Vous êtes sur la page 1sur 259


This page intentionally left blank

Moral Dilemmas, Emotions,
and Social Norms


New Yor k Oxfor d

Oxford Universit y Press
Oxford Ne w York
Athens Aucklan d Bangko k Bomba y
Calcutta Cap e Town Da r e s Salaam Delh i
Florence Hon g Kong Istanbu l Karach i
Kuala Lumpur Madra s Madri d Melbourn e
Mexico City Nairob i Pari s Singapor e
Taipei Toky o Toront o
and associate d companie s in
Berlin Ibada n

Copyright 1995 by P. S. Greenspan

Published by Oxford Universit y Press, Inc. ,
200 Madison Avenue, New York, New York 10016
Oxford i s a registered trademar k o f Oxford Universit y Press
All rights reserved. No par t o f this publication ma y b e reproduced,
stored i n a retrieval system, o r transmitted , i n any for m o r b y any means ,
electronic, mechanical , photocopying, recording , o r otherwise ,
without the prior permission o f Oxford Universit y Press.

Library of Congress Cataloging-in-Publication Data

Greenspan, Patricia S., 1944-
Practical guilt : mora l dilemmas, emotions, an d social norms /
P. S. Greenspan.
p. cm .
Includes bibliographica l references and index .
ISBN 0-19-508762-3; ISBNO-19-509090-X (pbk.)
1. Guilt . 2 . Emotion s (Philosophy ) 3 . Socia l norms. 4 . Ethics .
I. Titl e
BJ1471.5.G74 199 4
128'.3dc20 93-4006 7

Printed i n the Unite d States of America
on acid-fre e pape r
In memor y o f
The fox condemns the trap, not himself.
William Blak e
The Marriage of Heaven and Hell

The sources of my thinking on the several subjects of this essay go back many
years, but its first full draft was written durin g 1990-91 at the National Human -
ities Center i n Research Triangle Park , Nort h Carolina . I am indebted to th e
other fellows in residence at the center for historical and literar y discussion of
guilt, and to the center staff for sustaining the more upbeat emotions that allowed
for m y unusual productivity ove r th e year . Le t me als o than k th e Nationa l
Endowment fo r the Humanities for providin g my funding an d th e University
of Maryland, which provided supplementary support. Two terms off at Mary -
land during the preceding two years enabled me to write drafts o f some initial
sections an d to work ou t a detailed plan of argument. I owe special thanks t o
colleague James Lesher, then actin g dean of the Colleg e o f Arts and Humani -
ties, for arranging my leave during autumn 1988; I was awarded another leav e
by the Graduate Research Board at Maryland durin g autumn 1989 .
I am also grateful t o a number of people who supporte d m y work o n thi s
project or provoked m y thinking with correspondence o r conversation o n the
subjects it brings together. Alan Donagan, t o whom the book is dedicated, wa s
my colleague at the University of Chicago an d the mainstay of my grant refer -
ees for many years thereafter, switching over to this project despite his sharply
opposing vie w on moral dilemmas . Hi s letter s o n the subjec t determined th e
basic shape of my argument and its ultimate focus on metaethicsto use a term
still out o f favor for a set of issues in the foundations of ethics that concerne d
me most as a student of philosophy bu t became unfashionable at about the time
I left school. Much of this essay emerged in an attempt to answer Alan's insis-
tent an d impassioned question s abou t ho w ethics could accommodat e dilem -
mas while amounting to somethin g mor e tha n a simple codification o f emo -
Alan's prodding serve d to dra w me back t o a number of earlier interests ,
with dilemma s providing a helpfully narro w angle of approach to topic s to o
large to deal with effectively head-on . At one point during the planning of this
essay, upon reviewing some of my old student papers and other course materi-
als while packing up to move, I even found som e anticipation s of my specific
theses. There was a long-forgotten defense of guilt for the unavoidable written
up for a seminar at Harvard i n 1970 an d a critique of W. D. Ross for a 196 7
ethics course whose main line of argument I decided to incorporate into chap-
viii Acknowledgments

ter 4 , sectio n 3 , of this essay. Some of these topics, which at tha t poin t wer e
rather off-center, have since come into their own i n contemporary mora l phi-
losophy. However, they have mainly been linked with the Aristotelian approach
to ethics that stresses notions of virtue and character over obligation and action.
My argumen t her e is in part a n attemp t t o sho w wha t sor t o f role emotion s
also play on the modern approac h we associate wit h Kant , once we detach i t
from som e o f the mor e extrem e elements of a Kantia n approach, a s brough t
out b y my treatment of dilemmas.
Some features of this project are due to conversation s with Marylan d col-
league and department chair Michael Slote that encouraged me to bring my work
on dilemma s and emotion s t o bea r o n mor e mainstrea m issues in ethics. To
highlight th e connection , I decided to presen t m y argument a s it emerged, in
discussion of the views of a number of central contemporary authors, to whom
I also owe a debt of thanks for providing me with materials I sometimes use in
ways quit e othe r tha n wha t the y intended . Amon g author s o n dilemmas , a
particular influence was Ruth Marcus, whose paper I responded to at the 198 0
Chapel Hill Colloquium in Philosophy. Three conversations with Bernard Wil-
liams in October 198 9 als o helped me clarify his views on dilemmas and related
Besides Donagan an d Slote , a number of people wrote letters in support of
my numerou s gran t application s fo r thi s project : Annett e Baier , Simo n
Blackburn, Richard Brandt, John Cooper, Jonathan Glover , Mary Mothersill ,
Thomas Nagel, and Philip Quinn. What emerged from my year at the National
Humanities Center was a social view of the base s of ethics that I came to see as
the centra l result of my work o n this subject. Colleagues who provide d com-
ments o n draft s o f par t o r al l o f th e essa y includ e Kent Bach , David Copp ,
Jonathan Dancy , Stephen Leighton, Geoffrey Sayre-McCord , Walte r Sinnott -
Armstrong, the late Michael Woods, several anonymous reviewers, and students
in some of my classes, especially Lawrence Dobbs, Richard Fyfe, Scott Gelfand,
David Hull, and Stephen Tighe.
I also received comments on drafts of particular sections at a number of oral
presentations (a t Philosophy Department Colloquia , unles s otherwise noted) :
Chapter 5 , sectio n 1 , "Subjective Guilt and Responsibility, " wa s rea d at th e
University o f Rocheste r i n 1988 , the n a t Indian a University in 1989 ; par t of
chapter 4, section 1 , at North Carolin a State University in 1991 unde r the title
"Guilt an d Virtue" ; chapte r 4 , sectio n 2 , "Guil t a s an Identificatory Mecha-
nism," at the University of North Carolin a a t Chapel Hill and at the National
Humanities Cente r Seminar on the Concept o f a Person i n 1991; par t of chap-
ter 5 , section s 2-3 , a t a symposiu m i n honor o f Ruth Marcu s a t th e Pacific
Division meetings of the American Philosophical Association in 1992 unde r the
title "Perspectival Guilt"; and a selection from chapter 3, section 3 , at George-
town University and Queen's University (Kingston, Ontario) i n 1993 unde r the
title "Protagorea n Realism. " Chapte r 3 , section 1 , was discussed at a meeting
of the Georgetown / Maryland Moral Psycholog y Reading Group in 1991 .
Later drafts of some of these pieces, along with some related selections, have
appeared in print: "Subjective Guil t and Responsibility, " in Mind 10 1 (1992):
Acknowledgments i x

287-303; "Guil t a s an Indentificator y Mechanism," i n Pacific Philosophical

Quarterly 7 4 (1993): 46-59; "Guilt an d Virtue," in Journal o f Philosophy 9 1
(1994): 57-70 ; and a longe r versio n o f "Perspectiva l Guilt, " i n Modality,
Morality an d Belief: Essays in Honor o f Ruth Barcan Marcus, ed. W. Sinnott-
Armstrong (Cambridge: Cambridge University Press, 1994 [copyright Cam-
bridge University Press; reprinted by permission]).
For secretarial support at various stages, let me also thank Marsha Brown ,
Richard Fyfe , an d Kati e Kight. I owe particular thank s a t Oxfor d University
Press to Angela Blackburn for painstaking editorial help and advice.

Washington, D.C . P.S.G .

August 1993
This page intentionally left blank

Introduction, 3


1. Defusin g Dilemmas , 9
1. Mora l Dilemma s and Motivational Force , 1 1
2. Motivatin g Moral "Ought, " 23

2. Practica l Oughts and Prohibitions, 29

1. Practica l Oughts in Conflict, 30
2. Deonti c Strength and Value, 41
3. Problem s for Practical Ought-Systems, 5 2

3. Motivationa l Foundations of Conflict, 6 6

1. Mora l Realis m and Practical Phenomenology, 67
2. Internalis t Dilemmas, 77
3. Betwee n the Horns, 90


4. Mora l Residues, 109
1. Th e Moral Significanc e of Guilt, 111
2. Guil t as an Identificatory Mechanism, 126
3. Contrary-to-Dut y "Ought-to-Feel," 13 6

5. Unavoidabl e Guilt , 15 1
1. Subjectiv e Guilt and Responsibility , 152
2. Perspectiva l Appropriateness, 166
3. Objectiv e Guilt and Wrong, 17 6
xii Contents

6. Basin g Ethics on Emotion, 187

1. Th e Motivational Model, 188
2. Sensibilit y and Standpoints, 19 8

Notes, 211
Bibliography, 23 5
Index, 24 3
This page intentionally left blank

In what follow s I deal with a numbe r o f diverse issues of interest to philoso -

phers (an d others) in several areas and o f different intellectua l bent. I explain
the connections among these topics in chapter 1, but since my discussion quickl y
enters the thick of metaethical debate, it might be helpful at this point to present
an overal l ma p o f my argument s o that reader s ma y char t alternativ e paths
through it.
My argument begins with issue s in the foundations of ethics, with a n eye
to determinin g th e plac e o f mora l emotion . However , man y reader s migh t
prefer to begi n with my treatment o f a specific emotion, guilt, in part II , pos-
sibly doubling back later a s needed t o mak e full sens e of the general point s I
extract fro m that discussio n for moral dilemmas , toward th e end of chapte r
5, or the role o f emotion i n ethics, i n chapter 6 . I have provided index refer-
ences to cases discussed, along with other background informatio n that should
help t o clea r up an y initiall y obscure passages , bu t I think tha t th e bul k of
chapters 4 and 5 can be read independentl y by readers wit h a special interes t
in emotion .
For readers of part I, I shall give more specifi c advice as my discussion there
proceeds about how they might bypass the more technical treatment of deontic
logic in chapter 2 . It should als o b e possible t o proceed directl y to chapte r 3 ,
and thenc e t o chapte r 6 , for a n accoun t o f my proposals o n metaethic s that
bypasses details of the literature on moral dilemma s along with the treatmen t
of guilt. The index will supply references to some terminology explained in chap-
ter 1; and the second sectio n of chapter 1 might be read on its own, for a fulle r
account o f my overall argument i n relation t o metaethics. A t this point I want
just to give an outline that indicates th e main topics to be covered in each chap-
ter but leaves explanations unti l later .
I begin in chapter 1 with a review of philosophers' treatmen t o f dilemmas,
attempting to show why they are thought t o pose a problem fo r the coherency
of ethics. On my account they call into question the motivational force of "ought"
and similar deontic terms, taken as essentially practical or action-guiding. They
pose a metaethical dilemma of sorts, between "subject-independent" views of
moral motivation , whic h apparentl y mak e mora l dilemma s impossible, an d
"subject-dependent" views, which seem to be unable to capture the motivational
difficulty o f dilemmas.
4 Practical Guilt

In chapter 21 deal with some of the standard question s raised for dilemmas
in connection with deontic logic , though I come at them from a direction tha t
I hope will bring out their metaethical relevance. First I address questions raised
by Bernard Williams's dismissal of dilemmas involving practical oughts as con-
clusions of deliberation. I argue that an investigation of some presupposition s
about the weighing of moral reasons favors a redefinition of dilemmas in terms
of prohibitions rather than positive oughts. I then apply this negative formula-
tion t o th e choice dilemma s apparently force between standard deonti c prin-
ciples, mos t notabl y "ought"-implies-"can " an d agglomeration , wit h som e
comments abou t th e implication s o f th e choic e fo r th e unifiabilit y o f action-
guiding ethics, understood a s based on practical "ought."
Chapter 3 begins with a shift t o contemporary version s of the questio n of
motivational forc e introduce d i n chapter 1 . On th e basi s of some of Philippa
Foot's comments abou t mora l teaching , I suggest a way o f underminin g the
standard dichotom y betwee n internalis m and externalis m i n contemporar y
metaethics by connecting motivational forc e to th e general moral functio n of
"ought," as presupposed i n teaching the term, rather tha n t o it s meaning in a
given judgment. The resulting view, which I call "general internalism," amounts
to a modification of externalism to provide a nonaccidental link between moral
judgment an d motivation . I attemp t t o sho w tha t i t yield s a mor e plausible
account of various problematic metaethical issues than standard internalis m or
In particular, I argue that the view is needed to make sense of moral dilem-
mas. Standar d version s o f internalism , a s represente d b y th e view s o f Joh n
McDowell an d J. L. Mackie, see m to b e unable to accoun t fo r dilemma s ade-
quately. B y switching t o genera l internalism, however , I sho w ho w w e ca n
extract fro m elements o f bot h views an explanatio n o f dilemmas i n term s of
the conception o f ethics as a social artifact with a link to individual motivation
provided by emotion. Par t I ends with some of the implications of my proposed
metaethical alternative, which I defend as a version of moral realism .
Part II turns to guilt and related reactions to moral wrong, initially consid-
ered without distinction , a s emotional residue s of moral conflict or more gen-
erally of "mora l failure " i n the sens e of ought-violation. I n chapter 4 I sho w
the significance of guilt feelings as a link between virtue and duty ethics. I then
attempt a mor e specifi c accoun t o f th e natur e an d functio n o f guil t a s a n
identificatory mechanism , eventuall y distinguishing it fro m sham e an d othe r
similar moral feelings that do not necessarily involve identification with others,
though m y view allows fo r their overla p with guilt in some cases .
Next I turn t o question s abou t th e sens e in which guilt or an y emotiona l
reaction can be morally requiredas guilt seems to be in typical cases of moral
failurein ligh t of the principle that "ought"-implies-"can. " I n application t o
dilemmas, these are questions abou t a judgment of "ought-to-feel" tha t might
be thought o f a s offering a way outa n indirect way o f satisfying the ough t
that i s not acted onfor agents with enoug h control to generate the requisite
emotion. I argue, however, that this suggestion needs to be qualified in impor-
tant ways. Among other things, there are second-order dilemmas in some cases,
Introduction 5

in which guilt feelings would interfere with effective actio n on the ought that is
supposed t o be satisfied directly , by taking action .
In chapter 5 I address objections to classifying th e agent's reaction a s guilt
in situations of dilemma on the grounds that guilt implies a judgment of culpa-
bility. In application initially to nondilemmatic cases of guilt for the unavoid-
able, I defend a view of the prepositional conten t o f guilt according t o which
the emotion involve s an "a s if " version o f the corresponding evaluativ e judg-
ment, not necessarily an evaluation that the agent believes. On the question of
appropriateness, my approach yields an asymmetry between guilt and emotional
blame that I take to explain any reluctance we may have to impose guilt on the
agent in a dilemma. The differen t positio n o f the subjec t in relation to guil t as
opposed t o blame licenses a different sor t of assessment o f emotional warrant ,
so that les s is required to warrant guilt. I go on to defen d a corresponding ob -
jective notio n o f guilt in application t o mora l dilemmas , thought o f a s cases
where all the agent's options for action ar e wrong .
In chapte r 6 I focus agai n o n th e metaethica l position emergin g from m y
argument, as introduced in chapter 1 and developed in chapter 3. After drawing
out som e o f the implication s of my motivational interpretatio n o f the rol e of
moral emotion , I attempt t o sho w ho w th e "socia l artifact" versio n of mora l
realism can allo w for a n elemen t of expressivism and othe r view s commonly
thought of in contrast to realismmost notably relativismwithout entailing
acceptance of those views . On it s most genera l characterization, th e resul t of
my argument is thus a defense of commonsense ethics against the problems raised
by dilemmas, by way o f an extended treatmen t o f the base s of ethics in moral
This page intentionally left blank
This page intentionally left blank
Defusing Dilemma s

In the past twenty-five year s or s o a number of philosophers hav e argued fo r

the possibility of full-blown mora l dilemmas : cases in which al l of the agent' s
alternatives, through no fault of his own, turn out to be morally wrong. Other
authors have attempted to dismiss such cases as undermining the rational cohe-
rency of ethics, on th e moder n conceptio n o f ethics as essentially a system of
rules meant to guide action. A good specimen of the sort of case at issue in the
literature is provided by one of Michael Walzer's examples in his contribution
to a 1973 symposiu m on the rules of war with Thomas Nagel, Richard Brandt,
and R . M. Hare. 1 The case assumes that torture i s necessary to get a terroris t
to giv e away the whereabouts o f bomb s set to g o off throughout th e city but
that tortur e stil l remain s wrong unde r thes e circumstancesalon g wit h th e
agent's onl y other alternative, which involves letting the bombs destroy many
innocent civilians.
The case is one in which our ordinary moral code fails to yield coherent mora l
advicein Nagel' s word s i t set s th e agen t i n a "mora l blin d alley"since it
apparently rule s out al l alternatives.2 Brand t and Hare argue in opposition t o
this account tha t suc h acts as torturing th e terrorist, eve n i f normally wrong ,
do not reall y count a s wrong under the circumstances. 3 Instea d o f relying on
general rules such as the one prohibiting torture, we should appeal in a case of
conflict t o straightforwar d utilitaria n assessments of the acts in question. Bu t
note tha t someon e on either side of this dispute could agre e with Brand t an d
Hare tha t utilitaria n considerations outweig h principled objections to tortur e
in Walzer's cas e when i t comes t o a practical decision . Th e disput e concern s
not what to do in such cases but whether the action one must do, morally speak-
ing, can stil l be wrong.
For a case that poses the issue more sharply, we might set up alternatives of
roughly the same sort and attemp t t o brin g them into balance. Walzer's cas e
involves doing active harm i n order to sav e lives, but consider a possible case
from sp y fiction i n which someon e ha s stumble d across plans fo r a politica l
assassination an d is threatened with the murder of a family member unless she
remains silent. Here the agent may be made out as choosing between two fail -
ures to save a lifeon th e assumption that she is all set to inform o n the assas-
sins when they reveal that the family member is under threat and will be killed
unless she holds back. Each of her alternativesfailing to save the family mem-
10 Between the Horns

ber and failing to save the assassin's intended victimwould seem to be forbid-
den b y th e mora l rules . B y adjusting amounts o f har m tha t woul d ensu e
and similar morally significant features of the case, we could presumably make
both alternatives come out as about equall y wrong on the view that allows for
dilemmasor not wrong after all , apparently, on the view favored by Brandt
and Hare.
At issue in this "balanced" assassinatio n case , on m y understanding of di-
lemmas, is whether the term "wrong" makes sense apart from a comparison of
options i n which at leas t one act comes ou t a s right. Proponents o f dilemmas
interpret "wrong " as applying to som e property o f acts that i s not essentially
comparative, and the phenomenology of moral experience seems to bear them
out. Though in some cases the availability of a better alternative may be all that
makes us count an act as wrong, that is not our initial reason for holding in the
assassination case that it would be wrong to allow a murder. It also seems that,
whatever the agen t decide s to d o here , i t will be reasonable fo r he r t o fee l a
sense of guilt for allowing a murder. On the other hand, the logic of moral dis-
course ma y seem to requir e taking the comparativ e poin t abou t "wrong " a s
understood an d dismissing guilt feelings as inappropriate where an agent makes
the best choice possible under the circumstances .
I shall eventually argue, in the chapters that follow, that guilt can indeed be
made out a s appropriate i n such casesnot jus t an understandabl e spillover
from mor e normal cases but rather a warranted reactio n t o whichever act the
agent decides to do. This account will yield a way of making sense of the pos -
sibility of dilemmas, for n o compariso n o f alternatives seems to b e built into
the clai m that a virtuous agent would vie w some ac t a s a stai n o n her mora l
record. Just a s every one of a set of exhaustive alternatives can b e ugly or up -
setting or in some other way repugnant, so there is no strictly logical problem
in supposing that al l of an agent's alternatives warrant guilt.
There might seem to be a logical problem with conflicting ought-judgments
themselves, but i n the contemporary debat e ove r dilemma s to b e outlined i n
the present chapter this at any rate has been shown to be false: It begs the ques-
tion at issue in dilemmas to suppose that the prohibition of an act implies that
some alternativ e to i t is not prohibiteda s we would nee d to sa y in order t o
derive a straightforward logical contradiction fro m the prohibition o f all alter-
natives. However, even if dilemmas involve no contradiction, the y seem to make
impossible demand s of an agent . I n warranting guil t for unavoidabl e wrong,
they subjec t a n agent t o "mora l luck, " meanin g moral responsibilit y for fac -
tors beyon d hi s control.4 More fundamentally , they seem to pos e a threat t o
the very intelligibility of ethics as a system of rules meant to guide action. It is
unclear how moral judgments can be thought o f as telling an agent what to d o
in a case of practical deadlock, where everythin g he can do i s forbidden.
In this chapter, I hope to brin g out som e more particular question s raise d
by dilemmas for the view of ethics as essentially action-guiding. I take these t o
be variants of a historical problem about the motivational force of obligation .
In fact, I shall suggest that dilemmas also reveal the main lines of an answer t o
these more general metaethical questions in the role they assign to guilt. Let me
Defusing Dilemmas 1 1

first fil l i n some of the philosophical background on dilemmas and attempt t o

show how the issue bears on metaethics (section 1). I shall then be in a position
to giv e an ide a of my own intende d argument and it s implications for ethics
and mora l psychology (section 2). In general terms, I think we can use dilem-
mas to exhibi t the motivational structure of an ethic s that i n some sense rests
on psycholog y yet escapes between the horns of some standard way s of mak-
ing out th e relation between the two. It does so on my account by reference to
the social role and origin s of moral motivation .

1. Mora l Dilemmas an d Motivationa l Forc e

Dilemmas first surfac e a s problems for the coherency o f action-guiding ethics

in the modern literature, with its emphasis since Kant on notions of obligation.
Aquinas had earlier restricted conflic t to cases of prior wrongdoing on the part
of th e agent , apparently in answer t o question s o f moral luc k raised b y Gre-
gory the Great' s view of dilemmas as traps set by the devil. 5 If the devi l could
trap a n innocen t agent int o doin g wrong , then presumabl y not al l evildoers
would deserve eternal punishment. But the claim that all alternatives are wrong
is not thus problematic as applied to an author o f prior evil, already consigned
to hell by a just god .
Instead, th e clai m seem s t o rais e problem s fo r th e vie w of a n agen t a s
"bound" to act by obligation independentl y of some sanction like the threat of
divine punishment. Thus, Kant denied the possibility of moral dilemmas, seen
as conflictin g obligations, o n a notio n o f obligatio n a s "mora l necessity." 6
Though it might make sense to say of Aquinas's evil agent in a dilemma that he
has to do two incompatible acts in order to get to heavenat least as a way of
saying that heave n i s by now beyon d hi s reachthe clai m tha t h e "ha s to "
simpliciter would just seem incoherent. Such metaphors of necessity refer to obe-
dience t o la w as a kin d o f compulsion, an d fo r Kan t they also have a literal
application t o the holy or perfectly rational will, which is moral by nature. The
moral la w simply describes the natural behavior o f a holy will, so any obliga-
tions derived from it on Kantian assumptions must be capable of joint fulfillment.
Moral philosopher s afte r Kan t allow fo r an d ofte n emphasiz e conflict of
obligations, bu t not of a sort that clearly involves dilemmas. In most cases dilem-
mas are ruled out: Ross, for instance, limits conflict to prima faci e duties, only
one of which in any conflicting set can amount t o a n actual duty in the sense
that implies really being in force.7 The suggestion tha t there might be genuine
dilemmas seems to have emerged only in recent years as a product of the attempt
to formalize Kantian assumptions abou t moral necessit y in deontic logic .
Deontic logic , th e logi c of obligation, wa s se t up o n th e mode l o f alethi c
modal logic, the logic of necessity and possibility, with moral requirement taken
as analogous to necessary truth except for its failure to imply truth; but a num-
ber of paradoxical consequences immediately raised questions about the anal-
ogy.8 Undermining it opened the door to the possibility of dilemmas, not as a
further problem for deontic logicexcept insofar as the standard system lacked
12 Between the Horns

the resources t o capture dilemmasbut rather as an insight into the nature of

the ethic s o f duty , th e moder n approac h t o ethic s exemplifie d by Kant. Th e
suggestion was picked up by moral philosophers and interpreted a s in some way
problematic for ethics, but with disagreement about the extent to which it under-
mined ethical rationality. Let us take a look a t the highlights of that debate , in
enough detail for a reconception o f the problem .

Dilemmas in Contemporary Duty Ethics

Dilemmas were introduced into the nontechnical literature by a logician, E. J.
Lemmon, in an article published in The Philosophical Review in 1962.9 Lemmon
understood dilemma s as cases in which an agent both ought an d ought no t t o
do the same thing. In typical cases he took the m to be explained by the deriva-
tion o f oughts fro m thre e differen t sources : dutie s (base d on th e agent's posi -
tion or status), obligations (base d instead on acts of commitment), and general
moral principles. Thus , in Plato's well-known case of weapons borrowe d fro m
someone wh o then goes mad and demands their return, the agent has to choos e
between fulfillin g a n obligatio n based o n his promise an d satisfyin g th e mor e
general obligation t o prevent harm. 10
Lemmon treats Plato's case as resolvable by appeal t o a higher-order utili-
tarian principle, but he also brings up Sartre's case of a young man in occupied
France who ha s to choose betwee n joining the resistance an d stayin g home t o
support his dependent mother. 11 This case apparently involves a choice between
two ought s o f the same sort: duties based o n the agent's differen t role s as son
and citizen. Lemmon treats the evidence bearing on the choice as inconclusive,
but th e cas e provide s a t leas t the framewor k for a full-blown dilemma to th e
extent that it is not clearly decidable by appeal to the agent's preexisting moral
attitudes. I n a full-blow n dilemm a o n Lemmon' s account , th e agen t ha s t o
develop a new moral outlook i n deciding what to dowhich i n this case would
seem to mea n identifyin g himsel f with one of his conflicting roles .
Lemmon's accoun t leave s a number of questions unanswered : It is unclear,
in particular, whether the choice of a new moral outloo k i s supposed to resolve
a dilemma or to leav e it in force. At any rate, Lemmo n make s out the proble m
raised b y dilemma s a s a proble m fo r th e adequac y o f genera l philosophica l
approaches t o ethics : approaches h e sees as tailored t o "th e easy , rule-guided
moral situation."12 In strictly logical terms, he takes dilemmas to be unproblem-
atic, a s h e think s ca n b e show n b y contrasting "ought " wit h "must. " Tha t
"must" and "mus t not " ar e incompatible he takes to follo w fro m th e fact (or
presumed fact; I shall question it in chapter 2) that "must" implies "will." "H e
will" and "he will not" canno t bot h be true, so neither can "He must" an d "H e
must not." Bu t the argument does not apply to "ought": The possibility of vio-
lating an ought amount s to the basic point of disanalogy betwee n deonti c and
alethic modal logic .
The next major discussion of dilemmas occurs in Bernard Williams's defense
of th e consistenc y of conflicting oughts , initially published in the Proceedings
of th e Aristotelian Society in 1965.13 Williams argues that dilemmas violate stan-
Defusing Dilemmas 1 3

dard deonti c assumption s t o th e exten t tha t the y force a choice betwee n th e

principle that "ought" implies "can" an d a principle of deontic distribution that
he refers to as "agglomeration" an d decides to dropin contrast to Lemmon ,
who instead rejects "ought"-implies-"can."14 With O as an operator for "ought"
and A and B as variables describing possible acts, we may symboliz e agglom -
eration as : OA and O B imply O(A & B) . In dilemmatic cases, wher e A and B
are assumed to be incompatible, it will not be possible to satisfy their conjunc-
tion, so O(A & B ) will violate "ought"-implies-"can. "
Williams's articl e ha s bee n widely disputed i n the ethic s literatur e fo r hi s
apparent argumen t to th e existence of dilemmas from mora l emotio n an d fo r
the metaethical conclusions he uses dilemmas to support. H e sets up dilemmas
a bit differently fro m Lemmon , focusing on cases of contingent conflict, where
the several things one ought to do are not logically incompatible but are jointly
unrealizable in the worl d a s it happens to be . The case Williams presents a s a
full-blown dilemma is one in which there may be no uncertainty about what t o
do, assuming that one ought to do what is "for the best": the case of Aeschylus'
Agamemnon, who as military commander has to sacrifice his daughter Iphigenia
to secure the success of the Gree k fleet. 15
Despite its false religious assumptions, Agamemnon's case is well chosen t o
illustrate the basis of dilemmas in moral emotion. Having killed his daughter ,
Agamemnon i s expected t o fee l ba d abou t hi s actio n eve n i n th e absenc e o f
reasons fo r doub t abou t whethe r i t was fo r th e best . Hi s conflic t canno t b e
resolved "without remainder " insofar as it leaves a kind of affective residue that
Williams identifie s as regre t an d defend s as a rationall y appropriat e mora l
reaction. Wha t William s wants t o sa y on thi s basi s is that th e "ought " tha t
Agamemnon has to violate is still in force rather than bein g canceled or quali-
fied in the way that ethical theories often assume . Agamemnon's conflict there-
fore resemble s a conflic t o f desires , i n which th e rejecte d elemen t doe s no t
simply vanish but ma y reappea r i n affective formo r sometimes wit h a dif-
ferent object , as in cases involving obligations to make u p fo r a violation of
In mor e genera l terms , th e cas e o f Agamemnon i s supposed t o exhibi t a
disanalogy between ought-judgments and belief s that Williams takes in a later
article as undermining moral realism, understood a s involving commitment t o
an independent reality that makes moral judgments true or false.16 On Williams's
account, in contrast to a conflict of desires, a conflict of belief s is decided by
eliminating one of the conflicting elements. The conflict between beliefs means
that one of them must be wrong, unlike ought-judgments, which may both re-
main in force despite a conflict.
This point against moral realism has received critical attention from Philippa
Foot an d othe r authors 17 a s the upsho t o f Williams's treatment o f dilemmas .
But dilemma s surface often i n Williams's writings, alon g wit h othe r case s of
conflictin discussions of utilitarianism, "dirty hands" i n political life, and in-
commensurable valuesapparently as basic ethical data with divers e implica-
tions.18 I n the firs t instance , they are used just to provide an extreme contras t
to cases of conflict i n which one o f the conflicting ought s is canceled.
14 Between the Horns

On Williams's view, dilemmatic oughts are neither canceled nor overridden .

He later distinguishe s an intermediat e sort o f case that als o involves a moral
remainder, albeit one characterize d somewha t differentl y i n terms of feeling .
In contrast to dilemmaswhich are now set up negatively, as "tragic" cases in
which all alternatives are wrongthere ar e fairly common cases of utilitarian
trade-offs in political life in which right action still incurs a "moral cost." Though
the agent does not do wrong simpliciter, h e has to wrong someonethere is a
victim o f hi s action , wh o ha s a justifie d complaintan d hi s ac t retain s a n
"uncancelled moral disagreeableness," reflecte d in an appropriate reactio n o f
disquiet or distaste. 19 A deeper form of regret is apparently reserved for dilem-
matic cases, thoug h William s stops shor t o f ascribin g guilt to th e agen t i n a
Guilt may be thought to b e irrational in a case where, through no faul t of
his own, the agent cannot avoid doing wrong, on the assumption that the reac-
tion can b e justified onl y if one i s guilty. However, som e later essays by logi-
cians do defend guilt, in different ways , as the appropriate reaction to dilemmas.
Bas van Fraasse n take s appropriat e feeling s o f guil t t o depen d o n actua l o r
objective guilt, but he holds that guilt for the unavoidable is shown to be coherent
in objective terms by the doctrine of original sin; Ruth Marcus expresses doubts
about this but notes that th e objec t of guilt in a dilemma, the particular alter-
native tha t i s rejected, is not unavoidabl e in itself . Van Fraasse n i s one o f a
number o f authors who follo w Lemmon in attributing dilemmas to multipl e
sources or grounds of obligation, extendin g Lemmon's account with a notio n
of incommensurable values or reasons, whereas Marcus defends the possibility
of single-principle conflicts at leas t for deontological conception s of the right.
Van Fraassen takes dilemmas to call into question presuppositions of standard
deontic logic, seen as including an assumptio n linking "ought" to what i s for
the best. 20
A more detailed account in terms of incommensurability is provided by Nagel
in a discussion identifying fiv e fundamental types of valueand ultimately the
clash betwee n agent- an d outcome-centere d standpoint s o f evaluationas
sources o f dilemma. 21 On NageP s view, dilemmas undermine the unifiabilit y
of ethics conceived as the search for general principles, as opposed t o a "frag-
mentary" approac h t o th e subjec t relyin g on the exercise of judgment in par-
ticular cases.
However, dilemma s attributable t o multipl e standpoints, o r ground s of
obligation or social roles, could be handled by keying oughts to these differen t
sources, as in Hector-Neri Castaneda' s alternativ e system of deontic logic with
its subscripted version of O. 22 S o it is important tha t Nage l briefl y allow s fo r
conflicts within his several categories23 and henc e between unqualified (o r a t
any rate, similarly qualified) ought-judgments. The common reaction that dilem-
mas as thus understood woul d mak e ethics in some way incoherent o r incon -
sistent gets its first sustained argument in an article by Terrance McConnell. 24
McConnell cites John Rawl s along with Castaneda an d David Lyons to illus -
trate the widespread assumption that a n adequate moral theory must exclude
dilemmas; Marcus adds Donald Davidson to the list of well-known contempo-
Defusing Dilemmas 1 5

rary philosophers wh o treat conflictin g oughts a s evidence of a contradictio n

in the existing moral code.25 But against these authors Marcus defends the code
that yields dilemmas as indeed consistent .
For our purposes, wha t McConnell picks out as his second sens e of ethical
inconsistency is what is relevant: On th e standard assumptio n that a n obliga -
tion to do B entails an obligation to do whatever B requires, dilemmas involve
commitment to bot h O A and O~A . This i s to sa y that Williams' s contingen t
notion of dilemmas yields dilemmas in Lemmon's sense, in which the same act
is both required and forbiddena strange consequence that violates some highly
intuitive principles of standard deontic logic , thoug h i t may no t b e logically
questionable in itself.
On the other hand, Marcus essentially argues that the general rules that give
rise to contingent dilemmas need not b e inconsistent even in practical terms, so
that ethics conceived as a system of rules is not undermine d by the derivation
of incompatible directives. She does so by putting forth a definition of consis-
tency for rules that count s as consistent an y set of rules that can be obeyed in
all circumstances in some possible world, eve n if not in the actual world.26 Ou r
moral code will not be deficient as a guide to action in the actual world as long
as it meets this weaker requirement of consistency, since Marcus takes the code
to include a second-order regulativ e principle enjoining th e avoidance of con-
flict. This is the action required of us in dilemmatic cases, and feeling s of guilt
are justified b y the role they play in motivating ita t any rate in future cases ,
or before a given dilemma becomes unavoidable. Assuming that violating any
obligation incurs at least a small burden of guilta need to make explanations
and excusesMarcu s extend s th e accoun t o f dilemmas in terms of unerase d
obligations beyond those cases Williams acknowledges, t o include even trivial
ought-conflicts. The result would see m to b e a view of dilemmas and guil t as
pervading the moral life .
In defense o f a Kantian approach t o ethics, however, Alan Donagan refor -
mulates McConnell's charg e of practical inconsistenc y agains t proponents o f
dilemma who rejec t the principle of agglomerationincluding all authors afte r
Lemmon in this overviewby comparing th e ultimate moral authority on any
such accoun t t o Captai n Quee g i n Th e Caine Mutiny. 27 Captai n Quee g wa s
found mentall y incompetent, partl y on the grounds tha t he issued conflicting
orders. By the same token, moralit y would b e "absurd," even if not inconsis -
tent, i f it subjected us to conflicting commands. O n Donagan's rationalis t as -
sumptions, then , we may reject a moral syste m as "il l constructed" i f its pre -
cepts canno t b e agglomerated. Donaga n see s Marcu s a s instea d modifyin g
"ought"-implies-"can," sinc e she restricts the principle to first-order rules, but
as thus interpreted her appeal to the second-order regulativ e principle that tells
us to avoid conflict involves treating morality as an unreachable ideal.28 Essen -
tially, then, our choice between the two principles, "ought"-implies-"can " and
agglomeration, amount s to a choice betwee n th e action-guidin g function of
ethics and it s rationality or coherency as a product o f human thought.
The bulk of Donagan's argumen t against dilemmas amounts to an attemp t
to account for the facts of moral conflict within his rationalist preconceptions .
16 Between th e Horns

Like McConnell (followin g Aquinas) , he exempts the evildoer' s self-impose d

conflicts a s no threat t o ethical consistency, bu t h e handles other case s either
by challenging the validity of one of the conflicting considerations or b y deny-
ing that th e choice betwee n them can b e moral. Othe r author s hav e added a
positive accoun t o f th e emotiona l fact s o f conflict; the bes t know n i s Hare's
two-level utilitarian explanation in terms of general habits of emotional response
instilled i n u s as th e mos t efficien t wa y o f motivatin g mora l behavio r under
normal circumstances. 29 But the main charge against dilemmas seems to be some
version of incoherency of the sort illustrated by Donagan's Captai n Queeg case.
There are other, more specialized treatments of dilemma.30 I have presented
only highlights of the debate in order t o exhibit the central problem dilemma s
seem to raise for moral theory. In the first instance, it is a problem about whether
"ought" can play the strong action-guiding role, as a vehicle for expressing moral
commands, tha t appear s to be assigned to it by the modern ethics of duty. For
the assumptio n behin d the charg e of practical incoherenc y is that conflicting
moral judgments cannot mak e sense if taken at fac e valu e as telling the agen t
what to do, in contrast t o their interpretation i n the cases Aquinas accepted, as
essentially serving to punish an agent for prior wrong .
That the derivation of conflicting commands amounts to a perfectly intelli-
gible foul-up of an ethical system or its fit to th e worldone tha t ma y not b e
fully attributabl e to problem s in assessing value seems t o b e an overlooke d
alternative. Indeed, as I shall go on to indicate, even Williams fails to leave room
for thi s possibility . Marcus's view makes room fo r somethin g lik e it, bu t de -
clines to acknowledge it as a foul-up. I take it to be based on a view of ethics as
a necessaril y imperfect human product, bu t no more incoheren t than a hypo-
thetical pinbal l machine tha t registers "tilt " unde r circumstance s othe r tha n
player error. This results in a nasty surprise, as on the view of dilemmas as traps
of th e devil . On m y own account , though , dilemma s ar e explained b y limita-
tions of the mechanism rather tha n b y fiendish intention .
My overvie w here als o omit s some o f the meatie r case s o f dilemma: news-
worthy cases of "moral blackmail" on the model o f the choice between lettin g
hostages b e killed and encouraging further terroris m b y negotiating for their re-
lease, and variou s other real-lif e an d literary cases such as (in one example) the
woman's choice between family duty and duty to self in Ibsen's A Doll's House. 31
I focus instead throughout this essay on variants of a few cases dealt with at length
in the philosophical literature, plus some rather streamlined additions like the case
of Captai n Queeg , t o illustrat e the problem dilemma s seem to pose for action -
guiding duty ethics. I shall later argue that it is a soluble problem, i f we make a
number of important distinction s on the understanding of action-guiding status.
At thi s point, however, I want t o identif y i t a s a problem an d t o sho w ho w i t
bears on more general concerns i n the history of moral philosophy.

The Problem for Practical "Ought"

Donagan uses the Captain Queeg case against a command-based view of ethics
like van Fraassen's . Bu t it is worth notin g that a mora l legislato r responsible
Defusing Dilemmas 1 7

for dilemmas need not come out with straightforwardly conflicting commands
to swab an d no t t o swa b the deck, saybu t only with ought-judgment s that
imply such commands, give n further fact s about th e world and a strong inter -
pretation o f "ought." We would be less likely to question Queeg' s sanity or legal
competencemore likel y to attribute the conflict to change of mind, forgetful -
ness, an d simila r norma l menta l imperfectionsif he had issue d two contin -
gently conflicting commands, as on Williams's account o f dilemmas. The con-
flict migh t even be rather obvious : Imagine a harried mother' s commands t o
her children to clean up and (i n the same breath) to kee p still.
Donagan's analogy seems to mix together differen t sort s of criticism, then.
I want t o disentangl e on e o f them, th e on e correspondin g t o th e charg e of
incoherency, for discussion in what follows. His statement of the Captain Queeg
analogy in terms o f commands mask s a distinction between irrationalit y and
unreasonableness: If Queeg had simply come out with conflicting requirements
requirements for adequate performance in a certain rank, say (perhaps as crite-
ria for promotion)the fact that it would be impossible to act on both of them
might be seen as canceling the natural interpretation o f his statements as com-
mands. Issuing them would still undermine his authority t o some extent, since
it would show him to be an impossible person to satisfy. But it is another ques-
tion whethe r hi s statements mak e no sense , as one migh t want t o sa y of th e
conflicting command s issue d by the mother i n my example .
Of course, morality or the moral code might be said to be in a special position,
since it is not subject to the mental limitationsof knowledge and memory, of
change in perspective over timethat characterize human legislators and judges.
Morality presumabl y means what i t says, along wit h al l the consequence s of
what i t says, at an y given time. Similarly for God. Bu t Donagan need s the sor t
of distinction just illustrated in order to make an exception of Aquinas's dilem-
mas fo r th e evildoer . Ho w woul d Go d b e cleared o f incoherency i n his com -
mands at a given timewhere that implie s a failure t o mak e sense, or irratio-
nalityby the fact that the conflict between them is attributable to the agent's
prior moral error ? The agent i s punished in such cases wit h unreasonabl e re-
quirementsor with requirements that would be unreasonable if he were inno-
cent. Imposin g the m o n a n innocen t perso n i n full-fledge d case s o f dilemma
would presumably support an objection from unfairness , or moral luck, which
Donagan an d other s might see as involving an incoherency i n our moral con -
cepts or in the nature and purposes attributed to God as a moral judge. So this
more general incoherency rather than incoherency in the conflicting commands
themselves seems to b e what i s in question.
However, what if one thinks of ethics or moralityI shall use the two terms
interchangeablyas primarily an instrument of social rationality, eithe r man-
made o r designe d wit h huma n limitation s in mind? Dilemmas migh t the n be
explained a s side effect s o f the pursui t of perfectly coherent genera l purpose s
with a moral code tailored to reflec t thes e limitations, and we might deny any
possibility of appeal beyond them to som e more ultimat e level of moral trut h
that woul d resolv e the conflict. 32 With moral luc k thus accepte d a s a fac t o f
ethical life, dilemmas may still be said to undermine the authority of the moral
18 Between the Horns

code t o som e exten t b y revealing its fallibility a s a guide to choice . Bu t it is a

further ste p t o th e charg e tha t i t is incompetent t o legislate , on th e mode l of
Captain Queeg .
Williams is one author who seems to be eager to make room fo r moral luck,
and in recent writings he attempts to undermine the notion of moral blame ; yet
even he denies all-things-considered or "conclusive" practica l status to the ought
that i s not acted o n in a dilemma.33 This ma y surprise some readers; h e is usu-
ally and naturally interpreted a s holding that dilemmas involve action-guiding
oughts.34 I attempt t o she d ligh t on thi s issu e in chapter 2 . Fo r th e moment ,
however, le t us just note tha t othe r putativ e sense s of "ought, " distinguished
for other purposes, migh t be called into service to capture dilemmas but do not
seem to b e adequate t o captur e their problemati c aspect . A merely classifica -
tory ough t tha t migh t b e thought o f as "critical " o r "judgmental"labelin g
alternatives to a certain action as wrong but not actually telling the agent what
to dowould no t seem to capture the sense in which an agent i n the grips of a
dilemma is motivationally "torn. " Moreover, variou s weake r bu t stil l action-
guiding senses of "ought" might be distinguished from the strong or imperatival
sense that yields commands; thes e include commendatory or ideal "ought" and
a prim a faci e o r other "ought " tha t record s a commitment o r othe r practica l
reason but without fina l judgment as to whether it requires action. Just because
they are in themselves relatively inert in motivational terms, though, these sub-
stitutes for conclusive practical "ought" seem to yield too easy a picture of dil-
emmatic choice .
This problem i s brought ou t in its sharpest for m by what I call "balanced "
dilemmas, wher e th e alternative s in questio n ar e abou t equall y wrongan d
seriously so, enough to justif y takin g both o f the conflicting oughts a s conclu-
sive i n a moral sense . O f cours e the y cannot bot h determin e action , nor ca n
they coherently be meant to d o so in conjunction. This i s essentially why Wil-
liams denies them conclusiv e practical status . Bu t his view here seem s t o cu t
against the characterization o f dilemmas on Nagel's account , say , a s cases in
which there is "decisive support" for incompatible alternatives: If an ought tha t
seems to express "decisive and sufficient" reason s for action35 has practical or
motivational force at all, how can the rational agent who accepts i t forgo actio n
on i t in favor of its competitor?
Williams cannot mean to say merely that its competitor i n fact wins the day.
We shall have to as k what h e does (or should) mean by inquiring into some of
the notions of deontic compariso n tha t come up in an attempt a t explanation .
More generally , though, we need to ask whether it is possible to accommodat e
dilemmas within a coheren t motivationa l pictur e of the mora l "ought. " Th e
question that seems to emerge from the debate over the rationality of two com -
peting (an d in some sens e conclusive) practica l ought s ca n b e made ou t a s a
new version of an old question in moral philosophy about the motivational forc e
of obligation: How can the reason that "binds " a n agent to the performance of
an obligatory act b e seen as compelling action?
The question for dilemmas is how "compulsive " moral motivation can pull
in opposing directions. We might think of motivational force in terms of vectors,
Defusing Dilemmas 1 9

which i n dilemmatic cases apparentl y ar e no t cancele d o r eve n weakened b y

opposition. In mathematical terms, they do not combine to yield a single prod-
uct. To retain their problematic aspectt o defuse dilemmas as a threat to ethics
rather than merely debunk or deflate themwe need to retain this difficult moti -
vational property. It is unclear how we can do this, however, o n standard ac -
counts of the relation of ethics to psychology.
In motivational terms, standar d account s divide into those tha t mak e ou t
the motivational force of moral judgments as dependent on some extracognitiv e
psychological state of the agent, typically desire, and those that insist that belie f
is sufficien t t o generat e the necessar y motivation for action . Account s o f th e
latter sor t migh t b e said t o mak e mora l motivatio n "subject-independent, "
meaning that its source is independent of the particular mind that holds a moral
judgment (rather than minds generally). Subject-independent accounts are given
by authors who would be classified in contemporary terminology as "internalis t
realists" (sometime s "cognitivists"). Thes e authors hold, that is, both that the
motivation t o ac t o n a mora l judgmen t is implied b y its meaning, so tha t a
rational agent who holds it necessarily acts on it (internalism), and that mora l
judgments describe some subject-independent facts about the world (realism).36
Accounts of motivational force that mak e it subject-dependent, on th e othe r
hand, are given by authors who deny either one of these positionsholding either
that moral motivation is provided b y something besides the content o f a mora l
judgment (externalism) or that moral judgments either have no subject-indepen-
dent content o r are false (antirealism).
These terms are not without problems , bu t on the assumption tha t Hum e
falls int o th e externalist-or-antirealis t categor y insofa r as h e gives a subject-
dependent accoun t o f moral motivation , th e mai n positions o n motivationa l
force can be illustrated by the contrast betwee n his view and Kant's. 37 Because
of its link to obligation , which for Hume is secondary to virtue, the notion o f
motivational forc e comes u p mor e explicitl y in Kant, thoug h Hume' s tal k of
practical forc e an d o f the dependenc e of morality on sentimen t (including de-
sire) can be understood a s contrasting wit h Kant's position. Roughl y speaking,
then, I want t o say that th e choice betwee n Humean an d Kantian, or subject-
dependent and subject-independent, approaches t o moral motivation amount s
to a choice between making dilemmas implausibly easy on the agent in motiva-
tional terms and making them hard t o the point o f impossibility.
Let us begin with Kant, whose positio n i s more clear-cut . I t is summed u p
in general terms by his claim, shortly before he rules out conflicting obligations,
that "i n discussing practical law s of reason we do not take [moral] feeling into
account, since it does not concern th e ground of these laws but only the subjec-
tive effect whic h they have on ou r mind." 38 H e goes on t o give a version o f a
common objectio n to emotion-base d theorie s of ethicswhat I shall cal l th e
charge of "subject-relativity"i n view of the fac t that emotiona l motivatio n
varies from on e mind to the next. However , fo r Kantand in the general his-
torical tradition of discussion of motivational force, which goes back to natu -
ral la w theoriststal k of motivation is not limite d to subjectiv e causatio n o f
the sort the objection questions, but rather refers in the first instance to the deter-
20 Between the Horns

mination t o ac t b y considerations o f reason. 39 The poin t fo r ou r purpose s i s

that motivational forc e in this sense, if built into the meaning of a moral judg-
ment, would indeed seem to rule out belief in conflicting judgments on the part
of an agent who is fully rational. To hold bot h of two ought-judgments known
to be in conflict would be to attempt t o act on both of them and hence to attemp t
to do what on e knew to b e impossible.
At any rate, this holds for "conclusive" ought-judgments , on an all-or-nothing
interpretation o f their motivational force. But on Kantia n cognitivist assump-
tions, th e connection betwee n belief in and actio n o n a given ought-judgment
would not seem to allow for any independent variation in degree. So a less than
firm tendenc y t o act , o n th e par t o f a rationa l agen t (rationa l in a sens e tha t
rules out weaknes s of will), would have to b e explained in terms of some simi-
larly qualified beliefno t the sort of conclusive or "all-things-considered" judg-
ment we have in a full-blown dilemma.
Do dilemma s se t u p i n term s o f practica l "ought " far e an y bette r o n a
Humean o r other subject-dependen t approach? Her e we have a wider range of
possibilities to consider , bu t i f we limit ourselves for th e moment to standar d
forms o f subject-dependence, it is hard t o se e how the y can yield an adequat e
motivational pictur e o f dilemmas . What i s in question , le t u s assume , i s th e
strength o f an agent's desire s to ac t on eac h o f the ought s i n conflict. Subject-
dependent views divide, though, over the question whether this connection t o
desire is in some sense given in the content o f an ought-judgment. At this point,
let us assume that it is, in line with the standard readin g of Hume as an internalist
antirealist. Now , i f an agent were full y attune d i n these terms to bot h side s of
a conflict, the problem just outlined for subject-independent views would see m
to apply here too. S o instead we need to assume that the agent's motivation t o
act on one of the oughts in conflict is weak enough to permit action on the other.
In that case, on our current understanding of the content o f an ought-judgment,
at least one of the judgments in question woul d itsel f be weakened accordingly
to a prima faci e ought-judgment , and w e would los e the sense of "all-things -
considered" conflic t that accounts fo r the motivational difficult y o f dilemmas.
This is to say that dilemmas become to o unproblemati c o n a standard sor t
of Humean approach . The y apparently ar e assimilate d to case s of prima facie
ought-conflict. But the resultant picture of choice in a dilemma seems in a certain
sense to b e too easy : It is as if Agamemnon o r the agen t i n my balanced assas -
sination cas e i n section 1 had simpl y to weig h u p pros and con s an d decide ,
flipping a coin o r th e lik e to brea k an y ties , wit h variou s compensatory act s
and feeling s see n as calle d fo r b y the ac t thu s chose n bu t wit h an y opposin g
motives under th e circumstances canceled out . Th e agent ough t t o fee l some -
thing like horror a t the choice, perhaps; bu t this and similar moral reactions on
the standar d pictur e seem to b e felt mor e o r les s on the side , not a s part o f his
current motivation to take action but rather as contemplative responses to action
on the model o f an aesthetic reaction .
In fact , I shall go o n t o sugges t ways o f modifyin g this picture in defens e
of a roughly Humean approach, thoug h one whose conten t I shall understand
as subject-independent , in a sens e sufficien t t o answe r th e charg e of subject -
Defusing Dilemmas 2 1

relativity. At this point, however, I want to blur over some distinctions I make
later, to set up the problem my argument is meant to address. In short, the stan-
dard approache s availabl e in the literature to handl e moral dilemma s fac e u s
with a metaethica l dilemma: They make dilemmas either too problematic , i n
the sens e o f bein g practicall y difficult t o th e poin t o f impossibility , o r to o
unproblematic: too easy on the agent in motivational terms to capture th e dif-
ficulty o f dilemmatic choice.
The second horn of this metaethical dilemma covers several distinguishable
sorts of inadequacies in accounts of dilemma. As a further instance of it, suppose
we now try the other Humean view I allowed for and think of subject-dependent
motivational forc e a s extrinsi c to th e meanin g o f a n ought-judgment . Thi s
amounts in contemporary terms to a shift fro m internalis m to externalism. But
it would see m to attribut e to the agent i n a dilemma the kind of motivational
detachment fro m a moral judgment that i n extreme form amounts to motiva-
tional "amoralism " (sometime s called accidie)the lac k of any inclination to
do what one believe s to b e requiredhere explained jus t b y the impossibility
of satisfying all requirements. On the resulting account, dilemmati c oughts could
both b e recognized as all-things-considered, bu t w e would hav e to den y that
they could bot h b e practicalmeant to guide action, that is to say, as used by
the agen t wh o ex hypothesi hold s both o f them.
The agent might still feel morally torn between alternatives, but the point is
that this motivational effect woul d no longer b e attributable to the judgments
he applies to the case. It seems to be an accidental effect, dependen t on what a
particular agen t happen s t o feel , an d henc e to b e vulnerable to th e charg e of
subject-relativity. Even if we grant that a normal agent will feel pulled in bot h
directions i n the situation of dilemmatic choiceor i n Humean terms that i t is
part of human nature to feel tha t waywe cannot conclud e that such feeling s
are appropriate t o the situation or in some way called for by it.40 This versio n
of the Humean account , then, seems not to have the resources t o represent the
authoritative status of moral claims, as imposing requirements upon minds rather
than merel y allowing for whatever desire s mind s happen t o have. 41 So again
(but in another sense ) dilemmas seem to b e made too easy .
What if we respond to these difficulties i n the obvious way, by rejecting the
possibility of dilemmas? In fact, I think that problems similar to those just noted
can be seen to arise in another for m for an attempt to capture the psychologi -
cal phenomena of ought-conflict without recognizing dilemmas. Consider Hare's
attempt t o explai n the agent's reaction s i n cases o f conflict as appropriate t o
more norma l sort s o f cases, cases covered by the simpl e rules we learn a s chil-
dren an d generally find adequat e to the moral life , though w e also can appea l
beyond them to utilitarian considerations wher e the y conflict. In the rare situ-
ation of conflict on Hare's account, guilt, remorse, and similar moral reaction s
to wron g would actuall y be inappropriate i n representational termson e of
the agent's options woul d no t really be wrongbut th e feelings would stil l be
valuable as signs of a good mora l upbringing. 42 Indeed, some for m o f moral
distress a t what he has to d o seems to b e required of the agent in a dilemma if
we are to think well of him.
22 Between the Horns

That las t clai m ma y see m t o b e born e ou t b y Aeschylus ' treatmen t o f

Agamemnon, for instance. 43 I n general, Hare's account of the feelings expected
of a moral agent in many ways resembles my own intended account i n this essay,
but hi s metaethical presuppositions yiel d a different vie w of conflicta view I
take to be inadequate in the end to capture even the subjective aspect of dilem-
mas across th e board . T o se e this, le t us first as k how th e emotiona l require -
ments Hare would impos e on the agent in a case of conflict are supposed to fit
into his utilitarianism. Hare's discussion of moral education suggest s a justifi -
cation o f emotion a s a precondition o f personal virtue, but hi s overall view is
put fort h as a two-level version of act-utilitarianism. Emotions woul d seem to
come out, i n that case, as required only in light of the generally good effect s of
inculcating virtue. But it does not seem obvious that this approach woul d yield
a requiremen t o f emotio n i n al l cases o f putativ e dilemmaor specificall y a
requirement to fee l guilt or som e similar negative self-directed emotion o f the
sort that th e relevant cases assume .
We would no t b e satisfied, that is , by an Agamemnon wh o reacte d simply
with horror or some other form of anxiety at his action viewed externally, with-
out evaluativ e focus on hi s own rol e i n it a s agent . W e see m to deman d th e
negative self-focu s of guilt and relate d emotionsa t a minimum , Williams's
"agent-regret"in respons e to a seriously wrong act.44 But it is unclear whether
we can always justify this in utilitarian terms, in part just because the situatio n
of dilemm a i s so rare . A reasonable se t o f utilitaria n rules prescribin g moral
feelings woul d presumabl y make exceptions fo r cases tha t ca n b e seen not t o
support th e usual role of feeling as a goad t o moral behavior . Bu t there might
have been no reason to think that Agamemnon or anyone affected b y his actions
would be likely to suffer specificall y as a result of his failure to feel guiltenough
so, at any rate, to justif y th e unpleasantness of that emotio n as something re-
quired above and beyon d an y public atonement he might offer .
The mos t Hare's account ca n provide , i t seems, i n the wa y o f a require-
ment t o fee l guilt y in suc h cases i s another simpl e rule o n th e leve l o f thos e
that ar e supersede d b y act-utilitarian calculations i n cases o f dilemma. Th e
requirement to feel would itsel f be superseded, then, in the even t of conflict
as might well occur in a case lik e Agamemnon's, wher e guil t feelings woul d
be likely to undermin e the performance o f act-utilitarian obligations . More-
over, thes e remark s appl y no t jus t to agent s lik e Agamemno n wh o happe n
not t o fee l what is generally required of agents who hav e acte d similarl y but
also t o som e agent s wh o d o mee t th e norma l expectations . I t i s sometime s
possible, tha t is, to tal k onesel f ou t o f feeling guilty in cases wher e the emo -
tion is understood t o b e just an unfortunate side effect o f oversimplified rules
learned i n childhood bu t late r refine d so as not to apply . I n such cases, par t
of wha t on e doe s to ge t rid o f the feelin g i s to poin t t o th e evidenc e for it s
inappropriatenesswhere tha t mean s it s failure t o represen t accuratel y th e
particular situatio n a t hand . O n the sort of view Hare recommends, w e ap-
parently lose the sense that guilt or the like is a correct respons e to the agent' s
situation i n th e sens e that implie s accurate representation o f itwhat I call
rational (a s distinc t fro m mora l o r social ) appropriateness.45 S o where th e
Defusing Dilemmas 2 3

moral requiremen t to fee l i s superseded by act-utilitarian considerations, a n

agent ough t simpl y to forg o guil t if he can manag e it.
In my own argumen t I assume that th e rational appropriatenes s o f guilt is
in question unles s otherwise noted . I shall later have much more t o sa y about
the emotional requirements of cases like Agamemnon's, which might be thought
of a s yieldin g a subjectiv e notion o f dilemmas . Subjectivel y speaking , dilem -
mas involve the appropriateness o f guilt for all alternatives, taking guilt broadly
at thi s point t o includ e remorse an d simila r reactions t o a n ac t thought o f as
wrong, thoug h I later mak e some more detailed distinctions. Apar t fro m thi s
subjective sense, however, I do not intend to argue here for the existence of dilem-
mas. Instead, my central question is whether dilemmas as cases of unavoidable
wrong threate n the coherency o f ethics. To address that question, I shall often
assume in what follows that dilemmas in this objective sense exist, but the reader
is free t o conditionaliz e my argument, taking i t as an indicatio n o f the prob -
lems to which acceptance o f dilemmas would give rise.
In the first instance, then, my aim here is to escape the metaethical dilemm a
set up in this section by showing how a reasonably authoritative conception o f
ethics can coherently allow for dilemmas, without making them motivationally
flaccid o r in some other wa y implausibly undemanding. I shall not attemp t t o
convert th e reader t o a particular vie w on th e questio n o f dilemma s bu t just
to provide a sensible rationale for such a view. This is also my aim with respec t
to the broader questions, such as that of emotional appropriateness, tha t come
up in connection with my treatment of dilemmas. Since these broader question s
are important to my argumentthey eventually will displace dilemmas, in fact,
as its central focusthey require separate discussion .

2. Motivatin g Moral "Ought"

The general problem of motivational force was posed i n contemporary term s by

Elizabeth Anscombe in an argument for the abandonment of the specifically moral
sense of "ought" in modern dut y ethics.46 Anscomb e favored an interpretatio n
derived from th e Aristotelian ethics of virtue, and a contemporary mov e back to
Aristotle's approach, with evaluations of persons and personal traits replacing act-
requirements as the primary ethical judgments, now seems t o b e in full swing. 47
The problem just posed fo r dilemmas as practical conflict s fits int o this trend in
the literatur e to the exten t tha t it cuts agains t a strong action-guidin g sens e of
"ought." Anscombe's argument calls into question the moral sense of "ought" on
the grounds that i t is at this point merely "emphatic": It rests on a mere appear -
ance of force left over from the earlier use of the word on natural law accounts in
connection wit h divine judgment.48 Withou t th e view of ethics as based on laws
promulgated by God, s o that motivational force could be understood concretel y
in terms of a threat of God' s displeasure or o f punishment, there i s nothing t o
support th e heavy stress laid on "ought" a s a distinctively moral notion .
Moral dilemma s can now b e seen as providing a logically vivid illustration
of th e impossibl e demands o n th e mora l "ought " accordin g t o thi s vie w of
24 Between the Horns

modern ethics . Th e problem the y raise fo r the practical coherenc y o f ethics

taken a s somethin g ove r an d abov e problem s o f moral luckresult s from a
secular Kantian conception o f obligation as moral necessity. A strong interpre -
tation of "ought" in terms of divine punishment, though it would of course raise
questions abou t God' s justice, might i n principle be handled b y hesitating t o
attribute to God our notion of individual responsibility. O n the other hand, a
weaker secular interpretatio n a s merely judgmental or at any rate not conclu -
sively practicalas referring simply to a commitment, say, or to liability to social
condemnationwould mak e dilemma s in a sens e to o easy , i f my precedin g
argument is correct. I t is emphatic "ought " that i s in question here, too, bu t in
its action-guiding role, as providing the motivational force that Anscombe dis-
misses a s "purel y psychological " (p . 41), no w tha t th e ter m itsel f ha s bee n
drained of content .
The extended argument that follows amounts to a defense of "ought" against
the motivationa l proble m raise d b y dilemmas along with som e othe r case s i n
metaethical dispute. But rather than attempting to answer Anscombe's challeng e
by supplying the missing content o f "ought," I shall suggest a two-componen t
view of moral meanin g that separate s th e questio n o f the meanin g of a given
judgment from the question o f its motivational force. My argument on motiva-
tional forc e will allow fo r differen t interpretation s o f the conten t o f "ought "
rather than pinning it down precisely , though I shall also give an indicatio n of
my ow n favore d vie w if only t o exhibi t it s departure fro m Anscomb e o n th e
connection o f "ought" to natura l law .
Anscombe's claim that the motivational forc e of "ought" is undermined by
the abandonment of belief in natural law, the belief supporting its original inter-
pretation, assume s tha t motivationa l forc e i s determined b y meaning . Thi s
amounts to a version of internalism. Internalism seems to have emerged, in fact,
during the seventeenth- and eighteenth-century transformations o f the natural
law approach withi n British moral philosophy . Motivational force was initially
understood a s something like a causally effective version of what Philippa Foot
calls "reason-giving " force. 49 A moral judgmen t has reason-giving force for a
certain agen t i f it supplie s her wit h a reaso n fo r doin g wha t i t prescribes a
reason tha t ha s to make sense in light of her interests and desires , though sh e
need not therefore be motivated b y them to act on it. On natural law accounts ,
for instance , perhaps we could say that th e threat of divine punishment would
count as a reason for obeying the law, even though some basically rational agents
with insufficien t fea r o f divine punishment might no t b e motivated t o d o so ,
instead choosing t o incur the punishment . This allow s tha t a rational agent' s
desires and othe r motivatin g states suc h as emotions ma y fail t o reflect he r in-
terestsin other words, tha t ther e ma y be a gap between accepting an ought -
judgment and bein g motivated t o act on it.
However, the gap is closed by a different wa y of understanding motivational
force tha t seem s to hav e become standar d i n Anglo-American moral philoso -
phy b y the tim e we ge t t o th e contemporar y debat e ove r internalis m versus
externalism. The shift i s presupposed b y eighteenth-century moral philosophy,
especially "moral sense" theories, which began to focus on emotion as the critical
Defusing Dilemmas 25

factor in moral motivation, though this was soon overshadowed b y an empha-

sis on its role in moral knowledge. 50 In effect, reason-givin g force was made t o
depend on motivational force rather than the other way around: It was benevo-
lence or conscience or some other internal motive that supplied the agent with
a reason for action o n a moral judgment. But the reversal made it seem plau-
sible to take holding a moral judgment as implying motivation t o act on it.
Hence Anscombe finds "ought " empty o f content i f it lacks motivational
force; on the other hand, she assumes that its force must be more than psycho -
logical i n order t o captur e th e authorit y w e ascribe t o mora l judgment . She
quickly disposes of Butler, for instance, o n the grounds that he does not mak e
room fo r an immoral conscience.51 Her argument is essentially a version of the
Kantian objectio n fro m subject-relativity , but a central purpos e o f m y ow n
argument in what follows is to assign emotion a role in modern duty ethics that
escapes reasonable forms of this objection.
I shall approach the problem of motivational forc e through dilemmas, tak-
ing dilemmas as a way of posing the problem i n sharp form. My ultimate aim,
as I have noted, is not to convert the reader to my own view of dilemmas; rather,
I hope to use dilemmas as a way of exhibiting the connections among a number
of central issues in metaethicstaking the term broadly to refer to the study of
the foundations of ethics.52 I think we can establish a new angle of view on this
subject, awa y fro m it s previous focus on metaphysical , epistemological, an d
semantical issues, by assigning a central position to issues in moral psychology,
as raised by the question of dilemmas. I shall begin, then, in chapter 2 by con-
sidering some fairly narrowly defined questions about dilemmas and moral rea-
soning in connection with deontic logic, but my concerns wil l soon branc h ou t
from tha t starting point .
In effect, this essay will have three main topics: dilemmas, guilt, and metaethics
or, more specifically, the role of emotion i n the foundations of ethics, which
I take to be that of supplying motivational force. My initial focus will be rela-
tively narrow, with dilemmas serving as a way in and guil t as a bridge to my
ultimate topic, the role of emotion i n ethics. Similarly for authors considered:
I shall begin with Williams as my central exampl e since his work cover s th e
various topics I want to bring together, thoug h he often fails to connect them
clearly or to go as far as I would like on key issues. On two issues concerning
dilemmaswhether they are conflicts between conclusive practical oughts and
whether guil t as opposed t o som e weake r for m o f regret i s appropriate fo r
the ought not acte d onm y own position comes closer t o some of the other
authors discusse d in my initial overview, especiall y Nagel an d Marcus .
Once my argument turns to general metaethical issues, as it will fairly quickly
on the basi s of the opposition William s sets u p between dilemma s and mora l
realism, I shall discuss some other author s whos e overal l approaches stan d i n
sharp contrast to Williams's: Philippa Foot and John McDowell. J. L. Mackie's
view, which is one of the main contemporary examples of antirealism, provides
both a contrast t o McDowel l (i n a contemporary versio n of the Hume/Kan t
contrast in the preceding section) and some materials for constructing the sort
of two-component realist view that I want to defend here as able to handle dilem-
26 Between the Horns

mas. I refer to this view as "social artifac t realism." I t is realist at any rate in a
somewhat extende d sens e that seem s to fit current definitions.
The view essentially puts social rationality in the role Anscombe assigns to
a divine lawgiver. A moral code or other normative system (for simplicity's sake,
I shall mainly speak i n terms of a code o f rules ) i s something man-mad e bu t
subject t o constraint s dictate d b y social end s and th e natur e and function s of
societyas i f the standard Aristotelian arguments abou t virtu e were modified
to appl y to groups. It s content i s therefore not fixe d b y the rules a group hap-
pens to havethe norm-based alternative that Anscombe at one point considers
and rejectsbu t it is real in the way that artifact s are real, though thei r origi-
nal existenc e is of course dependent on minds. It is also imperfecta s a prod-
uct of human imperfection in general terms, bu t more specificall y becaus e it is
subject t o limitation s in the ordinar y huma n capacit y t o maste r rules . Fo r a
fundamental socia l constraint o n the cod e i s the requiremen t of teachability.
The mora l cod e o n this accoun t i s not unlik e what Har e describe s as the
simpler set of rules that we ordinarily go by, but i t is not supersede d i n moral
terms whe n i t yields a conflict . Rather, case s o f conflict involve real motiva -
tional deadlock , indicatin g a breakdown o f the moral code that m y own view
will tr y t o explai n a s a sid e effec t o f the mechanis m that allow s th e cod e t o
function properl y in normal cases. For what make s the rules action-guiding in
general is the sort of internal goad to action tha t i s provided by teaching them
in conjunction with guilt and similar moral emotions. Thi s motivational role of
emotion favor s what I call a "perspectival " notio n o f appropriateness tha t i s
weaker tha n the one Hare take s for granted, sinc e it assesses emotions as war-
ranted relativ e to a partial subset of the total bod y of evidence in light of which
we would asses s a corresponding belief . Thus, guilt ma y be appropriate a s a
feeling even where the agent is not objectivel y at fault all things considered, a s
we assume i n a case of dilemma.
I shall not give a detailed account of the nature and justification of the moral
code i n what follows , since I want m y argument t o appl y to variou s differen t
approaches to ethics . Those contemporary approache s I am familia r wit h all
seem to allo w fo r dilemmas at least on something like Hare's "intuitive" leve l
of simple rulesthough sometimes with a higher level deus ex machina (whether
act-utilitarianism o r God himself) brough t i n to sav e the day. O n a rule-utili-
tarian view, for instance, the rules that would have the best consequences over-
all if generally adhered t o presumabl y must b e consistent, bu t the y might still
come into contingent conflict as a result of some unusual event or in cases where
the assumption o f general adherence is not satisfied. On a divine command view,
we could get similar results by thinking of God as a kind of moral watchmaker
whose command s amoun t to a code o f rules meant t o serve for all time rather
than specific directives, without a n option o f direct appea l where they conflict.
And eve n on a Kantian view, applyin g the categorica l imperativ e t o a se t of
maxims simple enough to yield teachable rule s might also be held to generate
Such theories need not treat dilemmas as an embarrassment. Rather, dilem-
mas amount t o th e sorts o f exceptions that ca n be said to "prove " the moral
Defusing Dilemmas 27

rules. They exhibit the general mechanism of action-guidance b y blocking its

usual direct effects, displayin g a residue of guilt or some similar negative moral
emotion that usuall y functions i n anticipatory form, before action, t o ensure
compliance with the rules. This emotion is what yields the appearance of bind-
ing force in cases where compliance is reluctant.
The discussion of guilt, then, will connect my initial topic, moral dilemmas,
with th e genera l issue I expect i t to illuminate : the rol e o f emotio n i n mora l
motivation. Through most of this argument, I understand "guilt" rather broadly,
using the term at the outset to cover various distinguishable emotions suc h as
shame and remorse, an d even later for the most part ignoring it s religious and
psychoanalytic associations. N o doubt in reaction to one or both of the latter,
the ver y idea of guilt is an objec t of amusement or aversio n i n some quarters ,
and ver y few mainstream moral philosophers make mention o f it. But I think
there is something to b e said in defense o f the emotionor , rather , th e emo -
tional mechanism. For guil t o n th e vie w I shall defend here, eve n guilt in th e
narrow sense , i s not reall y a single emotion bu t a tendency to tak e on various
different identificator y emotions involving a negative self-evaluation.
Within philosophy, this view gets support from Jonathan Edwards' s explana-
tion of conscience in his 1755 treatis e on ethics, and i t is also borne out b y re-
cent psychological studies of guilt.53 In fact, the term "guilt " seem s not to have
been used as an emotion term at all until the late sixteenth centuryan d then
only in error, as a substitute for "sens e o f guilt," with "guilt " on it s accepted
use taken as referring to an objective state of affairs.54 Th e religious notion, a t
least in the firs t instance , involves an extraemotional stat e o f the self a vari -
ant of primitive ideas of "tainting," analogous to a disease that will spread unless
one takes steps t o preven t it, as spelled out fo r guil t by rules of ritual atone -
ment.55 Atonement as originally conceived might or might not involve a notion
of some feeling one ought to havefear of God's wrath, say, or contrition. But
the gradual internalization of religious focus associate d particularl y with th e
period of the Reformation seems to have led to the idea of a general feeling o r
sense of guilt as the appropriate response to wrongdoing. 56 Late r I attempt t o
show how the emotiona l mechanism I equate with (subjective ) guilt preserves
some of these historical associations .
I shall also have to say something about the pitfalls of the emotion, an d a t
least by implication those of emotional response generally, since guilt in some
of its manifestations seems to represent an extreme case of emotional uncontrol .
The widespread view of guilt as a personally destructive psychological force
a mechanism of social control that essentially inflicts damage on the individual
is due to Freud' s influenc e and . among philosopher s ma y b e traced bac k t o
Nietzsche.57 Eve n within the psychoanalytic literature, however , on e can fin d
the defense of a psychologically beneficial form of guilt in the work of Melanie
Klein.58 My own defens e o f the emotion i n common sens e terms will focus on
the social bases that give it value as a moral motivator .
My interpretation of guilt as an identificatory mechanis m will allow for its
attribution beyond Judeo-Christian religious culture and it s historical heirs to
cultures that lack our emphasis on guilt, sometimes thought of by anthropolo-
28 Between th e Horns

gists followin g Ruth Benedic t as "shame-cultures." 59 I shal l eventually argue

that guilt has some advantages over shame and other related emotions as a source
of moral motivation. To some extent, however, the central role I assign to guilt
is an accidenta l function o f my focus o n th e proble m of moral dilemmas . My
detailed accoun t o f it s role i n moral motivatio n i s intended as exemplary a
way into the general question of the role of emotion in ethics. It is a moral (or
specifically deontic) emotion par excellence, but it shares with other emotion s
on m y account a motivational functio n tha t stand s i n contrast to th e percep -
tual analog y applied t o emotion s o n othe r emotion-base d bu t nondismissive
accounts o f ethics. That is, Hume and other moral "sentimentalists " see m to
see moral emotions as recording evaluative information (or possibly misinfor-
mation) about the world. The standard contrasting accountbesides emotivism
and similar dismissive accountsis Mill's utilitarian treatment of "internal sanc-
tions," which assigns emotions a motivational rol e without effect o n the con-
tent of moral judgment.60 I hope in what follows to exhibit a further metaethica l
option, one that gets in between the horns of the standard alternative s with its
focus on the social standpoint o f evaluation. I shall do so by digging deeply into
the motivational question s raised by dilemmas.
Practical Oughts
and Prohibitions

Let us first tur n to th e problems raised by dilemmas for the logical principles
governing ought-judgment. Williams's early article focused attention on the clash
between the principle that "ought" implies "can" an d the principle of agglom-
erationthat OA and OB imply O(A & B)bot h of which seem intuitively to
characterize a strong action-guiding sense of "ought." The same can be said of
two other principles of standard deontic logic that later authors have brought
into conflict on the assumption that dilemmas exist: the principle that "ought"
implies "permissible" and the principle of deontic closure .
The latter principle essentially tells us that anythin g necessary to fulfil l a n
obligation is itself obligatory; with ~M as the alethic modal operator fo r possi-
bility, we may symbolize this as: OA and ~M( A 8c ~B) imply OB.1 Some such
principle along with agglomeration would see m to b e needed to suppor t th e
derivation of one ought-statement from another and hence the systematic project
of deontic logic . On the other hand, the two principles governing the implica-
tions of the term "ought" seem to be needed to support a notion of prescriptive
"ought," taken as an ought that is both practical and positive, or meant to tell
an agent what to do. We seem to be forced to choose, then, between the sys-
tematic aims of deontic logic and its relevance to action-guiding ethics.
In fact, I think that somethin g like this will turn ou t to b e true. However ,
we need not deal here in full detai l with the problems raised by the two pairs of
principles withi n deontic logic . Instead , le t us loo k a t som e centra l deonti c
notions an d principle s in application t o mora l reasoning , wit h deonti c logi c
understood as a failed attempt at systematization whose grounds for failure may
be illuminating. I shall occasionally use the resources o f deontic logic to sym-
bolize the principles and other assumptions unde r scrutiny; but the main thing
I expec t m y argument on thi s subjec t t o revea l i s that a n attemp t t o handl e
dilemmas b y working ou t som e alternativ e deonti c syste m would a t bes t b e
impossibly complicated.
I shall begin by focusing on the problem posed in chapter 1 for conflicts be-
tween practical oughts (sectio n 1). The principle that "ought " implie s "can "
would seem to be defensible by appeal to the notion of a practical ought as one
that is intended to guide action: What would be the point, in short, of trying to
30 Between the Horns

guide an agent i n a direction he cannot go ? Yet Williams's explanation o f con-

clusive practical "ought " apparentl y make s the notion inapplicabl e to dilem-
mas. It has to apply in some form, however, in order to make sense of our rea -
soning about th e fulfillmen t o f conflicting oughts at times before they actually
come into conflict.
My argument involves a closer look at the interpretation of practical "ought"
and relate d notions , includin g notions o f deontic weigh t suc h a s "all-things -
considered." I n section 21 raise some more general questions about the picture
of the logical structure of ought-conflict that is presupposed b y deontic logic in
common with much of contemporary mora l philosophy, as derived from Ross's
account o f the balancin g of prima faci e duties . Amon g other things, I hope t o
bring out a way o f assessing th e comparative strength o f oughts tha t favor s a
negative characterization o f dilemmas in terms of prohibitions, or action-guiding
judgments of moral wrong , rather tha n positiv e ought-judgments. In short, I
argue that acceptin g th e principle tha t "ought " implie s "permissible" mean s
taking positive ought-conflicts as merely prima facieand to that extent grant-
ing Williams's pointwherea s cases of exhaustive prohibition ma y stil l come
out a s dilemmatic in the fulles t sense .
The defens e of negative action-guiding dilemmas is the mai n resul t of my
argument in this chapter. In response t o the literatur e on dilemmas, however ,
the secon d hal f o f the chapte r turn s t o somewha t mor e technica l discussion .
For those who prefer to shortcut detail s this is summed up in a final subsectio n
beginning o n p . 62 . M y ful l argumen t i n section 3 focuses on th e change s i n
deontic logi c an d ultimatel y in our pictur e of moral reasonin g tha t woul d b e
needed to accommodate dilemmas on my account. To let the operator O cover
our ordinary ought-judgments, with "ought " taken in a strong action-guiding
sense, I retain the two ought-implication principles "ought"-implies-"can" an d
"ought"-implies-"permissible." I n light of my treatment of dilemmas, the latte r
principle is limited to positive ought-judgments, so closure can also be retained
with appropriat e limitations . O n th e othe r hand , I attempt t o sho w ho w we
might indee d fin d ground s fo r dropping th e othe r ought-derivation principle ,
agglomeration, an d wha t result s the change would hav e for the logica l struc-
ture of ethics. Broadly speaking, on the view that will emerge here, ethics comes
out a s fragmentedas Nagel put s it in his treatment o f dilemmas i n terms of
incommensurable values . My ow n accoun t wil l involve a specificall y deonti c
form o f fragmentation, a splintering into practica l subsystems; but on e line of
attack on the coherency of ethics as thus construed will be met by showing th e
rationale behind th e denia l of agglomeration .

1. Practica l Ought s i n Conflic t

In this section I want to defend the practical status of dilemmatic oughts agains t
some implications of Williams's view.2 I take it that denyin g the practica l sta-
tus of dilemmas would leave us without the aspect of motivational conflict tha t
we need in order to capture what is troubling about them as problems of moral
Practical Oughts and Prohibitions 3 1

choice. Firs t I focus on a particular argument that seems to show tha t practica l
force has to be attributed to dilemmatic oughts in order to account fo r the force
of som e ought s derive d fro m them: ought s prescribin g actions neede d t o pre -
pare t o satisf y th e firs t set , o r wha t I shall call "preparatory " oughts . At any
rate, this conclusion follows from a natural understanding of practical forc e in
terms of meaning. I go on to defend that interpretation a s an alternative to tak-
ing practica l forc e a s a functio n of th e speaker' s intentions . I then begi n t o
respond t o the various reasons m y discussion bring s to ligh t for denying "all -
things-considered" statu s to conflicting practical oughts .
Throughout thi s argument , I assume tha t ther e ar e o r a t an y rat e ca n b e
genuine dilemmasmeaning (for purposes of the present discussion) dilemmas
of th e "balanced " variet y i n which th e tw o ought s i n conflict ar e o f roughly
equal weight. I also assume that the cases under consideration involv e "time -
bound" oughts , i n contras t t o th e timeles s obligation s sometime s take n fo r
granted i n discussions of deontic logic. 3 This means essentially that whether a
given ought i s in force depends o n when it is evaluateda time that need no t
be the sam e as the dat e (implici t o r explicit ) on it s object, or wha t i t tells the
agent he ought to do. If yesterday I promised, fo r instance, to do act A tomor-
row, it is already true today that I ought to do A tomorrow. In my central argu-
ment in this section I hope to show that these assumptions favor an interpreta -
tion o f dilemmatic "ought" as practical, in order to make sense of the advance
deliberation that fulfillin g i t may entail.

Deliberation in Dilemmas
Consider th e cas e o f dilemm a se t u p b y Sartre: Th e ma n wh o mus t choos e
between joining the French resistance and staying home to support hi s depen-
dent mother might very well represent both of his options a s things he ought t o
do. That is, we can imagine him prescribing each of them, as he weighs the rele-
vant considerations. O f course he would not com e out with a conjoint ought -
statement prescribin g both, but we can follow Williams and othe r author s on
dilemmas in allowing for a distinction here by rejecting the principle of agglom -
eration. However , on e might b e moved to as k how th e agent i n Sartre's case
could coherently prescribe even each of his options without a change of mind
in the same breath, as it weregiven that his combined prescriptions would be
Should we avoid this problem b y representing the oughts i n conflict as not
really meant to elicit action an d i n that sense not practical oughts? Perhaps we
might characteriz e dilemma s instea d i n term s o f ought s tha t simpl y classif y
various possibl e act s wit h respec t t o mora l reasons , tellin g us that ther e ar e
decisive moral reasons fo r eac h o f two exclusiv e alternatives rather tha n pre -
suming to tel l the agen t wha t t o d o unde r th e circumstances . However , thi s
would leave out the kind of active motivational conflict that moral reasons seem
to generate in typical cases of dilemma. It would make dilemmas too easy : We
would los e the sens e of the rationa l agent as practically "torn," o r subjec t t o
contrary motivationa l vectors, t o th e exten t tha t sh e appreciates the reason s
32 Between the Horns

bearing on her choice of action. Instead , th e case would b e assimilated to one

in which an agent could not decide what to do, at any rate on moral grounds
a case in which she was intellectually torn, though not becaus e of any defect in
her practical reasoning .
We would als o seem to lack the resources to capture a n agent's earlie r rea-
soning i n cases o f dilemma . That is , we sometime s hav e t o deriv e practica l
directives from dilemmatic ought-statements befor e they come into conflict
for instance, where their fulfillment require s advance preparation. I f the oughts
in a dilemmatic pair were not practical, however , i t is unclear how they could
support th e derivation o f practical preparatory oughts . A n attempt t o take the
dilemmatic oughts as practical only at an earlier time when they yielded such
preparatory ought s withou t conflict would seem to undermin e the very poin t
of practical forc e by letting it lapse unaccountably before the time the ought -
statement i n question assign s to action .
This is my argument for dilemmatic practical "ought" in a nutshell; now let
me illustrate it with a version of Sartre's case . For the agent t o be able to sup -
port hi s mother financially , say, in 1942taking a simple-minded view of the
sort of support tha t is in question here just for purposes o f easy illustrationit
might be necessary for hi m to star t savin g in 1932 . S o in 193 2 w e apparentl y
could deriv e a claim that h e ought t o sav e some money. 4 Bu t then on e o f the
ought-statements tha t conflict in 1942 woul d seem to involve a practical ought
at least from the standpoint of 1932. Moreover , i f we say that its practical force
somehow lapse s b y 1942, tha t woul d see m to undermin e any practical forc e
ascribed to i t earlier. Presumably, the reason fo r the earlier advic e was just to
enable the agen t to suppor t hi s mother later , bu t supportin g he r i n 194 2 re -
quires further action a t that time . What woul d b e the point i n practical term s
of giving advice that will just be withdrawn before the time to act on it arrives?
Someone raisin g objections to thi s argumen t might defen d the notio n o f
earlier practical force that simply lapses as familiar enough in other cases: What
about obligation s tha t becom e unfulfillabl e befor e th e tim e o f action , say
assuming a tensed version of the principle that "ought" implies "can" a s apply-
ing to practica l "ought" ? However , sinc e "can " i n the principle implie s tha t
fulfillment i s possible through th e agent' s ow n efforts , I would repl y that th e
obligations i n question her e are those tha t th e agen t essentiall y makes unful -
fillable by doing or failing to do something earlier, as opposed t o any that lapse
because of events he cannot prevent, including actions of other agents. For in-
stance, we might suppose that our agent's obligatio n to support hi s mother in
1942 would lapse in practical force before then if he did not save money in 1932 ,
since on our hypothesi s the later obligation is fulfillable onl y with ten years of
savings behind him. Here a practical ough t lapses b y the tim e assigned to ac -
tion becaus e the agen t fail s t o ac t i n light of it a t earlie r times. In the cas e of
dilemma, by contrast, o n the assumption of lapsed practical force, it would seem
that the agent could follow our advice to the letter until the time for action arrives
and stil l have it withdrawn at th e last minute.
My argument can also be extended to meet various more sophisticated ob-
jections. I do not wan t to pause for a full treatment , but i t is worth noting that
Practical Oughts and Prohibitions 3 3

the oughts the argument turns on are not supposed to be relative to the agent's
state of knowledge, in particular whether the dilemma can be foreseen. For in-
stance, someone might suggest that we could explain the practical force of the
1932 ought s in terms of some sort of epistemic claim about the 194 2 oughts
about thei r probable practica l force, say, relative to what i s known i n 1932
without taking the 1942 ought s to be practical a s well. However, i t is not clear
why the agen t actually (an d not jus t probably) ought to prepar e t o satisf y a n
ought tha t is merely probable.
The claim that an ought is practical, though, whic h I interpreted abov e as a
claim that it is meant to elicit action, naturally suggests that it is offered a s advice
to a given agent (possibly by the agen t himself) , s o that it s force would vary
with contextual factors including what i s known a t the time. However, unless
we are concerned with the agent's blameworthiness for failing to act (with sub-
jective rather than objective "ought," in the usual terminology), a practical ought
is assumed to hold whether or not the agent knows or has reason to think that
it does. Whether an ought is practical or "action-guiding" i n this sense depends
only o n ho w i t i s intended, not o n whethe r circumstance s ar e suc h tha t th e
intention is fulfillable, includin g whether it gets through to the agent, as required
for actuall y guiding his action. In order to cover oughts in the third person w e
might think of this as an ought that is "suited to " action-guidance .
The problem with dilemmatic oughts, it seems, is with the speaker's state of
knowledge: The fac t that h e knows that two o f his intentions ar e no t jointly
fulfillable woul d seem to make a rational speaker retract one of them. My present
argument is not designed to answer this problem but to show that denying prac-
tical forc e to dilemmati c oughts doe s no t yiel d a satisfyin g answer , give n the
other thing s we want to sa y in such cases. M y answe r t o the problem i n later
chapters wil l involv e assessing th e rationalit y o f ethicso f a mora l cod e o r
system of norms (what actually gives rise to thes e oughts , even if in the delib-
erative voice of some individual speaker)in social terms. What on e is assess-
ing, in short, is something general: a set of general social rules or guidelines de-
signed to b e teachable o n the basi s of general emotional respons e tendencies.
This is the source of various specific practical directives that may not always be
rational considered in themselves, as utterances of some individual agent or other
The practica l force o f the dilemmati c oughts i n m y exampl e ca n thu s b e
understood a s derivative from th e moral syste m tha t yield s them. T o sa y that
they are meant to elici t action, then, i s ultimately to sa y something abou t th e
intent of the systemabout the role such a system assigns to moral judgments
rather than about the intentions of a particular speaker . I now want a t least
to allow for some of the larger points I have in mind by defending my interpre-
tation of practical force as something that is not simply supplied by the speaker .

Practical Force and Meaning

My argument above from preparator y "ought" seems to depen d on thinking
of practical "force" i n a linguistic sense, as a function o f meaning. This is what
34 Between the Horns

makes it puzzling to think o f a practical ought as derived from anothe r ough t

that is not practical: Where could it get its practical force except from the ought
it i s derived from? Bu t if practical o r action-guidin g force i s a function o f th e
speaker's intention , on e migh t wan t t o thin k o f i t a s contextual: a matte r of
what the speaker in a given case uses an ought-statement to do, which is a func-
tion of what he thinks it can do and varies with the circumstanceswith whether
the ought is fulfillable i n itself and whether it has competition a s in dilemmas.
On thi s accoun t a speake r supplie s practical forc e t o a n ough t tha t i s itsel f
motivationally inertso that h e might just supply it to a derivative ought like
those in my example. It is his "speech act" rathe r than the meaning of the term
that makes "ought" practical .
This account makes sense in light of all the cases in common languag e that
fail to obey the principle that "ought " implies "can." There are other senses of
"ought" i n play beside s the practicalmos t notabl y idea l "ought," which is
meant t o commen d some actio n o r stat e o f affairs . I n another versio n o f the
case from Sartre, for instance, w e might want to say that the agent ought to be
less attached t o hi s mother, without supposin g that h e has very much control
over his degree of attachment an d hence without meaning to get him to change
it. However, we need not assum e that these different use s of "ought" are dis-
tinct enough to count as different senses . We have something more general than
a particular speech act on a particular occasio n (a n utterance) to appeal to a
recognized use (meaning a mode of use rather than an instance) obeying logical
principles o f it s ownas a n alternativ e t o descriptiv e meaning. We may still
think o f thi s a s a for m o f "meaning " i n the wide r sens e of general linguistic
intent suggested a t the end of the preceding subsection.5 That is, a term may be
linked nonaccidentall y t o the pursuit of certain purposes b y a general role or
function (eve n i f one amon g several ) in th e language , in a wa y tha t control s
what a given speaker uses it to do rather than simply emerging from hi s speech
act. If we think of the language as already fixed b y past usage, we can say that
the term is "meant," in the sens e of "designed," to fulfil l a certain function .
This notion of a general recognized us e of a term, which itself has a kind of
functional meaning, will accommodate third-perso n practical oughts as instances
of practical "ought" that are not actuall y intended t o fulfil l it s defining func -
tion. What makes them practical, we want to say, is the fact that they are suited
to action-guidancecapable of eliciting action b y virtue of their general role in
the languagea s would no t b e the case, for instance , i f they were no t really
about action bu t rather were meant to commend some personal trait or state of
affairs. O f course , i t may sometime s b e indeterminate whethe r a given third-
person ought is an instance of practical "ought" in this sense. The only test would
seem to be a more general sort of appeal to the intentions of a given speaker
what h e would tr y to d o with i t in the presenc e o f the agent , say, at an y rat e
under ideal conditions (where interference would not be resented and so forth)
and speakers ' intention s ma y b e indeterminate. Bu t although the notio n of a
general use of the term is something we extract from such particular instances
much as we extract the notion of a language from a history of individual utter-
Practical Oughts and Prohibitions 3 5

ancesit may b e set up i n normative terms as something independent, a sys-

tem dictating correct us e on a given occasion .
This suggestion leaves it open that an ought may count a s practical even if
it is not i n fact use d to fulfil l th e definin g function o f practical "ought " o n a
given occasion , an d eve n if for extrinsi c reasons i t could no t b e so used . I t is
enough that i t be part o f a systematic use of the term that has that function i n
general. Whatever the limitations on our ability to tell when this is so, the point
gives u s a roug h argumen t fo r countin g dilemmati c oughts a s practical , just
insofar as they are logically linked to clear-cut cases of practical "ought" in the
way that makes them part o f the same general use of the term .
We also want to sa y more than this, for the practical force of the prepara -
tory ought in my example seemed to be something imposed on the agent or other
speaker, not somethin g he supplied, even by choosing a certain form of words.
To withhol d i t in light of the late r dilemma would hav e been an error , tanta-
mount to plumping for one side of the dilemma in advance. In effect, I argue in
chapter 4 that there is a way of taking dilemmatic oughts in most cases as prac-
tical in a fulle r sensealbei t an indirec t sense, i n which emotion ma y b e elic-
ited a s a substitute fo r actioneve n on th e leve l o f individual utterance. Fo r
the moment, however, let me reinforce my general suggestions on the meaning
of practical "ought" with some speculations on the origins of the notion .
One possible source is our talk of the practical "force" o f moral terms, used
more or less interchangeably with "motivational" force , though it refers to the
speaker as opposed to the agent. This usage traces back most notably to Hume' s
discussion in the Treatise of morality as practical, i n the sense of being meant
to influence action; the notion o f "force" tha t later authors hav e supplied here
seems essentially to combin e Hume's talk o f influence or impulsio n with th e
linguistic concerns o f our ow n times. 6 The lin k was effecte d b y noncognitivist
accounts of moral meaning in terms of practical function but i s by now a fea-
ture of the term s in which the metaethica l debate i s set up, eve n if one reject s
What Hum e ha d i n mind, however, was a branch o f philosophy tie d to a
certain facult y o f the mind : practical a s opposed t o speculativ e or theoretica l
reason. Bu t practical "ought " als o ha s a plausible interpretation i n terms of
practical reasons that might be thought to reflect this earlier notion rathe r than
the idea of "force" a s something a term might have by virtue of its function in
the language. Whether a given use of "ought" counts as a practical reason may
indeed seem to b e something that i s settled case by case, depending in the firs t
instance on the intentions of the speaker. Thus, Williams, taking the first-person
use as primary, understands a practical "ought" as one that plays a certain role
in deliberation; i t answers th e deliberative question, "Wha t ough t I to do?" 7
What i s in question here in denying this function to the oughts i n a dilemma is
whether the agent can reasonably use them to conclude deliberation unde r cer-
tain circumstanceswhere he cannot ac t on bothan d no t whethe r they are
oughts of a sort tha t play that role in general.
This understanding of practical "ought" makes sense in connection with the
36 Between the Horns

attempts o f some contemporary author s t o define "ought" in terms of reasons.8

However, th e intuitiv e appeal o f the denia l of practical "ought " i n dilemma s
may als o depen d o n th e commo n vie w o f a (conclusive ) practical reaso n a s
motivationally sufficienta s providing sufficien t reaso n fo r actio n i n causa l
termsassuming rationality.9 A practical reason o n this interpretation require s
nothing furthe r to produc e action . Th e interpretatio n come s ultimatel y fro m
Aristotle o n deliberation, especiall y his claim that the conclusion o f the practi -
cal syllogism is an action.10 We might want to say that a n ought tha t is practi-
cal i n this sens e i s one tha t woul d produc e actio n i n a rationa l agen t i f only
certain externa l condition s wer e met . I n tha t case , i f conflicting oughts wer e
both practical , the y would produc e incompatibl e actions . Sinc e that resul t is
impossible, i t follows that conflictin g oughts cannot bot h b e practical .
This argumen t rests o n taking "practical" a s meaning something lik e "suf-
ficient t o determin e the will to action, " wit h the phrase understoo d t o impl y
motivational effectivenes s i n the absenc e of external barriers. The phrase ech-
oes Kant; and the reading of "practical" seems to be borne out by Kant's iden -
tification o f practical reason wit h the will. 11 But I think we can see that it would
not surviv e the shift fro m Kantianand originally Aristoteliantalk of a gen-
eral facult y of practical reaso n t o th e contemporar y discussio n o f a practica l
reason an d o f ought-statement s a s practical . O f practica l reaso n i t migh t o f
course b e said tha t i t would no t b e the facult y i t i s unless it i n fac t produce s
action. Bu t it is a further step to mak e the same claim of a given practical rea -
son. To say that a n ought-statement i s practical or action-guiding need not be
to say that it actually motivates actio n i n a given case bu t just that i t is meant
toin a sense distinct from what it s speaker ha s i n mind, a s well as from th e
descriptive meaning of the statement .

"All Things Considered"

The notion o f motivational sufficienc y suggest s an idea of a practical ought a s
representing the output o f deliberation, or an "all-out" judgment , on the model
of the unconditiona l judgment in Davidson's treatmen t o f weakness o f will. 12
It may be natural to equate this with the all-things-considered or "all-in" ought -
judgment on Williams's account, 13 bu t Davidson interprets the latter a s condi-
tional. I t essentially sums up the evidence, or the inpu t of deliberation, on th e
assumption tha t al l th e fact s ar e in . Bu t i n betwee n inpu t an d outpu t fall s
the shado w o f redeliberation: deliberatio n from contrary reasons . I n cases of
dilemma, moreover, redeliberation need not be irrational. Assuming that dilem-
mas exist, then, a process of deliberation that is rationally sufficientsufficien t
to justify action and in that sense completeneed not be motivationally so, even
in a rational agent , i n the sense that implie s causal efficacy .
This conceptio n o f practica l reasonin g ca n b e reconciled wit h Aristotle' s
vieweven his view of the practical syllogism as entailing action, the source of
Davidson's unconditional reading of the "all-out" judgment . John Cooper , for
instance, treats the conclusion of practical reasoning as a decision to perform a
specific type o f action, with th e practica l syllogism a s a reconstruction of th e
Practical Oughts and Prohibitions 37

further perceptua l processe s neede d to produce a particular action. 14 It woul d

thus be possible to reach a conclusion of practical reasonin g or deliberation an d
yet fail to act, as happens in cases of weakness of will but also at least arguably
in som e case s of moral conflict. 15 The conclusio n of deliberation nee d not be
taken a s a here-and-no w judgmen t prescribing "this " actio n bu t rathe r a s
one that narrows thing s down t o a n act of a certain kind , with som e tim e lef t
before actio n fo r the agent to appl y or fai l t o apply the judgment to what lie s
before him. In that case, it seems, there will be room for contrary reasons to get
a grip.
On m y account , i t i s Kantia n necessitarianismwha t migh t b e calle d
"rational" necessitarianism, though in application t o moral judgments it yields
the notio n o f mora l necessitytha t stand s i n th e wa y o f makin g sens e of
dilemmas. In the present connection, Kantia n talk o f the will or practical rea -
son as "determining" actio n is often used without questio n eve n by opponent s
of Kantian ethics. I shall have more t o sa y later o n this general issue, but her e
I want to focus on the notion o f an "all-things-considered " practica l ought. In
response t o the argument just outlined, one might want t o sa y that th e claim
that deliberatio n yields an all-things-considere d conclusio n implie s a kin d of
rational completeness that rules out conflicting conclusions; it amounts to a claim
that there are no further reason s to take into account. So deliberation in a case
of dilemma may fail to yield an all-things-considered conclusion, bu t it canno t
yield two suc h conclusions.
What happens, then, in a case like Sartre's? One might be tempted to handle
the case by taking even an arbitrary resolution o f the conflict to turn on some -
thing like a choice of "projects" 16an implicitl y general act of self-legislation
with implications for the agent's future choicesthat could be represented as a
further step in practical reasoning. If the agent chooses t o stay with his mother ,
say, he will effectively b e committing himself to a future mode o f life contain -
ing further choices in harmony with that one. He will be ruling out options such
as quitting his job the following year to pursue his goals as an artist. It is impor-
tant, however, that this sort of deliberative conclusion would b e limited to the
first person , wit h the implication that other s could not prescrib e fo r the agen t
the choice h e is entitled to mak e for himsel f in the situation of dilemma. 17
Still, there seem to be cases unlike Sartre's cas e whose resolutio n doe s no t
or should not involv e a decision of principle, even an arbitrary decisio n wit h
implications for the future. Conside r the case of dilemma in the novel Sophie's
Choice, where Sophie is forced by a concentration cam p guard t o choose on e
of her children, lest both be taken to the gas chamber.18 Eve n if, morally speak-
ing, Sophie mus t choos e on e child to save , bu t supposin g tha t th e choic e of
her son ove r he r daughte r canno t reall y be justified b y appeal t o som e rele -
vant distinguishing feature of the sort she in fact relies on, the choice she makes
should not b e taken as constraining future choices. She is not henceforth com -
mitted t o actio n consisten t wit h i t such a s some sor t o f specia l attentio n t o
male childrenor for that matter , to action with the opposite tendency , as if
to compensate .
The poin t i s that no t ever y decision to act , an d henc e not ever y practical
38 Between the Horns

resolution of a dilemma, involves appeal to an ought-judgment. Thus, Williams

allows for an alternative version of the deliberative question, a s "What am I to
do?",19 which would seem to apply to decisions bearing on only one case. With
the deliberative question frame d i n terms of "ought," though, so that the con-
clusion of deliberation is supposed to b e an ought-judgmentwha t we might
think o f as a "principled " rathe r than a n arbitrar y answe r to th e deliberative
questionwhy shoul d w e no t sa y that Sophie' s deliberatio n terminates i n a
The denial that dilemmatic ought-statements can both hold all-things-con-
sidered is a standard response from opponents o f dilemma to cases like Sophie's .
What the y woul d sa y is that th e conflictin g ought-judgments a t issu e in th e
caseone prescribin g that sh e save her daughter an d th e othe r tha t sh e save
her sonar e no t fina l judgment s but merel y prima facie ; s o the cas e may b e
assimilated to Ross's cases of prima facie duties. The only duty that Sophie has
"all thing s considered" i s the disjunctiv e duty t o sav e either he r daughte r o r
her son, fo r it is of overriding importance tha t sh e save one of them, and under
the circumstances she cannot sav e both .
It is odd to fin d Williams in agreement on this point wit h the opponents of
dilemma, even if only with reference to practical oughts . In application to cases
of the sort I call "weighted," o n the model of Agamemnon's case, Williams wants
to say that neither ought in conflict is overridden: Though all things considered
it may be better to fulfil l on e of them, its fulfillment "doe s not adequately meet
the claim s involved in the conflict," 20 s o the othe r ough t remain s in force. In
balanced cases like Sophie's, then , where neither alternative is better, we would
seem to have all the more reason to say that th e two oughts in conflict are nei-
ther overriding nor overridden. 21
Williams briefly suggests a reason for denying that they are in force all things
considered whe n he notes tha t "th e proces s o f deliberation . . . involves nar-
rowing down , b y rejection, the answers t o 'Wha t ough t I to do?'" 22 O n thi s
account, a n all-things-considered answer would evidentl y be an answer base d
on ruling out all other answersall other alternative s to action, tha t iss o of
course ther e can b e only one . I n a case o f balanced dilemma , the onl y candi-
date fo r this status is the disjunctiv e ought that prescribe s fulfilling on e or th e
other of the two ought s i n conflict.
However, I think we should look again at this apparently trivial point. There
is another way of making out an all-things-considered answernamely, as one
based on ruling out all other answer s that ca n be ruled out. I f the evidence in a
given case does not support a unique answer, the n two answers to the delibera-
tive question ma y be said to hol d al l things considered. In light of all the evi-
dence, tha t isth e set o f "principled " appeal s availabl e to justif y a n ought -
judgmentwe seem to have sufficient reaso n fo r concluding both tha t Sophi e
ought t o save her daughter and that sh e ought t o sav e her son .
Both pieces of action-guidance are independently warranted, in shorta fac t
that would not be captured if we stopped a t the judgment that Sophie ought to
save one child or the other . Williams's explanation of why neither of the con -
flicting oughts in weighted cases counts as overriding also reveals the inadequacy
Practical Oughts and Prohibitions 3 9

of this stronger disjunctiv e ought. Presumably , each child has an independent

claim on it s mother to b e savednot just a claim for fai r consideratio n in the
choice of one to savethat the disjunctive ought fails to answer. Of course the
disjunctive ought may still be said to b e in force "all things considered," bu t it
would b e question-begging to assum e that i t i s in competition wit h the con -
flicting oughts for that status. So we have no real justification fo r letting it dis-
place them as deliberative conclusions. Thi s also holds fo r a n ought resulting
from a n arbitrary decision procedure that might be used to satisf y th e disjunc -
tive ought: the outcom e of a coin-flip, say.
What I think we ultimately need is another look at notions of deontic weight.
I shall attempt thi s in the next section, bu t a t this point we can already see at
least i n roug h outlin e how all-things-considere d status ma y appl y to eac h of
two conflicting practical ought-judgments. First, we need to recognize that even
reasons already acknowledged as bearing on a given case may be reconsidered
from anothe r evaluativ e standpoint. Ther e can b e different way s of assessing
reasons, tha t is , both o f which take i n all the reasons , a t leas t as background
considerations, if we allow for "gestalt shifts" determining which reasons stand
out agains t th e overal l background . Th e all-things-considere d evaluation of
action need not b e governed by a single standard o f what i s "for th e best, " on
the model that Williams applies to Agamemnon's case, 23 but can also appeal to
various independent standards, as Williams and othe r proponents o f dilemma
Second, w e shoul d not e tha t th e "core " meanin g o f a n ought-judgment
involves a negative evaluation a judgment that al l alternatives to actio n ar e
ruled outso that deliberation involves narrowing down one's options in a more
fundamental sens e than William s indicates. It does not jus t involve rejecting
answers to the question "What ought I to do?"; rather, an answer to this ques-
tion itself rest s o n rejecting the agent's other options . I f there can b e two such
answers, then , tha t togethe r would rul e out al l the agent's options, we would
have a case in which the same total body of reasons, assesse d accordin g to tw o
different standards , yield s conflicting negative all-things-considere d ought -
Let us return to Sartre's case for a simple example that allows for the plau-
sibility of an appeal t o reasons . O n a n account tha t denie s practical status t o
the oughts in a dilemma, what the agen t in such a case is doing is considering
only a limited subset of the reasons bearin g on his action a s he frames his two
conclusions: "I'v e got to join the resistance," say , and " I can't leav e Mother."
However, on e coul d jus t as easil y represent th e sam e chai n o f reasonin g a s
involving an evaluative "gestalt shift" in his view of the full set of reasons bearing
on the case. That is, instead of reaching different conclusion s on the basis of a
limited set of reasons about whether i t would be better to join the resistance or
to support hi s mother, the agent might be seen as concluding that a failure t o
do either would b e bad enough.
It would b e ba d enough to coun t a s morally unthinkable, let us say, for a
negative version of the "satisficing" account of moral deliberation that has been
suggested b y recent authors on utilitarianis m and relate d subjects. 24 The case
40 Between th e Horns

of cours e is one i n which the agen t has to decid e in practical terms to d o th e

unthinkable; but this decision need not rest on the sorts of reasons whose con-
clusion can b e translated int o a third-person ought-judgment . A general prin-
ciple of action o f the sor t that in some extramoral sense resolves the dilemma
would b e limited to th e firs t perso n i n the wa y note d earlie r and ma y some -
times involve a step beyond deliberation in terms of "ought."
This negative model of deliberation in dilemmas will let us see the agent as
motivationally torn rather than simply left u p in the air, as we might expect if
deliberation failed to terminate. However, it is important that "ought" in stan-
dard deontic logic also has positive implications insofar as it is assumed to imply
"permissible."25 I n fact , this seem s to m e to reflec t th e most common us e of
"ought" i n everyday speech an d t o b e tied t o th e prescriptiv e function ofte n
assigned to the term. Examples like those just given for Sartre's case"I've got
to join the resistance" and " I can't leave Mother"suggest that the point does
not hold for "must," though "must" is sometimes thought of as a stronger form
of the same concept.26 At any rate, with "ought" understood as prescriptive in
this sense , i t woul d indee d seem to b e impossibl e for practica l reasonin g t o
generate conflicting ought-judgments as conclusions.
This is one of those point s that ar e obvious once seen, and i t may explain
the plausibility of the rejection of practical "ought" in dilemmas. It is the com-
bination embodied in the meanin g of prescriptive "ought" of a positive claim
that its object is permissible with the negative "core" claim that everything else
is impermissible that disallow s conflict . I n intuitive terms: To sa y that a n all-
things-considered review of the reasons rules out everythin g is just to sa y that
nothing is permissible. But in that case, the requirements of prescriptive "ought"
will not b e met.
So it is for prescriptiv e "ought," no t jus t practical "ought"fo r "ought "
that is assumed to b e positive in its practical role, to th e extent tha t it points
the agent toward som e actionthat Williams's denial of practical "ought " in
dilemmas seems to be defensible. In that case, however, in order to capture the
motivational conflict involved in dilemmasto make them out as involving more
than a conflict between prima facie ought-judgmentswe can move to a nega-
tive characterization of them in terms of prohibitions, or action-guiding judg-
ments of moral wrong.
Williams himself seems in later writings to prefe r a negative characteriza-
tion.27 Taking the distinction to have the implications for practical "ought" that
my argument here suggests means violating some othe r assumption s o f stan-
dard deonti c logic, i n the first instanc e that "forbidden " implie s the permissi-
bility of some alternative to action. Bu t those assumptions essentially just beg
the question-o f dilemmas, presumabl y on ground s o f systematic simplicity. I
examine these deontic assumptions in more detail later in this chapter, but we
can see the significance of my positive/negative distinction in nontechnical terms
at this point i n relation t o Williams's deliberative question "Wha t ought I to
do?" In one sense, indeed, this question would not b e answered adequately by
a prohibitionit would b e raised again, jus t a s i t would be if answered wit h
two conflicting positive oughtsunless it is understood in relation to a particular
Practical Oughts and Prohibitions 4 1

act, s o that th e context supplies a positive ought-judgment. In that cas e "Yo u

ought no t to do A" will amount t o advice to do B, where A and B exhaust th e
field o f alternatives. Bu t this implicatio n o f positive practica l forc e trades o n
our ordinar y assumptio n tha t some act of those ope n t o th e agen t i s permis -
sible, which of course i s just what dilemma s call into question .

2. Deonti c Strengt h an d Value

The attemp t t o confin e moral conflic t to conflic t between prim a faci e ought -
judgments harks bac k t o Ross' s theory. Despit e man y authors' objection s t o
the term "prim a facie " and Ross' s own qualms about his use of it as an adjec-
tive modifying "duty," th e Lati n ter m seem s irreplaceabl y handy fo r makin g
an easy switch to adverbial status i n application to "ought" and related expres -
sions. We can speak of an ought or an ought-judgment as merely prima facie
or within one, of what its agent ough t prima facie to do. The term is also famil -
iar from its use in law with reference to partial evidence. Ross at one point gives
"parti-resultant" a s a more informativ e substitute for it. 28 We apply i t on th e
assumption tha t the reasons o r evidence bearing on an ought-judgment can be
broken dow n int o distinguishabl e grounds , pr o an d con , s o that a judgment
based o n onl y some o f them ca n b e said t o resul t from a partial subse t o f th e
total bod y o f evidence.
Although Ross himself contrasted prim a facie with actual or absolute duties,
in the contemporary literatur e the term seems to have two contraries, use d more
or less interchangeably in application t o ought s that result from a weighing of
all relevant evidence: "overriding" an d "all-things-considered. " Thes e term s of
deontic weight or strength are equivalent, however, only on the assumption tha t
there are no full-fledge d action-guidin g dilemmas. In this section I show ho w
the underlying notions come apart and examine the consequences of prising them
apart fo r the resolution of dilemmas.
Contemporary account s o f the meanin g of "ought " sugges t tha t man y of
the nondilemmatic cases for which "ought" has more or less become canonical
in moral philosophy are really more appropriately described by "must." 29 But
my discussion here of deontic dominance , as we might call the propert y o f an
ought that wins the weighing process, wil l also yield a more complicated pic -
ture o f this stronge r for m o f "ought. " I n applicatio n t o a n ough t o f norma l
strength, my discussion of dominance will bring out reasons for the differentia l
treatment o f positive and negativ e dilemmas that bega n to emerg e i n the las t
section. I then conside r som e question s about th e resolvabilit y of dilemmas
conceived i n positive or negativ e terms. B y singling out negativ e dilemmas as
irresolvable in a stronger sense , I shall be defending a conception o f dilemmas
that fit s thei r use in logic and mathematic s in addition to the everyday picture
of them a s practical choice-conflicts. Dilemmatic reasoning involves showing
that all alternatives have contradictory or otherwise unacceptable consequences.
Similarly, as a moral concept the notion has its clearest application to case s in
which all alternatives count as wrong.
42 Between the Horns

Forms of Dominance
The difference betwee n all-things-considered an d overridin g ought-judgments
can be brought out most sharpl y by way of a look at their application to judg-
ments o f permission. Th e operato r P (for "it i s permissible that") i s taken a s
primitive in one approach t o standard deonti c logic, with O defined in terms of
it as denying the permissibilit y of the negatio n o f the objec t of obligation. An
axiom o f the standar d syste m yield s the principl e that "ought " implie s "per -
missible," s o that "ought " is doubly linked to permissibility: negatively, since
Op = ~Pp, and positively , since Op implie s Pp. However, a positive judgment
of permissio n ca n b e thought o f a s holdin g al l thing s considere d bu t no t a s
overriding: It just makes no sense to say that something is overridingly permis-
The reason fo r this seems to b e that two differen t kind s of deontic weight
or dominance are assessed by the two notions. For "all-things-considered" th e
relevant sort o f weight i s at leas t partly evidential: the weigh t o f the evidence
or reasons for holding the deontic judgment in question considere d agains t th e
background o f the tota l bod y o f evidence. 30 This applie s readily enough t o a
positive judgment of permissibility, if only b y negating it s application t o th e
negative judgmen t of impermissibility. The positiv e judgmen t may b e said t o
hold "al l thing s considered" a s long as the negative judgment does not .
By contrast, for a judgment of overridingness the relevan t sort of weight is
specifically practical, since it measures something like the importance o f acting
on the deonti c judgmen t in question. Thi s ma y b e thought o f as the practica l
"strength" o f a requirement or prohibition . However , i t does no t appl y t o a
permission, as a judgment that simpl y allows som e action . A n action an d th e
alternatives to it may both be permitted, an d there is no inference from a denial
that a n action i s prohibited with such-and-suc h a strength t o a claim that th e
action is permitted with the same strength. Permissions have no particular prac-
tical strengt h bu t woul d see m just t o hol d i n the absenc e o f any prohibition ,
however weak, that meets the requisite level of strength to count as "all-things-
This distinction between evidential backing and practical strength seem s to
hold up, moreover, if we understand ought-judgments in terms of practical rea-
sons. A permission woul d then amount t o a denial that there are such reason s
in sufficient strengt h to yield an ought. Reason s for i t could not b e practical in
the relevant sense, since there is nothing in particular that amounts to action in
fulfillment o f a permission. Besides reasons fo r belie f i n a permission, o r a s a
way o f analyzin g such reasons, ther e migh t be said t o b e reasons fo r issuing
a permissio n tha t ar e practica l i n a mor e genera l or indirec t sense; they may
appeal, say , to th e valu e of morally unconstrained choic e wit h respec t t o th e
action in question. But these are not reason s fo r action i n the sense that applie s
to a requirement or prohibition, where what i s in question is action on the part
of th e agent to whom th e judgment applies, as opposed t o the speaker's act in
issuing the judgment.
The distinctio n between the two sort s o f assessment in application to per -
Practical Oughts and Prohibitions 4 3

missions introduce s complexities that are increased i n application t o the com-

pound judgments taken as corresponding t o positive oughts. Presumably , both
elements of the judgment Op, ~Pp and Pp, would hav e to b e weighed in an all-
things-considered assessment , wherea s if permissions cannot b e overriding, only
the negative "core" element of the meaning of the judgment can be taken int o
account b y a claim of overridingness. But the point for my purposes her e is that
on the assumption of dilemmas the all-things-considered assessment of the tw o
elements wil l itsel f com e apart. Supposin g that all alternatives in a given cas e
are prohibited strongly enough to count as impermissible all-things-considered
and that the case can be described (granting closure) as one in which bot h O p
and Op holdthe negative but not the positive element of the meaning of each
of th e ought-judgments bearing on the case would see m to b e satisfied.
That all alternatives are ruled out all things considered amounts , of course,
to th e sor t o f assumptio n whos e coherenc y is in questio n i n the debat e over
dilemmas. But I think we can now see a way of interpreting "all things consid-
ered" tha t allow s fo r it s coherency. Le t u s thin k o f a n all-things-considered
prohibition i n the firs t instanc e as one that i s important enougho r th e rea -
sons fo r whic h ar e importan t enought o stan d i n ligh t o f the tota l bod y of
evidence. This amounts to a kind of perceptual o r figure-ground dominance
of th e reason s fo r prohibitin g somethin g agains t th e genera l backgroun d of
reasons bearin g on actionand hence allows fo r gestal t shifts i n a way that is
not possibl e for practical motives. The reasons fo r prohibiting somethin g and
the reasons for prohibiting its contrary may both be important enoug h to stand
out agains t th e field , tha t is , assuming that neithe r overrides th e other , a s in
cases of dilemma.
In the case drawn fro m Sartre , fo r instance, i t seems to mak e perfect sense
for th e agen t t o sa y that th e reason s agains t abandonin g hi s mother an d th e
reasons agains t lettin g down th e resistance are importan t enoug h t o b e unaf-
fected b y each other. Both prohibitions hold, that is, in light of all the evidence
bearing o n actionassumin g (as the cas e demands ) that th e resistanc e need s
this particular agent no less than his mother does. Neither prohibition i s over-
riding, on our assumptio n tha t the case is a balanced dilemma , so each prohi-
bition is sufficiently seriou s to remain in force despite th e conflict.
The "weighted " cases that Williams discusses may not see m at first glance
to allow fo r a shift i n evidential perspective; bu t the y turn out t o b e amenable
to the same general treatment. Williams wants to say that neither obligation is
really overriding in a case like Agamemnon's, even though one of his options is
clearly better , on the grounds tha t the reasons fo r preferring one to th e other
fail to answer adequately the reasons on the other sideor in Williams's terms
"the claim s involved i n the conflict." 31 Tha t is , as utilitaria n considerations ,
they fail to answer the nonutilitarian reasons for the ought that is not acted on;
and presumabl y the same could be said o f any resolution o f the dilemm a that
appealed to reasons different i n kind from those the agent chooses to act against.
If Agamemnon decided that his duties as military commander were more impor-
tant unde r the circumstances than his role as a fatherwhethe r on utilitarian
grounds o r something elsethat would still leave his paternal obligations with
44 Between the Horns

enough importance , of a sort incommensurabl e with that o f his military role,

to remai n in force despite the conflict.
The basi c notion that lies behind the all-things-considered weighing of evi-
dence seems to b e that of a sufficiently serious reasonto prohibi t something,
in the first instance , with permission define d i n terms of the absenc e of a suffi -
ciently serious reason to prohibit something and obligation defined a s indicated
above. This notion mus t have comparative content o f the sort that comes into
the notio n o f overridingness , i f we assum e that som e reason s ca n b e seriou s
enough to cancel out others , at leas t below a certain threshhold leve l of abso-
lute seriousness. It cannot b e purely comparative, though, o n the assumptio n
that i t yields dilemmas for prohibition s tha t ar e supporte d b y reasons abov e
the threshhold levelreasons that are serious enough. If we understand dilem-
mas in terms of the notion, we can rule out the sorts of conflicts among trivial
obligations that are plausibly dismissed as prima facie.32 I appealed to the notion
at the end of my preceding section as providing a way aroun d Williams' s ban
on conclusive dilemmatic oughtssupposing that oughts include "ought-nots, "
or prohibitions.
Our intuitiv e talk of practical necessit y corresponding t o the modal auxil-
iary "must"for whic h "ought" often function s a s a weaker substitute in the
discourse o f moral philosophyseem s to fi t the notio n rathe r nicely . Talk of
what one "cannot" do on moral groundsin simplistic terms, the idea of some-
thing "taboo" (withou t the overtones the word sometime s ha s of absolute or
groundless prohibition)may be taken a s indicating the absolut e seriousness
of a prohibition. The prohibition i n question is serious enough, that is, to per -
sist i n the fac e o f extrem e barriers to actio n i n accordance wit h it , including
moral barrier s of the sort one encounters in dilemmas.
It does sound natural, as I noted a t the end of the last section, fo r the agen t
in Sartre's case to use variants of "must" to express the two ought s bearing on
his decision: "I've got to join the resistance" an d " I can't leave Mother." Each
statement evince s a kind of moral urgency ; and neither , not eve n the positiv e
statement, seem s to imply "permissible." On e can easily imagine the agent , in
response t o a revie w of the reason s agains t joinin g the resistance , protestin g
that h e has to. "Must " i s stronger tha n "ought " i n the suggestio n i t gives of
moral catastroph e i f unsatisfiedas opposed t o doin g les s than one' s best in
moral termsand i n cases of dilemma there is catastrophe eithe r way.
However, som e author s hold tha t "must " has strong enoug h implications
to provid e a way ou t o f dilemmas. The vie w is suggested b y Lemmon's argu-
ment for their logical coherency, which depends on contrasting "ought " with
"must," an d b y Williams's accoun t o f practical necessity , i n which he argues
for a version of Lemmon's assumptio n tha t "must " implie s "will," o r actual -
ity.33 Bu t Williams's account is importantly limitedt o intentiona l action, as
he notes, bu t also to what he calls "incapacities of character."34 Th e latter rep-
resent onl y on e sor t o f applicatio n o f "must, " th e sor t exemplifie d by th e
Sophoclean tragi c hero that h e picks out a t th e beginnin g of hi s treatment a s
one o f the tw o mai n examples of its moral usewith the othe r example pro-
vided b y the Kantian moral agent. Faced with a dilemma the Sophoclea n hero
Practical Oughts and Prohibitions 4 5

cannot do otherwise, given his character, than act on a certain one of the oughts
in conflict. The claim that he cannot do otherwise, then, is a claim that his charac-
ter i s such that h e cannotwhich would b e falsified , o f course, b y his doin g
otherwise after all.
It should be obvious that there cannot be two incompatible actions of which
this claim is true, for reason s resemblin g the argument from motivationa l suf-
ficiency that I considered an d rejecte d toward th e end of my preceding section
as an interpretation o f Williams's notion o f conclusive practical "ought." Does
the interpretatio n instea d fi t the notio n o f "must, " whic h William s takes as
stronger? It could fit the notion in general terms, I think, only if all decisions in
cases of dilemma could be made out as determined by the agent's preestablishe d
character. But in Sartre's case the causal relation between character an d choice
is supposed t o run in reverse. At any rate, reference to character in a sense that
allows for development over the time in questionas at least partly constituted
by the agent's current projects and the like, assumed to preexist hi s choice but
also to be subject to change in the situation tha t requires choicecannot play
a causal role here. Does the truth of the agent's claim that h e "had to " choos e
a certain option simpl y depend, then , on what he goes on to do, in something
like the way that a claim of knowledge depends on truth: not becaus e his char-
acter make s him ac t a s h e does, bu t rathe r becaus e our notio n o f i t refers in
part t o future action ?
It seems wrong, firs t o f all, to rejec t a s false a n agent' s clai m that h e "ha s
to" d o something he later fails to do. Consider " I have to go now," a s said by
a guest who is persuaded t o linger on. Though w e might respond i n conversa-
tion with a denial of his statement, our counterclaim can also be read as a way
of pointing out that practica l necessit y does not impl y necessity. We need no t
be denying, that is , that th e guest' s reason s fo r leavin g are stron g enoug h t o
support the statement; we might just be telling him to ignore them or acknowl-
edging after the fact that he did ignore them. Similarly, the reason why the agent
in Sartre's cas e cannot sa y at a later tim e that h e had t o take th e option h e in
fact rejected i s not that on e of his original statements has turned out to be false
but rather tha t the moral importanc e of action on it has become a thing of the
past. Mora l catastroph e ha s alread y occurred an d bee n assimilate d into th e
background of action .
In an y case, ther e ar e "must"-claim s tha t d o no t fi t the characte r mode l
at all. "I have to pa y my rent o n the firs t o f the month" nee d not b e taken as
attributing to it s speaker a high degree o f conscientiousness in financial mat -
ters. Instead, it is naturally read as elliptical for a claim that something bad will
happen i f the ren t i s not pai d b y the tim e indicated . Thi s ma y b e somethin g
specificthe agen t wil l be subject t o a fine, saybu t it need not be . Perhap s
the only penalty for tardy payment is failure to meet a certain standard o f per-
formance in financial matter s or failure to conform to the rules, something that
is perfectly possible, o n the model of "moral catastrophe. "
A simplification of deontic logic sometimes called "escapism" applie s this
reading to oughts generally, with the role of the threatened bad state of affair s
assigned to "the sanction, " symbolized by a constant S referring to whatever is
46 Between the Horns

entailed by the nonfulfillment o f obligation and presumably is escaped by fulfill -

ment.35 O n thi s account a statemen t lik e the one just above might b e read a s
"If I don't pa y m y rent b y the fift h o f th e month , I incur the sanction. " Th e
further "escapist " assumptio n that the sanction is avoidable makes this approach
unsuitable for our purposes, sinc e it begs the question of dilemmas. However ,
it provides a reasonable model for interpreting individual "must"-claims of the
sort just illustrated: as ought-statements that threaten the agent with a morally
catastrophic sanction .
Will the sanctions model le t us assimilate "must"-claims t o the perceptua l
picture of figure-ground dominance I have sketched in application to all-things-
considered prohibition? Not withou t some differences, i t seems; for besides its
failure to impl y "permissible" bu t for similar reasons "must " also fail s t o im-
ply "can " i n intuitive terms. The agent may protest tha t h e has to suppor t hi s
mother, fo r instance , i n response t o reason s fo r thinking it impossible at thi s
pointwithout ten years of savings behind him, sayas well as in response t o
reasons for some conflicting obligation. I would explain this by taking "must"
as resting on a partial view of the evidence rather than on an all-things-consid-
ered view. That is, it rests on reasons that are thought to be important enough
to stand o n their own, blottin g out contrary reasons rather than simply domi-
nating the field . Th e agent' s protes t amount s to a denial of any contrary rea -
sonsfor th e moment , anywaya s he asserts th e strengt h o f the reason s h e
chooses t o focu s on. I t isolates th e latte r from the evidentia l background, in-
stead o f pickin g them ou t fro m a backgroun d tha t remain s withi n view , a s
indicated by the fact that he could not have said, in light of the reasons for join-
ing the resistance, "Even so , I can't leave Mother."
On this account "must" is not just a strengthened form of "ought" but rather
presupposes a different wa y of dealing with the reasons for an ought-judgment,
where these are assumed to b e particularly strong. If we also assum e that the
same bod y o f evidence admits o f gestalt shifts , o n th e othe r hand , th e "eve n
so" statement jus t above would be acceptable as rephrased in terms of "ought":
"Even so , I shouldn' t leav e Mother." No r doe s it s naturalness depen d o n a
weaker readin g o f "ought " a s merel y recommending rathe r tha n requirin g
action; it would soun d most natura l as "Eve n so , it would b e wrong t o leave
Mother." This grants the contrary reasons rather than denies them, but assert s
in the fac e o f them tha t th e prohibitio n i n questio n hold s "al l thing s consid -
Can the prohibition b e practical if the contrary prohibition is, too? I see no
barrier t o takin g contrary prohibition s a s meant t o guid e action, whe n the y
follow fro m rules designed for tha t purpos e i n general terms, though th e spe -
cific directives derived from the rules on a given occasion are not jointly fulfillable
and henc e would b e unreasonable a s commands. Wit h agglomeratio n ou t o f
the picture, they are not therefore logically incoherent in combination. For that
matter, they may in fact guide action at least indirectly, on the sanctions model ,
by making out nonfulfillment a s morally unacceptable or as incurring some sor t
of specific sanctio n rather than b y directly bringing about fulfillment . O n this
interpretation, "Don't do X" i s elliptical for "Don't do X or you will incur S."
Practical Oughts and Prohibitions 4 7

In cases of dilemma, of course, th e agen t wil l have no wa y o f avoiding S alto-

gether; i n opting fo r a particular horn of the dilemma, however, h e chooses t o
accept a particular instance of S. Putting up with a sanction, then, or acceptin g
itor something more active, such as subjecting oneself to itmight be thought
of as an indirect way of satisfying th e ought tha t loses out .
However, thi s possibilityt o be considered i n mor e concret e for m later ,
when I turn t o guil t as an exampl e o f what philosopher s cal l "interna l sanc -
tions"can hol d onl y fo r negativ e oughts o r prohibitions , i f my precedin g
argument is correct. Insofar as a positive ough t als o implies that fulfillmen t i s
permissiblein other words, insofar as it does not simply reduce to a negative
oughtit rules out the competing all-things-considere d ought as interpreted o n
the sanctions model , with a statement o f permission take n a s denying that it s
object will incur a sanction.36 Further, on the account just given, the correspond -
ing "must," positive or negative, turns a blind eye to contrary reason s i n a way
that undermines the suppor t fo r it s competitor. B y contrast, a negative ough t
exhibits a kind of perceptual dominance that admits contrary reason s int o th e
field o f view and henc e allows it s competitor a basis in it as well.
In maintaining a prohibition "all-things-considered " eve n in light of equally
strong prohibition s of alternatives, we are exhibiting what migh t be called the
intractability of moral wrong: the relative imperviousness of judgments of wrong
to th e sort s o f furthe r practica l calculations that affec t positiv e o r nonmora l
action-guidance. The claim that there are important enough reasons for requir-
ing some act, b y contrastor o n the other hand, for ruling it out in prudential
terms as irrational o r the likewill not make sense as an all-things-considere d
judgment in combination wit h an admission that there are equally strong rea -
sons on the other side . The evidential assessment of negative ought-judgments
thus seems to allo w them uniquely the sort of noncomparative conten t tha t is
needed fo r full-fledged mora l dilemmas.

Resoluability and Wrong

Williams and othe r proponent s o f moral dilemma s make the m ou t a s resolv-
able only nonmorally, by appeal to a practical ough t reflecting the agent's per-
sonal projects and the like. Morally speaking, they are irresolvable, in the sense
that moralit y fail s t o provid e a basi s for resolving them. Onc e w e distinguish
between positive and negative dilemmas, however, I think we can see that full -
fledged dilemmasthos e that fi t the negativ e conception i n terms o f exhaus -
tive prohibitionswill be irresolvable in a stronger sense : The y wil l admi t o f
no morally acceptable resolution , whether o r not morally based. The agent has
to violat e mora l norms , no t jus t t o reac h beyon d them , i n order t o resolv e a
negative dilemma in practical terms , a s he must i n one way o r another , sinc e
the prohibitions bearin g on his decision ar e assumed to exhaust the field .
For dilemmas characterized positively, on the other hand, we seem to have
to yiel d t o th e insistenc e of opponents o f dilemmas that wha t w e really have
are prima facie ought s in conflict; otherwise, the same acts would come out a s
both permissibl e and impermissible . This follow s fro m th e assumptio n tha t
48 Between the Horns

separates positiv e from negativ e oughts: tha t "ought " implie s "permissible."
Whether or not one of the conflicting oughts can be taken as overriding, it seems
that neither can count as all-things-considered on this positive formulation. This
was th e kerne l of truth I found in Williams's argumen t fo r denyin g that bot h
can be conclusive. Rather than excluding dilemmas from the practical sphere ,
however, my account allows for a separate treatment o f dilemmas set up nega-
As noted i n my initial overview of the literatur e o n dilemmas , moral phi-
losophers hav e sometimes stressed the negative characterization, bu t as far as I
know th e distinctio n was no t give n any attentio n unti l my own treatmen t i n
1983 o f the case from Sophie's Choice. 37 The positive characterization i s empha-
sized i n th e literatur e on deonti c logic ; with th e exceptio n o f Von Wrigh t i n
1968 non e o f the classic papers on the subjec t discusses dilemmas in negative
terms, an d th e distinctio n is first note d i n a piece published i n 1987. 38 Here I
want essentially to revise my earlier argument for the distinction i n light of later
developments, especially my current suggestions on deontic strength and value,
by taking a fresh loo k at the notion of a disjunctive ought brought in to resolve
a dilemma. Readers who wish to bypass technical discussion should at this point
skip to p . 6 2 fo r the summar y and transitiona l comment s a t th e end o f this
In attemptin g t o resolv e Sophie's dilemma , then , on e migh t appea l t o a n
ought that prescribes saving at least one child, which under the circumstances
requires choosing a particular one. I think of this as a disjunctive oughton a
use of th e ter m tha t cover s inclusiv e "or"since it amount s t o a n indefinit e
prescription t o save one child or other. On the assumption tha t the dilemmatic
oughts that prescribe saving each child are in balance, the disjunctive ought might
naturally be taken as outweighing either of them. It is most important that Sophie
not let both children be killed, so on this account what she ought to do all things
considered i s to make a choice between them an d sav e one of themchoosing
one by some fair method . What the disjunctive ought prescribes , in effect, i s a
practical resolutio n o f the dilemma : action i n accordanc e wit h a tiebreaking
ought, presumably arbitrary, of the sort that migh t result from flippin g a coin
in circumstances that allow for it. In order to satisfy the minimal moral demands
of th e case , Sophie i s advised to appea l beyon d strictl y moral considerations .
However, Williams's view on overridingness seemed to rule out that account
of Sophie' s choice . Supposin g tha t a mothe r ha s seriou s an d separat e mora l
obligations t o preserv e each o f her children, no t jus t her progen y considere d
collectively, Sophie's satisfactio n of the disjunctive obligation woul d no t ade -
quately meet the claims involved in the conflict. Unlike Agamemnon's choice
between his daughter and the expedition unde r his command, Sophie' s choic e
might be said to involve commensurable (equal) claims, but on the assumptio n
that they do not permit trade-offs, a fair decisio n between them will still leave
one of them unsatisfied. It is not enough that both be taken into account on the
model of political representation.
On the other hand, we can now see grounds for accepting the conclusion of
opponents o f dilemmasan d Williams's conclusion for conflictin g practical
Practical Oughts and Prohibitions 49

oughtsthat neither of the conflicting positive oughts in Sophie's case is in force

all things considered. Thoug h neithe r is overridden, each implies a claim of per-
missibility that is contradicted b y the other. Can we then say that it is all-things-
considered statu s rather tha n overridingnes s that distinguishes the disjunctiv e
ought? There ma y be no ought in force all things considered t o contradict th e
implied clai m that i t is permissible for Sophi e to sav e one chil d or th e other .
To say this, we would have to reject at least some strong form s of the prin-
ciple of deontic closure, assuming that the negative oughts or prohibitions bear-
ing on th e cas e are in force all things considered. A s the cas e i s set up, saving
one child requires letting one be taken to th e gas chamberwhichever one is
not picked out b y the tiebreaker Sophie uses. If the necessary means to satisfy -
ing an ought are themselves obligatoryas closure over causal relations would
have itand "ought" implies "permissible," th e disjunctive ought presumably
combines with th e tiebreake r (a t any rate afte r th e latte r i s decided upon) t o
imply the permissibility of failing t o sav e a particular child . So it cannot hol d
all things considered alon g with a prohibition o f that failure.
In any case, what the disjunctive ought gives us is not quit e a resolution of
the dilemma. On the positive account o f the case, we do not have a dilemma in
the first place but instead just a prima facie ought-conflict. And on the negative
account, th e disjunctiv e ought doe s not yiel d a morally acceptable resolution ,
despite it s moral basis , sinc e we also hav e all-things-considered prohibitions
ruling out eac h o f the action s tha t would satisf y it. Closure ove r causal rela -
tions would yield explicit prohibitions of saving either child, and even without
closure we have to conclude that any particular acts that would satisfy th e dis-
junctive obligation are morally prohibited under some other description. 39 They
may not be prohibited with equal forcewe can say that the disjunctive ought
is weightier than the dilemmatic prohibitions and hence in one sense "outweighs "
thembut th e prohibition s ar e seriou s enough i n absolute terms t o coun t a s
In short , though i t i s perfectly clear what Sophi e mus t do , an d o n mora l
grounds, i n respons e t o th e dilemma , ther e seem s to b e n o wa y ou t o f th e
dilemma that can count as morally permissible. I have framed this conclusion
in terms of "must" rather than "ought," since the use of "ought" migh t be taken
to entail that some act is permissible after all , on our assumption that "ought "
implies "permissible. " Whethe r i t doe s hav e tha t consequenc e depend s o n
whether w e accep t closur e over causa l relations . Fo r reason s lik e those just
indicated, i f we d o accep t closure , there wil l be nothin g tha t Sophi e morally
ought to do, at any rate all things considered, in response to the dilemmanot
even the disjunctiv e act of saving one child o r the other, despite it s overriding
moral importance, as one might say. At any rate, even without closure we would
have to deny that any particular act done to satisfy the disjunction can possibly
pass muster in moral terms .
There will be other balanced cases, like the assassination case in chapter 1 ,
that do not allo w for a disjunctive ought in the first place, since all the alterna-
tives in them, including inaction, are prohibited with equal force. Sophie's choice
might be made into such a case if we assume that she has to do something morally
50 Between the Horns

worse to either child in order to save the other than simply failing to save it. At
a certain levelif Sophie had to torture one child, saythe wrongness of these
acts might be thought o f a s balancing the wrongnes s of violating the disjunc-
tive obligation an d failin g t o sav e either one. However , eve n without a n ex-
haustively balanced case of exhaustive prohibition, dilemmas on th e negative
characterization will still have problematic consequences of the sor t just indi-
We can already begin to see some of the complications for deontic logic that
dilemmas introduce. I shall postpone t o m y next section further comment s on
closure and the other standard principles that dilemmas call into question. The
important thing to note at this point is that much depends on how we decide to
describe a given caseboth the choices and the reasons for them that are thought
to b e at issue in the dilemma . In the firs t instance , our decisio n will affec t th e
plausibility of the case as set up in negative terms: the plausibility of the claim
that there is reason enoug h fo r prohibiting all options i n it. Consider Sartre' s
case, once again: A serious question might be raised about counting as wrong
or impermissibl e in it s own righ t th e agent' s failur e t o joi n the resistance . It
sounds plausibl e in positive term s to sa y that h e ought t o joi n the resistance ;
and o n standar d deonti c assumption s thi s implies that i t would b e impermis-
sible for him not to . Bu t in this case the reasons for th e act i n question would
normally be taken a s primary, it seems, with reasons against omittin g the ac t
seen as derived from them. To get around this problem and continue to use the
case for purposes of illustration, I specified that the resistance needs this agent
in particular. To suppor t the claim that he would b e doing something wrong
by staying homein a sense that does not reduce to a failure to satisf y a posi-
tive dutywe might suppos e tha t h e is the on e person wh o possesse s certai n
skills with explosives, s o that th e resistanc e depends o n hi m n o les s than hi s
mother does .
Our ordinar y assumptio n abou t th e case, by contrast, woul d b e just that
the agent ought to volunteer for a task that is required of some indefinite set of
members of a group he belongs toin a commendatory sense of "ought" that
does no t reall y have the imperativa l force o f a n ough t base d o n prohibition .
We sometimes blur the distinction betwee n these different level s or strengths of
action-guiding force to ge t practical results. But it is important tha t with an y
weaker assumptions than those just indicated Sartre's case would no longer count
as a full-fledge d dilemm a i n m y terms, howeve r neatl y balanced th e alterna -
tives it presents, o r howeve r balance d they seem to b e if we conside r deonti c
strength independently of positive or negative value. The crucial thing for the
question o f resolvability is that th e agent' s alternative s each b e prohibited b y
strong enough reasons for the prohibitions t o stand i n light of each other .
On the other hand, the cases I have referred to as weighted dilemmas, such
as Agamemnon's case, will fit my account of dilemmas irresolvable by permis-
sible means, assuming that the prohibitions they involve are in force all things
considered. Even in ligh t of al l the evidencei n light o f wha t w e migh t sup-
pose to be the overriding importance of the success of the Greek military expe-
Practical Oughts and Prohibitions 5 1

ditionAgamemnon's sacrifice of Iphigenia remains prohibited; it still counts

as morally wrong, and indeed in the strongest terms, in a sense meant to guide
action. Here it is even clear which alternative the agent must choose for a prac-
tical resolution of the dilemma like that given by the disjunctive ought in Sophie's
case, whereas in Sophie's cas e th e disjunctiv e ough t neede d supplementation
by a tiebreaker selecting a specific option. S o the case does not involv e a moral
quandary, but it still seems to amoun t to a moral dilemma.
Since balanced dilemmas pose furthe r problem s of choice, they might seem
to b e dilemmatic in a fuller sense . But the term "fatal" dilemm a seems apt fo r
any cases that exhibi t the property o f allowing no morall y acceptable resolu -
tion. Fatal dilemmas are those tha t exhibi t Nagel's "mora l blin d alley": They
offer th e agent no way out of a forced choice among moral wrongs. To describe
even Sophie's case in these terms, however, requires some care. Her prohibited
alternatives have to be thought of as acts of failing to save either childstand-
ing idl y b y while a child is takenrather than a s the act s of saving one child
that they also amount to in fact, for the acts of saving also come out as permis-
sible under that description. Moreover , if it does not seem plausible that there
are strong enough reason s agains t eac h ac t o f omissio n fo r th e tw o prohibi -
tions to hold in light of each otherand even in light of the weightier disjunc-
tive requirement to save at least one of her childrenthe case could always be
strengthened. We might suppose that Sophie actually has to harm one child to
save the other, as in the exhaustively balanced variant case I described. My sug-
gestion i s just that on some version the case will exhibit the sort of fatal dilem-
matic structure that poses problems fo r the practical coherency of ethics.
For an example that doe s no t requir e compounding the wrongs th e agen t
has to choose between, we might turn back to my assassination case. Here the
agent faces a choice between allowing the assassination of an important politi-
cal figure by her silence and lettin g the assassins murder one of her relatives, if
she fails to cooperate an d informs on them. To balance the agent's responsibil-
ity to her family, we might suppose tha t the political figure represents a major
force fo r good under urgent circumstances. However, eve n if the case is not a
balanced dilemma , it wil l come ou t a s a fata l dilemm a along wit h Walzer' s
torture case , accordin g t o th e treatmen t jus t give n o f weighte d case s lik e
Agamemnon's. In common with Agamemnon's case and at least some variants
of Sophie's, it admits of no morally acceptable resolution; unlike all but exhaus-
tively balanced variants of Sophie's case, it does not eve n allow for the sort of
morally mandated, thoug h stil l impermissible, way ou t tha t i s provided b y a
weightier disjunctive requirement.
In short, no. matter what the agent does here, she does wrong, and neithe r
of her options seems to be any better or morally more important than the other.
What can it mean, then, to offer he r all-things-considered moral adviceor for
that matter, to decline to do so, stopping at the claim that each of her alterna-
tives is prima facie prohibited? The case seems to undermine not just the agent' s
choice of action but ou r ow n accoun t of it in action-guiding terms. It therefore
poses in particularly stark form the problem of practical "ought" in dilemmas.
52 Between the Horns

I want no w t o consider som e o f the general consequences for deontic logi c of

my own attempt t o make room for such cases within action-guiding ethics .

3. Problem s for Practica l Ought-System s

On th e vie w I se t out t o defen d here , th e statement s tha t see m t o appl y t o
dilemmas are odd and in many ways unsettling but not incoherent. They make
sense as results of applying teachable practical rules to situations of intractable
wrong, even if they may be said to mak e no sens e as concrete piece s of practi-
cal advice in those situations. I now want to pinpoint some of the ways in which
dilemmas on the accoun t jus t given are unsettling, particularly for systemati c
approaches t o moral reasoning that purport t o capture our ordinary notion of
practical "ought. " I shall focus o n a few central problems raised b y dilemmas
for deontic logic, without much technical detail on alternative systems. My aim
is just to indicate in broad outline the degree of complication tha t would resul t
from a n attempt t o stretch the materials provided by the standard syste m to fit
our ordinary moral thought.
I shall assume on the basis of my understanding of the action-guiding func-
tion assigne d t o "ought"o r t o th e particula r sense or us e of "ought " tha t
deontic logi c as here interpreted is meant to capturetha t bot h o f the ought -
implication principles picked out earlier do apply. For practical "ought, " that
is, it is essentially a matter o f definition that "ought " implie s "can " an d tha t
positive (and hence prescriptive) "ought" implies "permissible"; bu t thi s is of
course compatible with the claim that other uses of "ought" have different func -
tions and henc e do no t obe y the two principles. 40 M y central tas k i n this sec-
tion will be to suggest ways of handling the two further principles that seemed
to conflict with these in application to dilemmas: the ought-derivation principles,
agglomeration an d closure.
Closure will naturally come u p firs t becaus e of its relation a s indicated i n
my last section t o som e of the odder result s of the distinction I have defended
between positiv e an d negativ e dilemmas . In fact , befor e w e ge t t o th e clas h
between principles , we need to rais e som e question s abou t th e results of that
distinction fo r deonti c logic . S o far, I have blurred over problem s abou t per -
mission an d negatio n i n the standard syste m that no w nee d to b e considered.
Once the y are deal t with, I shall claim, dilemmas themselves will no t giv e us
reason t o restric t th e principl e of closure . However , ther e ar e othe r reason s
having to do with the interpretation of the deontic operator tha t will be worth
discussing here at least briefly .
On th e othe r hand , agglomeratio n wil l stil l b e calle d int o questio n b y
dilemmas, an d I shall go on to sugges t a rationale for droppin g i t that make s
sense on our assumption of a strong action-guiding interpretation o f "ought,"
now though t o f i n negativ e terms. Withou t agglomeratio n w e see m t o en d
up wit h a collectio n of subsystems of practical "ought" designe d to rul e ou t
dilemmas within any single system. Along with dilemmas, though, this picture
also seems to rule out the sort of interplay between ought-judgments that deontic
Practical Oughts and Prohibitions 5 3

logic has to capture if it is to represent any of the more interestin g cases of moral

Prescriptive "Ought"
Standard deonti c logic derives the principle that "ought" implies "permissible "
from a n axiom tha t essentiall y begs the question agains t dilemmas : P p v Pp. 41
Even without this assumption, however, a version of "ought"-implies-"permis -
sible" migh t seem to rule out the distinction underlying our negative definition
of dilemma s because th e principl e applies t o negativ e ought s a s well as posi -
tive. That is, if "ought-not" also implies "permissible-not," w e would seem to
have incompatible judgments of permissibility in negative cases as well as posi-
tive. In the assassination case , for instance, if the agent ough t neither to infor m
on the assassins nor to remain silent, we seem to have it that she is permitted to
do either, given that no t doing one amounts t o doin g th e other. But our nega-
tive ought-judgments presumably tell us that she is not permitted t o do either.
Since "ought" involves both a positive and a negative permission, i n short, the
case yields a contradiction, assuming closure. Even without closur e it yields an
implausible set of permissions .
Perhaps the simplest way of avoiding this problem would be to take "ought -
not," o r F (for "it is forbidden that"), a s primitive in deontic logi c rather tha n
O o r P and t o buil d "ought"-implies-"permissible" int o the definition of Op.
With Pp defined a s ~Fp, Op would then come ou t as Fp & ~Fp; and "ought" -
implies-"permissible" woul d not apply to Fp to yield Pp. The principle would
still apply to Op, but the point is that on this approach O p would no longer be
equivalent to Fp . It would no longer amoun t t o a negative ought, o r "ought -
not," a prohibition, bu t instead, one might say, to "ough t not-"with "not "
understood a s negatin g the objec t o f "ought, " s o tha t th e resultin g ought -
statement require s some alternative action, rather tha n a s yielding a negative
version o f the operator, s o that th e statement forbid s action .
This negative approach therefor e involves complicating on e of the underly-
ing grammatical assumptions of deontic logic , whic h treat s "ought-not " an d
"ought not- " a s interchangeable . However, i t doe s see m to fi t the pictur e of
all-things-considered practical oughts and reasons that emerged from my treat-
ment of dilemmas in this chapter. Th e notion o f sufficient reaso n fo r prohibit -
ing some action came out as fundamental in the sense of not reducing to reason s
in favor of some alternative action, since it is not undercu t by decisive reasons
against alternatives . This pictur e captures somethin g intuitivel y basic, more -
over, t o the extent that the notion o f a "taboo"ruling ou t certain actions on
the basis of authoritative commandmight be thought o f as a primitive "ought"
in historical o r developmental terms as well as logically.
On the other hand, the picture might seem to violate the standard interpre-
tation o f negation, derived from Frege , as modifying th e content o f a proposi -
tion rather than amounting to a further operator on propositionsdenial, say
on a par wit h assertion. For instance, Frege insists that no distinction between
positive and negative propositions (in his terms, "thoughts" ) i s needed to make
54 Between th e Horns

sense of any logical principle he knows of. 42 Would thi s rule out principle s of
deontic logic that distinguish between positive and negative deontic operators ?
In fact, Michael Dummett suggests that we add to Frege's stated remark s some
corresponding points about prohibition: To avoid overcomplication in the way
Frege ha s i n min d whe n h e takes th e denia l of a propositio n a s asserting it s
negation, we also nee d to tak e prohibition as commanding a negation. 43 Bu t
this is apparently to grant that Fp amounts to Op .
This applicatio n o f Frege' s vie w is less straightforwar d tha n i t appears ,
however. First, we should not e that th e deontic operators are not themselves
"force operators" on the model of assertion. They can be understood a s prepo-
sitional operators , forming compound proposition s fro m prepositiona l vari -
ables"it ough t t o b e the case tha t p " fro m "i t i s the case tha t p"s o tha t
negative force can stil l modify th e conten t of a proposition: I t would be built
into th e conten t o f th e compoun d propositio n forme d b y a negativ e deontic
operator. We now need to drop the assumption that this is expressible in terms
of th e negatio n o f the simple r component proposition , however . Instead, we
have to recognize a distinction for a statement with imperativa l force between
negating the description o f the state of affairs i t commands the agent to bring
about and issuing a negative version of the commandin other words, substi-
tuting "prevent" for "bring about."
We should grant, of course, tha t packin g negation int o th e operator a s on
this account will indeed complicate matters in a way that is contrary to the spirit
of Frege' s approach . Bu t agains t th e simpl e application o f Frege' s view s on
affirmation an d negation to the deontic modalities, we should note that the latter
are alread y complicated b y th e presenc e o f permissio n a s a n intermediate
possibility betwee n command an d prohibition . Ther e ar e tw o differen t way s
of negatin g the imperativa l content o f a command, on e migh t say: issuing a
countercommand an d simply retracting the command. 44
In fact, "entertaining" a proposition migh t be suggested in place of asser-
tion as providing a mentalistic parallel to command tha t more accuratel y rep-
resents the real complications to b e found in our us e of the notion , includin g
the logica l space i t leaves for permission. Here there seem to b e several possi-
bilities: (1) One can hold a proposition i n mind, or actively entertain it, by anal-
ogy to a command. (2) One can block it out or exclude it from consciousness
the analogue of prohibitionpossibly but not necessarily by entertaining some
other proposition tha t excludes it. But (3) one can also simply fail t o entertain
it, not necessarily by blocking it out but perhaps just because one is unreceptive
to i t in some more passive way. The sort o f assumption marking off the third
possibility from th e second is that being unprepared to think about something,
being uninterested in it, or the lik e is not alway s explainable in terms of some
competing mental activity. For similar reasons, (4 ) one can als o in some cases
do nothing to prevent a certain proposition fro m coming to mind and yet just
fail to entertai n it. So failing to bloc k out a prepositional though t count s as a
distinguishable possibilityamountin g t o th e analogu e of permissio n in m y
treatment of the logi c of practical "ought."
The complications introduced by my approach, then, are not without par-
Practical Oughts and Prohibitions 55

allel i n moral psychology , an d m y approach i s founded on mora l psycholog y

insofar a s i t rest s o n th e perceptua l o r figure/groun d assessment o f deonti c
weights that was outlined in the preceding section. We can see further reason s
for complicatin g this account i f we now take a look a t th e problems with clo-
sure, the principle called int o question b y dilemmas in standard deonti c logic
on the assumption that "ought" implies "permissible." Sinc e my account effec -
tively limit s the latte r principl e to positiv e oughtsthos e expresse d b y th e
operator O , including oughts wit h negated proposition s a s their objectsand
at the sam e time limit s dilemmas to negativ e oughts, i t keeps dilemmas fro m
posing any special problem with respect to closure. We no longer have to worry,
that is , about applyin g closure t o a statemen t grantin g the permissibilit y of
satisfying a conflicting oughtabout whethe r we could deriv e from i t a con-
tradiction of the statemen t o f impermissibility that follow s by definition from
the ought with which i t conflicts. For all-things-considered conflicting oughts
will now be limited to prohibitions, which do not imply "permissible. "
Closure may be questioned, however, on independent grounds that bring to
light important general problems with the interpretation of the deontic opera -
toror operators, we now shoul d say, since the ambiguities in our readin g of
O will be exhibited somewhat differentl y b y F, which does not provide a paral-
lel to the distinction betwee n "ought" and "obligation. " A s previously stated
in positive form, closure allows us to derive obligations to do anything neces-
sary to fulfil l ou r obligations . To th e extent tha t th e unrestricte d principl e is
plausible, however, it really seems to apply to oughts rather than to obligations
strictly construedat any rate, obligations whose objects amount to actsfor
reasons we may illustrate with the assassination case.
In intuitive terms, with "obligation " take n broadly to include requirements
imposed b y moral principles as well as specific commitments, the agen t in the
assassination cas e may be said to have an obligation to protect he r country or
even to save a certain political figure from assassinatio n in addition to the obli-
gation to protect he r family. To fulfill thi s obligation under the circumstances,
she is required to infor m on th e assassins . Fulfillin g th e latte r requiremen t in
turn imposes on her any number of further requirements in context, most nota -
bly that sh e fail t o prevent the murder of the family membe r under threat bu t
also, for instance, tha t sh e move her lip s and othe r trivia l presuppositions of
action on her original obligation that have nothing to d o with the conflict.
Whether O applies to any of these putatively derived obligations, however
whether we can speak o f them plausibly as "obligations"depends very much
on the terms in which we decide to interpre t the operator. The word "ought "
applies fairly readily to the results of closurewe may grant that the agent ought
to move her lips, saybut "obligation" appears to be sensitive to instrumental
remoteness and change s i n wording of a sort tha t woul d mak e it hard t o for-
mulate a qualified versio n of the principle. We might grant, fo r instance , that
the agent has an obligation that requires her to move her lipsor even that it is
obligatory that she move her lipsbut surely not that she has an obligation to
move her lips. On the other hand, I think we would allow that she has an obli-
gation to say something, for instance, even though this already involves going
56 Between the Horns

somewhat beyon d the stated content s o f her original obligation and certainly
of the general moral principle from which it was derived by closure.45 Th e bes t
we can say , I think, is that closur e doe s no t hol d generall y for obligation , a t
any rat e as applied directly to acts.
Standard deonti c logi c reads O i n a way that mixe s "ought" and "obliga -
tion," but as applied to prepositional variable s on assumptions that do seem to
allow for closurethough they also seem to keep deontic logic from capturing
practical "ought." The readings usually given for the operator ar e "it is obliga-
tory that " or "i t ough t t o b e (the case) that," take n as interchangeable. Both
phrases, bu t especially the latter, ar e naturally read i n light of the sort o f ideal
interpretation of ought-judgments that corresponds to the "ideal world" seman-
tics of standard deonti c logic. 46 Wha t ough t t o b e the case, accordin g to thi s
approach, i s what i s the case in all deontically perfect or idea l worlds, world s
in perfect conformity to th e mora l rules . In recognizing dilemmas, of course ,
we recognize that the real world i s not i n this sense ideal: Any world compat -
ible with wha t i s now irrevocabl y the cas e will have to b e seriously flawed in
moral terms. Even apart fro m dilemmas, moreover, our real-world obligations
will sometimes rest on the assumption of a less than perfect moral world t o the
extent tha t the y include "contrary-to-duty" obligations , obligations to mak e
up for moral wrongs. 47 Wha t i s in question here is what a given agent ought t o
do, not what ought to be, but although "ought-to-do" seems to cover more of
our ordinar y practica l ought-judgments , deontic logi c seem s t o fi t "ought -
Even read in terms of "obligation," then, but applied to propositional vari -
ables, O ca n b e expected t o suppor t closure , as it seemed t o i n the example s
drawn from the assassination case just above. That anythin g logically or caus-
ally require d b y what i s true o f al l idea l worlds wil l b e true o f the m a s wel l
presumably would follow from the definitio n o f a morally ideal world a s satis-
fying logica l and causal laws. As applied to acts, on the other hand, example s
like those just cited suggest that closure holds for "ought" but not for "obliga-
tion," s o i t seem s w e hav e t o choos e betwee n "ought " an d "obligation " i n
interpreting th e deonti c operator . Where the y introduc e a n infinitiv e phras e
rather than "that," terms based on "obligation" appea r to be more tightly tied
to the wording of the principles, commitments, or the like from which particu-
lar requirements are derived.
To avoid these complications, I have mainly stuck to "ought " i n this chap-
ter, supplyin g its missing noun for m wit h th e ver b even in some case s wher e
"obligation" migh t sound mor e natural . For we seem to nee d both th e prin-
ciple of closure and statement s of "ought-to-do" i n order t o capture the prac -
tical reasoning involve d in dilemmasincluding most notabl y the derivation
of preparatory oughts that came into my initial argument against Williams. In
a case lik e Sartre's, tha t is , even if there are n o positiv e dilemmatic oughts i n
force all things considered, we still want to be able to derive obligations requir-
ing th e agen t t o tak e an y necessar y and permissible mean s to satisfyin g th e
corresponding prohibitions. If it is forbidden fo r the agent to fail to support hi s
mother in 1942, say , and financia l a s well as personal support is intended, then
Practical Oughts and Prohibitions 57

he ought t o star t savin g as early as 193 2 o n the assumption tha t savin g in no

way interferes with the ability to satisfy his other obligations. That we now need
to spell out the latter assumption, of permissibility, is a complication introduce d
by my negative formulation of deontic principles. With permissibility granted,
however, closure does seem to apply to the case in negative form and to yield a
positive preparatory ought-judgment .
In negative "ought-to-do" terms, closure says that it is forbidden to do any-
thing that entail s doing something forbidden: FA and ~M( B & ~A ) imply FB.
The principle seems at least not to clash with intuition, whose deliverances are
less firm for the less familiar negative operator. "Forbidden" ma y sound odd in
application to acts whose logical or causal tie to a violation of obligation is rela-
tively remote, bu t it is not clearly ruled out, an d we seem to have no less awk-
ward phrase except "ought not," whic h does apply. It also is worth noting that
"forbidden" clearl y means something stronger than "nonideal, " even where it
is used in a phrase that suggests "ought-to-be." Similarly for "prohibited," which
I shall continue to us e here.
The distinction betwee n "ought-to-do " an d "ought-to-be, " considered a s
extending to "forbidden" an d related negative ought-terms, might now help us
to deal with some of the odd results noted (an d some others that were blurred
over) i n the precedin g section. O n th e presen t accoun t w e would see m to b e
limited t o a negativ e version of the disjunctiv e ough t tha t was brough t i n t o
resolve Sophie's dilemma at least in practical terms: a conjunctive prohibition
on letting both childre n be killed, saythat would no t b e subject to "ought" -
implies-"permissible." Otherwise , th e principl e o f closure , o r closur e plu s
agglomeration, woul d apparentl y classif y th e sam e ac t a s bot h permissible
and forbidden : A positive disjunctive ough t woul d impl y the permissibilit y of
saving a particular child in accordance wit h some sort of tiebreaker, while the
original dilemmatic prohibitions would imply that this is forbidden.
Without th e disjunctiv e ought , though , ther e wil l be nothing tha t Sophie
ought t o do. We now hav e to deny , that is , that sh e ought (al l things consid -
ered) to save at least on e childjust becaus e there is no permissible means t o
doing soeven assuming that the prohibition on letting both be killed is clearly
stronger tha n either of the dilemmatic prohibitions. The point is that Sophie is
given no moral permission to act in this case. Any particular act she might per-
formmeaning an y act-token (whethe r or not we should tak e thi s to include
disjunctive an d othe r compoun d acts)i s prohibited wit h sufficien t indepen -
dent strength t o remai n i n force even in ligh t of the conjunctiv e prohibition.
Although this result may sound odd, w e can mute its oddness b y granting
that i t still ought to b e the case that Sophi e save at leas t one child, since there
ought to be some permissible way of doing soeven in the less-than-ideal world
in which she has to act to save a child at all. There is still a distinguishable type
of act that is required of her (and hence permitted) under the circumstances; it
is merely a contingent matter that this coincides with a prohibited act-type , as
her choice-situatio n is in fac t se t up . However , prohibitio n might b e sai d t o
"dominate" permissio n for "ought-to-do": If an act-token is prohibited under
some description, then it is prohibited tout court, whether or not i t is also per-
58 Between the Horns

mitted unde r some other description . What thi s means is that al l of Sophie' s
particular options come out a s prohibited. At the same time, there still will be
some to which "ought-to-be" applies , to yield not just a version of the disjunc-
tive ought (an d its particular conclusion by way o f the tiebreaker ) but als o of
the original dilemmatic oughts whose positive "ought-to-do" versions we had
to drop. We can also say that it ought to be the case that Sophie save each child,
that is, given that "ought-to-be " ca n be limited fairly naturall y to act s under
different descriptions .
With differen t backgroun d condition s presuppose d b y differen t ought -
statements as on dyadic systems of deontic logic, "ought-to-be" should accom-
modate al l of the variou s statements we want t o mak e abou t case s o f moral
dilemma.48 Relative to the fact that Sophie has to act to save one child, that is,
it ough t t o b e the cas e that sh e doescompatibly with th e fac t tha t i t reall y
ought no t t o b e the case (and would not b e the case in a world tha t was ideal
without qualification ) tha t sh e have to ac t to sav e one child. Different condi -
tions would yiel d different an d sometimes conflicting obligations, but o n this
account the statements expressing them would essentially be insulated from each
other b y their different presuppositions in a way that would make it impossible
for the m to capture the motivational conflict characteristic of dilemmas.
This sort of fragmentation o f deontic logic into separate subsystem s is the
main threa t pose d b y dropping the principl e of agglomeration, a s I go o n t o
argue. I n any case, i t is "ought-to-do" rathe r than "ought-to-be " tha t woul d
seem t o hav e even a chance o f capturing action-guidin g dilemmas. It i s also
"ought-to-do" that coincides with the sort of everyday practical use of "ought"
that deontic logicians generally seem to b e after, despite their usual reading of
the operator. The basis of standard deontic logic in ideal conceptions of "ought "
seems to undermin e any claims it might make to capture th e elements of ordi-
nary moral reasoning. Nonstandard approaches i n the field see m to b e able to
capture the m onl y by disconnecting them, cuttin g of f possibilities of mutual
influence, as we shall see.

Deontic Fragmentation
Williams defends his decision to drop the principle of agglomeration by noting
that many of the act-evaluations that migh t b e thought t o entai l ought-state-
ments"desirable," "advisable," "sensible, " an d "prudent" are his examples
do not obe y the principle. 49 Tw o acts can eac h b e good i n one of these ways,
we might say, and ye t not b e good i n combination, since their goodness rest s
on reason s tha t cance l each othe r out . Bu t this argumen t has force onl y for a
use of "ought" whose link to "good" makes it weaker than the imperatival sense
that is in question here for cases of action-guiding dilemma. It suggests just the
sort of quick departure from dut y ethics in the fac e o f moral conflic t that my
overall argumen t here is meant to resist. Fo r th e sor t o f ought tha t rules out
alternatives, agglomeration does seem plausible: If anything but a certain act is
unacceptable, an d the same is true of another act, how ca n it not b e true of the
two i n combination?
Practical Oughts and Prohibitions 59

Since conflicting positive all-things-considered oughts hav e been eliminated

from ou r picture of dilemmas, our problem is now with agglomeration i n nega-
tive form. The reformulate d principle tells us that F A and F B imply F(A v B),
but i n a cas e o f exhaustiv e prohibition th e latte r formul a cannot b e action -
guiding sinc e its disjunctive object hold s necessarily . It disobey s the negative
correlate of the principle that "ought " implies "can": that "ought-not" implies
"can avoid." However , we should now raise a more general question about what
the apparent reference to agency here can amount to on an "ought-to-be" inter -
pretation o f the principlean d how muc h action-guidanc e i t really involves.
Interpreted i n terms o f ideal worlds th e principle that "ought " implie s "can "
essentially guarantees that a deontically perfect world i s also a possible world,
but i t does not ensur e that actio n i s needed to brin g such a world about . S o it
allows fo r ought-judgments without an y action-guiding point, suc h as a judg-
ment that one ought t o obey the laws of logic.
Odd consequence s of this trivial sort ar e ofte n tolerated i n the interest s of
systematization.50 "Ought"-implies-"ca n avoid " (a s opposed to "can"), thoug h
presupposed b y the Kantia n approach fro m whic h w e ge t "ought"-implies -
"can," i s not assume d i n standar d deonti c logic . Requirin g the inevitabl e is
thought to be harmless enough, since conformity to it is automatic. Bu t of course
this sort of failure with respect to action-guidingness may not be so harmless to
whatever claims deontic logic makes to represent ordinary moral discourse. An
ought that gives reasons fo r action, rathe r tha n just for preferring some inde-
pendent outcome , woul d appl y onl y to state s o f affair s tha t depen d o n wha t
one does .
However, the main variant of deontic logic for "ought-to-do," Castaneda' s
system, with O applied to "practitions" (intentions or prescriptions), allows for
logically harmless but practically pointless use s of "ought" to th e extent tha t
it still makes out obligation as a kind of moral necessity.51 Castaneda also accept s
the principle of agglomeration, which he at one point defends in "ought-to-do"
terms b y appeal t o a vie w of ought-statement s a s restrictin g a n agent' s free -
dom, since the extent of the restriction on freedom and the reasons for it would
have to be the same for O(A & B) as for OA 8t OB. 52 My remarks here suggest,
though, that one should consider the point of such a restrictionits importance ,
or what it achievesas part of what i s meant by the reasons for it and hence as
determining whether it constitutes a real or a tenable restriction. Ther e migh t
be a separate point in each of the restrictions on action whose combination turns
out to b e pointless in cases of dilemmato move now beyon d logically harm-
less cases on the assumption that there is also no point in requiring the impos-
sible. By analogy, there might be two restrictive dietary regimens, each offerin g
certain health benefits in return for forgoing certain foods but jointl y ruling out
so much that th e combined regimen would b e unhealthy.
We can think of Castaneda's restriction s on freedom as based o n a practi-
cal interpretation o f "ought" i n terms of reasons fo r action-guidanc e that im -
proves o n Williams' s reference t o desire s and relate d notion s o f th e good .
Castaneda seems to b e on the right track, moreover, in switching deontic logic
to "ought-to-do. " However , I think he still retains too muc h of what i s wrong
60 Between the Horns

with standar d deonti c logi c b y retaining its Kantian basis in moral necessity .
One reason h e may need to d o so, apart from his emphasis on systematic sim -
plicity, is that h e does mak e referenc e to desire s insofar a s he makes ou t th e
truth of an ought-statement in terms of its relation to the agent's optiona l ends. 53
The strengt h o f a requiremen t i n contras t t o a desireth e binding force of
"ought," in other wordsha s t o come fro m something other tha n it s source ,
on this sort o f end-based account ; th e usua l alternative is some sor t o f struc-
tural feature analogous t o logica l necessity as on Kant's view.
The intermediate "escapist" account I have suggested in this chapter essen-
tially replaces the mode l o f necessit y with tha t o f needs, strengthenin g desire
by reference to som e sor t o f harm o r ba d stat e o f affairs a s contingent upo n
nonfulfillment. 54 Th e change allows us to continue to speak o f oughts a s im-
posing restrictions on an agent's freedoma moral ought may be said to leave
one wit h n o rea l alternativ e i n moral term s t o it s fulfillmentbu t w e no w
lack the unifyin g assumptio n tha t the agent must be left wit h somethin g he is
free t o do. Needs can quite conceivably be unsatisfiable i n combination, tha t
is, as things work ou t in some particula r situation, though they still press just
as urgently for satisfaction when they conflict. They do not see m to undercut
each other in felt practical force , one might say, even where they cannot bot h
result i n action .
Considered individually , needs of the sort i n question here would see m to
imply "can " a t leas t in the sens e of holding out hop e of avoiding the harm in
question if they are satisfied. Since a set of needs does not impl y a correspond -
ing ability to satisfy al l of its members in conjunction, however, this version of
escapism provide s u s with a reasonabl e paralle l to "ought " tha t ma y indeed
violate the principl e o f agglomeration . Needin g each o f two thing s doe s no t
imply needing both of them, i f only because the har m eac h averts migh t con -
ceivably be allowed o r eve n brough t on b y the pair. Consider , fo r instance, a
patient with two disorders wh o needs two medications that interact badly , so
that the result of taking both medications to avoid ill health would be ill health.
When we switch to the negative version of agglomeration, w e can see essen-
tially the same thing in terms of the model of perceptual dominance of reasons
against various alternative actsthe negative basis of the sanctions-model , in
effectthat I appealed t o earlier. There may be strong enough reasons agains t
each of two exhaustive alternatives to rule out each of themfrom moral con-
sideration, le t us sayeven if we den y that i t make s sense to spea k o f ruling
out th e pair of them, or al l possibilities, in a single breath. This suppose s tha t
the notion of strong enough reasons for ruling something out rests on a figure-
ground relation : selection fro m a backgroun d o f further possibilities , s o that
any given act of selection ha s to leav e something over.
That tw o figure s each ma y b e capable o f standin g ou t agains t th e back -
ground, i n short, does no t impl y that the y can do s o in combination, jus t be-
cause there may be no further backgroun d to contrast wit h them; each may serve
as backgroun d for the other , as one's vie w undergoes gestalt shifts. Th e "fig -
ures" her e amount t o act s rule d ou t b y strong enoug h reasons, o r forbidde n
with sufficient strengtho n the assumption required by agglomeration as inter-
Practical Oughts and Prohibitions 6 1

preted i n "ought-to-do" terms, tha t act s related b y disjunction or the like can
be thought o f as forming compound acts .
In "ought-to-be" terms, agglomeration ma y be dealt with differently. Stron g
enough reasons agains t doing some act need not b e seen as implied by the rea -
sons wh y it (or the state o f affairs tha t consist s i n doing it ) ought no t t o be , if
the strength o f reasons for a prohibition depends on its practical point, as I have
suggested. An act that is inevitably going to be done, say, might on this account
be one that ought no t to beeve n though ther e cannot b e strong enough rea -
sons fo r prohibitin g it , sinc e no possibl e reason s coul d satisf y th e poin t o f a
prohibition b y keeping it from bein g done. I n application t o dilemmas , then ,
where the overal l choice-situation ough t not t o be , the disjunction of exhaus -
tive alternatives the situation allows for may be thought o f as forbidden in ideal
terms. I n practical terms , th e sam e prohibitio n would violat e agglomeration :
There will be strong enough reasons to prohibit each of the alternatives but not
their disjunction.
To allo w fo r agglomeratio n o f nondilemmati c oughts, w e would nee d t o
complicate deonti c logic further b y qualifying th e principleadding a clause ,
say, tha t restrict s i t i n positive for m t o ought s wit h compatibl e objects . Fo r
dilemmatic cases , o n the othe r hand , w e have to dea l wit h a kind o f deonti c
fragmentation that i s evidently not limite d to cases of incommensurable value,
if there are balanced cases like Sophie's in which only one morally relevant value
is at stake. However, a s we saw earlier, a more pervasive sort of fragmentation
into subsystems already results from the attempt t o represent conflicting oughts
within a n "ought-to-be " versio n of standard deonti c logic as provided b y the
dyadic system.
The sam e ma y now b e said o f Castaneda's "ought-to-do " syste m wit h it s
assignment o f conflictin g oughts t o differen t subscripte d contexts o f ends .
Castaneda's approac h make s at least initial sense for many of the standard case s
of dilemma, which involve a clash between different sort s of values, principles,
or th e like . However, i f a basi s i n differen t end s is enough to insulat e ought s
from eac h other, we would see m to have a distinct subsystem for every distin-
guishable source of obligation.55 Thus , i n Sophie's case obligations t o differen t
children might be assigned different subscript s to avoid intrasystematic conflict,
even though the y are grounded i n the sam e moral concerns . Bu t just becaus e
they both are grounded in moral concerns, it is unclear that this move excludes
dilemmas from morality. In any case, by disallowing logical interaction betwee n
conflicting oughts, Castaneda's approac h would apparently keep deontic logi c
from capturin g the problematic aspec t o f dilemmas.
This i s not t o sa y that a logical syste m or se t of subsystems purporting t o
represent ethic s can b e expected t o captur e everythin g interesting about real -
life ethical cases. As a proposal for structuring deontic logic, Castaneda's treat -
ment of dilemmas seems to m e to b e on the mark a s long as it is not take n a s
ruling them out except fro m a certain version of deontic logic. His own view in
fact seem s to b e that moralit y essentially is an attemp t t o resolv e conflicts, in
the first instanc e between oughts arising from differen t agents ' ends.56 At least
in rough terms, his approach resembles the account of morality as a social arti-
62 Between the Horns

fact that I shall go on to defend in chapter 3 . However, i n setting up moral sub-

systems i n terms o f contexts o f end s that ca n b e "harmonized " int o a singl e
system, as he assumes, Castaneda essentially begins with a general ban on moral

What I have tried to do in this chapter i s not to give a detailed defense of a new
approach t o deontic logic but simply to show how an adequate approach woul d
seem to spli t into subsystems in the attemp t t o captur e th e notion o f practical
"ought." This notion allow s for dilemmas on the view I have outlined by virtue
of it s relation t o th e reason s fo r an d agains t action . I have use d a perceptua l
analogy t o understan d th e compariso n o f reason s tha t all-things-considere d
practical "ought " presupposes. I n the firs t instance , the notio n require s suffi -
cient reason fo r ruling out alternativ e actions; an d I have maintained that thi s
core elemen t of its meaning may b e separately applicabl e to al l alternatives in
a given situation.
This is enough to provide a rationale for dilemmas in negative form, as cases
of exhaustive prohibition. I n positive form, as cases o f exclusive requirement ,
they seem to be ruled out by the stipulation that positive or prescriptive "ought"
also implies "permissible," understoo d a s the absenc e of sufficient reaso n fo r
ruling out its object. To allow for the distinction, I have suggested that deonti c
logic be founded instead on prescriptive "ought. " The real complications thi s
change would introduce see m to me to be not much worse tha n those require d
in any case in order to shift fro m the ideal notion o f "ought" that the standar d
system cover s t o th e ordinar y practica l notio n tha t underlie s action-guidin g
We can now understan d why agglomeration fail s for practical "ought " by
noting that the perceptual account of reasons allows for a kind of "split image "
in certai n cases : Set s of reasons ma y eac h b e important o r seriou s enoug h t o
stand i n light of each other, eve n where they cannot stan d togethe r i n the way
that might seem to b e required by a picture o f ethics as harmonizing differen t
ends. O n th e perceptua l picture , reason s sometime s compet e fo r dominance ,
prompting gestal t shift s i n our vie w of the evidence, as captured b y the sort of
division into subsystems that Castaneda use s to avoid dilemmas. I take this alter-
native picture instead to indicate in evidential terms what i t means to say that
dilemmas ar e possible . Deonti c fragmentation , i n short , doe s no t undermin e
the coherency of ethics bu t jus t its uniflability a t the level of particular action -
guiding oughts. To the extent tha t th e moral code ca n be seen as a product o f
potentially conflicting (socia l and individual ) needs, i t provides a n intelligible
basis for dilemmas .
The mai n poin t o f m y treatment o f dilemma s i n connection wit h deonti c
logic i n thi s chapte r wa s t o allo w fo r th e coherenc y o f practica l "ought " i n
dilemmasthe sort o f ought tha t i s intended t o motivat e an d tha t therefor e
would seem to fal l subject to the problem of motivational forc e outlined in my
preceding chapter. I did s o by defending the negativ e conception o f dilemmas.
On this conception, which I introduced in section 1 as a way around Williams's
Practical Oughts an d Prohibitions 6 3

denial of practical "ought," dilemmas are thought of as prohibitions of all alter-

natives open to the agent under the circumstances rather than as conflicting re-
quirements or positive ought-judgments.
I defended the positive/negative distinction in the first half of section 2 with
reference to the perceptual picture of the weighting of practical reasons as evi-
dence for a n all-things-considere d ought, o r on e tha t holds i n light of all the
evidence, as distinct from on e that is "overriding" i n the sense of being strong
enough in practical weight to cancel out competitors. Thi s amounts to a figure-
ground notion of reasons that pass a certain threshhold o f seriousness as stand-
ing out agains t the general background in a way that allows for gestalt shifts .
It allows u s to thin k of negative dilemmas as involving exhaustive all-things-
considered prohibitions, because the reason against any given alternative open
to th e agen t ma y be strong enoug h in absolute term s t o pas s th e threshhold ,
even though i t cannot b e said to cance l the forc e of opposing reasons , which
also pas s th e threshhold . The notion applie s to weighte d a s well as balanced
dilemmasto cases like Agamemnon's as well as Sophie'son th e assumption
that even reasons of lesser weight in comparison with competing reasons ma y
still be strong enoug h not to b e canceled.
Another centra l idea of this chapter emerge d fro m the compariso n o f all-
things-considered "ought" with the stronger term "must," which apparently is
not subjec t to dilemmas even in negative form. As an alternative to Williams's
interpretation o f "must " a s implyin g "will " i n accordanc e wit h a mode l of
necessitation by character, I suggested a strengthened version of the "escapist"
or sanction s mode l that som e deonti c logician s have applie d to "ought. " To
say that one ought to do something is on this account to say that it is necessary
to avoid some sanction or ba d state of affairsperhaps just something on the
order o f moral wrong , a s the genera l sort o f sanctio n associate d wit h mora l
"ought." Accordingly , to sa y that on e "must " d o somethin g i s to refe r t o a
sanction s o bad as to blo t out opposin g reason s fo r action. Thi s i s not to say
that ther e ar e n o suc h opposin g reasonsan d statement s frame d in terms of
"must" and related notions d o seem to apply naturally to cases of dilemma
but just that they do not persist i n view in the way that the perceptual picture
of reasons allow s with respect to ordinary ought-judgments. What the stronger
term rules out i n such cases is the inclusion of all of the agent's reasons fo r ac-
tion within a single background o f evidence bearing on an ought-judgment. It
therefore might be said to capture the illusion that the moral sanction i s avoid-
ablethe assumptio n buil t into standar d version s o f "escapism " whic h my
account reject s for cases of dilemma.
My ensuing defense of the logical coherency of my account and its implica-
tions for deontic logic applied the sanctions model to an interpretation o f ought-
judgments in terms oineeds, as an alternative to Kantian "moral necessity" for
understanding the imperatival force of moral "ought." A need involves the threat
of a sanction that i s avoidable by satisfying it , thoug h perhap s no t avoidabl e
simpliciter sinc e there is no guarantee that a set of needs will be jointly satisfi -
able. One might think of "must"-statements accordingly as statements express-
64 Between the Horns

ing the dominance of some particular need or consistent set of needsperhaps

only a limited subset of the overall set of practical requirements imposed by the
On th e way to thi s application o f the sanctions mode l t o needs , my argu-
ment deal t i n somewha t mor e detai l wit h bot h th e assumption s o f case s o f
dilemma and som e centra l assumptions an d principle s of deontic logic , in de-
fense o f my negative conception o f dilemmas as cases in which nothing is per-
missible. In the second hal f of section 2,1 maintained that dilemma s on my ac-
count come out as irresolvable in the sense of not being settled b y appeal to any
morally acceptabl e considerations , eve n if one accept s a stronge r disjunctiv e
ought prescribing the satisfaction of at least one of the ought s i n conflictor,
in negative terms, forbidding the violation of both prohibitions. In Sophie's case,
for instance , this amount s t o a weightier prohibition o n lettin g both childre n
be taken t o the gas chamber. Wit h a certain allocation o f harms, moreover
for instance , if saving one child could b e accomplished only by inflicting some-
thing worse than death on the otherthe disjunctive ought might come out with
the same weight as those in conflict, for a case of exhaustively balanced prohi -
I then turned, in section 3, to the logical underpinnings of the positive/nega-
tive distinction, beginning with problems about the application of the principle
that "ought" implies "permissible" t o negative oughts or prohibitions. In non-
technical terms: On e migh t object that m y defense of the principl e in applica-
tion to prescriptive "ought " would see m to apply to prescriptive "ought " just
as well, as a practical ought enjoining the failure to perform a certain action. I
essentially removed the parallelism by distinguishing between a negative ought
and a n ough t tha t enjoin s some negativel y specified action. The former doe s
not point the agent toward some indefinite alternative to action bu t rather just
away from action, with no implication that there is a real alternative. My argu-
ment dealt with th e distinctio n b y way o f a proposal fo r a versio n of deonti c
logic that takes the operator F (for "forbidden") as primitive in place of O or P,
building "ought"-implies-"permissible" int o the definition o f O so that it cov-
ers only positive ought-judgments.
In response t o Fregean objections to buildin g negation int o the operator, I
argued that th e very real complications th e move would introduc e make sense
in ligh t of the complicatin g rol e o f permissio n a s a thir d possibilit y between
obligation and prohibition in deontic logic. The appropriate analog y to obliga-
tion versus prohibition is not really assertion versus negation but rather the more
complex arra y o f possibilitie s that emerg e i f we conside r th e alternative s to
actively entertaining a propositionor "thought, " in Frege's own mentalistic
termsbesides actively blocking it out .
An important resul t of my suggestions for deonti c logi c i s the retentio n of
the principle of closure: that O A and ~M ( A & ~B ) imply OB. My view avoids
any conflict between closure and "ought"-implies-"permissible," o f the sort that
has been held to resul t from dilemmas , since i t restricts the latte r principle to
positive "ought" and reject s positive dilemmas. However, I noted independent
problems with closur e as applied to ordinar y ought-judgments. The principle
Practical Oughts and Prohibitions 65

seems to hold reliably only for statements phrase d i n terms o f "ought," since
"obligation" i s more sensitive to the wording of particular commitments, prin-
ciples, and the like , at an y rate a s applied vi a an infinitiv e phras e t o som e re-
quired action.
On th e othe r hand , standar d deonti c logi c is not reall y set up t o captur e
ought-judgments about action , wit h its reading of the operator O as "it ough t
to be the case that" (used interchangeably with "it is obligatory that"). I argued
that its claims to represent practica l "ought " depend on capturing "ought-to -
do" (i n Castaneda's terms ) as opposed t o the statements of "ought-to-be" cor-
responding to ideal-worl d semantics . In the secon d half o f section 3 I consid-
ered th e extrem e complicatio n o f deonti c logi c tha t woul d resul t fro m th e
modifications neede d t o exten d i t to dilemmati c "ought-to-do." I provided a
rationale for rejecting the principle of agglomerationthat O A and O B imply
O(A & B)which dilemmas bring into conflict with "ought"-implies-"can" a s
a principle governing practical "ought. " My claim was that ought-statement s
do not count as practical unless they actually have some point as pieces of action-
guidance, and that two statement s with suc h a point considere d individually
may lose it when agglomerated. Bu t without agglomerationor with the prin-
ciple limited to different practica l contexts i n the manner of Castaneda's "ought -
to-do" version of deontic logicoughts seem to fragment into logically isolated
subsystems. So, again, deontic logi c fails to capture practical "ought."
The model of needs for interpreting practical "ought" comes up in illustrat-
ing the failur e o f agglomeration. Ito r the sanctions model generallyyields
a way of understanding oughts in terms of negative reasons, with motivational
force provided by a threat, the prospect o f some sort of sanction on nonperfor-
mance. So my argument here extends beyond the issue of the logical structure
of dilemmas , t o sugges t a conceptio n o f mora l "ought " tha t migh t begi n t o
answer the more general problems raised in chapter 1 . The model of needs has
an obvious subjective interpretation correspondin g t o m y perceptual analogy
in this chapter, with felt need s taken as involving emotional awarenes s o f the
cost o f doing without som e object . In the case of action-guiding moral needs ,
this would seem to amount t o anticipatory awareness of a sanction on nonper-
formance. Now I want to as k whether the sort of view suggested b y the model
can accor d dilemma s and ethic s generally any "real " basis in the sens e of one
that i s not merel y subjective.
Motivational Foundation s o f Conflic t

With dilemma s now formulate d negativel y in terms o f practical "ought"i n

terms, tha t is , of prohibitions that ar e mean t to guid e actio n i n some appro -
priately stron g sense I wan t t o tak e a close r loo k a t thei r implication s fo r
the question of moral realism . Williams's treatment of dilemmas as analogou s
to conflicting desires rather than belief s was supposed t o indicate, in effect, that
ought-statements canno t describ e moral facts: facts about th e world, indepen -
dent of our moral judgments and accounting for their truth as on standard defi -
nitions of realism.1 Bu t Williams does not take account of evaluative facts : facts
about the goodness o r badness of some object. To exclude these from the realm
of facts is to be g the questio n agains t realism before the issue of conflict even
comes up . Evaluative beliefs can certainly attribute the sorts of properties tha t
Williams cites as nonagglomerativeproperties describe d by variants of "good"
like "desirable" an d "prudent"to contingently incompatible objects without
more than pragmatic conflict about which object to choose. I t would b e another
story, o f course, t o appl y the corresponding overal l assessments t o incompat -
ible objectsto pronounce eac h of them best or most desirable, say, or the only
prudent objec t of choice unde r the circumstances. 2 Bu t unless we assume non -
cognitivism, evaluative beliefs still seem to provide at least a small foothol d fo r
conflicting ought-judgment s within the category of belief .
This is essentially a version of Philippa Foot's counterargument to Williams. 3
But Williams's problem fo r belie f migh t see m to re-arise in application t o th e
motivational aspects o f dilemma. Capturing the difficulty o f dilemmatic choic e
as something impose d o n th e agent b y the ought-statement s i n conflict seem s
to require an internalist view of moral motivation, according to which it is im-
possible for a rational agen t t o hold a moral belie f without bein g motivated t o
act o n it . I n the cas e o f dilemmas , though, th e attemp t t o ac t o n al l of one's
moral beliefs would be self-defeating. So conflicting all-things-considered ought -
judgments would seem to be irrational in combination; with motivational "vec -
tors" thought o f as part of the world suc h judgments describe, they would als o
rule out the sort of independent reality presupposed b y moral realism .
Foot's comments on dilemma s seem to b e limited to a fairly wea k sens e of
action-guidance that escape s this problem a t th e cost o f failing t o suppor t a n
account o f th e motivationa l force of mora l "ought." 4 However , I think tha t
Foot's remarks elsewhere on moral teaching contain materials for an interme-
Motivational Foundations of Conflict 67

diate position on internalism that combine s with realis m to yield a defensible

answer t o th e problem o f dilemmas and motivation . I n this chapter I extrac t
that view from Foot along with some other recent authors on metaethicsnone
of whom, i t seems to me, can account adequatel y for dilemmas as things stand.
I begin by usi'ng Foot's comments on moral teaching to identify a version of
externalism that ties motivational force in general terms to the moral import o r
function o f an ought-statement, thoug h not to its specific meaning (section 1).
This vie w might be thought o f a s "genera l internalism, " thoug h i t would b e
classified as a form of externalism as these terms are usually defined. Like stan-
dard externalist views, it allows for the coherency of motivational "amoralism,"
a hypothetical moral stance that involves holding a moral belief without being
motivated to ac t on it. On the other hand, lik e internalist views, it provides a
nonaccidental connection between the meaning of moral terms and their moti-
vational force.
I shall go on to illustrate this nonaccidental connection by offering a specu-
lative account of the teaching of "ought" by reference to emotion. In phenom-
enological terms my suggestion wil l be that th e "deman d quality " associate d
with action-guiding moral "ought" rests in part on a tendency to generate antici-
patory guilt feelingsbroadly construed as self-directed emotional discomfor t
at th e though t o f responsibility for a wrong. Thi s sor t of reference to feelin g
might seem to violate realist assumptions; bu t my view is meant to allow fo r a
subject-independent accoun t o f the conten t o f any given moral judgmen t and
hence for a variant of moral realis m in conjunction with a subject-dependent
account o f motivational force. It thereby solves a problem for current concep -
tions of realism, which seem to forc e a choice between dismissing amoralism
and slighting the motivational purposes o f moral judgment.5
I attempt t o make a space for this intermediate view by examining the ver-
sions of internalism that we find i n the writings of two contemporary authors ,
John McDowel l an d J . L. Mackie, whos e position s roughl y reflec t th e Kant /
Hume contrast o n motivationa l force set up i n chapter 1 . I exhibit problem s
with bot h view s but als o extrac t fro m the m som e element s of an alternativ e
position i n section 2. I then fil l i n my own proposed vie w and defen d i t against
some objections , wit h specia l attentio n t o variant s of th e charg e o f subject-
relativity in section 3. What I hope eventuall y to put together , beginning here
and continuing in my final chapter, i s a view of the basi s of morality in a set of
socially instituted norms enforced on the individual level by a link to emotion .
Despite a conventionalist elementand a partial basis in emotion, somethin g
usually associated wit h subjectivis t approachesI think there i s an argumen t
for takin g this view as a version of moral realism.

1. Mora l Realism an d Practica l Phenomenolog y

At leas t three kinds of forc e come u p i n th e literatur e on mora l motivation :
motivational, practical or action-guiding, and reason-givin g force (sometimes
referred to as "normativity"). A (first-person) moral judgment with motivational
68 Between the Horns

force is one that actually moves the agent who accept s i t to act, providing that
she is rational. O n the other hand, a judgment with practical forc e is meant to
get her to act , an d one with reason-giving force offer s he r a reason fo r action.
Anscombe called into question the motivational force of moral "ought " on the
grounds that, as currently understood, withou t referenc e to a divine lawgiver,
it i s merely psychological. Foot, on th e othe r hand, take s ai m i n some o f he r
arguments against the reason-giving force of moral judgments, particularly in
response to the Kantian insistence on "binding" obligation. 6 However, in a treat-
ment of Hume on moral judgmen t Foot has some related comments about th e
notion o f practical force, 7 and I think that her arguments i n both place s have
implications for motivational force. In the present discussion, then, I shall blur
over the distinctions between these notions, except where they seem to make a
difference, i n the hopes of using some of Foot's remarks to brin g out the point
I have in mind.
Anscombe took mora l "ought " to b e marked of f at thi s point onl y b y its
emphatic quality. I shall try to dea l with this in psychological terms by way of
a relate d phenomenologica l propert y sometime s calle d "deman d quality, " a
property I actually take to b e shared b y nonmoral ought s tha t exhibi t a com -
pulsive hold on us, perhaps by reference to a deviation from some aesthetic or
other notion of "fitness." 8 (Consider , for instance, the feeling that one ought to
straighten a slanted picture on the wall.) The contrast i s to oughts that merely
recommend some actionas the best way of promoting ou r aims , say. What is
special abou t th e moral "ought " i s presumably something abou t th e psycho -
logical sanctions such as blame that a violation o f it would incur . For presen t
purposes, however , I shall be content t o capture somethin g broader : what we
might call the stron g imperativa l sense of "ought, " thought o f as picking out
an ought wit h demand quality.
Foot's central aim in her work on reason-giving force is to debunk the spe -
cial bindingness that Kant attributed to moral "ought " as an illusionan illu-
sion foiste d on us by the way we are taught to us e moral language. In particu -
lar, as a result of what we are taught to say about morality , amoralis m come s
out soundin g impossible: I t seems to make n o sense for someone t o acknowl -
edge the truth o f a moral judgmen t and ye t fail t o acknowledg e it s force as a
reason bearing on his action. However, I want to emphasize the positive results
of this an d relate d illusions , including some partiall y self-fulfilling result s fo r
the special force of moral judgments .

Moral Teaching and Illusion

My point about self-fulfillin g illusio n can b e seen by considering Foot' s treat -
ment o f the overridingness of moral judgments. 9 As with reason-givin g force ,
overridingness may b e thought o f a s an instanc e o f the kin d o f binding forc e
that Kant had in mindin this case, involving enough force to win any conflict
with nonmoral considerations, however important the latter are thought to be.
In both cases Foot wants to say that the property i n question is not a necessary
Motivational Foundations of Conflict 69

property o f moral judgments in the sense of applying to an y given occasion o f

their use; to thi s extent the bindin g force of morality is an illusion.
However, despit e the negative conclusion she wants to draw from it , Foot's
explanation o f th e illusio n of necessar y overridingnes s seem s t o involv e the
inculcation of habits of practical reasoning that d o accord mora l judgment s a
special place or priority. Moral considerations seem to be necessarily overriding,
on Foot's account, simpl y because we are taught t o handl e moral judgments
differently i n the fac e o f conflict, modifying th e rule s to accommodat e excep -
tions rather than treating them as "rigid rules that it is sometimes right to ignore "
on the model of etiquette.10 Thus , i f our verdic t in a certain case is that an ac t
that would otherwise b e morally forbidden is required on nonmoral grounds ,
we are taught to withdraw the judgment that it is morally forbidden rather than
treating the judgment as overridden. The result, o f course, i s that no judgment
called moral is ever overridden.
It is important to see that the illusion in question here is not simply the view
of moral judgments as overridingthey are effectivel y made overriding , afte r
all, by the way we are taught to handle thembut rather of overridingness as
some kind of intrinsic or necessary property o f moral judgments . Foot points
out that there are deviant cases, as where a code of personal hono r such as one
that prescribes duelling is given greater weight than moral proscriptions. 11 Th e
special priorit y o f moral judgment s is something w e accord t o the m i n ordi -
nary cases rather than something we discover them to have; so it is something
we can also take back .
Foot does not say that the illusion of necessary overridingness is in any way
useful. However, the attribution of a product o f moral teaching to morality itself
makes sense as a way of strengthening the habits that moral teaching inculcates.
The point als o yields a defens e o f practical illusion on Foot's main "binding -
ness" issue , the issu e of reason-giving force: I t might well be useful i n getting
people to obey moral rules without attention to their desires or interests to rep-
resent morality as necessarily reason-giving. Foot in fact suggests something of
the sortalong with a basis in moral teaching similar to the one she ascribes to
overridingnesswhen she appeals to the social purposes o f moral language to
explain why "people are taught to take moral considerations as reasons for act-
ing, without an y reference to wha t the y want, or wha t thei r interest s are." 12
Again, the upshot of such teaching would seem to be that moral consideration s
generally do exhibit the property w e ascribe to them necessarily , and th e illu-
sion of necessity may help make that so.
To extend the point to motivational force, let us take a look at some of Foot's
earlier comments on Humean practical force. In arguing against what amounts
to an antirealist account of moral judgments as needed to explain ho w moral -
ity is necessarily practical Foo t acknowledges a kernel of truth i n the claim of
necessary practicality:

It is not that this is false, bu t that one may easily insist on too close a connexion
between moral judgment and the will. . . . [W]e take it as part of the meaning
70 Between the Horns

of what we call 'moral terms' that they are in general used for teaching particu-
lar kinds of conduct; though nothing follows abou t what any particula r indi-
vidual who use s the terms must feel o r do. 13
These comments sugges t a view that migh t be thought o f as a general variant
of internalism, though o n the usual definition of this contemporary terminology,
it would com e ou t a s externalist. Something about thei r practical force is pre-
supposed by the meaning of moral terms and may even be said to apply to them
necessarily, but onl y in a collective sense. It concerns a didactic functio n tha t
they have on the whole , tha t is, rather tha n their effec t o n any given use. Pre -
sumably, it is a condition o f meaningful mora l discourse that th e terms actu -
ally fulfill thei r didactic functio n wit h some regularityenough to ensure that
moral judgments have practical forc e in general, meaning "b y an d large. " Bu t
"general internalism" (a s I shall call this view) still leaves room fo r Foot's ver-
sion of the amoralist: an agent who without irrationality claims not to be moved
by moral considerations sinc e they fail to connect with her desires or interests. 14
I want i n what follows to supplemen t this suggestion b y showing ho w we
might usefull y teac h mora l judgment s in a way tha t assign s the m a n illusor y
kind o f compulsiv e forc e i n individua l motivation. Foot' s argumen t agains t
reason-giving forc e a t on e poin t briefl y suggest s wha t I hav e i n mind , albei t
dismissively, with a mention of some illusory feelings. Her point is just that our
sense that "w e 'must do' o r 'have to do' something whatever ou r interests and
desires" lack s any basi s in belief:
[J]ust as one ma y fee l a s if one i s falling withou t believing that one i s moving
downward, so one ma y fee l a s if one ha s to do wha t is morally required with-
out believing oneself to be under physical or psychological compulsion, or about
to incu r a penalty if one does not comply.15
However, "a s if " feeling s o f compulsion o f the sor t tha t Foo t bring s in here
would see m t o b e partially self-fulfillin g i n psychological term s to th e exten t
that the y impos e a penalt y o f emotional discomfor t o n noncomplianc e tha t
makes it difficult no t to comply. 16 Onc e again, Foot's legitimate objection is to
a clai m abou t necessityi n this case, a litera l reading o f ou r feelin g tha t w e
"must" comply, or one that makes out compliance as strictly necessary to escape
some external sanctio n o r penalty .
I shall go on to sketc h an account of the teaching of practical "ought" that
makes out suc h "a s if " feeling s of compulsion i n terms of an internal sanctio n
of anticipatory guilt , taken not merel y as an anticipation o f (later) guilt at vio-
lating some prohibition bu t als o as current guilt at the thought o f a future vio-
lation "a s if " alread y committed . Eve n without supposin g tha t suc h feelings
necessitate action, I think they have an important role to play in filling out Foot's
mainly Wittgensteinian linguisti c account o f mora l teachin g wit h a kin d o f
Humean psychological glue. It will not be impossible on my suggested accoun t
to resist the force of a moral judgment, but it will be psychologically difficult i n
most cases . M y accoun t ca n stil l leav e room , however , fo r Foot' s rationa l
amoralist as someone who has managed to talk herself ou t of the feelings asso -
ciated wit h mora l teachin g or someon e o n whom mora l teachin g never quite
Motivational Foundations o f Conflict 7 1

took emotiona l effect i n the first place: There i s also no necessary connectio n
on any given occasion o r for any particular agent between moral teaching and
the feelings it sets up as part of a general mechanism for eliciting action. All my
account require s is that ther e b e some suc h mechanis m i n operation mos t of
the timeand that most of the time it operate effectivelya s a presuppositio n
of action-guiding moral "ought. "
One might be tempted t o question, though, whether this requirement really
allows for "the" amoralis t as someone completely immune to the force of moral
considerations, i n contrast t o the many normal agents who see m to exhibi t a
kind of local amoralism for certain circumscribed areas of moral judgment. That
someone migh t believe that, say, eating factory farm animals is wrong, with-
out an y desire or interes t that pull s against it , seems plausible enough, bu t i t
does not follow that a particular agent could keep morality on ice, motivationally
speaking, with respect to everything it asked of him. A certain degree of moral
motivation migh t b e thought t o b e required b y sufficient participatio n i n th e
institution o f mora l discours e t o b e said to hol d mora l belief s a s opposed t o
merely parroting moral statements. I shall not take a position o n this issue, but
I consider i t an advantag e of the genera l version of internalism that it can ac -
commodate suc h limitation s o n th e possibilit y of amoralis m withou t rulin g
amoralism out entirely.
Essentially, though, what w e hav e on m y account o f compulsive motiva -
tion in terms of emotional discomfort are short-term psychological needs in place
of Kantian moral o r rational necessit y to supply the "binding " forc e of moral
obligation for individual agents. To say that this glue is merely psychological is
not to say that it is only accidentally linked to the meaning of moral judgments;
general internalism gets between the alternatives offered u s in the standard di -
chotomy between internalism and externalism by rejecting the assumption tha t
the "force " o f a given moral term is either part of its meaning or a mere con-
comitant of it. Instead, the view holds that motivational force in a general sense
is presupposed b y a term's moral meaning , or it s role i n a certain norm-gov -
erned linguistic institution, as a condition of any specific meaning it may have.
We may think of this as analogous, say , to th e way the practical implication s
underlying th e lega l meaning of certain term s ma y depen d o n thei r role i n a
legal system. (Consider, for instance, a phrase used to affi x lega l penalties like
"in contempt o f court.") Le t us now take a speculative look a t the way such a
link between moral meaning and motivational force may be set up on the basis
of illusion.

Teaching Practical "Ought"

In teaching a child moral language we begin with the strongest instances, leaving
qualifications an d refinement s until later. Thi s mean s that "ought " doe s no t
itself figur e i n the earlies t cases I shall consider, whic h fo r th e mos t par t use
"must" ("Mustn' t d o that!" ) an d straightforward imperatives or imperatival
variants ("Don't!" or simply "No!"). "Ought" is reserved for ages at which a
child is able to exercise some judgment. Use of the weaker term leaves room fo r
72 Between the Horns

reasons on the other side: "You shouldn't d o that" advises against a given action,
perhaps emphatically, but even where i t may be said to have imperatival force ,
it does not mean the same as "Don't!" On my proposed motivational account,
then, "ought " i s introduced onl y afte r w e use stronger term s t o establis h th e
initial link to moral motivation. I f deontic terms are in question, what w e need
to see in the first instance is how we teach "must"or rather , "must not," sinc e
the negative formulation also is teachable at a less advanced stage, when a child
has begun to exhibi t action o n it s own bu t cannot ye t reliably follow instruc -
On its earliest uses, though, "mus t not" applie s after th e performance o f a
forbidden action . In advance of or coincidin g with punishmento r as itself a
form of punishment insofar as it expresses disapprovalwe tell a child that he
"must not " perfor m the act in question. Befor e action we might more naturally
use the imperative. Both expressions fi t cases where the child seems already t o
be planning to do some forbidden action an d we mean to warn him of impend-
ing punishment. Imagine a child who seems to be about to touch som e delicat e
or dangerou s object . "Mustn't touc h tha t [vase], " say , migh t b e uttered i n a
singsong, threatening tone a t this point. But a warning to the same effect som e
time in advance of action woul d see m to b e unintelligible at the earliest stage,
before th e child can even understand positiv e act-descriptions .
I conclude that practical "ought" (which I shall take as a generic term cover -
ing stronger an d negative formulations along with statements framed in term s
of "should" ) ha s its origins i n a situatio n i n which tempora l distinction s ar e
blurred i n the interest s of early moral teaching . This fits i n with my proposed
account o f the teaching of "ought" in terms of anticipatory guilt, for the accoun t
will depen d o n blurrin g the barrie r betwee n pas t an d futur e i n psychologica l
terms an d extending guil t in a backward directio n t o a time preceding action .
Another fac t that seems to fit is that i n the time right befor e or after actio n
when "must not" get s its primary use we would naturally turn to an evaluation
of the agent: "Ba d boy!" o r the like. Later, in chapter 4, I shall examine in some
detail the role of emotional guil t as a link betwee n dut y and virtue ethics. Fo r
present purposes , le t us just note tha t guil t is normally for an act , bu t a n ac t
seen a s in some wa y "tainting " the perso n wh o perform s it; it amounts, on e
might say, to a feeling ofpersonal unfitness . We encourage guilt in the teachin g
situation just described, then , essentially just by telling the child that violating
a practical ough t ha s earned hi m condemnation .
The personal evaluatio n may be linked to punishment, but punishmen t of
the usual sort might just be seen as accentuating a more general emotional threa t
conveyed by our succession of verbal utterances. Their tone is one of mounting
anger turning to all-ou t ange r with th e performance o f the forbidden act . But
anger is an emotion tha t itself ca n serv e as a kind of punishment fo r others to
the extent that it involves focusing negative attention o n them. The underlying
threat here seems to be rejectionexclusion in emotional term s from the fam-
ily or othe r socia l groupas somethin g that prompt s act s of expiation o f th e
sort associated with adult guilt: apologies and various compensatory acts in an
attempt at reparation. In this case our anger is likely to abate after punishment ,
Motivational Foundations o f Conflict 7 3

or after our expression o f anger if there is no other punishment. On the assump-

tion tha t early emotional learnin g results fro m imitation, however , one fairl y
immediate result would seem to be self-directed identificatory anger as a pro -
totype of guilt.
We may think of guilt rather broadly at this point, picking it out just as some
sort of unpleasant reaction to one's own putative offenses. Even when it is refined
(both i n childhood developmen t an d i n the fulle r versio n o f my argument o n
this subject in chapter 4), its phenomenological aspec t will be various. But the
variants all involve self-blame. The scenari o jus t sketched fo r learnin g "must
not" involve s a kind of verbal ritual acting out anger, and my suggestion is that
it is meant not jus t to modif y th e child's over t behavior , linguistic and moral,
but more fundamentally, t o modify hi s emotion tendencie s as motives toward
future behavior. It elicits guilt as a reflection of the reactions o f authority figures
or objects of childhood dependency, though th e emotion will later take differ -
ent forms via emotional identification.
A case involving the infliction o f harm on another chil d may help illustrate
this last pointand th e possible origins of a distinction between "must " an d
"ought." Consider our likely response when the child hurts his younger brother.
If our aim were just to prevent the misdeed in the future b y getting the child to
exert behavioral control, the "must not " scenari o would seem to be sufficient .
However, ou r mor e likel y tac k i n all bu t th e mos t extrem e case s i s to tr y t o
instill empathy , or a tendency to identif y wit h th e victi m of the misdee d tha t
will serve as a barrier to harmful acts in the future. 17 So instead we make a fus s
over the child who ha s been hurt, thereby showing how we identify wit h him.
Sometimes this amounts to punishing the aggressor wit h a two-person for m of
social exclusion, as we play up the fact that ou r loyalties lie with his victim; in
any case, it reinforces the sorts of direct expressions o f anger jus t described
to instil l a variant of guilt involving empathy with the victim's distress. 18
If our emphasis in the new situation were on blame (after action ) or simple
prevention (before) , we might still use an ought-statemen t frame d i n terms of
"must": "You mustn't be so rough with him!" Thi s sounds too sharp, though,
for the usual sort of case, where our primary aim is to encourage a form of prac-
tical reasoning that rests on the ability to se e things from anothe r imaginative
standpoint. To acknowledge the validity of two standpoints, or the presence of
reasons o n bot h sides , a weake r statemen t frame d i n term s o f synonym s of
"ought" sounds more apt: "Yo u shouldn' t b e so rough."
What one seeks from a child by way of overt expiation in these cases is likely
to be an apology, and here it can be directed at an injured party, a victim of his
offense rathe r than a n authorit y he has offended . W e rehearse him i n certain
linguistic rituals after th e forbidden action"Tell hi m you'r e sorry"an d i n
both sort s of cases we thereby encourage a furthe r kin d of mentalistic ritual,
involving expiation accomplished by undergoing an unpleasant emotion plus a
reading back of later emotional reactions int o the standpoint of deliberation.
The point for present purposes is that, along with simple empathy with the vic-
tim in the more complex case, we also prompt a child to fee l guil t in advance of
action i n future situation s where he contemplates doing the same thing.
74 Between th e Horns

This experience of anticipatory guilt provides a basis for our illusion of prac-
tical compulsion o n my account. A s an unpleasant feeling , guilt can b e said t o
motivate action t o ward of f a kind of emotional self-punishmen t to th e exten t
that it constitutes a motive for its own relief. The account i s not meant to imply
that guil t is the only emotion appeale d t o in teaching practical "ought " or tha t
practical "ought " in the strong or "binding" sense yields the only or the best or
highest form of moral motivation. However , i t is worth noting that the accoun t
makes out guil t as not s o clearly distinct from lov e and othe r emotions some -
times ranked higher.19 With its link to empathy or emotional identification, guilt
might b e seen as involving a kind of reflection of lov e in self-directed negative
affect. M y clai m i s just that w e nee d som e suc h elemen t of negative affec t t o
supply the demand qualit y of "ought."
In defense of my suggestion o f temporal illusion , note that we have seen two
ways in which the emotional overtones of practical "ought " may serve to extend
it beyon d th e earlies t cases pairin g i t with guilti n th e for m o f "mus t not "
applied to an act the child may already have performedto ought-statement s
preceding action, perhaps at some temporal remove. First, the anticipation with
an ought-statemen t o f an act the child seems about t o performth e singson g
warning followe d b y ange r i f he act s anywaywoul d naturall y result i n th e
association o f remembered guilt to the temptation t o perform future acts of that
sort. Second, another way in which "ought" is extended vi a guilt to acts at some
distance in the future extends it also to positive ought-statements bu t with differ -
ent objects. I have in mind the sorts of "contrary-to-duty" obligation s impose d
on the child as expiation fo r a forbidden actionobligations , tha t is , to make
amends for acting contrary t o dutyon the basis of whose fulfillment hi s punish-
ment may be lifted. These get their motivational force from guilt insofar as they
rest on a threat o f continued emotional discomfortanxiet y about social rejec -
tion o r th e likeunles s and unti l he fulfill s them . O n m y proposed account ,
this experienc e of guilt as a forc e for futur e actio n ma y b e read int o th e very
content o f th e emotionguil t become s anxiet y about no t havin g ye t mad e
amends o r th e likean d the n ma y b e rea d bac k int o th e initia l situation o f
forbidden action . S o the two source s of motivational forceanticipator y emo -
tional punishmen t and contrary-to-duty obligationcombin e t o yield a single
mechanism capable o f operating in advance .
The resul t in adult life is recognizable as an element of anxiety accompany -
ing the thought o f unfulfilled obligation . T o th e extent tha t thi s common feel -
ing amounts t o anticipator y guilt , it involves a kind of illusion, I want t o say ,
with failure to perform the required act so far conflated in emotional terms with
failure to perform it. That is not to say that the agent believes he has failed irre-
vocably, bu t jus t that th e though t tha t h e has come s t o min d a s an objec t of
discomfort i n moment s precedin g reflection . O n th e vie w I apply t o guil t i n
chapter 5 , an emotion wit h generall y beneficial consequences ma y even count
as appropriate wit h this illusory sort o f object .
The illusion here resembles an optical illusion that the perceiver understands
as such but stil l is visually misled byin the way tha t on e might be said t o be
kinaesthetically misled by the sense of falling in the passage I quoted from Foo t
Motivational Foundations of Conflict 75

debunking moral compulsion. My suggestion is that guilt in anticipatory form

amounts to an "as if" feelin g with a beneficial role as a reason for action t o the
extent that it makes it psychologically harder for the agent to violate an obliga-
tion.20 Thus, for instance, in a case where action is needed to keep a promise, a
sense of having t o act can b e seen as painfully absorbin g attentionth e agen t
is constantly lookin g over his shoulder, one might sayin a way that is due to
childhood experienc e of guilt and relate d emotions. The result is a situation in
which failure t o ac t incur s a cost i n discomfortand therefore becomes a less
tolerable option , eve n if still possibleso that th e illusor y feeling o f compul-
sion fulfill s itsel f t o som e exten t b y constituting a penalty for noncompliance .
I should stress that my suggested account of childhood experience purposely
lumps togethe r element s of practica l though t tha t ar e late r distinguished . In
particular, there seem s t o b e no clea r contrast betwee n practica l an d specifi-
cally moral uses of "ought" in the early teaching of the term in connection with
guilt. My initia l "mus t not " scenari o woul d wor k jus t as well fo r touching a
light plu g a s a vase , s o my accoun t her e applies more generall y to practica l
"ought," even though it s foundation in identification with authorit y gives it a
primitive kind of moral basis . Since guilt is thought o f as a distinctively moral
emotion, on e might be tempted t o ask how it gets extended t o nonmoral cases .
However, I think we really should sa y that th e feelin g i s not extende d bu t th e
reverse: Its moral us e is picked ou t wit h time from an undifferentiate d cluster
of cases in which it plays a role in teaching behavioral norms. We initially lump
together cases of morality and etiquette, for instance, using the threat of group
exclusion to teac h bothan d usin g shame as well as guilt for both , ignoring
later emotional distinctions .
Apart from refinements, however, the general upshot of this account i s the
way it connects th e force of "ought" to its meaningnonaccidentally, I want
to say, but not by simply building motivational force into the meaning of indi-
vidual moral terms or judgments as on standard internalist accounts. Reference
to guilt and similar emotions i s an essential part of the teaching of moral terms ,
something that underlie s their meanin g as a precondition o f their moral use ,
taken as their us e in eliciting behavior. The resul t is a picture o f the "magne -
tism" o f mora l languag e as actio n a t a tempora l distance. 21 It s influenc e o n
behavior depends on emotional demands we make in early moral teaching: what
we might distinguish fro m the meaning of a given moral ter m as its "didacti c
In ligh t o f th e wa y suc h term s ar e taught , tha t is , they retai n emotiona l
overtones fro m th e initia l teaching situation . Bu t although thi s penumbr a of
associated discomfort is essential to the motivational purpose of moral language,
it can be canceledboth i n cases of defective emotional learning and a s a result
of critical reflection. In an emotional sense we can easily see how amoralis m is
possible: A rational agent may grant that she ought no t to do a certain act and
yet contemplate doing it without guilt or any similar motivating emotion. Indeed ,
emotion will drop ou t eve n in normal cases where one acts morally just out of
habit. My suggestio n i s not tha t w e experience guilt or som e similar emotion
on typical moral uses of the word "ought" ; rather, because of their role in moral
76 Between the Horns

teaching, suc h feeling s are normall y availabl e as back-u p response s i n case s

where mora l actio n i s not automatic . A s states o f emotional discomfor t the y
can functio n motivationally i n potential form , simply by providin g a threat .
What genera l internalism tells us, in short, is that this threat i s not par t of the
meaning of a moral term , though it plays a role in setting up moral meanin g in
the firs t plac e and i s in most case s carried alon g with it .
My argument fo r genera l internalism is meant to leav e room fo r a version
of mora l realis m that assign s a fundamental motivational rol e to mora l emo -
tion. I would also like to use it to suggest something more general: the reconcep-
tion of metaethics as a branch o f moral psychology. Some recent author s have
attempted t o locat e withi n th e usual boundaries of the subject i n metaphysics
and epistemolog y a n essentiall y perceptual account o f moral judgmen t that is
realist despit e subjectivis t elements.23 Their vie w makes ou t mora l response s
on the mode l o f color perceptio n an d thu s represents the conten t o f ethics as
subject-independent in the sense of being independent of any particular subject ,
while granting its dependence on th e existenc e of minds generally. Even if we
accept something of this sort for basic value concepts, however , the analogy to
color perception seem s not to fit those responses that underlie specifically deonti c
concepts lik e moral "ought. "
Feelings of guilt, that is, and related feelings associated wit h the thought of
wrong d o no t appea r t o exis t i n unrefined for m with a certai n se t of natural
objects on th e mode l o f untutored sense-perceptio n i n an infant' s initial reac -
tive apparatus. I f the accoun t I have offered her e is even roughl y right , suc h
feelings are based on a tendency to identify emotionall y with others, including
objects o f childhood dependency , who therefor e have the powe r t o shap e the
resulting emotions. Mora l emotion s lik e guilt thus seem to be subject to socia l
manipulation o f a sort that make s cross-cultural convergence on a perceptual
model uncertain . However , I hope to hav e exhibited another wa y o f allowing
for a subjectivist element in moral realism by understanding moral emotions as
supplying the motivational forc e o f moral "ought. "
The view makes out motivational force as indeed something psychological ,
something distinct fro m reason-giving force, which I shall say more about later.
This mean s tha t ethic s nee d no t b e rationally undermined , thoug h i t lose s a
certain psychologica l prop , if one sees through the illusions that originally got
it going, assumin g on e ha s developed sufficient insigh t by that stag e t o mak e
out independen t reasons fo r it. On my proposed account , emotions an d othe r
motives in the sens e o f interna l causes o f behavio r often reinforce moral rea -
sons i n adult life ; an d sinc e mora l behavio r must b e taught a t a stag e befor e
one can adequately discern the reasons for it, this form of motivation ha s to be
taught first , in conjunction wit h simple rules. At the later stage , however , on e
can kick away the motivational ladde r and still have access to reasons capabl e
of influencin g action : the aestheti c reason s fo r no t riskin g damage t o a vase,
say, and the social or interpersonal reasons for not causing harm to others. Wha t
is illusory according to m y argument here is not th e force o f moral reasons fo r
an agent assumed to be rational but just the impression that it amounts to "bind -
ing" force , understood a s a kind of extrapsychological compulsion.
Motivational Foundations of Conflict 77

My psychologica l account o f motivational forc e also explain s i t as some -

thing essentially social. It is something with which morality is invested by a social
groupnot particularly society at large but rather the various overlapping face-
to-face groups that teach and thereby shape moral emotions. In what follow s I
use that basic scenario of moral teaching to construct a general metaethical view
that i s capable of making sense of moral dilemmas.

2. Internalis t Dilemmas
Standard approache s t o mora l motivatio n see m to be unable as they stand t o
accommodate moral dilemmas, for reasons that also yield an unsatisfying treat -
ment of amoralism. Externalist views of the usual sort and antirealis t views
the approaches I grouped togethe r i n chapter 1 as "subject-dependent"ca n
be said to make dilemmas too easy. Whether the agent in a dilemma is motiva-
tionally "torn" by the choice he has to make depends on whether h e happens
to b e moved b y both of the ought s in conflict. If someone i s not, th e fac t that
others ar e or even that their motivational propensities are generally of greater
moral valu e tha n hi s own (a s on Hare' s view ) say s nothin g t o challeng e his
response to that particular situation. The fact that we might justify the response,
if it did occur, b y appeal to its ordinary function doe s not impl y some deficiency
where one manages to avoi d it. Similarly , on a n externalis t account , nothin g
essential to moral belief seems to be lacking to the amoralist: I t comes out no t
just as possible but even as unproblematic how someone ca n accept an ought -
judgment and yet feel n o inclinatio n to ac t on it.
Internalism is linked in the first instance to noncognitivist versions of anti-
realism, which interpret the content of a moral judgment in terms of its intended
practical functio n an d henc e i n subject-dependen t terms. S o o n standar d
antirealist accounts, assuming that genuine ambivalence is possible with respect
to moral motivation, dilemmas in a subjective sense will come out a s possible
too. Thei r motivational opposition, though, will amount to nothing beyond the
agent's ambivalence . Since the conten t an d th e motivationa l forc e of a mora l
judgment will both presumably be supplied by the same mental state, whethe r
a situation is a genuine dilemma will depend on the agent's reaction to it. Thi s
leads us by a different rout e to the problem just noted fo r externalism. O n th e
question o f amoralism, however, we get the opposite result: Just becaus e th e
content o f a moral judgmen t supplies its motivational force, it will be impos-
sible to accept on e without the other.
There is another sort of antirealist view in the contemporary literature, J. L.
Mackie's "erro r theory," o n which dilemmas also come out as impossibleand
at the same time, one might still say, as too easy on the agentjust because all
moral judgments are taken to be false. On this approach, whic h assumes inter-
nalism, moral judgments do mean something subject-independent and action-
guiding, but the combination is impossible: All there really is to bac k them u p
motivationally is the agent's menta l states. The view is therefore subjectivist ; I
argue that it might be reconstructed as a form o f realism, however, by correct-
78 Between the Horns

ing standard internalis m to avoid what Mackie's theor y take s to b e our usual
moral error.
I eventually present this move as a way out o f the problem set up in chapter
1. First, though, I want t o loo k a t som e feature s of the internalis t version of
realism that now seems to hold the field. This is John McDowell's cognitivism ,
defended i n opposition to the desire/belief model of intentional action insofar
as it takes moral beliefrathe r tha n belie f supplemente d b y a furthe r menta l
state, typically desireas sufficient t o generate action. 24 Her e to o I think that
refining the view to make it accommodate dilemma s and amoralis m will yield
something close r t o m y own position . I n bot h cases , fo r tha t matter , I think
that the necessar y refinements can be constructed largel y out o f materials the
author provides. McDowell's remarks at one point suggest that it might be belief
as an object of current attentio n rathe r tha n belie f alone , a t leas t i n the ordi -
nary sense, that constitutes the motivating cognitive state he has in mind. This
is something that m y own vie w interprets i n terms of emotion, vi a an evalua-
tive analysis that als o seems to accommodat e McDowell' s characterizatio n of
moral insigh t as a specia l kind o f sensitivity. However , m y understandin g of
emotion will depart fro m McDowell's perceptual model with an account of the
tie between emotions and actio n tha t allow s for moral dilemmas .

Cognitivism and Motivational Sufficiency

Though McDowell' s vie w is developed i n a numbe r of article s o n differen t
subjects, it s main point s relevant to motivationa l issue s can b e foun d i n a n
early critical piece on Foot alon g with a footnote allusio n to dilemma s in an
account of virtue.25 As in my treatment of Foot, I shall not attemp t a detailed
exposition o f McDowell's overal l metaethical view but instea d shall bring in
particular points an d position s a s they affect th e issue s under discussion. In
brief: McDowell defend s an Aristotelian notion of virtue as based o n a kind
of perceptio n o f reasons fo r actio n i n opposition t o Foot' s view of morality
as simply not giving reasons to someone wh o has no independent motivation
to act morally, of the sort provided by desire on the standard model . Reason -
giving forc e o n McDowell' s accoun t doe s no t depen d o n desir e as a n inde -
pendent factor ; rather, i t is something tha t the virtuou s person perceive s as
applying to the particular situation, with desire taken as following from tha t
McDowell agrees with Foot, then, that a n agent who is unmoved by moral
considerations i s not necessaril y irrational bu t instea d make s her out a s mor -
ally blind: She fails to se e the reason tha t morality indeed provides. What she
lacks i s an accurat e vie w of th e practica l requirement s o f th e situatio n tha t
McDowell identifies with virtue on the assumption that it depends on a similar
appreciation of all potentially competing requirements and hence amounts to a
general perceptual capacity. It is in reference to his discussion of this version of
the doctrin e of the unit y of the virtues that McDowell includes a footnote rel-
evant to dilemmas.
In defense of a claim that the virtues of kindness and justice presuppose each
Motivational Foundations of Conflict 79

other, since the agen t wit h eithe r virtu e ha s to b e able to pic k ou t situation s
calling for its exercise fro m thos e that cal l for the other, McDowell notes :
I do not mea n to suggest that there is always a way of acting satisfactorily (a s
opposed t o makin g th e bes t of a ba d job) ; nor tha t there is always on e right
answer to the question what one should do. But when there is a right answer,
a virtuous person should be able to tel l what it is. 26

McDowell's firs t clause here is as close as he comes to taking a position o n the

possibility of dilemmas. But it might be read as substituting an explanation short
of dilemmatha t sometimes al l alternatives are morall y objectionable in th e
sense o f nonidealfo r the sort s o f cases o f impermissibl e alternatives that a
proponent o f dilemmas would cite. I take the rest of the passage to imply a rejec-
tion of genuine dilemmas. In such cases there might be said not to b e one right
answer (a s in McDowell's secon d clause ) to th e questio n what to do , bu t this
may be understood i n a way that does not allow for the substitution (in his next
sentence) of no right answer, meanin g no requirement of the sort that a virtu-
ous person apprehends .
McDowell's substitutio n rests on the assumption tha t a right answer to the
question what one should do amounts to one that specifies a right act. This works
well enough for the positive conception o f dilemmas on the assumption tha t a
right answe r is an ought-judgment prescribing a permissible act. W e can no w
see that the sor t of problem abou t th e motivationa l sufficienc y o f conflicting
ought-judgments that might seem to undermine McDowell's cognitivismth e
problem I used Kant' s vie w to illustrat e in chapter 1doe s no t aris e for th e
positive conception . W e do not hav e to dea l with case s i n which each o f tw o
conflicting all-things-considere d ought s yields sufficient motivatio n to act, o n
McDowell's accoun t o r Kant's , sinc e such cases ar e rule d out b y the assump -
tion that "ought " implies "permissible." Bu t the problem can easily be formu-
lated fo r exhaustive prohibitions .
On McDowell's account , the virtuous agent must be sensitive to any and all
act-requirements imposed by his situation, including of course negative require-
ments or prohibitions; accordin g to standard internalism , his sensitivity would
seem to be sufficient b y itself to generate action. Bu t then, if there were cases of
genuine dilemma, McDowell's notion s of sensitivity and motivational sufficienc y
would apparentl y impl y that th e virtuous agen t perform s tw o incompatibl e
actions. I t follows that McDowell's accoun t cannot accommodat e dilemma s set
up a s case s o f exhaustiv e prohibitio n unles s it someho w modifie s standard
internalism. His footnot e evade s this problem , i n effect , b y putting the ques -
tion in positive terms and equating the existence of a right answer with that of
a right o r satisfactor y thing to do . With th e question se t up negatively, how -
ever, there will be two righ t answerstwo prohibitionswhic h togethe r yield
the result that ther e is no right thing to do .
It migh t b e though t tha t McDowel l doe s mea n t o modif y standar d
internalism, though in a way quit e different fro m wha t I have in mind, insofar
as he relies on a notio n o f motivationa l force that connect s i t t o th e forc e of
reasons. Tha t is , there would see m to b e no distinctio n betwee n motivational
80 Between the Horns

and reason-giving force on his viewor on Nagel's Kantian version of internalist

realism, though Nagel does accept dilemmas. Perhaps an internalist in this sense
can simply deny that motivational force has to involve a push toward action
has to determine the will, in Kantian terms. Instead, in some cases i t might just
involve appreciating th e practical relevance of a reasonor, in the case of dilem-
mas, it s decisiveness. On thi s account, whethe r on e actuall y i s moved t o ac t
would depend o n the absence of all-things-considered competing reasons .
Both Nagel an d McDowel l clearl y want motivationa l forc e to impl y more
than an intellectual recognition of the reasons bearing on action, however. The
notion i s supposed t o replac e desir e in explaining the generatio n o f action b y
reasons.27 T o withdra w an y pus h towar d action , then , i n the fac e o f seriou s
moral conflict, would mak e dilemmas too easy on the agent. That is, the agent' s
conflicting reasons would apparently just be rendered inert by conflict, reduced
to a mere list of negative features of action tha t leav e him with nothin g to do .
Rather tha n bein g motivationally "torn, " or impelle d in opposing directions ,
he would seem to be in a condition o f practical stalemate and henc e simply frus-
trated at his inability to settl e on a course of action.
I shall eventually propose a way of modifying McDowell's Aristotelia n view
in connection with the issue of amoralism that will yield a preferable approach
to mora l dilemmas . But let us note firs t tha t certain moves that migh t b e sug-
gested fo r applying his view to dilemma s by requiring a lesser degree of moti-
vational force are not reall y open t o him, given the kinds of demands hi s view
makes on the notion o f a virtuous agent. One might want to object, for instance,
that the motivational vacillatio n or even deadlock that would see m to be man-
dated b y cases o f exhaustive prohibitio n coul d b e explained b y attributing t o
the agent motives o f less than full strength . We should see the virtuous agent in
such cases as moved t o some exten t b y each of two competin g prima facie rea -
sons. To say that, however, woul d jus t be to say that th e case doe s not consti -
tute a dilemma; we should remind ourselve s of the assumption tha t dilemma s
involve conclusiv e o r all-things-considere d ought s i n conflict . Presumably,
moreover, th e virtuou s agen t woul d hav e to b e sensitive to eac h o f them i n a
degree that reflects their moral importance in absolute as well as in relative terms.
Agamemnon, for instance, falls short of virtue because he does not sufficientl y
register the mora l horro r of the sacrific e of his daughter an d henc e is insensi-
tive to one side of his dilemma. On the other hand, we cannot say that th e vir-
tuous person trie s but fails to act on two conflicting reasons that she is sensitive
to, for she is assumed t o b e rational, an d rationalit y rules out actio n a t cross -
I shall suggest a way o f handling this problem; bu t les t it be thought eas y
enough for McDowell t o handl e it just by rejecting dilemmas , we should als o
take note of a related proble m with respect to amoralism. McDowell' s answe r
to Foot's treatment o f the amoralist in effect interpret s th e amoralist a s some -
thing elsesomething othe r tha n th e motivational amoralis t unde r consider-
ationby making him out as morally blind, or unable to appreciate moral rea -
sons o n a cognitiv e level . McDowell' s alternativ e to th e desire/belie f mode l
Motivational Foundations o f Conflict 8 1

essentially rests o n expandin g the notio n o f what i s involved at th e cognitive

level so that it would not be possible to share the same view of the circumstances
as someone who i s morally sensitive and ye t not b e similarly motivated.28 T o
the extent that the "view" i n question here involves seeing a certain action in a
favorable light , it implies seeing a reason t o perform the action and hence has
reason-giving force.
McDowell doe s not distinguis h motivational from reason-givin g force, as I
have noted . Presumably he would coun t i t as eithe r impossibl e o r irrational ,
and hence as impossible for a virtuous agent, t o recognize a reason fo r actio n
and ye t not b e motivated. Bu t the amoralist n o les s than th e virtuou s agent is
supposed to be rational. McDowell's answer to Foot, then, essentially involves
denying that the amoralist really recognizes the same moral reason s fo r action
as the rest o f usor i n the firs t instanc e a s the virtuou s agent, whose reason s
we share when morally motivated. This response may be difficult t o fault intu-
itively just because the problem of amoralism a s set up in global terms rests on
acceptance of a counterintuitive kind of moral personality. To the extent tha t
we can mak e sense of it (b y appeal to literar y examples, say) , it is sufficientl y
foreign t o pu t th e amoralist' s cognitiv e stat e somewha t beyon d imaginative
reach. In response t o example s of our ow n "local " amoralism, howevermy
earlier exampl e wa s eatin g factor y far m animalsMcDowel l woul d hav e to
say that we could no t reall y share the virtuou s person's view of eve n th e iso -
lated sor t o f situatio n tha t i s in question. A s a moral "blin d spot, " thi s als o
involves failure to see .
An argumen t over thi s sort o f case woul d typicall y focus on th e questio n
whether th e agen t professin g motivational amoralis m coul d possibl y be ac-
quainted i n a ful l enoug h sense with al l relevant information, especially first-
hand or imaginative information such as that giving insight into the misery of
animals on factor y farms. However, i t is important tha t thi s lin e of respons e
would move beyond the specific focus on belie f a s the beare r of moral knowl-
edge in the desire/belief model from which McDowell pulls away. It is indeed a
kind of "perceptual " acquaintanc e with th e factsbut in a more litera l sense
than McDowell has in mind in his use of perceptual imagery for moral insight
rather than simpl e intellectual comprehension tha t is likely to make a motiva-
tional differenc e i n the case cited.
McDowell himself might just say that th e supposed amoralis t abou t eating
factory far m animalsth e agent who claim s to recognize that it is wrong an d
yet not to be motivated to abstaincould not really understand what it was to
"recognize" a practical reason. The notion entail s acknowledging the bearing
of some consideration o n one's choice of action, which effectivel y undercut s a
distinction between intellectual and motivational acknowledgment. However ,
it is not clear that the only way to recognize the practical relevance of a reason
is to be inclined to act on it. One might simply use it to fudge actionacknowl -
edging that eatin g factory farm animal s is morally substandard, say, without
feeling impelle d t o liv e up to a moral standard so rigorous a s to deman d that
one abstain . To captur e the decisivenes s of a serious moral reason , w e could
82 Between th e Horns

understand th e relevan t sort o f motivational amoralist as measuring his own

and other agents' virtue in light of the standard, deferring to vegetarians as moral
exemplars but still without motivatio n t o joi n them.
Besides dropping desire , then, I think McDowell's exclusio n of amoralism
must turn on an expansion o f the sorts of cognitive possibilities allowed for by
the desire/belief model. "Cognitive " come s out, i n short, as a broader categor y
than "intellectual " on an approach tha t would support hi s talk of moral sensi-
tivity and the like. The move is masked by McDowell's us e of the standard ter -
minology: He evidently wants to include some conative elements within belief
rather than questionin g whether belie f i n the usua l sense is sufficient. T o dis -
cuss the mov e in terms tha t relate i t to m y own treatmen t o f emotion, I shall
use "belief" in what follows more narrowly than McDowell, though one might
also substitut e som e mor e qualifie d notion suc h a s "belie f i n it s judgmental
aspect" fo r th e sak e o f faithfulnes s t o McDowell' s preferre d blurring of th e
standard distinctions .
McDowell's inclusio n of extraintellectual elements within cognition is in fact
suggested b y some brie f comments o n incontinenc e at th e en d o f his reply to
Foot.29 Since he no longe r ha s a purely conative category like desire available
to explain continence, McDowell ha s to make use of the cognitive, and here he
appeals beyond belief to attention, which he interprets as part of the continent
person's "conception " of the situationthe most intellectual of the terms replac-
ing "belief " i n his substitute fo r th e desire/belie f model. (H e sometimes use s
"perception," apparentl y withou t distinction, 30 bu t hi s most commo n term ,
encompassing both, is "view.") O f incontinent or weak-willed agents he writes:
Their inclinations are aroused, as the virtuous person's ar e not, b y their aware-
ness of competing attractions : a lively desire clouds or blur s the focu s of their
attention o n "th e noble." 31
The comment hark s back to McDowell's precedin g account of Aristotle's vir-
tuous person as exhibiting temperance or single-minded moral motivationas
distinct fro m mere continence, or strong-willed triumph over contrary motiva-
tionby "silencing " immora l inclinations. 32 I n McDowell's exampl e i n tha t
discussion the virtuous person simpl y lacks any desire for illicit sex rather than
manages to overcome one. In a case of weakness, then, desire can be made out
as an interfering factor, though McDowell' s cognitivis t view deprives it of any
independent role in virtuous motivation .
What doe s i t interfere with i n cases o f weakness, an d how ? If we remind
ourselves of the distinctio n betwee n a belie f an d a thought tha t on e actually
entertains o r hold s i n mind , w e ca n replac e McDowell' s Aristotelia n tal k of
attention t o the noble with a reference to whatever mental mechanism serves
to fi x moral judgment s in mind . Thi s ma y o r ma y no t b e purely cognitive; I
take i t typically to involv e the sor t o f cognitive/affective mi x tha t w e fin d i n
moral emotion . A t any rate, it need not b e seen as involving the sort of sepa-
rable or independent noncognitive element that McDowell means to exclude in
dropping desire. On the other hand, in most cases it involves something beyond
belief. Belie f i s just what th e weak-wille d agent is assumed to shar e with on e
Motivational Foundations o f Conflict 8 3

who is virtuous: what the person overcome b y adulterous passion lose s sight of
but stil l retains. H e know s wha t i s right bu t fail s t o ac t on his knowledge jus t
because he does no t hav e it fixed firml y i n mind.
This nee d no t mea n tha t th e weak-wille d agen t i s not thinkin g abou t th e
moral truth in question, even in relation to the noble; he may very well be acutely
aware tha t h e is acting against hi s beliefs . I t is a fulle r kin d of practical atten -
tion tha t is presumably lacking. Somethin g simila r might no w b e said o f th e
amoralist, however. His moral sensitivity has been "silenced," t o use McDowell's
term, not b y desire but in this case simply by a failure to generate the appropri -
ate moral emotio n a s needed t o supply motivation .
This i s not t o sa y that a moral emotio n i s always needed to motivate. M y
speculative accoun t o f the teachin g o f "ought " i n section 1 allowed tha t th e
moral behavior we inculcate via emotion normally becomes habitual. In typical
cases habitual moral behavior is not quite automatic but can be said to involve
an emotional residue of the teaching situation, in the sense of arousal sufficien t
to fix a belief in mind, securing attention. However, i t need not involve a specific
moral emotion; McDowell' s "sensitivity " will do. I would also grant that belief
alone can motivate, meaning "can i n some cases," even without attentiona s
when th e habi t o f acting on som e mora l belie f become s s o ingrained as to b e
discharged in rote fashion. Habit in such a case would no t amount t o a further
mental determinant of action needed to supplement belief but just a pattern o f
action on belief. It need not be taken strictly as "determining" action ; the com -
mon inferenc e fro m a claim that som e menta l state or facult y ca n motivate t o
the view of it as sufficient i n something lik e a causal sense is an instance of th e
Kantian "necessitarian " approac h t o moralit y and motivatio n tha t m y argu -
ment here is meant to question. It is not obvious that anythin g with the power
to motivat e mus t d o s o unles s checked, o r tha t ingraine d habit s ar e alway s
compulsive. In any case, the amoralist is assumed to lack normal habits of moral
behavior; so at least in short-range terms, motivating him requires an emotion .
Whether th e amoralis t o n thi s accoun t ma y b e said t o shar e th e virtuou s
person's "view " o r "conception " o f the situation"perception " i n a broa d
sense, roughly equivalent to "apprehension," withou t any sensory overtones
depends on whether we interpret such notions as covering the full range of cog-
nitive and auxiliary responses t o the moral facts, including attention an d emo -
tion, or as limited to belief. At any rate, the notion of emotional appropriatenes s
that I explain in chapter 5 will allow fo r the failure to fee l an appropriate emo -
tion compatibly with rationality. Appropriateness doe s not mandate feeling , in
short. Bu t this point wil l effectively driv e a wedge betwee n reason-givin g an d
motivational force .
Although McDowell doe s not explicitly question the standard categories for
explaining action that he has inherited from the philosophers h e criticizes, many
of his comments i n his reply to Foo t and elsewher e suggest my broader inter -
pretation of the cognitive. Here, for instance, he at one point brings in our under-
standing of th e "meaning " o f som e morall y significan t circumstanc e as illus -
trated b y a statemen t w e migh t mak e i n trying to ge t someon e t o shar e th e
requisite view of things: "You don' t kno w what i t means that someon e i s shy
84 Between th e Horns

and sensitive." 33 Someon e wh o made this statement woul d typicall y be claim-

ing not jus t that th e hearer lacke d some factua l information abou t th e behav-
ioral or other implication s of shyness but rather tha t he had no idea what shy -
ness was like. This is what shyness means to the one who is shy, and the reaso n
why i t inhibits him behaviorally. Without experiencin g shynes s "fro m th e in-
side," on e is at leas t arguabl y unable to understan d it s outer meanin g in the
sense of its behavioral significance. Real knowledge of this sort, however, seems
to requir e imaginatio n an d emotiona l empath y a s part o f the recommende d
"cognitive" vie w of things .
More generally, meanings in the sense indicated include saliences of the sor t
that emotions register : the notion o f what i s important o r significan t about a
situation, o r in cognitive terms, what i s worth attention. 34 McDowel l hasten s
to point ou t that the appeal the statement abou t shynes s makes is not to "pas -
sion a s opposed t o reason, " or to feelin g take n a s something "quit e ove r and
above one's view of the facts," 35 th e sort o f independent element of desire that
he wants to eliminate. My suggestion here, however, is just that emotional evalu-
ations ar e part o f one's view of the facts. On th e account o f emotions I defend
later in application to guilt, emotions includ e evaluative thoughts a s objects of
comfort o r discomfort . I n McDowell' s perceptua l imager y o f silence d an d
unsilenced reasons, they may be thought of as an important menta l mechanism
for amplifying reasons , o r for amplifying an d recordin g them for futur e refer -
enceregistering them a s objects of practical attentio n b y loading the m wit h
positive o r negative affect .
We can no w se e how m y account migh t le t McDowell handl e cases o f di-
lemma. Here, on any adequate picture, we have two emotionally amplified rea -
sons as a norm for the appropriately sensitive moral agent, not just as a descrip-
tion o f how some agent s happen to react . O f course, i t is impossible t o act on
two conflicting moral beliefs, but the possibility of emotional ambivalence allows
for a form of practical attention t o both o f them.36 Wit h the agent's conflicting
reasons fo r action seen as given in the evaluative content o f the two emotions ,
we can make him out as motivated by both o f two conflicting ought-judgments.
Action on one of them will be blocked, of course; but a "residue" emotio n suc h
as guilt will remain as a sign of its motivational forc e deflected onto emotion .
I shall later argue that this function of guilt depends on its component o f nega-
tive affect .
Making McDowell' s view accommodate dilemma s requires modifying hi s
commitment t o th e motivationa l sufficienc y o f cognitiv e commitmen t t o a n
ought-judgment, if that means its causal sufficiency fo r action. M y account ac-
complishes this by abandoning strict internalism in favor of the general version
defended i n section 1 , according to which an agen t can understan d a n ought -
judgment and fai l t o registe r i t motivationally. M y accoun t take s emotio n a s
the primar y beare r o f motivationa l force an d henc e introduces a furthe r ele -
ment besides belief to explain moral motivationsomething added onto belief
in typica l cases and potentiall y detachable from it , even i f not quit e indepen-
dent of it on the model of desire. Belief may fai l to motivate either because this
further elemen t is missing or because its motivational influence i s blockedthe
Motivational Foundations o f Conflict 8 5

latter pertainin g to cases of dilemma, where not al l oughts ca n be satisfied, so

that emotion serves as a second-best substitut e for action.
A version of McDowell's persona l mora l standar d migh t b e said to pla y a
role in my own view to the extent that the weaker normativ e link my view sets
up betwee n mora l judgments and emotio n stil l depend s fo r its application t o
dilemmas on th e possibility of conflicting emotions i n a rational agent . How -
ever, my account of the role of guilt in chapter 4 will depart from the Aristotelian
standard o f perfect virtue on which McDowell relies . I n any case, rather tha n
presupposing a form of virtue ethics of the sort McDowell has in mindI shall
later mak e some alternative suggestionsmy defense of dilemmas admits of a
coherent interpretatio n i n terms of general principles of duty : the conceptio n
of ethics that McDowell's perceptua l imagery is meant to replace with a "par -
ticularist" moral theory. Let me now turn to Mackie's view for a notion o f the
rule-based structure of morality, with morality reconstructed t o answer Mackie' s
arguments against realism.

Protagorean "Social Artifact Realism"

The ordinar y conceptio n o f rea l perceptua l propertie s associate d wit h
McDowell's versio n of realism actually makes it see m eas y to accommodat e
dilemmas. In cases of exhaustive prohibition, al l we need are noncomparative
negative properties tha t apply to all alternatives, and th e list of possible paral-
lels i s a lon g one . Conside r th e propertie s o f being horrible o r hideou s o r re -
voltingor eve n (on certain assumptions) the sort of thing that would disgust
an ideal observer. A moral term with simila r visceral overtones tha t woul d fit
my perceptual treatment of sufficient reason s against action in chapter 2 is "in-
tolerable." Th e possibilit y of dilemmas as expressed i n such terms jus t means
that every option in some cases can fail to meet some minimal standard of moral
What threaten s th e intelligibilit y o f thi s vie w i s th e presume d action -
guidingness of moral judgments, as attributed to the moral facts to which they
refer on an internalist version of realism. This is essentially the basis of Mackie' s
"argument fro m queerness " t o th e denial of the objectivit y o f moral values. 37
Mackie use s the term "objectiv e prescript!vity" fo r the "queer " combination
of claims tha t h e takes to b e implied both b y moral discours e o f the ordinary
sort and b y the writings of most mora l philosophers . Prescriptivit y or action -
guidingness is a property of moral language , but insofa r a s we use moral lan -
guage to describe something objective, Mackie's accoun t make s us out as attri-
buting prescriptivity to something in the world. The account therefore amounts
to a n "erro r theory": In contrast t o noncognitivist account s tha t den y moral
statements descriptive meaning it interprets them as meaning something false. 38
However, Mackie' s comment s at some points suggest that the error migh t
be removable: Ordinary moral discourse could be revised to stic k to th e facts
(as Mackie conceives them) essentially by canceling its prescriptive force. Rather
than droppin g internalism , as thi s entails, Mackie evidentl y means t o revis e
ordinary moral discourse in the direction of noncognitivism. But an alternative
86 Between th e Horns

move of the sort I have in mind is suggested i n his discussion of naturalist defi -
nitions of moral terms with referenc e to a n assumed se t of purposes, which is
essentially a reworking o f Foot's treatment o f "good." 39 Mackie himsel f by-
passes the suggestion fo r all but a few cases involving fixed standards o f evalu-
ation; values would apparently be less than fully objectiv e in the sense he has in
mind to the extent that the determination of standards rests on potentially vari-
able purposes. Hi s ow n eventua l definition o f "good" makes indefinit e refer -
ence to the satisfaction of requirements, and he insists that statements o f moral
value must be taken as ascribing "intrinsic requirements" to the world: "require -
ments which simply are there, in the nature of things, without bein g the require-
ments of any person o r body of persons, eve n God."40 This is the assumption I
want t o challenge , for "ought " rathe r tha n for "good," alon g wit h Mackie' s
general assumption o f internalism.
Mackie ofte n refer s metaphoricall y t o th e questio n o f objectivit y as th e
question whether moral values are part of the "fabric" o f the world; his descrip-
tion of behavior as part of its "furniture" suggest s that h e has in mind a spatial
network o f relations amon g soli d objects. 41 Instead , w e might thin k of moral
values as analogous to the weave in a woven fabricsomething inseparable from
its threads, afte r allan d a t the sam e time a s relating behavio r to person s o r
minds to the extent tha t moralit y rests on a relation to their purposes, or their
harm and benefit. I take it that this latter sort of dependence on minds does no t
undercut realismdoe s no t constitut e "subject-dependence " i n th e relevan t
senseif i t leaves intact the role o f minds as knowers o f moral truths. Macki e
uses the term "objectivity " fo r the view he means to attack , an d h e attributes
the view to philosophers who base moral value on subjective states like pleasure/
pain, as well as to Plato an d Kant. 42 What mora l realis m rules out, however , is
subject-relativity, meanin g relativity to the putative subject of knowledge of a
moral judgment as opposed to the various subjects of experience the judgment
might be thought t o be about.
A realist view can even make moral valu e depend on the existence of minds
as knowers of some judgment s of the sort that i s in question, a s long as it does
not make a given judgment depend for its truth on someone's current commit -
ment to it . Since Foot's naturalist definition of "good" appeals t o a standar d
set of purposes, not necessaril y those of a given subject, it would see m to come
out as realist on this account. Moreover, McDowell's treatmen t o f moral value-
properties on the model of perceptual properties i s interpreted as realist, thoug h
it makes moral judgment s depend i n a general sense on ou r possessio n o f the
capacity fo r moral sensitivity. 43 I want i n what follows to suggest a further leve l
of general subject-dependence for specifically deontic properties as compatibl e
with mora l realism . I f we take fo r grante d eithe r o f thes e othe r account s i n
application t o evaluative propertiesmaking the m out as "subjective" onl y in
a metaethically harmless sensethe n I think we can also understand mora l re-
quirements as requirements of persons rathe r than intrinsi c requirements with-
out departin g fro m realism. On th e sort of interpretation I shall suggest, they
depend on prior social requirements within certain constraints imposed b y con-
Motivational Foundations o f Conflict 8 7

siderations of social value, including both rationa l and moral value as applied
to the comparison o f alternative moral institutions .
What I have in mind are the various social choices and attitudes that under-
lie the adoption an d maintenanc e of a moral code. I shall go on to extract my
proposed versio n of realism, which I refer to as "social artifact" realism, from
the picture of morality given in Mackie's own transition to his account of nor-
mative ethics. But my extension o f Mackie's view should also apply in general
terms to rather different conception s o f the basis of morality, such as a version
of divine command theory tha t takes God a s the sourc e o f moral ought-judg-
ments (perhap s with huma n welfare in mind) but no t o f judgments of good.
The view also might be taken to yiel d a contractarian basi s for morality that
avoids Kantian presuppositions.44 I t combines elements of virtue and duty eth-
ics, as we shall see. I put i t forth as a way of reconfiguring the structure of eth-
ics, with a different metaethica l basis insofar as it departs fro m standar d posi -
tions on the question of internalism.
After denying moral judgments objectivity, Mackie proceeds to a treatment
of normative ethics in terms of the function of morality as a system of constraint s
on conduct designed to protect others' interests. 45 Moralit y on this account is
needed to counteract limited sympathies, or a tendency toward self-interest, and
its content depends on what will best promote tha t cooperative social end. But
this suggests that evaluativ e judgments of a sort that Mackie apparently find s
unproblematic, about the effectiveness of means in promoting ends, might help
provide an objective basis for some moral judgments even if not thos e in ordi-
nary moral discourse. 46 I f so, morality could presumabl y be reconstructed t o
avoid the error o f attributing action-guidingness to something external.
Perhaps Mackie did not thin k of this position a s "objectivist" jus t because
it assigns an important rol e to moral emotion. The mythical account of the his-
torical basis of morality that he takes from Plato' s Protagoras splits it into tw o
elements: aidos (shame or respect; a word with broader implications that Mackie
decides to translate as "moral sense" ) and dike (law or justice).47 Mackie inter-
prets dike t o cove r formal rules and politico-lega l "devices " se t up t o secure
the aims of morality, bu t the term would also see m to appl y to various infor -
mal social practice s with th e sam e end, such as promising, that his preceding
argument includes under the umbrell a term "institutions. " These com e u p as
central devices of morality when Mackie fills in the Protagorean vie w with criti-
cal discussions of Hobbes's contract and Hume's artificial virtue of justice, con-
cluding with a rejection of the suggestion that ethics abandon rules in favor of
virtues. We might think of dike, then, on Mackie's interpretation, a s the basi s
of morality in rules.
Mackie seems'to restrict aidos, on the other hand, to the sources of motiva-
tion to conform to rules, dispositions t o act as well as emotions. Bu t it would
be naturaland consistent with Plato's use of the Protagorean viewt o extend
this term to the looser sorts of emotional and behavioral habituation that under-
lie moral virtue. I shall chop things up somewhat differently fro m eithe r Plato
or Mackie, however, by also allowing for an extension of dike to looser norms
88 Between the Horns

or standards o f behavior of the sort that virtu e ethics stresses. On this reading ,
dike an d aidos amount to external and internal aspects of moralitysanctions,
standards, o r what have youcorresponding in my own account to the subject-
independent and -dependen t components o f moral meaning.
On a les s mythological historica l account tha n w e fin d i n the Protagoras,
perhaps th e tw o notion s shoul d initiall y b e taken a s combined in some mor e
primitive idea such as that of a "taboo," understoo d a s a particular emotion -
laden rule. For m y purposes here , what i s important i s just that bot h pla y a n
essential role in morality as it has developed; dike i s conceptually distinguish-
able at an advanced stage as a moral code (broadl y construed) and aidos as the
source of motivation to act on the code, which cannot b e objective in Mackie's
terms since motivating states are subjective. Mackie's internalismhis assump-
tion that motivation must be built into the content of moral judgmentskeeps
him from accommodatin g th e Protagorean accoun t within a version of moral-
ity that would escap e his charge of error. However , withou t that assumption ,
Mackie might be seen a s allowing for a for m o f moral realis m in his own re -
marks on the aspect of morality that h e takes to b e worth pursuin g further.
The Protagorea n accoun t woul d fi t in with Mackie' s earlie r attempts t o
understand the basi c moral term s i n a way that cut s acros s position s o n th e
issue o f objectivity . We ca n se e how i t migh t b e defende d a s a n externalis t
form o f realism in connection wit h hi s comments on "ought. " His first us e of
the notio n o f a n institutio n come s fro m Joh n Searle' s attemp t t o clos e th e
Humean "is/ought " gap with a n argument, 48 vi a "institutional facts " abou t
promising, fro m th e fac t tha t someon e ha s mad e a promis e t o a n ought -
statement requirin g its fulfillment. I n answer t o Searle, Mackie distinguishes
between th e mer e reporting o f facts about a n institutio n suc h a s promising ,
or describing it from the outside, an d on the other hand, speaking from within
it, thus in effect endorsin g it, as in Searle's conclusion. The distinction is pre-
sented a s independent of any belie f i n objective prescriptivity, so it amount s
to a way of understanding the prescriptive force of "ought" without attribut -
ing any queer combinations of properties t o objects. Instead, the motivatio n
to act on an ought-statement apparentl y comes fro m involvemen t in an insti-
tutionsomething it is logically possible t o opt ou t of , thereby blocking the
inference fro m "is " t o "ought. "
On Mackie's later extensio n o f the notio n o f an institutio n to an y kind of
group practice, these remarks apply beyond promising to morality in general
and we might also say, to the use of terms like "ought" in moral teaching. Mackie
himself presumably thinks of an amoralist a s someone who reject s the institu -
tion of morality altogether, though h e is sufficiently awar e of it in reference to
other people' s behavio r to b e able to describ e i t from th e outside . But on my
general version of internalism, there is room for a motivational amoralist con -
ceived as someone wit h one foot in and one foot out of the moral institution t o
the extent that he uses moral language with a meaning set up by childhood mora l
teaching. The latte r o n my account amount s to th e practice of loading mora l
terms with emotional sanctions. The sanctions themselves can be dismissed as
"kid stuff" i n adult life without abandoning some or all of the other habits they
Motivational Foundations o f Conflict 8 9

were used to teachat a minimum the linguistic habits that support th e mean-
ingful us e of moral terms in the case of the amoralist .
In his account of "ought" in terms of reasons, Mackie applies the notion of
an institutio n t o th e practice of taking other people' s interests a s reasons for
action, which he represents as "an establishe d wa y of thinking, a moral tradi-
tion" that makes certain demands of an agent. 49 Institution s create moral rea -
sons, then , an d thoug h Mackie' s discussio n make s i t clear tha t the y ar e no t
therefore "artificial " creations in all cases, hi s later accoun t o f Hume i n con-
nection with the Protagorean myt h indicates that they sometimes come under
Hume's "artificial virtue" of justice. They are to some extent products o f social
convention, that is . But the flourishing o f a social group that he mentions here
briefly a s a kind of overarching group interesth e apparently equates it with
survivalwould seem to count as a further sourc e of reasons promoted by the
link between dike and aidos, between rules favoring cooperation an d the moral
sentiments that motivate action on them. To rejec t the link is essentially to opt
out o f human society by refusing t o shar e its aims.
Or so one might add on Mackie's behalf; he unfortunately fails to connect his
treatment o f "ought" and institution s with his later Protagorean accoun t o f the
function o f morality. I shall have more to say in the next section abou t ho w th e
notion of group flourishing might be made to yield a realist account of the sort of
binding force he attributes to "ought." For the moment, the thing to note is that
the resulting view does not involve ascribing any special motivational properties
to situations in the world around us. Both dike an d aidos can be described from
without simpl y by describing natural and artificia l (sociall y created) facts abou t
the world, in a sense that covers human behavior and responsesincluding the
kinds of nonmoral evaluative facts that Mackie would accept as objectively based,
such as facts about the "best" means to one's ends. "The facts " on a Protagorea n
version of realism, then, will be the same as those i n Mackie's subjectivis m but
with a connection to moral emotion explaining their prescriptive force as some-
thing that pertains to statements made within the social institution thus described.
The resultin g view, social artifac t realism , allows for a n understandin g of
morality as real even though inventedthat is, for Mackie's ow n understand -
ing of it as essentially man-made to fulfill a certain purpose. For artifacts surely
deserve a place in the fabri c o f the worldo r eve n as part o f its "furniture, "
like the tables and chairs we point to i n classroom discussion s of the reality of
physical objects . Morality o r th e mora l cod e i s real on thi s account i n some -
thing like the way that an artifact is: It is dependent on minds for its existence
and purpose and therefore at least to some extent for its form; on the other hand,
it is subject-independent in the sense of not bein g malleable at will. It can even
be viewed as imposing 'requirements on minds in a way analogous to the pos -
tural demands made on the body by a certain kind of chair. But we need to ask
whether this is enough to answer the charge of subject-relativity with a notion
of the authority of ethics of the sort that Mackie faults naturalis t accounts for
not being able to provide. 50 Let us now take a step back from th e view that ha s
emerged from discussio n of Foot, McDowell, and Mackie to see how the result
can b e reshaped to giv e us what we want fro m mora l realism.
90 Between th e Horns

3. Betwee n the Horn s

With it s assumption o f general internalism , socia l artifac t realism promises a
solution t o the problem of action-guiding dilemmas becaus e i t affords a sense
in whic h th e norma l agent' s emotiona l reactions ar e impose d o n hi m b y the
situation o f conflict rather than simply being accidental features of his response.
Its two components i n the Protagorean stor y were already in play in the account
of th e teachin g of moral language via emotion tha t I used at the beginnin g of
this chapter to establish a motivational link between ought-judgments and guilt.
My notion of "didactic import" should now enable us to make out guilt or some
similar moral emotio n a s justified i n cases of dilemma in a sense that is strong
enough t o mee t reasonable versions of the charge of subject-relativity, as well
as being weak enoug h to allo w fo r cases o f abnormal mora l respons e o n th e
order of amoralism .
One way i n which my proposed accoun t i s not subject-relative , we should
now note, amounts to a departure from Foot's version of externalism: An agent's
reasons for action do not depend solely on his own desires and interests. At any
rate, the social artifact view allows us to say that they do not, sinc e it allows us
to appeal to the notion of a group standpoint, with group flourishing as a further
end an d a source o f further interest s beyond those o f individual members bu t
requiring action of them in at least some cases. Morality o n this more thorough -
going version of externalism may "give" m e reasons that I do not accepttha t
I choose t o loo k awa y fro m o r simpl y miss, in the wa y tha t one might fai l t o
take a hint. It may give me reasons to act on others' behalf for the sake of social
harmony, saymuc h as an appeal to group interests might b e thought t o give
me reason to enlarge the gene pool by having children, even if that would in no
way promote my own interests, real or perceived. We can still say that reason s
are motivationally sufficient o n this account in the sense of being able without
supplementation to give rise to the requisite behavior, whil e denying that they
must do so, even on the part o f a rational agent who i s aware o f them and has
no countervailing reasons.51 Instead, they are available to motivate behavior if
the agent attends to them in emotional or other practical termsif he sees things
from th e standpoint o f the whole.
This yields a version of McDowell's vie w of moral insigh t but withou t th e
interpretation o f motivational sufficiency tha t keeps hi s view from accommo -
dating dilemmas. On my account, dilemmas arise because the code of rules best
fitted i n general terms to th e socia l an d individua l aims o f moralit y i s simply
not adequate to all possible cases; but it is not therefore superseded, at any rate
within the mora l poin t o f view, when it yields a conflict. Though man-made ,
the rules of morality are not lik e rules of a game that can b e tinkered with t o
correct any defects, or abandoned alon g with th e game i f they turn out not t o
be perfectible. They sometimes resul t in deadlock, wit h emotiona l response a
substitute for action on them, as provided by their general motivational under-
pinnings. The view drawn from Macki e makes ethics subject-dependent to the
extent that i t does interpre t motivational force in terms o f a lin k t o emotion .
We now nee d to as k bot h how th e vie w can b e seen as a for m o f realism and
Motivational Foundations o f Conflict 9 1

whether it can stand u p to the complaint tha t a subject-dependent view is too

easy on the agent in a dilemma.

Constraints on Moral Codes

What w e mainl y wan t fro m realism , I tak e it , i s a reasonabl e basi s fo r th e
authority o f ethics . Th e poin t o f distinguishing the tw o component s o f th e
Protagorean accoun t i s to separate out such a metaethical basis from th e sub-
ject-dependent elements of morality. I take them to correspond roughly to the
stages of developing moral consciousness produced b y Aristotelian habituation
in early moral life, as a child is made to follow rules in rote fashion, on the basis
of which, in conjunction with reward an d punishment , he learns moral emo -
tion tendenciesthe latter in turn yielding moral motivation at a more advanced
stage.52 More fundamentally, though, dike an d aidos represen t no t tempora l
stages but components o f my proposed explanation of moral meaning in terms
of the institution of a moral code: the social creation o f the code itsel f and th e
provision of psychological backing for it by connecting it to emotions. Though
they are distinguishable for explanatory purposes , however, i t is important t o
see how the two component s interact . The content o f the mora l code will be
influenced b y a requiremen t of teachability, mos t notably , tha t bring s in th e
general facts of individual psychology. So the order of developmental stages in
individual psychology may be reversed on the social level or in relation to ques-
tions of explanatory priority .
On th e othe r hand , th e content o f the moral cod e shoul d no t depen d o n
individual psychological variations; an agent cannot jus t opt ou t o f the moral
rules by appeal to a deviant or deficient conscience or mora l sensibility . This
gives i n a nutshel l the sens e i n whic h ethics ma y stil l b e sai d t o b e subject-
independent on the Protagorean account , as I shall continue to think of it despite
differences fro m historical Protagoreanism, difference s exploited an d expanded
in my ensuing discussion.
We still need to ask whether the view is undermined by the social variability
it presumably allows for. What is to keep it from applying to just any socially
accepted code of rules that might be used in moral teaching? Since the result of
moral teaching via emotion would seem to be a sense in which socially induced
emotion gives us access to moral truthshaving been taught to us in conjunc-
tion with them and indeed in conjunction with the use of moral languagethis
amounts to a social version of the charge of subject-relativity. Is the Protagorean
view viciously subject-dependent, after all ?
Interestingly, the same sorts of social facts may be used to support another
objection pulling in the opposite direction. The extent to which individual con-
science depends on social indoctrination might seem to undermine the autonomy
we expect of a moral agent. This amounts to the independence of the mind from
external forces, including social forces. The objection may be compounded by
a commo n philosophica l vie w of emotions themselve s as alie n psychological
forcesas standin g outside the min d or self , adventitiou s additions to i t im-
planted by external objects of desire or b y social training or suggestion .
92 Between the Horns

Both of these objections, from subject-relativit y and fro m autonomy, rais e

questions about the content an d interaction of the two componentsquestion s
about th e role playe d b y moral emotions , alon g with mor e genera l question s
about the presuppositions o f the moral code. At the outset, though, we should
note that it is not so obvious that emotions are in any worse position than belief s
as bases of conscience. Mackie' s interpretatio n o f aidos i n terms of emotion s
presumably has to d o with thei r genera l usefulness a s a motivational mecha -
nism; on m y own account , emotions lik e shame and guil t effectively punis h a
failure t o act with discomfort . Bu t this is part o f what seem s to pu t the m out -
side the agent's control: They are typically less malleable than beliefs in response
to either individual deliberation or the particular feature s of the agent's situa-
tion. For much the same reason, however, emotions may in some way s be bet-
ter insulated than beliefs against the kind of social manipulation that is now in
question. The y may als o hav e a biologica l basis i n innat e responses tha t set s
more limits on early childhood training.
On the other hand, even my realist version of the Protagorean view is meant
to allow fo r some degre e of relativity to social convention as the source of the
basic social arrangements presupposed b y justice. I do not want to insist that it
restricts u s to a unique outcome in the choic e o f a moral code ; t o coun t a s a
version of realism, however , i t has to constrai n th e choice sufficiently t o pre -
serve a sense of the authority of ethics.53 Le t us ask, then, whether we can fin d
something in the two components o f the view that might serve to keep the choice
of codes within reasonable boundsthat might impose limits, for instance, on
the fairness of an acceptable mora l code. I shall consider the questio n i n illus-
trative form by asking what might prevent a society from allowin g or encour -
aging indifference t o the welfare of a certain subclass of its members. Why no t
have an "outcaste" group, in short?
A full treatment of issues of justice is of course beyond the scope of my argu-
ment here . What I want t o argu e i s that a realist version o f the Protagorea n
view allows for the usual sorts of moves on the subjectplus one that is some-
what differentwithi n a broadly "communitarian " framework . Thes e move s
have well-known limitations, bu t my aim here is just to indicate how a socially
based view can avoid any special problems associated with communitarianism .
Though derive d from Mackie, m y own suggested version of the view rests on a
use of the notion of group flourishing as a kind of social parallel to Aristotelian
happiness, providin g a standard fo r the correction o f moral codes. Followin g
Aristotle on individual flourishing, I interpret the notion in a way that allows it
some moral or other evaluative contentso that it is not a strict welfare notion
but also a notion of group excellence or virtue. This may require departing some -
what fro m Mackie' s metaphysicall y auster e presuppositions , bu t I think th e
deviation will turn out to be minimaland largely defensible in terms of moves
that Mackie seems ready enough t o allow .
Mackie himself , despite hi s ow n us e o f relativit y as a n argumen t agains t
objective values, attempts to handl e fairness issue s by modifying th e principle
of universalizability. 54 The firs t thin g we ought to not e is that there is nothing
about a realist version of the Protagorean vie w to keep us from simpl y accept-
Motivational Foundations o f Conflict 9 3

ing some suc h principle as a basic constraint o n moral codes. I t is specifically

moral propertiesthe requirements of particular situations as in McDowell' s
version o f realism ; what Macki e think s o f a s "intrinsi c requirements"tha t
apparently would violate a Mackiean ban on "objective prescriptivity." But what
about rational constraints , suc h as those se t by facts about th e bes t mean s t o
group flourishing on Mackie's ow n view ?
Even if universalizability or some other principle of fairness cannot be com-
pletely justified i n instrumental terms, we might take it as a kind of constitutive
constraint on groups we would think of as flourishing. Our preference for groups
that satisf y th e constraint migh t be compared t o ou r preferenc e for simplicity
in the comparison o f rival scientific theories.55 We might think of universaliz-
ability, i n fact, as a moral versio n o f simplicity to th e exten t tha t i t involves
treating simila r cases similarly. To tak e this line i s to thin k o f fairness in th e
first instanc e as a featur e w e valu e in groups, whateve r it s consequences fo r
"prescriptivity," a s a source of principles of action. O r better , perhaps: I t is a
feature w e value in moral codes , allowin g for their nonarbitrary extension t o
new cases.
With fairness taken as a property of codes, universalizability also comes ou t
as promoting some independently defined purposes of morality. Besides accom-
modating unforeseen circumstances, a moral code governed by some such prin-
ciple will provide a way of resolving interpersonal conflicts that is defensible to
all parties.56 Even with fairness applied directly to groups, universalizability can
be seen as promoting socia l harmony and stabilit y to th e extent that i t makes
the group standpoint attractiv e to it s members and hence more likely to serve
as a source of individual motivation.
However, eve n if it served no furthe r purpose , mora l simplicity of the sor t
that rules out arbitrar y distinctions shoul d not b e dismissed as something we
just happen t o value . One migh t ask at thi s point whethe r the appeal t o sim-
plicity in the choice among scientific theories would b e seriously undermined if
we came up with peopl e or cultures with a taste for complexity and epicycles.
I take it we would say that people with deviant preferences of the sort suggested
just do not participat e i n the scientific enterprise; a mode o f explanation that
stresses elegant complexity would be put into some category other than "scien-
tific." To say something similar of "moral" codes would not be to rule out mora l
diversity. For that matter, it would not rule out caste systems, since moral codes
may be conjoined with religious and other nonmoral requirements. My sugges-
tion is rather tha t fairnes s may be taken a s a norm o f social rationality an d t o
that extentpossibly along with competing normsas governing the construc-
tion of moral codes .
Morality on this account involves assessment i n light of certain basic ratio-
nal value s governing th e ai m o f promotin g grou p flourishin g such a s social
harmony and stability . If one thinks of the notio n o f flourishing as applied t o
plant life, it seems plausible to suppose that it builds in certain basic evaluative
constraints such as symmetry or balance, other things being equal: A plant whose
lower leave s all fall off , say, might be thought of as a health y specimen if it is
the standard specimen of its kind, but presumably not i f others commonly flour-
94 Between th e Horns

ish i n a mor e robus t sense . O r conside r preference s in th e huma n form : The

bizarre things that other cultures and our own have done in the interests of physi-
cal attractiveness or to mark differences in status need not make us hesitate about
the ver y basic value assumptions implici t in the idea l of a healthy body. Simi-
larly, our notio n o f a "viable " mora l code, meanin g one that adequately pro-
motes grou p flourishing , may rely on certain basic constraints a s presuppose d
by social rationality without thereby imprinting our ful l se t of moral values on
the world o r imposing them on other cultures .
Constraints o f ordinar y instrumenta l rationality o f th e sor t wit h whic h I
beganour preferenc e for the most effectiv e wa y of promoting a n end (rather
than th e subtlest , say)shoul d themselves be viewed as evaluative. Taking a
straight lin e as the "best " way of connecting points require s deciding agains t
various alternative values that favor indirectness. But we can also move beyond
the instrumenta l model i n the way I indicated, by including moral constraint s
on flourishing: not further constituents of the social end, to be promoted a s part
of it, but minimal standards i t has to meet i n order t o be worth promoting .
Despite its general teleological structure, then, the social artifact view should
be able to accommodate nonconsequentialis t theories of normative ethics. It is
not necessaril y utilitarianany more tha n Aristotle' s vie w is egoist. A s with
Aristotelian happiness, the relevant constraints on the moral end may be thought
of as built into the notion of flourishing rather than as imposed o n it from with-
out. Thus, a condition o n group flourishing as the defining purpose o f a moral
code migh t be held to b e due regard for social subgroupss o that a caste sys-
tem could be rejected on moral grounds as "unbalanced," lik e otherwise healthy
physical growth tha t i s stunted i n one area .
The particular lines of argument I have brought in so far to handle fairness,
from universalizabilit y and from social symmetry or balance, still might allow
for a moral justification of caste systems as ways of assigning different sort s of
people differen t function s in a well-arranged social order. Bu t the justification
will be made no easier by the social basis I have assigned to ethics . As I intend
the notio n o f group flourishing , it does not refe r simpl y to the flourishing of a
group in the sense of a collective entity distinguishable from its individual mem-
bers; i t also and primaril y entails the flourishing of group members , o f indivi-
duals in a group. O n an y account I would fin d acceptable , thi s i s what socia l
groups ar e for , thoug h onc e se t up the y may tak e o n purpose s o f their own ,
sometimes i n conflict wit h individua l interests. M y vie w count s a s "social "
because i t takes socia l norm s a s basi c elements of morality; i t does not tak e
notions o f social valu e as prior t o individual interests .
Division accordin g t o functio n would thu s requir e a defense addressed t o
group members in an attemp t t o satisf y reasonabl e demand s o f a scheme de-
signed t o promot e thei r good . An argument o n these normativ e issue s might
assume ignoranc e o f an individual's place i n the social order, of the sort asso -
ciated with Rawlsian contractarianism; o r it might consider features of the social
order as "collective goods" of value to group members on the whole, for a teleo-
logical defens e of basi c social values. 57 The usua l strategies o f bot h dut y an d
virtue ethic s wil l b e availabl e as resources, sinc e m y vie w essentially supple-
Motivational Foundations of Conflict 95

ments agent-base d virtu e ethics with a socia l notio n o f virtue, as a sourc e of

rules of duty .
I have more to say in defense of social artifact realism in chapter 6. Though
I want i t to accommodat e differen t approache s t o normativ e issues , th e view
has genera l implication s for bot h th e structure an d th e conten t o f normativ e
ethics, some o f which will come ou t i n what follows . Indeed , in chapter 6 we
shall begin to have a glimpse of some ways in which the view might even affec t
practical issues. Its main implication for metaethics (narrowl y construed) is that
an argument in terms of group flourishing relocates th e area of potential moral
disagreement to the level of basic values. It is here that one might have to appea l
to one's own moral perception as in McDowell's versio n of realism. But we can
at an y rat e limi t such appea l t o abstrac t discussio n o f th e element s o f mora l
code construction an d othe r higher level social normsboth norm s o f instru-
mental rationality and constitutive constraints on the moral enterpriseas dis-
tinct fro m specifi c requirement s of action. This has th e advantag e o f making
the fac t o f moral disagreemen t more intelligibl e and o f avoiding the applica -
tion o f special motivationa l properties t o situations i n the world aroun d us
the sort of thing that Macki e find s "queer. " Instead , ou r fundamenta l moral
disagreements concer n personal an d socia l ends: envisioned states o f self an d
society, whose properties ar e quite reasonably seen as motivational .
This is not t o say that their motivational force is irresistible, even assuming
rationality. In fact, there is a further kind of instrumental argument besides those
commonly noted tha t ough t to be brought in at this point a s a way of limiting
the appeal to moral perception. The project of constructing a viable moral code ,
that is, will be subject to material limitations, as well as to forma l constraint s
like simplicity, to the extent that viability depends on a connection to individual
motivation. T o strengthe n th e cas e fo r basi c mora l value s such a s fairness , I
suggest tha t w e supplement the perceptual mode l o f realism wit h som e refer-
ence to the emotional mechanism that underlie s moral teaching .

The Role of Moral Emotion

The main point t o not e i s that mora l emotion s see m to b e formed from othe r
responses o n th e basi s of natural identificatory processes tha t are not directly
sensitive t o socia l o r sociall y marked distinction s amon g peopleo r eve n t o
differences i n natural endowment of the sort that might be used to support socia l
stratification accordin g t o function . Instead, on th e sor t o f picture I relied on
earlier in my speculative account of the teaching of moral "ought " in conjunc-
tion wit h guilt , our inculcatio n of moral emotion s presuppose s sensitivit y to
personal and behaviora l similarities and relationships such as contact an d per-
sonal dependency .
Early on , infant s exhibi t a crying response t o th e soun d o f another infan t
crying.58 On e ca n se e how som e genera l tendency o f th e sort , t o imitat e the
emotional behavior of other species members, might be useful i n an animal herd
as a way of communicating a quick response to a threat t o the group perceived
by one o f its members. In humans the empathetic response serves as the foun -
96 Between the Horns

dation fo r a way o f communicating different sort s o f emotionne w feeling s

built on a n infant' s original stoc k o f emotional responses, a s associated wit h
the behavio r it imitates. A kind of "imprinting " o f emotiona l behaviorand
with i t emotions themselvesfrom the mother o r other objec t of dependency
reinforces the infant's natural responses and regroups them into new ones. The
infant's natura l sympath y in this initiall y behavioral sense i s shaped int o th e
adult tendency to tak e on others ' emotion s partl y by seeing others imitate its
own feelings, with imitation cultivated as a kind of imaginative play. The result
is a quick way of communicating more complex feelings, including specifically
moral feeling s (fo r instance, guilt), b y initiatin g the gam e i n situation s tha t
prompt th e child to entertain certain thoughts .
On this rough accoun t o f emotional learnin g as founded on imitation and
imagination, identification is not itself an emotional response among others fro m
a certain putative range of sensibility but a mode of communicating responses
from preexistin g ranges and constructing new responses o n the basi s of them.
The important questio n in connection with moral teaching, then, is not on e of
natural convergence of emotional response on the model of perceptionor even
of "corrected " respons e o n th e mode l o f perceptual judgmentbut rather of
its feasible allocatio n to object s in support o f a moral code.
Behavioral similarity would seem to be enough to set off the mechanism, so
the attemp t t o limi t it to member s of a social subgroupo r t o chec k i t with
some opposing tendency, even one that i s arguably natural such as fear o f the
strangermight b e taken as yielding an unstabl e combination. I t i s not a t all
an impossible combination, but its two elements are apparently in tension, and
the more inclusive element would seem to be the one that determines the scope
of emotional interaction.59 A rigid stress on the exclusive element thus may be
challenged a s impeding the construction of a viable moral code.
My claim about caste systems is not that they are ruled out b y human emo-
tional nature but jus t that a basis in social exclusion pulls against our natural
response tendencies in a way that tends to undermine moral code construction.
Agents would risk acting at cross-purposes in not taking adequate moral account
of the sufferings o f people they understood to be human. At the very least, then,
human emotional nature affects th e social and psychological costs imposed by
alternative moral codes.
This approach t o mora l psychology might be thought o f as "bottom-up "
since it assigns emotion a role in determining the shape of a viable moral code
rather than simply tacking it on to a predetermined code a s a source of moral
motivation. My account therefore differs fro m other motivational accounts, most
notably Mill' s utilitaria n appeal to "interna l sanctions, " a s well a s from th e
perceptual o r aestheti c mode l that w e find , most notably , in Hume. 60 Mora l
emotion as I conceive it does not just register in affect a n independently grounded
(or an ungrounded) moral code but is part of the system of penalties and payoff s
that makes a code operative in the first place. It plays a special role even among
features of individual psychology insofar as occurrent emotion provides the sort
of controllabl e episodic basis that allow s for th e teachin g of moral language.
On m y account, negative moral emotions such as guilt may b e thought of
Motivational Foundations of Conflict 97

as felt needs registering in affect ou r morally "binding" reason s for action. Bu t

they are in many cases artificial or manufactured needs set up to reflect the group
standpoint in individual response tendencies . Their purpose, in short, is to har -
ness egoisti c energie s t o mora l ends . S o their nature a s the availabl e suppor t
mechanism for a moral code also imposes limits on the nature of a viable code:
The code must be such as to allow for support fro m a manageable extension o f
our natura l stoc k o f emotions .
Is it odd to think of this view as realist? The term may seem to presuppos e
a perceptual model, but a moral code is of course not something "ou t there " t o
be seen. Besides being instituted by minds, it now appears t o b e constrained by
their response tendencies. In any case, on reasonable versions of the view I have
offered, th e standard appealed t o may not be the actual code in force but rathe r
one that is corrected i n certain ways as needed to promote grou p flourishing. 61
So it is an idealized artifact of human choice and emotion. Bu t I take it that the
hypothetical fact that the actual code would promote grou p flourishing if cor-
rected in certain ways is "real" enough for our purposes an d is connected clearly
enough to human harms and benefits to count as a "moral" fact.62 I shall allow
the term "realism, " then , for views that answer the charge of subject-relativity
by appeal t o suc h facts, see n a s governed b y constraints o f the sor t just indi-
The term has a number of misleading connotations, bu t i t seems t o b e the
term in use as a single expression summing up the alternative to both metaethical
noncognitivism or nondescriptivism and Mackie's "erro r theory," two ways of
denying that moral judgments state facts: the former by denying that the y pur-
port to b e factual; the latter b y denying their truth. M y two-component vie w
essentially serves to separate th e factual content o f morality from the question
of its force over behavior. It is not put forth as realism in the metaphysical sense
that would make out moral judgments as ascribing special properties to objects. 63
Rather, it s conception o f moral fact s a s hypothetical about unrealize d states
of affairs on the order of group flourishin g suggests a way of reinterpreting the
notion of "moral observation " tha t is often assumed, by analogy to perceptual
judgment, to amoun t t o acquaintanc e wit h particula r moral fact s of the sor t
given by "This is wrong."64 Moral error and disagreement will be easier to under-
standand for that matter, a certain degree of skepticism will be reasonable
if we relocate the observation leve l in ethics (i n contrast t o science ) to tha t of
abstract imagination.
Though morality on this two-component versio n of realism may depend o n
the actual moral code as something set up and held in place by minds, its content
will no t b e determined b y the desire s o r decision s o f any particula r agen t o r
even those ope n to the group as a whole at a particular historical moment. The
notion o f a constitution a s a written cod e of rules that constrains furthe r legal
decision bu t i s itself subject to correction i n light of its general social purpose s
provides a helpful legal analogy for understanding the authority of a moral code ,
as a product o f human will and inventio n that nonetheles s constrains human
choice. To gran t it authority in that sense , however, i s not t o suppos e tha t it
somehow has the power t o compel obedience, even among agents assumed t o
98 Between the Horns

be rational; it constrains choice only within the relevant framework. For morality
(as opposed t o a constitution), the relevant framework is not simpl y the par -
ticular code or system of norms in force but also something more general back-
ing it up: the standpoint of group flourishing. It also involves something more
particular that can be harnessed to the social code in psychological terms: guilt
or som e similar motivational mechanism involving moral emotion. A n indi-
vidual ma y have to pa y a price for abandoning the grou p standpoint , i n the
manner of an athlet e who risk s her health in pursuit of more specialized self-
I shoul d not e tha t th e socia l artifac t view is not pu t fort h a s a thorough -
going analysi s of moral term s o r judgments . I t i s meant t o captur e onl y th e
descriptive content o f ethics, at this point considered as a whole: the social prac -
tices, behaviora l and emotional , b y virtue of which a moral code may be said
to be "in force " a s a way of promoting th e characteristic end of life i n groups,
along with corrections and extension s o f the code in light of that general end.
This amount s to th e backing for our mora l judgments , but i t leaves out thei r
prescriptive force as something that depends on adopting the normative stand-
point tha t i t tries onl y to understan d from outside . An interpretation of wha t
we say from the inside would of course have to do more, and the view is meant
to be compatible with different way s of attempting t o do more. It provides only
a roug h accoun t o f morality as a socia l institution , plus an indicatio n of th e
role of emotion a s its link to individua l motivation.
Norms of emotional response are fundamental to morality, then, insofar as
they regulat e the interna l sanctions and som e o f the chie f externa l sanction s
that actually keep the moral cod e i n force. I tried to illustrat e with my earlier
remarks on fairness ho w they play a role in determining the nature of a viable
code. They are essential to it in general terms because they build the group stand -
point into individual psychology, providing the individual with a way of access-
ing it from hi s own immediat e psychological reactions . Thi s yields a sense in
which moral emotions let us "perceive" mora l facts, but it is an indirect sense,
parasitic on their primary role as motivators .
As the basis of moral teaching, norms of emotional response set up a particu-
larly exacting standard o f individual moral sensitivity. However, the standar d
in question need not b e seen as an ideal of individual virtue on the Aristotelian
model. It would make amotivational use s of moral terms count as in some way
deficient thoug h no t therefor e deviant i n meaning, and no t i n all cases a ba d
thing. There are times calling for actioneven moral actionwhen the demands
of perfect moral sensitivity would ge t in the way; political "dirty hands" case s
provide important examples. 65 Bu t there is another point to be made here: Per-
fect virtue in fact might be said to require the ability to free oneself in moments
of adult reflection from th e emotional baggag e o f childhood mora l instructio n
in order t o take a critical look a t the accepted mora l code.
This point rests on some assumptions worth noting. It places a value on the
capacity for moral growth, o n the assumptio n that a perfect mora l education
of the sort Aristotle describes, say, may or may not yield perfect moral sensitiv-
ity, dependin g o n th e surroundin g society' s codea s witness , o f course ,
Motivational Foundations of Conflict 99

Aristotle's view s on women and his acceptance of slavery. A "blind spot" may
be held in place by emotional distractions, including those set up by the moral
code. In fact, as I argue in the next chapter, Aristotle' s treatmen t o f shame in
connection wit h th e virtue s suggests that hi s view makes insufficien t accom -
modation fo r personal moral imperfection as something an otherwise virtuous
person might be called upon to change. When the moral code itself is in need of
improvement, though, its underlying system of emotional response may consti-
tute an inertia l force agains t reasonable change.
A side effect o f allowing for detachmen t from mora l emotio n i s the possi -
bility of amoralism: the rationally intelligible use of moral language without its
usual motivationa l props . Anothe r sid e effec t i s the possibilit y o f a kin d of
emotional self-criticism that can be used to reconcile emotion-based ethic s with
the valu e placed o n autonomy . On e ca n ris e above one's personal syste m of
emotional response , a s well as the system set up b y the moral code, even sup-
posing tha t it constitutes an integral part of oneself. Indeed, the point i s to ask
whether it constitutes the self one wants to be. Thus, we might discourage cer-
tain emotions in ourselves, perhaps on moral groundsand perhaps including
the very emotions that allowe d us as children to perceiv e moral grounds : fea r
of social disapproval, say, or an eagerness to be accepted. Emotiona l self-criti-
cism may itself be seen as emotion-basedas motivated, for instance, by admi-
ration o r disdain for one's current self, as something partly constituted b y cer-
tain emotions.66 However, my suggested account of the emotional basis of ethics
is not mean t to yield a monolithic view of its foundations but to accor d emo -
tions a serious place amon g them, despite th e notoriou s pitfall s of emotiona l
motivation. At least in some cases, emotional self-criticism conceivably involves
rising abov e emotion. I do want t o deny , though, tha t thi s mean s risin g to a
level where the usua l emotional support fo r ethic s is annulled.

The Intractability of Wrong

One might think that a corrected versio n of the actua l moral cod e o f the sort
that is now in question could not contain conflicts. In defense of universalizability
or some simila r principle, I appealed to simplicity considerations a s favoring a
code that can be extended readily to new cases. But surely, it might be said, the
need for "i n principle" decision s uniting different case s counts agains t a cod e
that allows for irresolvable conflicts of the sort at issue in dilemmas. In the face
of conflict, what we ought to do is to detach ourselves from the emotional reac -
tions that support our ordinary set of moral rules and bring in some alternative
principle that yields an acceptable resolution .
This way.of ruling out dilemmas is associated with Hare's account of moral
thinking as taking place o n two levels : the "intuitive " level , consisting i n our
everyday stock of emotion-based rules , and th e "critical " level, a higher court
of appeal fo r deciding what rules to put o n the lower level and how to resolve
cases in which the rules turn out to b e inadequate.67 On Hare's view what we
appeal to is a single principle that he takes to be derivable from universalizability :
the principle of utility. But Hare's approach t o normative ethics need not con-
100 Between th e Horns

cern u s here; we may grant for purposes of argument tha t th e practical resolu-
tion to a moral dilemma should involve "acting for the best"as Williams puts
it in considering th e case of Agamemnon 68and that thi s mean s maximizing
utility. Th e questio n raise d b y dilemma s is whether th e mora l rul e tha t w e
thereby ac t agains t mus t b e seen as superseded. What I need t o argu e her e is
just that we may still take it to be in force compatibl y wit h a coherent vie w of
ethics, on assumption s lik e those jus t defende d as yielding a version of moral
The question of realism initially came up in negative form in connection with
dilemmas: Williams's accoun t of them was presented in part a s a way of under-
mining realism. Bu t if I am right , the view turns out t o pla y a positive rol e in
defense of dilemmas. Realist assumptions allow us to put a limit on the respon -
siveness of our intuitiv e principles to considerations o f simplicity: to distinguish
the sorts o f corrections t o the actual code that ar e needed t o promote its gen-
eral practical purpose s fro m other s that might be introduced solel y for the sake
of various theoretical end s such as systematic neatness.
In contrast t o Hare's prescriptivism, social artifac t realis m does no t make
out mora l principle s as simply chosen by us within certain logica l constraints .
Our correction s t o th e mora l cod e tha t i s socially in force are mean t t o le t it
capture somethin g subject-independent , even i f not unique : a versio n o f th e
actual code that would do an adequate job in promoting th e end or complex of
ends summed u p as "group flourishing." On this account, a viable moral cod e
is constrained b y the nature of its defining social end as well as by the means to
it, including most notabl y its motivational basis in emotional learning. By con-
trast, Hare claims that reasoning at the critical level permits n o appeal t o sub-
stantial moral intuition s but only to logical intuitions such as universalizability.
With regard t o dilemmas, I think there is reason t o insist against Hare that
the emotiona l underpinning s of a viable moral cod e would i n fac t b e under-
mined by canceling the application of its principles to cases of conflict. In a word,
it is important t o the general purposes of the code that a morally sensitive agent
both registe r emotionall y the principle he has to act against an d be seen as re-
sponding thereby to a practical requirement imposed by the situation. This point
rests o n a n accoun t o f moral educatio n outline d previousl y that i s not unlike
Hare's, except tha t Hare apparentl y relies solely on considerations o f personal
virtue to mak e roo m fo r our intuitiv e emotional responses , a s needed fo r his
account t o cove r response s tha t ma y not b e justified i n utilitarian terms. I n a
case of conflict, Hare's account allow s us to say that a person wh o experiences
no guil t o r remors e fo r violatin g the overridde n intuitiv e ough t woul d b e a
morally worse personor a morally worse educated person, as he also puts it 69
whom w e reasonably think less well of for the lack. But Hare's account appar -
ently does not le t us say that one morally ought to feel guilt or remorse i n such
a case a s a substitute fo r action, o r to satisf y a "contrary-to-duty" obligation ,
perhaps eve n in contexts wher e the emotion would d o more har m than good .
Nor doe s it let us think of the emotion a s rationally appropriate.
Of course, consideration s of virtue and the Aristotelian picture of virtue as
based o n genera l habits of emotional response also underlie the notio n o f the
Motivational Foundations o f Conflict 10 1

ideally sensitive agent that my own discussion took from McDowell . I attempt
to sho w ho w guil t fits int o thi s picture a t the beginnin g of chapter 4 , an d my
own account wil l diverge further fro m both Aristotle's and McDowell's. It has
already begun to pul l away fro m the m by setting up a special standard, mor e
exacting tha n normal , for mora l teaching . A n importan t poin t o f agreemen t
with them, however, is the view of an ideall y sensitive agent a s responding t o
something realt o morall y significan t feature s of th e situatio n rathe r tha n
merely to feature s tha t ar e i n general morall y significant. It i s Hare's willing-
ness to drop this point, I want to claim, that keeps him from making adequate
sense of dilemmas.
Consider on e of Hare's illustrations of the role he assigns to guil t in moral
thinking: On a visit to Czechoslovakia at the height of the cold war, he says, he
would have lied to avoid being expelled if he had bee n asked by officials abou t
the purpos e of his visit; despite the belie f tha t he ought t o lie , he would have
felt guilty. Hare handles the case by allowing for conflict on what he elsewhere
distinguishes as the intuitiv e level of moral thinking; he here refers to " a sens e
of 'thinkin g that I ought' in which . . . [f]eeling guilty is inseparable fro m . . .
thinking that I ought not. " Bu t on the critica l level, there i s no mora l dut y t o
back up this conflicting thought; i t is justified simpl y as something expecte d of
a morally good agent: " I should be a morally worse man i f I were not affecte d
in this way."70 Note that this justification is not utilitarian: Though ther e might
be generally beneficial consequence s of emotional sensitivit y on mora l issues,
those agent s who woul d b e able to forg o it in this case without effec t o n their
general tendencies would be well advised to d o so.
Now contras t th e cas e wit h tha t o f Agamemnon . Agamemno n i s indeed
considered morally deficient fo r failing to b e affected wit h guilt or som e simi-
lar emotion b y the sacrific e of his daughter. Bu t to say only this i s to trea t th e
case as fundamentally indistinguishable from one that involves an act that would
be wrong under other circumstance s but is not wrong under those tha t obtain ,
as with Hare's imagine d lie to a n officia l o f a n unjus t government . Th e ideal
response o n Hare's account i s treated a s appropriate no t t o th e situatio n bu t
rather for a personas i n this case failing t o fi t the situation i n a way that is
characteristic of someone with a good moral upbringing, given the fact that emo-
tions rest on general habits of response .
Hare is willing to apply his position to some morally serious cases, but they
all involv e slighting some members of a group in order t o fulfil l a more basic
responsibility to the group as a whole.71 These are what Williams calls cases of
"moral cost, " withou t th e sor t o f unanswere d clai m tha t i s a t issu e i n
Agamemnon's sacrifice o f his daughter. In all the cases Hare deals with, then,
the ought that i s not acte d o n is on his account clearl y overridden. His reason
for expecting a morally good person to react to it has to do with the limitations
of human psychologythe need to construct our everyday moral sensibility from
the limite d materials provided by general emotion tendenciesconsidere d as
distinct from th e morall y significant feature s of the situation.
My ow n approac h t o morall y serious cases appeal s to a n idea l of moral
sensitivity no t jus t a s the produc t of the righ t sort o f upbringing but als o as a
102 Between the Horns

standard o f correct response : the more exactin g didacti c standard tha t under-
lies moral teaching . Thi s ma y b e thought o f essentiall y as a standar d o f du e
attention t o what on e is talking about, a s registered i n emotion. I do not hol d
that an agent has to conform t o it insofar as he holds moral judgments but that
otherwise his responses ar e i n a certain sense deficient. They d o no t coun t a s
full response s sinc e they fail t o expres s th e motivation to actthoug h i n nor-
mal cases it is enough that the agent does act, as he may do automatically. In a
case like Agamemnon's, however, th e emotional components o f a full respons e
will be morally required as a substitute for action o n one of the ought s i n con-
flict. This was my suggested modificatio n of McDowell's cognitivis m i n order
to allow for dilemmas.
The point of the further departures from McDowell's vie w that I took from
Mackie but defended as compatible with the aims of realism is just that we can
still insist along with McDowell tha t Agamemnon's requisite emotional response
is one that is merited by features of his choice-situation, a s opposed to simply
being meritorious on his part.72 Guilt or remorse i s justified o r appropriate i n
Agamemnon's circumstances, no t just understandable or even admirable as the
result o f a goo d mora l upbringing . However, I d o no t wan t t o sa y tha t th e
emotion amount s t o a perceptio n o f som e specia l motivationa l property o f
the situation; i t is enough that i t be backed up b y subject-independent facts in
the relevant sense. This sense allows for facts about a code of rules set up an d
sustained by minds for subject-dependent purposes. What it excludes is relativ-
ity to minds as knowersor a s the source o f the prescriptions that substitut e
for knowledg e on Hare's noncognitivist account .
In short, the moral code is not simpl y stitched together case-by-case , and it
cannot b e tailored t o fi t th e requirement s o f a particula r cas e a s o n Hare' s
account o f dilemmas. Where conflicts occur, we may indeed have to appeal t o
considerations of utility. I count thi s as a moral appeal , bu t it does no t yiel d a
moral resolution of the conflict on the view I have defended: It leaves intact the
intuitive judgments that Hare would dismiss as lower level. Given the interaction
of the two component s o f morality that m y own vie w takes as elements in its
explanation, we can rest this refusal to rise above emotion on the general moral
importance o f registering the standpoin t o f the ought that lose s out .
Indeed, we may even appeal t o utilitarian considerations i n support o f this
departure from Hare's utilitarianism. In part I I I shall use the example of guilt
to defen d a notion o f appropriate emotio n tha t build s in appeal t o consider -
ations o f general adaptiveness. I t is not limite d to a utilitarian reading, a s we
shall see. At this point, however, let me confin e myself in arguing against Har e
to an example of the general utilitarian benefits of emotion i n cases of dilemma.
In a case like Agamemnon's, guil t or remorse as an identificatory response ca n
be said to exhibit the agent's mora l convictions in feelings an d act s of expres -
sion that are superior to statements o f belief to the extent that they are beyond
full voluntary control an d hence harder to simulate. They therefore are able to
reassure us about the agent's general response tendencies more than any claims
he might make about his preference for avoiding the conflict and the act he would
have chosen i n its absence.
Motivational Foundations o f Conflict 10 3

The agent can also be said to subjec t himself to discomfor t b y undergoing

guilt or remorse, to the extent that his emotions are under some voluntary con-
trol. I shall later supplement the picture of emotions a s useful fo r purposes o f
communication with an appreciation o f the motivational and symbolic signifi-
cance of self-subjection to discomfort. For our present purposes, w e can already
see that expressions o f belief of the sort just indicated would also be oddly re-
mote under Agamemnon's circumstances"hollow" o r "wooden" in a way that
contrasts wit h the immediacy of emotion. The agent brings the situation home
to himself and gives others a reason for taking his response seriously insofar as
he manifests some emotion on the order of guilt. The emotion serves to drama-
tize his commitment to ends that hi s action i n this case cannot serve . The fact
that it is unpleasant both fits the negative content of the moral belief it expresses
and counts a s a reason agains t undergoing the emotion i n its absence.
The conclusion I want t o dra w fro m these points i s that a moral code that
yields a dilemma on the level of intuitive moral thinking should not be corrected
in the way Hare suggests, by anulling the application of the rules to that case in
favor o f critical thinking. It is not enoug h to bas e the same emotional require-
ments on considerations of virtue, as Hare has in mind. To see this, let us take
another loo k a t what an agent migh t reasonably think or fee l about his emo-
tional responses i n a situation of conflict of the sort that Hare imagine s on his
trip to cold-war Czechoslovakia . It would b e compatible with the good mora l
education evidenced by the guilt he feels about a lie that is in fact justified sim-
ply to "laug h off" the feeling, dismissing it as a representation of the real moral
requirements of the situation, perhaps eve n with a degree of pride for hi s ad -
vanced moral sensibility.
This is not what we want from Agamemnon, needless to say. Nor coul d he
delude himself about th e situatio n an d stil l b e capable of critica l thinking. I
conclude, then, that th e appeal to simplicit y as a reason for limiting the num -
ber of rules relevant to the case would be undercut by the need to explain what
the agen t i n i t ough t t o feel . I n any case , the sor t o f simplicity I appealed t o
earlier had a practical justification as allowing for a comprehensible method of
extrapolating fro m th e preexisting rules to new cases. What is in question here
is just a n attemp t t o swee p away mora l clutte r in th e interest s of metamora l
neatness. The problem with its application t o dilemmas is that it also seems to
discard our sense of the intractability of moral wrong.
I brought up this notion earlie r in connection with m y claim that a mora l
prohibition is not erased by considerations on the other side, in favor of action.
At that point I had in mind nonmoral or positive considerations; but what is in
question here is something like higher level considerations: th e view of things
from a n impersonal standpoint of judgment, to be distinguished from bot h the
all-things-considered and the group standpoints tha t have come into my argu-
ment. The notion o f intractability will have further, more particula r applica -
tions later, but i n the present context, it can b e used as a general way o f sum-
ming u p m y answe r t o th e proble m I bega n wit h abou t dilemma s an d
motivational force. It is what makes our mora l intuitions resist the kind of sys-
tematization that many philosophers would impose on them, by ensuring that
104 Between the Horns

a sufficiently seriou s moral reason against some act is not blotted out by a stron-
ger reason i n favor of it. At bottom, then, it involves a kind of intractability to
argument that keeps an appeal to critical thought from doin g what Hare wants.
The notion of intractability connects my account o f dilemmas to specifically
moral o r deonti c notions , a s distinct fro m Hare' s accoun t o f their emotiona l
basis in terms of virtue and other authors' attempt s t o explain them a s clashes
of incommensurable values. The claim that an agent must do something wron g
in a case of moral dilemma adds something to the claim that he must forgo some
good. It eve n adds somethin g to th e clai m that eithe r choice h e makes is bad
insofar as the link to emotion ensures that "wrong " is not just a motivationally
inert labelo r perhaps even one with positive force o n the mode l o f the Afri -
can-American slang use of "bad" i n American English or of the tongue-in-cheek
extension to adult behavior of "naughty" an d similar terms for reproving chil-
This i s to sa y that a us e of "wrong " withou t motivationa l forc e is a sub-
standard use , not jus t that the speaker falls short of a standard o f perfect virtue,
though dependin g on the reasons for it the use may be morally substandard an d
the speake r mor e tha n linguisticall y deviant . In application t o dilemmas , this
means that it is not lef t t o an agent's psychological makeu p to tell him how t o
reactwhether to feel "torn" or simply to grant that both alternatives are wrong
while responding t o on e o f them with indifference . T o b e unmoved b y either
alternative is to violate a norm of full response ; sinc e being moved require s at
least undergoing an unpleasant feeling, we have a sense on this account in which
a dilemma is hard on the agent but still possible. The agent in a dilemma is torn
between tw o alternative s that ar e both i n emotional term s ba d fo r him . This
yields a solution t o the problem posed i n chapter 1 .
Or rather , i t yields the outline of a solution. The summary of my results in
this chapte r raise s in roug h for m a se t o f furthe r question s mor e specificall y
about guilt and other moral emotions: about the response mechanism I want to
use t o explai n mora l motivation , it s assessmen t fo r appropriateness , an d it s
relation to judgments of responsibility and wrong. These include questions about
the moral worth and the practical usefulness of guilt as a source of moral moti-
vation bu t als o mor e genera l questions about norm s governin g emotional re-
sponse an d their relation to moral action .
I want to treat guilt in dilemmatic cases essentially as a substitute for action
rather than as a way of registering the perception of moral wrong, o r a mark of
moral virtu e or a good mora l upbringing , at leas t i n the firs t instance . But it
may see m to b e a rather poor substitute for an y number of reasons. On e tha t
might come up here, in light of my criticisms of Hare's account, i s its degree of
self-focus: An agent who i s motivated by guilt might be said to be acting out of
concern for his own moral worthor eve n for his own state of emotional com-
fort or discomfortand hence to be no less morally deficient than one who dis-
misses hi s guilt feelings i n a case of conflict.
Moreover, a n accoun t o f such feelings a s appropriate in cases of dilemma
might seem to compromise the notion of intractability just defended. It would
apparently make out guil t as justified o n general utilitarian or othe r practical
Motivational Foundations o f Conflict 10 5

grounds in at least some cases where the corresponding evaluative beliefthat

the agent is responsible for wrongwould not b e warranted. But if this is true,
it would also seem to detach guilt from a judgment of wrong. Instead, the emo-
tion migh t be see n a s requiring only prima faci e evidenc e of wrongdoing . I n
that case, the claim that guilt is appropriate no matter what th e agent does in a
case of dilemma stops short of giving us what it seemed to jus t above: a view of
the agent as responding to something real in the situation. Le t us now attemp t
to deal with this question by way of an extended look a t guilt.
This page intentionally left blank
This page intentionally left blank
Moral Residue s

Guilt came int o m y argumen t i n part I in two connecte d roles : i n its primary
form, following action, in response to a contrary-to-duty obligation to feel some
appropriate emotion, an d i n anticipatory form , i n advanc e o f action, a s th e
emotional strut of the motivational force of moral "ought. " I have maintained
that the two role s allow the emotion t o serve as a kind of substitute for action
in cases o f dilemma. I t i s by no mean s a n adequat e substitute i n mora l term s
but i s enough to answe r metaethica l worries abou t th e sense in which both of
the ought s i n conflict can b e practical. Th e motivationa l rol e o f th e emotio n
also serves to justif y i t as a sometimes problematic after-the-fac t reaction-an
affective residu e of moral failur e that persists even when failure is unavoidable.
My treatment o f dilemmas as cases in which guilt is warranted fo r al l alterna-
tives is essentially a modification of Williams's claim that thei r practical reso -
lution leaves a moral "remainder," no t itsel f a feeling but marked by moral feel -
ings of regret, alon g wit h othe r way s of acknowledging that th e ac t the agen t
has to perform is still wrong. However, alon g with most philosophers, Williams
has little to say about guilt.
Williams's "agent-regret" migh t be thought o f as a general category meant
to include guilt along wit h othe r emotions , thoug h i t suggests a more passiv e
variant o f sadness. 1 Bu t Agamemnon's case counts a s a genuin e dilemma for
Williams just because the sacrifice of Iphigenia is still morally wrong unde r the
circumstances, eve n though i t is required by Agamemnon's dutie s a s military
commander. I t does not just involve a regrettable wrong don e t o his daughter
that i s analogous t o th e "mora l cost " of the thing s a politician ma y b e over-
ridingly required to do. Rather, its wrongness i s serious enough not to be erased
by th e balancin g of obligation s tha t make s i t com e ou t a t th e sam e tim e a s
required, perhaps eve n with stronger practica l weight. My vie w is that guil t is
needed to capture the force of "wrong" in this account as something more tha n
a motivationally inert label.
At the outset, however, I want to continue to work with a broad interpreta -
tion o f guilt as an "interna l sanction" o f the moral cod e involvin g some form
of discomfort a t the thought that one is responsible for a wronga feelin g tha t
may b e covered by other emotion concepts, mos t notably shame, in other cul-
tures. A distinction between guilt and sham e with some reasons fo r preferrin g
guilt, plu s a n explanatio n of th e pitfall s o f th e emotion , wil l emerge as this
110 Sensibility an d Standpoints

chapter proceeds. In chapter 5 I argue that th e standard way of distinguishing

guilt from variou s other self-directed contrary-to-duty feelings in our ow n cul-
tureby limiting its object to voluntary actiondoes no t really fit the facts of
emotional lif e o r eve n the norm s of emotional rationality . Williams bypasses
guilt in favor of regret a t least partly because he relies on this distinction i n his
treatment o f dilemmas. In fact, though, on Williams's account the substitution
of obligation for practical "ought" in dilemmas would apparently allow blame,
the third-perso n counterpar t o f guil t in emotional terms , fo r th e violatio n of
moral obligation i n such cases. 2 But appropriate blam e no less than guilt seems
to b e limited to voluntar y action. My ow n view will allow instead fo r weaker
grounds for guilt than for blame on the basis of the special motivational role of
the first-person emotion .
In this chapter, befor e attacking the question of voluntariness, I want to fil l
out th e accoun t o f guil t that bega n t o emerg e i n m y discussio n o f it s role i n
childhood moral teaching. My treatment of the moral significance o f the emo-
tion i n adult life (sectio n 1 ) will initially focus on the lin k i t provides betwee n
the ethics of virtue and of duty. Insofar as it captures the intractability of moral
wrongits resistance to the balancin g of obligationsin a case lik e Agamem-
non's, i t expresses th e sens e tha t som e act s ar e intolerable . In Agamemnon' s
case it essentially serves to direct feelings of "taboo" onto the self, as tainted or
stained b y wrong action . I t thereb y provides a n elemen t of mora l self-threa t
that is not present in more passive reactions lik e regret or shame or in less self-
oriented feeling s lik e remorse. I t also provides an alternative to a strictly Aris-
totelian approach t o virtue by making room fo r serious lapses from it within a
notion o f flawed or imperfec t virtue.
I eventuall y bring my remark s o n guil t and virtu e to bea r o n tw o genera l
contrasts tha t see m to underli e the motivational significance of guilt: between
positive and negative and between self - and other-directed emotions an d othe r
attitudes. My argument for an asymmetrical treatment of emotions in each pair
will lay the basis for my later defense of weaker grounds for guilt than for blame.
Here I shall use it to help answer questions raised by a problematic case for the
connection betwee n guil t and perfec t or idea l virtue, of the sort that seem s to
be possible in genuine moral dilemmas , understood a s cases in which the agent
is not responsibl e for th e situatio n o f conflic t and assumin g that h e doe s th e
best he can to ac t in light of it. The discussio n will lead to my account, i n sec-
tion 2 , o f guilt in the narro w sense a s an identificator y mechanism: a general
reaction pattern sometimes encompassing the other feelings typically contrasted
with guilt but distinguishe d by a different sor t o f connection t o th e self.
I expect m y remarks t o suppor t a distinctio n between guil t and shame , in
particular, that favors guilt as a moral motivator. Bu t I shall mainly be consid-
ering guilt in reasonable dose s and wit h reasonabl e objects, in sharp distinc -
tion to a common vie w of the emotion a s an overwhelming and uncontrollable
inhibiting factor. The pitfall s o f the emotio n will come u p i n my argument as
side effect s o f it s valuable features an d a s reason s for denyin g that a require-
ment to feel guilty provides a way of resolving a dilemma. In a defense of "ought -
to-feel" (sectio n 3), I address the general problems raised by requiring any feel -
Moral Residues 11 1

ing, o n the usua l assumptions abou t emotional control . I also consider a sec -
ond-order dilemm a raised by the possibility that guilt in a case like Agamemnon's
might interfer e wit h effectiv e actio n o n th e stronge r ough t i n the conflict . A t
best, if I am right, a requirement to feel guilty resolves the metaethical proble m
of dilemmas and practical "ought. " It cannot b e taken as offering a way out t o
the agen t i n a dilemm a or a s preserving th e ethic s o f virtue from mora l luck .

1. Th e Mora l Significance o f Guil t

Guilt i s usually thought ofb y thos e philosopher s wh o mentio n i t at alli n
connection wit h th e ethics of duty, which takes acts as the primary objects of
moral evaluation. 3 It is meant to attac h t o an act viewed as wrong, or a viola-
tion o f moral obligation . However , within duty ethics it is treated a s a second -
ary matter, a question to be postponed unti l after the determination of the right
act. I want to argue that the moral significance of guilt begins to come out whe n
we consider i t as an element of virtue ethics, or the earlier (and recently revived)
approach tha t instead emphasizes the evaluation of persons and personal traits
in connection wit h notions o f character an d personal mora l perfection .
Guilt can be seen as linking the ethic s of virtue to tha t o f duty more firml y
than i s accomplished, fo r instance, jus t by listing conscientiousness amon g th e
virtues or by laying special stress on the virtue of justice. If conscientiousness i s
just on e virtu e among others , it s requirements ma y sometime s b e slighted i n
favor o f others compatibl y with overall virtue. If justice, on the othe r hand , i s
the essential virtue, a single lapse from it presumably takes an agent out o f the
running fo r overal l virtue. Even if virtue is subject to degrees , it s degrees ar e
not calculated act by act, so a serious lapse may make it unclear what motiv e is
left to the agent for future virtuous action. By contrast, taking justice to involve
a requirement o f guilt for wrong actio n a s a precondition o f overall virtue has
the effec t o f imposing a repeatable emotiona l cost on lapse s from virtue.
It is important t o thi s suggestion tha t guilt both ascribe s somethin g nega -
tive to th e sel f an d i s itself a negative state o f feeling. It is not itsel f a virtue
nor i s feeling i t o r havin g a tendenc y t o fee l itbu t rathe r a requiremen t o f
imperfect virtu e and a goad t o futur e virtuous action. I shall be thinking of it
here as an emotion, no t a state of affairs o r a personal trait. To that extent, my
treatment o f guilt will be roughly in line with Aristotle's remark s on sham e in
the Nicomachean Ethics; sinc e the Greek s did not hav e a shame/guilt distinc -
tion, I shall tak e Aristotle' s remark s a s applying to bot h emotions. 4 Aristotl e
apparently consider s sham e importan t enoug h i n connection wit h th e virtue s
to b e included in his account o f them, bu t a t th e sam e tim e h e acknowledge s
that i t doe s no t reall y belon g o n th e list . A s a correlat e o f imperfec t virtue ,
shamemeaning th e sens e o f shame, o r sham e a s a dispositiona l emotion , a
tendency t o fee l certai n occurren t emotion s (includin g both sham e itsel f an d
fear of disgrace) under the right circumstancesmay be seen as a virtue in chil-
dren. For adults, it is neither a virtue nor a part of virtue on Aristotle's account,
which a t tha t stag e o f developmen t rule s ou t seriou s lapse s fro m perfection .
112 Sensibility an d Standpoints

My own use of "guilt" i s meant rathe r broadl y at this point, in a way that
does no t distinguis h it from mora l shame , o r sham e fel t fo r a n ac t viewe d a s
wrong. Even that limitation might seem to introduce a difference fro m Aristotle' s
notion insofa r as it connects guilt to duty ethics. Though Aristotle' s shame does
extend t o lapse s from dutyindeed, he ties it explicitly to voluntar y actions 5
it more directly concerns the personal disgrace to which they subject the agent.
One's sel f i s viewed as diminishe d by th e shamefu l wrong i t does , i n short ,
whereas fo r guilt , we ma y sa y b y contrast, th e sel f ma y merel y be threatened
with diminution: Wrong acts "taint" the agent, but, in normal cases he can erase
the taint through reparativ e action.
In what follows I argue for the importance o f guilt even within virtue ethics
as a negative response t o wron g action ; I thereby resist bot h Aristotle' s insis -
tence on perfection and a certain use of Aristotle i n contemporary mora l phi -
losophy a s a model for isolating virtue ethics fro m th e modern ethic s of duty .
Williams, mos t notably , favor s dropping strictl y moral notion s i n favor o f a
broader categor y of the ethical, which is not base d on blame. 6 While I welcome
the mov e awa y fro m a narro w concentratio n o n duty , m y ai m her e i s essen-
tially to hel p keep the ethics of virtue moored to that of dutyto kee p it, as I
would say , from drifting ou t t o sea . I also turn t o a case of moral dilemm a t o
illustrate an important distinction within virtue ethics introduced b y the exten-
sion o f emotional guil t to putative instances of moral perfection.

Guilt and Imperfect Virtue

Let us first not e that we are not simpl y returning to dut y ethics when we refer
to guil t as required b y imperfect virtue (in the ordinar y sens e of flawed as op -
posed to ideal virtue, with no reference to the Kantian distinction betwee n per-
fect and imperfect duties). Our claim at this point need not be that guilt is obliga-
tory i n the usual sensea possibility I shall in fact defen d later i n this chapte r
but rathe r just that undergoin g the emotion i s a precondition o f such virtue
as is still achievable under the circumstances. From the standpoint o f duty eth-
ics, this claim would seem to amount to a hypothetical imperative"Feel guilty
if you want to be virtuous" or the likeand guilt would no t b e commanded by
such a principle , except a s supplemented b y a dut y to exhibi t virtue . Within
virtue ethics, by contrast, th e claim just serves to recommend guil t as one of the
necessary ingredients of virtue in the cas e in question, with virtu e assumed t o
be a state w e want t o exhibit . It is a question, o f course, wha t forc e the clai m
would have against someon e wh o doe s not muc h car e whether h e qualifies as
virtuous; an answer might overlap with duty ethicsby appealing, for instance,
to th e claim on others' respec t that virtu e entails. But let us limit attention a t
this poin t t o agent s who d o hav e a n idea l o f overall virtue, though the y ma y
not ac t in accordance with it on all occasions o r even bother figurin g out what
it requires .
The questio n of overall virtue, as I interpret it, asks whether someon e i s an
admirable person, o r admirabl e on the whole, a s opposed t o bein g admirable
only with some qualificationin certain respects or in a certain role; as an art -
Moral Residues 11 3

ist, say . Moreover, an answer to the question involves more than simpl y sum-
ming the more limited sorts of virtue that an agent may displayeven with due
weight assigned to their importance, the degree to which they involve a display
of virtue, and so forth. Thus, for instance, we might deny that Richard Wagner
was an admirable person in view of his anti-Semitism or his betrayal of his friend
Von Billow , eve n granting that h e was a ver y admirable composer, and even
with some inclinatio n to say that his achievements as a composer wer e impor-
tant enough to outweigh moral failings. His moral failings may be outweighed
in some general scheme of things, that is to say, but not i n the determination of
overall worth as a person o n this account.
Indeed, a similar point might be made for cases in which only moral virtue
is under consideration a t ever y stage. If a moral laps e is sufficiently serious , it
will not be enough to make up for the lapse with good deeds; rather, the agent
must appreciate it s seriousness, in a sense not unlik e aesthetic appreciatio n t o
the extent tha t it rules out bein g left cold. A Raskolnikov, say, who goes on to
become a major philanthropist without a moment o f remorse for his murder of
the old lady would not thereb y have met the demands of overall virtue, even if
he also managed to convince himself intellectually that the murder was wrong.
Indeed, eve n a religiou s conversion would no t b e morally satisfyin g unless it
involved an appropriate elemen t of discomfort abou t hi s crimes. We might be
willing to infer this from th e strength of his later motivation to do good, take n
as an instance of the reparative tendency associated wit h guilt. But at least some
postulated negativ e feeling seem s t o b e needed t o assur e u s that th e agent' s
negative evaluation of his act affect s hi m personally .
Some such argument is familiar from discussions of legal punishment, reha-
bilitation, and surrounding issues. The hypothetical rehabilitated mass murderer
who goes on to lead an exemplary life but never feels remorse gives pause, quite
apart from deterrenc e considerations, even to those of us who would like to do
without revenge . No doub t w e insist on feeling , in real-lif e case s o f the sort ,
partly jus t to hel p us determine whether an y observed change s in personality
and mode of life are genuine and reliable, on the assumption tha t feeling is one
of the les s malleable signs of belie f an d behavio r tendencies. Bu t we insis t on
guilt feelings i n particular (taking guilt broadly to includ e remorse an d mora l
shame) becaus e it is also importan t to u s that such changes rest on emotiona l
For practica l purposes , i t would presumabl y be enoug h i f we could ge t a
mass murdere r t o fee l horro r a t he r past acts , pity fo r th e victims , and othe r
alternatives to self-directed discomfortsupposing, fo r instance, tha t w e had
some method o f rehabilitation that resulted in her ceasing to identif y wit h her
past self. Psychosurgeryor, for that matter, religious conversionmight have
that result, but I take it that in moral terms we would not b e satisfied. The dis-
continuity between the agent's pas t and present selves would seem to undercut
any judgment of overall virtueby substituting two selve s for one, as it were
rather than justifying a judgment of imperfect virtue. To the extent that we care
about the moral worth o f persons, we prefer a method of rehabilitating some-
one that does not essentially involve giving up on her. Other things being equal
114 Sensibility an d Standpoints

(as of course the y may not be) , we want th e agent to redee m her old lif e rathe r
than simpl y to launc h a new one .
It may be that wrongs as serious as those now in question have to be said to
rule out anythin g worthy o f the name "virtue, " eve n qualified. Guilt may no t
be sufficient, tha t is , even in combination wit h good behavior i n the future, fo r
a judgment of imperfect virtue that applies to the agent's life as a whole in such
cases. However, guilt still seems to be necessary to something we think morally
valuableand something that falls within the scope of virtue ethics to the extent
that it involves a judgment of persons and their characters an d lives rather than
being adequately handled by claims about what acts ought to be done. The result
may be said to be a "graded" notion of virtue, taken as covering diverse notions
of personal moral worth, not all of them covered in ordinary language (or stan-
dard philosophi c usage ) by the ter m "virtue. " Eve n in a case in which real re-
demption i s impossible, that is, we still seem to place a value just on facing u p
to th e past . We admire someone wh o insist s o n doin g s o at som e cos t t o his
own peace of mindand perhaps eve n to hi s effectiveness a s a moral agent in
certain cases. The notion o f a noble character seems to include a kind of height-
ened sensitivit y to one' s own moral wrongs . W e sometimes thin k o f this as a
nobler ideal than mora l purity, for that matter, s o that imperfect comes out as
better in a way than perfect virtue. Other terms might be substituted for "virtue "
"moral decency, " fo r instancewhere "virtue" o n its ordinary us e does not
quite apply.
Our idea l o f mora l self-sensitivit y can giv e ris e t o duty/virtu e conflicts
potentially dilemmatic choices between doing the right thing and displaying the
requisite emotional reactionbecaus e o f the cripplin g effect s o f guilt in some
cases. Ther e ma y eve n be cases in which overall virtue is not achievable , even
with qualifications , because o f a conflic t within virtue ethics , whic h o n thi s
account wil l be influenced by the contingencies of the agent's record o f moral
action to date. Also, of course, our ideal leaves open many questions about the
type an d degre e o f wrong fo r whic h guil t is requiredwhether jus t fo r "in -
character" violations , say , or rather fo r any major lapses fro m virtuous char-
acter o n the part o f the agent .
Even with some very rough edges, though, the ideal manages to fill two gap s
in standard virtue ethics. First, it yields a "time-bound" vie w of virtue that allows
us to ask what is still achievable in a life that may already include some serious
and irrevocable deviations from perfect virtue. By contrast, Aristotle's dismissal
of shame in virtuous adults underlines the uncompromising quality of his con-
ception o f virtue . The lis t of virtues derived fro m Aristotl e i s not reall y well
designed, one might say , to advis e an agen t in medias res an agen t decidin g
what to d o at some particula r poin t i n his lifeas oppose d to a n educator o r
someone else who is in a position to plan lives from the outset o r to judge them
as a whole. By supplying a notion o f imperfect virtue that includes some nega-
tive feeling s abou t oneself , we modif y Aristotle' s essentially prideful idea l by
building in serious gradations, instead of simply balancing it with an indiscrimi-
nate ideal of humility as on some standard religious extensions of the Aristote-
lian model.
Moral Residues 11 5

Second, a s a corollary o f the time-bound view, our idea l of moral self-sen -

sitivity provides a way o f representing withi n virtue ethics th e stricte r action -
guiding status of certain ought-statements though t of as commands rathe r tha n
simply recommendations . Tha t i s to say , i t let s u s preserve som e normativ e
notion o f binding "ought," conceived a s a requirement that th e agent canno t
get out ofb y developin g himself in other dimension s of virtue, say, or by com-
pensating withi n the mora l sphere . The idea l accomplishe s thi s withi n virtu e
ethics b y insisting on a kin d o f emotiona l reparatio n fo r mor e seriou s viola -
tions a s a condition of such virtue as is still achievable. From amon g th e many
things that a virtuous person woul d no t do , tha t is , it singles out som e a s in-
compatible even with imperfec t virtue unless the agent pays a certain affectiv e
price. By thus allowing for a wrong act's conditional compatibility with a lesser
degree o f virtue, it keep s virtu e ethics fro m pushin g the agen t i n suc h a cas e
beyond the ethical pale, declaring him incapabl e of any morall y estimable lif e
by din t o f hi s futur e behavior . I t ensure s tha t eve n the morall y flawe d agen t
will b e subject t o som e persona l norms , i f only norms o f imperfec t virtue, in
what he goes on to do .
Someone migh t sugges t tha t th e Aristotelia n model ca n secur e thi s result
just a s well by appealing act b y act t o th e standar d provide d b y the perfectly
virtuous person. That is, what a n imperfectl y virtuous person ough t t o d o on
any give n occasion wil l be jus t what th e perfectl y virtuous person would do ,
even though i t is no longer possibl e to attai n perfect virtue by so acting. How -
ever, I think we can see that differen t norm s sometimes apply to the imperfectly
virtuous. For one thing, consider th e Aristotelean ideal of proper prid e that was
just mentioned: If nothing else, Aristotle's own picture of what this requires will
have to admit o f gradations h e fails to recognize: some alternatives to the sort
of loft y self-regar d he expect s o f th e virtuou s person . Beside s this, there ar e
specific "contrary-to-duty " obligationsobligation s t o apologiz e o r to mak e
amends, most notably; bu t also other obligations based on changes i n the situ-
ation resulting from the agent's deviation s from ideal behaviorthat would no t
apply to the perfectly virtuous agent. Perhaps Aristotle could accommodate thes e
under "rectificator y justice," 7 i f the category were stretched t o cove r state s of
mind. M y poin t her e i s just tha t guil t a s a requiremen t o f imperfec t virtu e
amounts t o a compensatory stat e o f feelin g tha t doe s no t fi t the Aristotelia n
model without some stretching .

Guilt and Perfect Virtue

I now want to switch t o a different sor t of case, i n which guilt seems to b e re-
quired by perfect virtue. This is a case of moral dilemma, where the agent's recor d
of mora l actio n i s necessarily imperfect, since the circumstances leav e him n o
choice but to do something wrong , eve n though he is not to blam e for the cir-
cumstances an d doe s th e bes t h e can t o resolv e the dilemma . Le t u s fo r th e
moment ignore the various questions raised by insisting particularly on guilt in
such cases and not e tha t ou r intuitiv e treatment of them doe s seem to requir e
some mora l emotio n o f th e sort . A nice cas e fo r discussio n i s provided b y a
116 Sensibility an d Standpoints

passage fro m Bertran d Russell' s autobiograph y cite d b y Marcia Baro n i n a n

argument for the inadequacy of Williams's "agent-regret." 8 Russell limits him-
self i n the passage to an expression o f sorrowno guil t or other emotio n con -
cerning the moral quality of his actfor a time when his activities as a pacifist
during World War I required him to jil t an American woman h e had promise d
to live with i f she persuaded he r father to brin g her to England .
The case a s it stands ma y seem t o b e what Williams calls a case o f "moral
cost"of wrongin g someon e bu t withou t doin g somethin g wron g unde r th e
circumstancesthough Russell' s description o f it provides enough ra w mate -
rial fo r a full-scal e dilemma . Even , without modificatio n i t allow s u s t o rais e
some importan t question s abou t jus t ho w muc h our idea l of moral sensitivity
might require. Russell writes:
When sh e arrived I could think of nothing but the war, and as I had determined
to come ou t publicl y against it , I did no t wis h to complicate m y position wit h
a privat e scandal , whic h woul d hav e mad e anythin g that I migh t sa y o f n o
account. I felt i t therefor e impossible to carr y ou t wha t w e ha d planned . She
stayed in England and I had relation s with her from tim e to time, but the shock
of th e wa r kille d my passio n fo r her , an d I broke he r heart . Ultimately she fel l
victim t o a rar e disease, whic h firs t paralyze d her, an d the n mad e her insane .
. . . Before insanity attacked her , sh e had a rar e an d remarkabl e mind an d a
disposition a s lovable as it was unusual. If the war had not intervened , the plan
which we formed in Chicago migh t have brought great happiness to us both. I
feel stil l the sorrow o f this tragedy. 9
The case is conceived as one of clear moral choice: Russell's obligation to work
effectively agains t the war is presumably strong enough to make it right to brea k
a privat e promise. Bu t even assuming that the promis e itsel f wa s blameless
along with Russell's later behavior toward th e promiseethe promisee stil l has
a righ t to fee l aggrieved . The cas e a t leas t resembles a dilemm a to th e exten t
that it involves an unmet claim: The wrong done to the woman i s from her stand-
point not reall y canceled out b y the importance of Russell's antiwar activities .
We may grant that the latter ar e important enoug h to mak e the promise rela-
tively trivial and Russell' s act permissible, however.
As Baron points out , Russell' s reaction stil l seems disturbingly impersonal.
It is not quit e a cold reaction; rather, I would say (modifying Baron), it is warm
in a suspiciously sentimentalized way. Wha t sound s lik e some sor t o f heredi-
tary mental illness is inserted into the tal e as if it resulted from a broken heart .
Such regre t a s Russel l attributes to himsel f doe s no t see m to b e agent-regret,
since it is not directed toward his action. He says he feels the sorrow of the trag-
edy that befel l th e woman, but presumabl y he has i n mind the tragedy of her
overall lif e rathe r tha n particularl y hi s own contributio n t o it , which di d no t
itself result in tragedy. Williams would surely accept Baron's claim that our ideal
of overall personal virtu e requires more than this .
Does it require guilt, at least in a broad senseof discomfort at responsibil -
ity (leaving it open what sort of responsibility) for a wrong? As the case stands ,
we might want to say that Russell really ought to feel a twinge of guilt or remorse,
perhaps just as a stage on his way to overall sorrow, while thinking of the woman
Moral Residues 11 7

and hi s behavior toward he r as vividly as he is now doin g i n writing his auto-

biography. But he is under no obligation on other occasions t o dredge up those
thoughts i n the first placeor to consider the situation from the woman's stand -
point. Love may require this, but that is another story, irrelevant to the present
question, unles s we suppose tha t i t obligates th e agen t t o remai n i n a state of
love. Fro m a n all-things-considere d moral standpoint , i t seems, the woman's
claim on Russell was relatively unimportant, an d his behavior toward her was
justified. A t an y rate , le t u s grant thi s poin t fo r purpose s o f argumen t alon g
with others that might be questioned in Russell's own understanding of the case,
since our questio n i s what emotio n i s called for b y the cas e a s Russell under-
stood it.
The case could amount to a full-blown dilemma , as I understand the term,
only i f by jilting th e woma n Russel l had don e somethin g harmfu l enoug h t o
count as wrong even all-things-considered. A different sor t of case in which she
had alread y disrupted her life seriously in light of the promise (ruined her repu-
tation b y living with him , say) might yield a pacifist paralle l to Sartre' s case .
Another, more fanciful possibilit y is suggested by the link Russell may be imag-
ining between his act and the woman's late r illness. If heartbreak or something
similar ha d reall y helped caus e th e degre e o f har m tha t late r befel l her
particularly supposing that the link could have been foreseenwe might indeed
hold tha t an adequately sensitive moral reaction on the part of the agent in the
case would hav e to includ e some element of guilt.
This is not to say, of course, that the reaction has to last forever. Let us get
away from the question of what Russell should be feeling many years later while
writing his autobiography an d just note that we do now seem to have a case in
which eve n perfect virtue requires feelings o f guilt , as agains t wha t Aristotl e
suggests about shame when he limits it to children. Either that or our notion of
virtue has to be modifiedmade "extensional, " a s it wereto take in the moral
quality of an agent's life , the record o f what h e does in life, along with his per-
sonal qualities of character. I f Russell in this hypothetical case had been able to
pursue his antiwar activities only by doing serious wrongbut on the assump-
tion tha t thi s was the best choic e available to hi m under th e circumstances
such wrong a s he had t o do would not hav e been chargeable to flaw s o f char-
acter. Indeed, one might say, it sometimes takes strengt h o f character i n such
cases not to be inhibited from action by moral squeamishness. At the same time,
it seems that a truly noble character would lead the agent to reflect on his past
act at som e time or othe r an d underg o some variant of guilt.
Thus modified, Russell's case (as I shall still call its modified version) brings
out a number of important points and questions about ou r idea l of moral sen -
sitivity. As we just saw, i t suggests that w e need an extensiona l reading of the
notion of virtueas the agent's recor d o f moral merit , let us sayin order t o
accommodate ou r expectation s o f an agent i n a dilemma. Or better , perhaps :
There are two possibilities to be distinguished here, one of them in conflict with
the clai m that perfec t virtu e in such dilemmatic cases requires guiltthough
both o f the m i n differen t way s migh t see m t o constitut e departure s fro m
Aristotle's assumptions. On the one hand, we might think of merit as a distinct
118 Sensibility an d Standpoints

notion from virtue properthe alternative I have proposed. A judgment of merit

might be said to involv e an assessment of the agent's lif e story, or his life story
insofar a s it is active, in contrast t o both hi s character and th e record o f things
that happen to him. For we seem to need two notions of personal excellence to
capture what w e want t o sa y about Russell's casenamely, that virtu e in its
standard sense , or noble character, requires guilt on Russell's part in response
to a blot on hi s record o f action a s measured by our nonstandar d alternative
notion o f virtue as moral merit.
On the other hand, if we think of merit as replacing the standard notion of
virtue, we apparently have to drop the claim that perfect virtue requires guilt
welcome news, perhaps, to those who find the claim odd in its mixture of notions
of virtue and duty . Merit, it might be said, is simply virtue conceived as subject
to mora l luck , including a furthe r sor t o f luck besides that recognize d on th e
standard conceptio n o f virtue , which does no t allo w fo r dilemmas. 10 Perfec t
merit simply is not ope n to a n agent in the unluck y circumstances of Russell's
case, where through no fault of his own he has to do something wrong. I f merit
amounts to virtue, this just means that, in describing such dilemmatic cases, we
need to retreat to the claims made earlier about gradations of imperfect virtue.
In the end, though, I think that an appreciation o f the emotional requirements
of Russell' s case wil l tel l i n favo r o f retainin g a conceptio n o f virtu e as itsel f
untouched b y dilemmatic luck and i n more general terms will shed some light
on the appropriate connectio n between virtue ethics and th e modern ethics of
Up to this point m y argument on guilt and virtue has essentially bypassed
duty ethics by treating the requirement to fee l guilty as ideal rather than prac -
tical. The contraste d term s correspon d t o tw o alternativ e readings of ought -
statements: o n the one hand, ranking something highest, or commending it to
choice, an d o n the other hand , commanding an agen t t o choos e i t or brin g it
about. A n idea l ough t evaluate s but doe s no t prescrib e it s object , i n othe r
wordsexcept o n conditio n tha t th e agen t want s t o fulfil l th e ideal . S o in
application t o Russell' s case it tells us only what h e would fee l i f he were per-
fectly virtuous , not wha t feeling s (i f any) are required of him simpliciter. On e
might want t o sa y that fro m a general moral standpointa s oppose d t o on e
required by love, say, or some similar ideal that calls for imaginative participa-
tion i n the standpoin t o f the woman wit h who m h e was involvedguil t feel-
ings on Russell's part would be supererogatory. Self-subjectio n t o them consti-
tutes a condition of special merit or virtue, beyond what duty commands. M y
own view, to b e defended in section 3, is that in typical cases of dilemma guilt
feelings are indeed obligatory and that a reading of "ought" that prescribes them
is part of what keep s dilemmas from underminin g action-guiding ethics.
My ful l argumen t for this last point wil l depend o n th e general account of
guilt I go on to give in this chapter and the next. In response t o those (perhaps
Russell himself) who would argue that guilt is irrational or unnecessary for an
act of perfect virtue , it is possible to appeal beyond intuitions on specific cases
to a general view of the functio n o f the emotion. Morally speaking , emotions
like guilt have a special role to play in an approach t o ethic s that gives serious
Moral Residues 11 9

attention t o persona l standpoint s i n th e wa y recommende d b y many of th e

current proponents o f virtue ethics.11 In Russell's case, guilt serves essentially
to register the standpoint o f the woman h e had to jilther justified resentment
of his actionalongside the overarching moral standpoint tha t determined his
In the end, then, what counts in favor of guilt in a case like Russell's will be
its fit with an intelligible overall account of morality and emotion. Fo r purposes
of my more limited argument in this section, however, let me end with an attempt
to defend my treatment of Russell's case by detaching the claim I mean to make
about i t from various more questionable claims that migh t b e thought to fol-
low. The claim that Russell ought to fee l guilt y for having jilted the woman in
the case does not mean , firs t o f all, that a reasonable person woul d encourage
Russell to fee l guilty. Even from a moral standpoin t (an d there are others), we
may recogniz e overriding reasons t o forg o that ofte n debilitatin g emotion; a s
mentioned earlier with reference to other cases, feeling guilty might undermine
an effective decisio n to avoid the other horn of the dilemma by keeping Russell
from concentratin g on his antiwar activities . Feeling guilty later migh t not b e
an option , moreover . Fo r one thing, guilt might conceivably just be crowde d
out b y other appropriat e feeling s onc e th e woman' s horribl e fat e ha s over -
shadowed an y immediate harm done b y jilting her. 12
Secondly, a deepe r poin t t o not e i s that Russell' s cas e seem s t o involv e a
basic asymmetry between guilt and other-directed emotional blame as responses
to a less than perfect moral record. When we say that Russell ought to feel guilty
(or ought at some earlier point to have felt guilty) about jiltin g the woman, we
do not mean to recommend blame or personal anger toward someon e who acted
as he did, however understandable it may be on the part of the woman. Under
the circumstances (on our assumptions abou t th e case), his action towar d th e
woman was morally required, and h e was not a t fault fo r the fact that i t was .
In requiring that he also feel guilt y about hi s action w e would not b e assessing
him negativel y for i t in emotional term s bu t rathe r insistin g that h e so assess
himself. Nor would his self-assessment commit him to a similar emotional assess-
ment o f others i n his position. H e woul d b e committed onl y t o holdin g that
they too ough t to fee l guilt y for act s o f the sortno t t o a view of them as in
fact guilty and henc e appropriate objects of third-person blame . To the extent
that a feeling o f guilt on Russell's part involves a past practical ought , then
directed toward the act that he holds himself responsible for omittingwe might
seem to b e recommending that i t be limited to his own cas e i n a way that th e
logic of duty ethics is thought to rule out.
Even from the standpoin t o f virtue ethics considered i n isolation, th e case
serves to highligh t the self-directednes s or reflexivity of the conten t o f a first-
person ideal ought requiring guilt. The expression "hold s himself responsible "
just applied to Russell is significant because it brings out the fact that what we
are demanding of him is the impositio n of some sort of moral burde n on him-
self. This does not imply that we (or he) would have a right to impose that same
burden o n otherst o hold them responsible directly, that is , as distinct from
expecting them to take that stance toward themselves. It is important, too, tha t
120 Sensibility an d Standpoints

we do not thin k of those who satisf y th e ideal of perfect virtu e just as persons
of whom w e expec t something . Ou r notio n o f perfec t virtu e is specificall y a
notion of emotional or other burden s self-imposed.
We need to retain a distinct notion of virtue, then, as unaffected b y dilemma
in order t o capture within virtue ethics our sens e that the agen t i n a dilemma
acts blamelessly. Aristotle's notio n o f virtue is picked out b y its connection t o
praise and blame . What m y discussion here indicates is that du e attentio n t o
the special features of guilt as the first-person counterpart of blame introduces
a further notion of virtuean extensional notion subject to moral luck and hence
more tightly tied to happiness than virtue in Aristotle's sense seems to be.
More fundamentally, the guilt/blame contrast in Russell's case seems to rest
on two asymmetries with important bearin g on questions of moral motivation.
It depends not onl y on the fac t tha t one of the two emotion s contrasted i s self-
directed but also on the fact that both emotions are negativebad states of feel-
ing, that is, directed toward states of affairs evaluated as bad. The contrast would
apparently b e reversed if the tw o coul d b e seen a s emotiona l reward s rathe r
than essentiall y punishments, give n th e sam e sort o f differenc e i n reflexivity .
Consider how we would react to a positive self-directed feeling in Russell's case:
pride, say, at having been able to resis t romantic temptation in the interest s of
the antiwar cause. Even if a parallel feeling of admiration for Russell would be
justified o n our part, we would n o doubt thin k less of him for reflectin g posi -
tively on the same facts. An emotional reactio n t o th e case, i n short, involves
taking a kind of position o n the agent's moral record whose own moral assess-
ment depend s no t jus t on accurac y bu t als o o n wha t i t doe s fo r th e agent
whether it inflates o r diminishe s himand on whether the agen t o r someon e
else undergoes it. In the end, then, the importance of guilt to virtue depends on
the fac t tha t i t is itself somethin g bad .

Two Asymmetries
Let us at this point paus e to conside r i n their own righ t the two asymmetrie s
that see m to cu t of f guilt fro m eithe r prid e or blam e in Russell' s case. Thei r
general importance , I want t o say , lie s in their lin k t o th e tw o side s o f emo -
tional justification: the "backward-looking " assessmen t of emotions as appro-
priate or inappropriate to the surrounding situation and their "forward-looking "
instrumental assessment i n relation to action. 13 Wha t I shall call the "qualita -
tive" asymmetr y (positive versus negative, as illustrated by pride versus guilt)
bears on our treatment of guilt as a reason for action and hence on its forward-
looking justification, as I shall indicate. On the other hand, what I shall call the
"directional" asymmetry (self- versus other-directed, as illustrated by guilt versus
blame) bear s more immediately on the noninstrumental reasons we expect t o
find fo r guilt, or it s backward-looking justification.
I shall begin with th e qualitativ e asymmetry, since its explanation i n gen-
eral terms seems to focu s o n norms o f individual rationality in contrast t o th e
social norms that underli e the directional asymmetry. However, applying it to
the rational assessment of emotions wil l b e a complex matter that als o brings
Moral Residues 12 1

in socia l norms ; an d th e mor e tellin g contrast fo r presen t purpose s wil l tur n

out to b e not rationa l bu t moral. Th e effec t o f the two asymmetrie s i n combi-
nationto make a long story shortwill b e a tendency to justify guilt on rathe r
slim grounds in contrast to either pride or blame. But its explanation will make
reference to the moral role of emotions insofa r as it involves assessing them a s
rewards or punishments.
Consider firs t th e motivational significanc e of the qualitative asymmetry
in the present case, the greater forc e for action of a negative emotion lik e guilt
than a corresponding positiv e emotio n suc h as pride. Intuitivel y it seems tha t
pursuing and attaining some positive goal is optional i n a way that the drive to
escape something negative is not: Th e latte r serves not jus t as a goal o f actio n
but also as a goad to it. However, w e should note that this way of putting thing s
blurs over a certain ambiguity in the classification of emotional state s a s posi -
tive or negative . I t assumes that th e relevan t classification is affectivedivid-
ing good fro m ba d feelingsso i t may no t fi t certain emotions whose qualit y
as feelings seem s to clash wit h the evaluativ e points the y make.
Love is commonly thought o f as positive, for instance, becaus e it says some-
thing positive about it s object, not becaus e it necessarily feels good. However ,
my assumption i s that a n emotio n lik e love can be analyzed into several pair s
of affective an d evaluative components i n a way that yields a qualitative match
for eac h pair. 14 A n apparently positive emotion ma y thus turn out to b e nega-
tive because it depends on desire, with desire taken a s involving discomfort a t
a negative evaluation of the current situationof distance fro m th e object, i n
the case of lovebut as yielding comfort whe n satisfied , so that th e emotio n
may sometimes still be pleasurable overall .
The two levels of qualitative assessment, affectiv e and evaluative, will match
up readily enough for guilt and pride. However, distinguishin g them brings out
another possible source of confusion in understanding what I want to say about
the qualitative asymmetry. That is , insofar as its motivational significance rests
on the contrast betwee n positiv e and negativ e affect, th e qualitative asymme-
try might seem to presuppose th e directional asymmetry rather than operatin g
independently: The greater force for action of a bad feeling depends on its being
bad for the agent. However , fo r "self-directedness, " in the sense at issue in the
directional asymmetry, the emotion i n question also has to be about the agent
to evaluate him positively or negatively in the way exhibited by guilt and pride .
The qualitativ e assessment o f emotions i s necessarily self-involved, one migh t
say, but whether a feeling is self-directed is a distinct question that depends o n
its evaluative content, no t jus t its element of affect, o r "feeling " in the narro w
Considered i n isolation , then , th e qualitativ e asymmetry amount s t o a n
imbalance i n motivationa l forc e betwee n positive an d negativ e feelin g state s
whose quality is assumed to coincide with their value for the agent. Th e asym-
metry concerns rational motivatio n to the extent that it is attributable to a dif-
ference i n the forc e o f positiv e an d negativ e affective reason s for action . I t i s
often take n for granted , as I noted i n chapter 3 , tha t a reaso n (o r a goo d o r
sufficient "all-things-considered " reason) for action amounts to a consideration
122 Sensibility an d Standpoints

in ligh t o f which i t would b e irrational no t t o act . Bu t the assumptio n seem s

really to hol d only for negative affective (instrumental ) reasons i n my sense
the sens e tha t implie s affective badnessa t leas t from th e standpoin t o f basic
or minimal (as opposed to perfect or ideal) rationality. It seems to be quite com-
patible with basi c rationality to ignore th e corresponding sor t of positive rea -
son for action: a positive feeling state seen as requiring action fo r its perpetua-
tion o r attainment .
Thus, fo r instance, to say that prid e offer s m e a reason fo r moral behavio r
might just be meant as a claim that I can attain or sustain a certain particularly
good stat e o f emotiona l self-assessmen t only (o r mos t effectively ) b y mora l
behavior. We need not suppose , tha t is , that i n the absence of moral behavio r
I would fee l particularly bad. If not, though , m y reason fo r action nee d not b e
viewed a s one I would b e irrational to ignoreo r no t unles s we supplemen t
our intuitiv e conception o f basic rationality wit h something more demanding ,
most notably one that insist s on maximizing the good. If rational intelligibilit y
is in question, that is , turning down somethin g good require s no instrumental
justificationit i s enough to say "I don't need this"whereas puttin g up with
something bad makes sense only on the assumption tha t th e costs of changing
it would b e too hig h or that it is likely to improv e things overall.
Again, these points ar e meant to b e limited to things perceived as good o r
bad for the agent, in particular the agent's affectiv e states . However , the quali-
tative asymmetr y will extend t o nonaffectiv e reasons tha t refe r t o a n agent' s
interests b y way o f futur e feelin g states . A n agent' s recognitio n tha t mora l
behavior i s a presuppositio n o f his futur e moral good , or virtue , may indeed
provide hi m wit h a rationa l goa d towar d action , dependin g on ho w h e feel s
about a lapse fro m virtue and o n ho w h e thinks he would fee l i f he exhibited
one, assumin g a rational requirement to take into account futur e feelings.
Alternatively, an agent may simply choose t o act to achieve a positive goal
of mora l virtue ; but h e will not b e rationally compelled t o d o so , even on the
assumption that he has such a goal and that i t does demand action o n his part.
At most, rationality will require that h e either act on the goal or abandon it. 15
By contrast, a goad towar d actio n o f the sor t tha t negativ e affect provides
perhaps jus t as part o f the desire to achiev e some goa l or o f the state i n which
one envisions it or of some later state one envisions if one does not achiev e it
is not subjec t to the agent's will in quite the same way. One cannot simpl y get
rid of a goad, as opposed t o a purely positive end or goal, by focusing on some-
thing else; so rationality requires attention t o i t in practical terms as wellif
only t o indirec t means o f getting rid o f iteve n when i t stand s a t som e dis -
tance i n the future . B y contrast, th e natura l urg e to sustai n a present stat e of
positive feelingassumin g n o negativ e element suc h as that introduce d b y a
threat o f loss or withdrawalseems to be optional fro m the standpoint o f basic
In short : Goal s ar e optional ; goads compel . W e ma y stil l sa y that a goa l
provides "sufficient" reaso n for action in the sense explained earliermeaning
that no further reason would be needed to justify actio n on it, but not that action
must occur in light of it, even assuming all-things-considered status and practi-
Moral Residues 12 3

cal rationality. Th e qualitative asymmetry emphasizes th e bad ove r the good,

then, as providing the kind of "compulsive" motivatio n tha t undermine s free -
dom b y making it difficult t o d o otherwise. 16 W e recognize it , in effect, whe n
we focus on internal sanctions a s motivational props of the moral code .
The significance of the qualitative asymmetry in relation to dilemmas begins
to emerg e when we consider ho w th e instrumental role o f guilt as a source of
moral motivation might also affect ou r assessment o f the warrant fo r the emo-
tion. We seem to adjust the standards of backward-looking justification to reflect
forward-looking considerations: An emotional response ma y be encouraged for
its general social (o r other practical ) adaptiveness, essentiall y b y relaxing the
evidential demands we would make on a corresponding judgment. 17 In the case
drawn fro m Russell' s autobiography, i f guilt as opposed t o prid e i s generally
useful as a moral motive, that might explain why we expect Russel l to feel guilty
about a wrong essentially forced on him by the moral demands of the case rather
than proud o f himself for satisfying thos e demands .
However, th e pride/guilt contrast doe s not work out quit e so neatly. First,
it seems to be only in moral terms that we rule out pride on Russell's part. If we
distinguish rationa l from moral o r socia l appropriateness , prid e seem s a per -
fectly appropriate reactio n to the case, providing it has the right sort of object.
Pride at having jilted the woman would not be appropriate, jus t because its object
is not itsel f praiseworthy, bu t pride at havin g resisted romanti c temptatio n i n
favor o f duty (o r something similar) might be sufficiently justifie d b y the evi-
dence. We do not tighten th e standards o f evidence, at any rate beyon d a cer-
tain point, fo r an emotion tha t i s not to b e encouraged o n grounds of general
adaptiveness. Accordin g to the rather generous "perspectival" account o f ap-
propriateness tha t I defend wit h referenc e to guil t in chapter 5 , sufficient evi-
dence for a judgment of guilt according to the usua l standards o f warrant fo r
belief will always be enough to warrant the corresponding emotion . S o assum-
ing tha t Russell' s antiwar activitie s were importan t enoug h t o mak e hi s act
praiseworthyas i t might b e even on a dilemmatic construction o f the case ,
where i t also counts a s all-things-considered wrongpride wil l be rationally
even if not morall y appropriate .
Second, however, questions about any simple denial of motivational forc e
to pride seem to me to shift attentio n to the directional asymmetry as the main
source of our weake r evidential standard fo r guilt. We may begi n by granting
that pride does no t motivate action in quite the same "compulsive " sens e as a
negative emotion lik e guilt. But this assumes that the form of pride in question
contains no element of psychological need of the sort that can sometimes make
positive emotions har d t o suppress . Fo r instance, such pride a s Russell might
feel for his resistance to romantic temptatio n i n favor of duty must be thought
of as a "quiet" variant of the emotion tha t would no t make him at all likely to
boast about his deed or about his impersonal cast of mind or the like. But spell-
ing ou t thi s assumptio n suggests tha t th e reaso n prid e i s not encourage d in
Russell's case i s not necessaril y because it lacks motivational force bu t rathe r
at leas t partly becaus e whatever force it ha s woul d ten d t o mak e i t socially
124 Sensibility an d Standpoints

On the other hand, "quiet" pride may seem to be adaptive to the extent that
it lends a kind of passive reinforcement to an urge toward actionin this case,
an urge toward mora l action. The contrast with guilt turns out to be undercut,
that is, by a different wa y in which pride can have practical consequencesnot
precisely by "motivating" action , i f that means providing a reason tha t makes
the agen t act , bu t rathe r b y putting the agen t into a stat e o f mind tha t facili -
tates action. I t i s useful fo r it s mood-lifting and othe r energizin g effects a s a n
enabling factor in the backgroun d of action even if not a cause of action in any
stronger sense. Even without the admixture of negative feelings, then, pride will
have a rol e to pla y in moral (an d other rational ) motivation distinc t fro m its
role as an end of action or positive goal. So the qualitative asymmetry need not
make pride less practically useful tha n guilt, despite its less compelling force as
a practical reason.
In any case, we also relax the standards of evidence for guilt on the basis of
certain noninstrumental considerationsof how one has a right to trea t one -
self versus othersthat my later perspectival account will allow for, along with
considerations of general adaptiveness. Let us therefore now brin g in the direc-
tional asymmetry , initially by comparing prid e on Russell' s part wit h a posi -
tive reaction to hi m on the part of someone else. Even in moral term s it would
presumably be open to u s as readers of Russell's autobiography t o admire him
for doin g what h e had to d o in the service of the antiwar cause . On e wants t o
say that w e have a right, in a way that Russel l himself does not, t o ignor e the
harm his act involves for others and react positively to hi m as its agent just on
the basis of other features of it that in this case are assumed to b e more impor -
tant. I n making this comparative judgment , we rely on th e directiona l asym-
metry, essentially by demanding that a self-directed emotion be more responsive
to significan t subsets o f th e availabl e evidence than i s expected o f it s other -
directed counterpart .
By contrast, w e d o no t hav e the righ t to reac t negativel y to anothe r per -
sonto condem n someone , emotionall y speakingon the basi s o f a limited
subset of the evidenc e bearing on his act, however significant evidence of that
sort may b e i n general terms. S o blame for Russellth e other-directed coun -
terpart of the guilt we expect of himapparently is ruled out. Indeed , it seems
to b e rule d ou t rationall y a s wel l a s morall y in backward-lookin g terms
unlike pride, which , as noted jus t above , seems to b e ruled out onl y morally.
Rather than being insufficiently warrante d by the overall facts of the case, pride
apparently rests on shiftin g attention awa y from a partial subset of themthe
facts abou t th e har m Russell' s act doesto whic h a virtuous agent woul d be
sensitive, even where something else is morally more important.
The proble m with blame , understood a s other-directed persona l anger , is
not that it plays a lesser role than guilt in moral motivation. At least insofar as
it is negative, it plays a roughl y commensurate rol e a s a motiv e fo r actin g t o
change one's interpersonal environment. To the extent that i t involves a nega-
tive evaluatio n of someon e else, though, there are als o noninstrumenta l rea-
sons that count agains t relaxing the standard s of backward-looking justifica-
tion as with guilt. From the standpoint of minimal charity it would be too much
Moral Residues 12 5

to expec t other s t o liv e u p t o th e highe r standard s o f behavio r that on e may

justifiably impos e o n oneselfan d that th e virtuou s person o n th e vie w sug-
gested her e must impose o n himself.
In short, the view holds that the demands of perfect virtue include a kind of
moral fastidiousness: painstaking attention t o morall y significant detail. Sub-
jecting oneself to negativ e emotion a s a way o f focusing attention o n a corre -
sponding evaluative proposition serve s to meet these demands in circumstances
that do not allow for action. I n defense of dilemmatic guilt, then, reacting with
discomfort eve n to partial evidence of responsibility for a wrong may be taken
as showing special concern fo r its victim at the cost o f some emotional harm to
oneself. But moral fastidiousness as a requirement of perfect virtue would seem
to yield just the opposite resul t for blame insofar as it is other-directed and henc e
involves subjecting someone els e to a negative evaluation. Indeed, I take it that
emotional blame is ruled out b y our normal expectation of charity toward oth -
ers, no t just by the more exactin g demands o f virtue, at any rate for those no t
in the position o f victim.
On th e socia l vie w of ethic s an d emotio n defende d here, th e qualit y an d
direction of these and simila r emotions give s them a significance a s rewards o r
punishments for action tha t ca n b e used t o provid e motivationa l backin g fo r
the moral code. Guil t and blam e bot h com e ou t a s "punishing " t o th e exten t
that the y direc t negativ e attentio n towar d thei r objects , wherea s prid e an d
admiration coun t a s emotiona l rewards . I t i s th e fac t tha t prid e i s a self -
administered reward, then, that would see m to make us morally wary of it in a
case like Russell's, even granting i t sufficient backin g to coun t a s rational. A t
the ver y least, a self-administered reward i s unseemly under circumstances in
which one' s goo d dee d depend s o n doin g serious wrong, eve n apart from th e
instrumental danger of encouraging future indifference t o the harms one's actions
cause. So we have a kind of reverse parallel of my treatment o f guilt as a mor-
ally encouraged self-subjection to punishment for harm caused, bu t with com -
plications introduced by our differentia l treatmen t o f evidence for and agains t
rational appropriateness .
I have more to say in chapter 5 about the guilt/blame asymmetry in strictly
rational terms, but at this point w e can already see how the notion o f guilt as a
self-imposed emotional burden gives it a special moral role to play in cases like
the one drawn from Russell' s autobiography. I t captures our ideal of the noble
character a s requiring more o f itself , i n short. But it combines this element of
moral self-sufficienc y wit h a sens e o f connectio n t o others , o f sharin g i n th e
harms one does to them, t o the extent tha t it involves emotional identification
with the victim of wrong. It is this that seem s to be missing from Russell's sor -
row a t the unfortunate demise of the woman h e jilted. I now want t o go on t o
argue, however, that something more like guilt in the narrow sens e is best suited
to supply this lack.
I shall leave off further discussio n of Russell's case until I have set forth my
account o f guilt . I n advanc e of th e account , I would expec t som e reader s t o
recoil at th e suggestion that guil t best fits the case . I indicate a number of rea-
sons for this in my next section, but on e rather obvious reason I want to dispel
126 Sensibility an d Standpoints

at th e outse t ha s to d o wit h th e sexua l overtone s o f guilt in popular culture .

This may suggest to some readers that guilt is to be imposed on Russell for sexual
misconduct, bu t nothin g of the sort is intended. At any rate, shame, the main
competitor to guilt among after-the-fact mora l emotions, would seem to be in
even a worse position in that respectand also to be ruled out by our assump-
tion that Russell acted honorably in the case. He would also have acted nobly ,
my claim is, if he had allowe d himself a twinge of guilt. Remorse might seem
initially to b e less objectionable than eithe r shame or guilt, but one of my cen-
tral aims in what follows is to undermin e the view of these emotions as neces-
sarily distinct.

2. Guil t as an Identificator y Mechanism

On a standard philosopher' s picture , emotiona l guilt i s contrasted wit h alter -
native reactions such as shame and remors e on the basis of the way its evalua-
tive content brings in the self and connects the self to its acts and to other agents.
On Rawls' s account, guil t is said to b e directed in the first instanc e toward a n
act viewed as wrong rather than some bad trait of the sel f a s with shame. 18 It s
content i s seen as essentially moral i n a way tha t i s not tru e of shame, which
may sometimes rest on nonmoral traits, including one's relation to the traits or
acts of other agents. However, for purposes of setting up a neat contrast i n this
discussion, I shall restrict myself t o a subtype of shame with the same sorts of
grounds a s guiltwhat might be called "moral shame. " Thu s limited , shame
seems to focus more closely on the self than does guilt insofar a s the self rathe r
than a n ac t i s its primary object of negative evaluation.
Like moral shame but apparently unlike remorse, though, guilt does repre-
sent th e sel f a s somehow affecte d (tainted , stained, tarnished, or th e like ) b y
the wrong act attributed to it. As a practical motivevia the desire for repara-
tion that seems to separate guilt and remorse from shameguilt may therefore
seem to be self-regarding in a way that remorse is not.19 Perhaps partly because
of this contrast, guilt-motivation is sometimes taken to fal l on the egoistic side
of the current debat e about egoisti c versus altruistic sources of moral motiva -
tion.20 The usual emotional alternative considered in the debate is love, whose
object is another person viewed positively; but eve n among negative contrary-
to-duty emotions there may seem to b e more altruistic possibilities.
Indeed, the point may be applied even to guilt versus shame, though in one
sense (as just noted) shame seems to involve a closer focus on the self than either
remorse o r guilt . A common contras t betwee n guil t and sham e links guilt t o
approaches t o ethics based o n a notion of individual responsibility for action,
or autonomy. Sham e is tied more closely to th e ethic s of virtuenot particu-
larly to the good opinion of others, as was maintained in earlier versions of the
contrast but rather to something more like status in a moral community. 21 Bot h
emotions res t o n th e internalizatio n of others' reactions , but sham e seems t o
involve a thought of being viewed by others with scorn, whereas guilt involves
accusatory anger.22 However, the two emotions are necessarily mixed together
Moral Residues 12 7

as a result of overlaps in childhood teaching: Guilt is often inculcated b y a form

of group rejection , for instance, to the extent tha t disapproval involve s at leas t
a qualifie d withdrawa l o f love.
For tha t matter , an y neat egoism/altruis m contras t i s further undermine d
by some othe r importan t way s in which guilt may seem to be tied more closel y
to regard fo r others than shame is. Its characteristic desir e for reparation ma y
be said to move the agent toward other s insofa r as it prompts hi m to make up
for an y har m don e them , i n contrast t o th e desir e to hid e from other s tha t i s
linked with shame. In fact, the connection t o egoism/altruism just described may
seem to b e reversed by Rawls's accoun t o f guilt versus shame i n terms o f th e
agent's concern fo r the welfare of others as opposed to his own stat e of mora l
perfection.23 Thoug h Rawl s sees shame as a higher mora l motive than guil t to
the extent that it prompts act s of supererogation, guil t comes out on his account
as a more directl y altruistic motive: an emotion buil t on concern fo r others .
Rawls's vie w seems to accor d wit h recen t researc h i n developmental psy -
chology linkin g guilt with empathetic distress.24 I n what follow s I want essen -
tially to sugges t a reconception o f the natur e o f guilt that incorporate s a ver-
sion o f this approach, but broadene d t o cove r recalcitran t cases i n light of an
older traditio n o f understanding guilt in terms o f anger. A version o f the latte r
that bring s in a form of empathy can be found in Jonathan Edwards' s accoun t
of "natural conscience." 25 Th e view I derive from Edwards's accoun t essentially
interprets guil t not as a distinct emotion o n a par with shame, remorse, an d the
like but a s a particular emotiona l mechanism , o r a pattern o f emotional reac -
tion, making use of these more basic reactions among others. I take this to allow
for overlap s betwee n guil t and sham e o f a sor t tha t wil l bea r o n th e disput e
among anthropologist s an d others abou t th e differenc e betwee n culture s that
rely on guilt versus shame as instruments of moral control. 26 My more general
aim in this section, though , is to tie guilt to the self in a way that explains bot h
its special force and its peculiar pitfalls as a motive for action on behalf of others .

Guilt and Empathy

The current tren d among psychologists attemptin g t o understand guil t in indi-
vidual development bases the emotion o n empathy, usuall y conceived a s iden-
tification with others' sufferings, o r empathetic sorrow or distress.27 W e might
be tempted t o se e this work a s providing empirical confirmation fo r the con -
nection Rawl s sets up between guilt and concer n fo r others, bu t i n fact i t rest s
on similar definitional assumptions that deserve some scrutiny. Since "guilt" is
essentially a made-up emotion ter m in the firs t place , arising in English origi-
nally by mistake, it would seem to be fair game for artificial line-drawing , even
more tha n the othe r element s in our disorderl y stoc k o f emotion terms. 28 Bu t
we d o a t thi s poin t hav e some fir m intuition s about it s use that ough t t o b e
respected. I n particular, a wider notion of empathy seems to be needed to cover
the cases of guilt for simple rule infraction withou t envisione d harm to other s
that w e ofte n tak e as prime examples of the emotion .
One migh t feel a t leas t somewhat guilty, for instance , about various viola-
128 Sensibility an d Standpoints

tions of the duty of fidelityfrom flatterin g someone to revealing a confidence

to misrepresenting one's political opinions for the sake of social harmonythat
do no damage to anyone affected b y them and in some cases even are performed
just in order to benefi t those affected . The sort o f reference to general harm
to the usual effects o f such behaviorthat we might be tempted t o make her e
would apparently limit the emotion to agents who accept a utilitarian justifica-
tion fo r th e rule s they feel guilt y about violating . But surely it i s at leas t con -
ceivable that someone might undergo the emotion without theoretica l commit -
ment on the subject, and also for acts that seem to do no harm to others even in
general terms suc h as violations o f certain rules of religiou s obedience. O n a
definition that builds in reference to harm, guilt about victimless sexual behav-
ior, say , would requir e either attributing to the agent a mistake about th e con-
sequences o f his act or extendin g "harm" to cover cases of offense, conceive d
as involving distress about the rule violation per se.
In any case, there is still a question whether empathetic distres s is sufficien t
to captur e th e evaluativ e content o f guiltwhat the emotio n say s abou t it s
objectwhere the act in question is reasonably seen as harming others. Some -
thing else seems to be needed, that is, to distinguish feeling guilty from just feeling
sorry fo r th e perso n on e ha s harmed , o r eve n sorrowful o r distresse d o n hi s
behalf, perhaps t o a degree that i s heightened by the recognitio n o f one's own
causal role but without any element of emotional self-reproach. Consider wha t
one might feel, say, about the other side's casualties in a war. Adding in a mere
belief tha t on e i s responsible or eve n blameworthy migh t o r migh t no t affec t
the conten t o f one's emotion, turning sorrow int o guilt . Even with a shif t o f
emotional focu s t o th e agent' s actio n (a s with Williams' s notio n o f "agent -
regret") we might just have a kind of self-directed sorrow or some othe r emo -
tion such as anguish that intuitivel y seems to fal l shor t of full-fledge d guilt .
If we add i n a further emotion , thoughself-ange r or som e othe r negative
self-directed emotioni t might seem unclear why we shoul d no t tak e thi s a s
the cor e elemen t of guilt, detachable from empath y an d sufficien t fo r guil t in
its own right in the cases of simple or victimless rule violation just cited. Empa-
thy could still be seen as playing a critical role in the development of guilt because
it serves to exten d the emotio n beyon d suc h cases, even if we d o no t mak e it
out as part of the very concept o f guilt, let alone the essential part, or as neces-
sarily present i n the earliest cases in childhood emotiona l experience .
However, I think that the notion o f guilt as based on self-anger may in fact
allow fo r a broade r accoun t o f th e emotio n i n term s o f empathy . Conside r
Jonathan Edwards's use of the elements of empathetic anger along with a notion
of emotional consistenc y to mak e sense of conscience:
[W]hen a man' s conscienc e disapprove s of hi s treatmen t o f hi s neighbour , i n
the firs t place he is conscious, that if he were in his neighbour's stead, he should
resent such treatment from a sense of justice, or fro m a sense of uniformity an d
equality between such treatment and resentment, and punishment. . . . And then
in the nex t place, he perceives that therefor e h e is not consisten t with himself ,
in doing what h e himself shoul d resent in that case; an d henc e disapproves it,
as bein g naturally avers e to oppositio n to himself. 29
Moral Residues 12 9

At the initia l stage o f Edwards's two-stag e account, th e agen t identifie s with

the resentment of the perso n h e harms o r otherwis e treat s unjustly . I t is clear
from a preceding passage tha t Edwards means to say here that the agent actu -
ally puts himself imaginatively in the other person's shoe s in understanding his
resentment.30 We can think of an agent at this first stage, thenresolving some
ambiguities in Edwards's accounta s reproaching himself from the standpoint
of the victim of a moral wrong he has committed. Th e second stag e essentially
combines this element of empathy wit h the agent's sens e of his ow n contrar y
motives as the basis for a second-order emotiona l reaction of "aversion t o self -
inconsistence and opposition " tha t Edward s equate s with uneas y conscience
and takes to b e at odds with self-love. 31 We may simplify thi s fo r our presen t
purposes to the agent's uncomfortable awareness that his first-order empathetic
emotion is self-directed and negative.
Emotional guilt of the sort in question here seems to belong at the first stag e
of Edwards's account, as the basic empathetic ingredient of an uneasy conscience,
involving self-opposition bu t not yet the second stag e of emotional reactio n t o
the awareness of self-opposition. A s Edwards describes it, this stage seem s t o
involve a form of anger (as I take resentment to be) based on the agent's imagi-
native participation i n the standpoint of those affected b y his action. The account
can be stretched to cover psychologists' account s of guilt as based o n empathy
in the narrower sensethe sense suggested by Rawls's talk of concern for other s
if w e grant tha t ange r ma y b e projected ont o the person s affecte d b y one's
act. Tha t is , empathy i n the narrowe r sens e ma y sometime s com e t o includ e
anger b y considering wha t on e migh t fee l i f one were i n someone else' s posi -
tion, even if the other person i s not perceived as angry on his own behalf . This
rests on understanding empathy as emotional identificationwhich I interpret
to cover any emotions base d on sharing the evaluativ e standpoint o f anothe r
person, a s distinct from th e tendency to take a particular person a s an overall
model for one's own experienc e that psychologists sometimes hav e in mind by
the term . We ma y spea k accordingl y of guilt as an identificator y mechanism
with a different particula r basi s in empathy in different cases , some involving
self-anger added onto a more sorrowful (or other negative but passive) view of
things from another' s standpoin t an d some involvin g other self-directe d nega-
tive emotions .
This view also seems capable of covering cases of guilt for victimless wrongs,
if we allow for an independent source of self-anger or some othe r self-directed
negative emotion that is empathetic in a broader sense, since it is based on iden-
tification wit h authorit y figure s or wit h object s of dependenc y i n earl y life .
Empathy o n m y speculative account o f guilt and "ought " i n chapte r 3 is en-
couraged in children essentially by linking emotional imitation wit h the condi-
tions of social acceptance or acceptance in the family group. Disapproval o f a
forbidden ac t ma y b e communicated a t a stage befor e th e resentmen t of an y
victims of harm is in question, with the rol e of the victi m as a source o f anger
assumed instea d by something on the orde r of a judge. My propose d accoun t
of guilt is not mean t to settl e the ultimat e origin of the emotion, however ; nor
would I want to exclud e any possibilities by conceptual line-drawing. Rather,
130 Sensibility an d Standpoints

my suggestion that we understand guil t as an identificatory mechanism is sup-

posed t o allo w for emotional variation s as long as they all exhibit an elemen t
of Edwards' s "self-inconsistenc e an d opposition"wha t w e ma y thin k of a s
"self-alienation," borrowin g a term from Gabriele Taylor.32 That is, the guilty
agent is assumed to b e emotionally at odds with himself as a result of the kind
of identification with others that we find in the first stage of Edwards's account .
This i s the basi s for th e aversiv e reaction t o th e awarenes s o f self-oppositio n
that yield s conscience in Edwards's secon d stage .
In the typical sort of adult moral case, guilt rests on identificatory self-anger.
I interpret this broadly to involve discomfort at the thought of oneself as respon-
sible for a wrong (wit h the latte r ter m take n to includ e both unjustifie d har m
and simple or victimless rule-violation), typically as a basis for a reparative urge
that provides a parallel to the desire for revenge characteristic of other-directed
anger. Thi s urg e may b e absen t i n some case s tha t w e stil l coun t a s case s o f
guiltwhat I call "deficient" (a s opposed t o "full-blown" ) cases. 33 It s presence
in typical cases, however, yields a further, forward-looking sense in which guilt
is action-oriented i n contrast t o shame , beside s being based on the evaluation
of a past action and of the self only in relation to that action rathe r than as the
bearer of a certain trait .
In other cases, thoughincluding many of the sorts of cases psychoanalyst s
handleguilt may involve some other self-alienate d emotion, something mor e
passive though still negative and self-directed , of the sort that would result from
identification wit h a victim seen a s reacting with a passive sense of grievance
rather than active blame. Other forms of emotional self-reproach include self-
hatred (i n the sense of self-aversion) or eve n self-horror (at one's role i n caus-
ing the other's suffering, say) . What i s important i s that th e reaction involves
the kin d o f emotiona l self-punishmen t o n behal f o f other s tha t Edwards' s
account captures in terms of the sense of justice. This is to be distinguished from
internalized fear o f punishment by others, a s on Freud's account of the origin s
of the superego , thoug h i n fac t m y account seem s to fi t Freud's comment s o n
remorse.34 It s central featur e fo r ou r purpose s i s that i t make s guil t ou t a s
including remorse, alon g with othe r negativ e emotions, wher e they are used as
ways o f inflictin g punishmen t o n onesel f a s par t o f th e genera l sor t o f
identificatory mechanism that we find i n Edwards.
This accoun t seem s to m e to b e the one that bes t explains the overlaps we
find i n our intuition s about cases . I t is not jus t that emotion s lik e shame, guilt,
and remors e have indefinite boundarie s o r that our idea s of them are obscur e
or ill-defined . In my own experienc e of testing out th e standard distinction s in
application to imagine d cases, i t ofte n seem s clear tha t bot h guil t and sham e
are in play: I find myself more or less wincing and hanging my head at the same
time, say . Nor doe s it seem to b e enough t o sa y that guil t and sham e or guilt
and remors e coincid e i n application t o cases , a s they would i f they both inde-
pendently applied; rather, ther e ar e a t leas t some case s i n which the y d o no t
really see m to b e distinct. All three may b e part o f the sam e emotiona l reac -
tionor, mor e precisely, shame an d remors e ma y b e part o f guiltin a cas e
where an appropriately sensitive person with certain moral expectations of him-
Moral Residues 13 1

self find s tha t h e has accused someone unjustly , fo r instance . O n th e accoun t

suggested here shame and remorse , or self-alienated versions of them, ma y be
part o f the ver y punishmen t that guil t inflicts , i n contrast t o th e usua l treat-
ment of these alternative emotions as either interchangeable or quite distinct.
My account als o applies to forms of guilt that psychoanalyst s have distin-
guished as "survivor('s) guilt" or "separation guilt" and that our culture seems
to have in common with others that mainly impose shame for moral breaches. 35
Guilt at some form of benefit ove r or separatio n fro m th e othe r members of a
group with which one identifies may serve as a mechanism of group control in
a way that includes moral control, bu t that also seems to represent at least one
strand o f development of the notio n that i s not base d o n ideas of sin or pollu-
tion.36 The biblical sources of our own culture's emphasis on guilt actually deal
explicitly only with sin; to locate emotional guilt in the Bible 37 we seem to need
to apply the general sort of identificatory mechanism that we find in Edwards's
account to the relationship between man and his divine judge. What these dif-
ferent case s have in common seem s to b e an inner conflict that ma y be multi-
plied b y the multiplicity of objects of identificationfrom othe r people to th e
group a s a whole t o Goda s reason s fo r self-alienatio n that ar e themselves
capable of conflict and henc e can generate dilemmas. 38
I shall deal in my next chapter with the questions raised by dilemmas as cases
of unavoidable wrong. For the moment, I should note that the account of guilt
I have outlined here is not put forth as a definition in the sense of an analysis of
the content o f the emotion. W e may retain the standard definitio n in terms of
self-attributed responsibility for a wrong. By explaining guilt as an identificator y
mechanism, I am in effect addin g to this a certain view of how discomfort comes
to be directed toward tha t thought, thus constituting guilt as an emotion. What
I have picked ou t a s essentia l to th e developmen t o f th e feelin g i s the sor t o f
self-opposition tha t Edward s interprets in terms of identificatory resentment .
Typically this arises from som e form of anger, though i t also may be based on
a range of other emotions serving to distance the self from its act in response to
grounds for anger. In any case, it is phenomenologically distinct from standar d
self-anger t o the extent tha t i t is identificatory and henc e self-alienated.39
Even allowing for the phenomenological rang e of guilt feelings, I think we
can see that mere acceptance of others' anger without some identificatory emo-
tion would no t b e sufficient. Conside r th e stoc k exampl e of a wayward hus -
band who accept s his wife's anger a s valid from he r standpoint, whic h i s one
he has to put up with (even meekly, perhaps) for his own ends, though it is not
a standpoin t h e share s i n emotiona l terms . I think w e hav e to sa y tha t hi s
acquiescence in her anger as a punishment imposed from withouteven if reg-
istered in some sort of counterpart emotionfall s short o f felt guilt, though he
may of course hav e a reason t o obscur e th e distinction . Fel t guilt seems to be
more likely, in fact, in cases where it is clear that others will not impose punish-
ment from without b y feeling anger .
More generally, the account of guilt as an identificatory mechanism will help
us distinguish what intuitivel y seem to b e genuine cases o f the emotio n fro m
other possible "fixes" o n the thought of responsibility for wrong: satisfaction,
132 Sensibility an d Standpoints

say, at being the focus of others' angry attention. B y itself, a definition of guilt
would limit the relevant cases to uncomfortable emotions but still would allow
for emotions other than those we intuitively classify a s guilt: simple resentment
on one's own behalf, most notably; or self-pity at social rejection (perhaps with
a content that acknowledges its justificationsomething on the order of "I can't
do anythin g right.") I shall also go on t o sho w ho w th e accoun t draw n fro m
Edwards lets us explain guilt-motivationboth its advantages and its pitfalls
in a way that depends o n its basis in self-alienation .
A Puritan divine may seem an unpromising choice to appea l to , eve n with
modifications a s indicated , i n attemptin g t o convinc e contemporar y mora l
philosophers o f the valu e of guilt-motivation. One reaso n fo r th e aversio n t o
talk o f guilt , alon g wit h objection s to th e Freudia n account , i s of cours e it s
religious overtones: I n its more overwhelmin g forms the emotion i s associated
with a particularl y self-punitiv e extreme o f Purita n religious consciousness .
However, it should be clear from the passage quoted earlier that Edwards's treat-
ment o f guilt is philosophical rathe r tha n religious . In fact, Edwards reverses
the usual link to religion by considering conscience as a "natural" motive , mean-
ing pre-religious: a motive that is present i n us before our soul s ar e saved. As
such, of course, it is not the highest motive on his account; the motive required
for Edwards' s notion o f "true virtue" i s benevolence or lovejus t wha t som e
of the current opponents o f guilt-motivation have in mind. Moreover, Edwards' s
(and th e common ) characterizatio n o f ba d conscienc e a s "uneasy " stand s i n
contrast to th e familia r picture o f overwhelming guilt that w e fin d i n certain
depictions o f Puritan religious consciousnessor, fo r tha t matter , i n psycho -
analytic case histories or the modern novel . Guilt on the picture Edwards offer s
us is or ca n b e a moderate emotiona l reaction an d seem s perfectly sensibl e in
affective term s a s a wa y o f registerin g self-attribute d wrong. Le t us no w se e
what els e might b e said i n its favor.

In Praise of Guilt
On the identificatory account jus t presented, guilt is sometimes a passive feel -
ing, as I think we have to sa y in light of the ful l rang e o f cases. Bu t in the cen-
tral moral cases, it s tie to anger makes it activeif not affectively , i n the sense
of being aroused i n feeling tone, then at any rate evaluatively, to the extent that
its evaluative component typicall y supports a n urge toward reparativ e action :
discomfort a t the thought tha t one ought to make up for a wrong. Thi s means
that guil t exhibit s motivationa l forc e t o a degre e tha t i s not tru e o f shame .
Shamemeaning occurren t emotiona l shame , no t jus t th e genera l sens e of
shame, or a disposition t o exhibit the emotionis an inhibited or "downcast"
feeling i n phenomenological terms. S o cases of motivation b y shame turn ou t
for th e mos t par t to b e cases o f motivation t o avoid shame , base d o n (i f any
emotion) fear or even pride. By contrast, though guilt may be incapacitating in
excessive doses, th e agen t i n a state of feelin g guilt y i s typically motivated b y
that stat e to escape itin a way that makes guilt provide a potentially power-
ful mora l motiv e even afte r a moral lapse. It has a special role to play , then
Moral Residues 13 3

which is not t o say that it leaves no room for shamein a morality buil t on a
view of human nature a s imperfect but improvable .
For a nonmoral example , on e might consider wha t on e would b e likely to
do ou t o f guilt versus shame i n response t o a ba d piec e o f wor k on e ha d t o
publish: On the standard picture, someone motivated by guilt would get to work
on something else to make up for wasting his talents or his colleagues' tim e on
the las t effort ; someon e motivate d b y sham e woul d typicall y try t o hid e o r
obscure th e object of his emotion. H e might also tr y to mak e u p fo r it , bu t it
would no t see m t o matte r t o th e effectivenes s o f suc h compensator y actio n
whether it is connected to the object of shame in a way that would mak e it count
as "reparative"as repairing what was broken by the violation of professiona l
standards. Tha t is , the agen t wh o i s ashamed o f his past performanc e migh t
make up for it, not by improved performance in the same area, but just by stress-
ing other area s in which h e already shines.
Guilt in contrast t o shame , w e might say, i s intractable to summing . On e
might indee d attemp t t o wor k i t off by performance in other areas , bu t guilt
for a serious moral wrong wil l be expiatedif at allonly by action (including
mental action) that addresse s th e wrong done . In sublimated form, then, with
the thought o f wrong hidden from consciousness, i t yields a particularly pow -
erful motiv e because it is unappeasable. For much the same reason, it may some-
times be overwhelming. There ar e cases of wrong s o serious that guilt is inex-
piable: Agamemnon's sacrifice of his daughter and many of the other standar d
cases of moral dilemmas would seem to qualify. On the other side, though, psy-
chological studies indicate that shame is particularly incapacitating i n extreme
form, at least as induced in children.40 Someone overwhelmed by either shame
or guilt , of course, wil l not b e motivated to d o muc h at alla t an y rate , will
not b e effectively motivatedexcep t t o th e exten t tha t undergoin g th e emo -
tion itsel f count s as a kind of partial or substitute action. Sham e involves low-
ering oneself , or takin g th e submissiv e posture, w e ma y sayhangin g one' s
headwhereas guilt involves a kind of emotional self-punishment.
The structure of guilt-motivation is worth explorin g further in these terms ,
since the emotio n itsel f count s a s a for m of the punishmen t it demands. On e
way o f making up fo r a seriou s wrong, tha t is , involves anticipating others '
reactions b y inflicting punishmen t on oneself . So at leas t t o som e extent , th e
desire for reparation characteristic of guilt may be satisfied just by undergoing
that unpleasant self-directed emotion. This gives the emotion a special motiva-
tional role as a kind of ritual act of emotional self-punishment. Psychoanalysts
see self-punishment of various sorts (self-destructiv e behavior , most notably )
as a defense against guilt in certain cases.41 But the account suggested here builds
this function into guilt itself b y way o f its self-referential quality .
The point ha s complex motivationa l possibilities that seem to pul l in bot h
directionsreinforcing guil t as a motive but als o explainin g some o f the pit -
falls o f guilt-motivation. Though guil t i s unpleasant, it seem s t o b e on e o f a
number of unpleasant emotions that we characteristically "wallow" in, with a
need to dwell on their negative thought content. In anticipatory form, then, as
an attemp t to hea d off further o r wors e guilt b y self-punishment, i t may ver y
134 Sensibility an d Standpoints

likely fail . I t often has the opposit e result, i n fact"feeds" o n itsel f in a wa y

that can compound it s motivational force but also constitutes a powerful motive
for avoidin g the emotion jus t by shifting attention. Whereas sham e incapaci -
tates by inhibition, the arousal ofte n linked to guilt may involve hounding one-
self t o a degree tha t also interferes with action . Ther e ar e cases i n which on e
avoids even entertaining the thought o f actionof writin g a letter already to o
long postponed, sayout of an inability to face the emotion. O n my account,
we may say that a tendency toward compulsive repetition of ritual self-punish-
ment sometime s undermines its motivational effectiveness.
This complex accoun t o f guilt-motivation with its basis in talk of emotiona l
self-punishment gives rise to questions o f voluntariness that I shall postpone to
my next section . Fo r the remainder of this section, I want t o exhibit som e fur -
ther pitfalls of guilt as a motive, along with some of its advantages in compari -
son to shame , remorse , an d th e various othe r reaction s t o mora l wron g tha t
might b e substitute d fo r it . Th e centra l poin t tha t distinguishe s guil t fro m
shameto focus first on its main competitorseems t o be self-alienation. Shame
does no t see m to divide the sel f agains t itsel f in the way that guil t does o n th e
account draw n fro m Edwards . Thoug h i t may be seen as based on the expecta -
tion of contempt o r the like from othersperhaps onl y hypothetical others (o r
even jus t one' s hypothetica l idea l self ) i n th e cas e o f a n actio n hidde n fro m
viewit does not amoun t t o self-contempt. Th e agen t doe s no t identif y with
his judges, that is , in feeling what he feels; he is solidly in the inferior position
of th e perso n judgedno t at al l a comfortabl e positio n bu t on e that at leas t
assigns him a definite place . If he did identif y wit h hi s judges, to th e exten t o f
feeling things (at least ambivalently) from their standpoint, hi s emotion woul d
also com e ou t a s part of the guil t response, o n th e accoun t suggeste d her e of
guilt a s a n identificator y mechanism tha t sometime s use s othe r emotion s a s
With attention confined to cases in which the two emotions are distinct, then ,
there seems to b e one way in which guil t may be worse than sham e in motiva -
tional terms: Both emotions obviously interfere with self-esteem, but guilt may
be said to do so in a "conflicted" wa y that is arguably more incapacitating tha n
simple inhibition . However, w e hav e begu n to se e how thi s element o f inne r
conflict a s the basis for the self-referential qualit y of the emotion als o makes it
potentially more powerfu l as a motive. Guil t may play an irreplaceable mora l
role, in short, to the extent that it yields a self-perpetuating motivational mecha -
On mos t o f the other comparison s I can think o f between guilt and shame ,
we seem to fac e a choice between differen t sort s o f advantages an d disadvan -
tages. Sometime s i t is a value-laden choice: Fo r instance , it might be said tha t
guilt at least accords th e sel f a kind of dignity insofar as it is identified wit h it s
judges rather tha n simpl y being confined to the position o f the person judged .
The functio n o f th e sens e o f sham e seem s t o includ e keeping people i n thei r
placeit amount s t o th e knowledg e o f one's placei n a wa y tha t link s the
emotion to social stratification and group conformity. Similarly, moral sham e
may sometime s b e incapacitating to th e exten t tha t i t favor s acknowledgin g
Moral Residues 13 5

imperfections as characterizing oneself i n global term s rather tha n just in con-

nection wit h a past action .
Moral guilt is also commonly said to be tied to individualism because of its
reference to specific acts as falling within the agent's sphere of responsibility. 42
I raise questions in chapter 5 that undercut thi s point somewhatit introduce s
further issue s of voluntariness of the sor t that the presen t discussio n i s meant
to bypassbut my account stil l yields at least a rough contrast with shame on
the question of individual responsibility. Still , to the extent tha t the guilty self
identifies wit h th e victim s of the wron g i t does, as well as with it s judges, its
responsibility ties it to others; and th e connection i s reinforced, o f course, fo r
nonmoral variants of guilt such as survivor's guil t that are imposed fo r benefits
distinguishing an individua l from other members of a group. A s a reaction t o
someone else' s plight o r disadvantages , guil t is owed t o other s i n a way tha t
shame i s not.
In differen t ways , though , bot h guil t an d sham e involv e negative evalua-
tions focuse d more o r les s explicitly on th e self . I take i t that thi s is what set s
them apar t fro m alternativ e reactions t o moral wrong . Variants o f regret, fo r
instance, eve n i f directed towar d one' s ow n pas t acts , o n explicitl y mora l
grounds, d o no t involv e any kin d o f threa t t o th e selft o the estimatio n o f
character o r meriton the basi s of past acts , but on the contrary see m to use
negative feeling simply to distance the self from its past act s without an y inter-
vening stage o f self-punishment. In affectiv e terms , regre t ma y b e unpleasant,
but i t lacks the elemen t of negative self-evaluation that support s m y motiva-
tional accoun t of guilt.
The same might also be said of remorse, however , even though this involves
a wish to undo the wrong on e has done (a s guilt may not) alon g with the need
to make up for it by future action. In cases where it is distinct from guilt, remorse
is more tightly tied to a specific past action: Its associated reparativ e desire does
not amoun t to th e sort of general need t o clea r th e selft o expiate a wrong,
thereby erasing a "taint"that account s fo r the motivational power o f moral
guilt i n anticipator y form . Th e possibilit y o f "free-floating " guilt , thoug h a
source o f some devastating psychological problems , a t the same time gives the
emotion a practical scope tha t doe s not characteriz e it s alternatives.
Thus, in my earlier example of publishing a bad piece of work, neither regret
nor remors e woul d see m to motivat e late r wor k i n the wa y that we typically
see guilt as doing. The same seems to hol d fo r the various other after-the-fac t
"taboo-feelings" (a s I think of them) that might be suggested as alternative emo-
tional residues of moral ought-violation: horror, say; or a kind of moral anguish
that amounts to a variant of grief. On th e other hand , a feeling o f "compunc -
tion," which is sometimes mentioned a s a less problematic alternativ e to guilt
in case s o f moral dilemma , would no t reall y seem t o b e after-the-facto r at
any rate, it would not admit of the same after-the-fact temporal distancean d
thus i s naturally interpreted by Hare a s a variant of fear. 43
I conclude, then, that guilt is the best candidate among these negative emo-
tional reactions for supplying the motivational force of moral "ought." It can
also cause motivational problems insofar as it sometimes involves an obsessive
136 Sensibility an d Standpoints

inability to "let go"of act s one cannot make up for, relationships one cannot
improve, or tendencies one cannot brin g under controlbut i t seems to share
at least some of these problems with moral obligation. I t therefore seems well
suited to help explain the sense of compulsion we associate with "ought." What
guilt ha s ove r othe r emotiona l candidate s o n th e accoun t I hav e give n is a
potential future-orientation mediated by its identificatory focu s on th e self .
For the same reason, of course, we might question its application to Russell's
case and other moral dilemmas. There seems to be no need for the agent in such
a cas e to mak e u p fo r th e wrong h e has don e b y hi s future behavio r in other
areas of life beyond doing what he can to mitigate the effect s o f that particular
action, as he would if he felt only remorse. Presumably the wrong in question is
something he should have done overall. Why not say , then, that Russell ought
to hav e felt remors e a t jiltin g th e woma n bu t no t th e sor t o f negativ e self -
evaluation involve d in emotional guilt ? M y answe r t o thi s questio n will ulti-
mately depend on a fuller accoun t o f the standard s o f emotional appropriate-
ness. Befor e w e get to tha t issue , however, w e nee d t o as k wha t i s meant b y
saying that a n agent in a dilemma ought to react with one emotion o r another.
My view , in brief , i s that th e agent ought to fee l somethin g like guilt an d tha t
guilt in the narrow sense just distinguished is appropriate; bu t it remains a ques-
tion whether th e agent ought specificall y to fee l guil t in that sense . Let us now
look a t some of the more general problems raised by "ought-to-feel. "

3. Contrary-to-Dut y "Ought-to-Feel "

Both my account of guilt and th e role I assign to i t in cases of dilemma rest on

a view of emotions a s able to serve as substitutes for action. I speak of them in
some cases as self-imposed and as indirect ways of satisfying obligations to take
actionand for that matter as objects of obligation in their own right. Because
of both their motivational force and their independent symbolic significance as
rewards and punishments for action, they may sometimes be required to express
evaluative standpoint s o r commitment s tha t a n agen t canno t expres s mor e
directly, through action , unde r th e circumstances . B y contrast, th e recen t re-
surgence of interest in emotions i n the ethics literature seems for the most par t
to limit their role to virtue ethics. They serv e alongside action o n an Aristote-
lian accoun t a s objects of long-term trainin g in virtue, commendable both in
their own right and in support o f dispositions to virtuous action but not in any
stronger sense required. 44
The exclusion of emotions from dut y ethics is typically traced back to Kant;
Ross, fo r instance , i n extending on e o f Prichard' s arguments , appeal s t o th e
Kantian principle that "ought" implies "can" a s a reason for denying that action
from a particular motive (in a sense that includes virtuous feelings) can be obliga-
tory.45 Hi s argument , alon g with Prichard's , apparentl y drive s a hard wedg e
between dut y and virtu e ethics . In fact , though , elsewher e in the sam e wor k
Ross provides the reade r with sufficien t material s for a respons e to th e argu -
ment. I shall begin my treatment of "ought-to-feel" by constructing a response
Moral Residues 13 7

from Ross' s comments a s supplemented b y my own vie w of emotion. What I

hope to show is that emotions have an important role to play, albeit a comple x
role, as objects of obligation alongside action in the general approach t o ethics
derived from Kant, despite any excesses of Kant's own versio n of it.
My response t o Ross's argument will turn on the way we rely on voluntary
action to generate emotion i n standard cases. It will eventually be put to wor k
in defense of an interpretation of dilemmatic ought-judgments as guiding action
indirectly insofar as they imply various contrary-to-duty obligations, including
those prescribing guilt and simila r moral emotions. This view involves assign-
ing to action-guiding "ought" a kind of conditional imperatival force in appli-
cation to feeling, in addition to its literal or direct force as commanding action.
However, some care is needed in interpreting this view. For one thing, a major
pitfall o f guilt-motivation not previousl y discussed seems to b e a tendency t o
substitute the emotion fo r right action. "I' m such a miserable sinner" as a way
of fendin g of f reasonable change can b e sincerely fel t a s a more o r les s auto -
matic response to well-rehearsed cues, but this is obviously not what one has in
mind in requiring guilt. The requirement to fee l guilt y if one cannot manage to
act must not b e taken a s implying a permission to omi t actio n an d fee l guilty
Nor shoul d "ought-to-feel " b e taken as resolving dilemmas. It is important
in any case to distinguish the practical an d normativ e ethical problems raise d
by dilemma s from th e metaethica l difficulties t o whic h m y argumen t her e is
addressed. A wrong plus guilt will not mak e a right on an y view I would fin d
plausible in answer t o thos e difficulties . Bu t further, "ought-to-feel" wil l no t
always yield a way out of them. In some cases, like the one drawn from Russell's
autobiography, possible conflicts between required feeling and action might give
rise to second-order dilemmas . If the normative point o f the required feeling is
to protect virtue from the effects o f moral luck, my discussion will indicate that
this role is also subjec t to important limitations .

Requiring Guilt Feelings

In th e passage s tha t bea r o n "ought-to-feel, " Prichar d an d Ros s ar e mainl y
concerned to deny the possibility of being obligated to act from a certain desire
or motiveas opposed t o simply possessing it, or exhibiting it in some way or
other, which i s what concern s u s here. The y d o so , moreover , partl y o n th e
grounds that requirin g act-motive compounds rathe r tha n simpl y acts woul d
involve a regres s i f applied t o th e sens e of obligation, th e desir e to d o some -
thing because it is required, or Kantian conscientiousness.46 However, Ross also
brings in an argument from "ought"-implies-"can " tha t woul d see m to coun t
against requiring a motive in itself, and Prichard's argument apparently come s
up in defense o f a general limitation of "ought" to "ought-to-do, " taking it as
applicable to "action s an d actions alone." 47
I want to loo k closel y at Ross's us e of the Kantian principle. He writes:
It is not th e case that I can b y choice produce a certain motive (whether this be
an ordinar y desir e o r th e sens e o f obligation ) i n mysel f a t a moment' s notice,
138 Sensibility an d Standpoints

still less that I can a t a moment's notic e mak e it effective i n stimulating m e t o

act. I can ac t fro m a certain motiv e only if I have the motive; if not th e mos t I
can d o i s to cultivat e it by suitably directing my attention o r b y acting i n cer-
tain appropriate ways so that o n some future occasio n it will be present i n me,
and I shall be able to act fro m it . My present duty, therefore, cannot b e to ac t
here and now from it. 48
Our questio n (t o modify Ross' s last line here) is whether it can b e someone' s
present duty to have or display a certain motiveand even to display it "her e
and now" i n some wide enough sense of that expression to accommodate act s
of the norma l sort, a s the norma l objects of duty. I shall argu e that i t can be .
Though I shall not take the point further here , moreover, I think we could eas-
ily extend i t to yield the claim Ross explicitly wants to deny: that one can have
a duty to act from th e motive in question. We just cannot tak e this to rule out
a duty to perfor m the ac t i n itself, or regardless of motive.
The object of the duty to act from a certain motive, that is, should be under-
stood as being moved b y that desire , emotion, or whatever it may be, with the
duty t o perfor m th e ac t i n questio n see n a s independently fulfillable, no t a s
replaced b y the dut y t o displa y a certai n act/motiv e compound, a s o n th e
Prichard/ Ross interpretation o f Kantian conscientiousness. If acts alone could
not b e required, we might jus t avoi d th e worrie s thes e author s rais e b y sub-
stituting motivational prohibitions fo r requirements. A prohibition o f all less-
than-virtuous motives woul d o n this account b e taken a s ruling out an y act s
the agent would do out o f those motives, even if that means ruling out al l pos -
sibilities for performing those acts. O n the supposition that a t least som e act s
are required simpliciter, howeveri f only acts of omission like refraining from
murderwe need a more plausible response, i n positive terms, to Ross's argu-
ments agains t obligatory motives.
The relevant sort of motive for our purposes is not a long-term character trai t
on the model of the sense of duty but a short-term emotional state, including certain
motivating states o f fel t desire . These ar e clearl y covered b y Ross's argument ,
despite its focus on th e sorts o f motives Kant favored. However, th e argumen t
apparently limits our control ove r such states to the long-term strategies fo r cul-
tivating them in ourselves that come up in relation to the Aristotelian picture of
training in virtue. By means of acts of attention or overt acts that we can control,
we can manag e to produc e a certain feeling o r othe r motiv e in ourselvesbut
perhaps not "a t a moment's notice, " o r with the kind of immediate contro l tha t
Ross in the quoted passage takes "ought"-implies-"can" t o imply.
Ross's denial that w e can produc e a motive "b y choice" suggests tha t h e
may als o b e interpreting th e principl e a s insisting on direc t contro l o f a sor t
that is unlikely: On the Aristotelian picture, one becomes conscientious o r cou-
rageous, say , no t simpl y b y aiming toward bein g so but instea d b y directing
attention towar d somethin g else , such as the actions thes e motives requir e on
particular occasions. However , w e might begin by raising some questions abou t
this interpretation of "ought"-implies-"can" tha t will also bear on the issu e of
temporal immediac y and o n probable further backgroun d assumptions of the
Kantian principle.
Moral Residues 13 9

Are acts required only if they can b e accomplished directl y by choice? Pre-
sumably they must depend only on choices one can manage, bu t must these be
choices to perform the acts in question unde r the descriptions under which they
are required ? I do not se e why w e should say so, particularly if we remembe r
some o f th e difference s betwee n "ought " an d "obligation " tha t emerge d i n
chapter 2 and concentrate o n "ought," the term for which closure over causal
relations holds . There seem to b e cases in which our require d ends have to b e
promoted indirectly , that is , by means tha t entai l shiftin g attentio n t o some -
thing else. For instance, perhaps I ought t o hel p a frien d wh o wil l b e able t o
benefit from what I do for him (s o that I will indeed be helping him) only if I do
not se e myself a s helping him bu t rathe r a s participating in a joint project fo r
its own sake . We might suppose tha t h e needs help because of a loss in confi-
dence that would just be made worse by too much overt concern from others
and tha t if I thought of myself as helping him, my concern woul d be evident in
what I did.
It ma y b e important t o thi s sor t o f case tha t ther e i s something I can d o
directlynamely, whatever overt acts are required to help my friend. But even
supposing that this exhausts my strict obligations in the caseif indeed the term
"obligation" applie s to it strictlythe claim that I ought to help him seems ac-
ceptable. Simila r cases might be constructed fo r various activitie s that require
intellectual absorption, sometime s to th e point o f blotting ou t ful l conscious -
ness of one's motives. Indeed, a similar point is often made about the pursuit of
happiness, a s a backgroun d requirement of either morality or prudence . An d
there are examples of specific behavior required by morality or prudenceoften
described by reference to motives or other states of mindthat would ten d t o
be undermined by direct aim. Consider th e command t o sho w respec t say , o r
the advice to be assertive, and imagine how likely one would be to satisfy eithe r
of them adequately by just trying to.
So intuitive cases do not bear out an interpretation o f the Kantian principle
as limiting oughts to acts one can perform at will, if that means by willing those
acts as suchone way of taking "by choice" i n Ross's argument . What about
"at a moment's notice" ? There ar e certainly examples of overt acts that are fit
subjects of obligation and yet require preparationor tim e for completion. Some
of thes e involv e long-ter m cultivatio n o f an abilit y t o act , o f th e sor t tha t i s
suggested by the Aristotelian picture of training in virtue: Consider, fo r instance ,
my "time-bound " extensio n o f Sartre' s cas e i n chapter 2 , wher e th e agent' s
ability to discharge his obligation to support hi s mother depend s on ten years
of savings . Bu t there ar e othe r example s shorte r ter m tha n this , particularl y
where the act itsel f takes time.
Ross provides at leas t one good example , i n fact, when he turns to a n ex-
tended discussion of the nature of right acts, or the question how much is to be
included i n th e descriptio n o f a n objec t o f obligation. 49 O f a cas e i n whic h
returning a borrowed book can only be accomplished by mailing it, he concludes
that fulfillmen t o f the obligation to retur n it requires actually getting it to th e
other party, not jus t dropping it in the mail. 50 At the very least, if the mail s go
astray, one owes the other party an explanation of one's failur e to get the book
140 Sensibility an d Standpoints

to him . O f course, i n that case , the objec t o f obligation i s not somethin g tha t
can b e don e "a t a moment' s notice." On e ca n initiat e action a t a moment' s
notice, perhaps, bu t only in a sense that include s such preparatory menta l acts
as planning to bu y some wrappin g pape r an d th e like . In normal cases , how -
ever, the sam e can b e said o f the processe s tha t giv e rise to a certai n kin d of
motivationto an emotional reaction like guilt, sayif one allows for the point
about indirectnes s that came out just above .
Let u s take a loo k a t the questio n o f ho w w e generate emotional motiva -
tion. Fo r instance , consider ho w I might get myself to fee l guilt y about some -
thing I took to requir e the emotion . Suppos e I find mysel f indifferen t t o som e
harm I have done to Xor perhaps to some harm that has befallen X as a result
of my actions or in comparison t o some benefi t that comes to me. What might
I do to generate guilt? Note that w e do not hav e in mind here the general ten-
dency to fee l guilty , or proneness to guilt, even just with respect to X; perhaps
I hav e no deficienc y i n tha t regard , o r perhap s ther e ar e reason s agains t th e
general tendency. In any case , what w e want i s just a feeling o f guilt on som e
particular occasion wit h an object that is limited accordingly.
One thing I might do and might well be advised or advise myself to do is to
think abou t certain aspect s o f the situationto attend t o o r eve n to dwel l on
the harm done to X from X's standpoint, it s connection to my own action, and
the various things that connect me to X, suc h as benefits I have received from
him i n the past . Emotional discomfor t essentiall y serves as a way o f directing
or sustainin g attention towar d som e negativ e evaluation and arise s naturally
(albeit not inevitably ) in the act of fixing attentio n o n the thought i n question.
An attempt t o generate it more directlyjust by telling oneself to fee l guilty
would b e inefficaciou s o r eve n counterproductive, i n part becaus e i t direct s
attention to something else.
We should be careful t o cancel out an y suggestion tha t generating emotion
or emotion itself amounts to an act that is within the agent's controlany more
than i t is fully withi n the control o f the agen t i n Ross's case to ge t a boo k t o
someone throug h the mail. The indirec t methods fo r generating emotion ma y
be no more reliable than the mail. If the mail does go astray, of course, we might
withdraw the claim that th e agent really ought to hav e returned the book, o n
the grounds that "ought" implies "can"; but that does not mean that the claim
should b e withdrawn for all similar cases in which th e mail does not go astray
but might have. "Ought" holds on the assumption that "can" does ; but "can "
is adequately fulfilled by doing the act (or exhibiting the emotion or other motive)
in question. I t does no t impl y "by som e generall y reliable method"in othe r
words, that the agent possesses a general capacity to bring about th e object of
obligationany more than it implies "at a moment's notice. "
Our uncertai n control ove r our emotion s an d othe r motive s and state s of
mind may be indirect in a further way besides that indicated, because the exer-
cise of it typically runs in reverse temporal orde r from means-end s calculations,
as well as depending on th e substitutio n in attention o f another end . That is,
one ma y sometime s generate a required emotion b y aiming at performin g an
actincluding an act of attention, as in the example just givenwhich in fact ,
Moral Residues 14 1

if performed, would be motivated b y the emotion i n question. Though th e tem-

poral reversal pertains only to occurrent motivationby emotions and similar
episodic states rather than long-term dispositional motivesi t is illustrated by
one of Prichard's comments abou t courage . Prichar d allow s that on e ma y be
required to do something that involve s action fro m courage, a particular dan-
gerous act that one can perform only by working up some courage, compatibly
with hi s exclusio n o f a n obligatio n t o ac t fro m tha t motive. 51 However ,
Prichard's poin t depend s o n interpretin g courag e a s th e suppressio n o f a n
occurrent emotion , namely fear.
It i s worth takin g note i n general terms o f the sor t o f negative control o f
emotions to which this example appeals: our ability to block emotions , so that
we can also be said to allow the m to occur whe n we do no t bloc k them. But it
might be thought that guilt admits of this less than other emotions; in any case,
it should b e clear from m y previous discussion of the pitfall s o f guilt that I do
not mea n to den y that on e can b e overwhelmed b y guilt feelings. Indeed , i t is
possible (an d compatible with anythin g I want t o sa y about emotion s gener -
ally) tha t a degre e of "uncontrol " i s built into th e ver y concept o f emotions :
that what unites the states we think of as emotional is in part their resistance to
full control . More o r less by definition, that is , they are state s tha t w e canno t
just put ourselve s into either directly or b y entertaining certain thoughts . M y
claim here is that they are not therefor e disqualified as objects of obligation
of practical as opposed to ideal "ought"given the range of acts that have that
Emotions might be said to be midway between belief and action in the degree
of practical control they allow for. Belief is typically harder to control, an d our
claims abou t wha t belief s on e ough t t o hav e are presumabl y commendatory
for th e most part. But "ought-to-think" als o applies to occurrent acts of atten-
tionof fixin g a certain thought i n mindand through the m (i n the wa y just
indicated) to emotions. Acts of sustained attention ar e of course not full y con -
trollable either: One cannot simpl y command oneself to concentrate on some-
thing undistracted . More precisely , such a comman d ma y no t b e possible t o
satisfyor ma y in the next moment go unheeded. We do not therefore hesitate
to issue such commands, thoughto require someone to pay attentionin cases
that allo w fo r th e ordinar y sor t o f qualifie d control . Directin g and sustain -
ing attentio n seem s to coun t a s a long-ter m ac t o n th e mode l o f getting tha t
book bac k through the mail to its rightful owner . I t is similarly recalcitrant to
full control , despite the fact that i n this case the interferin g factors may al l be
Our eve n more qualifie d contro l ove r the emotional state s that aris e from
such acts of attention is sufficient reaso n no t t o coun t th e generatio n o f emo-
tion as an act: Though the concept o f action extends beyond intentional action
to behavio r that occurs more o r les s automatically (an example might be tap-
ping one's fingers while thinking), presumably this has to be behavior of a sort
that on e could produc e at will. My claim here is simply that, even i f emotions
are not lik e this, they are stil l controllable enough by way o f acts of attention
to b e suitable objects of practical "ought" in at leas t some cases.
142 Sensibility an d Standpoints

This does not mean we would issue such a practical ought i n conversation,
even mental conversation. Orderin g oneself or someone else to fee l guilty might
defeat it s point b y it s very directness, accordin g to th e argumen t I have out -
lined here; more than that, i t would seem to undermine the moral significance
of th e emotion. Guil t in the case s where it is required is supposed t o b e auto-
matic; only as an automatic response , unmediated by practical reasoning, can
it giv e u s th e kin d o f evidenc e of moral motivatio n that w e wan t i n cases of
dilemma, most notably . Russell, for instance, in the case previously discussed,
would not b e fully admirabl e in the terms we require if he had simpl y managed
to make himself feel guilty in response to the knowledge that guilt was required
by perfect virtue. So virtue imposes a kind of paradoxical limitatio n on practi -
cal "ought-to-feel. " Th e sam e sor t o f parado x i s abundantly illustrated by
"ought-to-think," though , in the sense that involves attention: Consider the self-
defeating result s in many cases of explicitly urging intellectual absorption o n
oneself a s a necessary means to satisfyin g th e requirements of one's studies or
one's job . I conclude tha t th e parallel problems for emotion d o no t limi t it t o
virtue ethics in the way that the arguments in Prichard and Ross seem to show.

Indirect Action-Guidance
It is important t o bea r i n mind that "ought-to-feel " i n a case lik e Russell's or
even Agamemnon's i s not quit e on the sam e level as the all-things-considered
practical oughts in conflict in a dilemma, despite my argument fo r taking it as
analogous to "ought-to-do. " I shall illustrate in a moment how, beside s being
conditional on the failure to satisf y on e of the dilemmatic oughts, the require-
ment to feel guilty may conflict with the other ought in certain cases. This means
that m y suggestion tha t emotion s serv e as act-substitutes in such cases i s not
put fort h a s a wa y o f resolvin g dilemmas. Emotion s d o no t provid e a full y
adequate alternative for satisfying th e ought the agent fails to ac t on but serve
instead t o exhibi t it s motivational force, along with othe r sign s of its impor-
tance to the agent, b y exhibiting the internal sanctions on moral failure .
My crucia l claim is that thi s i s enough t o mak e th e dilemmati c ough t i n
question adequately action-guiding for our purposes: It is meant to guide action
directly, though o f course it cannot d o so in combination with the other ough t
in conflict ; it may stil l guide action indirectly , however, t o th e exten t tha t i t
tells us as a second-best alternativ e to fee l som e appropriate emotio n suc h as
guilt. According to my preceding treatment of "ought-to-feel," thi s amounts t o
telling us to allow ourselves to fee l the emotionif necessary, to get ourselves
toby performing a mental act: directin g attention in such a way as to gener-
ate the feeling. In short, practical "ought-to-feel" amounts to a kind of indirect
"ought-to-do" with anothe r object .
On thi s account, emotions ca n b e seen as "residues " of moral choic e i n a
fairly litera l sense: They embody what is left o f action on a moral ought whe n
direct actio n i s blocked. 52 However , the y ar e themselve s subject t o ought -
conflicts. Indeed, I want now to consider what to say about cases in which they
Moral Residues 14 3

seem to allow for higher order dilemmas . Up to this point, for instance, I have
been workin g wit h a n understandin g o f th e cas e draw n fro m Aeschylus '
Agamemnon that take s fo r grante d som e assumption s inherited fro m th e lit -
erature on dilemmas. In particular, I have assumed that Agamemnon unprob -
lematically ought to fee l guilt or some similar moral emotion and that, insofar
as the cas e amount s to a dilemma , this would hav e been true eve n i f he ha d
chosen differently. However, I think there are problems lurking in any easy resort
to "ought-to-feel. "
For instance, now that we have distinguished some of the emotions that might
be urged on Agamemnon, a question might be raised as to whether the dilemma
Aeschylus depicts is genuinely moral. I take it that Agamemnon indeed ough t
to fee l guilt , at leas t prima facie , a s a self-directed variant of remorse for th e
murder o f his daughter. The crim e is serious enough to b e personally "taint -
ing," i n short, even if in a larger sense he had to commit it. But it is not so obvious
that Aeschylus represents the necessity in question here as moral or that he takes
some moral variant of agent-regret to be appropriate fo r Agamemnon's othe r
alternative, which involves failing in his duties as military commander (an d more
fundamentally as his brother's avenger).53 The conflict might also be interpreted
as a clash between moral and nonmora l requirements, with the latter taken as
imposed b y something like social or religiou s roles and expectationso r bet -
ter, perhaps, by an older ethical system now uneasily superseded by specifically
moral norms , i n a sense that entail s regard for the welfare of other agents .
The murder on this account o f things is the morally wrong act in question,
but Agamemnon' s alternativ e migh t still be thought o f as more importan t t o
avoidfor instance, on a view like Williams's that questions the finality of moral
norms.54 If this i s true, w e might want t o sa y that sham e or som e othe r con -
trary-to-duty emotio n rathe r tha n guilt in the narrow sens e would hav e been
appropriate had Agamemnon refused t o sacrifice his daughter. The case might
still be a "tragic" case to th e extent that it pits morality against another pow -
erful syste m of normspre-moral religiou s norms, saywit h no satisfactor y
resolution possible. But it would n o longer seem to be a moral dilemma in our
terms. I shall therefore ignore these problems of interpretation i n what follows
and continue to work with what we might call "the philosopher's Agamemnon" :
the received picture of Agamemnon as an agent in the grips of a moral dilemma ,
with bot h o f the choices h e faces assumed to b e morally wrong.
The received picture also apparently has it that Agamemno n ough t t o fee l
something appropriat e for the wrong he has to do. One might think of the case,
however, as involving another dilemm a on the level of "ought-to-feel"in th e
first instance a conflict between feelin g and action, the result of mixing "ought -
to-feel" wit h "ought-to-do." There are cases, tha t is , as I noted earlie r in dis-
cussing the pitfalls of guilt, where the emotion would be so overwhelming as to
undermine any possibility of effective action. Surely this is likely in Agamemnon' s
case: I f he allowed himsel f an appropriat e reactio n t o th e murde r o f his ow n
daughter, on e migh t suppose , h e would b e unable to functio n i n hi s rol e a s
military commander and henc e would be failing t o ac t consistently in light of
144 Sensibility an d Standpoints

the very reasons for the sacrifice. But then the ought requiring him to feel guilty,
no less than th e ought he has chosen to violate, would conflict with the ough t
he means to ac t on in performing the sacrifice.
If th e result amounts to a further mora l dilemma , I take that to sho w tha t
"ought-to-feel" leave s intact the problem s o f choic e raise d b y dilemmas an d
makes clear their unhappy upshot for both duty and virtue ethics. Feeling guilty
might itself com e out a s wrongor a s emotionally self-indulgent, insofar as it
interferes wit h require d actionso substitutin g emotion fo r actio n i n such a
case will not provide an effective wa y o f erasing the moral stai n on the agent' s
character. Guil t will still count as appropriate t o the case on the view of appro-
priateness I shall go o n t o defend , which build s in referenc e only t o genera l
adaptiveness in the sense that involves fulfillment o f some moral or other func -
tion in standard cases ; but i n this case its usual function will be undermined. If
we als o sa y that th e agen t morall y ought t o fee l th e emotion , th e cas e migh t
seem to yield a higher order dilemma that rule s out perfect virtue no less than
perfect performance of obligation .
There ar e ways around this conclusion that involv e denying, after all , that
Agamemnon i n our varian t case reall y ough t t o fee l guilty . However, o n m y
account o f the mora l significanc e of guilt in connection wit h virtue, the mos t
obvious move would stil l leave virtue ethics with th e problem s o f moral luc k
raised by dilemmas. We might take "ought-to-feel " i n the cas e as prima faci e
onlyat least supposin g that w e also take it as practical, a s we need to do t o
establish th e lin k between virtue an d duty . Apart fro m consideration s o f vir-
tue, that is, Agamemnon might not seem to have a moral reason against failin g
to fee l guilt y that is strong enoug h to stan d in light of the reasons agains t fail -
ing in his duties as military commander i n the wa y explaine d i n chapter 2 in
defense of dilemmas. So on this account "ought-to-feel " woul d not yiel d a fur-
ther dilemma within duty ethics, though b y the same token i t would not affor d
a wa y o f preserving virtue against the problem s raise d o n th e lowe r leve l b y
"ought-to-do." Agamemnon presumably ought to forgo virtue in favor of right
action, if he has to choose. Moral sensitivity as a requirement of virtue will make
virtue unachievable , in short , eve n wit h th e "time-bound " imperfection s I
allowed for i n section 1 .
A possible way around this conclusion might involve a return to the broa d
sense of guilt, with the suggestion that the agent substitute some less incapaci-
tating emotion suc h as remorse to avoi d a further conflict . Guilt in the narro w
sense would stil l be made out a s appropriate (i n the sense of rationally accept-
able), no matter what the agent does, but it would not therefore be required by
perfect virtue . Remorse coul d stil l serv e to clea r th e agent' s characte r t o th e
extent that it is a negative feeling even if not self-directed in quite the same way
as guilt, and th e requiremen t to fee l i t might be said to b e in force "all thing s
considered." However , remors e migh t als o b e maladaptive i n som e cases
enough so to give rise to a second-order conflic t of the sort just outlined. Though
it does not have the self-focus of either guilt or shame and hence might be thought
to interfer e les s with activities requiring self-assurance, the mere fact tha t i t is
normally a downcast feelin g might inhibit effective action . For that matter, its
Moral Residues 14 5

focus on a particular act in the past and on acts of reparation specificall y relate d
to it might distract attention from the agent's other obligations in an unhelpfu l
All that such second-order dilemma s would d o to undercu t virtue ethics is
to limit achievable virtue to something even further from moral perfection than
we saw i n section 1eve n on the leve l of sensibility, our substitut e for actio n
in light of first-order dilemmas. They do not exhibi t any sort of incoherency in
virtue ethics but simpl y subject i t to problem s o f moral luc k similar to thos e
that appl y to th e ethic s of duty. Th e poin t i s simply that virtu e ethics woul d
seem to b e in no bette r positio n tha n dut y ethics with regard t o mora l dilem-
mas, a t an y rate if we ris e up a level and conside r possibl e conflicts involving
This point rests on a suggestion for avoiding second-order dilemma s within
duty ethics , however , b y taking th e requiremen t to fee l guilt y as prima faci e
only, which now deserves scrutiny. The suggestion presumably would be meant
to save the claim that the statements of "ought-to-do" in dilemmatic cases guide
action indirectly via "ought-to-feel." Otherwise , that is, we could not make out
indirect action-guidance, or the substitution o f emotion fo r action, a s yielding
a conflict-free manifestatio n of motivational force for the oughts in conflict on
the lower level. So duty ethics would still seem to be threatened with incoherency.
In fact, though, I think this misconstrues th e relevanc e of "ought-to-feel " i n a
way that woul d leav e us with problem s o f explanatory coherency .
We can see this by considering the connection o f all-things-considered sta -
tus to the strength of reasons, an d henc e to reasons fo r attention, a s the foun-
dation o f "ought-to-feel." On the perceptual mode l of reasons fo r prohibitio n
defended i n reference t o dilemma s in chapter 2 , ou r assignmen t of all-things-
considered status to a statement such as "Sacrificing one's daughter is prohib-
ited" woul d amount to a claim that the moral reasons agains t sacrificing one's
daughter ar e stron g enoug h to stan d i n ligh t of opposin g reasons . Bu t this is
essentially a claim about "salience, " o r the appropriate allocatio n o f attention .
The strength or seriousness of a reason i n comparison t o others amount s to its
moral o r other practical importance. It is important enoug h to bea r in mind as
a reaso n fo r action, i n shortand henc e there is reason t o loa d i t with affect ,
thus giving rise to emotio n on my account a s the norma l wa y o f fixing atten -
tion o n some appropriate objec t of thought .
Now, thi s latter reasonthe reason fo r emotionmight sometimes be bal-
anced by stronger opposing considerations. There are many quite ordinary cases
in which emotion is not the best way of directing attention, despite the fact that
this standard function i s what sets it off as a distinct mental category. Emotio n
can sometime s undermine action o r eve n distract attentio n fro m th e nee d t o
act. I f we consider case s o f the sor t tha t cam e int o th e precedin g subsection,
where moral action is best promoted by looking away fro m th e moral reason s
for it , it seems that a prohibition might be outweighed in some case s even as a
reason for attention, since there may be stronger reasons for attending to some-
thing else.
I have been following Williams , however, in interpreting overriding status
146 Sensibility an d Standpoints

as implying more than this : that all claims on the other side will be met by act-
ing o n th e reaso n i n question. 55 Bu t surel y Iphigenia , a s th e victi m o f
Agamemnon's sacrifice , has a claim on his attention that would not b e met by
single-minded devotio n t o hi s dut y a s Gree k commander . I n somethin g lik e
McDowell's terms, he r claim is important enoug h not t o b e "silenced" b y any
allegedly more pressin g obligation s to others . I prefer the visua l metaphor o f
figure/ground: From a certain standpoint o f moral evaluation, her claim stands
out agains t th e backgroun d o f countervailing considerations . S o it counts a s
all-things-considered, even when w e grant tha t ther e ar e stronge r reason s fo r
directing attention elsewhere .
According t o much the sam e argument a s we applied t o first-order dilem-
mas, then , w e d o see m to fac e furthe r dilemma s o f "ought-to-feel. " Despit e
appearances, thi s will no t reall y compromise m y propose d resolutio n o f th e
metaethical proble m o f dilemmas. Th e keyston e o f the latte r wa s the general
version of internalism, according to which moral language by virtue of its mean-
ing has motivational forc e in general terms bu t no t necessaril y in any particu-
lar case . Emotio n cam e i n a t thi s metaethica l leve l as th e initia l vehicl e of
motivational force , relied on i n early moral teachin g t o connec t "ought " an d
similar terms to tendencies to required action. My current argument shows how
"ought" may extend t o emotion tendencies too, an d in particular to guilt, as a
way of exhibiting the motivational forc e of the first-order ought that the agent
has to violate in a dilemma. We should not expect i n this case either, however,
that the usual role of emotion woul d b e fulfillable i n every instance. There may
be cases lik e Agamemnon's i n which th e agen t als o ha s to violat e "ought-to -
feel." What thi s possibility undermines is not th e overall coherency of ethics in
the terms explaine d above bu t th e agent' s persona l recor d o f actionand hi s
virtue, which a t thi s leve l i s not s o clearly distinguishable from it , a s we ca n
now se e in application t o som e o f the normativ e problems raise d b y cases of

Making Up for Moral Luck

Even if dilemmas do not pos e a threat t o the coherency of ethics, they still are
troubling as cases of moral luck, or responsibility for factors beyond the agent' s
control. In chapter 5 I extend the guilt/blame asymmetry to show that they do
not thereby undermine the coherency of the notion o f responsibility, which can
be made out i n terms that do not impl y blameworthiness. M y emphasis in my
overall argument here is on the metaethica l problems raised by dilemmas. But
within normative ethics, dilemmas are problematic insofa r as they indeed saddle
an agent with a form of responsibility that undermines her efforts to make some-
thing morally worth y of her life .
Even without blame, that is, the agent in a dilemma will have to live with a
blot on her moral record. My defense of "ought-to-feel" effectivel y extend s her
moral record t o includ e her record of emotional reaction an d thus has norma-
tive implications. It casts guilt in a positive role as a way of making up for moral
luck: Because the emotion serves as a way for the agent to distance herself fro m
Moral Residues 14 7

the wrong she has to do via mental self-punishment, it effectively insulate s her
character, or virtue proper, from he r record of moral action , o r wha t I distin-
guished as "merit" in section 1 .
In the firs t instance , then , what dilemma s of "ought-to-feel " tel l us is just
that even this second-best substitute for action may sometimes be morally ruled
out. So it will not work a s a practical response to moral luck in all cases. More
generally, however, th e explanatio n o f why i t fail s seem s to m e to brin g ou t
what i s fundamentally a t issu e in the problem of moral luck : the fac t tha t th e
self t o whic h w e attribut e responsibility , conceived as what cause s action, i s
really a construct fro m our actions (along with the other things we take as basic
expressions o f moral character ) and henc e in a certain sens e is a fiction . It is
fictional as the sort of separable entity on which we can pin responsibilityin
a sense that require s a determinate moral character .
Kant of course assumed the existence of a separable noumenal self when he
made out a good will as compatible with thoroughgoing moral failure. 56 Even
without Kantian metaphysics, though, the ordinary notion of responsibility that
is threatened by moral luck seems to trade on an artificia l notio n o f the self a s
something one could i n principle characterize independently of what i t does.
However, this is particularly questionable when we get to th e leve l of feeling ,
understood accordin g to m y present argumen t in term s o f act s o f attention .
Consider Russell's case once again. According to my argument in section 1,
the act that he had to do in response to his first-order dilemma, jilting the woman,
could b e taken a s affectin g hi s record o f moral actio n bu t no t hi s character .
However, i f he als o face d a second-order dilemma , we woul d hav e to giv e a
similar treatment to something that intuitivel y seems more central to his char-
acter: his emotional response s in morally momentous circumstances.
One might attempt to cordon off the circumstances of dilemma, at any rate
for agent s who d o not exhibi t a similar general pattern of response. However ,
if we really suppose that the circumstances in Russell's case are morally serious
enough to yield a dilemmathat jilting the woman would cause her serious harm
(or we might turn back to Agamemnon's case for a deed that wears its serious-
ness on its face)it begins to be unclear how we can form a notion o f the agent' s
moral character that does not include his emotional act in the particular instance.
The act in question involve s a kind of refusal t o feelshiftin g attentio n away
from any objects of thought that would tend to generate the required emotion
rather tha n simpl e affectiv e numbness , in response t o a morall y momentou s
occasion. And it is a long-range act, extendin g beyond the occasio n i n tempo-
ral terms . Withholding the response fo r a short period , that is , would no t b e
enough to allow the agent to satisfy his primary obligation in the case. Nor could
he detach himself from his refusal t o feel, on our understandin g of the case, by
rising to yet a further leve l and punishing himself with guilt for itor not until
so much time has passed tha t th e resul t would seem t o coun t a s a chang e in
character. S o at this level, apparently, one is more or les s what one does .
That i s why i t ca n als o b e appropriat e t o blam e someone wh o i s simply
prevented b y his own emotiona l incapacity from satisfyin g "ought-to-feel"
not necessarily an agent confronted with a dilemma but a sociopath, say, who
148 Sensibility an d Standpoints

does not hav e it in her to fee l remors e fo r a vicious murder; or fo r that matter ,
the real-life Russell. My asymmetrical treatment of guilt and blame along with
various other "reactiv e attitudes"to us e Strawson's ter m in bringing them t o
bear o n fre e wil l issueswill therefor e no t yiel d any sor t o f uncomplicate d
overall asymmetry as applied to moral responsibility. 57 On the level of "ought-
to-feel" th e guilt/blame asymmetry may i n fact b e reversed: It may b e socially
adaptive on the whole to blam e others for certain morally crucial feelings they
cannot help , just because it expresses our own standpoint and commitments a s
moral agents. By contrast, guil t for one's ow n unavoidabl e feelings o r refusal s
to fee l might seem to have lost its usual moral point: By shifting attention awa y
from th e objec t of the require d feeling s t o one' s own failur e t o hav e them, i t
would simpl y set up anothe r for m of distraction.
The result of some personal deficiencies, in short, i s a deficient person, eve n
if h e had n o chance to b e other tha n h e is. What w e fee l i n response i s rightly
thought o f as blame, moreover, i n the sense of personal angernot just hatre d
or some other emotion tha t focuses on the person as distinct from what h e does.
On the account defended here, the agent does something morally condemnabl e
in such a case: He violates a statement o f "ought-to-feel." Eve n in cases of sec-
ond-order dilemma , where w e can presumably assign the cause of the agent' s
emotional deficiency to the moral demand s of the situation rather tha n to per -
sonal mora l coldness, w e would typicall y feel a n aversion to the cold-bloode d
agent o f harm that migh t b e seen as an emotional correlate o f "tainting."
If I am right, then, an understanding of moral luck and related free will issues
ultimately require s coming t o term s wit h wha t amount s t o somethin g lik e a
Humean versio n of the moral self : a fictionally independen t entity, invented to
serve as a stable object of praise and blame. 58 If this notion raise s problems fo r
the coherency of ethicsif its fictional status means that we have to retrac t i n
moments o f reflection what we need to sa y as functioning moral agentsthe y
are at an y rate no t problem s specia l to dilemmas.
For present purposes i t will be enough to make clear with reference to con -
crete cases like Russell's that claims about reactiv e attitudes like guilt and blam e
will not yield any simple way out of the problems raised by moral luck . We can
find som e further importan t complication s in variants of Russell's case that do
not amoun t t o dilemma s i n my terms bu t instea d raise issue s of whether an d
how action and its consequences "stic k to " th e moral self. Suppose we alter the
case as I have interpreted it and assum e that Russell was guilty of some element
of negligenceperhaps he inappropriately led the woman on, to get her to com e
to Englandtha t make s th e cas e fal l shor t o f perfec t virtue. I t woul d the n
amount to a dilemma of the sort that Aquinas recognized, a result of the agent' s
own wrongdoing , eve n if wrongdoing o f a relatively minor sor t tha t doe s no t
typically have such devastating consequences .
I take it that in such a case we would hold Russell at least partly responsibl e
for th e consequence s of his action, eve n if our blam e is mitigated somewha t i n
degree b y the fac t tha t the y were completely unforeseeable. W e woul d blam e
him fo r causing the woman's demise , that isno t just for leading her on , an d
not jus t t o th e limite d extent tha t w e would blam e him i f he ha d le d he r o n
Moral Residues 14 9

without suc h extreme consequences. Th e cas e woul d resembl e one i n whic h

driving home from a party slightly intoxicated results in a car accidenta vari-
ant of the case of an unavoidable car accident that I deal with in chapter 5, bu t
one that i s avoidable in long-range terms in the sense that the agent could an d
should hav e prevented the acciden t b y refrainin g fro m drivin g whil e intoxi -
cated.59 The point for present purposes i s that the agent's faul t a s we conceive
it in such cases does sprea d to th e consequence s of her action , s o we think it
reasonable to blame her in light of the harm she actually causes. We do not jus t
expect he r t o fee l guilt y about i t hersel f i n order t o distanc e herself from he r
record o f action an d thus preserve virtueeven whatever imperfect virtue she
can stil l clai m compatibly with wha t sh e didas in my treatmen t o f cases in
section 1 . She is damned by the harmfu l consequence s o f her deed . Th e poin t
also holds for Russell's case, supposing that harmful consequences would result
from a morally mandated response to the first-order conflict .
One might be tempted to correct fo r the oddity of these ordinary judgments
of responsibility by subtracting out moral luck factors, but any such attempt t o
tidy them u p for theoretical purposes woul d thro w the m ou t o f line with ou r
intuitive assessmen t o f th e correspondin g reactiv e attitudes. Consider , fo r
instance, the suggestion that an agent guilty of minor negligence without harmfu l
consequences ought properly t o fee l jus t a s guilty as i f she had cause d majo r
harm, since her character would be no different fro m tha t of the agent who did .
This amounts to another way of separating virtue from merit. It seems unrea-
sonable, though, where the harm in question is not even foreseeable in general
terms a s the sor t o f thin g that migh t wel l happen i f one commit s th e mino r
misdeed in question. This is how Russell' s case differ s fro m the cas e of mino r
negligence that might have caused a car accident but did not. I n the latter case ,
guilt is warranted no t jus t by the possible consequences of such negligence but
also b y their likelihoodthe risk the agen t actuall y (and knowingly) took b y
drinking while intoxicated, say. The standard way of generating guilt for some-
thing of the sort is in fact to imagin e the possible consequences as if they were
actual, and we can reasonably demand that the agent do sovividly enough to
induce greater care in the future. Bu t that i s nothing to what w e would expec t
of her i f the consequence s actually occurred, an d sh e did i n fact kill someon e
through negligence.
So the real-lif e Russel l who treate d th e woma n h e jilted somewha t badly ,
but with no very bad results, ought presumably to feel some guilt in light of the
normal sort s of wounds that can b e inflicted b y such behaviornot just for a
violation o f gentlemanly honor o r th e like . Some guilt may b e in order o n th e
same grounds, in fact, even in a case where the woman turn s out to suffer noth -
ing at allbecause she immediately takes up with someone else , saythough
here I think the demand for self-punitive emotion would agai n be mitigated in
degree, this time by events as they actually ensued. However, i t would be unrea-
sonable to expect Russel l in either of these cases to fee l what w e would expec t
him t o fee l i n the case of catastrophic moral luck.
That there are such cases is something I take to b e a fact o f moral life, to b e
understood bu t no t mad e any easier, even in a theoretical sense, by appeal t o
150 Sensibility an d Standpoints

our reactiv e attitudes on the sort o f account propose d here . For emotions als o
reflect moral luck. My account of guilt for the unavoidable in chapter 5 will be
relevant to the question, then, insofar as it lets us extend the emotion an d even
appropriate instance s of it beyond an agent's morally culpable acts. But it will
thereby offe r a way of accommodating th e phenomenon rathe r tha n attempt -
ing to dispel it or to reinvent ethics in light of it; for the Humean self that picks
up its character from its consequences is not simply a philosopher's notio n that
might b e rejected while maintaining the social practices that constitute moral -
ity. I take i t to b e our ordinar y notion , underlyin g our ordinar y treatment of
the moral sentiments .
Unavoidable Guil t

Several recen t author s o n guil t and relate d notion s apparentl y allo w fo r th e

rationality of guilt feelings in at least some cases o f unavoidable wrong. Their
views yield differen t decision s o n som e troublin g real-life case s rangin g from
the case of an unavoidable ca r accident that, without any fault o n the part of
the driver, results in the death of a child to cases of survivor's guil t and vicari-
ous or collective guilt. I now wan t t o tak e a new look a t thes e cases and oth -
ersand a critical look a t some o f the views of guilt proposedin a n attemp t
to wor k ou t a n understandin g of the content of guilt and th e ground s fo r its
appropriateness tha t wil l better suppor t th e general conclusion tha t detache s
grounds for guilt from grounds for blame .
This project rests on a rejection of the standard vie w of guilt as involving a
corresponding evaluativ e judgment that w e find, most notably , i n Rawls. 1 O n
this "judgmentalist " account , emotiona l guiltwha t w e ma y distinguis h a s
"subjective" guiltrequires a judgment (in the sense of a belief) that on e actu -
ally i s guilty, a judgmen t o f "objective " guil t o f the sor t tha t implie s mora l
responsibility. Views differ o n whether this requirement is to b e imposed onl y
on appropriat e instance s or o n all genuine instances o f the emotion, a s Rawls
apparently has in mind. At any rate, there is a way of accommodating guil t for
the unavoidable that remain s within th e judgmentalist framework bu t weak -
ens the content o f the requisit e judgment to an attribution o f causal responsi -
bility that extend s t o case s lik e the acciden t cas e while apparently rulin g out
survivor's guilt and othe r dispute d cases as impossible .
What I want t o defen d her e i n oppositio n t o thi s accoun t i s a nonjudg-
mentalist view of guilt that see s the subjectively guilty agent as "feelin g as if "
he were morally responsibl e (sectio n 1) . On m y own view , guilt amount s t o
discomfort with a certain evaluative prepositional objec t and hence may be said
to correspond t o a judgment, though one can undergo the feeling without holding
the judgment. Indeed, one can even undergo it appropriately, a s I argue in sec-
tion 2 . My view will allow u s to interpre t survivor' s guil t and othe r dispute d
cases as having the same sort of content a s moral instance s of guilt; it will also
give us a framework for fine r grained consideration o f questions o f emotiona l
appropriateness in application to the ful l rang e of cases.
My view takes guilt as adequately warranted b y a partial subset of the total
body of evidence bearing on its corresponding judgmenta perceptual "slice "
of the evidence, one might say, that on practical grounds is seen as sufficient t o
152 Sensibility an d Standpoints

warrant holding in mind the evaluative content o f the judgment. We may think
of this as a "perspectival " accoun t o f emotional appropriateness . Th e general
idea behind the view is that emotional discomfor t serves as a way of holding an
evaluative thought i n mindas distinc t fro m puttin g i t into storage , a s wit h
beliefso tha t warran t fo r a n emotio n i s properly affecte d b y nonevidentia l
reasons for attention to its evaluative component. Thes e include various instru-
mental functions that such attention serves , including social or moral functions,
along with any noninstrumental moral or other norms that affect its value under
similar circumstances. I argue that the view therefore supports a n asymmetri -
cal treatmen t o f guil t an d blame , a s form s of emotiona l punishmen t whos e
practical effect s diffe r becaus e guilt imposes th e punishmen t on th e on e wh o
undergoes the emotion whereas blam e imposes it on someone else.
In a word, guilt is sometimes appropriate, i n contrast t o blame, when we do
not hav e adequate warrant fo r th e correspondin g judgment . After defending
this point i n general terms, I attempt to teas e ou t som e o f its implications for
our dispute d cases of guilt fo r th e unavoidable . I then turn t o problem s tha t
arise in applying the view to cases of moral dilemma, as cases in which guilt is
unavoidable because the agen t mus t d o wrong, thoug h an y particular wron g
he does will be avoidable. The guilt/blame asymmetry explains our reluctanc e
to blame the agent in a dilemma, but i t also seems to exhibit the inadequacy of
a subjectiv e definitio n of dilemmas , a s cases i n which eithe r alternative war-
rants guilt. The argument allowin g for guilt without mora l responsibility also
apparently supports at least some cases of guilt without wrong as perspectivally
appropriate. A subjective definition of dilemmas would therefore seem to imply
only th e prim a faci e wrongnes s o f eithe r alternative . However, I shal l try t o
show ho w we can extract an objective notion fro m m y treatment of appropri -
ate guilt (section 3) by way o f the referenc e to "tainting " tha t come s int o ou r
explanation of the desire for reparation characteristi c of the emotion .
My account of guilt will have to b e complex enough to yield plausible dis-
tinctions betwee n dilemma s an d othe r cases , includin g our initia l cases o f
perspectivally appropriate guilt, where we want to deny that the agent really is
guilty. However , I shall begin b y narrowin g attentio n t o th e questio n o f th e
content o f guilt in cases of the latte r sort , especially the familia r cas e o f guilt
for causing the death o f a child in an unavoidable car accident . In effect, I shall
be arguing that th e simplest ways o f responding to suc h cases are too simple :
Either they rule out guilt as something other than the emotion it appears to be
or the y accep t i t a s unproblematicall y rationa l i n a wa y tha t seem s equally
counterintuitive. I shall eventually argue that it is rationalbut on the basis of
the mor e comple x "nonjudgmentalist " accoun t I shal l defend here , firs t i n
application to guilt itself and then t o the grounds for its appropriateness .

1. Subjectiv e Guil t and Responsibilit y

How shoul d we analyze emotional guilt? The feelin g would seem to rest on self -
blame, but we should note first that it does not always have the active or aroused
Unavoidable Guilt 15 3

quality of anger. Following the rough lines of Aristotle's definition I take anger
meaning persona l anger , th e varian t o f th e emotio n tha t amount s t o other -
directed emotional blam e as opposed to mer e frustrationto involve a desire
to inflic t som e sor t of punishment on it s object for a wrong. 2 I would analyze
the emotio n furthe r into affective an d evaluativ e aspects b y taking the desir e
for punitive action as involving discomfort at an unfulfilled actio n requirement
at th e though t tha t the agent , o r th e subjec t o f emotion, ough t t o punis h it s
object, meanin g tha t actio n o n he r par t i s still neede d t o effec t punishment .
Anger thus is seen as an aroused feelin g at least partly becaus e of its essen-
tial orientation towar d actionits threat o f continuing discomfort unles s and
until the agent does punish the objecteven though it may sometimes be satisfied
without action , o r without action o n the part of the agent, perhaps b y an apol-
ogy from the object. An apology might be seen as a kind of self-punishment t o
the exten t tha t i t involve s self-abasement, s o i t seem s t o coun t a s a n activ e
expression o f guilt . Fo r tha t matter , emotiona l guil t involve s a kin d o f self -
punishment in cases in which the agent has some control over whether he expe-
riences that unpleasant feeling. But it is important that, if anger is thought o f as
originally grounded i n an animal urge to attack , guil t comes ou t no t a s a self -
directed versio n o f the urgea n urge to attac k oneselfbu t as a les s aggres -
sive counterpart o f it, requiring in the first instanc e reparation, o r some way of
making amends. Guil t may thus involve self-punishment as a form of repara -
tion alon g with the readiness to submit to attack o r to other punishmen t from
others; in developed form, though, i t is not simpl y inwardly directed anger .
One might still say that the different desire s for action essentia l to guilt and
(personal) anger both hav e the same general end: a state of affairs i n which the
perpetrator someho w "pay s for" a prior wrong. They differ i n where they place
the burden of active responsibility for accomplishin g that end, each assigning
to a differen t agen t a requiremen t o f action enforce d b y .discomfort unti l th e
job is done. This difference o f course amounts to a limited structural similarity:
The agents in question here are different i n relation to the prior wrong, but they
are both subject s of the relevant emotion. O n the other hand, anger has a per -
sonal object, viewed as the perpetrator o f the wrong, whereas guilt also assign s
this prior sort o f responsibility t o th e subject . Insofar as they both hol d som e
such guilt y party t o account , though , bot h emotion s ultimatel y rest o n th e
attribution of responsibility for a wrong. S o on the assumption tha t emotiona l
evaluations amount t o judgmentsth e assumption I call "judgmentalism"
one might look fo r the basic content of both emotion s i n a judgment of respon -
sibility. The judgment yields emotional guil t and blam e as different specifica -
tions of the urge to right a wrong depending on the agent's practical standpoint .
I now want t o consider problem s with this view and ways of defending it in
application t o som e familia r cases of guilt in which we do no t see m to hav e a
judgment of moral responsibility: cases of guilt without fault. I shall work wit h
the cas e of guilt for a n unavoidabl e car acciden t an d construc t a variant of it
that seems to involv e guilt without agency t o sho w th e inadequacy of a recent
attempt t o defen d judgmentalis m by framing th e accoun t o f guilt i n terms of
causal responsibility . I think we ca n se e that thi s move commits the agen t i n
154 Sensibility an d Standpoints

such cases to a counterintuitive degree of irrationality. I shall attempt t o sho w

that a nonjudgmentalist account tha t preserves the reference t o moral respon -
sibility would square better with intuition, particularly as extended to more stan-
dard case s o f guilt without agenc y that anothe r recen t autho r make s ou t a s
rational but nonmoral. I shall end this section by extracting from the cases some
features o f the practical role of the emotion tha t see m to yield a rational basis
for case s o f guilt without blame.

Guilt without Fault

On th e vie w of the conten t o f guilt just described, the emotio n i s sui generis,
though it s evaluative structure links it closely to anger, with differences in desire
content bu t th e same evaluative basis for desire in a judgment of responsibility
for a wrong. It seems natural to thin k of the latte r a s a judgment of moral re -
sponsibility, for a version of this view that I shall call "naive judgmentalism."
A naive judgmentalist analysis of guilt makes feeling guilt y rest o n a straight -
forward belie f that one is guilty, a judgment simply asserting the evaluative con-
tent of the emotion. But this analysis seems immediately to be called into ques-
tion by cases of guilt without faultclear-cu t cases , in which the agent may be
assumed t o kno w tha t h e i s not a t faul t (a n assumption I shall often take fo r
granted i n what follows)suc h as guilt at causin g the deat h o f a chil d i n a n
unavoidable ca r accident . Eve n supposing tha t th e feeling s i n suc h case s ar e
irrational, it seems undeniable that they occur, and almost undeniable that they
amount t o guil t feelings.
The denia l that they amount to guilt is suggested b y the responses o f naive
judgmentalists to another sor t of case, in which the agent does not believ e that
some ac t o f his was wrong , thoug h he does accept ful l responsibilit y for it . A
standard example , found in Rawls, involves the violatio n of a religious taboo
in a religion the agent was taught a s a child but no w rejects. 3 What th e agen t
now feel s whe n h e violates the Sabbath , fo r instance , ma y b e a feelin g o f dis-
comfort persisting from childhood experience s of guilt, but on this view it can-
not b e a genuine case o f guilt.
It may see m that w e nee d some vie w of this sort t o distinguis h guilt fro m
various other emotions that migh t be confused with it, such as fear o f punish-
ment or other forms of anxiety that might be thought of as its childhood prede-
cessors. Eve n as supplemented by a desire to hea d of f or appeas e th e punish-
mentby performin g some ac t of the sor t tha t amount s t o reparatio n wher e
we do have guiltthis feeling needs to be conjoined with a view of the punish-
ment as somehow justified , to amount to (subjective ) guilt. Built into the notio n
of guilt, that isas distinct from, say, fear of persecutionis a kind of acknowl-
edgment of grounds for punishment as given in the corresponding judgmen t of
In general, it seems that emotions have an evaluative content that determines
their classificationdistinguishin g guil t from fea r an d variou s other unpleas-
ant reaction s that ma y no t alway s be so distinct from i t in affectiv e quality . I
take this to b e the main point i n support of a judgmentalist analysis, though I
Unavoidable Guilt 15 5

think it can also be used to construct a n evaluative view that does no t require
the content o f an emotio n to b e an objec t of strict belief . But before we aban-
don judgmentalism , we might tr y makin g room withi n i t fo r case s o f "emo -
tional inertia, " o r th e la g of feelings behin d their corresponding beliefs , as in
taboo cases of the sor t jus t illustrated.
In a n attemp t t o accommodat e tabo o cases withi n a versio n of judgmen-
talism, one recent author, Gabriel e Taylor, seem s briefly to bring together tw o
such strategies.4 First , the naive judgmentalist analysis of guilt might be retained
by simply weakening the notio n o f judgment o r belief , takin g it to cove r an y
thoughts that com e to mind, even if they are immediately rejected by "consid -
ered judgment." Thus, Taylo r say s that the act forbidden by a taboo retaine d
from childhoo d stil l "presents itself " as wrong in the circumstances of action ,
though no t when the agent considers it "from a more rational point o f view. "
Her introductor y remark s o n emotion s an d belief 5 indicat e tha t sh e woul d
interpret th e notio n o f belie f t o includ e the sor t o f mental stat e suggeste d b y
"presents itself " i n this passage: wha t i s sometimes calle d a n "a s if " feeling .
An agent's residual feelings from childhood religious belief would still count as
involving belief s o n thi s accoun t a s lon g a s the y hav e th e though t conten t
required for feeling s o f guilt.
Second, one might also allow for taboo cases by weakening that evaluative
content, the content o f the judgment in question, allowing for something like a
conventional interpretatio n o f "justified " punishment. Punishmen t migh t b e
thought to be justified i n some sense, that is, as long as it is imposed fo r violat-
ing an authoritative rule , meaning one whose authorit y is generally accepted,
even if the agent questions the reasons fo r it himself. What on e feels guilty for,
on thi s view, is just the violatio n of a taboonot necessaril y a moral wron g
and hence not enough to satisfy the naive judgmentalist analysis. Though Tay -
lor apparently combines a move of this sort ( I have restated he r version of it to
apply more clearl y to rules ) wit h the weakene d notion o f belie f tha t wa s just
noted, it would seem to be sufficient o n its own to allow for guilt in taboo cases.
Both o f these two genera l judgmentalist strategiesweakening the notio n
of belie f o r th e evaluativ e content o f the belie f require d fo r a guil t feelings
may b e extended t o th e cases with whic h we are concerned. Her e th e agent' s
responsibility i s in questio n rathe r tha n th e mora l evaluatio n o f hi s act , o r
whether i t amounts t o a wrong i n the sens e of something forbidde n as in th e
taboo cases . A different versio n of the second strategy , yieldin g what migh t be
called a "weak " judgmentalis t analysis of guilt, stand s behin d Taylor' s ow n
treatment of such cases. I want t o consider it here at some length , since it rep-
resents a possibility intermediat e betwee n naive judgmentalism and m y ow n
Taylor interpret s emotiona l guil t in terms o f a weaker sor t of judgment of
responsibility tha n tha t involve d in ordinary judgments of fault: causal rathe r
than mora l responsibility. 6 As applied to the accident case, thi s notion appar -
ently would make guilt feelings unproblematic, though it would also leave one
puzzled a s to ho w thei r rationality could be called into question. If subjective
guilt implies only the self-attributio n o f causal responsibility, feelings o f guilt
156 Sensibility an d Standpoints

at causing the death of a child, however unavoidably , would see m to b e clearly

appropriate a s well as authentic .
Of course it is not unthinkabl e that the analysi s of such complex emotion s
as guilt might yield a few conceptual surprises . Bu t the onl y advantag e o f the
weak judgmentalis t analysis seems to b e its straightforward treatmen t o f this
and othe r putativel y rational case s of emotion a s cases o f rational belief . Fo r
an idea of the problems it faces in application t o irrational cases we might com-
pare th e usual version of the acciden t case with on e not eve n involving causal
responsibility. Consider wha t the weak judgmentalis t account woul d hav e us
say about the distinction betwee n th e usual version of the case, in which harm
results from something the agent does (albeit unavoidably), and a case in which
the chain of causation lead s back without interruption to some prior cause. For
instance, what i f the agent's car had simpl y been propelled int o a nearby child
by anothe r vehicl e that hi t itan d that would hav e hit th e child, ha d hi s car
not been thereso that the accident i n no way resulted from his agency? Here
Taylor woul d apparently hav e to dismiss guilt as unintelligibleat any rate, in
an agent assumed to b e otherwise basicall y rationalsince the emotion coul d
not have even the weaker belie f content require d by her suggested modification
of its standard judgmentalis t analysis.
Intuitively speaking , however , I think w e hav e t o sa y that suc h case s ar e
possible. We can imagine someone goin g over in memory the sequence of events
leading up to the crash, pulling out a subset of them to focus on that is compat-
ible with the usual scenario, th e one that does involve responsibility, and feel -
ing guilty. His reason need not be uncertainty about what happened bu t rather ,
say, some irrational tendency to fix on the worst possible interpretation o f events
from th e standpoin t o f self-esteem . Perhaps h e wa s taugh t t o blam e himself
excessively as a child. I take this sort of tendency to be compatible with overall
cognitive rationality : Th e agen t on e ha s i n mind her e i s not cognitivel y con-
fused i n th e usua l sensehi s syste m o f settle d belief s i s not disruptedbu t
instead is subject to a relatively localized disruption of the normal response ten-
dencies. However, in order to accommodate suc h cases within judgmentalism
to preserve a foothold fo r real but irrational guilt feelingswe apparently need
to brin g in a version of Taylor's first strategy for handling emotional inerti a in
taboo cases, with a notion o f belief wea k enoug h t o allo w fo r cognitive delu-
Further weakening the content o f belief woul d mak e the case com e ou t a s
rational, that is; so to explain it as a case of irrational emotion o n this account ,
we apparently have to ascrib e to its agent a n irrationa l causal belief. We have
to grant that the agent i n the case believe s himself causally responsible fo r the
accident. The agen t clearly feels as if he were responsibleat an y rate, off and
on, o r at those time s when h e focuses on the subse t of his memories tha t sug-
gests that interpretatio n o f eventseve n thoug h h e knows h e i s not an d dis -
misses the feelin g as deluded. But the judgmentalist account o f the case insists
that th e feelin g implie s a deluded belief.
If we have to allow this much cognitive delusion, however, why not attribut e
to th e agen t a deluded belief i n moral responsibility ? The strateg y o f weaken-
Unavoidable Guilt 15 7

ing th e naiv e judgmentalis t analysis o f guiltth e strateg y tha t yield s wea k

judgmentalismmight seem to be unnecessary, in other words, when one con -
siders the weak sense of "belief" th e analysis still presupposes. W e might extract
a simila r point fro m th e taboo cases: An agent migh t sometimes fee l lingerin g
guilt about th e violation of a religious rule that is not actuall y accorded muc h
authority i n hi s adult lif e eve n i n conventiona l terms , thoug h i t wa s i n forc e
during childhood. Particularl y clear-cut example s migh t b e drawn fro m rule s
of conduct meant to be limited to children, such as prohibitions of naughty words
and the like, to show tha t th e weakening o f belief content t o handle irrationa l
cases does no t reall y represent a distinct alternativ e t o the reliance on a wea k
notion o f belief .
At most, the weak judgmentalist strategy serves to contain the extent of the
cognitive delusion that the naive judgmentalist strategy attribute s to the agen t
in such cases. I t makes out th e firs t version o f the acciden t case as undeluded ,
that is, and hence as clearly rational; o n the other hand, it seems to yield a treat -
ment of the second versio n of the cas e a s involving a deeper kind of delusion :
delusion as to the facts. Rather than merely being confused about the standard s
for mora l responsibilit y or somethin g similar , the agen t mus t on thi s accoun t
mistakenly believe that h e somehow cause d th e accident after all.
More precisely , he must both believ e this and believ e it to be false, either at
the sam e tim e o r wit h n o goo d reaso n fo r a chang e o f mind , sinc e o n ou r
hypothesis he knows he is not responsible for the accident, though he feels com -
pelled to dwel l on th e subse t o f events tha t suggest s tha t h e is. The cas e thu s
requires logical conflict; so the exten t of delusion require d by weak judgmen-
talism still seems excessive. It is minor onl y if one considers how littl e may b e
built int o th e notio n o f belie f on eithe r versio n o f judgmentalism : Any stra y
thought would seem to have to qualify, if all of the cases in question here are t o
count a s cases o f guilt.
On th e assumptio n tha t belie f interpreted i n any reasonabl y strict sens e is
governed b y a principle of logical charity , I would tur n instea d to a nonjudg-
mentalist analysis of emotion.7 Whereas the weak judgmentalist analysis of guilt
allows for a weaker judgment than the judgment of moral responsibility required
by the naive analysis, my approach would allo w fo r something weaker tha n a
judgment bu t stil l framed in terms o f moral responsibility . In general, instea d
of claiming that emotion s entai l evaluative beliefs, I take them sometimes just
to involve evaluative thoughts held i n mind by intentional states of comfort o r
discomfort. Thus , whe n an agent feel s guilty about th e death caused b y an ac-
cident he was involved in, he need not actuall y assent t o the evaluative basis of
his emotionthe though t o f himself as responsible for a wrongbut h e doe s
have to be discomfited by it in a sense that involve s entertaining the thought as
an objec t of discomfort .
This is not to say that the agent merely "entertains" such a thought; it s status
as an objec t of discomfort (with discomfort taken as a general state o f feelin g
of a sor t on e would naturally want t o escape ) is essential to th e motivationa l
cast of my account. Nor d o I want t o say, on the other hand, that the agent ha s
to entertain such a thought explicitly. Let me very briefly try to cancel out som e
158 Sensibility an d Standpoints

misleading suggestions of my account and t o indicat e the general rationale for

it before considering its application to cases. First, the agent's discomfort about
a certain evaluativ e thought is understood t o b e directed toward th e preposi-
tional conten t o f the thought , th e stat e o f affair s i t concerns, no t towar d th e
fact tha t i t occurs t o hi m or som e logica l or othe r featur e o f it as an object of
contemplation. Second , my analysis is meant to allow for unconscious emotions,
as involving conscious states of affect (comfor t or discomfort) but with evalu-
ative objects that th e agent cannot identif y correctly. 8
Further, though the analysis is understood t o be limited to occurrent emo-
tions, s o that guil t amount s t o occurren t discomfor t a t a certai n evaluative
thought, the thought content of an emotion need not itself be taken as occurrent.
There need be no episode of mental utterance corresponding to an unconscious
thought; and even on the conscious level , talk of thoughts might be replaced by
statements describin g relations amon g hypothetica l objects o f attention. Th e
claim that an agent' s discomfor t in a case of guilt, say, is occurrently directed
toward a thought o f himself as responsibl e might be understood a s meaning
that attention to his discomfort would lea d to awareness of that thought in the
absence o f various barrier s t o attentionincludin g barriers of th e sor t tha t
presumably make guilt unconscious.
It is tempting to equate a dispositional mental state of attention t o a propo-
sition, of the sort here in question, with a belief . However, I take the relevant
disposition t o be tied in a special way to prereflective features of "mental set, "
as part o f a preparatory for m o f practical reasonin g that sometime s turns on
remaining subliminal, immune from the intellectua l criticism that belie f i s ex-
pected t o withstand. 9 Emotiona l evaluations, in short, amount to pattern s of
attention, without the stability we expect o f beliefs. That belief implies a degree
of logica l coherency is suggested by some recen t treatments of more straight-
forwardly cognitiv e influences o n behavior . Stephen Stich, for instance, inter-
prets the less logically structured behavioral dispositions of nonhuman animals
as "subdoxasti c states, " an d i n a treatmen t o f developed practical reasoning
Michael Bratman uses the term "acceptance" fo r a state of taking some propo-
sition for granted without belie f relativ e to a given practical context. 10
What Bratman has in mind is the sort of assumptionabout the likelihood
of an earthquake, in one examplethat it is rational to make in some context s
and not others (depending, for instance, on the practical costs of error) in delib-
erating about what to do. It would not be rational on Bratman's account to let
belief var y with the contexta n instanc e of th e principl e of logical charity, I
take it, based on the primary tie of belief to theoretical contexts. I want to sug-
gest, further, that what I have called "as if " feelings in my argument above may
be picked out on the model of Bratman's states of "acceptance" b y their role in
practical reasoning. Their role is of course somewhat different; but like Bratman,
I shall be content to characterize them in terms of itin terms of a kind of ideal-
ized or normative functional role , in effectwithout attemptin g to say anything
about thei r nature in themselve s excep t that the y otherwis e resemble beliefs.
Generally speaking , on th e accoun t I have applied to guilt , emotio n add s
motivational force to our explicit reasoning from judgment s insofar as it directs
Unavoidable Guilt 15 9

attention t o certain thoughts by loading them with discomfort. The usual prac-
tical point o f guilt, after all , is to motivat e actio n o n a moral ought-judgment
by inflicting emotiona l punishment for the failure to act . A s we have seen, the
mechanism can operat e i n advance of action a s applied to th e mer e anticipa -
tion of moral failure : the thought of oneself as already responsible for a wrong,
even when thi s does not amoun t t o a belief . What it amounts t o i s a momen -
tary objec t of (dispositional) attention, held in mind an d allowed t o influenc e
thought an d behavio r as i f it were believed , though unlik e belief it woul d b e
discarded upo n a moment's reflection.
The point is to generate a readiness to act that resists reflection, along with
the ability to ignor e or explain away the practical urging s of judgment. Emo-
tional motivatio n on this account reinforces the usua l model of practical rea -
soning with the need to discharge discomfort as a reason for action beyond what
is provided b y the evaluative content of emotion, whic h may or may not b e an
object of belief. Like Bratman's "acceptance," a n emotional evaluation is treated
temporarily o r fo r certain purpose s a s if it were believed . This means that, in
discussing cases of the sort at issue here, we cannot take at face value the vari-
ous thoughts that might occur to an agent: Some mental contents that could be
seen as self-ascriptions of responsibilitythe agent's reflection tha t he should
never have gone out that night , for instancewould b e most charitabl y inter-
preted on that account as nonjudgmental. They may sometimes just amount t o
questions the agent puts to himself or thoughts h e considers and rejects; or they
may be held in mind on something other than a literal reading as responsibility
On some such assumptions, at any rate, we may say that guilt does involve
a thought of oneself as morally responsible but that i t need not alway s involve
the corresponding belief . The nonjudgmental analysis will therefore allow for
guilt without cognitive delusion even in the second version of the accident case,
in which the agent feels guilt y just as a result of passive causal involvement in
a child's death , though h e knows he lacks even causal responsibility for it . In
effect, th e analysi s limits his disturbance to th e leve l o f emotion b y detaching
emotion fro m belief . It also has the advantage, as I shall argue, of allowing fo r
guilt i n some case s where on e might b e tempted t o detac h th e emotio n eve n
from a thought of moral o r causal responsibility.

Guilt without Agency

A crucial test of my nonjudgmentalist analysis will be its ability in the next sec-
tion t o provid e a foundatio n fo r som e defensibl e distinction s o n th e issu e of
emotional appropriateness. Fo r present purposes, however , let us just attempt
to use the analysis to accommodate some apparent cases of guilt. First, in what
follows, I show how the analysis allows for clearly irrational cases like the one
just discussed (but without yet defending their classification a s irrational). Then
I appl y i t t o som e arguabl y rational cases tha t eve n the wea k judgmentalis t
analysis in terms of causal responsibility would seem to rule out. In effect, then,
in th e remainde r of thi s sectio n I shal l b e extendin g my defens e o f th e non -
160 Sensibility an d Standpoints

judgmentalist analysi s to a widening circle of casesand t o anothe r possibl e

way of weakening the notion of responsibilitywith implications for the justifi -
catory questio n o f guilt versus blame.
My analysi s differs fro m th e judgmentalis t analyses considere d s o fa r i n
allowing guilt to take on objects other than the agent's actsor thing s he does,
if that should be taken more broadly; or even, perhaps, event s involving him
for a less clear-cut dividing line than is usually drawn betwee n object s of guilt
and shame. O n the usual view, only shame applies to persona l traits , though t
of a s distinc t fro m conformit y to rules . Bu t i f there ar e norm s impose d o n
children's personal developmen t that a child might be expected t o come to live
up to without necessaril y doing anything, then one might very well grow u p to
feel guilt y about violatin g them. A n example might b e guilt about no t bein g
very bright o r ambitiousnot havin g what i t takes t o succeed i n the way that
one's parents may have had in mind. The object of guilt here amounts to some -
thing ove r an d abov e an y definit e omission s on e migh t b e sai d t o hav e per -
formedomissions sufficientl y localize d in time to count a s objects of guilt on
the usual account. It need not be seen as reducible, say, to a failure to go through
some particular stage in development that was supposed t o involve doing some-
thing: perfecting one's talents, working hard, or the like. Intuitively speaking,
it seems possible, whether or not it is rational, to feel guilt, not just shame, abou t
all sorts of uncontrollable inadequacies, inabilities, and trait s o f character o r
temperament discourage d b y parents an d other s bu t extendin g a s far bac k i n
one's history as the claim to a distinct personality or self.
This means that th e objects of guilt in certain cases will violate the assump -
tions o f Taylor's analysi s by failing to allo w for reasonable causal attributio n
to the agent in terms of either event-causation of the usual (Humean) sort or an
indeterministic notion o f "agent-causation." 11 I shall come bac k to these tw o
alternatives for making out causa l responsibility in a moment, but firs t le t me
introduce another sor t of case that might tempt us just to abandon referenc e to
responsibility or a t an y rate to weaken th e sense of the ter m furtherth e ap-
proach take n b y Herbert Morri s i n his defense o f "nonmoral" guilt. 12
Morris mark s off as nonmoral a subtype of guilt that does not involv e cul-
pability; he explains it instead, in effect, a s a form of "separation guilt, " base d
on the severing of personal tie s to member s of a group wit h whic h one identi-
fies. Although Morris's argument focuses directly on the notion o f appropriat e
guilt rather tha n startin g from a n analysi s of the conten t o f the emotion , i t is
relevant to m y discussion her e because it yields a kin d o f opposit e pol e fro m
Taylor's accoun t in application to cases of guilt without agency . Unlike my own
view, it does not requir e even the thought o f oneself as somehow i n the posi -
tion o f an agent .
The cases just cited of guilt for personal shortcomings might indeed seem to
involve another sor t of responsibility besides the sort at issu e in moral blame .
Children growing up are more or less held accountable by their parents for liv-
ing up to parental expectations. I would treat this as sometimes giving rise to a
tendency (short of belief) for an agent to think of himself as if he had don e some-
thing to make his personality what it is; but Morris would say that moral respon-
Unavoidable Guilt 16 1

sibility is not necessar y even in thought fo r a genuine (or even an appropriate )

case of guilt. By failing t o confor m t o th e norm s o f the famil y group , a child
may in a certain sense be breaking away from the group independently of any-
thing it does, jus t in virtue of basic temperament o r personality; and thi s is all
that a judgment of nonmoral guil t asserts o n Morris's account .
Morris doe s not appl y hi s view to cases of guilt for traits. However , thes e
may not be the strongest cases in favor of either Morris's view or my own. On e
might object that their intuitive plausibility as cases of guilt rather than sham e
seems to depend on whether they involve behavioral shortcomingsas oppose d
to the failure to meet norms of attractiveness and the likeand hence on a kind
of general reference to agency. But there is another set of examples among thos e
discussed b y Morris, o f guilt for a n undeserve d benefit 13 ranging from sur-
vivor's guil t t o guil t at bein g favored economically over otherst o undercu t
the suggestio n that a n objec t of guilt must a t leas t be manifested in behavior .
An agen t might know tha t ther e was nothing sh e could hav e done, fo r in-
stance, to keep from being favored over a sibling in her childhoodand noth -
ing she omitted then or since in attempting t o make up for the inequality . But
the inequality itself seems to b e a possible object of guilt. At most, the object of
guilt in such cases would seem to involve a passive eventbeing benefited over
others, say . But this i s something tha t happen s t o th e agent , a s in the secon d
version of the accident case.14 Responsibility for the unequal state of affairs tha t
results from i t would seem to shift bac k to its own causes, then, with n o causal
role fo r the agent .
It may not b e easy, though, eve n to fin d a n event involving the agen t tha t
can serve as part of a causal chain leading to the object of guilt in such cases, if
we follow Hume's insistenc e on logicall y distinct caus e an d effect. 15 In som e
cases, i t may not b e possible to pic k out an y very definite event. Consider, fo r
instance, guilt for one' s beliefs : Commo n tal k o f guilt feelings fo r a failur e of
religious belief would seem to apply to a case in which the agent discovers that
she never did have the kin d of faith tha t is required of her. Bu t then he r guilt
will no t b e attributable t o a n even t of ceasing to believeo r even, give n the
sort o f belief tha t i s in question here , t o a failur e o f belie f formatio n a t som e
particular tim e when belie f wa s calle d for . Perhap s sh e migh t se e herself a s
involved in some sort of ongoing ac t o f omission o f belief formatio n through -
out he r life . Bu t although some mov e of this sort might serve to locat e withi n
the agent's history a Humean event-cause of her lack of faith, it could not give
us the two events we need for a Humean clai m of causal connection. Tha t is, a
second applicatio n o f the move would no t see m to pick out a distinct even t to
serve as the objec t of guilt in this case. What th e agen t feel s guilt y about, if it
does amount to an act, seems to amount to the same act: that lifelong omission
of belief . Bu t then sh e would no t com e ou t a s causall y responsible fo r i t i n
Humean terms, so the case would still pose a problem for the weak judgmentalist
Similar case s ar e brough t u p b y Morri s unde r th e headin g o f guil t fo r
thoughts, as reasons against requiring culpability.16 By "thoughts" Morri s ha s
in min d mainly wishes, bu t i t is worth notin g that th e strengt h o f suc h cases
162 Sensibility an d Standpoints

depends on assigning them somethin g lik e the dispositional structure of belief

in the cases just considered. Actively entertaining a wish, that isa n occurrent
thought o f the sort tha t typically involves an ac t of attentionmay be volun-
tary and hence avoidable. At any rate, it is subject often enoug h to some volun-
tary control tha t m y nonjudgmentalist interpretation o f guilt as involvin g an
"as if" attributio n of responsibility seems to fit. On the other hand, the sort of
evil wish about someon e that an agen t migh t discover i n himselfperhaps as
something h e ha s carrie d aroun d fo r yearsma y also b e a n objec t o f guilt ,
whether o r no t har m actuall y befall s the other . Sinc e the proble m jus t note d
about distinguishin g cause fro m effec t woul d see m t o appl y to it , though , i t
would see m to count agains t wea k judgmentalis m in the way explained wit h
reference to beliefs .
An alternative strategy for handling the various cases of guilt without agency
that hav e surfaced so far migh t involv e appeal t o a notion o f substantial self-
causation that i s broader than agent-causation. Agent-causation is supposed t o
be limited to free acts, but perhaps some other things about me are importantl y
attributable to me in a sense that appeals neithe r to some distinct event involv-
ing me nor to my agency. To say this would not be to deny that there are event-
causes to b e found for character traits, beliefs, and dispositiona l wishesany
more tha n fo r ou r statu s i n relatio n t o othe r member s o f the variou s group s
that defin e us . But just because of the wa y suc h notions a s character ar e con -
structed, thi s consequence of determinism might be held to be compatible with
the view that certain traits an d patterns o f action, includin g especially mental
action, ar e attributable in a special sens e to th e self . They help define it s indi-
vidual nature, one might say, along with the sorts of morally self-defining act s
that surface d in chapter 4, bu t i n contrast to passive processes lik e the work -
ings of the digestive tract tha t do not express some basic property of the agent ,
though they may admit of indirect causal control .
This move beyond agent-causation would not solve all problems for the weak
judgmentalist view , however . O n Taylor' s causa l versio n o f th e view , fo r
instance, only shame, not guilt, is thought t o make sense as a "vicarious" emo -
tion, in response t o acts ascribed to othersfor the behavior of one's children,
say, or one's fellow citizens. 17 An d it is important t o note that the view entails
not jus t that vicarious guilt is irrational bu t tha t i t is conceptually impossibl e
without at least a deluded judgment of indirect causal responsibility: the belief
that one' s failing s a s a parent, for instance , mus t hav e been responsible fo r a
child's later misdeeds.
Intuitively speaking, however, suc h cases seem to b e even more common,
if anything , tha n thos e resemblin g th e acciden t cas e wit h whic h I began .
Whether rationall y or not , w e sometime s fee l a s i f we wer e responsible fo r
the acts of others whose doings would b e said to reflect on us just by virtue of
common grou p membershi p without reall y acknowledging an y causal con -
nection. A case in point migh t be white American guilt about slaver y or guilt
felt for various other national misdeeds that occurred before the birth of those
who fee l guilty.
Unavoidable Guilt 16 3

At least sometimes, moreover, suc h feelings see m to b e accepted a s appro-

priate or reasonable. Morris devotes considerable attention to cases of guilt felt
in respons e t o th e act s o f one's natio n a s cases i n favor o f hi s ow n account ,
which migh t be set up a s an alternativ e version o f weak judgmentalism. 18 In
place of a judgment of responsibility in the sense that involves culpability, Morris
appeals to our identificatory ties to other group members. Even without a judg-
ment abou t th e latter , though , i f we allo w fo r emotiona l reaction s base d o n
imaginative identification with others , w e can mak e out vicariou s guilt as n o
more problematic than a host o f vicarious emotions, including the empatheti c
reactions to harm don e to others that underli e ordinary guilt on the account I
gave in chapter 3 .
Taylor's shar p contrast betwee n guilt and sham e on this question actually
seems to rest on a misunderstanding of the notion of vicarious emotion. Sham e
is commonly felt for acts of others in a way that is not true of guilt. But strictly
"vicarious" sham e would amoun t t o sham e fro m th e other's imagine d stand-
point; it is "for" anothe r agen t in the sense of being felt o n hi s behalf . If I feel
shame in response to the misdeeds of my brother, say, normally my feeling will
not b e vicarious; what I am ashame d o f i s something connecte d t o mysel f in
real-life terms. My own status is assumed to be diminished by what my brothe r
does even without the assumption that I am responsible for it but just by virtue
of the fac t tha t statu s is partly a product o f interpersonal ties.
Identification o f cours e play s a rol e i n determinin g interpersona l ties .
Whether I fee l shame , fo r instance , a t th e misdeed s of a colleagu e i n ethic s
depends in part on the importance others assig n to our common categor y bu t
also o n m y own tendenc y to grou p mysel f with colleagues in my field. How -
ever, a simila r point applie s to guil t in case s i n which on e feel s guilt y abou t
activities attributable to a group to which one belongs: family, profession, nation,
race, o r world . I n fact, there i s a furthe r way o f generating vicarious guilt in
such cases that is worth distinguishin g from the usual mechanism of identifica-
tion wit h othe r individuals . On th e assumptio n tha t responsibilit y can some -
times be assigned to a group considered a s a whole, not just to its other mem -
bers, what on e feels ma y depend o n a general kin d of group identification , or
imagining oneself as participating in the group's collective actions. "Collective "
guilt i n this sense amounts t o th e feelin g o f guilt for involvemen t in a case of
collective responsibility.19 It is a prime case, though not the only case, of vicari-
ous guilt, but it seems to rest on the same general identificatory mechanism that
can also operate i n extending the bases for shame .
What is the function o f guilt in such cases? It can sometimes just amount t o
an unavoidabl e cost o f basin g individual identit y on grou p membership , a s
Morris wants to say. 20 However, one can often manag e to drop out of a group
in imagination. Particularly when the group act s in ways one is fundamentally
opposed to , it might seem unclear what the point is of feeling guilty on its behalf.
Why should I punish myself for something I would never dofor wars and witch
trials I know I would have resisted? Doe s the prid e or othe r pleasure I take in
group involvemen t at othe r time s somehow commi t m e to feelin g guilt y i n a
164 Sensibility an d Standpoints

case where the group goes astrayas a matter o f emotional "logic " or perhap s
as a kind o f recompense fo r the benefit s o f group membership?
Morris's comments elsewhere suggest grounds of general moral solidarity, 21
but i t is important t o not e tha t th e acceptance o f guilt as appropriate i n such
cases falls short of a strict requirement to feel guilty. I may be required on moral
grounds t o fee l something fo r th e America n internment o f Japanese civilians
during World Wa r II , say; and perhap s m y feeling shoul d in some way reflect
my membership in the group tha t committed th e crime. However, i t is unclear
why its very content mus t reflect that fact. Assuming that I am myself unlikely
to participate i n such collective acts or even to allow them, it might be enough
just to feel sympathy and outrage, perhaps to a degree augmented by my ties to
the event. The point i s to detac h myself , after all .
However, I think we can make out vicarious guilt as having a point, even if
not as specifically required, in just such terms. In cases where the identificatory
bases of the emotio n admi t o f control, i t can b e seen as a way of clearing one-
self of involvement and at the same time expiating the deed on behalf of others
by a kind of ritual self-punishment. One identifies with the perpetrator, whethe r
a group on e belong s t o or som e other individua l member of it, simply to dis -
tance oneself and the group as a whole symbolically from th e deed by submit-
ting in emotional term s to the punishment the deed merits .
Identification with the victims, that is, does represent an alternative way of
exhibiting moral solidarity in response to the case. What distinguishes the func -
tion of guilt is its self-punitive aspect as a negative self-directed emotion. Thoug h
sympathy and outrage may involve an initial layer of negative feeling, they need
not feel bad overall in the way that guilt does. I take it, then, that my nonjudg-
mentalist account serve s to supplement Morris's appeal to identificatory ties in
explaining th e distinctiv e role of guilt in suc h cases . Th e "a s if " feelin g tha t
one is morally responsible for a wrong and therefore deserves punishmentat
any rate, the emotional self-punishment of guilt feelingsseems to make more
sense here in intuitive terms, moreover , than Morris's denial that guilt in such
cases counts as moral. Morris's distinction between moral and nonmoral guilt
allows for a univocal account of the emotion insofar as moral guilt on his account
is understood t o rest on separation fro m the moral community.22 But by a kind
of transitivity of identification, this would seem to make vicarious cases of moral
guilt come out as moral too .
My suggested moral-but-nonjudgmental account of the various cases Morris
considers seems to me to be compatible with the substance of his view; it rejects
only his classification of separation guil t as necessarily nonmoral. I do recog -
nize cases o f nonmoral guiltguil t about goin g of f one's health regimen an d
similar examplesin which the rule s the agen t violate s are in fact nonmoral ,
though h e treats the m i n emotional term s a s if they had mora l force . Bu t the
distinction does not turn on the issue of culpability. The hardest cases to inter-
pret plausibl y as moral might seem to b e cases o f survivor's guilt. These ma y
involve a n undeserve d benefit tha t th e agen t clearly did nothin g to gai n an d
can do nothing to mak e up for; so on the assumption o f basic rationality, one
might want to question how he could see himself even in "as if" term s as mor-
Unavoidable Guilt 16 5

ally responsible. Morri s in fact suggests something of the sort himself, though,
when he includes these and similar cases in his discussion o f guilt about "unjus t
My own proposal for handling such cases can be strengthened b y appeal t o
the self-referential character of guilt. To the extent that guilt functions as a kind
of emotional self-punishment , that is, it goes some way toward fulfilling it s own
characteristic desir e fo r reparation. Bu t in that case , th e failur e to fee l guilt y
counts as a possible object of guilt itself. This effectively redouble s the motiva-
tional force of anticipatory guiltguilt about one's failure (s o far) to make up
for a wrong , a s th e basi s fo r a forward-lookin g varian t o f th e emotio n tha t
Morris also want s t o recognize. 24 O n m y account o f it in terms o f an "a s if "
feeling, the emotion i s sometimes self-generating in anticipatory form : One feel s
as if one already has done something wrong simpl y by failing to fee l guilty (or
guilty enough) yet.
This way o f compounding guil t may seem a fiendish trick. In fact, I think
it helps explain som e of the pitfalls of the emotion, i n particular it s obsessive
or unappeasabl e qualit y in many cases. Bu t can i t serv e a s a foundatio n fo r
guilt, or mus t on e assum e a more basi c negative self-evaluation as a reaso n
for th e emotion? I n cases of guilt for an undeserved benefit, I think we might
well begi n with anticipator y self-referentia l guilt, i f we appea l t o a deman d
for "leveling" : a requirement that th e agen t brin g himself down to th e leve l
of others i n a group wit h which he identifies, if only by subjecting himself t o
emotional self-punishmen t for exceedin g th e norm. What he should punis h
himself with , accordin g t o this suggestion , i s discomfort a t th e thought tha t
he has done something to deserve punishmentthough all he really may have
done i s to fai l to inflic t i t so far. Indeed, som e demand o f the sortrequirin g
in th e firs t instanc e unconsciou s self-referentia l guilt, whic h the n ma y b e
masked b y guilt with som e othe r (perhap s indefinite ) objectma y see m o n
occasion t o be imposed b y others as a condition o f participation i n the group s
to whic h w e ar e boun d b y mutual identification . Breakin g awa y fro m th e
family, for instancewhether b y one's own efforts or by the death of parent s
or other misfortunes of family members that do not befal l oneselfis a prime
source of separation guil t and ca n sometimes b e encouraged b y family mem -
Morris's discussio n bring s u p resentmen t an d indignatio n a s appropriat e
reactions on the part of others (alon g with self-reproach on the part of the agent
as involved in feelings of guilt) only in connection with moral guilt. 25 But others'
reactions in nonmoral case s may also include other-directed form s of blame of
the sort that demand guilt of the agent, sometimes in no less fiendish forms than
the one suggested. Part of showing that one identifies with others in a way that
makes inequalities unwelcome involves the willingness to make up for inequali-
ties with self-inflicted emotiona l distress. But this is an unachievable aim in many
cases, an d accordin g t o th e accoun t I have offered, i t i s based o n a n illusory
feeling of responsibility. I now want to argue, however, that there are some cases
in which guilt may b e rationally appropriate eve n without adequat e ground s
for other-directe d blame.
166 Sensibility an d Standpoints

2. Perspectiva l Appropriateness

Understanding the grounds on which we take some cases of guilt to be accept-

able in rational terms while dismissing others as unreasonable or unwarrante d
"emotional reactions " mean s taking at least a brief look a t the general notion
of emotiona l appropriateness . Th e ter m "appropriate " i s semitechnical: a n
artificially regimente d refinement of our loose r an d mor e varied talk i n com-
mon speech of particular emotions as reasonable or unreasonable in "backward -
looking" termsas fitting the agent's situation, accurately representing its salient
or significant features t o th e exten t that h e is aware o f them. I t is a notion of
justification relativ e to reason s rathe r than a truth-value of emotions, bu t i t is
meant t o serve as their primary representationa l value in a sense that implies
adequate performance of representational function.
Emotions are taken to have a representational function on the sort of account
I favor insofar a s they are mad e out i n terms of evaluationsmeaning evalua-
tive thoughts, possibl y unasserted. I take these a s belieflike state s o r preposi -
tional attitudes with a content give n by an evaluative judgment that might be
thought o f a s corresponding t o th e emotio n i n question , thoug h i t nee d no t
always accompany the emotion. The evaluative thoughts may not be connected
logically to our other settled attitudes, that is, in the way that serves to pick out
judgments as thoughts expressing belief , or attitudes of assertion; they are held
in mind in a more temporary way , but one that allows them a significant influ -
ence on behavior . Normally they are objects of attention, thoug h th e accoun t
also allows for unconscious emotions as cases in which the agent misidentifie s
the content o f the evaluation . What hold s the content i n mind i n the relevant
sense is emotional comfort o r discomfortsomething that can also be seen as
having a representational functio n to th e extent tha t i t reflects the positive o r
negative aspect o f an associate d evaluation .
It is essential to this account tha t comfort an d discomfort are taken a s gen-
eral intentiona l statesstates o f positiv e o r negativ e affect directe d towar d
evaluative propositionsrather tha n amounting merely to affective symptom s
of emotional evaluation, a s in the usua l list of sensory states characteristi c o f
emotions. Thus, guilt amounts to discomfort at or about the thought of oneself
as responsible for a wrongnot jus t to the thought plus an accompanying pang
of discomfort. On the other hand, it is also important that in justifying guilt all
we have to justify , a t leas t in the firs t instance , i s something on th e orde r o f a
"pang" of guilt: a momentary stat e of discomfort tending to convey attention
to its evaluative object in more o r less immediate terms rather than the sort of
settled behavioral tendency linked with belief .
Further, justificatio n of th e emotio n make s referenc e to it s motivationa l
function, with discomfort seen as a state from which an agent would naturally
want to escape.26 In justifying guilt, what we are justifying is essentially an escape
tendency: to change one's affective stat e b y acting to falsif y it s evaluative con-
tent, o r th e self-attributio n of responsibilit y for a wrong, see n a s imposin g a
requirement of reparative action. That is, the point of the emotional reaction is
to set up a need to act in certain ways that make up for and in some sense miti-
Unavoidable Guilt 167

gate responsibility. But I take this to mean that the representational rationality
of guilt (along with other emotions) will be properly influenced b y instrumen-
tal considerations: The standards of emotional appropriateness will be adjusted
to reflect fact s about the general practical adaptiveness of a given emotion ten -
dency, or its value as a motivator. Fo r guilt, my account will have the effec t o f
extending the emotion beyond the corresponding judgment of faultand beyon d
other-directed emotional blame, as we shall see, for an asymmetrical treatment
in rational terms of guilt and persona l anger .

Emotions and Cognitive "Fit"

The account just outlined might be said to mak e out emotion s a s "evaluative
affects." Comfor t o r discomfor t an d th e though t towar d whic h it i s directed
are not reall y separable parts of emotion, bu t I still speak of them as affectiv e
and evaluativ e "components," wort h distinguishin g conceptually in order t o
concentrate on the latter in considering justificatory questions. My strategy for
considering emotional appropriateness involve s packing the qualitative content
of an emotion into its evaluative componentincluding its positive or negative
aspect, eve n though this might seem to repea t informatio n given b y the affec -
tive componentto permit a n analog y to belie f warrant, o r evidential justifi-
cation fo r th e correspondin g judgment. 27 The centra l ai m o f thi s approach ,
though, i s to brin g into sharper relie f som e o f the limitation s on the analogy .
Let me try to capture thes e now a t leas t roughl y by appeal to a distinction in
current philosophy of mind between "direction s of fit."
The distinction arises in Searle's account of intentionality as a point of com-
parison betwee n belief s an d desire s (alon g with analogou s speec h acts ) tha t
applies onl y in a derivativ e sense t o emotions. 28 A belie f exhibit s "mind-to -
world" fit insofar as it is supposed t o correspond t o the world. By contrast, the
point of a desire is to get the world to conform to it; its direction of fit is "world-
to-mind." Thus , in the event of a disparity, a belief counts as deficient in fulfill -
ing its function, o r in the usual terms false, whereas the onus normally falls o n
the world (includin g the agent) for failin g t o satisf y a desire.
Emotions do not see m to fal l neatl y int o either category. They ar e assess -
able in rational terms as appropriate o r inappropriatewhat I take to provide
the analogy to belie f warrantbut thes e term s normally build i n reference to
the adequacy of the reasons for an emotion, not just the sort of fit to the worl d
that may be accidental. We may think of such reasons as facts about th e agent
and henc e as include d in "th e world, " a t leas t i f we tak e the m a s limite d t o
noninstrumental o r backward-lookin g reason s tha t ar e assumed t o charac -
terize things as they are independently of the mental state in question to allow
for ou r initia l contrast between states o f belief and desire . A desire is also sup-
posed to fit instrumental or forward-looking reasons, that is, insofar as these
count a s reasons for thinking its satisfaction good. This i s not fitnes s to th e
world, though , i n the sens e that applie s to beliefs . I n an y case , emotiona l
appropriateness still resists characterization in these terms. Even assuming that
it is rational appropriateness that i s in questionas distinct from th e various
168 Sensibility an d Standpoints

moral an d quasi-mora l o r socia l version s o f th e notio n t o whic h on e migh t

appealit may not b e clear that an emotion tha t fail s to fit the world i s ratio-
nally deficient .
Our assessmen t o f states o f sadness, fo r instance , t o star t wit h a simpler
example than guilt, will not be based solely on whether they correspond to the
facts o f the case as measured by a single evaluative standard. The furthe r fac t
that sadnes s tends to inhibit corrective actio n make s it a sometimes unhelpfu l
response to misfortune. For practical purposes, then , we may be justified in with-
holding the emotional response in circumstances that warrant it. But this point
also affect s ou r vie w of things fro m a representationa l standpoint : Forgoin g
sadness is taken as rationally appropriate i n a case in which sadness would also
be appropriate, so that we have an important contrast to the assessment of belief
On certai n assumptions , th e sam e ma y b e said o f substituting a contrar y
emotion suc h as joy. Just to be happy about the same misfortunes described in
the same way would of course be unreasonable; a certain caricature of an East-
ern mystic might be able to manage it, but she would presumably be sacrificing
reasonableness t o somethin g else . What i s not s o unreasonable , or unusual ,
though it also requires some mental self-trickery, is to focu s attention instea d
on a different perceptua l "slice " or cross-section o f the information available
in order to generate a positive emotion with preferable motivational effects. T o
the extent that the feeling thus generated involves attention to a thought that is
reasonably picked out from the background of evidence for evaluation, the emo-
tion wil l count a s appropriate. Th e point i s that this justification rests o n th e
assessment o f a mental actan ac t o f attentionan d henc e is quite properly
influenced b y practical considerations .
What emotions are supposed to represent on this account is the importance
of certai n piece s of informationreasons o f a certain sort, including reasons
for attention, rather than directly "the facts " i n some distinct sense. But atten-
tion is another elemen t of cognition beside s belief an d is also subject to a kind
of evidential assessment. Fitnes s to th e salient facts i s enough to warrant even
a maladaptive emotion. Accordingly , we may say that a n appropriate evalua -
tive objec t of affectwha t my accoun t represent s a s the conten t o f a n emo -
tionamounts t o on e that o n a reasonable vie w of things stands out agains t
the background of evidence. Gestalt shifts are possiblethere may be more than
one reasonable view of thingsso the notion is a tolerant one. It also has to be
left somewha t fuzz y t o correspond a t least roughly to our intuitive judgments:
What constitutes a reasonable subset of the evidence to pick out for attention
a reasonable perceptual "slice, " as I put itwil l often b e a matter of debate. It
is subject to mora l an d socia l norm s i n a way tha t allow s fo r som e degre e of
relativity to time and loca l convention. Consider , fo r instance, how we would
classify sadness in the middle of a happy event at the thought of man's ultimate
death and decay. The thought makes sense in a certain kind of religious culture
but unde r other circumstance s might be a matter fo r psychiatry.
The variability of the cases I have cited should indicate that we cannot expect
an account of emotional appropriateness to provide firm answer s to the many
Unavoidable Guilt 16 9

questions such cases raise. What I hope to provide is just an explanation of some
of th e relevan t questions . Ou r reaso n fo r picking out "slices " of the evidence
for purposes o f emotional reactio n (and hence for its assessment) has to do with
the function of emotions in directing attention: They serve to hold in mind rea-
sons of potentially immediate or isolated significanceby contrast with evalu-
ative belief , which essentially involves putting the sam e sort s of proposition s
into storage, connecting them with our other beliefs as settled response tenden -
cies. A s I noted earlier , this means that w e nee d onl y justify somethin g shor t
term i n justifyin g a n emotiona t an y rate , i f we limi t ourselve s to it s basi c
qualitative justification, ignoring many important question s o f degree.
The sam e point als o seem s to hel p explain wh y warran t fo r emotio n an d
belief will be affected differentl y b y the practical adaptiveness or instrumental
value of a given evaluative thought. On e should expec t emotio n with it s basis
in attentio n t o b e sensitive , most notably , t o fact s abou t th e usefulnes s o f a
thought i n motivating action. A s a quick response i n many cases, however , i t
cannot res t o n any sort of calculation of consequences in the case at han d bu t
at most ca n come t o registe r the general value of the kin d of thought i n ques-
tion. A belief, on the other hand, need not b e borne in mind, and the results of
including a certain evaluatio n among the objects of belief will be more unpre -
dictable insofar as beliefs are both less present to consciousness in standard case s
and longe r term . At any rate, as a normative matter, tie d to our acceptance of
representation a s its defining function, we exclude considerations of adaptiveness
from ou r weighin g of the evidence for belief. This amounts t o assignin g belief
its straightforward mind-to-worl d directio n o f fit . My alternativ e suggestio n
for emotion s migh t b e summed u p a s a clai m tha t emotion s exhibi t variable
mind-to-world fi t to the extent that their representational function incorporate s
reference to forward-looking practical reasons along with the sorts of backward-
looking reasons tha t constitute evidence. The standards o f evidence are raised
or lowered , tightened o r relaxed , i n ligh t o f facts about th e genera l practica l
adaptiveness of a given response tendency .
The resulting picture of emotional justification ties emotions to somethin g
more like perceptual standpoints than a unified conception o f "the world." The
range of allowable standpoints will be limited by facts of the sort that properly
govern objects of attentionfacts about the general significance of certain pieces
of informationbut since these are evaluative, they presumably do not fit into
"the world" in the sense intended. Even if one attempted t o include them, how -
everas a kind of "atmosphere" to the world, saythe point for our purpose s
is that one would hav e to allow for multiple worlds (or a world with multiple
overlapping atmospheres) to capture emotional appropriateness .
The notion of appropriateness tha t this account yields may be thought of as
"perspectival" becaus e it allows for the assessmen t o f emotions i n relation t o
particular subsets or perceptual slices of the total body o f evidence, as well as
in relation to the evidence overall. It rests on the general point that what we are
justifying in justifying an emotion is essentially an affectivel y mediate d tendency
to direc t attention rathe r than the sor t o f settled respons e tendency that i s at
issue in assessments of belief. The notion allows for appropriate emotion in cases
170 Sensibility an d Standpoints

of what I call "snap " evaluation, where we jump to a n emotiona l conclusio n

before all the facts are in, as well as in other "parti-resultant " cases , with emo-
tional deviatio n fro m warrante d belie f assume d t o reflec t the genera l signifi-
cance of some part o r aspect o f the evidence. For guilt, I now wan t t o say, the
relevant aspect may amount to something like one's own involvement in a vio-
lation of the moral norms , eve n if the involvement falls short of moral respon -
sibilityand even if the violation is not quit e a wrong acti n contrast to th e
judgment that one is guilty.
My thought is that in general terms the subset of the evidence for the corre-
sponding judgment to whic h one reacts i n these cases is important enoug h t o
justify a t least an initial outlay of attention. This is what i t means to b e "sensi-
tive to" th e mor e ramifie d set of considerations that woul d groun d th e judg-
ment. There ar e furthe r reasons i n the cas e of guilt for thinkin g that uncom -
fortable attention , i n particular , migh t b e warranted a s a wa y o f controllin g
future behavior , or o f making up fo r past behavior, if only by subjecting one-
self to guilt. Guilt seems to be morally useful to us as a motivator, in fact, largely
in anticipatory or symbolic forms that depend on perspectival appropriateness .
In cases where the emotio n i s warranted b y serious past wrongdoing o f one's
own, adequat e reparation is often jus t not possible. For that matter , it is ques-
tionable how often the emotion is felt in such casesapart fro m cases in which
the agen t unwittingl y failed t o se e things i n the requisit e light a t th e tim e of
action, to appreciate the wrong done or his own involvement, and hence might
seem to lack the fulles t sor t of moral responsibility . The point applies to case s
of collective responsibility, in which the main moral failing on the part of many
agents is likely to be just the sort of failure to pay attention that emotional sen-
sitivity on the account I have outlined is supposed t o prevent. It does so, if I am
right, primarily by focusing attention on something short of adequate evidence
for the corresponding belief .

Detaching Guilt from Blame

I now wan t t o us e the notio n o f perspectival appropriateness t o explai n wh y
guilt should be appropriate i n cases in which blamepersonal anger, its third-
person counterpar t i n emotional termsis not.29 Guilt functions as a general
way of keeping oneself alert to significan t subset s of the available evidence, in
the first instance for their potential bearing on future moral action. We can there-
fore justify the emotion a s a reaction to subsets of the evidence that do not simi-
larly warrant persona l anger. It would b e unjust, for instance, to blam e some-
one personally in a case of collective guilt in which the agent himself did nothing
wrong. The perceptual slice that he might reasonably focus onthe thought of
the dee d a s chargeable to a group i n which he claims membershipdoes no t
count as adequate reaso n fo r someone else to inflic t emotional punishment on
him. The ai m of keeping him on hi s toes, morally speaking, would no t justif y
the kind of treatment that one might reasonably accord oneself in a similar case
or demand of him.
The reason is partly just that on e is oneself, with more extensive sway over
Unavoidable Guilt 17 1

what on e may do to onesel f in emotional (o r other) terms. Guilt also serves a

positive function for onesel f t o th e extent tha t self-subjectio n to th e emotio n
counts a s a kind o f self-cleansing ritual: a way o f clearing oneself o f involve-
ment in wrong by emotional self-punishment. One would not similarly be ben-
efiting anothe r person , helpin g to clea r hi m o f involvement , by blamin g him
for the wrongs done by others in a group to which he belongs. To do so would
be to indulg e in a menta l variant of scapegoating tha t subject s him t o unde -
served punishment on others' behalf.
This brie f argumen t rests o n understandin g emotion a s essentially evalua-
tive, with negative personal evaluations seen as unpleasant for their objects, even
if their objects do not undergo the associated elemen t of unpleasant affect . Th e
argument does not rest, however, o n equating emotional blam e with blaming
others overtly : There ar e o f cours e al l sort s o f mora l o r socia l reason s fo r
restraining oneself from over t act s o f blame, but eve n if we limi t ourselves t o
mental acts, blam e requires a stronger justification tha n guilt. Besides the sort
of mora l reaso n jus t given , th e asymmetr y turn s o n th e wide r se t o f moti -
vationally adequate emotional alternative s to blame . In addition t o th e avail-
ability o f a collective object o f blame , w e ough t t o tak e not e o f a structura l
difference betwee n blame and guilt that explains why I specify "personal " anger
as the third-person counterpart o f guilt corresponding to blame. For anger also
has a n impersonal variant, frustration , whereas guilt is by definition directed
toward th e self. 30 In combination wit h m y perspectival accoun t o f appropri -
ateness, the availability of a less personally punishing alternative for anger means
that th e standards for justifyin g it s personal varian t will be higher.
Frustration count s as a deficient instance of anger, as I put iti n contrast
to th e full-fledge d case s tha t represen t th e most full y elaborate d instance s of
the emotio n i n adult life . S o there i s something els e to tur n t o whe n th e evi-
dence is not adequate for blame. There may also be deficient instances of guilt
passive cases, say, without the characteristic desire for reparationbut my point
here is that these will not b e deficient in quite the same way as impersonal an -
ger. They will not be directed toward something mor e diffuse tha n their defin -
ing personal object, the self; otherwise, they would amount to something othe r
than guilt, perhaps frustration or perhaps a form of sadness with an indefinit e
object (a s on som e account s o f depression). By contrast, anger i n the sens e of
generalized frustration, directed toward n o one in particular, provides us with
an emotiona l option t o perspectiva l personal ange r or blame , ange r tha t tar -
gets its object on an inadequate evidential basis for the corresponding judgment
of fault. So the standards for imposing blame will not b e relaxed a s they are for
guilt (along with other emotions), assuming that the general practical function
of anger would b e served well enough by frustration.
There are , o f course , case s i n which emotiona l blam e withou t adequat e
evidence for a judgment of fault is of psychological or other benefit to the agent
in the sense of the person undergoing the emotion; cases of scapegoating might
be thought to be chief among them. However, this does not affec t th e justifica -
tion o f personal anger in it s representational function; w e stil l spea k o f anger
as unwarrante d in suc h cases. The sort s o f fact s tha t affec t ou r weightin g of
172 Sensibility an d Standpoints

backward-looking reason s fo r th e emotio n hav e t o d o wit h it s general

adaptivenessin th e firs t instance , it s energizing function i n motivating cor -
rective action. Frustration involves a generalized urge to lash out, correspond -
ing to it s diffus e object , that shoul d serv e this en d a s well as personal ange r
when the evidence does not pick out a specific object. In any case, it is not just
benefit t o the agent that i s in question i n considering the general adaptiveness
of an emotion. To the extent tha t persona l anger has a role to play as a moral
corrective, by inflicting emotiona l punishmen t for injustice , an y instrumental
benefits o f scapegoating would b e undermined by the fac t tha t i t punishes an
innocent party.
Guilt seem s t o b e different , assumin g that w e hav e a righ t to scapegoa t
ourselvesto sacrifice ourselve s to some ai m of our owna t any rate, up to a
point. The emotion has a positive function in defense of moral self-esteem that
would not b e adequately served by other-directed emotions like frustration, at
least in general terms: One might indeed recommend some other-directed sub-
stitute on many or even most occasions i n the manner of various self-help manu-
als, but it is general adaptiveness that on my account lets us loosen our eviden-
tial demands on an emotion. This is meant to b e broadly construed, to exten d
to emotion s tha t offe r som e uniqu e benefits to th e agen t o r jus t to peopl e i n
general, even if they also inflict harms, perhaps even greater harms. Remember
that perspectival appropriateness i s taken a s compatible with the reasonable-
ness of suppressing a given emotion.
Let u s now attemp t t o se e what happen s t o som e cases of guilt when w e
apply the notio n o f perspectival appropriateness. First , conside r th e case s of
collective guil t that hav e surfaced at point s i n m y argument . My accoun t o f
appropriate emotion allows us to grant that no one should be blamed solely for
the crimes of others: It would b e unreasonable to hold someone responsible in
a backward-looking sense (as opposed t o insisting that she take responsibility)
for what her nation or other group may have done in her name.31 O n the other
hand, we may also sa y that sh e ought t o hol d hersel f responsible in the wa y
that is evidenced by emotion. Sh e may b e blamed if she fails to manage this
at any rate, in some form; there are emotional alternatives to guilt, albeit more
limited than thos e that appl y to blame .
The fac t tha t guil t i s appropriate, i n the sens e of "rationall y acceptable, "
will no t b e sufficient t o mandat e tha t reaction , as I have noted. I t does com e
out a s appropriate, though , i f we suppose tha t the agent' s involvemen t in the
group that counts as the collective agent in the case naturally sets up enough of
a lin k betwee n he r ow n agenc y an d th e ac t i n questio n t o justif y directin g
momentary discomfort toward a thought o f herself as responsible fo r it . Th e
thought may just be brought to mind with a question mark and rejected as soon
as she has a chance to reflect on it , but i t is appropriate a s an object of uneasy
attention a s long as there is reason fo r he r to rais e the questio n o f her mora l
responsibility. A crime committed by a group in the remote past, say, and clearly
excluded from its future behavior by changes in group character over time would
not naturally raise the question of responsibility in this sense, though of course
a lin k could be manufactured.
Unavoidable Guilt 17 3

The considerations relevan t t o decidin g such issue s are complex, an d th e

relevant concept s ar e properl y fuzzy . I shall b e conten t jus t to sketc h thei r
application to the range of cases at issue. Conside r no w the case in which one
feels guilt y for havin g caused the deat h o f a child in an unavoidabl e car acci -
dent. A central cause of guilt feelings in such cases, o r those we are inclined to
accept a s appropriate, i s just the agent's nee d to reassure himself that he is not
guilty. That is, after the accident the driver of the car that killed the child would
naturally feel impelle d to run throug h what happened, considering vividly his
own participation i n it, if only to b e sure that he did not d o anything to cause
it. He would do this by confronting himself with various particular subsets of
the total bod y of relevant evidence that considered i n themselves would tend t o
suggest that he is guiltyand to call up feelings o f guilt as they come to mind,
given the responsiveness of emotions t o partial evidence. In each case he would
be able to answer any self-accusation by bringing to mind further evidence. This
would presumably allay guilt, though i t would no t prevent the reaction, an d it
would no t prevent the reaction fro m recurring, at any rate for some time afte r
the accident : A morally sensitiv e person i n a case o f such moral consequenc e
could b e expected t o review the facts repeatedly an d i n rather exactin g detail ,
focusing first on the memory of this or that ac t of his that might be picked ou t
as a cause of the accident, the n searching again for the reasons fo r thinking its
results could not hav e been foreseen .
The need to clear oneself of real or objectiv e guilt can in this way generate
guilt feelings . Where the nee d is real, though, i n the sens e of being supporte d
by more tha n th e fact s o f the agent' s individua l psychology, th e feeling s may
well come out a s appropriate. I t is not irrelevan t to th e case under discussio n
that the victim is a child, someone an agent is supposed t o take special pains to
look after , so that i t is reasonable t o assume more tha n the normal amoun t of
responsibility even after-the-fact, and blame cannot easily be shifted to anothe r
agent. The agent reviewing his contribution t o the case may focus, say, on the
fact that he turned the car sharply in a certain directionthe n remind himself
that it was too dark to see the childor he may focus on things he could have
done differentl y whos e connection to the case is more remote. Goin g out at all
that night or going out with a trivial purpose migh t well come up as reasons fo r
self-reproach (no t just regret), at leas t unti l the fact s of the cas e mor e o r les s
settle in mind. My suggestion here is not that they should settle into a judgment
of guilt but that emotional guil t properly precede s adequate evidence for such
a judgment. Even when al l the evidence is at hand, guilt will be appropriate i n
reaction t o partial subset s o f it whose rol e i n producing th e reactio n serve s a
general practical purposein this case, as a way of assuming special responsi -
bility for those unabl e to fen d fo r themselves .
The appropriateness o f the emotion depends , then , o n the reasonablenes s
in instrumental as well as representational term s o f focusing attention o n cer-
tain thoughts, including the sorts of perceptual memories one might well retain
from a car acciden t and othe r bit s of reflective informatio n of a sort that ca n
come t o mind, be banished, an d the n come bac k again. This picture stands in
contrast t o the usual idea of even emotional guilt as imposed on the basis of an
174 Sensibility an d Standpoints

inner set of courtroom proceedings, deliberatin g to a fina l judgment . I would

fill it out, moreover , wit h a parallel in mentalistic terms to the primitive notion
of "tainting " i n which th e agen t struggles to clea r himself of a connection t o
some morall y disturbin g event b y holding i n mind th e variou s thoughts tha t
seem to lin k him t o i t an d severin g any putativ e link b y means o f emotiona l
What will be in question in disputed cases o n th e issu e of appropriatenes s
will normally be the general adaptiveness of guilt feelings in response t o a sub-
set of the overall evidence that constitutes prima facie evidence for a judgment
of guilt. However, in cases of irrational guilt where the emotion i s not base d o n
a natural or reasonable perceptual slice of the available information, we do not
have even prima faci e evidenc e for th e judgment. This i s how I would handl e
the variant of the accident case i n which a memory o f merely passive involve-
mentdriving a car tha t wa s propelle d into th e child by another ca r tha t hi t
itis trumped u p into a recurring thought of oneself as somehow a t fault, just
on the basi s of a neurotic habit of self-accusation.
Such cases are not so odd, nor are they limited to cognitively irrational agents,
if one includes as genuine feelings of guilt the pangs and othe r momentary sen-
sations that might b e felt just in raising the question o f warrant fo r the corre -
sponding judgmen t in cases where responsibility has unclea r boundaries. Th e
heightened responsibility that adults assume on behalf of children yields a fun d
of familiar examples, but one might also consider, for instance, the feelings that
arise as one ask s whether damag e to a borrowed boo k o r a sublet apartmen t
was something on e ought to have foreseen and avoided. In some circumstances
or state s o f mindwhere on e expect s to b e blame d unreasonabl y b y others ,
sayit woul d no t b e unnatural to pa y at leas t momentar y attentio n eve n t o
perverse subset s o f th e evidence , i f only t o rehears e th e argument s agains t
according them prima faci e weight. In general, it is important t o not e that no t
just any emotional respons e that might be made out as normal will come out as
appropriate o n the perspectival account, despite its appeal in standard circum-
stances to th e notio n o f a natural object of attention. 32
Some explanation o f irrational guilt along these lines may thus be applied
to mor e ordinar y case s o f guilt without faul t fro m th e literature . Le t u s see
how i t fit s tw o particularl y problematic sort s o f cases. Consider firs t taboo
cases suc h a s thos e discusse d b y Rawls: case s o f residua l guil t for violatin g
prohibitions fro m childhoo d o r th e surroundin g societ y tha t th e agen t cur -
rently rejects. These resemble the original accident case i n not involvin g self -
attributed faul t (meaning responsibility for a wrong), though unlik e the vari-
ant cas e they do involv e agency. Here, however , th e stres s i s on the denial of
"wrong" rathe r tha n "responsibility. " Taylor' s inclusio n o f taboo case s a s
cases o f genuin e guilt (contr a Rawls ) by appeal t o conventionall y accepte d
authority for the rule in question would see m to make them appropriate cases,
assuming no mistake about th e conventions. One might indeed provoke a pang
of guil t just by remembering or anticipatin g blame on th e par t o f others, s o
the response i s in that sens e natural. However, o n my own accoun t the emo -
tion would stil l come out a s inappropriate in a case in which the convention -
Unavoidable Guilt 17 5

ally accepte d rul e i n questio n jus t sets u p a n artificia l wa y o f sortin g moral

Proscribing somethin g purel y as a matte r o f convention, tha t is , withou t
reference t o huma n harm s an d benefit s or othe r reasonabl e object s o f mora l
attention (i f there are any), does not support th e appeal to general adaptivenes s
that result s i n our mor e relaxed standar d o f evidence for guilt than fo r blame .
The genera l mora l significanc e o f emotio n i n a broade r sensea s a wa y o f
enforcing conformit y t o whateve r rul e is in placewould seem t o b e relevan t
only where moral considerations favo r adherence to the particular rul e in ques-
tion. S o "mere" taboos in the sense of purely conventional prohibition s whos e
validity on e reject s will no t yiel d eve n prim a faci e suppor t fo r guil t feelings.
We have to say more, though, to handle taboos that may not be purely con-
ventional bu t arguabl y have som e basisinadequat e from th e standpoin t of
moral justificationin a natural human response. Examples include fea r of the
stranger a s a possible source o f racial taboos an d th e feeling s of hostility used
to punish deviation fro m standar d se x roles and preferences . O n th e assump -
tion that a certain behaviora l norm ought t o be modified, the feelings that serv e
to enforce it need no t com e ou t a s appropriate jus t becaus e it is in fact widely
accepted. However , rathe r tha n denyin g that guil t in such case s i s based o n a
natural way of sorting information , we can follow the model of blame and den y
the adaptiveness in general social terms of allowing ourselves to react to a cer-
tain portion o f the evidenc e in isolation .
On the other hand , there ar e rules on the order o f the ban on incest that are
commonly calle d taboos but ar e thought o f as essential t o th e foundation s o f
human society, albei t with details subject to convention. Here we may be will-
ing to accept guil t as appropriate eve n in a case like that o f Oedipus, wher e th e
violation is involuntary and hence does not merit blame. This sort of extensio n
of appropriate guil t beyond blam e has to be limited to the aspect of emotional
justification tha t appeals t o the practical reason s fo r attention t o a certain sub-
set of the evidence for the relevant evaluation. The subset itself has to be accepted
independently, that is, as a natural or reasonable object of attentionon ground s
I shall not attemp t t o pi n dow n herei n orde r t o avoi d th e kin d o f practica l
gerrymandering that woul d simpl y manufacture objects of guilt.
If we now turn to survivor's guilt and similar cases of guilt without agency
cases that migh t seem to be distinct in kind, involving the thought o f separatio n
from a group rather than of moral responsibility for a wrongwe ca n ask whether
there is a way of counting them as appropriate withou t multiplying senses of guilt
as in Morris's account. In fact, many of the moral case s just discussed typically
involve a component o f survivor's guilt: The justification of collective guilt com-
monly appeals to benefits received as a result of the harms done to others by one's
group; eve n in the secon d versio n o f the acciden t case, guilt might b e defended
simply fo r survivin g the chil d that dies . I interpret suc h feeling s as involving a
thought of moral responsibility that is not warranted b y the evidence as a whole.
Rather, on the perspectival account the justification for holding it in mind depends
on the practical value of guilt as an identificatory mechanism: its role in promot -
ing the sense of moral solidarity that Morris describes.
176 Sensibility an d Standpoints

This is to say that guilt in the relevant cases functions as a way for the agent
to clear himself via emotional self-punishmen t of any suggestion o f benefit from
being favored above the other members of a group with which he identifies. It
expresses the fact that he does identify with the group by bringing him down to
its level in emotional terms. On the perspectival account, hi s self-attribution of
responsibility also has to res t o n a natura l way o f sorting the availabl e infor-
mation. I think we can sa y that i t does i n serious cases o f survivor's guilta s
opposed to just any case of guilt for the ills of the worldinsofar as the though t
arises naturally as a question i n others' minds where people ar e genuinely "in
the same boat" in some respect or other.33 These are cases in which it is natural
to resent inequality.
However, th e account doe s no t accep t others ' persona l resentmen t o f the
agent a s warranted i n such cases . It matter s t o th e justificatio n of survivor' s
guilt that it is self-inflicted a s a way of expressing identificatory commitments.
Others ar e in no position t o decide for the agent what these are to be; the mos t
they can reasonably do i s demand tha t h e decide a certain way a s a conditio n
of grou p acceptance . Bu t even this doe s no t pic k ou t guil t i n particular: Th e
emotion come s out a s optional, morally speaking as well as rationally, to th e
extent tha t other feeling s suc h as those involve d in humility represent alterna -
tives for accomplishin g the sor t of leveling in question .
The cases I have brought u p s o far give us the beginning s of a continuum of
cases on the question of grounds fo r guilt short o f beliefand th e beginning s of
an argument from appropriat e guil t feelings to a kind of objective guilt suitable
for dilemmas . What varies in the cases is the strengt h o f the reason w e have for
attention t o the prepositional objec t of discomfort involved in guilt feelings: the
thought tha t the agent is responsible for a wrong. I n the original version of the
accident case, but not in the variant case or the typical taboo cases, we consider
it reasonable as well as understandable fo r someone i n the agent' s positio n t o
raise the question of guilt. Appropriate case s of collective guilt and survivor's guilt
seem to involve something stronger, however: Unlike either of the accident cases,
they are cases in which the question of responsibility also naturally arises in other
people's minds. There is a social as well as a nonidiosyncratic psychological basis
for th e emotion, tha t is a basis for claiming that the agent really is tainted by
what his society has done in his name, say; or that there is a real need to make up
for inequalit y of the sort that undermines group solidarity. The other case s that
Morris emphasizes, such as guilt for thoughts, see m to me to vary with the par -
ticular circumstances in respect t o where they fall o n the continuum. But I now
want t o extend the continuum in the direction of greater strength of reasons fo r
attentionto case s in which there i s reason no t jus t for raising the questio n o f
responsibility but also for reaching a positive conclusion .

3. Objectiv e Guil t an d Wrong

Moral dilemma s seem to b e cases of appropriate moral guilt without ful l mora l
responsibility. They are cases in which the agent cannot avoi d doing something
wrong, even if he can avoid the particular wrong he chooses to commit, since his
Unavoidable Guilt 17 7

choice situatio n offer s hi m onl y wron g alternatives . S o the case s d o involve

agencythey even require specifically moral agencybut they seem to count as
cases of guilt without fault , versions of the original accident case distinguished
by the agent's appeal to moral reasons for what he does. I take them to be cases
in which each alternative is prohibited on serious enough grounds to make it come
out as wrong all things considered despit e the lack of permissible alternatives to
it. With th e notio n o f perspectiva l appropriatenes s o n han d an d defende d in
intuitive terms fo r nondilemmati c guil t without fault , w e can se e dilemmas as
cases in which emotional or subjective guilt is appropriate for either alternative.
The thought is that the reasons for prohibiting a given alternative form a morally
significant subse t of the informatio n bearin g on a dilemma: a natural object of
attention tha t is important enoug h i n general moral terms t o warran t an emo-
tional reaction eve n without bein g balanced against countervailing reasons.
Indeed, we might be tempted to define dilemmas in subjective terms as cases
in whic h either alternative warrant s guilt . However, despit e m y dismissa l of
rational guilt in taboo cases, my notion o f perspectival appropriatenes s coul d
conceivably allow for cases of guilt without wrong. Perhaps guilt is sometimes
adequately warranted i n light of its general practical function i n a case of merely
prima facie evidence for counting some act as wrong, as well as for holding the
agent responsible for it. If so, what the agent does in a dilemma might not really
count a s wrong al l things considered unde r the circumstances.
Thus, o n this account, w e might sa y that th e ban on parricide violate d by
Agamemnon's sacrific e of Iphigenia has enormous general moral importanc e
that is registered in emotion bu t is not really in force as a guide to actio n in the
particular situationassumin g tha t i t i s even more importan t t o ensur e th e
success of the Gree k expedition. O r i n a "balanced " cas e lik e the on e draw n
from Sartre , where the agent faces a choice between abandoning his dependent
mother and failin g i n his duty to joi n the French resistance, w e might say the
same for a prohibition of either action: Neither is more important, s o both have
merely prima facie status, and all the agent morally must do is choosethough
guilt will be warranted whateve r choice he makes.
However, i f we accept thes e cases as genuine dilemmas, at an y rate as phi-
losophers understan d the term, we have to say more than this. My own view is
that the violation of prohibition in such cases is serious enough to count as wrong
all things considered; it is not in the usual terms justified, then , at any rate fully,
even if it is excused by the facts of the agent's choice-situation, so that he is not
subject to blame. 34 We can now see that the appropriateness of emotional guilt
does not imply this view of the wrongs of the case. I shall go on to argue , how-
ever, that i t does impl y something stronger tha n merel y prima faci e wrong
something we can understand a t least in rough terms by reference to the primi-
tive notion of "tainting." With some further backing, moreover, it can be made
to yiel d a notion o f objective guilt suitable for dilemmas.

Real Taints
My appeal to the notion of tainting in explaining the psychological function of
guilt in the last section suggested a way o f distinguishing among cases of guilt
178 Sensibility an d Standpoints

for the unavoidable. What we wanted there was a way of distinguishing between
appropriate an d inappropriate subjective guilton a notion of appropriatenes s
that di d no t reduc e to warran t fo r th e correspondin g judgment , of objective
guilt or fault . O f those case s that com e ou t a s appropriate o n the perspectiva l
account it seems that some do and some do not involv e what might be thought
of as a real tainta real need to dispe l the appearance o f fault, let us sayas a
result of involvement in the act in question. I now want to suggest that this notion
is sufficiently objectiv e to allow for a reasonable interpretation of the comments
of agent s i n survivor's guilt and relate d cases . I t also wil l hel p us understan d
what i s at issue in cases of dilemma.
I should firs t poin t ou t tha t th e agen t in the acciden t case is not tainte d b y
involvement i n a morall y offensiv e ac t i n th e wa y tha t Agamemno n is , fo r
instance. The cas e involve s a nee d t o clea r oneself, in response to a questio n
that naturall y arises about one' s ow n responsibility for what happened; but as
I understand it, the case assumes that the question does not arise in minds other
than one' s own . I t is taken a s obvious, that is, that th e agent in the case is not
responsible for a wrong; indeed, as the case is set up, no wrong was done. What
justifies attentio n t o th e questio n o f responsibilit y and henc e make s guilt an
appropriate reactio n is the adaptiveness on other occasions of the general emo-
tional mechanis m guilt brings into play . This i s to sa y that there i s no rea l or
objective correlate o f guilt in the particula r case at hand , though th e case ha s
important motivationa l implications for other cases .
What I want t o argue, though, is that more tha n thi s is involved in cases of
dilemmaand even in the sorts of cases that Morris cite s as examples of objec-
tive nonmora l guilt . As a firs t ste p i n the argument , w e migh t sa y that thes e
cases involv e an appearanc e o f faultsomething to whic h others beside s the
agent migh t be expected t o react, so that guilt functions essentiall y as a way of
heading off others' blame. The agent's nee d to clear himself i s based on some -
thing intersubjectively real, that is. I refer to it as a "taint" to mark what I take
to b e its nonaccidental resemblanc e to th e primitiv e notion, whic h involve d
various punitive or cleansing rituals acted ou t eve n on inanimate objects used
as instrument s of wrongdoing. 35 Bu t my point i s not jus t abou t wha t w e ar e
inclined t o sa y in cases of dilemma . For on e thing , the verba l point doe s no t
really apply so readily to al l such cases, as we can se e by considering what w e
would say about the agent in the case drawn from Sartre. Failure in one's duties
as a son or a citizen may not dra w th e kind of harsh reactio n fro m others that
leads us to speak of a taint as opposed to a serious moral shortcoming or other
flaw. M y point her e is just that there is something, whateve r it amounts to o r
should be called, that gives guilt more than individual psychological backing in
such cases .
A taint in this sense amounts at least to an intersubjective tendencyto raise
the question of responsibilitysomewhat on the order of a "suspicion" of objec-
tive guilt in the semiobjective sense in which that term is sometimes used. To be
tainted i s to b e "unde r a cloud" o f suspicion , or reasonabl y subject t o suspi-
cion on the part of others, as opposed t o just having reason oneself to raise the
question of responsibility, which is all we need for appropriat e subjective guilt
Unavoidable Guilt 17 9

as in the acciden t case. What we are doing here, one might say, is reinterpret-
ing the primitiv e notion i n psychological terms. A n agen t i s really tainted o n
this account wher e he has a real need to clear himself in the eyes of othersno t
just wher e other s woul d i n fac t b e suspicious bu t onl y where, i n ligh t of th e
facts or the evidence available (assumed to constitute a natural perceptual slice
of the overall evidence), they would have real reason t o be. At a minimum they
have reason to deman d an explanation. Emotiona l guil t is essentially a way of
heading the m of f b y raisin g the questio n oneselfan d reactin g t o it , a s a n
instance o f a genera l response tendenc y learned for it s motivationa l useful -
ness o n othe r occasion s tha t als o serve s on thi s occasio n a s a for m o f self -
punishment, symbolically distancing the agent from even nonculpable involve-
ment in moral wrong.
To se e how th e notion o f a real taint in the sense of intersubjectively war -
ranted emotional guilt applies to casesfillin g ou t th e rational portion o f the
continuum o f case s mentione d i n th e precedin g sectionwe migh t first con -
sider variants of the accident case in which the agent does have "something to
answer for " i n others' eyes, even if he would in the en d b e declared blameless.
One might suppose that the agent did something a bit irregular in his car upkeep
that required explanation i n light of what happened. Perhaps it was not really
a factor in the accident; or perhaps its explanation would amount to an excuse.
In any case, the very need to suppl y such a reason i s enough t o yield a limited
objective basis for guil t on the account I have offered .
By contrast , o n th e usua l interpretation o f the acciden t case , th e nee d t o
supply a reasonto defend oneself against charges of objective guiltis a matter
of individual psychology, albeit by no means idiosyncratic. In general terms: A
perceptual slic e of the relevant information that counts a s a natural objec t of
attention fro m the standpoint o f individual psychology may not have the same
status fro m a socia l standpoint. I n the origina l accident case , other s typically
would dismiss the agent's guilt feelings as making no sense to them, simply on
the basis of what is immediately evident. In other cases, further question s may
be in ordereven i f we sometimes suppres s them to preserv e social harmon y
or the agent's peac e of mind. Despite its social reference, that is , the notion o f
a tain t i s not merel y conventional: An agent may hav e something to explai n
independently of whether others in fact demand an explanation; his real appear-
ance of guilt in this sense will not depend on whether he actually appears guilty
in others' eyes .
What shoul d w e say, then, abou t survivor' s guilt and th e other cases that
Morris wants to treat as involving real or objective guilt? The word "taint " does
not apply naturally to such cases, nor do such cases seem to raise a serious ques-
tion o f moral responsibility . Where the y do involve some sor t of wrong, a s in
cases of being favored financially over others, i t is something that is clearly not
the agent's fault ; what i s objectionable to other s abou t hi m is supposed t o be
just that h e benefits fro m th e situation , if only passively. However, I think we
can still say that the agent in such a case has a need to prove something to others
to expres s hi s solidarity with them, bringing himself dow n t o thei r leve l by
way of the sort of symbolic leveling behavior described earlier. Guilt here may
180 Sensibility an d Standpoints

be seen a s a way o f keepin g social frustratio n fro m focusinga s i t naturally

mighton him . What it serves to eras e is a real appearance o f indeed benefit-
ing ("enjoying" benefits) , or grounds fo r envy.
The notion o f a taint does apply naturally to cases of collective guilt, and it
provides a way o f understandin g the rea l basis for collectiv e guil t in applica -
tion to individuals. In popular discussion s o f the question o f German guil t for
the Holocaust, for instance, there sometimes seems to be a tendency to dismis s
guilt feelings a s unreasonable to expect fro m peopl e born afte r the war o n the
grounds that it would indee d be unreasonable to hold them personally respon -
sible in a backward-looking sense for what happened . But assuming that ther e
is a real basis for a judgment of collective guiltin a sense, remember, that need
not appl y to individual s (as with corporations whos e environmenta l sins can-
not b e charged t o ever y employee, or i n some cases , t o an y employee in par -
ticular)it makes sense to speak of innocent individuals as tainted b y the acts
of others. This does no t mean that they are themselves morally flawed in some
way but that th e question o f moral responsibility naturally arises on the basi s
of thei r grou p affiliationwit h emotiona l guil t supplying a kin d o f answe r
(though not th e only possible answer) in practical terms .
The upshot o f this discussion of casesgradually narrowing dow n my ini-
tial cases o f guilt without faul t b y appeal to the perspectival notion o f appro -
priatenessis that ther e i s sometimes a rea l basi s for guil t feelings that fall s
short o f the judgmen t tha t on e i s guilty. I have referred to thi s a s a "taint, "
offering a n intersubjective reading of what I take to be a successor to the primi-
tive notion o f tainting. But we might just think of it instead as perspectival guilt.
It amounts t o somethin g midwa y between guilt feelings and objectiv e guilt in
the usual sense, as implied by a judgment of fault, and it seems to b e enough t o
explain thos e nondilemmati c case s o f guilt without faul t i n which w e woul d
not b e inclined to dismis s the agent's clai m that h e really is guilty. Instead of
reconceiving the cases as involving guilt in some other, nonmoral sense , we may
reinterpret thei r claim to objective status in terms of intersubjective backing for
guilt feelings.
Real taints in this intersubjective sense would see m to represen t th e stron-
gest consequenc e fo r dilemmati c options tha t i s derivable from a n argumen t
from mora l feeling , an argumen t base d solel y on th e appropriatenes s o f guilt
or some simila r emotion i n response to whatever the agent does . However , b y
adding i n referenc e to th e elemen t o f voluntar y control a t thi s poin t i n th e
argument, I think we can se e that dilemma s als o involv e something stronger ,
something closer to full-blown objective guilt, though still perspectival in a way
that accord s with my treatment o f the grounds for the emotion. The y involve a
real basi s fo r full-blow n objective guiltnot jus t fo r raisin g th e questio n o f
responsibility but als o for a positive answer in the form of a judgment of fault,
though th e evidenc e for tha t conclusio n i s overturned b y countervailing rea -
sons. In itself, that is , the evidenc e is sufficient fo r the attributio n o f responsi -
bility just insofar a s the agent intentionally did something he knew to be wrong
under th e circumstances ; but sinc e they were circumstances where h e ha d to ,
morally speaking, a broader vie w would show that h e is not reall y at fault .
Unavoidable Guilt 18 1

Thus, even long after Agamemnon's sacrifice of his daughter, the act requires
explanation i n moral terms. Objectivel y speaking rathe r than subjectivel y (if
he feels nothing) , there is a need for Agamemnon to satisfy himsel f and other s
that the act was morally required under the circumstances. But we also want t o
say more tha n this: that th e act lef t mora l traces o f a sort tha t are summed u p
in the imagery of staining. It is a mark against him, a blot on his record of moral
action, and a threat to his character or virtueunless the wrongness of the deed
(from a limited but morally central standpoint) can somehow b e "put int o per-
spective" by rehearsing the explanation fo r it yet again.
The way dilemmas apparently differ fro m th e cases of intersubjectively real
taints just considered lies in something like the fact that Marcus notes in defense
of dilemmati c guilt:36 They d o involv e responsibilit y for a particular wrong ,
insofar as the agent in them chooses a wrong option i n the knowledge that it is
wrong an d carrie s through o n th e choic e b y voluntary action. O f cours e th e
choice is in some sense involuntary or forced on him; he has no control ove r his
range of options, in particular the fac t that all of them are wrong. So he is not
responsible for doing something wrong, or for wrong o r wrongdoing a s such.37
But he is responsible for th e particular wrong h e does i n a sense that exceeds
mere involvement in moral wrong. We might say that hi s act is culpable as far
as it goesperspectivally, in other words, or considered from a certain stand -
point, a standpoint that is important enoug h to count as morally significant in
This is more than a claim of general moral significance, of the sort that yields
feelings o f compunction i n respons e t o prim a faci e wrong . M y suggestio n i n
Agamemnon's case is that there are strong enough reasons against the sacrifice
of his daughter to make it count as wrong all things considered in the circum-
stances in which it occurs, eve n on the assumption that there are stronger rea -
sons o n the other side. That thes e reasons for counting the act as wrong als o
tend t o undermin e the agent' s mora l wortho r metaphorically , that th e act
stains the selfcount s a s a n objectiv e basis for fel t guil t that stil l makes psy-
chological reference bu t i s not therefor e merely intersubjective.
Marcus's ow n treatment of dilemmatic guilt does not seem to be limited to
a first-person standpoint in the way I have in mind. She appeals, a s I have just
done, to the agent's voluntary control over what he does in support of the attri-
bution of responsibility, but she does not distinguish between grounds for guilt
on the part of the agent and grounds for blame, the corresponding other-directed
emotion. Our intuitive view, I take it, is that blame in the sense of personal anger
is not reall y warranted i n such casesat any rate, for what th e agent does , a s
opposed to a failure t o fee l wha t h e ought to fee l abou t his action.
It may be reasonable to shun an agent tainted b y a seriously wrong actan
act on the order o f parricide, say, as in Agamemnon's caseor to fee l various
related emotion s (possibl y taken as variants of blame), such as personal aver-
sion or horror. But if we really agree that he made the morally best choice open
to him at ever y stage, it would not b e appropriate, though it might be natural,
to blam e him i n the sens e that entail s feelin g th e desir e to punis h him tha t is
characteristic o f anger . From a third-perso n standpoint, o n th e perspectiva l
182 Sensibility an d Standpoints

account, hi s act comes ou t a s wrong bu t excused , eve n thoug h i n emotiona l

terms h e should no t excus e himsel f an d i s therefore subjec t to blam e fo r no t
feeling guilty . According t o th e argumen t I hav e outlined , thi s asymmetrica l
treatment o f blame and guilt rests on the differentia l practica l effect s o f react -
ing to a partial basis for assigning faultassessed in light of the different emo -
tional alternatives to that reactionin one' s own case versus that o f another.
On Marcus ' unamende d account , b y contrast, a n agen t i n a dilemm a woul d
come out as morally blameworthy for his actionfor doing what morally speak-
ing he had to do .
To complet e the tas k o f characterizing dilemmati c guilt perspectivally, let
me now attempt t o show ho w we might replace the metaphor o f staining with
appeal t o reasons, of the sort that would tel l us when a taint counts a s real. As
noted earlier , talk of taints i s not quit e applicabl e t o case s lik e the one draw n
from Sartre . On the other hand , it does apply to nondilemmatic cases of moral
conflict: We speak of money from a bad sourc e as tainted, fo r instance, even on
the assumption that i t would not be wrong t o make use of it for a good end. To
meet the challenge to my account pose d at the end of part I, we need to sho w
that perspectival guilt can capture the intractability of wrong. Bu t that require s
a notion strong enough to distinguish dilemmas from the various weaker cases
of moral conflict that migh t be confused with them. W e are now in a positio n
to approac h th e tas k i n light o f my discussio n o f the functio n o f guilt in thi s
chapter and the last .

Dirty Hands
If w e thin k o f emotional guil t as a self-cleansin g ritual, w e ma y sa y that th e
agent i n a dilemma has real grounds fo r it to the extent tha t th e ritual is really
required to cleanse him of responsibility for a wrong. Thi s means that th e rea-
sons supporting a negative self-evaluation, even if evidentially limited, are mor -
ally important enough in themselves to affect th e assessment of the agent's mora l
worth unles s checkedas they are when th e agent subjects himself to guilt. To
remove the metaphor o f tainting, then, le t us stipulate, first , that th e reason s
are significant enough for the evaluation to register in attention i n the way that
unpleasant affect allow s it to dothus warding off a more settled negative self-
judgment. Fo r dilemmas , however , w e nee d a secon d stipulation , on e tha t
explains why the need for the self-cleansing ritual recurs: From the limited per-
spective of the act under consideration w e do have enough reaso n fo r reaching
a judgment of fault; the judgment is undercut onl y in light of the total body of
evidence. So attention t o th e act (whic h is morally warranted) warrant s emo -
tional guilt . What w e hav e in the rol e o f objectiv e guilt in dilemmas, then, i s
essentially an unsatisfiable needa need to dispel the appearance of fault, base d
on a limited subset o f the evidenc e that i s important enoug h b y itself t o yield
moral wrong .
Guilt functions o n thi s account a s a way o f preserving virtue by providing
some furthe r reason s o n the othe r sidebalancin g those tha t threate n t o un -
dermine virtue with evidence that the agent takes them seriously enough to suffe r
Unavoidable Guilt 18 3

over them . T o th e exten t tha t th e sufferin g i s in some sens e self-inflicteda s

we may say in light of our indirect control ove r emotions via acts of attention
it amount s t o a way o f bringin g oneself dow n t o th e leve l o f a victim of th e
wrong in question. On m y understanding of guilt as an identificator y mecha-
nism, it essentially serves to register the standpoint o f the victim. But there are
other case s resemblin g dilemmas, sometime s marke d of f a s case s o f "dirt y
hands," wher e the agen t i s not require d to tak e tha t standpointperhap s h e
even ought not to take itnot just in order to avoid a second-order dilemm a of
the sort discussed in chapter 4 but rather becaus e of an obligation to see things
from th e standpoint o f the whole. Or so I want to maintain as a way of distin-
guishing such cases from dilemmas.
The term "dirty hands," which is based on the metaphor o f staining or taint-
ing, sometimes is used as a more general term covering dilemmas, though Wil-
liams an d othe r author s us e it to mar k of f cases fallin g shor t o f dilemm a a s
exemplified b y the unsavory compromises politician s sometimes have to make
to gai n an d retai n office. 38 Th e ter m als o extend s t o mor e seriou s cases that
arise in the conduct of office; one example is Winston Churchill' s reputed deci-
sion durin g World Wa r I I to allo w the cit y of Coventr y to b e bombed rathe r
than reveal to the Germans that the Allies had cracked their code. Like the word
"taint" and its derivatives in application to the cases in my last section, the term
"dirty hands" doe s no t appl y naturally to al l the relevan t cases at issue here.
But I shall take it as a semitechnical term covering all such political and relate d
cases alon g with genuine cases of dilemma.
Williams's discussion of the range of relevant cases in fact suggests another
term to distinguish nondilemmatic cases. The term, which has surfaced at sev-
eral points in my argument, is "moral cost." Williams uses it in application t o
cases where , thoug h ther e i s a right thin g to do , i t still involves "wronging"
someone. Ther e i s a victi m of th e action , tha t is , who ha s a legitimat e com -
plaintunlike the person whos e interest s los e ou t i n a case where on e ough t
simply overrides anotherthough on the other hand, this is not enough to make
the actio n wron g a s i n cases o f dilemma . The notio n introduce s a tripartit e
division of cases of conflictinto cases of overriding, moral cost, and dilemma
that apparently rests on a view of certain agents (politicians and others exercis-
ing authority on behal f of a group) as licensed t o mak e utilitarian trade-offs.
Their duties require sacrificing individual interests to the interests of the group
as a whole an d henc e involve special permissio n t o violat e the right s o f indi-
viduals. But the rights of individuals are not therefore simply canceled; they leave
emotional trace s o n th e orde r o f "disquiet " o n th e par t o f a n appropriatel y
sensitive agen t a s evidenc e of th e "mora l cos t attache d t o lettin g a righ t b e
overridden by consequences." 39
In short , then , th e notio n o f mora l cos t amount s t o a mora l remainde r
resulting from a clash between utilitarian and rights-based obligations in which
rights lose out fo r certain special reasons bu t ar e no t therefor e overridden on
the account Williams wants to maintain. For our purposes, questions about the
details of the account can be shelved in favor of the general point that Williams's
notion o f an uncancele d moral remainder in cases of conflict i s not limite d to
184 Sensibility an d Standpoints

dilemmas. He makes it out in nondilemmatic cases a s the remainder of a right,

as distinct fro m a n all-things-considere d moral reason a "claim " o f the sort
that figures i n his explanation o f the uncancele d element in a dilemma. 40
The notion als o seems to be linked to a distinction between the appropriat e
emotional residue s in such cases, with "regre t a t the deepest level" reserved for
dilemmas, whil e mora l cos t case s ar e characterize d b y referenc e to variou s
weaker emotional reactionsa s "distasteful, " fo r instance. 41 However , I think
it would b e a mistake to suppose that the different sort s of cases of conflict may
be understood adequatel y by appeal t o differen t variant s of agent-regretor ,
on the othe r hand , i n terms of quantitative differences i n emotional intensity.
I think we can improv e on Williams's appeal to the leve l of emotional involve-
ment with a finer specification of the possibilities in play here, if we think abou t
the way in which guilt and simila r emotional reactions ar e focused not just on
action bu t als o o n the self .
Consider Churchill' s case as described jus t above, o n th e assumptio n tha t
he did the right thing in sacrificing Coventr y to the Allies' war aims . If we als o
assume tha t th e conduc t o f hi s duties allows fo r emotiona l sensitivit y on th e
questionin contras t t o th e second-orde r dilemma s discussed i n chapter 4
the requisite reaction wil l involve discomfort at what h e had to do, but not the
sort o f devaluation of self that i s involved in the variants of guilt. Nor nee d we
suppose tha t the problem with guilt has to do with the crippling effects o f self-
alienation. Rather, a negative self-evaluation i s just not warrante d here , as it is
in a case of dilemma.
We might su m this up by noting that distast e an d disquie t and other emo -
tions of the sort Williams uses to characterize moral cost cases do not impute a
stain to the agent's self but at most something more peripheral, a s suggested by
"dirty hands. " The y als o see m t o diffe r fro m guil t i n no t requirin g quite th e
same focus o f attention o n the wrong i n question: I t would b e correct a s a dis-
positional attributio n o f emotion to say that someone feel s uneasy about some -
thing she did in the way these terms suggest even if she managed to avoid thinking
about i t and henc e never felt the relevant emotions. Similarly, a certain kin d of
avoidance i n thought woul d suffic e t o discharge an obligation to respond wit h
distaste or disquiet to an unsavory act. One need not dwell on the act and inflic t
some negativ e emotion o n onesel f i n occurrent terms . Bu t this i s the poin t o f
guilt feelings, in virtue of which they are sometimes required, on the account I
have offered .
We do sometimes us e the very malleable term "guilt " to refer to the agent' s
feelings i n moral cos t cases, bu t I would sugges t that what distinguishe s genu-
inely dilemmatic case s i s a requiremen t to fee l guilt y simpliciternot jus t o n
condition tha t one take a certain limited standpoint bu t in a sense that implie s
a requiremen t t o take the relevan t standpoint . W e can su m this u p b y saying
that the standpoin t i n question, a subse t o f the evidenc e bearing o n th e case,
has a claim on the agent' s ful l attention . B y contrast, the politician who mus t
wrong someone is obligated to include that person's standpoint within her overall
view o f thingsan d hence to qualif y an y satisfactio n sh e feel s o n th e whol e
with an element of discomfort. But she is not reall y obligatedindeed it might
Unavoidable Guilt 18 5

seem to be inappropriately "personal" of her, though it is sometimes treate d as

a mark of noble characterto drop out of the group perspective and dwel l on
the harm done to individuals. One expects politicians to give consideration i n
their overall judgments to the harm they have to do to individuals, but not nec-
essarily to give it the sort of isolated emotional attention that Sophi e owes, for
instance, to the child she cannot save .
In short , then, mora l cos t case s ma y b e distinguished from dilemma s i n
emotional term s a s requiring only peripheral attention t o th e wrong i n ques-
tionsomething secured well enough by dispositional affect o r b y mixing dis-
comfort into what one feels toward a more inclusive object. By contrast, dilem-
mas call for occurrent discomfor t as a way of focusing attentio n o n the threa t
of a negative self-evaluation o f the sort involved in guilt. We can illustrat e this
point b y comparing the origina l version of Russell's case with th e dilemmatic
case that I constructed fro m i t i n chapter 4 . The differenc e betwee n th e tw o
cases turned o n degre e of harm done: In the origina l version the woma n wa s
disappointed an d he r lif e was disrupted by Russell's failure to keep his prom-
ise, but th e connection betwee n his act and he r later mental illness and deat h
was merely incidental. However, I take it that with the same degree of political
reference (enough , let us grant, t o justif y Russel l in assessing hi s action fro m
the standpoint o f the whole, even though his political role i s unofficial)an d
without referenc e to requirement s imposed b y lovethis difference i n degree
of harm would make the difference betwee n moral cost and dilemma. It would
yield a general moral obligation to think about his act from the woman's stand-
point and to experience something on the order of emotional guilt .
This is to say that, with the case set up as a dilemma, Russell is required t o
assume the emotiona l standpoint o f a victim, the on e whose rights h e has t o
sacrifice i n favor o f consequences, a s a way of making up for the sacrifice . On
a mora l cost version of the case, though, if we consider thing s from a general
moral standpoint, h e is required at most to fee l guilty if he sees things from the
victim's standpoint. He is also required not to feel various unqualified positiv e
emotions suc h as unalloyed pride at his act as viewed from the mor e removed
standpoint. H e can get by without guilt in the narrow sense , even on a view of
things fro m he r standpoint , i f he substitutes a perspectival emotion wit h less
self-focus o n the order of remorse, however .
To justif y a n unconditional obligation to fee l guilt y in the moral cost case,
we would hav e to brin g in the particular requirements imposed o n Russell by
love as an interpersonal ideal, and I think we would also have to switch to "sepa-
ration guilt." That is , love for the woman presumably does commit Russell, at
least as a requirement of virtue, to seeing things from her standpointto iden-
tifying wit h he r i n a sense that shoul d indee d yield self-alienation if he has t o
sever th e relationship . I n contrast t o th e dilemmati c case, though , wher e h e
actually does the woman serious harm, I would not take the real-life Russell as
morally tainted by the ac t required of him.
I conclude, then, that the worries introduced at the end of the last subsection
can be met by a subtle enough treatment of the different emotiona l requirements
of different case s of conflict. A "real taint" in the fulles t sense involves real self-
186 Sensibility an d Standpoints

threat i n moral terms fro m a n all-things-considere d moral standpoint . Bu t in

Williams's cases from political life, the wrongs done to individuals for the sake
of more importan t political aims ar e assumed to b e insignificant fro m the all-
things-considered standpoin t an d t o tain t th e person wh o doe s the m onl y t o
the extent of making her morally impure. Though sh e is not a fit subject of perfect
pride, sh e i s also no t a n appropriat e targe t o f self-blam e lik e the agen t i n a
The basic point of difference betwee n my own and more standard emotion -
based approaches to dilemmas can be illustrated with reference to Russell's case
by pausing now to reflect on the epigraph I gave this book: "The fo x condemns
the trap, not himself. " If Russell's case is interpreted a s a moral dilemma , on e
might naturally ask why he should blame himself. After all , our assumptio n i s
that he made the best decision he could under circumstances that arose through
no faul t o f his own. I f anything was to blame , it was th e world fo r settin g u p
those circumstances , or perhap s th e mora l cod e tha t reject s all responses t o
themthe trap, not himself.
On th e vie w I have taken, though , wha t th e ideall y sensitive moral agen t
would fee l may not quit e fit what the fox would believe. Feelings are belieflik e
but properly part from belief s insofar as their primary function include s some
of the purposes o f action. In cases of dilemma, guilt serves as an act-substitute
to the exten t that i t goes som e wa y toward satisfyin g its associated desir e fo r
reparation. Her e it would essentiall y bring Russell down t o the position of the
person h e has to harm , makin g u p t o som e degre e for the deficiencie s of the
world an d the moral code as applied to action. In general, then, moral emotio n
plays a role within the moral lif e an d doe s no t merel y record it .
Basing Ethic s on Emotio n

My argument in chapters 4 and 5 was needed to dispel certain initial objections

to understandin g dilemma s in terms o f emotional guilt . A more fundamenta l
question no w remain s t o b e considered, i f we widen ou r focu s beyon d guilt ,
which wa s picked ou t for central treatmen t here because of its disputed status
in cases of dilemma and it s special motivational significance. This is the questio n
of how the acceptance o f dilemmas ties ethics to emotions i n general terms. An
argument to dilemma s from th e reactions expecte d o f agents like Agamemno n
as suggested b y Wiliiams's treatmen t migh t seem to tak e mora l judgment s as
simply serving to record mora l emotions .
I proposed a metaethical alternative in chapter 3 that counts emotion s amon g
the grounds of ethics in virtue of their role in moral motivation. The y constrai n
the shape of the moral code, that isthe corrected version of the social institution
of morality that counts as a real albeit man-made basis of ethicsin virtue of
the fact that any viable moral code has to be teachable in conjunction with emo-
tions. Bu t emotions o r othe r attitude s o r mental states migh t see m to exhaust
the bases of ethics on this account, which I put forth as a two-component varian t
of moral realism , though granting the oddity of the term. On e migh t sa y that
what it means for a moral code t o b e "in force " in a given society i s just for it
to b e held i n force by shared emotion s of moral assessment : A n action count s
as a violation of the rules, or as wrong, t o the extent tha t i t is seen as justifyin g
moral blame or some simila r emotion.1 M y guilt/blame asymmetry qualifies an
account in these terms but leaves intact its view of ethics as a projection of human
feelings onto the worl d o f objects.
This sor t o f "projectivist " vie w is often attribute d t o Hume. 2 O n it s usual
interpretation, however , i t depends on a roughly perceptual model of moral sen-
timent that seems to forc e a choice betwee n realism an d emotion-base d ethics .
This is what I now want to call into question. My aim in this discussion generally
has bee n to trac e a path betwee n the horn s o f various opposed alternative s in
metaethics, and o n this issue I want to say that a motivational mode l of the role
of emotio n offer s a wa y betwee n standar d position s o n th e questio n o f mora l
realism. I shall argue this by considering the view put fort h b y Simon Blackburn
as a "quasi-realist" version of projectivism.3 Blackburn's account comes close to
the one I have offered i n its resistance to opting for either of the standard alterna-
tives on the question in sharply opposed form . It provides a logical structure for
emotion-based ethical discourse that essentially mimics the structure of discourse
188 Sensibility an d Standpoints

about real objects. What I propose is a similar sort of intermediate account that
does not fi t Blackburn's characterization of his own vie w as a form of antireal-
ism, but tha t seem s to m e to yiel d a better resolutio n t o the proble m se t up in
chapter 1 for "subject-dependent" approache s to moral dilemmas.
I attempt t o fil l ou t m y explanation o f the motivationa l mode l i n this fina l
chapter b y showing how i t differs fro m standard projectivis m and b y exhibit-
ing its implications for the view of moral emotio n a s essentially a kind of per-
ception o f moral saliences (section 1). My aim is to pul l together an d t o place
into perspectiv e som e o f th e divers e points an d insight s that emerge d a s my
argument her e proceeded throug h th e rang e o f issues brought u p initiall y by
moral dilemmas. The problem of dilemmas, as I have understood it , centers on
a question abou t the rational coherency of a moral cod e that generates set s of
practical o r action-guiding oughts incapabl e of joint satisfaction. M y answe r
has interpreted moral action-guidance as based on the interplay between indi-
vidual and group standpoints. Emotio n function s primarily on my account a s
the source o f individual motivation t o conform t o a moral code set up i n the
first instance as an instrument of group flourishing. It is justified b y its general
motivational role in a way that allow s for unsatisfiability o r some othe r moti -
vational foul-up on a given occasion. On the other hand, its own social sources
block any simple foundationalist view of emotion i n relation to the content of
the moral code .
The metaethical role of emotion i s multileveled and complex i n ways that I
attempt t o bring out i n this chapter. Among the morally significant aspects of
emotion i s its capacity to take in other evaluative standpoints vi a identification.
I have made much of this capacity in my account o f guilt, as a prime exampl e
of a socially constructed emotion . Bu t we may als o see identification in more
general terms as the source of a kind of emotion-based knowledge, both of other
individuals and o f the interpersona l standpoint tha t inform s the mora l code .
This latter is important a s providing another way between two horns in recent
debate: the personal an d the impersona l standpoint s a t issue in discussions of
the value of ethical impartiality.4 It also lets us characterize dilemmas as a prod-
uct of the basi s of ethics in combining different standpoints . O n th e accoun t I
have defended, these come ou t no t jus t as conflicting personal standpoint s o f
the sort that ca n result from different role s or othe r source s o f obligation bu t
more fundamentall y a s complementary metaethical standpointso f th e indi-
vidual agent faced with a moral decisio n and o f the social end that governs his
behavior a s a moral agentwhose interactio n can go wrong i n the particula r
case. I end wit h a discussio n o f some problem s fo r th e genera l view of ethic s
(section 2) that has emerged from this attempt t o understand dilemmas as con-
flicts of action-guiding moral oughts .

1. Th e Motivationa l Model

On m y accoun t o f dilemmas , emotions functio n primaril y a s substitute s for

action, motivationa l traces of th e ough t tha t i s not acte d on . Thi s stand s in
Basing Ethics on Emotion 18 9

contrast t o th e perceptua l mode l o f emotion s a s putativ e bearer s o f ethica l

knowledge that w e find i n a number of contemporary authors. 5 On the moti -
vational model, moral emotions ma y still be seen as yielding a kind of indirect
knowledge of ethical properties a s an inference from our fel t tendency to act in
certain ways. But the difference can be brought out by contrasting different pos -
sible accounts of an example that comes up in the dispute between intuitionis t
realists and projectivists : perceiving the sadness i n a face. 6 Realists relying on
the perceptual model want to say quite simply that we are aware of sadness in
such a case as an irreducible property o f what we see, though one that depends
on ou r perceptua l makeu p in the wa y tha t colo r propertie s do . A standar d
projectivist account , b y contrast, woul d tak e u s as "projectin g onto " a fac e
perceived a s contorted i n a certai n wa y a n associate d feelin g fro m ou r ow n
Thus explained, our naive view of sadness as something we perceive in the
face comes out a s a kind of perceptual illusion. Mackie sees an analogous illu-
sion, o f "intrinsic requirement" o r "objectiv e prescriptivity," as undermining
ethical discourse, but on other version s of projectivism the illusion comes ou t
as benign. The poin t fo r my purposes i s that i t still apparently counts as per -
ceptual in a broad sense: Projectivism on its standard construal takes for granted
the model of emotion as primarily a mode of apprehension that i s assumed by
the realist view from which it departs. O n th e sor t o f view I want t o suggest ,
though, on e might say that what goes on when we see a face a s sad is that we
respond t o it s contortions wit h a trace o f empathetic sadnessnot a percep -
tion, real or apparent, o f its feeling bu t a feeling fel t o n its behalf.
The mora l analogu e to sadnes s o n th e view I shall defend, however, need
not be something we can attribute to external objects considered in themselves.
We might think of it as a relational property linking an act that is judged wrong
to emotional sanction s accepte d a s part of the institutio n of a moral code . I n
any case, it is the tendency to impos e th e sanctionsa menta l act tendency
that we are aware of when we seem to "perceive" mora l properties on this view.
But we now need to ask how the account differ s fro m more sophisticated ver -
sions of projectivism than Mackie's. Blackburn, for instance, at one point char-
acterizes his own account as "conative" i n contrast to th e realist's perceptua l
account.7 The main point of difference i n general terms is Blackburn's classifi-
cation of his account as antirealist, whereas in departing altogether from a per-
ceptual account, I hope to locat e an alternative to perceptual realism within a
form o f realism that assigns emotions a fundamental moral role .

Forms of Subject-Dependence
Blackburn's quasi-realism is essentially a way of reconciling projectivism with
the "ordinar y language " evidence for realism . We speak o f ethical statements
as true or false, most notablyas objects of doubt and o f knowledgeand we
use them in unasserted contexts such as conditional clauses without reference
to the speaker's current emotion or other attitude. Blackburn attempts to explain
such features o f ethical discourse in terms of a complex set of semantical con-
190 Sensibility an d Standpoints

straints meant to "fulfil l th e practical purposes fo r which we evaluate things."8

These ma y be said to give ethical discourse th e appearanc e o f describing some
independent reality, whereas in fact on Blackburn's account i t rests on the pro -
jection o f our ow n attitude s onto a world o f natural o r nonmoral facts .
Blackburn's antirealist approach, inherite d from Aye r and Har e and ulti -
mately attribute d to Hume , start s wit h a pictur e of moral judgmen t as base d
on an individual's emotional or other practical stance toward the world, though
recently Blackburn has attempted to allo w for a social version of the picture. 9
Like my own view, Blackburn's projectivist quasi-realism claims a kind of sub-
ject-independence compatibly with a basis in emotions or other attitudeswha t
Blackburn sometimes speaks of as personal "commitments." This is backed u p
by a refusal t o endorse statements expressing subject-dependence, 10 as opposed
to m y own restrictio n in chapter 3 of forms of subject-dependence relevant to
the issu e o f realism to thos e in which th e min d i n question i s seen as the on e
currently making a given moral judgment. Does Blackburn' s theory provide an
alternative way out of the problem outlined here for dilemmas and motivationa l
force i n chapter 1 ?
Let us first as k how Blackburn' s theory differ s fro m th e one I have offered .
On a general level, several important difference s see m to emerge from the fac t
that Blackburn's account i s set up explicitly as a version of expressivism.11 That
is, although accordin g to quasi-realis m moral judgment s exhibit belieflik e be -
havior, their primary function on Blackburn's projectivist view is really that of
attitudes. O f the sentences expressin g them h e writes:
They expres s something more to d o wit h attitudes, practices, emotions, feel -
ings arisin g in contemplating some kinds of conduct, with goal seeking , with
insistence upon normative constraints on conduct, and nothing to do with rep-
resenting the world. In the familia r metaphor , their "direction of fit" with the
world is activeto have the world conform t o them, rather than descriptive or

In othe r words , the worl d i n itsel f contain s n o mora l fact s against whic h w e
might measure our moral statements; a discrepancy does no t impl y inaccurate
description but rather some sort of failure to act properly. Yet it is perfectly apt
to spea k of such statements a s true or fals e o n Blackburn' s view insofar as we
construct truth an d simila r notions t o furthe r th e practical aims of moral dis -
One might want to say that an extended notio n of "the world " results from
our creation of values on this account. This is essentially the view I have taken,
with the moral code see n as a creation o f the human min d but once created as
taking o n a lif e o f it s own i n constraining furthe r moral choices . Th e rule s it
comprises exhibi t the world-to-mind directio n o f fit that Blackbur n attributes
to our moral statements in the passage just quoted. But moral statements them -
selves may on this accoun t b e understood a s having a descriptiv e function to
the exten t tha t the y describ e the working s o f th e code . Mor e precisely , they
describe human action i n terms of its fit to the codea s satisfying o r failing t o
satisfy th e mora l rules . The cod e itsel f ma y b e see n as a kind of projection of
Basing Ethics on Emotion 19 1

our deonti c attitudes , bu t ont o an intervenin g screen, a s it were, rathe r tha n

directly onto objects. The resul t is an artifact o f thought, allowin g for a new ,
relational way o f describing natural objects. It is in some respects like the fic-
tional object s that Blackburn discusses along with lega l codes a s examples of
how texts ma y constrain what we sayexcept that it does not purport t o rep-
resent anything. For a legal analogy, as noted i n chapter 3 , one might consider
the way the U.S. Constitution constrain s legislation, sometimes in a way tha t
conflicts wit h the values of current legislators.
Blackburn at one point in discussing subject-dependence speaks of the values
we create a s taking on a lif e o f their own.13 However, a s I intend this point, it
involves a break with the antirealist tradition that Blackburn's view is meant to
defendthe tradition from whic h he derives its classification as "expressivist. "
On the account I have suggested, the point of moral discourse is to describe acts
in relation to th e codeto an idealize d version of the actual code, correcte d i n
light o f the en d i t is meant to achieve : group flourishing , which essentially re-
places Blackburn' s appeal t o th e pragmatic aims of moral discourse . Sinc e the
code on this account is not the invention of any one mindor even of the domi-
nant socia l sensibilit y of any on e erait is perfectly possible to get its contents
wrong. The overarching end that constrains it is itself made out as something other
than a product o f human sensibility, though it depends on human ends. By con-
trast, Blackburn presumably wants to say that his own appeal to pragmatic aims
is simply an expression of some more fundamental set of attitudes .
This las t point reveals a basic difference betwee n Blackburn' s account an d
the one I described in chapter 3 . On the face of them, many of his comments on
moral sensibility and the like resemble my characterization of the corrected moral
code. In connection with the notion o f a "viable " code , fo r instance, consider
what Blackburn says in response to popular associations wit h "relativism" an d
"subjective" views of ethics:
Just as the senses constrain what w e can believ e about th e empirical world, so
our nature s and desires , needs and pleasures , constrain muc h of what w e can
admire an d commend , tolerat e an d work for . There are no t s o many livable,
unfragmented, developed , consistent, an d coheren t system s of attitude. 14

However, remark s like these about the "best possible set of attitudes" see m to
require interpretation themselves in terms of the speaker's attitudesa s endors-
ing a certain sensibilityo n Blackburn's account o f value. 15 Blackbur n often
brings in higher order emotion s an d other attitude s toward one's moral emo-
tions i n his detailed account of quasi-realism. They figure, for instance, i n his
explanation of the use of moral terms in unasserted contexts , with the speake r
understood a s endorsing logical and other connections amon g endorsements. 16
And they also underlie his explanation of moral truth to the extent that seeing
our moral judgments as false depends on the notion o f an improved standpoint
of evaluation from whic h we may admire or disavow our first-order emotional
propensities.17 Fundamental reference to a judgment of emotional appropriate-
ness take n a s descriptiv e could no t d o thi s jo b withou t compromisin g the
expressivist cast of Blackburn's theory.
192 Sensibility an d Standpoints

The same must be said, then, of appeals to group flourishing or even to par-
ticular moral ends such as fairness as constraints o n the moral code. I suggested
several alternativ e ways of making sense of such constraints i n chapter 3i n
terms of judgments of group value, for instance, on a version of the perceptual
model; but also in motivational terms, as supporting the teachability of moral
rules. An expressivist account , though, woul d seem to hav e to rel y on a basic
expression of preference at some point in the explanation. If it is expressivist at
the leve l of moral judgments , moreoverand I shall later sugges t that i t need
not beit would hav e to do without the separation into two components tha t
allows for a version of externalism.18 That is, from within morality on my own
account on e has to endors e the sort o f cooperative sensibility that is constitu-
tive of the moral enterpriseor at any rate, one has to endorse it by and large
but the content o f morality is given independently. So one can withhold the en-
dorsement and still acknowledge moral truths.
For an individual property analogous to group flourishing on my proposed
view consider health as an end that might be said to b e constitutive of medical
practice. The doctor's warnings about alcohol may be understood an d acknowl-
edged, with healt h seen as the standard of adequate functioning proper t o the
sort of entity that is in question, a body, even by someone whose current com-
mitments lie elsewhere. Similarly, group flourishing may b e seen a s a norma -
tive constraint governin g groups, spelling out the elements of group perfection,
but assesse d fro m a standpoint th e speaker may reject.
It is worth notin g tha t neithe r of these standpoints, medica l or moral , has
to b e taken a s settin g u p a n invarian t ideal: There ma y b e interna l conflicts
among th e elements i n question, so that a n individua l might hav e to choose ,
say, betwee n hig h energ y and placi d endurance , on th e mode l o f th e choic e
between societie s stressing cultural achievement and thos e conten t t o mak e a
relatively unperturbe d contributio n t o th e strea m o f life . Thi s ma y b e some -
what weake r tha n Blackburn' s view, with it s assumption tha t mora l trut h is
constructed t o serv e the pragmati c ends of moral discourse , favoring a single
answer to questions set up in moral terms.19 Blackburn's view apparently builds
in the philosopher's concer n with dialogue and debateor perhaps particularly
the linguistic emphases of his semantical approach t o metaethics . At any rate ,
on my own accoun t eve n the overarching end of morality need no t b e shared
by someone wh o grasp s the content o f a moral judgment.
Subject-independence o n m y accoun t rest s on mor e tha n a highe r orde r
positive attitude toward subject-independence as in Blackburn's quasi-realism.
It also seems to provid e a basis for a descriptivist interpretation o f moral lan-
guage. A moral judgment may be understood, not necessarily as expressing the
speaker's attitudes , bu t a s purporting t o describ e the world i n terms of some
standard o f morality tha t count s a s "real" enough fo r the purpose, whateve r
its origins as a projection o f attitudes. It is a real invention or artifact , some-
what on the model of a computer program, which once designed constrains what
the programme r ca n doincludin g his attempts to modif y it . Bu t the sor t of
reality in question here should not be confused with that claimed by metaphysical
realism: the vie w that abstrac t propertie s exist apar t fro m th e thing s whose
Basing Ethics on Emotion 19 3

properties they are. Existence apart from the mind of a putative subject of knowl-
edge is all that is in question, as with the color properties discusse d in the debate
over the perceptual model o f realism.
I hav e suggeste d "socia l artifac t realism " a s a nam e fo r thi s alternativ e
approach t o metaethics. However, i f the word "realism " pose s problems, w e
might just get rid of it and speak instead of social artifact "descriptivism" a s an
alternative to the expressivist version of projectivism that Blackburn draws from
twentieth-century metaethics in the tradition stemmin g from Hume. Th e view
still shares with Hume the insistence that moral properties ar e not simply "ou t
there" a s objects of perception, eve n though our moral language often suggests
as much. But neither are many of the acts we describe by reference to such prop -
erties. Instead of filling ou t th e Humea n accoun t o f moral language , the view
essentially expands Hume's accoun t o f the moral imagination b y taking some -
thing like the standard to which he appeals in making sense of justice as under-
lying more basic moral attitudes, to the extent that morality entails a principled
extension to new cases.
It is appeal to som e suc h standard tha t provide s the hardness o f the mora l
"must" and that thereby helps make dilemmatic cases hard enough on the agent
to avoi d th e proble m I se t u p fo r subject-dependen t view s i n chapte r 1 .
Blackburn's quasi-realist projectivism and m y own vie w both se e ethics as in
some sense subject-independent even though emotion-based . However , it was
subject-dependent motivatio n tha t wa s i n questio n i n chapte r 1 . The two -
component view allows for a split between content an d motivation, s o that the
motivational force of an ought with all-things-considered rational statu s nee d
not actuall y be strong enoug h to win the day. At the same time, th e self-pun-
ishing aspect of guilt as a substitute for action and its anticipatory role as a moti-
vator i n advance of actio n ma y b e said t o kee p th e agen t o n th e hook . If we
allow fo r "perspectival " guilt , with a real bu t limite d foundation i n a mora l
"taint" o n th e agent , th e emotio n wil l be warranted b y the ough t th e agen t
violates in a dilemma.
By contrast, thoug h Blackburn could of course allow for guilt or some othe r
unpleasant residue of moral conflict, he could no t provid e any firmer founda-
tion fo r it than ho w we ultimately happen t o feel. 20 That is, a requirement to
feel guilt y would seem to have to rest o n our higher orde r attitud e toward the
emotiontoward a certain sort of sensibility in reaction t o a dilemma. To avoid
my objections to Hare's treatment of moral residue s in virtue-ethical terms, th e
account migh t endorse ange r o r som e simila r emotion towar d a n agen t wh o
fails t o exhibi t the requisit e sensibility. But this would itsel f express a certain
sensibility. At no poin t coul d w e simply appeal t o th e moral demand s o f the
case, th e conten t o f the practica l ought s i n conflict, as dictating what to feel .
My own account departs from expressivism in the first instanc e b y analyz-
ing emotions in terms of ought-judgments and other evaluations. This may seem
to make moral judgment prior to emotionwith odd results for the attribution
of emotions to animals and humans that lack the capacity for moral judgment
but i t is important t o not e tha t the dependenc e is not one-way. 21 My detailed
account o f guilt migh t in fac t b e viewed as providing a particular example of
194 Sensibility an d Standpoints

how an emotion wit h a moral content might result by processes o f refinemen t

and social shaping from a childhood reactio n that is itself not specifically moral .
The child in my developmental account in chapter 3 reacts initially on the basis
of a pre-moral evaluation of its immediate environment as a source of pleasure
or painbut extended to include social pleasures and pains via the mechanism
of emotiona l identification . The eventua l result of the socia l sources o f emo -
tional learning in general terms is comfort or discomfort directe d toward fairl y
sophisticated sorts of prepositional objects , including moral act-requirements,
so that emotions serv e to register the importance o f moral reasons for action ,
as i n th e pictur e that emerge d fro m m y discussio n o f deonti c dominance i n
chapter 2.1 now want to approach th e results of that discussio n from anothe r
angle, in relation to the question of the sense in which emotions may be said to
amount to perceptions of moral truths .

Emotional Perception
A perceptual interpretation of moral emotio n ma y see m to b e implied by the
role we accord emotional reactions as evidence of the truth of moral judgments.
Despite the original use of "intuition" i n ethics to stand for an intellectual fac-
ulty, the term now ofte n seem s to b e applied t o any gut feelin g i n response t o
particular cases. In attempting to systematize our emotional reactions into gen-
eral principles, it might be said, wha t w e are doing is weeding out illusio n by
applying a test of coherency. What passes the test counts as a perception o f the
moral facts.
This accoun t o f things also ma y see m to fi t ou r preanalyti c treatment o f
emotions as a guide to moral decision-making. A feeling of discomfort at some
act I have managed to justify to myself or that others are trying to persuade me
to do is something we count as a reason fo r moral mistrust. The act just "feels
wrong." However, I would suggest that such appeals to emotion as a bearer of
moral knowledge may be understood withou t the analogy to sense-perception ,
in terms of the motivationa l role I have assigne d to mora l feeling . Th e us e of
emotion to motivate moral behavior lets us take emotional reaction s as data
in th e firs t instance , as clues to wha t we really think, to th e exten t tha t they
involve urges to act . O n m y account, tha t is , they have an evaluativ e content
that i n ordinary cases is keyed to ou r mora l beliefs rather tha n directl y to th e
facts themselves, as on perceptual versions of realism. But they may b e said to
"track" the relevant facts, at any rate roughly, and to provide the bases for an
inference to them, insofar as they are taught to us as motivational props of the
moral code. They register the values instilled in us as part of the process o f set-
ting up the code b y linking it to individua l psychology.
Emotions may also be said to provide a kind of moral knowledge that short -
cuts belie f i n som e cases . Thei r motivationa l functio n depends o n directin g
positive or negative attention toward thei r evaluative objects, which might not
always be accepted as objects of belief. We might think of empathy, for instance,
as providing a kind of nonmoral knowledge, of others' menta l states, b y way
of evaluativ e standpoint s understood no t t o appl y to ourselves . But to gran t
Basing Ethics on Emotion 19 5

this is not to say that we "perceive" others ' mental states via emotion i n a sense
that implie s direct awareness ; rather , ou r awarenes s i s mediated b y imagina-
tion. In any case, emotiona l knowledg e of moral truth s come s ou t a s a much
more complex matter on the sort of account I mean to suggest here than it does
on a perceptual model . I shall have more to sa y in my next sectio n abou t th e
sense in which empathy ma y affor d individual s a source o f specifically mora l
knowledge vi a identificatio n with th e group . Fo r th e moment , le t me take a
simpler example of emotion without belief.
Suppose I feel uneasy about some action that in fact seems justifiedgrad -
ing a paper I take to b e irredeemably bad, say, on the basis of just the first few
pages. Emotional discomfor t here serves to hold i n mind a general rule enjoin-
ing punctilious performance of all job-related duties that prescribes completing
the paper. The result in this case is a variant of guilt; but the mechanism in play
here extends beyond self-punitive or even negative emotions in a way that may
indeed make it seem to resemble perception. Perceptua l language came into my
discussion o f the evidential assessment of oughts in chapter 2, where I used the
image of figure/ground dominance to make intelligible the persistence of nega-
tive oughts i n th e fac e o f conflictthe source o f dilemma s on m y account i n
terms o f sufficiently stron g reasons. On e might suggest, then, that moral emo -
tions amount to perceptions o f the force of moral reasons .
There i s a kernel of truth i n this view to the extent that emotional comfor t
and discomfor t o n m y own accoun t o f emotion s migh t b e said t o registe r i n
affective terms their positive or negative evaluative content. However, we have
to b e careful no t to take "perception " too seriouslyeve n after cancelin g out
any sensory connotations, a s I have done. For our emotional access to the facts
is typically mediate d b y beliefspossibl y false beliefs , a s m y example here is
meant t o indicate . My "perception " o f the importanc e o f completing an irre -
deemably bad paper i s explainable as the residue of an emotion securing atten-
tion t o a general rule that i s thought t o b e important enoug h b y and larg e t o
justify a few spillovers to othe r cases. Eve n in veridical cases, moreover, wha t
emotional attentio n hook s ont o in the firs t instanc e may b e something to b e
added to the surrounding situation, not something found in it: the act to be done,
not som e property of the world by virtue of which it demands action. And the
emotion ma y be at som e remov e from th e situationlik e an allergi c reaction
to a food enjoye d a t the time but producing hives a few hours later .
One migh t object that sense-perceptio n als o depend s on ou r expectation s
and othe r belief s abou t th e situation. However, to pic k ou t mora l emotion s I
take it that we need to assign them a belieflike content, even if it is one that ha s
to be left indefinit e in some cases. Consider a case involving more in the way of
affective qualit y than the one just cited: moral horro r at the thought o f a seri-
ously wrong act such as Agamemnon's parricide. The sensory "feel" of this and
similar reactions may be part of what motivate s the analogy to perception, bu t
it seems to have sources in imagination that are distinguishable from th e simple
awareness of moral wrong. I would interpret it as an identificatory response t o
the victim's pain and feeling s of betrayal or t o society' s feelings o f aversion to
the perpetrator . T o assig n the feelin g a specificall y mora l content , though ,
196 Sensibility an d Standpoints

so that i t might seem to coun t a s the perception o f a moral property , w e need

to se e it as directed toward a n abstract thought : that the act in question is for-
bidden, say. It is in this sense, conceptual rather tha n causal , that mora l emo -
tions see m to b e mediated by beliefs.
On m y account, emotion s ar e reasons fo r actioneve n instrumenta l rea -
sons, since comfort an d discomfort, besides registering the force of evaluations,
themselves amoun t t o reason s fo r sustainin g or changin g one' s presen t state .
To see more concretely how emotions can be understood i n motivational terms,
we might turn away from the moral cases under consideration t o other cases of
emotional perceptio n tha t we would no t b e so ready t o explai n by appeal t o
special properties o f objects. An analog y I find usefu l fro m detectiv e fiction i s
the detective' s uneas y sense that somethin g "doesn't fit" : There i s somethin g
funny abou t th e fact s o f the cas e a s she knows themo r perhap s ther e wa s
something out o f place in a room sh e has just visitedbut it is unclear where to
locate the problem . W e need not explai n her perception o f "unfittingness " as
keyed to a corresponding property of the situation that she somehow i s in con-
tact with directly. I would sugges t instead that sh e is more immediatel y aware
of something like the interrupted scanning of her own perceptions or thoughts
her review of the facts of the case or the objects in the room, in contrast t o nor -
mal cases in her experience or to the room a s she saw it before. In short, she is
reacting to somethin g irregula r about he r own mental tendencies, though her
attention i s focused outward o n its causes.
I lik e this analog y partl y becaus e th e sense o f "fittingness " i s sometime s
brought i n to characteriz e th e positiv e inpu t of our mora l faculties . M y ow n
view is that the negative reaction, a "taboo feeling," is more basican d that it
need not have much specific affective content, beyond something like the detec-
tive's sense of uneasiness, in order t o count a s an emotion. It amounts to a so-
cially inculcated or reinforced aversion to certain acts, with an intellectual con-
tent whose analysi s may require as much special insight and skill as is required
of the detective to get at the causes of her uneasiness. In the moral case, of course,
in order to be effective th e basic reaction ha s to be widespread. Bu t its analysis
involves attention t o things too remote to count as plausible objects of percep-
tion: the mora l code , the ends i t is meant t o serve , an d th e way s it has to b e
corrected i n order t o serve them.
One need not understand these things in detail to use moral languag e intel-
ligibly: Our ordinar y description s of acts a s satisfying o r failin g to satisf y th e
moral code will be left indefinite . I shall have something t o say in my next sec-
tion abou t problem s th e socia l artifac t view face s t o th e exten t tha t i t ha s a
relativist basiswith "the " mora l code picked out in some way that implicitly
refers to the speaker. At this point, however , I want t o raise a related questio n
about th e basi c value constraints tha t the view presupposes. Thes e emerged in
the discussio n o f fairness in chapter 3 as possibly importin g referenc e to rea l
value properties i n something like McDowell's sense . Bu t the properties wer e
limited to such things as the rational way of assessing means to an end and the
proper compositio n o f the socia l end t o b e used i n assessin g the mora l code .
Should we now accep t at least these basic value constraints as indeed objects of
Basing Ethics on Emotion 19 7

emotional perception i n the broad sense just indicated? Can one simply see, for
instance, that harmony and stability are traits of a flourishing group in the way
that the detective recognizes the solution when it occurs to her?
The analogy is too simple, I think, for what we have to deal with in the moral
case are constraints on what is to count as a solutionas if the detective had t o
choose betwee n two partial solutions, both wit h some loose explanator y end s
but on e yieldin g a bette r accoun t o f how th e dee d wa s don e an d th e othe r a
fuller accoun t of the perpetrator' s motive . Suppose she opted fo r th e firs t ex -
planation o n the grounds tha t a murderer's mental states ma y no t alway s be
open to understanding or scrutiny. If we share her basic explanatory expecta -
tions, must we b e given pause by the though t tha t someon e els e migh t have
different explanator y expectationsmight insist, for instance, that understand-
ing an act requires understanding its reasons?
By admitting that the acceptance of a certain solution rest s on a basic deci-
sion on explanatory issuesrather than on some sort of perception o f explana-
tory adequacyw e would no t b e taking a statement of the solutio n simpl y as
expressing th e speaker' s explanator y expectations, an y mor e tha n a jur y i n
announcing its verdict is simply expressing the presumption of innocence. Simi-
larly, i n th e mora l case , ou r relianc e on a choice of basi c values at a certai n
level need not be taken as undermining descriptivism; our moral judgments may
be understood as describing acts in accordance with the values we have chosen
rather than simply as expressing the choice of valueseven if the choice intro-
duces a relativistic element. This account still gives us what we want from real-
ism as long as our value-choice is implicit in the moral code, as a social/histori-
cal artifact that is not modifiabl e at will. Whether it is thus implicit is a matter
for argument on occasions when we encounter a real disagreement about basic
values, about which I say more in the next section .
The motivationa l mode l defende d her e a s a wa y o f understandin g th e
metaethical relevance of emotion seems to me also t o yield a bette r notio n of
its normative ethical relevance. It allows for the importance o f the "forward -
looking" assessmen t of different emotion s in accordance with their motivational
effects b y letting it influenc e th e standard s o f emotional appropriateness . B y
contrast, the analogy of perceptual correction t o remove the distortions of per-
ception fro m differen t standpoint s tends to produce a "disinterested " idea l of
moral judgmen t as somethin g essentiall y impersonal. On e lin e o f argumen t
sparking the current revival of virtue ethics rests on the importance o f particu-
lar personal standpoint s o f evaluation.22 But the motivational mode l seem s to
provide a clearer picture o f the reasons fo r the moral assessmen t o f emotion s
that virtue ethics brings to center stage.
On a n accoun t o f moral training like the one drawn fro m Aristotle, emo -
tions ar e take n a s subject s of praise an d blam e not jus t because o f their ten -
dency to lead to overt action but for their own sake.23 They are typically under-
stood in terms of character traits, but the motivational model sees them as based
on acts of a sortacts o f attentioneven where they do no t lea d to over t ac-
tion. In admiring someone's courage, say, or his sympathy and concern , what
I value in the firs t instanc e is his tendency to hol d in mind certain thoughts in
198 Sensibility an d Standpoints

situations o f danger or harm t o others, often at some affective cos t to himself.

Emotional affec t i s explained as a way of holding thoughts i n mind by register-
ing their evaluative features in positive or negative feeling tone. This may seem
at firs t sigh t to fi t the perceptual analogy , but i t involves actively entertaining
one propositio n rathe r tha n another , no t simpl y receivin g information fro m
external reality.
There ar e furthe r possibilitie s within duty ethics, i f we accep t m y defense
of "ought-to-feel, " fo r basin g ethics on emotion i n normative terms: Motive s
can b e see n a s the prim e bearer s o f mora l valu e with over t act s judge d only
derivatively.24 O n th e metaethica l questio n o f th e emotiona l basi s o f ethics ,
though, there seem to be three possibilities to be distinguished, besides of course
the denial of an emotional basis as in Kant.25 Besides the different way s of under-
standing emotion s a s registerin g mora l judgments , tha t isperceptuall y o r
motivationallythere are familiar way s of understanding moral judgments as
essentially registering emotions. View s on the metaethica l question can b e di-
vided accordin g t o whethe r the y assign emotions o r other motive s a place in
the semantical analysis of ethical judgments. Blackburn's semantical approac h
is a new entry in the dominant twentieth-centur y tradition o f inquiry into th e
meanings of moral terms stemming from G . E. Moore's cognitivist account of
"good" as unanalyzable on the model of color terms.261 hope to have contrib -
uted i n this essa y the beginning s of both a fine r graine d understandin g o f th e
nature of moral emotion and a way of remaining neutral on the semantical ques-
tions tha t have dominated metaethica l discussion unti l now.

2. Sensibilit y an d Standpoint s
One way of basing ethics on a kind of emotional perception has indeed emerged
from m y argument here to the extent that it underscores the role of empathy in
moral judgment. The motivational significance of empathy as a way of binding
oneself to others was important t o my defense of emotional guilt , which I inter-
preted a s an identificator y mechanism, in response t o mora l dilemmas . Emo -
tional identification also can b e seen as yielding a form of knowledgeknowl-
edge o f others' mental states, o f the sort that came up i n the las t section, an d
also, we should now note, a form of specifically moral knowledge to the extent
that it enables us to take a perspective essential to morality, that o f the grou p
as a whole .
The picture of morality to which I have appealed rest s on this overarching
view o f things i n its reliance o n a notio n o f group flourishin g as guiding th e
correction o f the moral code. This ideal is meant to b e understood b y analogy
to the Aristotelian notion of individual flourishing or happiness. Though I have
allowed a standard wa y of referring to our estimate s of flourishing in terms of
the ascriptio n o f value-properties, it i s important tha t the propertie s i n ques -
tion are often unrealized . The sam e is true of many of the other properties that
we understan d vi a empathy: They involv e state s o f feeling , fo r instance , tha t
would resul t from som e futur e actio n w e are contemplating bu t ma y no t do .
Basing Ethics on Emotion 19 9

Perception of external objects is not the right model for such cases. In my argu-
ment up to this point I mainly took for granted ou r views about group flourish-
ing and saw moral emotions o n my motivational account a s imposed o n indi-
viduals in order to enforce obedience to a pre-existing moral code. But we now
might probe deeper , to questions abou t th e bases of the code.
Morality woul d of course b e undermined by a certain kind of basis in emo-
tion, whereas the epistemic role of emotions o n the perceptual realist model pose s
no threat. However, the tendency of this essay has been to uncover various ways
in which the motivational role of emotions affect s the content o f the moral code
and to reinterpret their epistemic role in terms of it. I want to complete my argu-
ment, in this section, by beginning to respond t o some problems that might be
raised for the view that has emerged here, indicating how I think they might be
handled b y extensions of it . In particular, I attempt t o sho w a t leas t roughly
how objection s concerning the view' s relianc e on a standpoin t tha t seem s t o
oppose individua l autonomy can be countered b y a notion o f self-identity that
brings out it s sources in group identification.
A variant of group identificationself-projection int o the standpoint o f the
whole (not to be confused with the impersonal standpoint)ultimately figures
as one of the sources of our insight into group flourishing. We understand what
makes a group flourish b y analogy to the way we understand individual happi-
ness in the case of another agent, puttin g ourselves into the imagined position
of the group in question. This positive appeal to imagination stands in contrast
to th e eliminatio n of personal biase s that produce s disintereste d moral emo -
tions on the usual perceptual model, bu t i t can still be seen as involving a cor-
rection o f individua l perspective . O n th e othe r hand , i t ma y seem to involv e
personifying th e group i n a way tha t calls to min d the collectivis t excesses of
nineteenth-century idealism. I want to suggest , instead, tha t grou p identifica-
tion underlies a certain kind of individual self-regard.

The Interpersonal "Ought"

My treatment of "ought" in chapters 2 and 3 essentially explained the sense of
obligation a s based on th e impositio n o f group need s o n th e individual , with
moral emotions originating as felt individual needs manufactured by the grou p
to enforc e compliance . But part I ended wit h a differen t sor t of appeal t o th e
emotional basis of ethics in response t o initia l worries abou t subject-relativit y
and relate d issues. Despite the socia l source s o f our adul t stoc k o f emotions ,
ethics on the account I offered ha s to rely on the shape given to behavior by our
natural emotion tendencies. I appealed particularly to our tendency to identif y
emotionally with others, as a barrier to views that put some subgroups beyond
the moral palenot an insuperable barrier, but one that impose s a social cos t
on exclusive caste systems .
I now wan t t o ad d to thi s some mor e genera l reflections on ho w m y view
brings together self-regard an d concern for others in support of moral motiva-
tion. Even apart from th e sources of particular emotions like guilt in identifica-
tion with others, emotional motivation may be seen as linking the self to other s
200 Sensibility an d Standpoints

to the extent that it focuses an agent's attentio n o n something external to him-

self whil e appealin g t o hi s concern fo r hi s own stat e o f comfort a s a furthe r
reason fo r action. 27 By using object-directed comfort o r discomfort as a way of
holding significant thoughts in mind, emotions can be said to incorporate som e
other-regarding reason s fo r action within self-interest.
The stres s o n socia l factor s in this account fit s i n with view s of the sel f a s
essentially a social product.28 Our firs t stirrings of self-awareness are plausibly
held to depend on identification with the reactions of othersto behavio r that
from others ' externa l perspectiv e seems to requir e explanatio n i n terms o f a
localized set of perceptions and interests or a distinct personality. Given the facts
of childhood dependency, our sens e of ourselves as having interestsor a self -
interest; or a selfrequires no t just a review of our menta l contents considered
in isolation but also some awareness of the changes in them consequent on action
on ou r behal f by others. We must come to see such action a s originating out-
side ourselves and in general to distinguish ourselves from th e group or groups
that provid e our initia l sources of identity.
Further, even in adult lif e our sens e of who w e are is in many ways consti-
tuted by our relations to othersby a potentially conflicting set of statuses and
roles, from our initial place in the birth order of a certain family to the position
we eventually achieve in society at large. But these fairly commonplace observa-
tions about th e social sources of the self can b e made to yield an alternative to
the relianc e of traditional account s of the role of emotion i n ethics on notion s
of individual flourishing, even those that buil d in some moral o r social value-
constraints on flourishing.
We might think of the self as originally extendedenlarged t o include vari-
ous others (in the first instance, the mother)but with shifting boundaries lead-
ing us to distinguish an individual self, over time, as a kind of intersection of its
multiple group extensions. O n this picture, the other end-poin t o f social com-
position i s also a kin d of explanatory construct: Th e overarchin g group, th e
group tha t yield s "the standpoin t o f the whole, " migh t be thought o f as th e
conflict-free idealizatio n of th e variou s small-grou p socia l source s o f self -
identity. But we have to resist any temptation to see the standpoint of the whole
as one among other personal standpointsa standpoint someon e or something
might b e said t o have rather tha n on e we ought t o striv e to take . It need no t
count a s a possible experientia l standpoint, a "perspective " i n the sens e tha t
implies a subjective point o f view, but rathe r jus t a s a limit o n our attemp t t o
reconcile the various group standpoints that potentiall y constitute a self.
Reconciling group standpoints, that is , means reconciling the aims of vari-
ous competin g groups , whic h presumabl y means reconcilin g the desire s and
interests of the various extended selves that would result from allyin g ourselves
to them . By taking this project to th e limit, we remove a kind of limit to indi-
vidual self-definition b y keeping open multipl e possibilities of group member-
ship. To th e extent that a subjective viewpoint emerges, it is that of imaginary
self-projectionin thi s case into a n enlarge d view o f thing s rather than int o
someone else's particular perceptual position. It is a viewpoint that centers on
the agent rather than tends toward neutrality . We d o hav e to adjus t ou r indi -
Basing Ethics on Emotion 20 1

vidual slant on things to attain it, as on the usual account of perceptual correc-
tion. But it is important t o the moral cast of mind the account yields that ou r
corrections are not meant to achieve an impersonal standpoint; rather, they aim
at a certain overarchin g pattern o f personal standpoints .
The resultin g imaginary standpoint, o f enlarged self-interest , is a psycho -
logically rewarding one to take. Any barriers to i t involve forgoing something
personally valuable, if only because they put limit s on th e pleasure s of imagi-
nation. I think we can use these reflections to get at some specifically social rea-
sons for adherence to moralitythus in effect expandin g the standard notio n
of autonomy a s self-rule t o includ e identification with the mora l community .
Such an ideal can be made to yield a modern version of traditional Aristotelian
approaches t o virtue ethics, with their reliance on individual flourishing a s an
aim that depends on relations to other s bu t can be all too easil y limited to an
elite subgroup. The aim serves as something that might induce an agent to behave
morally withou t necessaril y advancing his welfare. What I think my socially
based approach migh t be able to add to fill this roleparticularly in cases where
individual flourishing has to b e sacrificed, even in a moral sense, as in second-
order dilemmasi s a vie w of grou p flourishin g a s affectin g individua l self-
identity and henc e the worth o r meaning of an individual life in a sense distinct
from individua l flourishing.
I shall attempt to fil l ou t this suggestion in a moment, bu t its initial point is
to avoid a view of morality as simply imposed on the individual by processes of
social engineeringeve n small-scale engineering, of th e sor t involve d in m y
account of moral teaching via guilt in chapter 3. The basis of guilt in identifica-
tory processes tha t I went o n to describe in chapter 4 begins to move us away
from tha t model t o the extent that i t makes out eve n a socially manufactured
emotion lik e guilt as expressing a natura l emotional mechanism that i s itself
essentially social. My current suggestion involves extending that account to the
bases o f moral motivation . O n th e accoun t I have offered, guil t involve s the
absence of something positivea sense of community with others, let us say
that i n emotiona l term s ma y b e though t o f a s a for m o f pride . Prid e in thi s
communitarian form may be seen as a socially based element in our conception
of the meaningfulness of an individual lifeits worth in a sense that sometimes
detaches fro m individua l happiness or even virtue.
To put the matter simply: We derive self-worth in a fundamental wa y from
our identification with various overlapping groups and group projects. Actio n
on reasons give n by a particular group standpoint , whateve r its instrumental
costs o r benefits , has a self-fulfillin g expressiv e function fo r th e individual : It
enhances the self by identifying i t with something larger. Group identification
may be psychologically rewarding with respect to all sorts of group affiliations,
of course, includin g bad ones, but a s a source of value it presupposes a value-
laden notion o f group flourishing. It is a further mov e to the group standpoint
the standpoin t o f the whole , wha t I think of as the mora l standpoint , a s op -
posed to that of some elite group, or even of the various overlapping subgroups
an individua l in fac t belong s tobut I think it can b e mad e easier by under-
standing what i s achieved by social pride.
202 Sensibility an d Standpoints

That is , the valu e o f grou p identificatio n for th e individua l self i n more

mundane cases, such as ethnic pride, lies partly in the protection i t offers agains t
contingency, or the vagarie s of individual fortune. My il l fate will be less of a
loss t o m e i f I can se e my effort s a s serving some large r effor t tha n m y own ,
something whose success in pursuit of what I value is compatible with my own
failure. This sort o f self-transcendence, in one form o r another , i s the poin t of
group membership insofar as it affects self-regard . Groups als o have fortunes ,
however, and their aims may be frustrated or misdirected; for that matter, they
can mistreat or exclude some of the individuals who identif y wit h them. So an
individual's group commitments can also be failures. Insurance against il l for-
tune i s provided b y spreadin g ou t one' s commitment s t o variou s differen t
groupsand i s reinforced by commitment to an ideal community, the commu-
nity of the whole . Taking th e mora l standpoint i n this sense i s the bes t avail-
able bulwark against ba d mora l luck.
There are ways in which the moral projec t also can fail I shal l bring up a
possible metaethical failure i n the next sectionbut it is at an y rate cushioned
against the usual forms of bad luck, including moral luck, to the extent tha t it
looks beyond individual flourishing, or even the flourishing of a particular sub-
group. For instance, in the face of a moral dilemma or other situation in which
my ow n virtu e turns ou t t o b e irredeemabl y compromiseda second-orde r
dilemma, say, in which moral sensitivity is no help eitherit would stil l be pos-
sible, if I really did ac t a s well as one could unde r the circumstances , t o tak e
pride at least in that achievement. At the very least, pride gives me a reason for
not doin g worse. What I do stil l matters or has a pointthe intende d sense of
"meaning" here .
In more norma l case s wher e w e raise the questio n "Wh y b e moral?" th e
answer may be given by appeal to a form of pridenot specifically to the affec -
tive rewards of the occurrent emotion but to a presupposition o f its evaluative
content: the claim moral behavior gives us on other agents' regard. This is sup-
posed t o b e independent o f our actua l affiliations wit h others , an d o f others'
accurate perception o f us, as also subject to ill fortune. What one loses by aim-
ing at others ' actua l regar d a s oppose d t o wha t the y would thin k (o r feel ) if
they knew one's motives , say , is membership in an idea l community as a basis
for self-respect. 29
An overarching group idea l as an extension of individual self-worth might
suggest some sort of Nietzschean versio n of self-perfectionism. Bu t the notio n
of imaginary self-enlargement I have in mind is not a n inflatio n of self-regard;
rather, it amounts to the set of potential identifications with others that results
from removin g barriers to enlargement of one's own perspective. With th e self
seen as the intersectio n of an individual's multiple overlapping possibilities of
group membership, any two-person grou p including oneself count s a s a com -
ponent standpoin t i n the set that is potentially one's own, th e set of communi-
ties on e migh t ente r into . T o sligh t any on e o f these standpoint s i s thus, on e
might say, to limi t oneself by obstructing a possible mental route t o imagined
We ca n thin k o f th e interpersona l standpoint i n this sense a s a particula r
Basing Ethics o n Emotion 20 3

kind of group standpoint : a standpoint o f the whole that is constructed ou t of

all its parts, since it takes in all the overlapping socia l perspectives that poten -
tially constitut e the self . M y suggestio n i s that this o r som e simila r practica l
ideal afford s a motive toward mora l behavio r i n people whos e motivationa l
structure is roughly normal but who might still seek reassurance that they are
not losing out by acting morallyas opposed t o an amoralist, defined as some-
one who lack s moral motives. 30 M y discussion her e is just meant t o exhibi t a
self-regarding motive for seeing things from the moral standpoint, not to present
that standpoint a s a precondition o f rationality, as on Kantian accounts of the
moral "ought. "
The "ought " that emerges from m y account o f morality as a social artifac t
will be normatively "binding" onl y from the group standpoint; from the stand-
point of particular individuals it is binding only in the psychological sense given
by my treatment of motivational force in chapter 3. Societies have reasons, tha t
is, why their individual members must do what morality prescribes; their mem-
bers have those moral reasons only in a weaker sense that allows fo r a failur e
to act on them compatibly with rationality. Accompanying psychological sanc-
tions, a s instilled by the group t o generat e conformity to mora l reasons , giv e
the illusion of rational bindingness on the individual level and serve as goads t o
action in the individual case. It is only when an individual takes the overarchin g
group standpoin t tha t he is "bound" by moral reasons. What I have suggested
is that we do have a natural motive for taking that overarching view of things,
namely pride; but i t is not th e sort of motive that compels action.

Final Perspectives
I now wan t t o conside r som e furthe r objection s t o th e socia l vie w of ethics I
have proposed, o n the grounds tha t it s reference to th e surrounding commu -
nity gives it a conventionalist aspect . There ar e variou s way s o f defusin g th e
threat o f relativism. 31 Here I want t o sho w ho w w e might ward i t off, at an y
rate for terms o f serious moral condemnatio n suc h a s "wrong," by appealin g
to assumption s the socia l artifact view makes about hypothetica l correction s
to th e moral cod e i n force i n a given community. Various way s of avoiding a
relativist upshot ca n b e teased ou t o f the reference to a "viable " mora l code ,
particularly as understood i n light of possibilities of group overlap . William s
has recently defended relativism for groups tha t d o not com e into contact.32 I
shall eventually be using the possibility of group contact as one barrier t o rela -
tivism on the view I think of as roughly Protagorean: a s making man th e mea-
sure of morality, at leas t i n a collective sense, whethe r o r no t i t comes ou t a s
relativist on all definitions o r makes man the measure of "al l things."
The relativist element of the view seems to emerge when we call into ques-
tion the many references I have allowed to "the " mora l code. On my proposed
interpretation, moral judgments are to be taken as describing contemplated act s
in accordance with a corrected versio n of the code. The y therefore essentially
extend an d appl y a man-made standard whose creatio n amount s t o th e con -
ventional acceptance within society of certain practices of moral response, in -
204 Sensibility an d Standpoints

eluding but not limited to moral emotions. But we need to face a familiar prob-
lem about th e multiplicit y of conventionally accepte d standard s o n man y is-
sues, even within the same overall group.
The question turns on the possibility of conflicting group practices that can
result in basic disagreements over issues such as abortion o n which there may
be moral subcommunities in opposition. But an adequate answer for metaethical
purposes does not require picking out one such community as the source of "the "
moral code for a given speaker. The actual code may be thought of as indeter-
minate i n some case s withou t threa t t o wha t i s at stak e her e i n metaethical
termsthe existence of a right answer t o mora l questionswit h its basis in a
corrected versio n o f the code. There migh t very well be two righ t answers in
the sens e of conflicting ways o f correctin g th e code , bot h makin g it viable
insofar as they both would adequately promote group flourishing. Since viabil-
ity i s a "satisficing " rathe r tha n a "maximizing " notion , th e criterio n i t pro -
vides might be met in several equally acceptable ways .
On the other hand , this sort of conflict need not entai l metaethical relativ-
ismdepending on where we locate the initial indeterminacy that gives rise to
it and wha t w e take a moral judgmen t to say . A n indeterminacy in the cod e
itselfin ou r attempt to sum up the moral practices of the overall community
need not imply variation in the higher order standard appeale d to in correcting
it: what counts as group flourishing or as adequate promotion o f group flour -
ishing. On thes e questions of basic value, as opposed to specificall y mora l o r
deontic questions, m y view allows fo r a limited "perceptual " versio n of real -
ism. Bu t it is one that turns on a complicated feat of imagination (of hypotheti-
cal social practices and their effects o n the group a s a whole) rather than sim-
ply on "seeing" mora l properties i n the world around us. Much error is possible,
along with disagreement about how to tell what is an error, on such questions
as whether a group that restricts reverence for human life to postnatal forms of
it would be likely to flourish. But divergent views could be explained either by
different hypothetica l factua l conjectures o r b y problem s i n performin g the
imaginative acts required for moral perception .
However, a n explanation o f the latter problems might be thought to intro-
duce a highe r level sourc e o f indeterminacy . There ma y b e no singl e answer,
that is , to th e questio n how on e is to conside r thing s from the overall grou p
standpoint. Fo r instance, even assuming that a fetus is conscious at a given stage
of development , ther e will b e different way s o f weighing its interests agains t
the mother's. Nor i s it clear that only a conscious standpoint ca n matter to the
determination of group flourishing. The assignment of independent value to lif e
might be thought to b e part of group flourishin g in the way that, say, cultural
achievement arguably isnot just because of its effects o n huma n conscious -
ness but also for its own sake .
Perceptual realism on the model of G. E. Moore's theory 33 i s the standar d
way of settling such questions of multiple goods, and my view is meant to allow
for i t as one possibility. We can still minimize the need to accept special moral
properties by limiting ourselves to perceptio n of the element s of group flour -
ishing. Valu e properties o n this accoun t woul d b e attributed t o hypothetica l
Basing Ethics o n Emotion 20 5

social practice s and situations a s seen from a general group standpoint rathe r
than to particular situations in the world around us. They would not be seen as
intrinsically motivating in the way that Mackie find s objectionable, for it is left
open that a rational agent may not care to promote what she understands to be
a requirement of group flourishing. So my suggested versio n of realism can a t
least do without any distinctively moral faculties of perception and rely on ordi-
nary interpersonal insight as mediated by emotion an d expande d b y imagina-
tion. At a certain point, however, thi s version of the view may indeed require
insistence on what one "sees."
Still, we ca n avoi d adoptin g Moore' s intuitionis t mora l epistemolog y t o
explain such claims. We may suppose that there is a "truth of the matter" about
the constraints on group flourishingon the model of Aristotelian happiness
without holding that we have access to basic truths of a sort that warrants the
term "knowledge. " Instead , perhaps, a particular visio n of the moral aim can
be tested i n imagination, either by simple apprehension o r b y the mor e com -
plex sorts of operations embodie d in Rawls's method of reflective equilibrium,
but without any guarantee that imagination is vivid enough or free enough from
personal biases to arrive at the right conclusion. With perceptual realism pushed
to the level of basic constraints on moral codes, we could thus make better sense
of disagreement among moral agents assumed to b e rational, educated, atten-
tive, an d th e like , and a t th e sam e time necessaril y committed t o particula r
personal standpoints .
As an alternative to perceptual realism, however, I think we could also allow
an element of indeterminacy at this level. Different societie s and differen t sub -
groups within the same society may disagree on basi c questions of value with-
out any of them being in error, perhaps because there is no definite answer. We
might be tempted to sum this up with a claim that such issues must be decided
by convention. But although important socia l aims will be impeded if we can-
not ge t everyone to agre e on a convention, the metaethica l result need not be
relativism. Basic judgments of value within the rang e o f indeterminacy might
be subject to a relativist account, but there may be ways of avoiding that result
for particular deontic judgments. I want to end this discussion by sketching one
line of argument that attempt s to "drai n off " a n initia l element of value-rela-
tivism b y the wa y i t understand s judgments of mora l wrong . I t i s a comple x
argument that may or may not convince; but I do not put it forth as essential to
the defens e o f social artifac t realism. Rather, i t suggest s a purer for m o f th e
view, one that avoid s any appeal to mora l perception , but that still does no t
come ou t a s relativist, if the argument is correct.
The argument turns on interpreting moral judgments as general claims about
the viabilit y of alternative moral codes, not jus t the particular code t o whic h
the speaker happens to adhere. The result in a case of conflicting codes would
apparently favo r the mor e permissiv e judgment. That is , if we were t o gran t
that there could be a viable society whose code permits an act forbidden b y our
own, I take i t tha t w e woul d b e granting thereby that th e ac t i n question
polygamy, sayis not really wrong but just contrary to our own way of life, or
"locally forbidden." Similarly, in the abortio n cas e we might be said to have
206 Sensibility an d Standpoints

two alternative ways of assigning values, either of which would lead to a viable
version of the code, assuming for the moment that we agree on the criteria fo r
viability. But the judgment that abortion is wrong amount s on the present sug-
gestion to a judgment that abortion i s ruled out by the corrected mora l code
meaning an y versio n o f th e mora l cod e tha t woul d b e viable in the sens e of
adequately promotin g group flourishing . This would seem to be false, though ,
if w e suppose tha t there is at leas t one viable version of the cod e that permit s
abortion. The statement tha t abortion is morally permissible, which makes this
weaker claim , woul d com e ou t a s tru e whateve r th e persona l value s of it s
In that case, however , mora l dispute s like the one ove r abortio n will turn
out to rest on disagreement over basic issues of social rationality: on what count s
as group flourishin g or as adequate promotion o f it. But then we need to com -
plicate the abov e argumen t b y considering the possibilit y of higher level con -
flict over the assessment o f viability. For instance, perhaps an abortion foe would
insist that the developmen t o f fetal lif e itsel f ough t t o coun t i n the determina-
tion o f group flourishing , whereas other s woul d den y this. Unless the disput e
could be attributed t o a factual disagreement, the two groups would seem to be
talking at cross-purposes i n the way that i s characteristic of moral relativism.
Whether i t really amounts to relativism, however, depend s on whether we
allow judgments of the viability of a moral code to be settled by personal deci-
sion. Ther e may b e a trut h o f the matte r o n tha t levelo r an indeterminacy
that we leave unresolvedcompatibly with the interpretation of particular moral
judgments as referring to the speaker's moral code. For the social artifact view
does not have to take moral judgments as in some sense about the different evalu-
ative standpoint s fro m whic h the y are made . Thoug h i t give s a descriptivis t
account o f them by reference to a corrected version of the speaker's moral code ,
it need not interpre t the m a s describing the content o f such a code. Rather , a
moral judgment may be understood as referring indefinitely to the form or forms
the actual code would tak e if corrected an d describin g a certain act in relation
to it. A judgment of wrong, say, would describe an act as ruled out b y any cor-
rected version of the code. So the meaning of its reference to a "corrected" moral
code (as understood i n terms of viability) would no t var y with th e speake r i n
cases of conflicting moral subcultures .
However, what i f we were to gran t tha t it s truth-value does var y with th e
speaker: that th e criteri a fo r determining its truth-value are open t o persona l
decision o n basi c question s o f value ? Thi s sor t o f subject-relativit y (whether
individual or social ) may b e taken a s the definin g featur e of moral relativism,
at any rate a s applied t o cognitivist views of ethics of the sort under consider -
ation. Bu t ther e i s an analog y I brough t i n earlie r t o undercu t metaethica l
expressivism that might also be useful i n connection with relativismshowing
how my view could conceivably incorporate a n element of both relativism and
expressivism without amountin g to a version of either of them.
Suppose we take a basic judgment of social rationality as expressing a cer-
tain decision-making stance, on the model of a jury's presumption of innocence
in Anglo-Saxo n law . This i s what determine s which corrected versio n o f th e
Basing Ethics on Emotion 20 7

code a given speaker ha s i n mind i n making a particular mora l judgment . So

his moral judgmen t might als o b e said to expres s tha t stance , bu t onl y in the
extended sens e i n whic h a jury' s verdict might b e though t o f a s a particula r
expression o f the presumption of innocence. Its linguistic function may still be
seen a s descriptive: I t is meant to describ e a certain ac t i n terms o f th e appli -
cable syste m o f rules, jus t as a verdic t of "no t guilty " i s meant t o describ e a
certain agent in terms of the applicable system. However, o n our present hypo -
thesis of basic indeterminacy, it is as if we were tempted to equate "no t guilty "
verdicts reached unde r tw o differen t lega l systems. The verdict s can b e taken
as meaning the same thingthat the agent did not commit the act she is charged
withbut their referenc e to differen t way s o f weighing evidence undermines
the equation.
So moral judgment s on thi s accoun t hav e to b e understoo d relativ e to a
particular system of basic presumptions and other norms; and it seems that there
may be several such systems corresponding to differences i n the constraints on e
recognizes as governing the promotio n o f group flourishing. But let us ask a t
this point wha t th e socia l artifac t view could say about tw o distinc t societie s
with different codes . One test of the viability of a moral code would seem to be
its ability to surviv e contact wit h othe r way s o f life . I f maintaining a certai n
code in force in a community that has access to alternatives requires changing
some o f its rules, that change would count a s a correction introduce d fo r th e
sake of social stability. A version of this point als o would seem to appl y cross-
temporally, to codes that develop in ways that can be made out as historically
necessary. We might accept a superseded code as a standard fo r assessing judg-
ments within its limited temporal sphereand similarly for geographically iso-
lated societiesin the way that Williams suggests; this is essentially the model
of alternativ e frame s of referenc e provided b y my jur y analogy . Bu t w e als o
have a way of getting beyon d this to a more overarching social frame of refer -
In effect, on this account, moral codes will be tested by their tenability from
a large r grou p standpoin t i n light of the natura l conditions o f socia l life . Bu t
note tha t this is a negative test onlythere might be a code that could survive
knowledge of alternatives yet for some other reason fails to promote group flour -
ishingso it is compatible with competing accounts of the constraints o n group
flourishing. The way the test is formulated lets us avoid the conclusion tha t any
culture that i n fact win s out ove r others ha s the more viabl e morality. Fo r in-
stance, it would not necessarily rule out cultures or subcultures such as the Amish
in American society whose existenc e depends on protective barrier s limiting the
contact thei r members have with the outside world. What it rules out i s a soci-
ety that coul d no t surviv e contact with other s eve n with th e ai d of protectiv e
barriers compatible with group flourishing on the account suggested in chapter
3. This would disallo w unfai r barrier s such as those singlin g out a particular
social subgroup (for instance, women) for behavioral restriction.
It may still see m odd t o coun t a code or a group adherin g to i t as morall y
superior just because it wins a competitive test of viability. But note, first, that
a society that does win out over others need not have a morality that itself would
208 Sensibility an d Standpoints

pass the test, since the test also implies viability in absolute terms. Some feature
of the winning society's code, that is, including the very means by which it won
(unfair economi c practices with respect t o other cultures , say), may be incom-
patible with group flourishing, at any rate on the larger scale of the group tha t
would be formed by incorporating othe r cultures as subcultures.
Second, thoug h ther e ma y b e a sens e of "morall y superior " i n which th e
term applie s to cultures that may not be viable in competitive terms, I think we
should consider this as analogous to the case of an individual who exceed s the
standards fo r personal virtue in ways that undermine his personal effectivenes s
and henc e make his conduct come ou t a s wrong. Imagin e a teacher, say , wh o
adheres t o a n idea l code o f conduc t i n relation t o student s tha t assume s a n
unrealistic degree of honesty and reliabilit y on their partwith the result tha t
he is always being taken advantage of and hi s aid to deservin g students i s not
really helpful . W e d o sometime s spea k o f suc h individual s as "morall y supe -
rior," though their behavior in fact comes out a s morally defective in any bu t
the ideal world it presupposes. I t is easier to think of societies than of individuals
as operating in isolation, bu t th e point i s that this view of things no longe r fit s
the facts of the case and henc e never did fit the long-term facts. What i s at issue
in determinin g viability i s the suitabilit y of a moral code t o guidin g a societ y
through time. If a code cannot comman d the allegiance of its membersor if it
could not comman d suc h allegianc e in the fac e o f competitionthen whethe r
or not one admires the features of human nature that lead its members to favor
alternatives, it is simply defective in real-world terms, on the model of the peda-
gogical code that make s unrealistic assumptions abou t studen t behavior . Like
it or not , th e function of codes i n controlling behavior means that th e fact s of
human behavio r impose limits on what they can prescribe.
One might object that moral codes are embedded i n a host of other institu-
tions that might influence individuals to opt in or out of a given social group. A
particular code ma y gain adherents, say, not becaus e of its moral content bu t
because i t is conjoined with a n appealin g set of cultural values. However, w e
should not e tha t th e tes t o f competitiv e viability , despite it s basis in real-lif e
patterns o f choice, als o ha s a hypothetica l element. I t i s meant t o tur n o n a
culture's survivability under conditions of clear-headed choiceruling out mili-
tary conquest , say , along wit h subtle r kinds of encroachment, a s evidence of
moral superiority . On the other hand , i f a society's mora l cod e i s in some way
responsible for it s cultural appeal rathe r tha n simpl y being incidentall y con -
joined with it , then a widespread preferenc e for the cultural values in question
would indeed seem to count a s a reason i n favor of the code. The preference for
them would coun t amon g th e facts of human nature, that is , so that my argu-
ment on the issue of the real-life presuppositions o f the code would apply, even
if the cultural values themselves (ranking Disneyland alongside the Louvre, say)
might be faulted on aesthetic o r other nonmora l grounds .
In short, the test of competitive viability involves a mix of real-life and hypo -
thetical element s that a fulle r treatmen t would o f course have to spen d som e
time sorting out. I think we have them sufficiently i n hand, however, fo r pur -
Basing Ethics on Emotion 20 9

poses o f my present argument . Let us apply the test to th e extreme case of cul-
tural contact, wher e we have conflicting subcultures within the same society .
If a social consensus would naturally tend t o develop i n favor of one of th e
views in conflict, on the one hand, that would seem to mean that the other side' s
version of the corrected cod e is not really viable, at any rate under present his -
torical conditions . O n th e other hand , i f both side s could manag e to coexist ,
that would apparently undermine the claims of the abortion foe s in our example .
We again would have to deny, that is, that their version of the corrected cod e is
really viable. It would not yield group flourishing in its own terms, on our hypo -
thesis, where the opposing subgroup is able to follow its own code, so that there
are regular violations of the presumption i n favor of fetal life . This assumes that
the opposing subgroup is sizable, or would naturally become so under the imag-
ined circumstances, which allow for contact wit h social alternatives. So even if
abortion require d going elsewhere, the violations would be substantial enough
to undermin e the abortion foes ' standar d o f "adequate " flourishingunlike,
say, the occasional murders that a society banning murder can be said to toler -
What if we conclude that neither version of the corrected cod e is viable? O r
what if neither consensus nor coexistence i s possible and th e conflict so under-
mines social stabilit y that th e actual code ha s to b e viewed as incorrigible? In
that case, a moral prohibition could not hold true, assuming that it amounts t o
a claim about corrected version s of the code. Abortion woul d again come ou t
as permissible, but s o would everythin g else. This i s a disturbin g outcome i n
metaethical a s well as ethical terms. Th e poin t fo r m y purposes, however , i s
just that it is not relativism. The truth of a moral judgment would not vary with
the speaker . Rather , judgments of wrong woul d al l be falseas on Mackie' s
view but fo r differen t reasons : becaus e moralit y ha s turned ou t t o b e a faile d
project, not becaus e its basic materials were misconceived.
Note, too, tha t thi s scenario rules out eve n replacing the actua l code wit h
one that enforces a decision without consensus, as an extreme form of "correc-
tion." The abortion example illustrates nicely the possibility that any such ruling
would be too regularly violated to promote stability. The alternative in practi-
cal terms would seem to be the adoption o f a more permissive standard o f via-
bility, as setting a threshhold of adequate (rathe r than requiring maximal) pro-
motion o f group flourishing . "Our" mora l code , then, can b e understood a s
the code universally in forcepotentially, that is; with current indeterminacies
where we now disagree, but on the more optimistic assumption that somethin g
can be worked out . For the moral ai m would indee d seem to b e unachievable
and the project of morality misconceived if our allusion s to the corrected cod e
turned out not to refe r to even one possible social construct .
At any rate, the view I have outlined as a version of realism, though se t u p
partly i n relativistic terms, o n thi s accoun t manage s t o avoi d thoroughgoin g
relativism without any element of perceptual realism. Nor doe s it quite fit into
the other standard metaethical categories, though it combines elements of sev-
eral of them. In particular, it involves a partial basis in emotion that offer s a n
210 Sensibility an d Standpoints

alternative to "emotivism, " a form o f expressivism brought u p i n the firs t in -

stance a s part o f the logical positivist program.34 Emotivism was originally an
attempt t o push morality outsid e the pale of cognition b y making it an instru-
ment of emotional expression an d influence. Other metaethica l views set up in
opposition to emotivis m have made roo m fo r moral emotion , i f at all , on th e
model of "moral sense" theories , a s itself an instrument of cognition. I hope in
this discussion to have extracted a subtler alternative for understanding the emo-
tional basi s o f ethics fro m th e attemp t t o mak e sens e o f dilemmas and othe r
puzzles of the mora l life .

Chapter 1

1. Se e the followin g selections i n Wa r and Moral Responsibility, ed . M . Cohen ,

T. Nagel, and T. Scanlon (Princeton , N.J.: Princeton University Press, 1974): Michae l
Walzer, "Politica l Action : Th e Proble m o f Dirty Hands," pp. 62-82; Thomas Nagel ,
"War an d Massacre," pp. 3-24; R . B. Brandt, "Utilitarianis m and the Rules of War,"
pp. 25-45; an d R. M. Hare, "Rule s o f War and Moral Reasoning, " pp . 46-61.
2. Nagel , "Wa r and Massacre," p. 23.
3. Se e esp. Brandt, "Utilitarianism, " pp . 30-31; cf . Hare , "Rule s o f War,"
pp. 59-60.
4. Fo r a general discussion o f the notion, see Bernard Williams, Moral Luck: Philo-
sopfeca/Papers 1973-19SO (Cambridge: Cambridge University Press, 1981), pp. 20-39;
and Thomas Nagel , Mortal Questions (Cambridge: Cambridge University Press, 1979),
pp. 24-38 .
5. Se e the discussion in Alan Donagan, Th e Theory o f Morality (Chicago : Univer-
sity of Chicago Press, 1977) , p. 144 ; cf. G. H. Von Wright, An Essay in Deontic Logic
and the General Theory of Action (Amsterdam: North Holland , 1968) , p. 81, n. 1. The
reconstruction o f Aquinas's reasoning i n terms of issues of moral luck is my insertion ;
one might suggest grace as the applicabl e concept withi n his own framework .
6. Se e Immanuel Kant, Th e Doctrine of Virtue: Part II o f th e Metaphysic o f Mor-
als, trans. M . J. Gregor (Philadelphia : University of Pennsylvania Press, 1964) , p. 23 .
Kant allows for conflict between th e grounds o f obligationwhich, according to com -
ments he makes elsewhere, apparently are such as to become obligations if certain con-
ditions (presumably including the absence of conflicting grounds) are satisfied; cf. idem,
Lectures o n Ethics, trans. L . Infield (Ne w York : Harpe r & Row , 1963), p . 19.
Kant's views on th e holy will, which I use subsequently to give some internal sup-
port fo r his position o n moral conflict, are not explicitl y brought t o bea r o n the issue,
as far a s I know, thoug h th e connectio n seem s obviou s enough. Se e esp. idem, Foun-
dations o f th e Metaphysics o f Morals, trans . L. W. Beck (Indianapolis: Bobbs-Merrill,
1959), pp . 30-31. But note that th e form o f necessitation i n question woul d amoun t
to obligatio n onl y for imperfect being s on Kant' s account; cf . the distinctio n betwee n
forms o f necessitation i n point o f freedom in Lectures o n Ethics, pp. 27-30 .
7. Se e W. David Ross , Th e Right an d th e Good (Oxford : Clarendon , 1930) ,
pp. 19-20 , 28, 41.
8. SeeG . H. Von Wright, "Deonti c Logic, " Mind 6 0 (1951): 1-15; cf. A. N. Prior,
"The Paradoxe s o f Derive d Obligation, " Mind 6 3 (1954) : 64-65. Cf. als o Roderic k
M. Chisholm, "Contrary-to-Duty Imperatives and Deontic Logic," Analysis 2 4 (1963):
212 Notes t o pages 12-15

33-36, an d G. H. Von Wright, Norm and Action (London: Routledge 8 t Kegan Paul,
1963). A general account is provided i n Dagfinn F011esdal and Risto Hilpinen, "Deontic
Logic: An Introduction," i n Deontic Logic: Introductory an d Systematic Readings, ed .
R. Hilpinen (Dordrecht , Holland: D . Reidel, 1971) , pp. 1-35; se e esp. pp. 8-9,13, and
9. Reprinte d as E. J. Lemmon, "Mora l Dilemmas," i n Moral Dilemmas, ed. C. W.
Gowans (Ne w York: Oxford Universit y Press, 1987) , pp . 101-14 ; cf . idem, "Deontic
Logic and th e Logic of Imperatives," Logique et Analyse 8 (1965): 45-51, fo r a fulle r
account o f the implication s of dilemmas for deonti c logic .
10. Se e Plato, Republic, I , 331c5-9.
11. Se e Jean-Paul Sartre , "Existentialis m I s a Humanism, " trans . P . Mairet , i n
Existentialism from Dostoevsky t o Sartre, ed . W. Kaufman n (Ne w York: Meridian ,
1957), pp. 295-96.
12. Lemmon , "Mora l Dilemmas," p . 113; cf. pp. 111-12.
13. Reprinte d a s Bernard Williams, "Ethica l Consistency, " i n Moral Dilemmas,
ed. C. W. Gowan s (Ne w York: Oxford Universit y Press, 1987) , pp. 115-37 .
14. Lemmon , "Mora l Dilemmas," p. 107, n. 2.
15. Se e Aeschylus, "Agamemnon," 204-52 .
16. Se e Bernard Williams, Problems o f th e Self: Philosophical Papers 1956-1972
(Cambridge: Cambridge University Press, 1973) , pp . 204-5.
17. Se e Philippa Foot, "Mora l Realism and Moral Dilemma, " i n Moral Dilemmas,
ed. C . W. Gowan s (Ne w York : Oxfor d Universit y Press , 1987) , pp . 250-70; also,
Samuel Guttenplan, "Moral Realism and Moral Dilemma, " Proceedings of the Aristo-
telian Society 8 0 (1979-80): 61-80, and Walter Sinnott-Armstrong , Moral Dilemmas
(Oxford: Basi l Blackwell, 1988), pp . 189-214 .
18. Se e Bernard Williams, Morality: An Introduction t o Ethics (Ne w York: Harper
& Row, 1972), pp . 92-93, and idem, Moral Luck, pp. 60-61, 74, and 78-79.
19. Se e ibid., pp . 60-63; cf. p. 74 , n . 2 (cf . also n. 3 for Williams's identification
of "tragic " cases with dilemmas).
20. Se e Bas C. van Fraassen, "Values and the Heart's Command, " i n Moral Dilem-
mas, ed. C. W. Gowans (Oxford : Oxford University Press, 1987), pp . 138-53; and Ruth
Barcan Marcus, "Mora l Dilemmas and Consistency, " i n ibid., pp . 188-204.
21. Se e Thomas Nagel, "Th e Fragmentation o f Value," i n Moral Dilemmas, ed.
C. W. Gowans (Oxford : Oxford Universit y Press, 1987), pp. 174-87 . For another view
of ethics as based on conflicting standpoints, close r i n some ways to m y own view , see
Stuart Hampshire, Morality and Conflict (Cambridge , Mass.: Harvard Universit y Press,
1983), pp. 140-69 . Cf. also th e discussion o f incommensurability in connection wit h
dilemmas in Joseph Raz , Th e Morality o f Freedom (Oxford : Oxford University Press,
1986), pp . 357-66.
22. Se e Hector-Neri Castaneda , Thinking an d Doing: Th e Philosophical Founda-
tions o f Institutions (Dordrecht , Holland: D . Reidel, 1975), pp. 191 , 195-201.
23. Nagel , "Fragmentatio n of Value," p. 175.
24. Se e Terrance C . McConnell, "Mora l Dilemma s and Consistenc y in Ethics," i n
Moral Dilemmas, ed . C . W. Gowan s (Oxford : Oxfor d Universit y Press, 1987) ,
pp. 154-73 ; cf . also idem , "Mora l Dilemma s an d Requirin g the Impossible, " Philo-
sophical Studies 2 9 (1976) : 410-11 .
25. Se e McConnell, "Mora l Dilemma s and Consistency," p. 171, n. 2, and Marcus ,
"Moral Dilemmas and Consistency," p. 190. Cf. John Rawls,A Theory o f Justice (Cam-
bridge, Mass.: Harvar d Universit y Press , 1971) , pp . 133-34; David Lyons, Forms and
Limits o f Utilitarianism (Oxford : Oxfor d Universit y Press , 1965) , p . 21; an d Donal d
Notes t o pages 15-18 21 3

Davidson, Essays on Actions an d Events (Oxford : Clarendon, 1980) , p . 34. Fo r mor e

on Davidson's views , cf. Frank Jackson, "Davidso n o n Moral Conflict, " in Actions and
Events, ed . E. Lapore and B. Mclaughlin (Oxford: Basil Blackwell, 1985), pp. 104-15 .
26. Marcus , "Mora l Dilemmas and Consistency," pp . 194-95 .
27. Se e Alan Donagan, "Consistenc y i n Rationalis t Mora l Systems, " i n Moral
Dilemmas, ed. C. W. Gowans (Oxford : Oxford Universit y Press, 1987) , pp. 278-81;
cf. Herman Wouk , Th e Caine Mutiny (Garde n City, N.Y.: Doubleday , 1952) .
28. Donagan , "Consistency, " pp . 280-81.
29. Se e R. M. Hare , "Mora l Conflicts, " i n Moral Dilemmas, ed. C. W. Gowan s
(Oxford: Oxfor d Universit y Press, 1987) , pp. 208-10; cf. also Ear l Conee , "Agains t
Moral Dilemmas, " i n Moral Dilemmas, ed. C. W. Gowan s (Oxford : Oxford Univer -
sity Press , 1987) , pp . 241-43.
30. Furthe r proponents o f dilemma include Philip L. Quinn, who extends the notion
to conflicts between moral and religious requirements in "Moral Obligation, Religious
Demand, an d Practical Conflict," in Rationality, Religious Belief, an d Moral Commit-
ment, ed. R. Audi and W. J. Wainwright (Ithaca, N.Y.: Cornell University Press, 1986) ,
pp. 195-212 , an d Michael Slote , wh o extends it to utilitaria n case s in Beyond Opti-
mizing (Cambridge , Mass.: Harvar d Universit y Press, 1989) , pp. 99-123; cf . idem ,
"Utilitarianism, Mora l Dilemmas , an d Mora l Cost, " American Philosophical Quar-
terly 3 2 (1985): 161-68 .
Foot, "Moral Realism and Moral Dilemma," doe s not seem to take a definite posi -
tion o n dilemmas, at leas t as here interpreted (se e esp. pp. 265-66), although her ar -
gument does allow for something weaker along the same lines (cf. pp. 254-55,267-68).
I omit a number of other intermediate or hard-to-classify views from this brief overview
of central contributions. For a detailed summary of the relevant arguments in the litera-
ture along with bibliographical information, see Sinnott-Armstrong, Moral Dilemmas.
31. Se e Henrik Ibsen , The Works ofHenrik Ibsen (Ne w York: Blue Ribbon Books,
1928), 4: 165-215. Among philosophers' treatments , see esp. Terrance C . McConnell ,
"Moral Blackmail, " Ethics 9 1 (1981): 544-67, and Martha Nussbaum , Th e Fragility
of Goodness: Luck an d Ethics i n Greek Tragedy an d Philosophy (Cambridge : Cam-
bridge University Press, 1986) , pp . 25-82.
32. Christophe r Chernia k and Roy Sorensen (personal communications) have each
independently called my attention to the theory of heuristics as offering a similar account
of decision-making within cognitive limitationsin that case meant to explain appar-
ent irrationalities in probability assessment; see, e.g., Amos Tversky and Daniel Kahne-
man, "Judgment unde r Uncertainty: Heuristics and Biases,"Science 185 (1984): 1124 -
31. Sorense n ha s suggeste d two nice analogies to m y proposed vie w of dilemmas as a
tolerable side effect o f man-made (or "man-tailored") morality . First is the practice of
overbooking flights , somethin g that airline s do systematicall y in the expectation tha t
not every passenger will show upa reasonable expectation, albeit sometimes violated.
Second i s an optica l illusion , which might be considered a n inevitabl e result of what ,
on th e whole , i s the idea l visual system. In th e mora l case , however , I mean to deny
that we have the kind of independent standard needed to dismiss such consequences as
"illusions"; a n appeal to God , say, even if it yielded a religious solution t o a dilemma,
would stil l leave it unresolve d in strictl y mora l terms . Th e sam e hold s fo r practica l
solutions o n utilitaria n or simila r grounds.
33. Se e Williams, Moral Luck, pp . 118-19 , 124-25 ; cf. idem , "Ethica l Consis -
tency," pp . 134-36. For hi s views on mora l luck and blame , cf. idem, Moral Luck,
pp. 20-39, and idem, Ethics and the Limits o f Philosophy (Cambridge , Mass.: Harvard
University Press, 1985) , pp . 176-77.
214 Note s to pages 18-20

34. See , e.g., Foot, "Moral Realism and Moral Dilemma," pp. 251-57, and Michael
Stocker, Plural an d Conflicting Values (Oxford : Clarendon , 1990) , p . 124 ; cf . als o
Christopher W. Gowans, "Mora l Dilemmas and Prescriptivism," American Philosophi-
cal Quarterly 2 6 (1989) : 187-97, for th e defens e o f a position simila r t o Williams's ,
apparently without awarenes s of the overlap.
35. Nagel , "Fragmentatio n o f Value," p . 175 .
36. Fo r the terms "internalism" an d "externalism, " se e W. D. Falk, "'Ought' and
Motivation," Proceedings of the Aristotelian Society 48 (1947-48): 492-510; cf. W. K.
Frankena, "Obligation an d Motivation in Recent Moral Philosophy," in Essays in Moral
Philosophy, ed. A. I. Melden (Seattle: University of Washington Press, 1958), pp. 40-81.
A more precise version of the roug h characterization of moral realism I rely on her e is
provided in David O. Brink, Moral Realism an d the foundations o f Ethics (Cambridge:
Cambridge University Press, 1989) , p . 17 .
Also roughly speaking, internalist realism can be identified wit h an Oxford schoo l
of contemporar y metaethic s develope d mos t systematicall y by follower s o f Joh n
McDowellsee, e.g. , Davi d McNaughton , Moral Vision (Oxford : Basil Blackwell,
1988), pp. 46-50 (cf . p. 15 for the clai m t o represen t McDowell' s views, although
McDowell avoids this terminology himself) .
On th e other hand , standar d externalis t realism seems to correspond t o a Cornell
school of realism; cf., e.g., Brink, Moral Realism and Foundations of Ethics, pp. 37-80.
As Brink points out (p . 78), internalism is linked in the first instance with noncognitivist
versions of antirealism, for which the practical function o f morality or th e connectio n
with emotio n tha t support s i t actuall y supplie s the meanin g o f a mora l judgment .
However, the view has come to b e linked with versions of realism that attemp t t o in-
corporate a Kantian treatment of practical reason. See, e.g., Thomas Nagel , Th e Pos-
sibility o f Altruism (Oxford : Clarendon , 1970) , pp . 7-14 , an d idem , Th e View from
Nowhere (Ne w York : Oxfor d Universit y Press, 1986) , pp . 139ff ; cf . Christin e M .
Korsgaard, "Skepticis m abou t Practica l Reason," Journal o f Philosophy 8 3 (1986) :
5-25. Fo r the alternativ e approach, an d anothe r versio n of the internalism/external-
ism distinction tha t Korsgaar d and other s rel y onapplie d to reason s rather tha n t o
moral judgmentscf. Williams, Moral Luck, pp. 101-13.1 shall rely on Falk's specifi-
cally moral version of the distinction in what follows; though it is sometimes explained
in terms of Williams's version, the two pul l apart for views that allow moral judgments
to be true in application to an amoralist or other agent to whom the y do not give rea-
sons in Williams's sense. Philippa Foot's view, for instance, which I deal with centrally
in chapter 3, section 1, seems to be naturally classified as externalist on moral meaning
but internalis t on reason-givin g force; cf. Brink, Moral Realism an d th e Foundations
of Ethics, pp. 39 , 43 , 61 .
37. Se e David Hume, A Treatise o f Human Nature (Oxford : Clarendon, 1964) ,
pp. 470-76; cf., e.g., p . 457. Cf . Nagel, Possibility o f Altruism, pp. 10-11. The clas-
sification tha t follow s is designed t o avoi d problem s abou t th e prope r interpretatio n
of Hume's view.
38. Kant , Lectures o n Ethics, p. 20.
39. Se e ibid., p. 28, fo r a distinction betwee n two sorts of practical necessitation ,
per stimulos and per motivos, with emotiona l motivation in the former category, and
the latter construed a s broader, includin g also the sort o f objective practical necessita-
tion that applies to actio n on obligation. Cf. Kant's comments on moral motive s and
binding grounds of obligation on pp. 18-19 .
This broader notion of motivation clearly extends beyond the psychologist's sens e
relied o n fo r emotiona l motivatio n in P . S . Greenspan, Emotions an d Reasons: A n
Notes to pages 21-25 21 5

Inquiry into Emotional justification (Ne w York: Routledge, Chapma n and Hall, 1988) ,
e.g. p. 153; cf. R. S. Peters, The Concept o f Motivation (Ne w York: Humanities, 1980) ,
esp. pp . 37-51.1 think that th e psychologist's sens e is at this point th e common one,
so I shall use Philippa Foot's alternative terminology ("reason-giving force" ) fo r th e
Kantian notion; but note that some current authors, e.g., Korsgaard, "Skepticism, " d o
not mak e this distinction in discussing motivational force.
40. Cf . Ronald De Sousa's apparent equatio n of normal and appropriate emotiona l
response i n Th e Rationality o f Emotion (Cambridge , Mass.: Bradford Books, 1987) ,
p. 202. Hume himsel f allow s onl y for a limite d belief-base d distinction betwee n rea-
sonable an d unreasonabl e emotional respons e tha t woul d no t suppl y what w e wan t
here; cf. Hume, Treatise of Human Nature, pp. 415-16. My argument in what follow s
will work fro m m y own accoun t of appropriate emotio n i n Greenspan, Emotions and
Reasons, pp. 83-107. I reexplain this notion her e in simpler terms, however; se e chap-
ter 5, section 2 .
41. Cf . McNaughton, Moral Vision, p , 48 .
42. Se e Hare, "Mora l Conflicts, " pp. 209-10.
43. Fo r discussion o f the textual evidence tha t the chorus faults Agamemnon fo r
not reactin g appropriately , se e Nussbaum, Fragility o f Goodness, pp . 36-37ff .
44. Se e Williams, Moral Luck, p. 27 .
45. Se e Greenspan, Emotions and Reasons, e.g. , p. 96; cf. Williams, Moral Luck,
p. 63.
46. Se e G.E.M. Anscombe, "Modern Mora l Philosophy," i n Collected Philosophical
Papers, Vol. 3 , Ethics, Religion an d Politics (Minneapolis : University of Minnesota ,
1981), pp. 26-42.
47. Se e e.g. Philippa Foot, Virtues and Vices and Other Essays in Moral Philosophy
(Berkeley: University of California Press, 1978) ; John McDowell, "Virtu e and Reason,"
Monist 6 2 (1979): 331-50; Alasdair Maclntyre, After Virtue (Notr e Dame, Ind.: Uni-
versity o f Notre Dam e Press , 1984) ; Williams, Ethics an d th e Limits, of Philosophy;
Stocker, Plural and Conflicting Values; and Michae l Slote , From Morality t o Virtue
(New York: Oxford University Press, 1992) . A n early influence on some of these writ -
ers was Iri s Murdoch; cf . Murdoch, Th e Sovereignty o f Good (London : Ark, 1970) .
48. Anscombe , "Moder n Mora l Philosophy, " pp . 29-33.
49. Se e Foot, Virtues and Vices, pp. 150-51, 162-63; cf. pp. 178-79. The latte r
term seems to capture current talk of "normativity," a s sometimes equate d with moti -
vational force, following Kant; cf. note 3 9 this chapter. In historical term s I take reason-
giving force to amoun t to th e rationally "binding" forc e of obligation, essentiall y the
justificatory force of the moral "ought." For an early contemporary statemen t i n terms
of normativity, see Stephen L. Darwall, Impartial Reason (Ithaca , N.Y.: Cornel l Uni -
versity Press, 1983), pp. 19ff . The notions of reason-giving and motivational forc e now
need to b e pried apar t in classifying view s as internalist; cf . Brink, Moral Realism an d
Foundations o f Ethics, pp. 38-40.
50. For an interesting historical overview, see Louis I. Bredvold, The Natural His-
tory o f Sensibility (Detroit : Wayne State University Press, 1962). For the philosophica l
account I rely on here , see Stephen L . Darwall, "Motiv e an d Obligatio n i n the British
Moralists," Social Philosophy an d Policy 7 (1989) : 139-40 . A helpful discussio n o f
the relevan t conceptua l framewor k i s provided i n Charlott e Brown , "Mora l Sens e
Theorists," i n Encyclopedia o f Ethics, ed. L. C. and C. B. Becker (New York: Garland,
1992), 2: 862-68. For a historical view that bears a certain structural resemblance to
the on e I go o n t o defend , cf . als o Jerom e B . Schneewind, "Pufendorf' s Place i n th e
History o f Ethics," Synthese 7 2 (1987) : 144-46.
216 Notes t o pages 25-29

51. Se e Anscombe, "Moder n Moral Philosophy," p . 27; cf. Joseph Butler , A Dis-
sertation o f th e Nature o f Virtue (London : SPCK, 1970), par. [2], p. 148.
52. Fo r a fairly standard definition of "metaethics" tha t would cover relevant issues
in psychology and other empirical subjects, see William K. Frankena, Ethics (Englewood
Cliffs, N.J.: Prentice-Hall, 1973), pp. 4-5.
53. Se e Jonathan Edwards , Th e Nature o f True Virtue (An n Arbor: University of
Michigan Press, 1960), p . 63; cf. the revie w of recent work i n developmental psychol-
ogy in Martin L. Hoffman, "Development of Prosocial Motivation: Empathy and Guilt, "
in Development o f Prosocial Behavior, ed. N. Eisenberg-Ber g (New York: Academic
Press, 1982) , pp. 281-313.
54. Se e 5d under "guilt " in Th e Oxford English Dictionary (Oxford : Clarendon,
1970), 4: 496; cf . 6a-b under "guilty " o n p . 497 fo r use s of the adjectiv e i n applica-
tion t o conscience , mind, feelings , and th e like , going bac k t o Shakespear e a century
55. Fo r a historical account of "tainting" in relation to legal culpability, see George
Fletcher, Rethinking Criminal Law (Boston : Little, Brown, 1978), pp. 343-50. Cf. the
account of religious ideas of pollution i n Mary Douglas, Purity and Danger (London :
Routledge 8c Kegan Paul, 1966).
56. A colleague at the National Humanities Center, Michael MacDonald (personal
communication), brought to my attention an English account from 162 1 of the pitfall s
of emotional guilt, or the guilty conscience, described in terms of the "sense o f . .. God's
anger justl y deserved" ; se e Robert Burton , Th e Anatomy o f Melancholy (Ne w York :
Tudor, 1927) , pp. 400-404. Note tha t Burton declines to attribute emotional guil t to
Catholics; despite the gradual internalization of religious feeling throughout the Middle
Ages an d th e influenc e o f th e Reformatio n on bot h religiou s traditions, som e differ -
ence in sensibility might be thought to result from the clearer steps laid out in the Catholic
tradition fo r penance and forgivenes s of sin. At any rate, on the accoun t offere d here ,
the mere lack of a word fo r emotional guilt (in Latin, for instance) will not b e decisive.
Thus, Augustine , writing abou t hi s adolescent thef t o f pears, ha s t o wor k withi n
the confines of a sham e vocabulary; cf. St. Augustine's Confessions, trans . W . Watt s
(New York: Putnam, 1922) , 1 : 77-79. Give n that h e describes himself a s undergoing
some emotion associated with the internalized sense of God's anger, my account in what
follows wil l allow Augustine emotional guilt . I have some furthe r comment s i n later
notes on these often fascinating historical and cultural issues, although I cannot attemp t
a full-scal e treatmen t here.
57. Se e esp. Sigmund Freud, Civilization and Its Discontents, trans. J. Strachey (New
York: Norton, 1961), pp. 83-96 ; cf. Friedrich Nietszche, O n th e Genealogy o f Mor-
als, trans. W. Kaufmann an d R. J. Hollingdale (New York: Vintage, 1969), pp. 57-96.
58. Se e esp. Melanie Klein and Joan Riviere, Love, Hate and Reparation (New York:
Norton, 1964), pp . 65-66.
59. Se e Ruth Benedict, The Chrysanthemum an d th e Sword: Patterns o f Japanese
Culture (Boston : Houghton Mifflin , 1946) , pp . 222-23ff .
60. Se e John Stuar t Mill , Utilitarianism (Ne w York: Bobbs-Merrill , 1957) ,
pp. 36-43 .

Chapter 2
1. Se e Sinnott-Armstrong, Moral Dilemmas, pp. 108-68 , for a discussion of tw o
patterns of argument against dilemmas from th e tw o pair s of principles. For ou r pur -
Notes t o pages 30-34 21 7

poses, on e fairl y compac t argumen t wil l b e sufficient t o exhibi t both conflicts . I f w e

allow M fo r physical possibility, the assumption tha t there are dilemmas as interprete d
in term s of conflicting positive ought-judgments comes ou t a s the claim that ther e ar e
acts A and B such that O A and O B and ~M( A & B) . But on this assumption, th e prin-
ciple of closure would yiel d O~B and O~A . Closur e essentially tells us that a n obliga -
tion als o applie s t o anythin g required t o fulfil l it , an d dilemma s amoun t t o case s i n
which each obligatory act requires the negation o f the other. So now we have two pair s
of ought-statement s wit h contradictory objects : OA & : O~A an d O B & O~B . Eve n if
they are practically inconsistent, however, neithe r pair of statements amount s to a truth-
functional contradiction . The same holds for the pair that conjoins their objects t o yield
obligations with contradictory object s according to the principle of agglomeration: O( A
& ~A ) and O(B & ~B) . But the standard syste m of deontic logic takes impossibl e state s
of affair s a s impermissible ; with P as it s operato r fo r permissibility , that is , it let s u s
conclude fo r an y A that ~P( A & ~A) . (Cf. (C4') i n F011esda l and Hilpinen , "Deonti c
Logic," p. 13. ) Th e assumptio n tha t "ought " implie s "permissible " wil l therefore let
us deriv e a contradictio n withi n th e standar d system , sinc e in applicatio n t o eac h o f
the conjoint ought-statements jus t derived it yields a statement of the form P(A & ~A).
Alternatively, a contraposative versio n o f "ought"-implies-"can " yield s ~O(A & -A )
for an y A and henc e a direct contradictio n o f our derive d ought-statements .
2. Williams' s positio n essentiall y rests o n applying to practical "ought, " taken a s
a conclusion of deliberation, a principle of "exclusivity" tha t combines th e two deonti c
principles he thinks dilemmas force us to choose between: "ought"-implies-"can " an d
agglomeration (see Moral Luck, pp. 118-19) . A version of this point occurs in his ear-
lier treatment o f what h e calls deliberative "ought"; see idem, "Ethica l Consistency,"
pp. 134-36 .
However, I shall not deal here with Williams's argument on this subject in full detail ,
partly because its terms shift in a way that makes it resist perspicuous treatment. Besides
equating practical "ought" with "ought" as a conclusion of deliberation (a point I shall
deal wit h late r i n thi s chapter), William s apparently make s differen t assumption s i n
different place s a s t o whethe r a practica l ough t i s by definitio n "conclusive, " o r all -
things-considered (cf . Moral Luck, pp. 118-19 , 124 , n . 3).
For the defens e o f a position simila r to Williams' s on practical ought i n dilemmas ,
see Gowans, "Mora l Dilemma s and Prescriptivism, " an d idem , Innocence Lost: A n
Examination of Inescapable Moral Wrongdoing (Ne w York: Oxford Universit y Press,
3. Se e P. S. Greenspan, "Conditiona l Ought s an d Hypothetica l Imperatives,"Jour-
nal of Philosophy 7 2 (1975): 259-79; cf. idem, "Derive d Obligation : Som e Paradoxe s
Escaped," Ph.D . dissertatio n (Cambridge , Mass.: Harvar d University , 1972).
4. Thi s depend s on a version o f th e principl e of deonti c closure , whic h i n roug h
form tell s us that anythin g necessary t o satisf y a n ought i s itself obligatory. I shall later
suggest some limitations on the principle, but even in modified form it should still allo w
for the inference in question here. In general terms , my present argument brings to mind
Nagel's argumen t o n reason s i n Possibility o f Altruism, p. 36 . Bu t not e tha t Nage l
considers only th e direc t derivatio n of earlie r instrumenta l reason s fro m a statemen t
dated a t the time assigned to action, whereas m y suggested accoun t date s ought-state -
ments at the time of utterance and allows fo r distinct dates o n their objects , or the act s
they require.
5. I n Grice's terms, what I ascribe to a general use of "ought" might be thought of
as part o f it s "conventiona l meaning" a s distinc t fro m "conversationa l implicature, "
or the shiftin g commitment s of individual speakers; see Paul Grice, Studies i n the Wa y
218 Note s to pages 3 5-3 7

of Words (Cambridge , Mass.: Harvard University Press, 1989), p. 41. An analogy might
be the way the word "but " implicates contrast betwee n tw o conjuncts, except tha t ou r
example migh t be said t o involv e two differen t conventiona l uses .
Among example s no t relate d t o "ought " itself , the bes t I can thin k o f hav e to d o
with litera l versus recognized nonlitera l uses of the same form of words. For instance ,
the expression "Go d damn!" as used today no longer prescribe s divine condemnatio n
but instea d function s mainl y to expres s anger. I t therefore "implicates " ange r on th e
part o f the speaker, a s a function o f conventional meaning. Bu t it can also be used with
almost th e opposite intent , in a hoot o f victory (when one's team scores a touchdown ,
say). It might even be thought tha t this latter use is common enoug h t o count a s a fur-
ther conventiona l meaning . I n either case , th e expressio n serve s a t leas t on e genera l
emotive functio n i n the language that is distinct from it s literal meaningand possibly
from a given speaker's persona l intent.
Assuming that practical force involves a similar kind of functional o r conventiona l
meaning, w e might want t o sa y that i t "implicates " rathe r tha n implies "can " in any
strict logica l sense . Remembe r tha t Gric e invente d th e forme r ter m t o contras t wit h
logical implicatio n (see pp. 24-25; cf. pp. 121 , 341). The important poin t for our pur-
poses, however, i s that thi s does not make "can" follow simply as a matter o f conver-
sational implicature , as Sinnott-Armstrong maintains (Moral Dilemmas, pp. 121-26).
Rather, th e principle would see m to b e definitive o f practical "ought, " a s designed t o
serve a particular (action-guiding ) function i n the language .
6. Se e Hume, Treatise on Human Nature, p. 457; cf., e.g., Foot, Virtues and Vices,
pp. 78-80 .
7. See , e.g., Williams, Moral Luck, p. 119.
8. See , e.g., Hector-Neri Castaneda , Th e Structure o f Morality (Springfield , I11.:
Charles C. Thomas, 1974) , p. 64; Gilbert Harman, The Nature o f Morality (Ne w York:
Oxford Universit y Press, 1977) , pp. 117-19 ; an d Josep h Raz , Practical Reason an d
Norms (Princeton , N.J.: Princeton University Press, 1990) , pp. 29-32.
9. A version of this view seems to surface, e.g., in Williams, Moral Luck, pp. 10 9 -11;
cf. als o Nagel , Possibility o f Altruism, pp. 8-9, an d Foot , Virtues an d Vices, p . 15 2
(though I gather that Foot would now reject this argument [personal communication]).
10. Se e Aristotle, Nicomachean Ethics, 1147a27-31; cf. idem, Movement o f Ani-
mals, 701all e t circa, and idem, De Anima, 433al7 et circa.
11. Se e Lewis White Beck , A Commentary o n Kant's Critique o f Practical Reason
(Chicago: University of Chicago Press , 1960) , p . 39 , fo r a reconciliation o f Kant's ref-
erences t o practica l reaso n a s determinant of and a s identical to the will.
12. Se e Davidson, Essays o n Actions and Events, p. 39 .
13. Williams , Moral Luck, p. 119.
14. Se e John M . Cooper , Reason an d Human Good i n Aristotle (Indianapolis:
Hackett, 1986) , pp. 23-46.
15. Se e Stocker, Plural and Conflicting Values, pp . 51-84, fo r a n argumen t tha t
Aristotle himsel f makes room fo r conflict. However, m y use of Aristotle's view of prac-
tical reasonin g her e is not mean t to commit hi m to dilemmas.
16. Cf . Williams, Moral Luck, e.g. p. 12.
17. Cf . Alasdair Maclntyre, "Wha t Moralit y I s Not," in The Definition o f Moral-
ity, ed . G . Wallac e an d A.D.M . Walke r (London : Methuen , 1970) , pp . 27-31. See
Greenspan, "Conditiona l Oughts and Hypothetical Imperatives," p. 275, n. 13 , for an
argument against one natural way o f interpreting the result s of a first-person ought as
binding fro m a third-perso n standpoint.
18. Se e William Styron , Sophie's Choice (Ne w York: Bantam, 1980), p . 589; cf.
Notes t o pages 38-48 21 9

P. S. Greenspan, "Mora l Dilemmas and Guilt," Philosophical Studies 4 3 (1983): 117 -

19. Williams , "Ethica l Consistency, " p . 136.
20. Williams , Moral Luck, pp. 78-79; cf. p. 74.
21. Cf . Sinnott-Armstrong, Moral Dilemmas, for th e us e of this compound over -
ridingness notion i n defense o f dilemmas.
22. Williams , Moral Luck, p. 119.
23. Williams , "Ethical Consistency," p . 123, and idem, Moral Luck, p. 79; cf. Moral
Luck, p. 125.
24. Se e Slote, Beyond Optimizing; cf . Stocker , Plural an d Conflicting Values,
pp. 314-15.
25. Th e two principle s of standard deontic logic that togethe r hav e this resul t are
given as (Cl) and (C2 ) in F011esdal and Hilpinen , "Deonti c Logic, " pp. 8-9; but not e
that (C2) , which denies the possibility of exhaustive prohibition, beg s the question fo r
our purposes .
26. Se e esp. Roger Wertheimer, The Significance o f Sense: Meaning, Modality, an d
Morality (Ithaca , N.Y. : Cornel l Universit y Press , 1972), pp. 82-83, 110-11.
27. See , e.g., Williams, Moral Luck, p. 60.
28. Se e Ross, Right and the Good, p. 28.
29. Se e Wertheimer, Significance o f Sense, pp. 117-18 ; cf. pp. 104 , 109 .
30. Not e tha t th e "total " bod y o f evidence on a standar d readin g mean s al l rel-
evant reasonsnot jus t all available reasons o r all reasons actuall y held b y the agent ,
as on Davidson' s interpretatio n of "all-things-considered " i n Essays o n Actions and
Events, p. 40. Taking the latter notio n as evidential thus should not rule out an objec-
tive notion o f dilemmas. To sa y that tw o competing prohibition s hol d al l things con-
sidered o n thi s account will not simpl y tell u s about th e agent' s epistemi c position
that he is not i n possession o f adequate reasons for choice or the like (which in weighted
cases, o f course , ma y wel l be false)but rather abou t th e requirement s o f hi s situa-
tion. M y perceptual imagery is meant to capture something abou t th e objective deter-
mination of rational weight, not t o sugges t that i t depends o n a n observer .
31. Cf . Williams, Moral Luck, pp. 74 , 79.
32. Cf . Marcus, "Mora l Dilemma s and Consistency, " p . 193 ; my own vie w here
comes closer to Williams's. Note that the initial examples of conflicting oughts in Foot's
critique o f Williams seem to b e cases o f overriding that ar e bette r suite d t o Marcus' s
view; see Foot, "Moral Realis m and Mora l Dilemma, " p. 251.
33. Lemmon , "Mora l Dilemmas, " p. 106; Williams , Moral Luck, pp. 124-30 .
34. Williams , "Ethical Consistency, " p . 129.
35. Se e Alan Ross Anderson, "The Formal Analysis of Normative Systems, " in The
Logic of Decision and Action, ed. N. Rescher (Pittsburgh: University of Pittsburgh Press,
1967), pp. 147-213 . Cf. A. N. Prior, "Escapism: The Logical Basis of Ethics," mEssays
in Moral Philosophy, ed . A. I. Melden (Seattle : University of Washington Press , 1958) ,
pp. 135-46 .
36. Tha t is, it denies that th e object of permission wil l incur any deontic sanction .
Even if OA and O B are interpreted as referring to two differen t sanctions, S A and S B, as
needed t o allo w for a n escapist account o f dilemmas, fulfillment o f either one of these
oughts i n a dilemm a would see m t o incu r some sanctione.g., sinc e A entails ~B , it
would incu r SBwhich would see m to b e enough to mak e it impermissible.
37. Se e Greenspan, "Moral Dilemma s and Guilt," p. 118 ; cf. pp. 121-22 . I pushed
the argumen t a bi t furthe r i n connectio n with som e assumption s o f deonti c logi c in
"Sophie's Choices : More o n Exclusiv e Requirement" (unpublished) .
220 Notes t o pages 48-58

38. Se e Von Wright, Essay in Deontic Logic, pp. 78-81; cf. Peter Vallentyne, "Pro-
hibition Dilemmas and Deontic Logic," Logique et Analyse 117-1 8 (1987) : 120, n. 5.
The distinction is made more explicitly and defende d at length in idem, "Tw o Type s of
Moral Dilemmas, " Erkenntnis 3 0 (1989): 301-18.
39. Thi s assumes that the tiebreaker does not cancel out any of the original dilem-
matic prohibitions and tha t prohibitio n applies to all particular act s satisfyin g a given
description, or tokens of a certain act-type. This i s another difference fro m permission ,
which tell s us only tha t ther e is nothing objectionabl e about a n ac t insofa r as it satis-
fies a given description, or amount s to a token o f a certain type , so that th e particular
act i n question may stil l b e prohibited.
40. Cf . Sinnott-Armstrong, Moral Dilemmas, pp. 121-26, 160-61, for th e limi-
tation o f th e ought-implicatio n principles to conversationa l implicatur e as extende d
beyond practica l uses of "ought. "
41. Cf . (C2 ) in F011esdal and Hilpinen , "Deonti c Logic, " p . 9 .
42. Gottlo b Frege , "Negation, " i n Translations from th e Philosophical Writings
of Gottlob Frege, ed. Pete r Geac h an d Ma x Blac k (New York: Philosophica l Library,
1952), p . 125.
43. Se e Michael Dummett , Frege: Philosophy o f Language, 2n d ed . (London :
Duckworth, 1981) , p . 317.
44. Cf . ibid., p . 335.
45. Cf . Ruth Barca n Marcus , "Mor e abou t Mora l Dilemmas " (unpublished ; de -
livered a t th e Chapel Hill Coloquiu m in Philosophy, 1980) , pp . 14-15 .
46. Se e Fellesdal and Hilpinen, "Deontic Logic, " p . 17 . Note that Marcus' s second -
order regulative principle to avoi d conflicts seems to make best sense in these terms
since at the time assigned to action in a dilemma it may well be too lat e for the agent t o
avoid it , though hi s doing s o would stil l count a s the idea l state o f affairs .
47. Cf . esp. Chisholm, "Contrary-to-Duty Imperatives. " Cf . Greenspan, "Derive d
Obligation," for m y discussio n o f many of these issues.
48. Cf . F011esdal and Hilpinen , "Deonti c Logic, " pp . 26-31. I shoul d not e tha t
this seems not to be true of a more recent attempt t o handle dilemmatic oughts as cases
of defeasibl e reasoning , b y abandoning modal logi c altogether (i n favor of "nonmon -
otonic" logic) as the appropriate foundation for deontic logic. See John F . Horty, "Mora l
Dilemmas and Nonmonotonic Logic," Journal o f Philosophical Logic 23 (1994): 35-65;
cf. idem , "Deontic Logi c a s Founded on Nonmonotonic Logic, " Annals of Mathemat-
ics and Artificial Intelligence 9 (1993): 69-91. The analogie s illustrating Horty's pro-
posal (propositions roughly true but conflicting as stated, such as "Birds fly" and "Pen -
guins don't fly," on th e assumption that penguin s are birds) suggest tha t it would no t
really captur e dilemmas, i n th e sens e o f specifi c conflicting directives fo r action , bu t
rather just general ought-statements whose conflic t in application t o a particular cas e
presumably would be handled by withdrawing one of them. (W e would no t conclude ,
for instance , that Tweet y th e Pengui n does fl y qu a bir d a s wel l a s failin g t o fl y qu a
penguin, but rather that "Bird s fly" is not reall y applicable to Tweety's case.) Horty's
point, as I understand it, is that we cannot appea l t o the content o f such statements
to built-i n exceptio n clauses and th e liketo tell us which on e t o withdraw . Hi s pro -
posed syste m of deonti c logi c would cove r conflictin g rules an d henc e woul d yiel d a
logic applicable to our ordinar y deontic judgmentsperhaps one that more accurately
reflects ou r ordinar y use of them in moral reasoning. But it would no t see m to yiel d a
logical systematization of our deontic judgments, or a "deontic logic " i n the usual sense,
except a s limited t o th e genera l level.
49. Williams , "Ethica l Consistency," p. 132.
Notes t o pages 59-69 22 1

50. O n dilemmas and simplicity, see Bernard Williams's comments on Isaiah Berlin's
value pluralism in his introduction t o Concepts an d Categories: Philosophical Essays,
by I. Berlin (London : Hogarth, 1978) , pp . xvi-xvii .
51. Se e esp. Hector-Neri Castaneda , "O n th e Semantic s of th e Ought-to-Do, "
Synthese 2 1 (1970): 451; cf. idem, Thinking an d Doing, pp. 207-8, 248-53.
52. Se e Castaneda, Thinking an d Doing, p. 224.
53. Fo r Castaneda' s overal l account (whos e ful l complexit y I cannot attemp t t o
capture here) , see esp. ibid., pp . 154-79 . Castaneda' s notio n o f th e "Legitimacy " o f
practitions as the value analogous to the truth of propositions, and o f deontic trut h as
necessary Legitimacy , is summed u p i n a very helpful overvie w of Castaneda' s views ;
see Michael E. Bratman, "Castaneda' s Theor y o f Though t an d Action, " i n Agent,
Language, and the Structure o f th e World, ed . J. E . Tomberlin (Indianapolis: Hackett,
1983), pp. 152-55; fo r mor e o n th e connectio n t o a n agent' s ends , includin g some
problems with cause-effect relations , se e pp. 155-59 .
54. Cf . the account of needs in contrast t o desires given in David Wiggins, Needs,
Values, Truth (Oxford : Basil Blackwell, 1987), pp . 5-16 .
55. Cf . Bratman, "Castaneda' s Theor y o f Thought an d Action, " p . 156 , where
oughts on Castaneda' s accoun t ar e taken as indexed to a specific promise .
56. Se e esp. Castaneda's accoun t of the thre e "dimensions " o f morality in Struc-
ture of Morality, pp . 175-226 ; cf. his account of the consistency of normative systems
by analog y to legal systems in Thinking an d Doing, pp. 225-28.

Chapter 3
1. Williams , Problems o f th e Self, pp . 204-5; cf. Wiggins, Needs, Values, Truth,
pp. 87-137 .
2. Cf . also P. S. Greenspan, " A Case of Mixed Feelings : Ambivalence and the Logic
of Emotion, " i n Explaining Emotions, ed. A . O. Rort y (Berkeley : University of Cali -
fornia Press , 1980) , pp . 223-50. I there deal with cases that exhibit a logical structure
Williams bypasse s (see Williams, "Ethica l Consistency, " p . 117) , along wit h th e sor t
of logical behavior that he takes to mark off desires. That is , the unqualified evaluative
beliefs that reflec t emotiona l ambivalenc e may b e said to b e retained i n residual form
by being qualified rathe r than simpl y eliminated in the fac e o f conflict.
3. Se e Foot, "Mora l Realism and Moral Dilemma, " esp. pp. 254-57, 265-67 .
4. Ibid. , p. 262; cf. pp. 267-68 .
5. Se e McNaughton, Moral Vision, pp . 13940 , 48-50; and Brink , Moral Real-
ism and Foundations o f Ethics, pp. 43-50 .
6. Se e Foot, Virtues and Vices, pp. 157-73 ; cf. p. 17 9 for the distinction from moti-
vational force. For a study of Foot tha t overlaps on important points with the account
that follows but without the implications for general internalism, see Simon Blackburn,
"The Fligh t to Reality," in Virtues an d Reasons, ed. R. Hursthouse and G . Lawrence
(Oxford: Oxfor d Universit y Press, 1995) .
7. Foot , Virtues and Vices, pp. 74-80 .
8. Cf . the account in terms of "fittingness" i n Maurice Mandelbaum, The Phenom-
enology o f Moral Experience (Baltimore : Johns Hopkin s Universit y Press, 1969) ,
pp. 59-71 .
9. Foot , Virtues and Vices, pp. 181-88 .
10. Ibid. , p. 186.
11. Ibid. , pp. 183-85 .
222 Notes t o pages 69-78

12. Ibid. , p. 153.

13. Ibid. , pp. 79-80.
14. Ibid. , see, e.g. p. 152.
15. Ibid. , p. 163.
16. Cf . P. S. Greenspan, "Behavio r Control an d Freedo m o f Action," Philosophi-
cal Review 8 7 (1978): 25-40.
17. Cf . Ben Spiecker, "Education and th e Moral Emotions," i n Philosophical Issues
in Moral Education and Development, ed . B. Spiecker and R. Straughan (Milton Keynes,
U.K.: Open Universit y Press, 1988) , pp . 43-63.
18. Fo r a review of the psychological literature suggesting a developmental account
of guilt as based on empathy (but omitting the earlier stage I include here), see Hoffman,
"Development of Prosocial Motivation, " pp. 297-305.1 discus s this evidence further
in chapter 4.
19. Se e Laurence Thomas, Living Morally: A Psychology of Moral Character (Phila-
delphia: Temple University Press, 1989) , pp . 76-80 , for a negativ e view of guilt as a
self-interested an d henc e a n essentiall y antimora l motive. It i s compatible wit h any -
thing I have in mind, I should say, that we would prefe r to have children (an d friend s
and others ) act out o f some motiv e other tha n guilt ; but I take i t that the sam e hold s
for mora l ought-statements .
20. Fo r m y general account of emotions a s reasons operating b y way o f the nee d
to escap e discomfort , se e Greenspan, Emotions an d Reasons, pp . 153ff . For a n em -
pirical account of emotions that pays less attention to conceptual issues but otherwise
overlaps with mine , see Paul L. Harris, Children an d Emotion: The Development o f
Psychological Understanding (Oxford : Blackwell, 1989); with reference to the sugges-
tions I develop here , se e pp. 44-45, 93-94, 98.
21. Cf . C . L. Stevenson, Pacts and Values (Ne w Haven, Conn. : Yal e University
Press, 1963) , p. 13.
22. Thi s might be thought of as a variant of Grice's "conversationa l implicature "
but with the ordinary purposes o f moral discours e replacing the communicative aims
of an ordinary speaker. See Grice, Studies i n the Wa y of Words, pp . 26-28.The notio n
of "didacti c import " i s not mean t to cove r jus t any emotiona l overtones o f a wor d
resulting from the way it is commonly taught, such as shame in connection with word s
for th e genitals, bu t just those essential t o settin g up it s role in the language. I charac-
terized the latter as its "functional" meanin g in chapter 2, section 1 ; in Grice's terms it
seems to amoun t t o "conventional " meanin g (se e my note 5 , chapter 2) . There i s no
institution of genital discourse designed specifically to control behavior . To claim that
moral discours e is another matte r is not t o rul e out th e possibility of a societyor in a
nice exampl e that wa s suggeste d t o me , som e ultralibera l parents i n California (pre-
sumably speaking their own language)that did without terms set up for their action-
guiding function. The point i s just that the y would thereby be doing without "moral"
language i n our (conventional ) sense. This i s not t o sa y that mora l discours e ha s n o
further functions that they might retain, as part of an institution that could still be called
"moral," albeit in a somewhat differen t (mor e "laid-back") sense .
23. Se e esp. Wiggins, Needs, Values, Truth, pp. 185-211; see also John McDowell ,
"Values an d Secondar y Qualities, " i n Morality an d Objectivity: A Tribute t o J. L .
Mackie, ed. T. Honderich (London : Routledge & Kegan Paul, 1985), pp. 110-29. Note
that Wiggin s considers "realism " somethin g of a misnomer for hi s view, although he
would accep t McDowell' s "cognitivism " (se e p. 209).
24. Cf . McNaughton, Moral Vision, pp . 21-23, 46-47, 106-13.
25. Se e John McDowell , "Ar e Moral Requirement s Hypothetical Imperatives?,"
Notes t o pages 79-89 22 3

Proceedings o f th e Aristotelian Society Suppl . 52 (1978) : 13-29 , an d idem , "Virtu e

and Reason," 331-50. For help in understanding McDowell's view , I am indebted bot h
to McNaughton, Moral Vision, and to Jonathan Danc y (personal communication), who
provided comment s on th e ensuin g discussion. Se e Dancy's Moral Reasons (Oxford :
Basil Blackwell, 1993), which advances a version of McDowell's view but which reached
me too lat e for consideration here .
26. McDowell , "Virtu e and Reason," p. 348, n. 5.
27. Cf . Nagel, Possibility o f Altruism, pp. 29-32.
28. McDowell , "Mora l Requirements, " p. 16.
29. Ibid. , p. 28.
30. McDowell , "Virtu e and Reason," p . 333.
31. Ibid .
32. McDowell , "Mora l Requirements, " pp. 26-27.
33. Ibid. , p. 21.
34. Cf . McDowell, "Virtu e and Reason," p . 344. Annette Baier, in her comments
on m y Emotions and Reasons (America n Philosophical Associatio n Pacifi c Divisio n
meetings, March 1990) , noted that the point about emotions can be traced to Descartes' s
treatment i n "The Passions o f the Soul" ; se e Th e Philosophical Works o f Descartes,
trans. E. S. Haldane and G.R.T. Ross (Cambridge: Cambridge University Press, 1970) ,
Vol. 1 , Pt. Second, Art. 74 , p. 364.
35. McDowell , "Mora l Requirements," p. 22.
36. Cf . my defense o f ambivalence as rationally appropriate i n Greenspan, "Cas e
of Mixe d Feelings" ; cf . idem, Emotions an d Reasons, pp. 109-36.
37. Se e J. L . Mackie, Ethics: Inventing Right and Wrong (London: Penguin, 1987) ,
pp. 38-42.
38. Ibid. , pp. 23-24, 40, 35; cf. pp. 23, 40.
39. Se e Foot, Virtues and Vices, pp. 132-47 ; cf. pp. 100-5 .
40. Mackie , Ethics, pp. 51-52, 59.
41. Ibid. , pp. 15 , 16.
42. Ibid. , esp. p. 31.
43. Se e McDowell, "Value s and Secondar y Qualities, " pp . 120-22 . McDowel l
acknowledges th e limitations of the secondary qualities analogy o n p. 120; the poin t
of the analogy seems to be the distinction between general and particular subject-inde-
pendence (independenc e of minds generally vs. the speaker) , with th e latte r take n as
sufficient fo r a claim to b e part o f "reality." Cf . also McDowell's discussio n of related
issues in "Aesthetic Value, Objectivity, and the Fabric of the World," in Pleasure, Pref-
erence and Value, ed. E. Schaper (Cambridge: Cambridge University Press, 1983) , pp .
44. Cf . John Rawls , "Kantia n Constructivism in Moral Theory, " J ournal o f Phi-
losophy 7 7 (1980) : 515-72. Since Rawls's late r politica l version of constructivism is
set up to accommodate intuitionist realism, it should also be compatible with the alter-
native I propose; cf . John Rawls , Political Liberalism (Ne w York: Columbi a Univer-
sity Press, 1993) , p . 95 .
45. Mackie , Ethics, see esp. pp. 106-15 .
46. Ibid. , cf. pp. 27-29, 35, 48-49, 65-66, 75.
47. Se e Plato, Protagoras, 320c8-323a4 , and Mackie , Ethics, pp. 108 , 113-15.
48. Cf . John Searle, "How t o Derive 'Ought' fro m 'Is'," Philosophical Review 7 3
(1964): 43-58; cf. Mackie, Ethics, pp . 66-72, esp. p . 67.
49. Mackie , Ethics, p. 79; cf. pp. 73-79, 81, 111-13 .
50. Ibid. , p. 33.
224 Note s to pages 90-95

51. Se e my discussion o f "having " a reason i n P. S. Greenspan, "Unfreedo m an d

Responsibility," i n Responsibility, Character, and the Emotions, ed. F. Schoeman (Cam-
bridge: Cambridge University Press, 1987) , p. 75, for a way of retaining the claim that
the agen t i n suc h a cas e "has " a reason t o d o wha t moralit y requires . Cf . Williams ,
Moral Luck, pp. 106-7 , fo r the basi s of the denia l of such "external " reasons i n the
insistence that a reason necessaril y motivates a rational agent . Extendin g externalis m
to th e version tha t applie s to reason-givin g a s well as motivational forc e allows fo r a
fuller answe r t o Mackie' s "argumen t fro m queerness" ; cf . Richard Garner , "O n th e
Genuine Queerness o f Moral Propertie s an d Facts, " Australasian Journal o f Philoso-
phy 6 8 (1990) : 137-46 .
52. Cf . Aristotle, Nicomachean Ethics, esp. Bk. II, Sees. 1 and 3. Note that ordi -
nary moral consciousnes s i s in question here . For philosophers an d other s inclined t o
reflect o n and analyz e morality, we might of course include a further stageth e evalu-
ation o f mora l code s accordin g t o metaethica l norms , whic h include s standard s o f
viability of the sor t I go on t o describe.
53. Fo r a recent antirealist attempt t o provide a limited kind o f objective basis for
ethics, with particula r referenc e t o th e issu e o f authority, see Allan Gibbard , Wise
Choices, Apt Feelings: A Theory o f Normative Judgment (Cambridge , Mass.: Harvard
University Press , 1990) , pp . 153-250 . O n Gibbard' s view , however , th e authority in
question comes out as that of a speakeror of particular moral utterancesrather tha n
of certain sorts of claims or considerations independentl y of their source , which I take
to b e the issue relevant to ethics .
54. Mackie , Ethics, pp. 36-39, 83-92.
55. Cf . the discussion of "explanatory values" in Geoffrey Sayre-McCord , "Mora l
Theory and Explanatory Impotence," i n Essays on Moral Realism, ed. G. Sayre-McCord
(Ithaca, N.Y.: Cornel l University Press, 1988) , p. 279.
56. Suc h appeal provides the basis for contractarian arguments ; cf. Rawls, Theory
of Justice, pp . 15 , 13 .
57. Cf . Raz, Morality o f Freedom, pp . 198-203 , 250-55. Alternatively , to th e
extent tha t social groups ar e susceptible to treatment a s moral agents , on e might take
them as having duties to their members, including duties o f fair treatment , perhaps as
an element of the sor t o f social atmosphere of tolerance and simila r libera l values that
Raz commends. On a contractarian approach , a group's clai m on its members' adher -
ence might be based on performance of such duties. An argument along these lines was
suggested t o m e b y th e treatmen t o f societ y a s the primar y beare r o f th e dut y t o ai d
needy individuals in David Copp, "Th e Right to an Adequate Standard o f Living: Jus-
tice, Autonomy, and the Basic Needs," Social Philosophy an d Policy 9 (1992): 231-61.
Just as the duty to aid on this account gets parceled out to individual members via taxa-
tion, moreover, othe r social duties to individuals might be held to transfer to them too,
for a socially based accoun t of "duties t o oneself " an d simila r issues.
My central argumen t here is meant t o b e independent o f these suggestions. How-
ever, for a systematic defense of some of the social notions I rely on, se e Copp, "What
Collectives Are: Agency, Individualism and Legal Theory," Dialogue 23 (1984): 249-69,
and idem , "Th e Concept o f a Society, " Dialogue 3 1 (1992) , 183-212. I n Morality,
Normativity, an d Society (Ne w York: Oxford University Press, 1995) , Cop p defends a
social artifact vie w that overlap s in many ways with my own. One poin t o f difference ,
though, is Copp's focus on the overall society as the source of moral norms, as oppose d
to th e variou s smaller group interaction s that suppor t th e teachin g of moralit y via
emotion o n m y account.
58. Se e Hoffman, "Developmen t of Prosocial Motivation," p. 282. My own attempt
Notes t o pages 95-97 22 5

to make philosophic sense of empathetic emotion in Greenspan, Emotions an d Reasons,

pp. 62-79, resembles th e accoun t o f "fello w feeling " in Adam Smith, The Theory o f
the Moral Sentiments, i n British Moralists: 1650-1800, ed. D. D. Raphael (Indianapolis:
Hackett, 1991) , 2: 201-6. Recent discussion s i n cognitive science also dea l wit h th e
topic unde r the headin g of "simulation" ; see , e.g., Robert Gordon , Th e Structure o f
Emotions (Cambridge: Cambridg e Universit y Press, 1987) , pp. 149-55.
59. Not e that the reference to emotional interactionmeanin g mutual influence ,
of the sort that supports emotional learning , as described in section 1may be used to
limit the moral implications of our more extended sympathy with animals. Indeed, emo-
tional identification sometimes extends even to inanimate objects: I cry out in pain when
my car hits a pothole, say ; see also the findings reported i n Alvin Goldman's APA Paci-
fic Divisio n Presidentia l Address , "Empathy , Mind , an d Morals, " Proceedings an d
Addresses o f th e American Philosophical Association 6 6 (1992) : 27-28.
On th e other hand , an argument of the sort I offer fo r not excluding members of a
subclass o f one' s ow n societ y fro m th e scop e o f ou r mora l obligation s als o ma y b e
extended to members of other societiesfo r instance, by noting that the habits of mind
encouraged by subjecting them to hostile treatment would b e likely to turn inward when
it lacked a target, thus threatening the stability of the moral code. However, these issues
(and others I omit) deserve a fulle r treatmen t than I can provide here.
60. Se e Mill, Utilitarianism, pp . 36-43, an d Hume , Treatise o f Human Nature,
p. 456; cf. p. 470. Mill holds that human emotional sensibilit y in no way restricts th e
content o f a moral code. Hume' s account, o n the other hand , restrict s sympath y to a
later stag e i n the establishment o f justice as a moral virtue; see, e.g., p. 498.
Cf. also Rawls , Theory o f Justice, pp . 453-512, for an account of moral emotion s
relevant t o contractarianis m tha t I take t o b e essentially "top-down. " Th e accoun t
presented her e overlaps at important point s with Rawls's defens e of the "stability " o f
the principles of justice, but Rawls's treatmen t of moral psychology in connection wit h
stability seems to b e intended onl y a s confirmation o f the principles , not a s an inde -
pendent basi s for them. Thus, o n Rawls's view , any feasible principles of justice must
be capable o f support fro m a natural set of moral sentiment s amounting to ou r sens e
of justice; so they would have to satisf y th e requirement of teachability and other gen-
eral presuppositions tha t ma y brin g i n fact s abou t huma n emotional nature . Bu t th e
choice o f principle s i n Rawls' s "origina l position " i s justified withou t mor e specifi c
reference t o huma n motivationin contrast t o m y own picture of moral emotion s a s
providing a particular episodic basis for the teaching of moral language and behavio r
that generates moral norms .
61. Cf . David Wong, Moral Relativity (Berkeley : Universit y of Californi a Press,
1984); cf. also Castaneda' s appea l t o correctio n o f actua l moral code s i n ligh t o f a n
ideal of morality in Structure o f Morality, pp . 185-89 . See also Richard Brandt, "To-
ward a Credibl e Form of Utilitarianism," i n Morality an d th e Language o f Conduct,
ed. H-N. Castaneda and G. Nakhnikian (Detroit: Wayne State University Press, 1965) ,
pp. 107-43 , for a similar use of the notion o f an ideal moral code, thoug h without th e
historical referenc e that thes e othe r view s presuppose. M y own suggestion s migh t b e
thought of, for that matter, as yielding something like a social version of Brandt's rela-
tivist "idea l observer " vie w in metaethics ; see, e.g., Richard Brandt , Ethical Theory
(Englewood Cliffs , N.J. : Prentice-Hall, 1959) , pp . 173-74 .
62. A similar basis for a version of realism in facts abou t social rationality is pro-
vided by Peter Railton, "Moral Realism, " Philosophical Review 9 5 (1986): 163-207;
however, Railton's approach is tied to utilitarianis m in a way I want to avoid. I leave
it as an exercise to the reader, incidentally, to decide whether my own view also counts
226 Notes t o pages 97-109

as a versio n o f "naturalism. " Th e issu e i s complicated b y m y willingness to gran t

that th e notio n o f flourishin g i s partly evaluative , possibly eve n in a sens e w e ca n
think of as "moral." McDowell has recently questioned th e interpretation of "natu -
ral" i n relation to Aristotle; see "Reason an d Nature" (unpublished ; delivered at the
meetings o f th e America n Philosophical Association , Easter n Division , Decembe r
1992). M y ow n hunc h (whic h I hop e t o explor e i n some furthe r work) i s that th e
categories of the natural and the normative are not so distinct at the social level, even
with "natural " take n in its more modern sense. In any case, the issue will be affecte d
by the option s I leave open fo r the metaethica l underpinnings of social artifact real-
ism i n chapter 6.
63. I n partial explanation of the tendency of moral philosophers t o interpret real-
ism i n thi s way (cf. , e.g., Gibbard, Wise Choices, Ap t Feelings, pp . 33-34) , I offe r
Keynes's observation on G . E. Moore's literal-mindedness: "Moore had a nightmare
once in which he could not distinguish propositions fro m tables. But even when awake,
he could not distinguish love and beaut y and trut h from th e furniture" (J. M. Keynes,
Two Memoirs [London , Ruper t Hart-Davis, 1949] , p. 94). Keynes's comment seems
to b e meant admiringly , bu t one result of Moore's influence is that his hypostatizin g
tendency sometime s sets th e standar d fo r wha t philosopher s hav e in min d by "real -
ism" an d simila r terms in ethicsas witness, e.g. , the characterizatio n o f objectivist
ethics as attributing moral values to "th e furniture o f the world" i n Mackie , Ethics,
p. 16 . My own mov e away from a perceptual model also means giving up a common
analogy to scientifi c realism , but I take i t that th e convergence of the two side s in the
current realism/antirealism debatee.g., the overlap between my view and Gibbard's
actually serves to undermin e antirealism's challenge to ou r naiv e phenomenology (cf.
Brink, Moral Realism and Foundations o f Ethics, pp. 23-24).
64. Se e esp. Harman, Nature o f Morality, pp . 3-10, fo r a n antirealis t argument
resting on this assumption; cf. Nicholas L. Sturgeon, "Moral Explanations," i n Mora-
lity, Reason and Truth, ed. D. Copp and D. Zimmerman (Totowa, N.J.: Rowman and
Allanheld, 1985) , pp . 49-78, for a realist reply .
65. Cf . esp. Walzer, "Political Action"; Williams, Moral Luck; and Nagel, Mortal
66. Cf . Simon Blackburn, "Rule-Following an d Moral Realism, " i n Wittgenstein:
To Follow a Rule, ed. S. H. Holzma n an d C . M. Leac h (London: Routledge & Kegan
Paul, 1981) , p. 175.
67. Se e esp. Hare, "Mora l Conflicts, " pp . 206-19; cf. idem, "Utilitarianis m and
the Vicarious Affects," i n The Philosophy o f Nicholas Rescher, ed. E. Sosa (Dordrecht,
Holland: D . Reidel, 1979) , pp . 141-52 .
68. Williams , "Ethical Consistency," p . 123.
69. Hare , "Mora l Conflicts, " pp . 209-10.
70. Al l quotes this paragraph from ibid., p . 210.
71. Ibid. , p. 209.
72. Cf . McDowell, "Value s and Secondar y Qualities," pp . 118ff .

Chapter 4
1. Thoug h Williams contrasts agent-regret with a number of other contrary-to-duty
reactions (se e Moral Luck, pp. 30-32), as far a s I know his published discussions do
not connect it to guilt. My remarks here are based in part on some conversations with
Williams i n 1989 , particularly in discussio n of preliminar y draft s o f chapter s 2-3 o f
Notes t o pages 110-123 22 7

Shame and Necessity (Berkeley : University of Californi a Press, 1993) ; th e boo k itself
came out to o lat e for consideration i n this argument, however .
2. Cf . Williams, Moral Luck, p. 121 .
3. See , e.g., Marci a Baron , "Remors e an d Agent-Regret, " i n Midwest Studies i n
Philosophy, ed . P. A. French, T. E. Uehling, Jr., and H. K. Wettstein (Notre Dame, Ind.:
University of Notre Dam e Press , 1988) , 13 : 259-60; cf. Herbert Fingarette , "Feelin g
Guilty," American Philosophical Quarterly 1 6 (1979): 159-64 , and Herber t Morris ,
"Reflections o n Feeling Guilty," Philosophical Studies 4 0 (1981) : 187-93 .
4. Se e Aristotle, Nicomachean Ethics, 1128blO-35.
5. Ibid. , 1128b29 .
6. Se e esp. Williams, Ethics and the Limits o f Philosophy; cf. also Slote, From Mora-
lity to Virtue.
7. Aristotle , Nicomachean Ethics, 1131b25-1132b20 .
8. Se e Baron, "Remors e an d Agent-Regret, " p . 270. Baron' s favore d alternativ e
to agent-regret is remorse, although she also acknowledges tha t guilt applie s to many
of her cases, at least in a broad sense; cf., e.g., p. 260. Her argument is apparently meant
to cut against virtue ethics, however, rathe r tha n t o support modifications i n the stan-
dard versio n of the theory, a s suggested here. Note that he r definitio n o f virtue ethics
is an odd one, sinc e it apparently insists on incompatibility with duty ethics and rules
out takin g conscientiousness as a virtue; cf. p. 259 .
9. Bertran d Russell, The Autobiography o f Bertrand Russell: 1872-1914 (Boston:
Little, Brown, 1967) , pp . 329-30.
10. Bu t cf. Stocker, Plural and Conflicting Values, p. 65 , fo r a n interpretatio n of
Aristotle o n whic h bas e acts , includin g even act s performe d unde r duress , rul e ou t
eudaimonia, an d with it presumably virtue. Stocker takes the point as applying to cases
of "dirt y hands"like my "weighted" dilemmas , except tha t th e alternatives are no t
said t o be wrong al l things considered; see , e.g., ibid. , p . 10 .
11. Fo r examples of this approach, alon g with the works by Williams cited earlier ,
see esp. Lawrence A. Blum, Friendship, Altruism and Morality (London : Routledge &
Kegan Paul, 1980), and Michael Stocker, "The Schizophreni a of Modern Ethical Theo-
ries," Journal o f Philosophy 7 3 (1976): 453-66.
12. Withou t our independent information about Russell's somewhat limite d emo-
tional range, I should add, this might be a more charitable explanation of his reaction
in the passage cited. Perhaps I ought to remind the reader, for that matter, that the case
in which I have made out guil t as obligatory i s substantially modified from Russell' s
real-life case. As another alternativ e in the real-life case, o n a view that allows for un-
conscious guil t as a n emotio n th e agen t misidentifie s (cf. Greenspan, Emotions and
Reasons, pp. 25-30), one might instead argue that Russell really did feel guilty on some
level, despite the fact that the passage acknowledges onl y sorrow. The element of irra-
tionality commonly introduced b y unconscious emotion migh t help make sense of the
confusion I have noted i n Russell's account o f the reason s fo r hi s feeling .
13. Se e Greenspan, Emotions and Reasons, p. 4; cf. pp. 83-136, for the develop-
ment and defens e of the general account o f emotional appropriatenes s tha t I begin to
apply to guilt in this section, which is preliminary to my argument in chapter 5, section
2, in this volume.
14. Se e Greenspan, Emotions and Reasons, pp . 31-32; cf. pp. 55ff .
15. Cf . Greenspan, "Conditional Ought s and Hypothetical Imperatives, " pp. 272 -
16. Cf . Greenspan, "Behavior Contro l an d Freedo m o f Action," an d idem , Emo-
tions and Reasons, pp . 153 , 165-66 , 173 .
228 Note s to pages 123-131

17. Cf . ibid., pp. 83-136.

18. Th e basi c account outline d here i s culled from a numbe r o f sources; se e esp.
Rawls, Theory o f Justice, pp . 440-46, 479-85; Herbert Morris , O n Guilt an d Inno-
cence: Essays in Legal Philosophy an d Moral Psychology (Berkeley : University of Cali-
fornia Press , 1976) , pp . 59-63; Gabriel e Taylor, Pride, Shame, an d Guilt: Emotions
of Self-Assessment (Oxford : Clarendon, 1985); and Fingarette, "Feeling Guilty." I shall
postpone unti l chapter 5 a discussion o f standard assumption s abou t th e connectio n
of emotional guil t to th e self-attributio n of responsibility.
19. Cf . Taylor, Pride, Shame, an d Guilt, p. 101; cf. pp. 99-100.
20. Cf. , e.g., Thomas, Living Morally, pp . 76-80.
21. Cf. , e.g., Taylor, Pride, Shame, an d Guilt, pp. 54-57ff .
22. Cf . Gibbard, Wise Choices, Ap t Feelings, pp. 136-40 .
23. Se e Rawls, Theory o f Justice, pp . 484-85; cf. pp. 445-46.
24. Se e esp. Hoffman, "Developmen t o f Prosocia l Motivation, " fo r a revie w of
the recent literature .
25. Se e Edwards, Nature o f True Virtue, pp . 61-74 .
26. Se e esp. Benedict, Chrysanthemum an d the Sword; cf., e.g., Millie R. Creighton ,
"Revisiting Sham e and Guil t Cultures : A Forty-Year Pilgrimage," Ethos 1 8 (1990) :
27. Se e esp. Hoffman, "Developmen t o f Prosocia l Motivation, " pp . 296-97.
Hoffman doe s no t mak e a clea r shame/guil t distinction; but cf . June Pric e Tangney ,
"Moral Affect : Th e Good , th e Bad, an d th e Ugly, " Journal o f Personality an d Social
Psychology 6 1 (1991) : 598-607 , for results supporting hi s central claim s in applica -
tion specificall y t o guilt .
28. Se e The Oxford English Dictionary. Othe r languages apparently have no noun
for emotiona l guil t per s e but instea d mus t refe r t o th e sens e of guilt or t o feeling s of
guiltwhich presumably can include feelings of moral shameto distinguish the emo-
tion fro m th e stat e o f being guilty or a t fault . Alternatively , they can us e an adjectiv e
for "guilty " wit h "feels," t o mea n "feel s a s if' a t fault .
29. Edwards , Nature o f True Virtue, pp . 66-67. Edwards's notio n o f consistency
is apparently a "sentimentalist" modificatio n of the then-current "intellectualist" views
of Samue l Clarke ; se e Norman Fiering , Jonathan Edwards's Moral Thought an d It s
British Context (Chape l Hill: Universit y of North Carolina , 1981) , pp. 87-93.
30. Edwards , Nature o f True Virtue, p . 64 .
31. Ibid. , cf. pp. 61ff.
32. Se e Taylor, Pride, Shame, an d Guilt, p. 92; cf. pp. 90, 106.
33. Cf . Greenspan, Emotions and Reasons, pp . 48-55.
34. See Freud, Civilization and Its Discontents, pp. 85, 94ff. Freud apparently uses
"remorse" fo r guilt fel t fo r a n actua l deed. I n any case, hi s developed accoun t of th e
sense of guilt that he takes to precede action seems to me to allow for the identificatory
interpretation I give here. In fact, Freud suggests something of the sort in his character-
ization of the supereg o i n terms of aggression directe d toward th e ego it develops as a
part of; cf., e.g., p. 84 . See also p. 95 for the characterization of the sources o f guilt in
terms of love/hate ambivalence; cf. Klein and Riviere , Love, Hate an d Reparation.
35. Se e esp. Arnold H. Modell , "O n Havin g the Right to a Life: An Aspect of the
Superego's Development, " International Journal of Psychoanalysis 46 (1965): 323-31;
cf. Takeo Doi , The Anatomy of Dependence (Tokyo : Kodansha International, 1973) ,
pp. 48-57, fo r a Japanese psychiatrist's account o f a simila r notion o f guilt i n Japan
(at th e deat h o f parents and th e like ) a s a reactio n t o th e sens e that on e i s betraying
one's group ties. I shall attempt a fulle r discussio n of such cases in chapter 5 .
Notes t o pages 131-139 22 9

36. Cf. , e.g., Douglas, Purity an d Danger. Wha t is commonly referred to as "Jew-
ish guilt" seems to me to b e in large part a variant of separation guil t with th e features
I focus on here rather tha n anything specifically religiou s as one might suppose . (Not e
that th e notion of original sin is distinctively Christian, indeed Western Christian , du e
to Sain t Augustine) Another elemen t ma y b e something like Nietzschean "debt" ; cf.
Nietszche, Genealogy o f Morals, pp . 62-63ff . Th e one element I can locat e tha t doe s
seem to hav e biblical sources i s the tendency to explai n ba d event s a s results o f one' s
own (or in the biblical case, the group's) misdeeds . According to my argument i n chap-
ter 5 , eve n raising th e questio n o f (objective ) guilt"Wha t hav e I done t o deserv e
this?"can b e enough t o generat e guilt feelings .
37. Cf. , e.g., Ezekiel 6:9 and 36:31 for expressions of self-loathing that see m to be
plausibly classifie d a s guilt; I owe thes e references to biblica l scholar Davi d Halperi n
(personal communication).
38. Cf . Sharon Bishop , "Connection s an d Guilt, " Hypatia 2 (1987) : 7-23 .
39. Cf . Gibbard, Wise Choices, Apt Feelings, pp. 139-40 . Gibbard's ow n accoun t
of guilt as a counterpart to ange r i s naturally read i n the rather wea k term s I go on t o
discuss, bu t it also allows for fillin g out i n the way I propose here.
40. See , e.g., June Price Tangney, Patricia Wagner, and Richard Gramzow, "Prone -
ness to Shame , Proneness t o Guilt , and Psychopathology, " Journal o f Abnormal Psy-
chology 10 1 (1992): 469-78 .
41. Se e Michael Friedman, "Toward a Reconceptualization of Guilt," Contempo-
rary Psychoanalysis 2 1 (1985) : 540; cf. pp. 535, 539.
42. Cf. , e.g., the account of the Ancient Greek shift from "shame-culture " t o "guilt -
culture" i n E. R. Dodds , Th e Greeks an d th e Irrational (Berkeley : University of Cali-
fornia Press , 1951) , pp. 17f., 28-63.
43. Hare , "Mora l Conflicts, " p . 208.
44. See , e.g., L. A . Kosman , "Bein g Properly Affected : Virtue s an d Feeling s i n
Aristotle's Ethics," i n Essays o n Aristotle's Ethics, ed. A. O . Rorty (Berkeley : Univer-
sity of California, 1980), pp . 103-16 . For other treatments of the issu e of responsibil-
ity for emotion (bu t mainly in a backward-looking sense of "responsibility"), se e Rob-
ert Merrihew Adams , "Involuntar y Sins," Philosophical Review 94 (1985): 3-31 , an d
Edward Sankowski , "Responsibility of Persons fo r Their Emotions, " Canadian Jour-
nal of Philosophy 7 (1977): 829-40 .
45. Se e Ross, Right an d the Good, pp . 4-6, and H. A. Prichard, "Doe s Moral Phi -
losophy Res t o n a Mistake?, " Mind 2 1 (1912) : 33 . Fo r a mor e recen t exampl e o f th e
standard vie w of Kant on emotions , se e esp. Blum , Friendships, Altruism and Morality;
cf. Nancy Sherman, "The Place of Emotions in Kantian Morality," i n Identity, Charac-
ter, and Morality, ed. O. Flanagan and A. O. Rorty (Cambridge, Mass.: MIT Press, 1990) ,
pp. 149-70 , for anothe r sid e o f this story. However, I take i t that Kant's exclusio n of
emotions from the grounds of moral judgment as required by my own argumen t remains
intact, even if he makes room fo r the m in other ways as components o f the moral life .
46. Cf . Prichard, "Moral Philosophy," p. 27, and Ross, Right and the Good, p. 5.
47. Prichard , "Mora l Philosophy, " p . 24.
48. Ross , Right an d th e Good, p. 5.
49. Ibid. , pp. 42-43.
50. Not e tha t Ros s late r changed hi s mind on this point, holdin g instead tha t th e
duty to fulfil l a promise was a duty not to effec t a certain result but to try to d o so; see
W. David Ross, Foundations o f Ethics (Oxford : Clarendon, 1939), p . 108 . The change
would reinforce my argument in what follows for treating "ought-to-feel" on the model
of "ought-to-do. "
230 Notes t o pages 141-160

51. Prichard , "Mora l Philosophy," p . 33.

52. Cf . Marcus, "Mora l Dilemmas and Consistency, " p . 198.
53. Cf . Aeschylus, "Agamemnon," 217-18 ; cf. Nussbaum, Fragility o f Goodness,
pp. 35-37 .
54. Williams , "Ethical Consistency, " p . 135.
55. Williams , Moral Luck, p. 179.
56. Se e esp. Kant, Foundations o f th e Metaphysics o f Morals, p . 10 ; cf . Nagel ,
Mortal Questions, p. 24. Wha t I suggest her e is not unlik e Nagel's ow n diagnosi s o f
the problem of moral luck, which he explains by appeal to his distinction between inter-
nal and externa l points of view; cf. pp. 36ff .
57. Se e Peter Strawson, "Freedom an d Resentment, " i n Free Will, ed. G. Watso n
(Oxford: Oxfor d Universit y Press , 1982) , pp . 59-80; cf. Susan Wolf, "Asymmetrica l
Freedom, "Journal o f Philosophy 7 7 (1980) : 151-66. The issu e is also complicated by
features of the rational and moral assessment of these emotions noted elsewhere in this
chapterfor instance , in my discussion o f asymmetries involving pride.
58. Se e Hume, Treatise o f Human Nature, pp. 251-55; cf. pp. 409-10. The fic-
tion I have in min d as bearin g on responsibilit y issues is a bi t differen t fro m th e on e
Hume discusses in his treatment of personal identity (cf . esp. p. 255). I t essentially in-
volves making out the self as distinct from an d prior to its actions, in order to cast i t in
the rol e of responsible cause. Although a ful l discussio n o f these issues i s beyond th e
scope o f this discussion, I shall hav e a bi t mor e t o sa y about the m i n chapter 5 . Fo r
related comments on the notion of causation b y character, see Greenspan, "Unfreedom
and Responsibility" ; cf. Adams, "Involuntary Sins. "
59. Cf . Greenspan, "Unfreedom and Responsibility," pp . 69-70. Nagel discusse s
a similar case, of an accident resulting from failur e to have one's brakes checked, wit h
reference t o moral luck in Mortal Questions, p. 29 .

Chapter 5

1. Se e Rawls, Theory o f Justice, p . 482. For general objection s to judgmentalism,

see Greenspan, Emotions and Reasons, pp. 15-80.
2. Se e Aristotle, Rhetorica, Boo k II, Sec. 2,1378a31-1378b5; cf. Greenspan, Emo-
tions an d Reasons, pp . 48-55.
3. Rawls , Theory o f Justice, p . 482.
4. Se e Taylor, Pride, Shame, an d Guilt, p. 86.
5. Ibid. , p. 1.
6. Ibid. , p. 91.
7. Se e Greenspan, Emotions an d Reasons, pp. 18-20; cf. idem, "Emotions a s Evalu-
ations," Pacific Philosophical Quarterly 6 2 (1981) : 158-69.
8. Se e Greenspan, Emotions and Reasons, pp. 25-30. But note, i n relation to my
present argument, the reasons offered agains t any easy inference to unconscious evalu-
ations o n pp. 19 , 24.
9. Cf . ibid., pp . 153-62.
10. Se e Stephen P. Stich, "Beliefs and Subdoxasti c States, " Philosophy o f Science
45 (1978) : 499-518, and Michae l E. Bratman, "Practical Reasonin g an d Acceptanc e
in a Context," Mind 10 1 (1992): 1-15 .
11. Fo r the latter notion, see, e.g., Donagan, Theory o f Morality, p . 45 .
12. Se e Herbert Morris, "Nonmora l Guilt, " i n Responsibility, Character, and th e
Emotions, ed. F. Schoeman (Cambridge: Cambridge University Press, 1987), pp. 220-41.
Notes t o pages 161-172 23 1

13. Ibid. , pp. 232-37.

14. Taylo r ha s a puzzling comment abou t th e parallel sort of economic case : "I f I
feel guilty about m y privileged position i n society due to circumstances of birth then I
see myself as an agen t causall y involved: it i s my birt h whic h ha s brough t abou t th e
state o f affairs whic h is my privileged position" (Pride, Shame, an d Guilt, p. 91). But
the agent's birth is no less passive than bein g benefited economically or the like; it just
is less easily distinguished from th e perso n t o who m i t happens . S o it is questionable
how this could yield the requisite sort of causal responsibility on Taylor's account with-
out a n implausibl e attribution o f delusion t o th e agent . Presumably , what i s in ques-
tion i s causation b y the agent in some sensei f no t throug h he r act or agenc y then a t
least through some other form of activity on her part. But surely birth does not qualify .
15. Se e Hume, Treatise of Human Nature, p. 139.
16. Morris , "Nonmora l Guilt, " pp . 226-32.
17. Taylor , Pride, Shame, an d Guilt, pp. 91-92 .
18. Morris , "Nonmora l Guilt, " pp . 237-40.
19. Cf . Joel Feinberg , Doing an d Deserving: Essays in the Theory o f Responsibil-
ity (Princeton , N.J.: Princeton University Press, 1970) , p. 233. Taylor apparentl y misses
this distinction when sh e cites Feinberg's poin t tha t vicarious objective guilt is impos -
sible, in support of her own view that "feeling s of guilt . . . cannot aris e from th e deeds
or omissions o f others" (Pride, Shame, an d Guilt, p. 91; cf. n. 5 and Feinberg , Doing
and Deserving, pp . 231, 237).
20. Morris , "Nonmora l Guilt, " pp . 239-40.
21. Ibid. , pp. 232, 237.
22. Ibid. , p . 226.
23. Ibid. , p. 232.
24. See , e.g., Herbert Morris, "Th e Decline of Guilt," Ethics 9 9 (1988) : 66.
25. Morris , "Nonmora l Guilt, " 226.
26. Se e Greenspan, Emotions and Reasons, p. 31; cf. pp. 153-62 .
27. Cf . ibid., esp. pp. 83-107, for my extended inquir y into appropriateness , re -
sulting in a nonquantitative criterion i n terms o f belie f warrant .
28. Se e John Searle , Intentionality (Cambridge : Cambridge University Press, 1983) ,
pp. 31-35; cf., e.g., p. 8 for his statement of the point a s a claim that emotions have no
direction o f fitmeaning, I take it , emotion s themselve s a s distinct fro m desire s an d
beliefs. Searle sees emotions as combining elements of desire and belief, with an emphasis
on desir e (but as only partially characterized b y his analysis; cf. p. 36), while making
out appropriatenes s i n term s o f belie f satisfaction (pp . 8-9) . Note, however , tha t h e
uses the category of desire to capture evaluative content.
29. Cf . Gibbard, Wise Choices, Ap t Feelings, p. 139. In th e ensuin g argument I
hope t o sho w ho w the standards for guilt and ange r might divergecompatibly with
the valu e placed accordin g t o Gibbar d o n meshin g guilt an d ange r (cf . p. 299). See
Sinnott-Armstrong, Moral Dilemmas, p. 107, for the application to dilemmas of a simi-
lar claim of asymmetry between remorse and blame , though without the explanation I
provide here; cf. also Greenspan , "Mora l Dilemmas and Guilt. "
30. Cf . Greenspan, Emotions and Reasons, pp. 48-55, 188 , n. 3.
31. Cf . my distinction between backward- and forward-lookin g notions o f respon-
sibility i n Greenspan , "Unfreedo m an d Responsibility, " pp . 64 , 71 . On e migh t be
tempted to take personal blam e as appropriate jus t in virtue of its usefulness i n getting
individual member s to exer t forward-lookin g responsibility and refor m the relevant
group. However, general adaptiveness would seem to b e well enough served by blame
imposed for failing to take responsibilityi n that way and als o in the emotional terms
232 Notes t o pages 174-189

that I go on t o justify . Somethin g simila r applies, moreover, t o any independent sym-

bolic function that might be assigned to blame, in expressing a strong stand against the
group behavio r in question .
32. O n th e othe r hand , th e standard s o f appropriatenes s ma y incorporat e fact s
about norma l emotional developmen t to the extent tha t thes e affect bot h th e natural-
ness of an object of attention an d th e general adaptiveness of an emotional response t o
it. This point ma y afford a deeper explanation of the notio n o f "paradigm scenarios "
in D e Sousa, Rationality o f Emotion, pp. 181-84 , as early stages i n emotional learn-
ing that ar e give n specia l stres s i n determinin g the rationalit y of a n emotio n i n adul t
life. In view of facts about emotional "inertia, " o r the tendency of emotional responses
to outliv e changes in their corresponding judgments , we would expec t som e stages in
childhood developmen t t o affec t adul t respons e patterns . Fo r instance , it migh t be
thought that anger in response t o injurie s t o oneself , a s the natural childhood stru t of
the sense of justice, is important enoug h i n developmental terms to warran t blam e on
the par t o f the victi m in a case like Russell's, even in the absenc e of adequate grounds
for third-perso n moral blamea possibility I tried to allo w for in chapter 4. But there
is als o roo m fo r a goo d dea l o f empirica l disagreemen t o n thes e issues; i t shoul d b e
evident tha t m y treatment of cases here rests o n a heavy element of conjecture.
33. Cf . Morris's somewhat les s discriminate acceptance o f "share d guilt " i n O n
Guilt and Innocence, pp. 111-138.
34. Cf . Sinnott-Armstrong, Moral Dilemmas, p. 42; unlik e Sinnott-Armstrong , I
take it that " I had no choice," eve n if meant as an appeal to moral necessity , counts as
an excuse .
35. Cf . Fletcher, Rethinking Criminal Law, pp . 343-44.
36. Marcus , "Mora l Dilemmas and Consistency," pp . 193-94, 198-99.
37. Thi s point distinguishe s dilemmas from th e case s o f responsibilit y for inevi-
table states of affairs cite d i n John Marti n Fischer and Mar k Ravizza , "Responsibility
and Inevitability," Ethics 10 1 (1991): 258-64. Note that the agent in a dilemma would
have done otherwis e (avoide d doing wrong) if he had bee n able to; cf. the contrastin g
cases cited on pp. 264-65.
38. See Williams, Moral Luck, p. 54; cf. p. 41. Cf. also Stocker , Plural and Con-
flicting Values, pp . 9-19 .
39. Williams , Moral Luck, pp. 63, 61.
40. Ibid. , p. 79.
41. Ibid. , p. 63.

Chapter 6
1. See , e.g. , Richar d Brandt , "Th e Scienc e of Ma n an d Wid e Reflectiv e Equilib-
rium," Ethics 10 0 (1990) : 263 . Th e genera l pattern o f definitio n derive s fro m Mill ,
Utilitarianism, p. 60cf. also Gibbard, Wise Choices, Apt Feelings, p. 41though Mil l
does no t singl e out emotiona l blam e but ha s in mind also overt act s o f social censur e
and punishment.
2. Se e esp. Hume, Treatise o f Human Nature, p. 469 .
3. Se e esp. Simon Blackburn , Spreading th e Word (Oxford : Clarendon , 1984) ,
pp. 167-71 , 181-223 .
4. Se e esp. Nagel, View from Nowhere, pp. 171-75 .
5. Se e esp. De Sousa, Rationality o f Emotion; cf. also, e.g., Morto n White , What
Is an d What Ought t o B e Done (Ne w York : Oxford Universit y Press , 1981) , which
Notes t o pages 189-197 23 3

assigns emotions a role in moral knowledg e analogou s t o tha t o f sense-perception i n

scientific knowledge. McDowell's wor k als o seems to assign emotions a role I think of
as perceptualin the broad sense I have in mind here (meaning, roughly, "representa -
tional"), not on a specifically sensor y interpretation, a s might seem to fit the "second -
ary qualities" viewdespite a very differen t accoun t o f moral knowledge ; cf. esp. m y
discussion of McDowell's Virtue an d Reason in chapter 3, section 2, in this volume .
6. Se e Simon Blackburn, "How to Be an Ethical Anti-Realist," in Midwest Studies
in Philosophy, ed . P. A. French, T. E. Uehling, Jr., and H . K . Wettstein (Notre Dame ,
Ind.: University of Notre Dam e Press, 1988) , 12: 365; cf. p. 374, n. 5, for furthe r ref -
7. Ibid. , p. 365; cf. Simon Blackburn, "Errors an d the Phenomenology of Value, "
in Morality an d Objectivity: A Tribute t o J. L. Mackie, ed . T. Honderic h (London :
Routledge & Kega n Paul, 1985), pp. 1-22 , fo r a careful accoun t of the ways in which
moral judgment fails t o fi t a perceptual analogy in detail (as on McDowell' s "second -
ary qualities" view).
8. Blackburn , Spreading th e Word, p . 195 ; cf. pp. 186 , 207.
9. Cf . Blackbur n "Flight t o Reality. " Fo r a n explicitl y socia l versio n o f anti -
realism, cf. Gibbard, Wise Choices, Ap t Feelings. Fo r simplicity' s sake, I shall limit
my remark s her e t o Blackburn , bu t I tak e the m t o appl y a t leas t roughl y t o an y
expressivist view.
10. Se e Blackburn, Spreading the Word, pp . 217-19.
11. Ibid. , pp. 167-71 .
12. Simo n Blackburn, "Attitudes and Contents, " Ethics 9 8 (1988): 504.
13. Se e Blackburn, Spreading the Word, p . 219, n. 21 .
14. Ibid. , p. 197.
15. Ibid. , p. 21.
16. Ibid. , pp. 193-95 .
17. Se e Simon Blackburn , "Truth , Realism, an d th e Regulatio n o f Theory," i n
Midwest Studies in Philosophy, ed . P. A. French, T. E. Uehling, Jr., and H. K . Wettstein
(Notre Dame , Ind. : Universit y o f Notre Dame , 1980) , 5: 357-58, and idem , "Rule -
Following and Moral Realism, " p . 175.
18. Blackbur n at som e point s seem s t o b e ope n t o a genera l interpretatio n o f
internalism, however; see his Spreading the Word, p. 189. But cf., e.g., idem, "How to
Be an Ethical Anti-Realist," p. 363, for the mor e standard view .
19. Blackburn , Spreading th e Word, pp . 198-201 .
20. I n a recent pape r o n dilemmas , Blackbur n indeed rest s th e specia l nature of
moral dilemmasbut set up as positive ought-conflicts and interpreted in terms of quan-
daries about what to doon nothing beyon d a quantitative difference i n residues; see
his "What I s Puzzling about Dilemmas?" in Understanding Moral Dilemmas, ed. H. E .
Mason (Ne w York: Oxfor d Universit y Press, 1995) . I saw a n earlie r version o f th e
paper, as delivered at a conference on dilemmas at the University of Minnesota in April
21. Cf . Blackburn, "Flight to Reality. " O n the social base s of emotion in connec-
tion with my own suggestions on this issue, cf. the "socia l constructionist " view s cur-
rent in anthropology and related subjects , as represented, e.g. , in Rom Harre ed., The
Social Construction o f Emotions (Oxford: Basi l Blackwell, 1986). However, o n som e
versions the approach apparentl y rules out an y innate emotions, a s I do no t mea n t o
do; m y assumption is rather that innat e responses ar e shape d int o ne w emotions via
social learning.
22. Se e esp. Stocker, "Schizophreni a o f Modern Ethica l Theories, " pp . 453-66.
234 Note s to pages 198-210

23. Cf . Kosman, "Being Properly Affected. "

24. Se e esp. Robert M . Adams , "Motiv e Utilitarianism, " Journal o f Philosophy
73 (1976) : 467-81 .
25. Cf . Michael Moore , "Th e Moral Wort h o f Retribution, " i n Responsibility,
Character, and the Emotions, ed. F. Schoeman (Cambridge: Cambridge University Press,
1987), pp. 199-202ff .
26. Se e G. E. Moore, Principia Ethica (Cambridge : Cambridge Universit y Press,
1903), pp. 6-15 .
27. Cf . Greenspan, Emotions an d Reasons, pp. 155ff. , and idem, "Fooling th e Moti-
vational Meter : A Reply t o Roberts " (unpublished) . I t migh t b e usefu l t o thin k o f
emotional affec t as also providing a second-order reason for action, in the sense of Raz,
Practical Reason, pp. 39-40albei t a positive or reinforcing reason o f a sort Raz by-
passesto th e exten t tha t i t provide s a reason fo r attentio n to a n evaluativ e reason
(the evaluative content o f emotion). I read thi s boo k to o lat e to influenc e m y discus-
sion here; however, it might also provide some helpful terminolog y for the ideas abou t
oughts an d reasons expressed i n chapters 2-3 .
28. See , e.g., Josiah Royce, The Philosophy o f Josiah Royce (Indianapolis: Hackett,
1982), pp. 189-96 ; cf. George Herber t Mead , O n Social Psychology: Selected Papers
(Chicago: Universit y of Chicago Press , 1956) , pp. 39-42, 199-246 . The brie f reflec-
tions tha t follo w are my own.
29. Cf . Royce's idea l communitarianism (e.g., Philosophy o f Royce, pp. 367-87).
For that matter, there are elements of a communitarian view in the Kantian ideal of the
realm of ends; cf. Kant, Foundations o f th e Metaphysics o f Morals, pp . 51ff.
30. Cf . Rawls's discussio n o f "congruence " i n Theory o f Justice, pp . 567-77 . A
more ordinar y argument fro m pride than the one I suggest here, bu t limited to agent s
with a certain kind of motivational structure, might point out that whatever one cared
to achieve in life on a n individual level would be cheapened or compromised b y being
achieved at the cost of violating moral requirements. Its achievement would b e compa-
rable to winning a fight that was "fixed"a s th e agent would be aware, eve n if no one
else was.
Note that m y argument is not mean t to show tha t moralit y is the overriding end,
but jus t that i t i s among th e ends , o f a rational agen t wit h th e usua l sor t o f motiva -
tional structure . It may have to compete with individual achievement and other forms
of value (some arguably moral) whose focu s i s more directl y on th e individual.
31. Cf . esp. Wong, Moral Relativity, an d Geoffre y Sayre-McCord , "Bein g a Real-
ist about Relativis m (in Ethics)," Philosophical Studies 6 1 (1991) : 155-76 , for recen t
defenses o f the possibility o f combining relativism and realism .
32. Se e Williams, Ethics an d th e Limits o f Philosophy, pp . 160-66 .
33. Se e Moore, Principia Ethica, Ch. 6.
34. Cf . A. J. Ayer, Language, Truth an d Logic, London: V . Gollancz, 194 6 (rev .
ed., Dover , n.d.) , pp. 102-13.

Adams, Robert Merrihew. "Involuntary Sins." Philosophical Review 9 4 (1985): 3-31.

. "Motiv e Utilitarianism." Journal o f Philosophy 7 3 (1976): 467-81.
Aeschylus. "Agamemnon. " I n Th e Oresteian Trilogy, trans. P. Vellacott, 41-100.
Harmondsworth, Middlesex , England: Penguin, 1959.
Anderson, Alan Ross. "Th e Forma l Analysis of Normative Systems." In The Logic o f
Decision and Action, ed. N. Rescher , 147-213. Pittsburgh: Univers