz, David, Partiality and Weighing Harm to Non-Combatants, Journal of
Lenard, Patti Tamara, Motivating Cosmopolitanism? A Skeptical View, Journal of
Lenman, James, How to Live, What to Do: A Critical Study of Allan Gibbard, Thinking
How to Live, Journal of Moral Philosophy 3(3) (2006), pp. 35969.
Liao, S. Matthew, Time-Relative Interests and Abortion, Journal of Moral Philosophy
4(2) (2007), pp. 24256.
Liao, S. Matthew, The Basis of Human Moral Status, Journal of Moral Philosophy 7(2)
(2010), pp. 15979.
Lippert-Rasmussen, Kasper, Publicity and Egalitarian Justice, Journal of Moral
Philosophy 5(1) (2008), pp. 3049.
Lippke, Richard L., Imprisonable Offenses, Journal of Moral Philosophy 3(3) (2006),
pp. 26587.
Lippke, Richard L., Punishing the Guilty, Not Punishing the Innocent, Journal of
412
bibliography
Ludwig, Bernd, Kant, Garve, and the Motives of Moral Action, Journal of Moral
Philosophy 4(2) (2007), pp. 18393.
Lynch, Sterling, The Fact of Diversity and Reasonable Pluralism, Journal of Moral
Philosophy 6(1) (2009), pp. 7093.
Mahoney, Jon, Public Reason and the Moral Foundation of Liberalism, Journal of
Marino, Patricia, Expressivism, Deflationism and Correspondence, Journal of Moral
Philosophy 2(2) (2005), pp. 17191.
Marino, Patricia, Moral Rationalism and the Normative Status of Desiderative
Coherence, Journal of Moral Philosophy 7(2) (2010), pp. 22752.
Martin, Rex, Twp Concepts of Rule Utilitarianism, Journal of Moral Philosophy 5(2)
(2008), pp. 22755.
Martin, Wayne, Hegel and the Philosophy of Food, Journal of Moral Philosophy 7(2)
(2009), pp. 27990.
Matravers, Matt, Whos Still Standing? A Comment on Antony Duff s Preconditions
of Criminal Liability, Journal of Moral Philosophy 3(3) (2006), pp. 32030.
McKeever, Sean and Michael Ridge, Turning on Default Reasons, Journal of Moral
Philosophy 4(1) (2007), pp. 5576.
Merritt, Maria W., Aristotelian Virtue and the Interpersonal Aspect of Ethical
Character, Journal of Moral Philosophy 6(1) (2009), pp. 2349.
Mookherjee, Monica, Feminism and MulticulturalismPutting Okin and Shachar in
Question, Journal of Moral Philosophy 2(2) (2005), pp. 23741.
Moyar, Dean, Unstable Autonomy: Conscience and Judgment in Kants Moral
Philosophy, Journal of Moral Philosophy 5(3) (2008), pp. 32760.
Mulnix, M. J., Harm, Rights, and Liberty: Towards a Non-Normative Reading of Mills
Liberty Principle, Journal of Moral Philosophy 6(2) (2009), pp. 196217.
Naticchia, Chris, The Law of Peoples: The Old and the New, Journal of Moral Philosophy
2(3) (2005), pp. 35369.
Nolan, Daniel, Consequentialism and Side Constraints, Journal of Moral Philosophy
6(1) (2009), pp. 522.
Norman, Richard, Particularism and Reasons: A Reply to Kirchin, Journal of Moral
Philosophy 4(1) (2007), pp. 3339.
Nussbaum, Martha C., Radical Evil in the Lockean State: The Neglect of the Political
Emotions, Journal of Moral Philosophy 3(2) (2006), pp. 15978.
ONeill, Martin, The Facts of Inequality, Journal of Moral Philosophy 7(2) (2009),
pp. 397409.
ONeill, Onora, Experts, Practitioners, and Practical Judgement, Journal of Moral
Philosophy 4(2) (2007), pp. 15466.
ONeill, Onora, Normativity and Practical Judgement, Journal of Moral Philosophy
4(3) (2007), pp. 393405.
verland, Gerhard, Self-defence among Innocent People, Journal of Moral Philosophy
2(2) (2005), pp. 12746.
verland, Gerhard, Poverty and the Moral Significance of Contribution, Journal of
verland, Gerhard, Conditional Threats, Journal of Moral Philosophy 7(2) (2010),
pp. 33445.
Papdaki, Lina, What is Objectification? Journal of Moral Philosophy 7(1) (2010),
pp. 1636.
Peterson, Martin, The Mixed Solution to the Number Problem, Journal of Moral
Philosophy 6(2) (2009), pp. 16677.
bibliography413
Pink, Thomas, Normativity and Reason, Journal of Moral Philosophy 4(3) (2007),
pp. 40631.
Primoratz, Igor, Patriotism and Morality: Mapping the Terrain, Journal of Moral
Philosophy 5(2) (2008), pp. 20426.
Redondo, Mara Cristina, Legal Reasons: Between Universalism and Particularism,
Reitan, Eric, Defining Terrorism for Public Policy Purposes: The Group-Target
Definition, Journal of Moral Philosophy 7(2) (2010), pp. 25378.
Richardson, Henry S., Our Call: The Constitutive Importance of the Peoples Judgment,
Ridge, Michael, Anti-Reductionism and Supervenience, Journal of Moral Philosophy
4(3) (2007), pp. 33048.
Riley, Evan, Libertarian Self-Defeat, Journal of Moral Philosophy 7(2) (2010),
pp. 20026.
Riley, Patrick, Kant against Hobbes in Theory and Practice, Journal of Moral Philosophy
4(2) (2007), pp. 194206.
Riley, Jonathan, Ethical Pluralism and Common Decency, Journal of Moral Philosophy
1(2) (2004), pp. 21121.
Ronzoni, Miriam, Constructivism and Practical Reason: On Intersubjectivity,
Abstraction, and Judgment, Journal of Moral Philosophy 7(1) (2010), pp. 74104.
Rosebury, Brian, Reply to Silcox on Moral Luck, Journal of Moral Philosophy 6(1)
(2009), pp. 10913.
Sable, Andrew, Virtue for Pluralists, Journal of Moral Philosophy 2(2) (2005),
pp. 20735.
Scarre, Geoffrey, Corrective Justice and Reputation, Journal of Moral Philosophy 3(3)
(2006), pp. 30519.
Scarre, Geoffrey, The Banality of Good? Journal of Moral Philosophy 6(4) (2009),
pp. 499519.
Segall, Shlomi, How Devolution Upsets Distributive Justice, Journal of Moral
Philosophy 4(2) (2007), pp. 25772.
Seglow, Jonathan, Associative Duties and Global Justice, Journal of Moral Philosophy
7(1) (2010), pp. 5473.
Shafer-Landau, Russ, Moral and Theological Realism: The Explanatory Argument,
Sheinman, Hanoch, Raz on the Social Dependence of Values, Journal of Moral
Philosophy 3(1) (2006), pp. 7787.
Silcox, Mark, Virtue Epistemology and Moral Luck, Journal of Moral Philosophy 3(2)
(2006), pp. 17992.
Silcox, Mark, Reply to Rosebury, Journal of Moral Philosophy 6(2) (2009),
pp. 24548.
Simpson, Matthew, A Paradox of Sovereignty in Rousseaus Social Contract, Journal of
Sin, William, Trivial Sacrifices, Great Demands, Journal of Moral Philosophy 7(1)
(2010), pp. 315.
Singleton, Jane, Neither Generalism nor Particularism: Ethical Correctness is Located
in General Ethical Theories, Journal of Moral Philosophy 1(2) (2004), pp. 15575.
Slomp, Gabriella, Kant against Hobbes: Reasoning and Rhetoric, Journal of Moral
Philosophy 4(2) (2007), pp. 20722.
Smith, M. B. E., Does Humanity Share a Common Moral Faculty? Journal of Moral
Philosophy 7(1) (2010), pp. 3753.
Smith, Stephen R., Keeping Our Distance in Compassion-Based Social Relations,
Sobel, David, The Limits of the Explanatory Power of Developmentalism, Journal of
414
bibliography
Spector, Jessica, The Grounds of Moral Agency: Lockes Account of Personal Identity,
Stark, Susan, A Change of Heart: Moral Emotions, Transformation, and Moral Virtue,
Stern, Robert, The Autonomy of Morality and the Morality of Autonomy, Journal of
Stone, Alison, Essentialism and Anti-Essentialism in Feminist Philosophy, Journal of
Talisse, Robert B., Does Value Pluralism Entail Liberalism? Journal of Moral Philosophy
7(2) (2010), pp. 30320.
Taylor, Robert S., Self-Realization and the Priority of Fair Equality of Opportunity,
Telfer, Elizabeth, Animals Do It Too!: The Franklin Defence of Meat-Eating, Journal
Thomas, Alan, Practical Reasoning and Normative Relevance: A Reply to McKeever
and Ridge, Journal of Moral Philosophy 4(1) (2007), pp. 7784.
Tideman, Nicolaus, Secession as a Human Right, Journal of Moral Philosophy 1(1)
(2004), pp. 919.
Timmermann, Jens, Good but Not Required?Assessing the Demands of Kantian
Ethics, Journal of Moral Philosophy 2(1) (2005), pp. 927.
Timmermann, Jens, Simplicity and Authority: Reflections on Theory and Practice in
Kants Moral Philosophy, Journal of Moral Philosophy 4(2) (2007), pp. 16782.
Tropman, Elizabeth, Renewing Moral Intuitionism, Journal of Moral Philosophy 6(4)
(2009), pp. 44063.
Tyler, Colin, Brian Barry and Writings on Social Justice from the Left, Journal of Moral
Philosophy 5(2) (2008), pp. 30112.
Upton, Candace, Virtue Ethics, Character, and Normative Receptivity, Journal of
van Zyl, Liezl, Agent-based Virtue Ethics and the Problem of Action Guidance, Journal
Vargas, Manuel, Taking the Highway on Skepticism, Luck, and the Value of
Responsibility, Journal of Moral Philosophy 6(2) (2009), pp. 24965.
Ward, Lee, Locke on Punishment, Property and Moral Knowledge, Journal of Moral
Philosophy 6(2) (2009), pp. 21844.
Webber, Jonathan, Virtue, Character and Situation, Journal of Moral Philosophy 3(2)
(2006), pp. 193213.
Weirich, Paul, Utility Maximization Generalized, Journal of Moral Philosophy 5(2)
(2008), pp. 28299.
Wenar, Leif, The Unity of Rawlss Work, Journal of Moral Philosophy 1(3) (2004),
pp. 26575.
Westphal, Kenneth R., From Convention to Ethical Life: Humes Theory of Justice
in Post-Kantian Perspective, Journal of Moral Philosophy 7(1) (2010), pp. 10532.
White, Heath, Fitting Attitudes, Wrong Kinds of Reasons, and Mind-Independent
Goodness, Journal of Moral Philosophy 6(3) (2009), pp. 33964.
Wiland, Eric, On Indirectly Self-defeating Moral Theories, Journal of Moral Philosophy
5(3) (2008), pp. 38493.
Williams, Reginald, Morality and Privilege, Journal of Moral Philosophy 5(1) (2008),
pp. 11835.
Wolff, Jonathan, Equality: The Recent History of an Idea, Journal of Moral Philosophy
4(1) (2007), pp. 12536.
bibliography415
Zaibert, Leo, The Fitting, the Deserving, and the Beautiful, Journal of Moral Philosophy
3(3) (2006), pp. 33150.
Zaibert, Leo, The Paradox of Forgiveness, Journal of Moral Philosophy 6(3) (2009),
pp. 36593.
Ziegler, Rafael, Tracing Global Inequality in Eco-space: A Comment on Tim Haywards
Proposal, Journal of Moral Philosophy 4(1) (2007), pp. 11724.
index
abortion 4, 156, 292, 299, 31528, 330,
333, 336, 339, 3478, 350, 353
Andreou, Chrisoula 1
animal/animals 11, 223, 745, 82, 216,
319, 321, 323, 32730, 3367, 339,
3445, 349
Anscombe, G. E. M. 80, 86
anti-essentialism see essentialism
Aristotle 18, 88, 175, 183, 207, 210,
2138, 222, 234, 261, 2737
Aristotelian 4, 64, 209, 271
Ashford, Elizabeth 374
attitudinal reason see reason/reasons
Brink, David 132, 242
Brooks, Thom 45, 287, 361
Broome, John 1, 64, 317, 330, 335,
3578, 362, 3656
buck-passing122
character 4, 116, 190, 201, 204, 20713,
21528, 2302, 247, 25964, 26873,
2768, 2813, 387, 38997
common sense 2, 989, 238, 323
Confucius 34, 1758, 1802,
1989, 201
consequentialism 66, 175, 232, 239, 248,
289, 304, 307, 375
Crisp, Roger 2, 10710
Dancy, Jonathan 2, 95107, 1656, 205
death 4, 141, 194, 288, 292, 298, 300,
3179, 326, 328, 3323, 336, 350
DeGrazia, David 318, 320, 326, 328
Doris, John 210, 235, 25971, 273, 277
essentialism 5, 38598, 404, 4067
anti-essentialism 385, 392, 395, 405
experimental psychology 4, 210
evil 4, 134, 215, 226, 2334, 257,
399, 401
Feizi, Han 34, 17380, 182, 184205
feminism/feminist 5, 38598, 4057
Freyenhagen, Fabian 1, 56
friendship 18, 69, 207, 214, 217,
220, 254
generalism 2, 95, 979, 1056

God 447, 96, 101, 103, 106, 109, 1258,
1301, 13643, 2513, 267
Hall, Timothy 4
happiness 668, 138, 2401, 256
harm 4, 301, 81, 193, 28793, 295304,
30615, 328, 347, 374
Harman, Gilbert 123, 1920, 32, 128,
1312, 136, 209, 235, 261
Herman, Barbara 287, 372
Hill, Thomas 2379, 244, 2479, 256
Hills, Alison 2, 125
Hirose, Iwao 35767, 369
holism 2, 967, 99100, 1048, 112,
1658
Hubin, Donald 13, 1920
Hume, David 69, 153, 251
Humean 134, 1921, 23
Hursthouse, Rosalind 1823, 205, 207,
219, 2345, 347
Hutton, Eric 34, 175, 2112, 235
ignorance 2830, 32, 34, 62, 64, 128, 135
intention 11, 146, 189, 30, 32, 356,
437, 52, 545, 60, 65, 6775, 80, 89,
1334, 142, 144, 152, 1545, 159, 164,
174, 209, 212, 222, 225, 230, 232, 234,
302, 3213, 340
Irigaray, Luce 388, 393
Kamm, Frances Myrna 2878, 297, 299,
3578, 362, 365
Kamtekar, Rachana 175, 20910, 221,
264, 271
Kant, Immanuel 712, 85, 878, 23757
Kantian ethics 5, 58, 81, 232, 234,
239, 2434, 25267
killing 57, 70, 82, 142, 246, 275, 288,
2902, 295, 298301, 303, 307,
30910, 3145, 31734, 3378, 353
Korsgaard, Christine 20, 25, 589, 64,
67, 1201
Liao, S. Matthew 45, 327, 331, 339,
3523
Lillehammer, Hallvard 1
418 index
Mackie, J. L. 127, 153, 161
Majors, Brad 125, 132, 147, 1623
Marcus, Ruth Barcan 91, 113
McDowell, John 88, 183, 205, 2746
McKeever, Sean 2, 1112, 147
McMahan, Jeff 4, 28795, 309, 314,
31729, 3313, 335, 3378, 344,
353, 371
Mencius 1768, 182, 1856, 188, 1924,
1967, 201
Merritt, Maria 4, 209, 2601, 268,
270, 277
metaethics 1, 3, 57, 126, 147
Milgram, Stanley 2656
Miller, Christian B. 209, 262, 2689
moral facts 3, 12738, 1404,
154, 209
moral non-naturalism 147
moral realism 1, 3, 1256, 1302, 134,
1368, 1534, 156, 159
moral status 4, 289, 293, 298, 301,
303, 3135, 326, 333, 339, 346,
3513, 356
Rabinowicz, Wlodek 25
rationality 1, 256, 2832, 345, 3745,
4752, 545, 59, 81, 105, 145, 210,
224, 2402, 276, 319, 340
reason/reasons
attitudinal reason 302, 347
practical reason 50, 54, 57, 5968, 73,
76, 99100
Ridge, Michael 23, 1112
Ross, W. D. 98, 167
Nagel, Thomas 31
non-Western philosophy 3
Norman, Richard 2, 1167, 121
norms 2, 7785, 87, 8992, 125,
209, 2112, 226, 229, 231,
397, 402
normativity 2, 59, 812, 11920
Nozick, Robert 305, 3089, 336
number problem 5, 3579,
3612, 369
utilitarianism 66, 84, 95, 100, 238, 245,

35960
ONeill, Onora 2, 92
Otsuka, Michael 287, 307, 357
particularism 1, 95, 979, 101, 106,
1112, 116
Peterson, Martin 5
Plato 11, 165, 210
practical judgement 2, 78, 85, 8792
practical reason see reason/reasons
punishment 26, 178, 189, 193, 400
Scanlon, T. M. 40, 457, 57, 335,

3379, 373
Shafer-Landau, Russ 3, 11, 1479,
1546, 15864, 166, 168
Sin, William 5
social psychology 2101, 21921, 223
Stark, Cynthia 11
Stone, Alison 5
supervenience 3, 30, 32, 38, 41, 43, 51,
14759
Timmermann, Jens 5
vegetarian/vegetarianism11
Velleman, David 5860, 6667
virtue ethics 1, 34, 1745, 1823, 201,
20511, 219, 2325, 2602, 264, 271
Webber, Jonathan 4, 208
Wedgwood, Ralph 29, 125
Wiggins, David 88, 276
Williams, Bernard 13, 1920, 613, 88,
100, 183, 189, 1912, 195, 1978,
2013, 213, 276, 3356, 372
Xunzi 1734, 1778, 184, 188, 194,
199201
Young, Iris Marion 386, 392, 395
Zimmerman, David 132, 1345

Ethics and Morals Phylosophy

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ethics and Morals Phylosophy

Uploaded by

Copyright:

Available Formats

Ethics and Moral Philosophy

Studies in Moral Philosophy

Thom Brooks (Newcastle University)

Chrisoula Andreou (University of Utah)

The titles published in this series are listed at brill.nl/simp

Ethics and Moral Philosophy

This book is printed on acid-free paper.

Ethics and Moral Philosophy

Notes on previous publication

notes on previous publication

Timothy Hall, Doing Harm, Allowing Harm, and Denying Resources,

Michael Ridge is Professor of Moral Philosophy at Edinburgh

Alison Hills (2007) considers how we might determine which theory

Standards, Advice, and Practical Reason

even though, according to his standards, his request is inappropriate

Gilbert Harman, Moral Relativism, in G. Harman and J.J. Thomson, Moral

standards, advice, and practical reason13

For it might seem as though there is something to the following line of

system of standards. If, however, I am right, and the nature of sincere

standards, advice, and practical reason15

standards, advice, and practical reason17

standards, advice, and practical reason19

According to the popular Humean view, practical reason judgments

standards, advice, and practical reason21

standards, advice, and practical reason23

Does Rationality Consist in Responding

As I shall put it, rationality is equivalent to responding correctly to

As I shall put it, rationality entails responding correctly to reasons.

does rationality consist in responding correctly? 27

Let us go more carefully. Suppose you ought to F. I take it for granted

It makes equivalence come down to:

One side of it is:

Since I am not interested in the explanatory or counterfactual

This schema is supposed to hold for all substitutions of a verb phrase

does rationality consist in responding correctly? 29

salmonella, but, without fault, you do not believe this constitutes a

Instead, they assume that a rational persons response to reasons has to

In each of these claims, the reason is said to be an attitude of yours or a

does rationality consist in responding correctly? 31

Here is why. Take R3 as an example. For p substitute Emissions of

I have stated these conditions only approximately. For example, you

does rationality consist in responding correctly? 33

umbrella. To do so, you do not need to have a second-order belief that

whether you can respond to an attitudinal reason without believing

Equivalence might be true if it was restricted in the same way:

That is to say, rationality is equivalent to responding correctly to attitudinal reasons.

does rationality consist in responding correctly? 35

There is a lacuna in this argument. It shows that you can respond

does rationality consist in responding correctly? 37

This seems plausible to me,6 if attitudinal reasons exist. But since

That is to say, rationality entails responding correctly to beliefs about

Thanks to Julia Markovits for pointing this out to me.

But I shall not accept modified equivalence under any interpretation.

This condition encounters no objection from the supervenience of

does rationality consist in responding correctly? 39

This condition constitutes an interpretation of modified entailment.

In section 4 I insisted that many other conditions are necessary for

I would be happy to accept some weakened formula such as this.

does rationality consist in responding correctly? 41

I explained in section 4 that this condition is not true universally,

Compare modified entailment from section 4. I repeat it here:

does rationality consist in responding correctly? 43

the conditions themselves. Replacing them with the direct enkratic

does rationality consist in responding correctly? 45

standards, advice, and practical reason13

standards, advice, and practical reason15

standards, advice, and practical reason17

standards, advice, and practical reason19

standards, advice, and practical reason21

standards, advice, and practical reason23

does rationality consist in responding correctly? 27

does rationality consist in responding correctly? 29

does rationality consist in responding correctly? 31

does rationality consist in responding correctly? 33

does rationality consist in responding correctly? 35

does rationality consist in responding correctly? 37

does rationality consist in responding correctly? 39

does rationality consist in responding correctly? 41

does rationality consist in responding correctly? 43

does rationality consist in responding correctly? 45

does rationality consist in responding correctly? 47

does rationality consist in responding correctly? 49

does rationality consist in responding correctly? 51

does rationality consist in responding correctly? 53

does rationality consist in responding correctly? 55

practical reason, value and action59

practical reason, value and action61

practical reason, value and action63

practical reason, value and action65

practical reason, value and action67

practical reason, value and action69

practical reason, value and action71

practical reason, value and action73

practical reason, value and action75

normativity and practical judgement79

normativity and practical judgement81

normativity and practical judgement83

normativity and practical judgement85

normativity and practical judgement87

normativity and practical judgement89

normativity and practical judgement91

ethics without reasons?97