The principal aim of this book is to develop and defend an analysis of theconcept of moral obligation. The analysis is neutral regarding competingsubstantive theories of obligation, whether consequentialist or deontolog-ical in character. It seeks to generate new solutions to a range of philo-sophical problems concerning the concept of obligation and its application.Among the topics treated are deontic paradoxes, the supersession ofobligation, conditional obligation, prima facie obligation, actualism andpossibilism, dilemmas, supererogation, and cooperation. By virtue of itsnormative neutrality, the analysis provides a theoretical framework withinwhich competing substantive theories of obligation can be developed andassessed.

CAMBRIDGE STUDIES IN PHILOSOPHY

The Concept of Moral Obligation

CAMBRIDGE STUDIES IN PHILOSOPHY

General editor Ernest Sosa

Advisory editors

Jonathan Dancy University of KeeleGilbert Harman Princeton University

Frank Jackson Australian National UniversityWilliam G. Lycan University of North Carolina, Chapel Hill

Sidney Shoemaker Cornell UniversityJudith J. Thomson Massachusetts Institute of Technology

RECENT TITLES

Andrew Newman The Physical Basis of PredictionDavid Owens Causes and CoincidencesJaegwon Kim Supervenience and Mind

Warren Quinn Morality and ActionMichael Jubien Ontology, Modality, and the Fallacy of Reference

John W. Carroll Laws of NatureHenry S. Richardson Practical Reasoning about Final Ends

Joshua Hoffman and Gary S. Rosencrantz Substance among Other CategoriesM. J. Cresswell Language in the World

Noah Lemos Intrinsic ValueLynne Rudder Baker Explaining Attitudes

Robert A. Wilson Cartesian Psychology and Physical MindsMichael Devitt Coming to Our Senses

Barry Maund Colours

The Concept ofMoral Obligation

Michael]. ZimmermanUniversity of North Carolina, Greensboro

CAMBRIDGEUNIVERSITY PRESS

CAMBRIDGE UNIVERSITY PRESSCambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo

Cambridge University PressThe Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.orgInformation on this title: www.cambridge.org/9780521497060

© Cambridge University Press 1996

This publication is in copyright. Subject to statutory exceptionand to the provisions of relevant collective licensing agreements,no reproduction of any part may take place without the written

permission of Cambridge University Press.

First published 1996This digitally printed version 2007

A catalogue record for this publication is available from the British Library

Library of Congress Cataloguing in Publication data

Zimmerman, Michael J., 1951—

The concept of moral obligation / Michael J. Zimmerman,

p. cm. - (Cambridge studies in philosophy)

Includes bibliographical references and indexes.

ISBN0-521-49706-X(hc)

1. Duty. 2. Responsibility. I. Title. II. Series.

BJ1451.Z56 1996 95-19155

170-dc20 CIP

ISBN 978-0-521-49706-0 hardbackISBN 978-0-521-03874-4 paperback

Contents

Preface page ixAcknowledgments xiii

1 Groundwork: Some distinctions 11.1 Moral and nonmoral obligation 11.2 Binding and nonbinding senses of "ought" 21.3 Overall and prima facie obligation 51.4 Objective and subjective obligation 101.5 Act-evaluation and other types of evaluation 20

2 Moral obligation: An analysis 212.1 Value maximization 21

2.1.1 Two problems 212.1.2 The analysis 252.1.3 Extension of the analysis 312.1.4 Perfect and imperfect duties 38

2.2 Underlying metaphysical and conceptual issues 402.2.1 "Can" and control 402.2.2 "Can" and accessibility 452.2.3 "Can" and knowledge 492.2.4 "Can" and control again 502.2.5 The ontological status of actions 532.2.6 Indefiniteness 57

2.3 Implications of the analysis 632.3.1 Certain deontic implications 632.3.2 Certain deontic paradoxes 722.3.3 Analysis and neutrality 75

3 The dynamics of obligation 793.1 "Ought" and "can" 79

3.1.1 An internal criticism 793.1.2 "Can" and "can avoid" 81

3.1.3 An alleged redundancy3.1.4 Past and present obligations3.1.5 Alternate possibilities3.1.6 Psychological incapacitation3.1.7 The obligatory and the ideal3.1.8 Excuses3.1.9 Self-imposed impossibility

3.2 Immediate and remote obligation3.2.1 Self-imposed impossibility3.2.2 The time at which a wrong is done3.2.3 The time at which an obligation is satisfied3.2.4 The cancellation of obligation3.2.5 The supersession of obligation

Conditional obligation4.1 Varieties of conditions on obligation4.2 The analysis4.3 Certain deontic paradoxes4.4 Subsidiary obligation

4.4.1 Factual detachment4.4.2 Levels of obligation4.4.3 Overriding4.4.4 A rival account4.4.5 Level of wrongdoing and seriousness of

wrongdoing4.4.6 Level of wrongdoing and level of obligation

Prima facie obligation5.1 A tale of two obligations5.2 Defective analyses5.3 The analysis

5.3.1 "All else being equal"5.3.2 Three objections5.3.3 Detachment5.3.4 Overriding

5.4 Ross on prima facie obligation5.4.1 Conditional and unconditional prima

facie obligation5.4.2 The self-evidence of prima facie obligations5.4.3 Prima facie obligation and overall obligation5.4.4 Residual obligation

5.5 Rights5.5.1 The correlativity of rights and obligations

828285899093959697

100105107111

114114117122128128131135136

138139

141141143145145154160162163

163164169174176176

VI

5.5.2 Absolute and prima facie rights 1815.5.3 Rights and overall obligation 187

6 Actualism and possibilism 1896.1 The dispute 1896.2 Objections to actualism 191

6.2.1 Certain theses denied 1916.2.2 Sanctioning wrongdoing 1936.2.3 A nondiachronic case 1956.2.4 A slightly amended version of actualism 195

6.3 Objections to possibilism 1966.3.1 Arbitrariness 1976.3.2 Advice 1986.3.3 Detachment 2006.3.4 Act-individuation 203

6.4 What it is best to do 203

7 Dilemmas 2077.1 The nature of moral dilemmas 2077.2 The case against basic moral dilemmas 2117.3 The case for basic moral dilemmas 2177.4 Nonbasic moral dilemmas 225

8 Supererogation 2328.1 The nature of supererogation 232

8.1.1 Going beyond prima facie obligation 2328.1.2 Going beyond perfect obligation 2338.1.3 Going beyond overall obligation 234

8.2 The possibility of supererogation 2378.2.1 Supererogation and discretion 2398.2.2 The division of values 2448.2.3 Continuity 248

9 Cooperation 2549.1 The problem 2549.2 Nonsolutions 258

9.2.1 Unrequited cooperation 2589.2.2 Procedural cooperation 2599.2.3 Group obligation 2609.2.4 Counterfactual cooperation 2639.2.5 Mere openness 265

vii

9.3 The solution: Unintrusive transigence 2689.4 Interpersonal moral dilemmas 274

Postscript 277Appendix: List of propositions 278List of works cited 282Index of names 293Index of subjects 297

vin

Preface

On October 18,1993, the News and Record of Greensboro, North Carolina,carried the following item:

Dear Ann Landers: I am 37 years old and have never married. I dated a 35-year-old divorced man for five years. "Jack" and I were very much in love. I worehis engagement ring for a year, and we talked of marriage, but we bickered alot. I assumed this was just our style of communication.

We decided to get premarital counseling about our never-ending arguments.One counselor said our values were too different. Another counselor said wewere made for each other. I postponed the wedding because Jack wouldn'tcontinue with the counseling. Also, he wanted to elope, and I wanted a churchwedding.

Last January, I went over to Jack's house and caught him with his secretary.It turned out this 21-year-old gorgeous thing had moved in with him. Jackinsists that he doesn't love her, but she cooks and cleans for him, which I neverdid.

Ann, we've been seeing each other behind the secretary's back. We agreethat we love each other, but he's afraid he hasn't the will-power or self-disci-pline to remain true to me. I still want to marry him.

I went to a doctor who told me I was depressed. I've tried dating other men,but I have no desire for them. What should I do?

Indianapolis Dilemma

I record this for two reasons. First, its comic value. Perhaps, as you werereading it, a slight smile played over your lips. I hope so; the rest of this bookis dead serious and will afford very little opportunity for amusement.

The second reason is this. A preface should let the reader know what toexpect from a book, and I want you to know that you should not expect ananswer to Indianapolis's question, even if the "should" in her "What shouldI do?" is a moral one. Nor should you expect answers to any questionsremotely like hers. Some people seem to think that the task of moral phi-losophy is simply to provide answers to such questions; if you are among

ix

these, then this book will sorely disappoint you and I advise you to read nofurther. Once, many years ago, I wrote a paper on the relation betweenintrinsic value and the appropriateness of pleasure, and I submitted itto a leading philosophy journal. One referee recommended rejection,saying:

Actual cases of moral questions and issues are hardly mentioned; yet what elseis morality about except Joan wondering whether to leave her husband or havean abortion, Albert wondering whether to tell his grandmother what her doc-tor just told him, or whether a new bike would spoil his son? That Albert andJoan would find this paper totally irrelevant to their dilemmas, even if they werephilosophers and understood it, seems to me to sit very differently than the ideaof the ordinary man who knows his coffee-cup is real not being interested inThe Theory of Knowledge or Science, Perception, and Reality... If I were advisingJoan, I would tell her to go and read The French Lieutenant's Woman, and notthis paper, and I consider that a wholly legitimate criticism of a piece of philo-sophical ethics.

I hope you are as appalled by this as I still am. (If not, then, again, pleaseread no further.) I have found that this referee's attitude is by no means anisolated one, but I find it quite baffling. Joan's situation, whatever it is, maybe a difficult one, and certainly moral philosophers may legitimately trytheir hands at resolving it; but it had never been my intention in the papereven to pretend to be concerned with the resolution of such matters, andto claim that such resolution is the only legitimate task of moral philosophyis utterly preposterous. At any rate, this book attempts something quite dif-ferent and much less ambitious.

A great deal of contemporary moral philosophy is concerned with find-ing answers to such general substantive questions as what it is that makesright acts right, what the various virtues and their interrelations are, whatconstitutes a proper excuse for vile behavior, and so on. These are all veryimportant questions, and moral philosophers (some of them, anyway) areespecially well equipped to handle them. But, again, it is not with suchambitious questions as these that this book has to do.

No, the focus of this inquiry is a concept: the concept of moral obliga-tion. The main question that is addressed is not what it is that Joan (orAlbert, or Indianapolis) ought to do in particular, nor what it is that makesan act obligatory in general; it is, rather, what it is for an act to be obligato-ry. The task that I have set myself is thus one of conceptual cartography, ofmapping the contours of the concept of moral obligation. The goal is sim-ply that of understanding this concept better. This may be a modest goal,but I have found its achievement difficult.

What follows, then, is an unabashed exercise in metaethics. Althoughyou won't find in it an answer to Indianapolis's predicament, you will findanswers to these questions (and many more besides): What is the relationbetween "ought" and "good"? If someone ought to do each of two things,ought he to do both? If he ought to do both, ought he to do each? If some-one ought to do something but cannot do it without doing something else,ought he to do the something else? What are imperfect duties? Does"ought" apply only to actions, or does it have broader scope? Ought oneto be perfect? Does "ought" imply "can"? If so, can someone divest him-self of an obligation simply by rendering himself unable to fulfill it? Cansomeone succeed in sloughing off an obligation through sheer laziness? Cansomeone be obligated to do something he cannot avoid doing? Can some-one have an excuse for doing something that it was not wrong for him todo? Can something that it was obligatory to do become wrong? Can some-thing that it was wrong to do become obligatory? Are there gradations ofobligation and wrongdoing? Can there be obligations without rights? Cansomeone be obligated to do, or to cause himself to do, wrong? If one per-son ought to advise another to do something, ought the latter to do thatthing? Can someone (Indianapolis, say) be in a genuine moral dilemma?What is the relation between wrongdoing and guilt? Can one go beyondthe call of duty? Can it be right to prevent someone from doing what'sright? Can two wrongs make a right? If questions like these interest you,please read on.

(Interlude on political correctness. In the last paragraph, I have used"he" in what is intended as a gender-neutral manner. I am quite aware —who these days could not be? - that some find this objectionable. In reply,let me simply quote Judith Thomson, who has recently stated, with cus-tomary eloquence, exactly what I would want to say:

On the other hand, "she" is no better. Indeed, it is in two ways worse. In thefirst place, prose should be transparent, like a pane of glass through which onesees the thoughts behind it; the use of "she" for these purposes is like a smudgeon the pane - it captures the attention. In the second place, those who now use"she" in this way are making a moral point in doing so, a moral point I thinkentirely right; but it is annoying to have that moral point introduced (with theback of the hand, as it were) into matters it has no connection with — one feelsnagged. All the same, "he or she" very soon yields impossible clutter.1

I would merely add that "one" also very soon yields impossible clutter,while "they" simply generates barbarisms.)

1 Thomson (1990), p. 3, n. 1.

xi

It is customary for prefaces to include an outline of the text to follow,but frankly I don't think I can say, in brief, much that is useful beyond whata perusal of the Table of Contents will reveal, and so I will not try to do so.Instead, let me just deal with one other preliminary matter, namely, the sys-tem of numbering that is adopted in the text. Numbers will be used to referto chapters, sections, subsections, charts, figures, and tables. The practicethat I shall adopt is illustrated in the following examples: Chapter 1, Section1.1, Subsection 2.1.1, Chart 1.1, Figure 2.1, Table 1.1.1 hope this practiceis self-explanatory. In no case will such numbers be used alone; in everycase it will be explicitly stated whether reference is being made to a chap-ter, section, chart, and so on. I shall also use numbers to refer to proposi-tions (and portions thereof). Here the numbers will stand alone, but theywill always appear within parentheses. (Thus, for example, "(1.1)" refers tothe first proposition so treated in Chapter 1.) When, in referring to propo-sitions, I make use of a prime or primes, this indicates a variation on someother proposition. (Thus (2.41') is a variation on (2.41).) When I make useof a lower-case letter, this indicates a variation on part of some other propo-sition. (Thus (2.41'a) is a variation on part of (2.41').) In addition I haveemployed the capital Roman numerals (I) — (XVI) to refer to certain coreanalyses of concepts having to do with obligation; (I) — (XII) are introducedin Chapter 2, (XIII) - (XVI) in Chapter 4. Finally, in Chapter 1 I also usethe capital letters (A) - (E) to refer to certain portions of propositions thatI there call "act-categories." All of this is, I confess, rather cumbersome, butI have sought in vain for a more wieldy method of reference.

xn

Acknowledgments

It is my pleasant duty to acknowledge the help that I have received fromothers in the writing of this book.

I thank the students in my graduate seminars on obligation given atBrown University in 1984 and 1987. They helped me deal with the mate-rial in its early stages of development.

I was awarded a research leave by the University of North Carolina atGreensboro for the fall semester of 1993. This enabled me to complete afirst draft of the book. I believe I would still be engaged in writing the bookhad I not received this leave, and I am very grateful for it.

Several anonymous referees, both for Cambridge University Press andfor certain journals, provided penetrating criticism and useful advice con-cerning portions of the manuscript. I am very thankful for this help. I amthankful, too, for permission to reproduce (often with modification) por-tions of previously published works. These works are listed in the List ofWorks Cited as: Zimmerman (1986) - copyright 1986 by D. ReidelPublishing Company, reprinted by permission of Kluwer AcademicPublishers; (1987) - reproduced by permission of the editor of the AmericanPhilosophical Quarterly; (1988b) - reproduced by permission of the editorof Philosophical Papers; (1990b) - copyright 1990 Kluwer AcademicPublishers, reprinted by permission of Kluwer Academic Publishers; (1992)- copyright 1992 Kluwer Academic Publishers, reprinted by permission ofKluwer Academic Publishers; (1993b) - reproduced by permission of theeditor of the American Philosophical Quarterly; and (1995a) - copyright 1995Kluwer Academic Publishers, reprinted by permission of Kluwer AcademicPublishers.

I received a great deal of help from a number of other people. It is a dis-tinct pleasure to acknowledge my indebtedness to them here. They are:Neera Badhwar, Randolph Clark, James Coley, Earl Conee, David Copp,Ishtiyaque Haji, David Heyd, Thomas Hill, Jr., Thomas Hill, Sr., JoshuaHoffman, John King, Jarrett Leplin, Alfred Mele, Gregory Mellema,

John Pollock, Gerald Postema, Philip Quinn, Andrews Reath, GaryRosenkrantz, Geoffrey Sayre-McCord, Ted Sider, Ernest Sosa, LynneTirrell, and Gregory Velazco y Trianosky.

Finally, there are four individuals whom I wish to set aside for specialmention, since their contribution to the present work has been soextensive.

Peter Vallentyne read a version of the entire manuscript and providedme with numerous detailed comments. He stopped me from making anumber of errors and pointed the way to a number of improvements, forall of which I am most thankful.

Terrance McConnell read various versions of the entire manuscript,some portions several times, and always provided helpful and judiciouscomments. He has set himself a dangerous precedent. As an expert on grat-itude, he knows how grateful to him I should be. I think that in thisinstance, if not in others, I have met the demands of morality.

Roderick Chisholm has been a source of tremendous inspiration to methroughout, and indeed prior to, my professional career. He gave me soundadvice and staunch encouragement when I needed them most. His supporthas been invaluable, and I extend to him my deepest thanks.

Fred Feldman has shaped this work more than any other person. Hiswork on obligation is the source of a great many of the ideas expressed inthis book; anyone acquainted with his work will immediately see howmuch I have lifted practically wholesale from it. In addition, he is a won-derful teacher and an extremely incisive critic ("cruel but fair," as MontyPython might say). Even though he will doubtless find much still tocriticize in this book, it is to him that I dedicate it with great gratitude andrespect.

xiv

1

Groundwork: Some distinctions

I believe that one ought to do the best one can. You may be inclined todisagree. If you are, I think it's because you don't yet know what I mean.Let me try to explain.

1.1 MORAL AND N O N M O R A L OBLIGATION

I am using the term "ought" to express moral obligation. It is of coursesometimes used to express what is required not by morality but by pru-dence, or law, or aesthetics, or the rules of chess, or whatever. "You oughtnot to have used so much red," for example, is likely to constitute aesthe-tic rather than moral criticism. "You ought to watch your pennies" is like-ly to constitute prudential rather than moral advice.

This well-recognized fact gives rise to a profound puzzle that I shallmention only to put aside. Suppose that you have investigated your optionsand reached the disconcerting conclusion that, although morality requiresyou to do one thing, prudence requires another, and the law yet a third (andaesthetics a fourth, and so on). You might find it hard to decide what to do.You might turn to a friend for advice and say: "This is my situation. Whatought I to do?" But what do you mean? "What ought I to do from the moralpoint of view?" Surely not, for you've already settled on an answer to thatquestion. Similarly, you've already settled on answers to the correspondingquestions having to do with the various nonmoral points of view.Presumably, then, you mean, "What ought I to do, period?" This is verypuzzling. It appears to presuppose that the various points of view, moraland nonmoral, are in some manner commensurate with one another, sothat they may be weighed one against the other and your question receivea principled answer. The answer is given in terms of this "ought, period,"which, it seems, is supposed in some way to transcend each of the variouspoints of view, moral and nonmoral.1 For my part, however, I find that I

1 Cf. Feldman (1986), pp. 212-15.

1

have no grasp of any such transcendent "ought"; I do not understandhow it could be that the various points of view, moral and nonmoral, arecommensurate with one another.

But perhaps this is just a deficiency in me. At any rate, as I said, I shallput the matter aside, for even if it makes sense to say on occasion that oneought, period, not to do what one ought morally to do, this would of coursenot alter the fact that one still ought morally to do what one ought moral-ly to do. My concern here is exclusively with what one ought morally to do,and my contention is that, morally, one ought to do the best one can.

In saying this, I don't mean to deny that nonmorally one ought to dothe best one can (in some sense), but only to assert that morally one oughtto do the best one can (in some sense to be elaborated on later). It may wellbe that nonmoral obligation is, in many instances, to be analyzed in termssimilar to those in which moral obligation is to be analyzed.2 Indeed, itwould be surprising if this were not so. But whether or not it is so is not myconcern here.

1.2 BINDING AND NONBINDINGSENSES OF "OUGHT"

"Ought" doesn't always express requirement or obligation, even withinmorality. Sometimes it expresses an ideal, a desideratum rather than a moralnecessity. Here we might use "should" instead of "ought"; we wouldprobably not use "must." Examples of this use of "ought" are: "Everyoneought to be happy," and "Little children ought not to have to suffer."

Sometimes the ideal, nonbinding sense of "ought" is called the "ought-to-be," while the binding sense of "ought" is called the "ought-to-do."3

This is suggestive but a bit misleading. The idea is that, when discussingwhat is ideal or desirable, we can express our thoughts in the following way(to use the illustrations just presented): "It ought to be that everyone behappy," and "It ought to be that little children not have to surfer"; where-as, when discussing what is obligatory, we can express our thoughts in thefollowing way (to employ fresh illustrations): "Tim ought to do what he'sbeen told," and "Pru ought to do what she promised." The trouble is thatthe "It ought to be that" locution can be applied to what is done, too (asin: "It ought to be that Helen do what she can to help out"), and anywayit is surely arguable that certain actions are sometimes morally desirable but

2 Cf. Feldman (1986), Chs. 5 and 7.3 See, e.g., Broad (1985), pp. 225-35; Chisholm (1974), pp. 11-13; Castaneda (1975),

p. 207. Cf. Marcus (1966) on "evaluative" and "prescriptive" senses of "ought."

not obligatory.4 In addition, it seems that what is obligatory need not besomething that is done (as evidenced in: "Mary ought to be with her ailingmother").5 It is better, therefore, to talk of what is binding or obligatoryand to contrast it with what is nonbinding or ideal.

Many philosophers have contended that there is no nonbinding sense of"ought (morally)," or that, at best, such a nonbinding sense is derivativefrom the binding sense (as in, perhaps: "God ought to prevent the suffer-ing of little children").61 am inclined to think that this is mistaken, to think,that is, that there is a perfectly respectable nonbinding sense of "ought(morally)" that is not to be analyzed in terms of some binding sense of"ought (morally)."7 Consider, for example, what Roderick Chisholm hasto say about supererogation: "The status of supererogation might be sum-marized in this way. If I seek advice, concerning an act which would besupererogatory, and ask 'Shall I do this?', I may well be told, 'You ought to,but you don't have to' — it is advisable, but not obligatory." Of course, thereare other readings to be given to "You ought to, but you don't have to."Someone might mean by this, "You are nonmorally (e.g., prudentially)required to do so, but you are not morally required to do so," or "You aremorally required to do so, but no one has any moral claim on you that youdo so." Fair enough. (I shall have more to say about the latter possibility ina moment.) What I don't see is that someone cannot also quite properlymean, "It would be morally desirable if you were to do so, but you are notmorally required to do so (or even to try to do so, or to do somethingrelated to doing so, or...)."

Here, though, someone might object: "What can it mean to say that Iought morally to do something if I am not morally obligated to do it?" Theidea underlying this objection is not that it is improper to use "ought" inan ideal, nonbinding sense, but that such an "ought" cannot be said to bea moral "ought." If "Little children ought not to have to suffer" implies nomoral obligation, how is its "ought" a moral one? The objection concernsthe proper extension of the term "moral." This is a huge issue, which, onceagain, I mention only to put aside. I have no ax to grind here and, indeed,have some sympathy with the objection, although on balance I am inclined

4 Cf. Mellema (1993).5 Broad himself notes these facts in Broad (1985), pp. 233-4. Cf. also Humberstone (1991),

p. 146.6 See Moore (1922), p. 319; Prichard (1949), p. 4; Ross (1939), pp. 45-6; Broad (1985),

p. 235.7 Cf. Castafieda (1975), p. 207.8 Chisholm (1968), p. 417.

to think it perfectly legitimate (whether or not it is true) to say that thereare certain goals which, from a moral point of view, are ideal or desirablebut which do not imply any obligation on anyone's part. Whether or notthis is legitimate, however, is not germane to the present inquiry. Howeverthe ideal "ought" is to be analyzed9 and whether or not it is properly saidto be a moral "ought," I shall confine my attention to the "ought" of moralobligation.

But ought the ideal "ought" to be so casually set aside? Might it not bethat, while it is not to be analyzed in terms of the "ought" of obligation,the reverse is true? Of course, I concede that if, as some contend,1 moralobligation were to be so analyzed, the notion of an ideal "ought" could notbe so easily dismissed in such an inquiry as this. But I deny the contention.The account of moral obligation that I shall give will not make anyreference to the ideal "ought."

Might there nonetheless not be some interesting connection betweenthese two "oughts," so that a full account of the "ought" of obligationwould still require that greater attention be given to the ideal "ought" thanI propose to give? Consider this example provided by Bernard Williams.Suppose that

(1.1) someone ought to help that old lady,

but also that

(1.2) Jones is the only person who can help her;

it follows (Williams claims) that

(1.3) Jones ought to help her.

Williams contends that (1.1) does not express an obligation but that (1.3)does, and that (1.3)'s derivation from (1.1) and (1.2) indicates that there issome very close connection between the binding and nonbinding senses of"ought."11 If he is right, then my failure to pay attention to the ideal"ought" in my attempt to account for the "ought" of obligation might wellseem rash. But I don't think that he is right. An indication that somethinghas gone awry in the derivation of (1.3) from (1.1) is that, as Williamsintends for them to be understood, (1.1) (which involves the nonbinding

9 Forasamplingofanalyses,see:Chisholm(1974),p. 11; Jackson(1985);Feldman(1986),Section 8.2.

10 E.g., Chisholm (1974), p. 13; Jackson (1985), p. 193.II Williams (1981), p. 116.

sense of "ought") is what might be called "passive-transformable," whereas(1.3) (which involves the binding sense) is not. (Williams mentions this factbut appears not to regard it as a problem for the derivation.) That is, (1.1)may be transformed, without loss of meaning, into its passive equivalentas follows:

(1.1') that old lady ought to be helped by someone;

but (1.3) cannot be similarly transformed without loss of meaning asfollows:

(1.3') that old lady ought to be helped by Jones,

for it is Jones who has the obligation, not the old lady. Of course, (1.1) and(1.3) could be read differently, either as neither being passive-transformableor as both being (more on the former possibility in Chapter 9), and theargument's validity might thereby be salvaged, but this will not provideWilliams with the link that he wants between the binding and nonbindingsenses of "ought."

1.3 OVERALL AND PRIMA FACIE OBLIGATION

Ever since the publication in 1930 of The Right and the Good by W. D. Ross,it has been common practice to distinguish between two species of moralobligation. Ross himself used a variety of terms to refer to these species. Onthe one hand there is what he called "absolute obligation," "actual obliga-tion," "duty proper," or "duty sans phrase" \ on the other hand there is whathe called "prima facie obligation," "prima facie duty," "conditional duty,""duty ceteris paribus," or "what tends to be duty."12 None of these terms isparticularly felicitous, as Ross himself acknowledged. More appropriate,perhaps, is to talk of "overall obligation" or "obligation all things consid-ered" for what Ross meant by "absolute obligation," and this is what I shalldo. It would also be more appropriate to talk of "obligation some thingsconsidered" (or "in some respects") or of "pending" or "pro tanto"13 obli-gation for what Ross meant by "prima facie obligation." Nevertheless,since the term "prima facie obligation" is by now so well entrenched, it willbe easiest to use it.

The distinction between the two species of obligation is very familiar,even if not very well understood. One can find oneself in a situation wherea variety of moral considerations pertain. What one ought to do insofar as

12 Ross (1930), pp. 18-20, 28, 30.13 This is the suggestion in Kagan (1989), p. 17.

some of these are concerned may not be what one ought to do insofar asothers of them are concerned. What one ought on balance to do will, pre-sumably, be some function of these various "insofar as" "oughts"; to put itdifferently, what one's overall obligation is will be some function of whatone's various prima facie obligations are (if any).

It is worth repeating that what is presently at issue is two species of moralobligation. To say that overall obligation has to do with what one ought"on balance" or "all things considered" to do is, in this context, to say thatit has to do with what one ought to do all morally relevant things consid-ered.14 The question that was briefly raised above concerning the com-mensurability of various points of view, both moral and nonmoral, is notat issue here. We are now working entirely within the moral point of view.(This is not to say that there is no question about the commensurability ofdifferent moral considerations; there is. It is one that I shall address in thenext chapter.)

My primary concern in this work is to provide an account of overallmoral obligation, although I shall also provide an account of prima facieobligation in Chapter 5, an account according to which it is closer to thetruth to say that prima facie obligation may be understood in terms of over-all obligation rather than the reverse.15

Sometimes moral obligations or duties are said to correspond to moralrights or claims. Ross distinguishes four different theses having to do withsuch alleged correlativity. They are:

(1.4) a right of A against B implies a duty of B to A;

(1.5) a duty of B to A implies a right of A against B;

(1.6) a right of A against B implies a duty of A to B;

and

(1.7) a duty of A to B implies a right of A against B.16

Each thesis has two main readings, one where "duty" expresses a primafacie obligation (and hence, presumably, "right" expresses a prima facieright), the other where "duty" expresses an overall obligation. Ross him-

14 Indeed, it is to say that it has to do with what one ought to do all moral-obligation-relevant things considered, inasmuch as some moral considerations, as noted in thelast section, are relevant not to what is obligatory but to what is ideal.

15 Two possible attempts to provide such a "reverse" account are to be found in Ross(1930), pp. 41 and 46-7, and Broad (1985), pp. 171-2.

16 Ross (1930), p. 48. Cf. Feinberg (1973), p. 61.

self (who employs the "prima facie" reading) says that (1.4) is clearly truebut that, since we can have duties to (nonhuman) animals but no rightsagainst them, (1.7) is false.17 Our having duties to animals but their havingno duties to us also leads Ross to note that not both (1.5) and (1.6) can betrue. He is uncertain which to reject but is inclined to say that animals, notbeing moral agents, do not have rights, and therefore that (1.5) is false. (1.6),however, he accepts as true.18

This rejection of (1.5) and (1.7) may be questioned. We may agree thatwe have duties towards or concerning animals, but do we have duties to themin the requisite sense? Ross thinks so, saying: "I suppose that to say we havea duty to so-and-so is the same thing as to say that we have a duty, ground-ed on facts relating to them, to behave in a certain way towards them."19

But it may be that this account of what it is to have a duty to someone orsomething is too broad. It would appear to imply that certain potential ben-eficiaries (such as the homeless) have a right not just to nonmaleficence butto beneficence, if it is agreed that we have a duty to act not just nonmalef-icently but also beneficently towards them. Although some may not balkat this (either because they reject the view that we have a duty to act bene-ficently towards such potential beneficiaries, or because, like Ross, theyaccept the view that such persons have a right to our beneficence), otherswill balk at it and claim that it is possible to have a duty towards or con-cerning somebody or something without owing that duty to that person orthing.21 In this way one might try to support not just (1.4) and (1.6) but alsoto defend (1.5) and (1.7) against Ross's argument. For one could argue thatall and only moral agents are such that duties are owed to them (in the req-uisite sense), and all and only moral agents are such that they owe duties toothers.

My own, tentative view is neither that of Ross nor that just mentioned.Contrary to Ross, and in keeping with the view just mentioned, I believethat there is indeed a difference between having a duty towards or con-cerning someone or something and having a duty to someone or some-thing. The latter, I believe, is correlative to a right held by the someone orsomething in question, whereas the former is not. Hence I accept both (1.4)

17 Ross (1930), pp. 48-9.18 Ross (1930), pp. 49-50, 53-4.19 Ross (1930), p. 49.20 Ross (1930), pp. 52-3.21 Cf. Lyons (1979c), pp. 60-1; Feinberg (1979), p. 79. Also cf. Hart (1979), pp. 17-19,

Feinberg (1980), p. 139, and Hill (1979), pp. 122-3, on the ambiguity of "to" in suchcontexts.

and (1.5). But I reject both (1.6) and (1.7), because, contrary to Ross andto the view just mentioned, I'm inclined to think that, although one mustbe a moral agent to have duties to others, one need not be a moral agent tohave rights against others. In particular, certain animals appear to me to haverights against us but duties to no one.

At this point it might be objected that, while it is possible that one ought(morally) to do something without one's having a (moral) duty or obliga-tion to someone (or something) to do that thing, it is not possible to be obli-gated to do something without one's owing that obligation to someone todo that thing. Consider this example, provided by Joel Feinberg:

[CJonsider how the word "ought" differs from the word "duty." Suppose astranger approaches me on a street corner and politely asks me for a match.Ought I to give him one? I think most people would agree with me that Ishould...

Now suppose that Jones is on the street corner and another stranger polite-ly requests a light from him... He brusquely refuses to give the stranger a match.I think we can agree that Jones's behavior ... does not constitute an ideal forhuman conduct under such circumstances; ... that it was not what Jones oughtto have done.

If we reproach Jones, however, ... he may present us with a vigorous self-defense. "Perhaps I was not civil," he might admit. "But surely I was under noobligation to give a match to that man. Who is he to me? He had no claim onme...; I don't owe him anything. It may be nice to do favors for people; but afavor, by definition, is nothing that we are legally or morally required to do."22

In a comment on his own example, Feinberg says:

Jones's defense makes me think no better of him. Still, ... [e] very thing Jonessaid ... was true. The moral I draw from this tale is that there are some actionswhich it would be desirable for a person to do, and which, indeed, he ought todo, even though they are actions he is under no obligation and has no duty to do.It follows logically that to say that someone has a duty or an obligation to do Xis not simply another way of saying that he ought to do X.23

But this is all very complicated. There are a number of possible positionsto adopt concerning Jones's behavior (let alone Jones himself) in this story.Even if we disregard the distinction between prima facie and overall oblig-ation, still eight different positions suggest themselves. Using the terminol-ogy that I have introduced, one may put these positions as in Table 1.1(where A is the action of giving the stranger a match). Positions 3 and 7

22 Feinberg (1968), pp. 392-3.23 Feinberg (1968), p. 393.

Table 1.1

Position

1

2345678

Jones ought, inthe nonbindingsense, to do A

yesyesyesyesnononono

Jones ought, inthe binding

sense, to do A

yesyesnonoyesyesnono

Jones has anobligation to thestranger to do A

yesnoyesnoyesnoyesno

may be immediately dismissed, since one cannot have an obligation-to todo something that is not something that one ought, in the binding sense,to do. But which position is it that Feinberg accepts? Clearly he does notadopt any position according to which Jones has an obligation to thestranger to give him a match (where this is understood to mean that Jonesowes it to the stranger to do so), and so he does not accept any of positions1, 3, 5, or 7. It seems clear, too, that he does not accept any of positions5-8, according to which it would not be ideal (under the circumstances)for Jones to give the stranger a match.24 This leaves positions 2 and 4, andI don't know which of these it is that Feinberg accepts. His remarks seemto me ambiguous between the two of them, and perhaps for good reason;for even if one agrees that the stranger has no claim on Jones that Jones givehim a match, but that something has gone awry, morally, when Jones failsto give him a match, still it is hard to know which of positions 2 and 4 toaccept. (Position 2, with its binding "ought," implies that Jones did wrong innot giving the stranger a match, whereas position 4 does not.) Of course, ifone rejects the possibility that there be a binding "ought" without therebeing an obligation-to, then one will reject position 2.

As before, however, I raise all this only to put it aside. Feinberg wishesto reserve the term "obligation" for what I have called "obligation-to"; ifhe accepts that position 2 is possible, he would still not express it by saying

24 Some people might now believe that common courtesy would require not giving thestranger a match, due to considerations of health. It is worth noting that Feinberg firstpresented his example in 1960, long before health-related condemnation of smokingbecame common.

that Jones has an obligation, but not an obligation-to, to give the strangera match. That is fine by me. I choose, however, to use the term "obliga-tion" simply as a nominalization of the binding "ought" and to say that, evenon position 2, Jones has an obligation to give the stranger a match. Onemust adopt a certain terminology, and this is the terminology that I, likemany others, adopt. What matters in the end is not what terminology oneadopts but that one get clear as to what the concepts at issue are. That iswhat I am trying to do. My purpose in this work is to clarify the conceptexpressed by the binding "ought." It is important to recognize that, for pre-sent purposes, I am neither presupposing nor denying the view that posi-tion 2 (or position 6) is possible, that is, that it is possible that there be anobligation (whether prima facie or overall) that is not an obligation-to(although, as noted earlier, I am in fact inclined to believe that this is pos-sible). My concern here is to provide an account of obligation without ven-turing into this territory of "to," that is, to provide an account of what maybe called "mere obligation." If it is possible to have an obligation that is notan obligation-to, then my account is designed to say what this amounts to.I shall write henceforth under the assumption that this is possible. But if itis not, still every obligation-to is of course an obligation, and my accountcan then be construed as being designed to say at least to that extent whatobligation-to amounts to.

1.4 OBJECTIVE AND SUBJECTIVE OBLIGATION

So far I have, in effect, been noting the ambiguity of "ought" and trying tonarrow down that sense of it that is my main concern here. This process ofspecification may be represented as in Chart 1.1.25 We have made our way

moral nonmoral

binding nonbinding

overall prima facie

mere obligation obligation-to

Chart 1.125 This diagram is incomplete, in that the various subspecies that have been identified on

the left-hand side have analogues that are not shown on the right-hand side. Thus, non-moral obligation may be subdivided into binding and nonbinding obligation, nonbind-ing obligation may be subdivided into overall and prima facie obligation, and prima facieobligation may be subdivided into mere obligation and obligation-to.

10

to mere obligation. Does this suffice? According to some, it does not; thereis, they allege, ambiguity even here.

To assess this claim of ambiguity, consider a distinction that BertrandRussell draws between different categories of acts. We may distinguish, hesays, between

(1.8) that act which will have the best consequences

and

(1.9) that act which will probably have the best consequences.

He calls the first act the one that is, under the circumstances, the "most for-tunate," while the second is the "wisest."26 Ross and H. A. Prichard drawa similar distinction between

(1.10) that act which best suits the objective features of the agent's situa-tion (that is, which best suits the agent's actual situation)

and

(1.11) that act which best suits the subjective features of the agent's situa-tion (that is, which best suits the agent's situation as the agent con-ceives his situation to be).27

If we ignore the question as to whether what makes an act "best" has to dowith its consequences or with its suitability to the agent's situation (or,indeed, with something else), we get this threefold distinction:

(A) that act which is best,

(B) that act which is probably best,

and

(C) that act which the agent believes to be best.

(1.8) and (1.9) correspond to (A) and (B), respectively, whereas (1.10) and(1.11) correspond to (A) and (C), respectively. (This is all decidedly rough,but further refinement is not pertinent to our present purpose.)

It is on the basis of such distinctions that some philosophers have claimedthat "ought" is ambiguous, even when it has been narrowed down to mereobligation. Russell, for example, distinguishes between what he calls"objective obligation" and "subjective obligation." Perhaps surprisingly,

26 Russell (1970), pp. 12-13.27 Ross (1939), pp. 146-7; Prichard (1949), p. 18.

11

what Russell says corresponds to objective obligation is not (A) but (B), andwhat he says corresponds to subjective obligation is not (C) but this:

(D) that act which the agent believes to be objectively obligatory.

Others, too, have adopted this terminology, but sometimes it is applied dif-ferently. For example, Fred Feldman says that it is (A) and not (B) that cor-responds to objective obligation and that what corresponds to subjectiveobligation is neither (C) nor (D) but this:

(E) that act which it is most reasonable for the agent to believe is objec-tively obligatory.29

Others have used somewhat different terminology. C. D. Broad, forinstance, distinguishes between what he calls "material obligation," "for-mal obligation," and "subjective obligation," saying that (A) correspondsto what is materially obligatory, whereas (C), read in one way, correspondsto what is formally obligatory and, read in another way, corresponds towhat is subjectively obligatory.30 And it might seem that Ross and Prichardwould say that (A) corresponds to what is objectively obligatory and (C)corresponds to what is subjectively obligatory.

Interestingly, though, Prichard does not say this, and Ross appears todeny it too. At one point Ross says the following:

It is clear that when we call an act right we sometimes mean that it suits theobjective features of the situation, and sometimes that it suits the subjective fea-tures. And when people express different opinions about the rightness orwrongness of an act, the difference is often due to the fact that one of them isthinking of objective and the other of subjective rightness. The recognition ofthe difference between the two is therefore in itself important as tending to rec-oncile what might otherwise seem irreconcilable differences of opinion. Butthe question remains, which of the characteristics — objective or subjectiverightness — is ethically the more important, which of the two acts is that whichwe ought to do.31

Here Ross seems to grant the distinction between objective and subjectiveobligation (or, as he puts it in the quoted passage, "rightness"; he draws nodistinction between rightness and obligatoriness) only to snatch it awayagain by asking what the agent (really) ought to do. Prichard is more force-

28 Russell (1970), pp. 13-14.29 Feldman (1986), p. 46.30 Broad (1985), pp. 125-7, 131-2, 144, 172, and Broad (1946), p. 557.31 Ross (1939), pp. 146-7.

12

ful still. He distinguishes between what he calls the "objective" and"subjective" views of obligation and insists that the former view is mis-taken and the latter correct.32 He thus denies that (A) captures a (proper)sense of "obligation," insisting that only (C) does. More recently, JudithThomson has also insisted on unambiguity here, although her view is thatthere is no subjective sense of "ought" (which in her view would, if it exist-ed, correspond to a variation on (D)).33 But others have insisted just asstrongly on ambiguity.34 What are we to think?

I'm inclined to think that we should side with Prichard and Thomson(and perhaps Ross) here and deny that there are several senses of "ought"at issue. Certainly there is no need to think that each of (A)-(E) (and per-haps other act-categories) corresponds to some distinct sense of "ought."Accepting that (A)-(E) are distinct (as one ought to do) does not requireaccepting that each expresses some sense of "ought." Still, I grant that morethan one sense of "ought" may be at issue here. Consider the claim, whichis not uncommon, that one ought to do what one thinks one ought to do.Sometimes this is offered as advice to someone seeking to discover what heought to do. As such, it seems to me likely to be singularly unhelpful,although it may be that it is on occasion appropriate; for it may be that theadvisor takes the advisee, despite the latter's uncertainty, to be a fine judgeas to what he ought to do and is simply telling him that, in this case as inothers, his (the advisee's) judgment can be trusted. At other times, how-ever, the claim seems to be offered as an account of what it is to be such thatone ought (in some sense) to do something. Here the claim seems to bethis: one ought (in some sense) to do what one thinks one ought (in someother sense) to do. If (D) expresses a sense of "ought" - a "subjective" sense,let us say - then it constitutes one version of this claim. But why think thatthere is any sense of "ought" such that one ought (in that sense) to dowhat one thinks one ought (in some other sense) to do? Is it appropriateever to speak in this way? I'm inclined to think not. It seems to me thatanyone who puts matters this way would do far better to put them asfollows: one is to blame if one fails to do what one thinks one ought (in theonly respectable sense) to do. Rather than distinguish two senses of "ought,"as (D) does, and define one (the "subjective" sense) in terms of the other(the "objective" sense), it seems to me preferable to declare that there isonly one sense of "ought" (at this level of inquiry — the level of "mereobligation") and to say that blameworthiness (and other concepts of

32 Prichard (1949), Ch. 2.33 Thomson (1986), p. 179.34 See Broad (1985), pp. 126-7.

13

responsibility) are to be understood in terms of it.35 Still, if it is insisted thatthere are two senses of "ought" here, we need only say in reply that we arepresently concerned just with one of them, namely, that in terms ofwhich the other is allegedly to be understood. If we were using the terms"subjective obligation" and "objective obligation" to draw this distinction,then the point could be put this way: our present concern is to give anaccount only of objective and not also of subjective obligation.36

But I don't like this way of putting it, and not just because I'm inclinedto deny the ambiguity of "ought" at this level. I don't like it because it pre-supposes that "ought" (in the sense that we are concerned with) is in someimportant way "objective" (whatever that may mean, exactly), and,although I do believe this, it is not something that I wish to presuppose forpresent purposes. For Prichard's question is still an open question: even if"ought" (in the sense that we are concerned with) is not to be understoodin terms of (D) (or (E), or any other account that itself makes reference to"ought"), how is it to be understood? For all that I've said, it may be thatPrichard is right and that it is to be understood in terms of (C); and it wouldbe highly misleading- as Prichard would surely agree - to say that "ought,"so understood, is "objective."

Here, though, it may seem to you that I've made a mistake. I began bysaying that one ought to do the best one can. Surely, you may think, thiscorresponds to (A) and not (C) (and not (B) either), and so the "ought" thatI'm after is properly called objective. But if you think this, you probablystill don't know what I mean. Let me try to set matters straight.

I am using the term "best" very liberally. In saying that one ought to dothe best one can, I am contending merely that there is some way in whichwhat one ought to do is superior to any other of one's alternatives. I shallcall such superiority deontic. My claim, then, is that, to have the deontic sta-tus of being obligatory, an act's deontic value must be greater than that of anyother alternative action. (Note that I am here restricting the use of "deon-tic" to the moral realm, whereas others sometimes use the term more wide-ly. An act's deontic value, then, although not necessarily itself a species ofmoral value, is that which, in conjunction with the deontic values of its

35 Zimmerman (1988a) contains a detailed elaboration of this view.36 It should be acknowledged at this point that some philosophers accept that "ought" is to

be defined in terms of a belief about "ought" and yet deny that there are two senses of"ought" at issue. Strawson (1986), for example, claims that, as a matter of definition, onecannot do wrong unless one believes that one is doing wrong — where the term "wrong"is used univocally as the contrary of "ought" (p. 220). This claim strikes me as incoher-ent, but there is no space to pursue the matter here.

14

alternatives, determines the act's deontic status — as morally obligatory,right, or wrong — which is a moral feature of the act.) For reasons that willemerge shortly, I shall steadfastly decline to specify what it is that deonticvalue consists in; I wish merely to commit myself to the formal view thatone ought to perform an action if and only if that action is, in some way,ranked first relative to all one's other options. This very abstract (and stillvery rough) account of "ought" is compatible, not just with the view that"ought" corresponds to (A), but also with the views that it corresponds to(B) and (C), each of which involves some more substantive, and so less lib-eral, construal of "best." To see this, suppose that we adopt, merely for thepurpose of illustration, the position that "best" in (A), (B), and (C) refers tothe act's instrumental value. Then, according to the view that "ought" cor-responds to (A), what is deontically superior is what is instrumentally best,while according to the view that "ought" corresponds to (B) what is deon-tically superior is what is probably instrumentally best, and according to theview that "ought" corresponds to (C) what is deontically superior is whatis believed instrumentally best by the agent. On this understanding of"best," then, the view involving (A) incorporates the substantive moralview that deontic value just is actual instrumental value, while the viewinvolving (B) incorporates the substantive moral view that deontic valuejust is probable instrumental value, and so on. Insofar as what is actuallyinstrumentally best need not be what is probably instrumentally best, whichitself need not be what is believed instrumentally best by the agent, thesethree views will on occasion issue in three different deontic rankings of theagent's options. Nonetheless, each of these substantive views is a species ofthe purely formal view that an act is obligatory if and only if its deontic valueis greater than that of (that is, the act is deontically superior to) any of itsalternatives.

Note how modest this formal view is. It merely states that what is oblig-atory is unique relative to its alternatives and that this uniqueness consti-tutes some sort of morally pertinent superiority; there is no commitment toany substantive view as to what this superiority consists in.37 In this book Ishall deal in detail only with this formal claim; I shall not inquire whether,on a more substantive level, it is to (A) or (B) or (C) (or, indeed, somethingelse) that "ought" corresponds.

37 Cf. Moore (1903), p. 147: "It is plain that when we assert that a certain action is ourabsolute duty, we are asserting that the performance of that action at that time is uniquein respect of value." Of course, Moore is here talking about the instrumental value of theact in particular, whereas, as I have just tried to explain, I am not committing myself tothe view that deontic superiority consists in instrumental superiority.

15

Still, perhaps I should say in passing that I find the view that "ought"corresponds to (C) wholly implausible. Prichard argued for this view,38 asI have noted, and Ross, persuaded by Prichard, abandoned his earlier view(which was that (A) captured what "ought" consists in)39 in favor ofPrichard's view.40 The reasons given by both Prichard and Ross seem quiteinadequate, but there is no need to go into this here, since the presentaccount does not presuppose that "ought" does not correspond to (C).4 Isuspect that, frequently at least, someone who is attracted to (C) in thisregard confuses it with (D) and also fails to distinguish between blame wor-thiness (and other concepts having to do with moral responsibility) on theone hand and wrongdoing (and other concepts having to do with moralobligation) on the other.

I find it much harder to decide whether it is more plausible to say that"ought" corresponds to (A) or that it corresponds to (B). What is it that aconscientious person, who is trying to determine what he ought to do,seeks to discover: that act which is best (in some substantive sense) or thatact which is probably best? There is reason to think the former. Suppose thathe believes that he has a duty to help those in distress and that he comesupon the victim of a car accident. He must choose whether to leave thevictim where he is or to pull him free from the wreck. Suppose that he pullsthe victim free, believing that this is probably best. Does he believe that heis doing what he ought to do, or does he believe only that he is probablydoing what he ought to do? Quite possibly the latter. After all, he may laterdiscover that it would have been better to leave the victim where he wasand thus come to believe, regretfully, that (however innocently, and how-ever much the evidence pointed the other way) he in fact did the wrongthing.

Yet there is also reason to think that it is more plausible to say that"ought" corresponds to (B) rather than (A). Consider this case recently pro-posed by Frank Jackson: Jill, a physician, can prescribe one of three drugs,X, y, and Z, for a certain illness and knows that X will probably relieve butnot cure the ailment, while Y will either cure it or kill the patient and Zwill either cure it or kill the patient; but Jill doesn't know which of Y andZ will cure and which kill.42 If "ought" corresponds to (A), then presum-

38 Prichard (1949), Ch. 2.39 Ross (1930), p. 31: "There is therefore much truth in the description of the right act as

a fortunate act."40 Ross (1939), Ch. 7.41 For a lucid exposition of what goes wrong with Ross's case for (C), see McConnell

(1988b).42 Jackson (1991), pp. 462-3.

16

ably Jill ought not to prescribe Xbut either ought to prescribe Yor oughtto prescribe Z. Jackson claims that this proposal is "intuitively wrong."43

I'm not sure that he is right, but neither am I sure that he is wrong.44

It is tempting to try to resolve such a puzzle by diagnosing ambiguity.But, again, I resist this here. Surely Jill, as a conscientious physician, is ask-ing just one question when she says: "What is it that I ought to do in thiscase?" But if there is ambiguity, then I shall simply say this: the "ought" thatI'm after is the one that conscientious people like Jill are concerned with;also, it is the one which is such that, if someone fails to do what he believeshe ought to do, then he is to blame.45

Here, though, I don't have to decide between these views about (A) and(B). As I've noted, the account of "ought" that I shall give is compatiblewith each of them. In fact, it is compatible with a wide variety of otherviews not yet mentioned. I asked whether it is more plausible to say that"ought" corresponds to (A) or to say that it corresponds to (B), but perhapssome other view is more plausible still. Certainly other views have beenproposed. According to some philosophers, what one ought to do is a func-tion not simply of what one can do but of what one knows that one can do(so that, for example, one does not do wrong in failing to rescue someoneof whose plight one is nonnegligently ignorant).46 According to otherphilosophers, what one ought to do is a function of what one intends, andthis may or may not have anything to do with what is, or is probably, or isbelieved to be, best (in some less-than-fully-liberal sense). Kant is, ofcourse, the obvious figure to cite here,47 but one needn't be a Kantian toagree that intended harm constitutes a more serious wrong than unintend-ed harm - a matter of adding insult to injury48 - and it is difficult to sub-sume this satisfactorily under either (A) or (B) or (C), at least as these wouldordinarily be understood.49 Again, many philosophers have contended thatwhat one ought to do is a function of respecting persons' rights;5 othershave contended that what one ought to do is a function of certain so-called

43 Jackson (1991), p. 466.44 Cf. Prichard (1949), p. 29; Ross (1939), p. 152; Thomson (1986), Ch. 11.45 See Zimmerman (1988a), Ch. 3.46 Cf. Prichard (1949), pp. 22-5; Lemos (1980), p. 301.47 Kant (1964), pp. 67-8.48 Cf. Hume (1888), p. 349; Rousseau (1992), p. 144.49 Intention also plays a central role in the oft-invoked principle of double effect, a princi-

ple which, as commonly interpreted, concerns the (im)permissibility of acting in certainways.

50 Almost any deontologist could be cited here, such as Kant (1964), p. 96, Donagan (1977),p. 65, or Fried (1978), pp. 28-9 and 81-6.

17

agent-relative values;51 and so on; and again it seems strained, at best, to tryto subsume these under (A) or (B) or (C), at least as these would ordinari-ly be understood. But my account, I claim, in abstracting from all that issubstantive in any ordinary sense of "best," is compatible with (almost)52

any such view; for (almost) any such view is a view according to which whatis obligatory is superior in some morally relevant way to its alternatives. Forpresent purposes, then, the principle by which alternatives are to be rankedcan be left wholly unspecified, as long as what is morally obligatory isacknowledged as morally unique (in some positive way) relative to itsalternatives.

It is important to note that I am not seeking here to effect some sort ofreconciliation between consequentialist and deontological views, as mightbe attempted, for example, by a consequentialist who ascribes intrinsicvalue to justice, or promise keeping, or the respect of rights.53 That wouldbe a substantive moral enterprise. Rather, I am simply claiming that, on amore abstract level, there is something common to (almost) all substantivetheories of obligation, whether consequentialist or deontological (or some-thing else, if "deontological" is narrower than "nonconsequentialist"), andthat is that they share this view about the superiority of what is obligatoryto what is nonobligatory. That which is superior maximizes some value and,I contend, (almost) all substantive theories of obligation can be appropri-ately cast in a maximizing mold. Thus, besides consequentialism ("It isobligatory to do that which maximizes intrinsic value"), each of rule utili-tarianism ("It is obligatory to do that which maximizes rule utility"), Ross'searly theory ("It is obligatory to do that which maximizes prima facie strin-gency"), Kant's theory ("It is obligatory to do that which maximizes com-pliance with the Categorical Imperative"), the divine command theory ("Itis obligatory to do that which maximizes compliance with God's com-mands"), and rights theory ("It is obligatory to do that which maximizesnonviolation of persons' rights"), and so on, can be put in terms of maxi-mization - in terms, that is, of doing what is "best" in some sense.54

This is not a new observation.55 Still, it might strike some as peculiarlyunhelpful to represent (almost) all substantive theories of obligation interms of maximization. Why do it? Because, as I hope to show in ensuingchapters, it is extraordinarily fruitful; we can come to a finer understand-

51 E.g., Nagel (1986), Ch. 9.52 The reason for the qualifier "almost" will be given in Chapter 2.53 Cf. Feldman (1986), Section 3.3; contrast Slote (1989), pp. 91-2.54 Cf. Lewis (1973), p. 76.55 See Nozick (1974), p. 29n. Cf. Sen (1983), p. 128ff., and Dreier (1993), pp. 24-5.

18

ing of the conceptual structure of obligation if we conceive of it in maxi-mizing terms. Michael Stocker claims that, if putting a theory of obligationin terms of maximization is to be informative, then the value that it is saidshould be maximized must be specifiable independently of the determina-tion of what one ought to do. This condition appears to be satisfied by cer-tain theories (such as consequentialism, where in principle what isintrinsically valuable is determined first, and then it is said that what oughtto be done is that which maximizes intrinsic value) but not by others (suchas Kant's theory, which presumably accords no value to a certain maxim'sbeing universalizable independently of its being such that one ought to acton it; or such as the divine command theory, which presumably accords novalue to certain actions independently of their being commanded by God).And Stocker says:

It might seem that the most that could be charged against maximizers makingthis move [of viewing all substantive theories of obligation in maximizingterms] is that they have trivialized maximization by making it conceptually true.Many maximizers would ... take this as an encomium, not a charge.

But I think they would be mistaken in seeing matters this way. For maxi-mizers are now open to the charge of shirking the needed honest toil... Afterall, one important task in ethics.. .is finding which objects of care are such thatmore of them is better.56

But while I certainly grant that "making maximization conceptually true"does indeed leave much work in moral theory left to be done, I deny thatit amounts to "trivialization." For much work can be accomplished by thismove.

Some have charged that such a move is not just trivial but "gimmicky."57

The term appears to be a pejorative one. But while I concede that the movedoes not do much to illuminate certain substantive theories with respect tohow they are to be contrasted with others (does it help us to understandwhat is distinctive about Kant's theory to put it in terms of maximization?Surely not), I do believe that it helps illuminate the concept of obligationitself. And I concede too that, if a certain theory (such as Kant's, perhaps,or the divine command theory) implies, when put in terms of maximiza-tion, that that which is obligatory is infinitely deontically superior to thatwhich is wrong (so that, we might say, that which is obligatory has a valueof 1 and all other alternatives a value of 0), then nothing is really gained inthis instance by using the terms of maximization. But neither is anything lost.

56 Stocker (1990), pp. 299-300.57 Nozick (1974), p. 29n.; Vallentyne (1988).

19

Moreover, what such a lack of gain shows is simply that such theories arein fact extremely impoverished, conceptually. We shall see that, wherefiner deontic rankings of alternatives are permitted by a theory, we can pro-vide an account not just of mere obligation but also of conditional obliga-tion, prima facie obligation, and what I shall call "subsidiary obligation,"and also that what I shall call the "dynamics" of obligation are illuminated.Theories with minimal rankings (theories that rank what is obligatory first,all else last) are themselves limited, in that they cannot express these otheraspects of the concept of obligation.

1.5 ACT-EVALUATION AND O T H E R TYPESOF EVALUATION

Much of ethics has to do with the evaluation of agents, their characters,their motives, and so on. Some of ethics has to do with the evaluation ofacts. An inquiry, such as the present one, into the concept of obligationmight be classified as an inquiry into the evaluation of acts. It is certainlyquite distinct from an inquiry into some types of evaluation of agents, theircharacters, their motives, and so on. Still, the lines are hard to draw withprecision. We might describe the present work as an inquiry into what it isfor an act to be obligatory, and then the classification of it as an inquiry intothe evaluation of acts might seemjustified. But, although I shall often write,and have already written, as if such classification is appropriate, we mightequally well describe the present work as an inquiry into what it is for anagent to be obligated to do some act, and then such classification might seemless appropriate.58 The fact is that I am not really concerned with what theproper classification in these terms is; I hope simply that I have done enoughby now to give you a fair idea of what the subject of inquiry is. I proposenow to turn to that inquiry.

58 Cf. Prichard (1949), pp. 36-7; Ross (1939), p. 56.

20

Moral obligation: An analysis

2.1 VALUE MAXIMIZATION

2.1.1 Two problems

At his doctor's request, Sam sticks out his tongue and says: "Aah." In orderto vote, Vincent drives his car to the polling station. Puzzles lurk withinthese mundane examples.

The claim that one ought to do the best one can (where, as explained inthe last chapter, "ought" expresses what I have called "mere obligation,"and "best" refers to deontic value) is perhaps most naturally, indeed stan-dardly, understood to mean the following: of all those actions which onecan perform (one's "alternatives," as it is often put), one ought to performthat action (if any) whose deontic value would be greater, if it were per-formed, than the deontic value of any other, if it were performed. But thereare at least two problems with so understanding it.1

The first problem was exposed by Hector-Neri Castaneda.2 Supposethat it is deontically best that Sam comply with his doctor's request.Suppose, that is, that

(2.1) Sam ought to stick out his tongue and say: "Aah."

Then presumably

(2.2) Sam ought to stick out his tongue, and Sam ought to say: "Aah."

But, on the current understanding of what it is to be obligated to do thebest one can, the first part of (2.2) implies that

(2.3) the deontic value of Sam's sticking out his tongue would be greater,if it were performed, than the deontic value of his saying "Aah," if itwere performed;

1 For a detailed discussion of this issue, see Feldman (1986), Ch. 1.2 Castaneda (1968).

21

while the second part of (2.2) implies that

(2.4) the deontic value of Sam's saying "Aah" would be greater, if it wereperformed, than the deontic value of his sticking out his tongue, ifit were performed.

But (2.3) and (2.4) are manifestly inconsistent with one another.There are five main responses to this puzzle. One might, first, deny that

it is legitimate to talk of compound events (such as that of Sam's stickingout his tongue and saying "Aah") as actions or, at least, as the sort of thingone ought to do. But this is surely a revisionary proposal. Second, onemight deny that (2.1) implies (2.2). But, although some do deny this, it issurely counterintuitive to do so.3 Third, one might deny that Sam's stick-ing out his tongue is a genuine alternative to his saying "Aah." Although Ihave some sympathy with this position, it is of no help to the thesis that oneought to do the best one can, on its present interpretation. For, on thisinterpretation, the thesis concerns simply all those actions that one can per-form, and certainly each of these actions is such that Sam can perform it.Perhaps it is a mistake to call these actions "alternatives," but that is strict-ly irrelevant. Fourth, one might deny that (2.3) and (2.4) are inconsistent.But this seems patently false. Fifth, one might hold that the thesis that oneought to do the best one can requires a somewhat different interpretationfrom the standard one. This, I believe, is the correct response to thepuzzle.

The second problem reinforces this response. Suppose that the deonticvalue of Vincent's voting would be greater, if he were to perform it, thanthe deontic value of any other action that he can perform, if he were to per-form it. In short,

(2.5) Vincent's voting would be deontically superior to his not voting.

Then the theory, on its current interpretation, implies that

(2.6) Vincent ought to vote.

But suppose also that, in order to get to the polling station, Vincent mustdrive his car. That is,

(2.7) Vincent cannot both vote and not drive his car.

Suppose, finally, that

(2.8) Vincent's driving his car would be deontically inferior to his notdriving it,

3 More on this later, in the present chapter and in Chapter 6.

22

and thus that, on the theory's current interpretation,

(2.9) it is not the case that Vincent ought to drive his car.

If we also assume that

(2.10) if one ought to perform a certain action A and cannot do so with-out performing a certain action B, then one ought to perform B,

we reach a contradiction.Again, a number of responses are possible. One might deny the gener-

al principle embodied in (2.10), but this is surely counterintuitive.4 It isworth noting here that (2.10) implies the claim that (2.1) implies (2.2);denial of this claim would therefore require denial of (2.10).5

Second, one might question the consistency of (2.5), (2.7), and (2.8).How can it be that Vincent's voting is better than his not voting, and hisdriving worse than his not driving, when he cannot vote without driving?The answer must, of course, rest on just what deontic value consists in, butcertainly this is perfectly possible on some substantive views about suchvalue. Consider, just by way of example, the view that an action's deonticvalue is a function of the sum intrinsic value of its consequences. It couldwell be that the consequences of Vincent's driving (including those conse-quences which are the action of voting and its consequences) would beintrinsically worse than those of his not driving, whereas the consequencesof his voting (now that the driving, for good or ill, has occurred) would beintrinsically better than the consequences of his not voting. Or consider,by way of another example, the view that an action is deontically superiorto its omission if and only if the action's maxim is universalizable whereasthe omission's is not. As far as I can tell, there is no guarantee that, if themaxim for Vincent's voting were universalizable and its omission's not,then the maxim for his driving would be universalizable and its omission'snot. Or, finally, consider the view that an action is deontically superior toits omission if and only if the action does not involve the infringement ofanyone's rights whereas the omission does. Again, one might wonderwhether there is any guarantee that, if Vincent's voting involved no suchinfringement, his driving would also involve no such infringement.

4 Again, more on this later and in Chapter 6.5 Let A be the compound action of Sam's sticking out his tongue and saying "Aah" and B

be the action of his sticking out his tongue. Since Sam cannot perform A without per-forming B, (2.10) implies that, if he ought to perform A, then he ought to perform B.

6 Some rights-theorists in fact insist on such a guarantee. See Thomson on the "MeansPrinciple for Claims" in Thomson (1990), pp. 156—7. This is in contradistinction to herearlier view, in Thomson (1986), p. 109, that there is no such guarantee.

23

A third response to the second problem is this. One might take note ofthe passage of time involved and contend that it is a mistake to claim thatboth (2.6) and (2.9) are true. At the time at which it is true to say that it isnot the case that Vincent ought to drive his car, it is not yet, and may neverbe, true that Vincent ought to vote. Whether or not he ought to vote willdepend on whether he drives. Only if he drives will it be the case that heought to vote. But while this response may seem plausible, there are rea-sons to think it mistaken; I shall go into these in Chapter 3. More to thepoint here, the response is irrelevant, since the problem at issue does notrest on the passage of time. Consider a parallel case in which

(2.5') Vincent's voting for Smith would be deontically superior to his vot-ing for no one, which itself would be deontically superior to his vot-ing for Jones,

(2.7') Vincent cannot vote without pulling a certain lever,

and

(2.8') Vincent's pulling the lever would be deontically inferior to his notdoing so

are all true, and where the reason that (2.8') is true is that, if Vincent wereto pull the lever, he'd vote for Jones.

A fourth response is this. We should not talk of individual acts beingobligatory; we should talk only of entire courses of action being obliga-tory. Hence it is not the case that Vincent ought to vote. Rather, either heought to drive-and-vote-and-X or he ought to not-drive-and-not-vote-and-Y (presumably we can rule out its being the case that he ought to drive-and-not-vote-and-Z), where X and Y (and Z) represent whatever theremainders are in the respective courses of action. But this is surely muchtoo revisionary a proposal. It would imply that almost all, and probablyall, of our everyday assertions concerning what we ought to do are simplyfalse (since they do not concern entire courses of action but only segmentsthereof).

A final response is this: we should revise the standard interpretation ofthe thesis that one ought to do the best one can. Once again, it is thisresponse that I endorse. The revision that I think these (and other such)problems call for is the following.7

7 What follows borrows heavily from Feldman (1986), Ch. 2.

24

2.1.2 The analysis

We should say that what one ought to do is a function of the deonticallybest course of action (and concomitant events) open to one. We shouldnote that, whenever we are in a position to make a choice as to how to act,it's not just that we can do action A or action B or action C (or some com-bination thereof), but that we can affect how the world will be. This maysound grand, even melodramatic, but it's the truth. For example, if I canbreak a certain store window or not (by throwing a brick through it, or not)— if I have a choice as to whether or not the window breaks; if it is up tome whether or not the window breaks — then it's also true that I can makethe world be such that it contains a certain broken window (or not) — I havea choice as to whether or not it contains this; whether or not it contains thisis up to me. Of course, much - most - of what happens in the world is notup to me, but that does not alter the fact that how the world as a whole willbe is (often) up to me. Another, and by now fairly well-established way ofputting this point is as follows: at any moment of choice a number of dif-ferent possible worlds are accessible to me; which of these will be the actualworld is up to me.

The situation may be very roughly illustrated as in Figure 2.1. Here theagent, 5, has a choice at time T1 between doing A and not doing A. (Alreadythis is oversimple, in that there would presumably be a number of ways ofnot doing A — by doing E, or F, or G, and so on, instead. Note that how-

Figure 2.1

25

ever not doing A is effected, it of course involves the failure to do A. Doingsomething else, E, may not be sufficient for not doing A, since it may bethat S can do both A and E; thus, if not doing A is effected by doing E, Emust be done instead of A.) If S does A at Tu he will have a choice at T2 asto whether or not to do B. If S doesn't do A at Tt, his only "choice" at T2

will be not to do B. If S does B at T2, he will have a choice at T3 whetheror not to do C, but his doing D will be a foregone conclusion. If S doesn'tdo B at T2, his not doing C will be a foregone conclusion, but he will havea choice at T3 whether or not to do D. The nodes in the figure thus repre-sent points of choice and the paths W-W6 represent different possibleworlds (each sharing the same history up to T?) accessible to S as of Tx bymeans of his choosing one way or the other. Of course, huge amounts ofinformation are left out of the picture: what happens before T?, what hap-pens after T3, what happens between T, and T3 apart from the small num-ber of actions to which reference is made. Nonetheless the picture ishelpful, I think, in bringing home two important points. First, as noted, atany moment of choice our choice is, in effect, a choice between worlds.Second, as time progresses, our options never increase but in fact grownarrower, in that with each choice we cut ourselves off from worlds whichhad hitherto been accessible to us; what is presently accessible to us alwayswas accessible to us (during our lifetime), but it may not remain so forlong.

Clearly, what has just been said is far from precise and needs furtherwork. I shall attend to that shortly. But at this point I want to lay out myinterpretation of the thesis that one ought to do the best one can; for thenthe refinements to what has just been said will, I think, be more readilyappreciated. The main gist of the interpretation is simply this: an agent5 ought to do an act A if and only if (a) 5 can do A, (b) 5 can refrainfrom doing A9 and (c) any accessible world in which S does not do Ais deontically inferior to some accessible world in which S does A. Moreprecisely:

(I) agent 5 ought, at time Tin world W, to do action A at time T ' iff8

(a) there is a world W' such that W' is accessible to S from Wat Tand Sdoes A at T ' in W'\

(b) there is a world W" such that W" is accessible to S from Wat Tand5 does not do A at T ' in W"\ and

(c) for all worlds W" such that W" is accessible to S from Wat Tand Sdoes not do A at T ' in W'\ there is a world W' such that

8 "Iff' is short for "if and only if."

26

(1) W' is accessible to 5 from Wat T,(2) 5 does A at T' in W\(3) the deontic value for S at Tof Wf is greater than the deontic value

for 5 at T of W , and(4) there is no world W"' such that

(i) W"' is accessible to S from Wat T,(ii) 5 does not do A at T' in W"\ and

(iii) the deontic value for 5 at Tof W"r is greater than the deon-tic value for S at Tof W'.9

Clauses (a) and (b) combined imply that A is optional for S, in the sense thatS can do A and am also refrain from doing it. We may call this the personalsense of "optional." Of course, these clauses do not imply that A is morallyoptional for S, in the sense that 5 may, morally, do A and may also refrainfrom doing it. (On the contrary, if S ought to do A, then it is not the casethat S may refrain from doing it, as will be made clear below.) Clause (c)doubtless appears more complex than necessary. In particular, is subclause(c4) needed? Yes, it is (or, at least, caution on a certain matter dictates it). Ishall explain why this is so in Subsection 2.2.6 below.

Perhaps it is as well to repeat here that "ought" is intended to expressoverall moral obligation and that no substantive theory is being presup-posed as to what deontic value consists in. The latter fact is important, since,as explained in Section 1.4, (I) is intended to accommodate almost all sub-stantive theories of obligation. Thus, once again, not only is (I) intended tobe compatible with consequentialist theories, but it is intended also to becompatible with a wide range of deontological theories. It is for this reasonthat the values of the worlds are said to be the values that they have^br 5 atT. The values are thus both agent- and time-relative; whether or not theyare significantly so depends on the particular substantive theory in question.If the theory is consequentialism, then the relevant value is intrinsic value.Such value is standardly understood to be both agent- and time-neutral.

9 Note that clauses (b) and (c) jointly imply clause (a), so that, strictly speaking, the latteris redundant. Nonetheless I include clause (a) in order to emphasize the fact that oneought to do the best one can.

Also, it may be that S should be said to be not just an agent but a moral agent (what-ever exactly that amounts to), for I can imagine that some people might want to say thefollowing: some (nonhuman) animals in fact have a choice between various alternativeactions, and these actions are in fact such that some are better (even deontically) than oth-ers, but it is false that such animals morally ought to choose the best alternatives, becausesuch animals are not moral agents. (See, e.g., Thomson, 1990, p. 215.) There may besomething to this; I shall not commit myself one way or the other here.

27

Here, then, the relativization of the values o£W', W", and W'" is insignif-icant and can be ignored, insofar as the values for S at T of these worlds arethe values that they have for any possible agents at any possible times. Buton other theories the relativization may be significant.

Consider Kant's theory, for example. It might at first be thought thatsuch a theory resists accommodation by (I), but this is in fact not so. If thesubstantive mark of obligation is compliance with the CategoricalImperative, then the deontically best world open to an agent is simply onein which such compliance is maximized. This statement is perfectly legiti-mate and should not be thought to be at odds with what a Kantian wouldwant to say. To see this clearly, consider the (oversimplified) situationdepicted in Figure 2.2. Suppose that each of A, B, and C would constitutecompliance with the Categorical Imperative by the agent, whereas none of~^4, ~B, and ~C would. A "minimal" Kantianism might declare any fail-ure by the agent to comply with the Categorical Imperative categoricallywrong and thus deem W t infinitely deontically superior to any of the otherworlds, so that (as we might put it) it has a deontic value of 1 whereas allthe others have a deontic value of 0.10 A somewhat richer Kantianism mightdeclare every noncompliance as deontically bad as any other but allow forthem nonetheless to "add up," in which case the deontic ranking would bethis: first, Wt (no noncompliances); second, W2, W'3, and W5 (one non-compliance each); third, W4, W6, and W7 (two noncompliances each); andlast, W8 (three noncompliances). A still richer Kantianism would declare

Figure 2.2

10 Cf. Vallentyne (1992), p. 120, including n. 10.

28

some noncompliances deontically worse than others, and this would allowfor a still finer deontic ranking of the worlds. This way of picturing mattersis also compatible with saying (as I think almost all Kantians would) that itcan sometimes happen that a person can avoid someone else's noncompli-ance only by way of his own noncompliance. Think, for instance, ofBernard Williams's celebrated case of Jim and the Indians: Jim cannot getthe captain not to kill the Indians except by killing one himself.11 A Kantianwould say: either Jim will fail to comply with the Categorical Imperativeor the captain will. In such a case, a Kantian would presumably say that itis right for one to comply and wrong for one not to comply (even at thecost of the other's noncompliance). This means that the relativization ofthe values of worlds to agents is significant, in that S's noncompliance has aspecial disvalue relative to S that it does not have relative to someone else.

This point perhaps requires elaboration. I am envisaging the Kantian assaying the following. Jim's killing the Indian would constitute a violating-by-Jim of the Categorical Imperative, and hence it would be morallywrong for Jim to do it. Admittedly, Jim's not killing the Indian would guar-antee (in the sense that Jim could not avoid them, even if the captain could)several violatings-by-the-captain of the Categorical Imperative, and thuswould guarantee that (in some sense) more serious wrong is done, but itwould not itself be a wrongdoing by Jim. It is for this reason that the agent-relative valuation of worlds is significant. In some nonrelative sense wemight well want to say that some Jim-accessible world in which Jim killsthe Indian is better than any such world in which he doesn't, since theformer involves only one violation of the Categorical Imperative whereasthe latter involves several. But for the Kantian such a valuation of worlds,while perhaps accurate and informative for certain purposes, is not the rel-evant valuation for determining what it is obligatory for Jim to do. Whenit comes to determining Jim's obligations, some Jim-accessible world inwhich Jim does not kill the Indian is better than any such world in whichhe does.12

Thus all, or almost all, Kantians will say that deontic value is significantlyagent-relative. It may be that they would also say that it is significantly time-relative; this will depend on whether they accept that it is possible that aperson can avoid his own noncompliance with the Categorical Imperativeonly by way of some other noncompliance by him. Suppose, for example,that Joe can at T1 do A at T2 and can at T1 do B at T3 but cannot at Tt do

11 Williams (1973), p. 98.12 See Scheffler (1984) for an interesting internal critique of this view.

29

neither. Suppose also (if this is possible) that A would constitute a minorinjustice and B a major injustice, and that, for this reason, each would con-stitute a violation of the Categorical Imperative. I take it that the Kantianwho accepts that this is possible would say that Joe may not do A, even ifonly thereby can he avoid the greater injustice, B. This means that anyaccessible world containing his not doing B at T3 is deontically worse, rela-tive to Joe at Tj, than some accessible world containing his not doing A atT2; and this is so even if, by some other measure, every ~^4-world accessi-ble to Joe at Tj is worse than some ~B-world accessible to him then. In sucha case, then, the time-relativization of deontic value is significant; agent-relativization alone would not suffice to yield the deontic inferiority ofevery ~B-world to some ~^4-world, since the actions in question are bothJoe's.

This last case also serves to reinforce the need to talk in terms of the val-uation of worlds rather than that of acts. An art-based theory might wellimply that Joe ought at Tt to avoid A at T2, since avoiding A would be bet-ter than doing A, and that he ought at Tt to avoid B at T3, since avoidingB would be better than doing B. But, as with the case of Vincent present-ed in the last subsection, this would be problematic, since Joe cannot at Tt

both avoid A at T2 and avoid B at T3. The world-based theory developed inthis chapter implies that it cannot be both that Joe ought at Tt to avoid A atT2 and that he ought at Tt to avoid B at T3, since {ex hypothesi) he cannotat Tj avoid both.

Thus Kantianism appears perfectly compatible with (I). So too do allsorts of other deontological theories. Still, at this point, certain limitationsto the analysis must be acknowledged. First, it is at best only a partial analy-sis. I have said that the value at issue in (I) is deontic value; I have not triedto give an account of this notion in turn. Nor shall I. Clearly, deontic valueis a type of morally relevant value, but there are other types of morally rele-vant value.13 A fuller analysis would give an account both of how morallyrelevant value is to be distinguished from morally irrelevant value and ofhow deontic value in particular is to be distinguished from other types ofmorally relevant value. I have no such account to offer. Second, it must berecognized that (I) does not accommodate all substantive theories of oblig-ation. After all, consequentialism has traditionally been put in terms simi-lar to those given at the outset of this chapter, and it was argued that such

13 What I have in mind here is the possibility that two worlds be such that the first is supe-rior to the second with respect to deontic value, whereas the second is superior to thefirst with respect to some other morally relevant value - for example, with respect to theincidence of virtue and vice.

30

terms are unacceptable, involving a misunderstanding of the very conceptof moral obligation. Perhaps, too, a similar problem besets Kant's empha-sis on the universalizability of maxims, as this is traditionally conceived. Atany rate, any substantive theory based on such a misunderstanding will ofcourse not be compatible with (I). Nonetheless, (I) is surely compatiblewith the basic view underlying traditional consequentialism (namely, thatobligation is some rather straightforward function of intrinsic value), and Isee no reason to think that the basic view underlying Kantianism and otherdeontological theories should not be thought to be similarly compatible.Still, there is this limitation. Consider the sort of theory that says that oneought to do what one believes to be best — in some less than fully liberal senseof "best." This can be accommodated by (I) only if the theory is under-stood to concern what, among those actions the agent can in fact perform,is believed by the agent to be best; (I) does not accommodate a theory thatis concerned with what, among those actions the agent believes he canperform, is believed by the agent to be best, unless the agent's beliefsconcerning his options are accurate.

2.1.3 Extension of the analysis

The thesis that moral obligation is to be understood in terms of the com-parative values of possible worlds is by no means new. Jaakko Hintikka,among others, has made such a proposal; and Holly Goldman, MichaelMcKinsey, and, especially, Fred Feldman have provided accounts of oblig-ation that are very closely related to the account given in (I). Each is some-what different, though, and, I believe, in some way inferior.14

For a further illustration of the application of (I), consider Figure 2.3.Let the numbers 1—3 on the right denote the value-rankings of the respec-tive worlds.15 (Worlds la and lb are thus of equal deontic value, relative tothe agent at Tv) Then, given (I), we may say the following: the agent, 5,ought (at Tt) to do A (at Tt) and also ought (at Tt) to do C (at T3). Ought(at Tj) he to do B (at T2)? No, for clause (c) of (I) is not satisfied (in that,although there is a top-ranked world containing B, there is also a top-ranked world containing ~B).

14 See Hintikka (1970), pp. 70-1; Goldman (1978), p. 202; McKinsey (1975), p. 385; andFeldman (1986), p. 38. There is no space here to make a detailed comparison betweeneach of these accounts and the account that I have to offer.

15 (I) does not presuppose that there is in every case an uppermost limit to the values ofaccessible worlds, but it is easier, for purposes of illustration, to suppose this.

31

-C&D

Figure 2.3

Figure 2.3 clearly contains additional information, though. To help usexpress it, let us develop the account further.

First of all, we can say that S may do A if and only if (a) S can do A,(b) S can refrain from doing A, and (c) any accessible world in which S doesnot do A is not deontically superior to some accessible world in which 5does A. (The term "may" thus allows for ties in what is deontically best,whereas the term "ought" does not.) More precisely:

(II) S may, at T in W, do A at T ' iff(a) there is a world W' such that W' is accessible to 5 from Wat Tand

S does ,4 at T ' in W'\(b) there is a world W " such that W " is accessible to 5 from Wat Tand

S does not do A at T" in W "; and(c) for all worlds W " such that W " is accessible to 5 from Wax. Tand 5

does not do A at T' in H/ ", there is a world PF' such that(1) W' is accessible to S from Wat T,(2) 5 does A at T ' in W' , and(3) the deontic value for S at To£W" is no greater than the deontic

value for S at T of W\

(The difference between (I) and (II) consists, first, in the difference in theirrespective subclauses (c3) and, second, in the fact that (II) lacks a corre-sponding subclause (c4). As to the former point: this concerns the issue, justnoted, of possible ties in maximal deontic value. As to the latter: I shall have

32

something to say about this in Subsection 2.2.6 below.) On Figure 2.3,then, not only ought S to do each of A and C, he may do each of these aswell. Moreover, he may do B, too, although it's not the case that he oughtto do it; for he may also do ~J3, that is, fail to do, or refrain from doing, B.(B is, in other words, not just personally but morally optional.)

Further:

(III) S ought, at Tin W, not to do A at T'[i.e., S may not, at Tin W, do A at T'] iff[as in (I), except that "does A' and "does not do y4" are reversed]

Thus, on Figure 2.3, S ought not to do D; also, he ought not to do ~A andhe ought not to do ~C.

Further:

(IV) S may, at Tin W, not do A at T' iff[as in (II), except that "does A" and "does not do A" are reversed]

Thus, on Figure 2.3, S may (at Tj) not do B (at T2). Of course he also may,because he ought, not do each of the following: D, ~A, and ~C.

It will already have become apparent that English terminology posescertain problems in this context. "May not" is ambiguous, as is illustratedin the statements of (III) and (IV); sometimes it is used to express the oblig-ation not to do something, sometimes merely the permission not to dosomething. Further problems of this sort lie ahead. To help us deal withthem, some symbolism will prove useful. Using O for "is obligated" (or"ought") and P for "is permitted" (or "may"), we may say that (I) gives anaccount of what is expressed by O(A); (II) gives an account of what isexpressed by P(A); (III) gives an account of what is expressed by O(~y4);16

and (IV) gives an account of what is expressed by P(~A).Conspicuous for their absence from the foregoing are the terms "right"

and "wrong." How do they fit in? Fairly easily, I think.17 What it is wrongfor one to do is that which one ought not to do; in symbols: W{A) iff

16 It is perhaps worth reemphasizing that ~A expresses the failure to do A, not merely thedoing of something other than A (which may be compatible with the doing of A). ThusO(~A) expresses the obligation to refrain from doing A, not merely the obligation to dosomething other than A.

17 Here I am concerned exclusively with the adjectival form of these terms. The nominalforms of "right" and "wrong" (as in "Rachel has a right to be left alone" and "You havedone Rachel a wrong in not leaving her alone") and the verbal form of "wrong" (as in"You have wronged Rachel") are pertinent to moral theory but are not at issue here,although they will be discussed in Section 5.5.

33

O(~A). What it is right for one to do is not quite so clear. Sometimes peo-ple talk of "the right" thing to do under the circumstances; here what ismeant, I think, is that which one ought to do. Sometimes people talk of whatit is "all right" to do; here what is meant, I think, is that which one may do.Sometimes, perhaps, what people mean by "right" occupies some middleground between what is expressed by O(A) and what is expressed by P(A);if so, I'm not sure what precisely is meant, and I shall therefore not seek toaccommodate it explicitly in the present account.

The account may be extended further. First:

(V) O(A&B):S ought, at T in W, to do both A at T" and B at T " iff(a) there is a world Wf such that W' is accessible to S from Wat Tand

S does A at T" in W' and S does B at T " in W'\(b) there is a world W" such that W" is accessible to S from Wat Tand

5 does not both do A at T ' in W" and do B at T " in W"\ and(c) for all worlds W" such that W" is accessible to S from Wat Tand S

does not both do A at V in W" and do B at T " in W", there is aworld W' such that(1) W' is accessible to S from Wat T,(2) S does A at T ' in W' and 5 does B at T " in W\(3) the deontic value for S at Tof W' is greater than the deontic value

fo rSa tTof W", and(4) there is no world W'" such that

(i) W'" is accessible to S from Wat T,(ii) S does not both do A at T ' in W"" and do B at T" in W " ,

and(iii) the deontic value for S at T of W'" is greater than the deon-

tic value for 5 at Tof W'.

That is, roughly: S ought to do both A and B if and only if (a) S can do themboth, (b) S can refrain from doing them both (whether by doing neither ofthem or by doing one but not the other), and (c) any accessible world inwhich S does not do them both is deontically inferior to some accessibleworld in which S does them both. (Clearly this can be extended even fur-ther to accommodate not just two actions but any number of actions.) OnFigure 2.3, then, not only ought S to do each of A and B and C (that is, notonly is each of these such that he ought to do it); he ought to do all of them(that is, he ought to do (A&B&C)). It might be asked what the differencereally is between saying, for example, that O(A) and O(B) and saying that

34

O(A&B). Well, there is a difference, even though, on my account, the twoare close to being strictly equivalent; that is, it is almost true that (O(A) &O(B)) iffO(A&B).18 But this near-equivalence may mask the difference, soconsider first of all this:

(VI) P(A&B):S may, at T in W, do both A at T ' and B at T " iff[as in (II), with changes as from (I) to (V)]

On Figure 2.3, S may of course do (A&B&C); and it is also the case that hemay do A, he may do B, and he may do C. But notice, as mentioned ear-lier, that he may also do ~B. Clearly, though, it's not the case that he maydo (B&~B) (because he cannot). Thus it is not in general the case that P(A)& P(B) iff P(A&B). This inequivalence obviously marks a difference, andI hope now that it will be clear how it is that there is a difference betweensaying that O(A) & O(B) and saying that O(A&B), even though they arenearly equivalent.

We may of course also say the following:

(VII) O(~(A&B)) [W(A&B)]iS ought, at T in W, not to do both A at T ' and B at T "[i.e., S may not, at T in W, do both A at T ' and B at T"][i.e., it is wrong, at Tin W, for S to do both A at T ' and B at T"]iff[as in (III), with changes as from (I) to (V)]

(VIII) P(~(A&B)):S may, at T in W, not do both A at T ' and B at T " iff[as in (IV), with changes as from (I) to (V)]

Thus, on Figure 2.3, S ought not to do both ~B and D; this is because heought not to do D (although it remains true that he may do ~B). Also, Smay not do both A and B (in the sense of (VIII); he is permitted not to dothem both); this is because he may do (A&~B) (although it remains truethat he ought to do A).

A final set of four statements concerns the tricky connective "or." As faras I can tell, the following best captures prevailing English usage:

(IX) P(AorB):S may, at T in W, do either A at T ' or B at T " iff(a) S may, at Tin W, do A at T ' ; and(b) S may, at Tin W, do B at T" ;

18 See the discussion of (2.40') in Subsection 2.3.1 below.

35

(X) O(AorB):S ought, at Tin W, to do either A at T' or B at T " iff(a) S may, at Tin W, do either A at T ' or B at T" ; and(b) 5 ought, at Tin W, not to do both -A at T ' and ~B at T "

[i.e., it is wrong, at T i n W, for S to do neither A at T ' nor B

at T"] ;(XI) O(AnorB):

S ought, at Tin FF, to do neither A at T ' nor B at T "[i.e., S may not, at Tin W, do either 4 at T ' or B at T"]19 iff(a) S ought, at Tin W, not to do A at T '

[i.e., it is wrong, at Tin W, for S to do A at T'] ; and(b) 5 ought, at Tin W, not to do B at T "

[i.e. it is wrong, at Tin W, for S to do B at T"] ;(XII) P(AnorB):

S may, at Tin W, do neither 4 at T ' nor B at T " iff S may, at TinW, both not do A at T ' and not do B at T".20

Thus, on Figure 2.3, we may say each of the following (among otherthings): S may do either A or B; S ought to do either (A&B) or (^&~B>); 5ought to do neither ~ C nor D; and S may (in the senses of both (XI) and(XII)) do neither ~ C nor D.

Some other points concerning English usage should be made here. First,it seems especially important to point out that the statements "S may doeither A or B" and "S ought to do either A or B" are to be standardly under-stood as interpreted in (IX) and (X), respectively, and not, as some logicianswould have it, in terms of mere disjunction.21 For this reason I have used"or" and "nor" as part of the symbolism rather than resort to the usual log-ical symbol for disjunction, v. The point here is this. For any propositionsp and q, \ip is true, then so is p or q (p V q). Thus, where p and q concernactions (as where p is the proposition that S does A and q is the propositionthat 5 does B), then, if S does A, it is also true that either 5 does A or S doesB. Some have gone on to infer from this that, if S does A, then S does eitherA or B. Now, it may be that there is a sense of "or" that permits this infer-ence, but, even if this is so, it seems to me quite contrary to normal usage

19 Another possible English equivalent: 5 may, at T in W, do neither A at T' nor B at T".But what this phrase means here is to be sharply distinguished from what it means in (XII)below. In such cases, the symbolism that has been provided is particularly helpful.

20 That is, P(AnorB) is equivalent to P(~A&~B) and not simply to P(~Aor~B).21 See, e.g., Castafieda (1981), p. 64.

36

to go on to say that, if S may (or ought to) do A, then S may (or ought to) doeither A or B.

Second, frequently, where P(AorB) or O(AorB) is true, so too is ~O(A)& ~O(B). But this is not necessarily so. Similarly, it is frequently, but notnecessarily, also the case that 5 cannot do both A and B.

Third, (X) can be used to capture what is at issue in such common claimsas "Robin ought to return the book by next Wednesday," "Robin oughtto return the book between Tuesday and Wednesday," and so on. Let A bethe act of returning the book and Tx... Tn be a (perhaps discontinuous) seriesof times, with T7 being the time of "ought" and Tn being Wednesday.Then what's at issue in such examples is this: Robin ought at T} to do eitherA at T2 or A at T3 or.. .or A at Tn. Similarly, (V) can be used to capture whatis at issue in such common claims as "Marsha ought to maintain silencethroughout (or during) the exam." Let A be the act of maintaining silence andT2... Tn be a series of times, with T2 being the beginning of the exam andTn being the end. Then what's at issue in this example is this: Marsha oughtat T1 both to do A at T2 and.. .and to do A at Tn.

Fourth, the English term "ought" is peculiar with respect to tense. First,it has no future tense. But this is of no philosophical significance; one cansimply use "will be obligated" to express future obligations.22 Second,"ought" appears to have no past tense.23 But, again, this is of no philo-sophical significance; one can use "was obligated" or "ought to have" toexpress past obligations.24 At least one philosopher has been misled by thesegrammatical peculiarities of "ought" and been moved to formulate the"Principle of the present-tenseness of ought," according to which, the verb"ought" always being in the present tense, it follows that, once S ought todo A, always thereafter 5 ought to have done A. Although this may seemcorrect (indeed is correct, given that "ought to have" expresses a past oblig-ation), what in fact is meant here is that, if 5 ought at Tt to do A at T2, then5 ought at all times subsequent to Tt to do A at T2.

25 This, I submit, is patentnonsense, and it is ruled out in (I) by clause (b); after A is done at T2, thereis no longer an accessible world in which A is not done at T2.

22 See Broad (1985), p. 225.23 I say "appears," because I'm not exactly sure how to classify grammatically the phrase

"ought to have."24 An alternative to "ought to have," perhaps more revealing but nonetheless nonstandard,

is "didn't ought," which some British speakers sometimes use to express past obligations.25 See Castafieda (1981), p. 61, where this example is given: "Yesterday is when I ought,

now, to have visited Mary." (Cf. Castafieda, 1975, Section 7.12.) Feldman is commit-ted to something like this, too, but not on the basis of a misunderstanding of what "oughtto have" means. See Feldman (1986), p. 43.

37

2.1.4 Perfect and imperfect duties

Sometimes a distinction is drawn, following Kant, between "perfectduties" and "imperfect duties." Kant's famous four examples of duty con-cern the duty not to commit suicide, the duty not to deceive others byvirtue of making a false promise, the duty to develop one's talents, and theduty to act charitably.26 The first two are said to be perfect duties (the firstto oneself, the second to others) and the last two are said to be imperfectduties (the first to oneself, the second to others). Just what the distinctionis supposed to consist in is controversial. In a note Kant says that a perfectduty is one that admits of no exception in the interests of inclination,whereas an imperfect duty admits of such exception, but just what thismeans is unclear.

The matter seems to me very complex. First, it is worth remarking thatit is natural to express the perfect duties negatively (as the duties not to dosuch-and-such), as I have done, while the imperfect duties are naturallyexpressed positively. Is this significant? It's hard to tell. Certainly, somehave thought that there is a significance to the distinction between positiveand negative duties, but it's not easy to see why one should think so.Indeed, it might seem that perfect duties could be expressed positively (inthis case, as the duties to preserve one's life and to behave honestly towardsothers) and imperfect duties negatively (in this case, as the duties not to letone's talents go to waste and not to behave meanly towards others),although it must be acknowledged that these alternative renderings mightnot quite hit the mark: isn't there, after all, a difference between (positive)action and (negative) omission or refraining?28 Whatever the truth of thematter, the present account is designed to accommodate both positiveduties (see (I)) and negative duties (see (HI)).29

Something else that Kant might have had in mind is this. A perfect dutyis fully specific, whereas an imperfect duty leaves it to the agent to deter-mine just how the duty is to be fulfilled. For example, a duty not to killoneself is just that, whereas a duty to be charitable can be satisfied either bybeing charitable towards A or by being charitable towards B, and so on.

26 Kant (1964), pp. 89-91.27 E.g., Foot says that negative duties are more stringent than positive ones, in Foot (1980),

p. 162ff.28 The literature on this topic is huge. For just a small sampling, see Brand (1971), Gorr

(1979), Moore (1979), Walton (1980), and Zimmerman (1984), Ch. 8.29 If a distinction is to be drawn between simply not doing A and refraining from doing A

and/or omitting to do A, etc., so be it. Then the formula "S ought to refrain from doingA (or omit to do A, etc.)" can be understood in terms of (I), with "refrain from doingA" being substituted for "do A."

38

This distinction, I think, is problematic. Hardly any duty is fully specific inthe sense indicated; after all, one can refrain from suicide by doing one thing(e.g., by going to the movies instead), or by doing another, and so on.30 Atany rate, the present account is adequate to the expression of what is at issuehere. If S ought not to commit suicide, then every accessible world inwhich 5 commits suicide is one that is deontically inferior to some worldin which he does not. If 5 ought to act charitably, then every accessibleworld in which S fails to act charitably is deontically inferior to some worldin which he acts charitably. If there is more than one top-ranked accessibleworld in which S acts charitably, then S may enjoy a certain sort of latitudewith respect to how to satisfy the duty to act charitably. Perhaps in one suchworld he is charitable towards A but not towards B, in another towards Bbut not towards A, and so on. In this case S ought either to act charitablytowards A or to act charitably towards B (or...) (see (X)), but it's not thecase that he ought to act charitably towards A and it's not the case that heought to act charitably towards B. Thus it is not (morally) up to S whetherto act charitably, but how to do so is up to him.

Another thing that Kant might have had in mind is this: a perfect dutyis a duty-to (with a correlative right), but an imperfect duty is merely aduty.31 The present account is not designed to accommodate this distinc-tion explicitly. As noted in Section 1.3,1 seek here only to account for whatI have called "mere obligation," that is, for what it is to have an overall duty(or obligation) to do something; how to supplement this with an accountof what it is to have a duty-to (or obligation-to) is something that I shallleave aside.

Sometimes it is said that what Kant had in mind is this: a perfect duty isa duty to perform some act (or to refrain from doing so), whereas an imper-fect duty is a duty to adopt a certain maxim for action.32 It may seem thatthe present account cannot accommodate this, since it has been put in termsof " 5 ought to do ^4," etc. I shall address this point in the next subsection.

Finally, it might be thought that what Kant had in mind is this: a perfectduty ought always to be satisfied, but an imperfect duty only ought some-times to be satisfied.33 What is meant by this? Presumably the following. Tosay that S has a perfect duty at Ti to do A is to say, on this understanding,that

(2.11) for all times T (at which 5 exists) not earlier than Tu S ought at Tt

to do A at T.

30 Cf. Chisholm (1968), p. 416, n. 6.31 This is how Mill understands the distinction in Mill (1957), Ch. 5.32 Cf. Hill (1971), p. 57.33 Cf. Hill (1971), pp. 56-7.

39

To say that S has an imperfect duty at T1 to do A would then be to say that

(2.12) for some but not all times T (at which S exists) not earlier than Tu

S ought at Tt to do A at T.

(This way of putting things presupposes that S can do A on more than oneoccasion - that is, that A is "repeatable." This view of action is controver-sial; I shall address it in Subsection 2.2.5 below.) Of course, Kant presum-ably meant that the duties of which he spoke are duties had by everyone (orat least by everyone capable of having duties of the sort in question). In thissense, a perfect duty to do A would be understood to consist in the truthof the following:

(2.13) for all agents S and times T' and T (such that Tis not earlier thanT' and S exists at these times), 5 ought at T' to do A at T.

Indeed, a so-called absolutist such as Kant34 might mean that a perfect dutyis one which ought always to be satisfied by everyone "no matter what."This phrase is unclear but presumably expresses some sort of necessity.Thus, on this understanding, a perfect duty to do A consists in the truth ofthe following:

(2.14) necessarily, for all agents 5 and times T' and T (such that Tis notearlier than T" and S exists at these times), 5 ought at T' to do A atT.

The present account shows how these statements are to be understood.

2.2 UNDERLYING METAPHYSICAL ANDCONCEPTUAL ISSUES

2.2.1 "Can" and control

According to the analysis of obligation just presented, S ought to do A onlyif S can do A and 5 can refrain from doing A. To be obligated to do A, then,S must be in control of whether or not he does A; it must be personallyoptional for S. That "ought" implies "can" is of course controversial; I shalladdress the matter in Section 3.1. Here I want to attend to a preliminaryquestion. The present account may seem to restrict itself to obligationsconcerning actions ("5 ought to do A"; "5 can do A"). Is such restrictionwarranted?

34 Or Donagan (1977) or Fried (1978).

40

Strictly, it is not. The sort of control, or freedom, expressed by "can" inthis context may be either direct or indirect. It is controversial just whatsort or sorts of item are such that an agent may have direct control overthem, but it is clear that there must be some such sort or sorts for there tobe any control at all. Consider the example given earlier: that of breakinga certain store window by throwing a certain brick through it. Let us saythat this is both something that I can do on a particular occasion and some-thing that I can refrain from doing. I thus have control over whether or notthe window breaks. This must surely be only indirect control, though; Ican achieve the window's being broken only by way of achieving some-thing else. What is this something else? Moving my arm, for one thing; andreleasing the brick; and so on. On my view (although the truth of this claimis not essential to the present account of obligation), the first item in this list,that is, that item over which I have direct control, is a choice or decision tothrow the brick (or some suitably related decision — one that will result inthe window's being broken). I have direct control over the decision andthereby indirect control over its consequences. These consequences includemy arm's moving, the brick's being released, the brick's flying through theair, the brick's hitting the window, the window's breaking, the alarm'ssounding, and so on.

It is worth noting that none of the items on the list given in the last twosentences is an action - unless a decision is itself an action, which I supposewe may grant in a minimal sense of "action."35 Yet I have control over allof them. Where do nonminimal actions themselves fit in? As follows, Ibelieve.36 The action that is my moving my arm just is (roughly) my decid-ing to do something relevant and this decision's causing my arm's moving.Thus I have a sort of "hybrid" control over this action: I have direct con-trol over the decision involved and (thereby) indirect control over its con-sequence, and my action just is this decision plus this consequence.Similarly, my releasing the brick is this same decision coupled with the con-sequence of the brick's being released; my breaking the window is this samedecision coupled with the consequence of the window's breaking; and soon; and I have a sort of hybrid control over all these actions.

This applies to omissions, too. Perhaps I have made a promise to stayindoors and water the plants; when I go outside and throw bricks, I omitto water the plants. I have control over this omission. How? Once again, Ibelieve, my control can be traced to a decision that I make, a decision whose

35 Some contend that only decisions are actions (see, e.g., Prichard, 1949, p. 190, and Davis,1979, p. 41). This is surely too restrictive.

36 See Zimmerman (1984), Ch. 4, and (1988a), Ch. 2.

41

consequences are incompatible with my remaining indoors and wateringthe plants. My not remaining indoors and watering the plants may thus besaid to be a consequence37 of this decision.

It may be that certain items over which we have control don't fit easilyinto this decision-plus-consequence picture. I am thinking primarily ofcertain mental phenomena, such as desires, motives, intentions, and othersuch attitudes, as I shall call them. (Obviously, there are important distinc-tions to be drawn among these phenomena, but doing so is not necessaryfor my purposes here.) Sometimes, it is clear, the adoption of an attitude isthe result of acting in a certain way (I may have had to school myself, atlength, in order to gain my newly acquired tolerance towards others whoare in certain ways different from myself; I may have had to work hard torid myself of my desire for cigarettes; and so on), and this fits in well enoughwith the picture in question. But it may seem that at other times I can adopta certain attitude directly, that is, without doing so by way of deciding oracting in a certain way. Perhaps, when Kant urges the adoption of a certainmaxim, he means (in part) that we ought to adopt a certain attitude - anattitude of preparedness to act in certain ways. (I shall say something simi-lar concerning cooperation in Chapter 9.) And perhaps it is possible to dothis directly. I suspect not, but I'm not sure. At any rate, to repeat, it is notessential for present purposes that the foregoing decision-plus-conse-quence picture be accepted. What is essential is only that the distinctionbetween direct and indirect control be accepted, and this distinction seemsto me unassailable. Surely one cannot have control over anything withouthaving direct control over something; and surely not everything one hascontrol over is such that one has direct control over it. Moreover, it is sure-ly plausible to think of some things (on my account, nonminimal actions)being such that one's control over them is a hybrid of direct and indirectcontrol.

We thus have three categories of control: direct, indirect, and hybrid.And so we have three candidates for what is obligatory: those items overwhich we have direct control, those items over which we have indirectcontrol, and those items over which we have hybrid control. I see no rea-son in principle to reject any of these candidates.38 That is, it seems to meperfectly proper to say, for example (and now I'll resort to the account ofaction just given), that it is obligatory for me that I decide to stay indoors

37 Of a certain sort. See Zimmerman (1984), pp. 58-9, and (1988a), pp. 23-4, 30, and 93.38 As Feldman (1993) demonstrates, on that type of account of obligation that both he and

I advocate, no inconsistency arises if the class of items that can be obligatory is claimedto be wider than the class of actions.

42

and water the plants; it is obligatory for me that I stay indoors and that Iwater the plants; and it is obligatory for me that I be indoors and that theplants receive water. Similarly, it is obligatory for me that I not decide tobreak a certain window; it is obligatory for me that I not break the win-dow; and it is obligatory for me that the window not break. Part of the rea-son why all of these things are obligatory for me is that I have control overthem, although only some of these things are actions — in the minimal sense,those over which I have direct control, and, in the nonminimal sense, thoseover which I have hybrid control; those over which I have only indirectcontrol are not actions in any normal sense of the term.

Nonetheless, I shall continue to talk in the abstract in such terms as "Sought to do A," "S can do A" and so on, where A ranges over things wedo. The reasons for this are first that doing so simplifies exposition, and sec-ond that, even though we cannot properly be said to "do" the conse-quences of our decisions, they are nonetheless something that we bringabout — and it cannot ever be the case that a certain consequence be oblig-atory for S without its being the case that it is obligatory for S to bring itabout, that is, so to act (whether nonminimally or only minimally) that theconsequence occurs.

At this point, though, some would demur. In particular, H. A. Prichardand W. D. Ross would claim that only those items (in their view, what theycall "volitions" or "self-exertions") over which an agent has direct controlare such that they may be said to be obligatory. Prichard and Ross at timestalk of actions as being obligatory, but they claim that, in the strict sense,only volitions are actions; items that include the consequences of volitions(as, on my view, watering the plants would include the plants' receivingwater) are not, strictly, actions39 and not, strictly, obligatory.40 The reasonthat they say that only volitions can be obligatory is apparently that theybelieve that only volitions are, strictly, in an agent's control.41 But this can-not be right; after all, the window's being broken (or not) is in my control,and this isn't a volition. Of course, this isn't in my direct control, and I thinkthat what Prichard and Ross have in mind is that anything that is only in anagent's indirect control is not fully in the agent's control. For example, afterI make the decision to throw the brick, what happens is not up to me butrather, as it is sometimes put, "up to nature."42 And nature may not coop-erate. I may decide to throw the brick, but I may be stricken by a sudden

39 Prichard (1949), pp. 20, 190.40 Prichard (1949), pp. 34-5.41 Ross (1939), p. 160.42 See Davidson (1980), p. 59.

43

paralysis, or the brick may be heavier than I thought, or the brick may slip,or my aim may be off, or I may have miscalculated the distance 'twixt handand window, or the window may be made of plexiglass, and so on, in whichcase the window will not break despite my decision to throw the brick; andwhether or not any of these things is so is not (at the present time) in mycontrol.

Now, it is surely correct to say that whether or not the window breaksis, in some sense, not just up to me but "up to nature." There is a sense,then, in which the window's breaking is not fully in my control; whetheror not it breaks is, to some extent, a matter of luck. But so what? Under thecircumstances (which include the fact that nature is cooperative), whether ornot the window breaks is up to me. Why not, then, say that I ought not tobreak it? Prichard and Ross appear to think that what's obligatory must befully in one's control, and they seem to think that that and only that whichis in one's direct control is fully in one's control. But this is a mistake.43

Whether or not I exert myself to do something is itself to some extent amatter of luck, just as whether or not my exertions bear fruit is to someextent a matter of luck; for such self-exertion requires the cooperation ofall sorts of things over which I lack control (e.g., that I not be distracted,that I not suffer a heart attack, that I be in a position to exert myself, and soon).44 If'full control were required for obligation, nothing would be obliga-tory; but some things are obligatory, and so full control is not required.Some measure of control is still required, however; but both direct andindirect control constitute some measure of control, and so I see no needto say that, strictly, only those items over which we have direct control areobligatory.45

The obvious fact that our control is limited (not simply in the sense thatwe are never fully in control of anything, but in the sense that some thingsare such that we are not in control of them at all) means that we may dividethe future into two portions. Castaneda calls these the "Future Framework"

43 Broad makes this mistake too. See Broad (1985), pp. 133-4.44 See Zimmerman (1988a), Section 4.11.45 The same holds true of moral responsibility. One may be morally responsible (e.g., to

blame) for those things over which one has (had) direct control and for those things overwhich one has (had) only indirect control. Still, there is an important difference here.Only those things over which one has (had) direct control are relevant to the extent ordegree to which one is to blame — or so I argue in Zimmerman (1988a), p. 54ff. andSections 4.3 and 4.11. There is no analogous claim to be made about obligation, for itdoes not come in degrees in this way.

For more on inadequacies in the position of Prichard and Ross, see McConnell(1989).

44

and the "Future Zone of Indeterminacy."46 That which lies in the FutureFramework is that over which we have no control; it is fixed; it is featuredin every world accessible to us. That which lies in the Future Zone ofIndeterminacy is that over which we have (some measure of) control; it isnot fixed; it is featured in some but not every world accessible to us. Ofcourse, what's fixed for one person may or may not be fixed for another.It is fixed for each of us that the sun will rise tomorrow; but although it is(probably) fixed for me what Bill Clinton will do tomorrow, it is not fixedfor him what he will do tomorrow.47 This is an important point: what oth-ers do may not be fixed for them, but it may be fixed for me, from whichit would follow that what worlds are accessible to me is (in part) up to them.(This is crucial to the question of cooperation, as we shall see in Chapter9.) Of course, sometimes what others do may not be fixed for me; I maybe able to influence their actions. In general, then, the actions of others maybe a part of my Future Framework or be a part of my Future Zone ofIndeterminacy, just as many other things may be (such as a certain win-dow's being broken).48

2.2.2 "Can" and accessibility

What exactly is the relation between "can" and accessibility? So far, I havehardly mentioned the latter when discussing the former, and yet it is interms of the latter that the formal account of obligation has been given.

In one respect, the relation is straightforward. We can say:

(2.15) S can at Tdo A at T' in W iff there is a world accessible to S fromWat Tin which S does A at T'.

Or more generally (to cover those items over which one may have controlbut which aren't strictly actions):

(2.16) it is personally possible for S at Tin Wthatp be true iff there is aworld accessible to S from Wat Tin whichp is true.

(Here p may itself be an entire world, that is, a possible proposition thatentails every proposition or its negation.) Although true, however, thesestatements fail to reveal certain underlying complexities.

The concept at issue — that which I have called "personal possibility" —is a very familiar one. It is crucial to a number of philosophical debates,

46 Castaneda (1975), p. 135.47 Barring, of course, the truth of fatalism or hard determinism.48 See Feldman (1986), p. 22; compare Lyons (1965), pp. 75, 82-3.

45

including the debate concerning freedom and determinism (which I shallnot discuss) and the debate concerning whether "ought" implies "can"(which I shall discuss in Section 3.1). "Can" can, of course, be used toexpress a number of different concepts of possibility, and so it would bedesirable if some further account of the particular concept of personal pos-sibility were given here. Unfortunately, I haven't much to offer. I'vealready indicated that, when it is personally possible for S to do A and it isalso personally possible for 5 to refrain from doing A, then S is in control ofhis doing A. Still, doubtless a variety of concepts of control might be dis-tinguished. Perhaps an illustration will help. There is a sense in which CarlLewis can run 100 meters in under 11 seconds but in which I cannot. Butit would be surprising if, in the sense of personal possibility, Lewis can dothis right now; for that to be true, there must be no obstacle in the way ofhis doing so right now. Still, it is true, even in the sense of personal possi-bility, that he can do so later (or, more accurately, that he can now do solater); for, I assume, he can, right now, so act that he will later, right then,run 100 meters in under 11 seconds. In other words, although there is noworld now accessible to me in which at some time (now or later) I run 100meters in under 11 seconds, there is a world now accessible to Carl Lewisin which he does so.

But these are just other words. I don't mean, by virtue of (2.15) and(2.16), to define personal possibility in terms of accessibility. Althoughsome come close to advocating this,491 do not. If anything, the notion ofaccessibility seems to me a philosopher's artifice that is itself to be under-stood in terms of the notion of personal possibility. I suggest the following,which is explicitly based on the view (already acknowledged to be contro-versial, and certainly not to be defended by me here) that direct controlconcerns decisions.

To say that someone can (in the sense of personal possibility) do some-thing is to say this: either he can directly do it, or he can indirectly do it.And to say that he can indirectly do it is, roughly, to say this: there is some-thing else that he can directly do such that, if he were to do that, then he'dbe one step closer to doing this. That is, taking the first step would enablehim directly to take a second step, and doing this would enable him direct-ly to take a third step, and so on, for however many steps are needed. Thisis in keeping with the branching pattern of Figures 2.1-2.3: a move by theagent from one point of choice takes him to the next, where he is able tomake another choice; a move from this takes him to another; and so on.

49 See Feldman (1986), pp. 17, 23. Certainly Feldman shrinks from defining accessibility interms of personal possibility.

46

A first stab at rendering this more precise is the following:50

(2.17) S can at T do A at T" in W iff there are times Tv..Tn anddecisions Di...Dn such that:(a) Tis identical with T?;(b) T ' is identical with Tn;(c) Tj is not later than T2 and.. .and Tn_i is not later than Tn;(d) S can directly at Tt make Dj at T? in W7;(e) if Dj is not identical with Dn, then for any D, and 7) (where i < n),

if 5 were to make D, at 7], then S could directly at Ti+1 make Di+1

at T)+J; and(f) if S were to make Dn at Tw, then S would thereby do A at Tn.

This is pretty complicated; unfortunately, it requires further complication.There are two problems with the proposal as it stands.

First, it may be that someone can now do something later without beingnow able to do anything now. After all, there are worlds accessible to meeven as I sleep; for I will wake up, and then, at some point, I will be direct-ly able to do something that will enable me to do something else, and soon. This point is, I think, easily handled by revising clause (a) so that T issaid not to be identical with Tt but simply not later than it (as long as 5 existsatT).

The second problem is trickier. Suppose that Chris can cross the road;he can do this because he can make a decision to step off the curb, and soon. But suppose that Chris is an inveterate daredevil and in fact would makethe decision to cross the road only if there were a lot of traffic about, in andout of which he could (he thinks) dart; and suppose that, if Chris were tomake this decision under those circumstances, he would in fact get run overand never reach the other side of the road. In this case, clause (e) of (2.17)is not satisfied. What we need to say, I think, is that there are certain con-ditions (that in fact obtain) such that, if Chris were to make the decision tocross under these conditions, then that would enable him to do somethingelse, and so on.

For these reasons, I think that (2.17) should be replaced by the follow-ing:

(2.18) 5 can at T d o A at T" in W iff there are times Tv..Tn9 decisionsDv..Dtt, and conditions Ct...Cn such that:(a) S exists at Tand Tis not later than T7;(b) T' is identical with Tn;

50 Compare Chisholm (1976), pp. 62—4, and Zimmerman (1984), pp. 235—6, for similarapproaches.

47

(c) Tj is not later than T2 and.. .and Tn_1 is not later than Tn;(d) iS can directly at T1 make Dt under Ct at T1 in W;(e) if Dj is not identical with Dn, then for any D(, C,, and Tj (where

i < n), if S were to make D( under C, at TJ, then S could directly atTi+1 make Di+1 under Ci+1 at Ti+1; and

(f) if 5 were to make Dn under Cn at Tn, then 5 would thereby do AatTn.

This statement concerns only actions, of course, whereas it has been notedthat other sorts of things may be under an agent's control. To accommo-date this, we may generalize (2.18) as follows:

(2.19) S can at Tso act at T' in W that p is true iff[as in (2.17) through clause (e), and](f) if S were to make Dn under Cn at Tn, then p would be true.

We may also say:

(2.20) it is personally possible for 5 at Tin W that j9 be true iff for sometime T', S can at Tso act at T' in Wthatp is true.

And, whether or not p is an entire world, we can also put this as follows: pis accessible to 5 from Wat T. Thus personal possibility just is accessibili-ty, and (2.20) declares it to be analyzable in terms of "can." Furthermore,this analysis establishes the truth of (2.15) and (2.16), once it is recognized(as noted in Subsection 2.1.2) that, where S can do A, S can so act that theworld contains his doing A, and vice versa.

I have said that nothing ever becomes accessible that wasn't alreadyaccessible. In a sense, then, whatever one can do one could always do. Butthis might seem to fly in the face of an obvious fact: one can learn to dothings one couldn't do before; one can gain an ability. But here we needonly note that personal possibility is to be relativized to times; in particu-lar, "can" has a double time index. What is directly in one's control is, tem-porally, immediately so;51 but what is indirectly in one's control is,temporally, only remotely so.52 Given this, the acquisition of a "new abili-

51 At least, this is typically so with respect to its initiation. Whether it is necessarily so isunclear to me. Perhaps even the exercise of direct control is or can be a noninstantaneousprocess, such that one can have direct control at some time over an event that begins tooccur only at a later time.

52 At least, this is typically so. Again, whether it is necessarily so is unclear to me. Perhapsit is possible that one have immediate direct control over some event E, that E have F asa consequence, and that E and F occur simultaneously, so that one's indirect control overF is not remote but immediate.

48

ty" may be understood in terms of something's becoming immediately pos-sible that was hitherto only remotely so (or, perhaps, in terms of some-thing's becoming close to immediately possible that was hitherto a long wayfrom being so). If I can now speak French now, having learned to do so, Icould always speak French now. Of course, this doesn't mean that, at allearlier times of my existence, I could then speak French then.

2.2.3 "Can" and knowledge

Sometimes "can" is linked with "know-how." I can't open the safe becauseI don't know the combination to the lock; I can't get my VCR to workbecause I don't know which buttons to push in what sequence; and so on.I think this comes very close to saying that one cannot do something unlessone can do it intentionally.

There is certainly such a sense of "can." The illustrations just given con-firm that. But it is not this sense of "can" that I'm working with here. Onthe contrary, I can open the safe, and I can operate the VCR, in the pre-sent sense of "can," and this is simply because I can (easily) perform thosephysical movements which would result in these things being done.53 Ofcourse, there is a sense in which I'd be very lucky if I managed to do thesethings; but in this sense of "luck" it is not that their accomplishment isstrictly out of my control, but rather that, among the very many things thatI can apparently do, I have no reason to choose to make this particular setof movements (that would result in the safe's being open or in the VCR'sworking) over the many other sets of movements (that would be fruitless).

So I can open the safe. Can I likewise win the lottery? Perhaps I can winit; that depends on how it works. If I would win it by inscribing a certainnumber on a card, and I can inscribe the number, then I can win it. But ifto win the lottery I have to purchase a certain ticket with the winning num-ber already imprinted, then it is highly unlikely that I can win it. It is high-ly unlikely, that is, that there is a world accessible to me in which I win thelottery; for there are only a relatively few tickets that I can purchase (giventhat my purchasing power is limited and that there are only a relatively fewstores that sell the tickets and that I can reach in time), and the likelihoodthat one of these tickets has the winning number is very small. There is adistinction to be drawn, then, between its being unlikely that I will do oneof the many things that I can do (as in the safe, VCR, and card-inscribingcases) and its being unlikely that I can do a certain thing (as in the pre-imprinted ticket case).

53 Cf. Feldman (1986), pp. 24-5.

49

Many, indeed most, of the things that I can (in the present sense) do arethings I know nothing about. Those worlds that are presently accessible tome are such that, at best, I only know a tiny fraction of their features.Perhaps I have a fair idea about several (though certainly not all) of myimmediate and close-to-immediate personal possibilities; but the moreremote the possibilities, the less I will tend to know about them. UnlessI die within a year, I can now do many things a year hence; but, exceptfor the most mundane among them, I have little idea what these thingsmay be.54

In saying that the "can" that "ought" implies has no interesting tie toknowledge, I am attempting to accommodate those substantive theories ofobligation (act utilitarianism, as traditionally understood, is one; certainrights-theories are others) that allow for the possibility that an act be oblig-atory even though the agent doesn't know that he can perform it or howto perform it. Thus it is not consistent with the present approach to insist,as some do,55 that an act can be obligatory only if the agent knows that hecan perform it or how to perform it. Still, it is consistent to make a claimthat comes close to this, namely, that unknown personal possibilities con-fer no deontic value on the accessible worlds that include them and thus donot affect the relative rankings of these worlds. If this were true, then,although some such possibilities could nonetheless be obligatory, theirobligatoriness would be wholly parasitic on the obligatoriness of knownpossibilities.

2.2.4 "Can" and control again

What precisely is the relation between "can," control, and freedom?A person must do (cannot not do) something, as of a certain time, if he

does it in every world accessible to him as of that time. A person does dosomething if he does it in the actual world. And a person can do something,as of a certain time, if he does it in some world accessible to him as of thattime. Hence, since the actual world is accessible to everyone, whatevera person must do he does do, and whatever he does do he can do. Nosurprises so far: whatever is necessary is actual, and whatever is actual ispossible.

But there is a puzzle lurking here. Sometimes, when talking of what ispersonally possible, people use the term "free to" instead of "can." "S can

54 See Feldman (1986), p. 24.55 Cf. Prichard (1949), pp. 22-5; Ross (1939), pp. 149-51; McCloskey (1973), p. 62;

Lemos (1980), p. 301.

50

do A" and "5 is free to do A" are sometimes used interchangeably. Is thisappropriate? The answer must be no, I think, at least if the following twopoints are conceded: first, that "can" is related to "muse" and "does" in theway just indicated; and second, that, if a person does what he is free to do(while he is free to do it), then he does it freely. For if one puts both pointstogether, and if one uses "can" and "free to" interchangeably, then oneobtains the result that whatever anyone must or does do he does freely. Andthis is surely unacceptable in any standard use of "does freely" — and here Imean to be invoking that common, if not very well understood, sense ofthe term that is at issue in the debate concerning freedom and determin-ism.56 And yet I declared earlier, in Subsection 2.2.1, that the "can" impliedby "ought" expresses control or freedom. How can this be so, given whathas just been said?

The fact is that what I said earlier was said too loosely, and I must nowput matters right. I have already noted (and this turns out to be crucial) thatthere are two "can"-statements implied by the fact that 5 ought to do A,namely, that S can do A and that S can refrain from doing A (i.e., there area world accessible to S in which S does A and a world accessible to S inwhich S doesn't do A).57 For what I think is true is the following:

(2.21) S is in control at T of his doing A at T' iff 5 can at Tdo A at T'and also can at T refrain from doing A at T';

(2.22) 5 is in control at T of his doing A at T" iff S is free at T to do A atT' and is also free at Tto refrain from doing A at T'.

Thus, perhaps surprisingly, even though "can" in the present sense does notimply "free to," the following is true:

(2.23) S can at T do A at T" and can also at T refrain from doing A at T"iff S is free at T to do A at T' and is also free at T to refrain fromdoing A at T'.

And so, I believe, "ought" implies "can," not just in that sense of "can"with which I have so far been concerned and which I have put in terms of

56 Actually, I believe that there is more than one sense of the term at issue in the debate,but there is no need to pursue this, since the present point holds, I think, on all relevantsenses.

57 Feldman's account of obligation (1986, pp. 38, 43) does not have this same implication.According to him, "ought" implies "can do" but doesn't imply "can refrain." This leadshim to say that whatever is inevitable is obligatory - including all that lies in the past. Thisis very odd. Still, he does also note (p. 43) that a second, related sense of "obligation" canbe accounted for simply by including the point about "can refrain."

51

accessibility, but in a new, stronger sense of "can" that might be put interms of "free to." And, given the relation between "free to" and "doesfreely" mentioned earlier, we may also say:

(2.24) if S ought at Tto do A at Tand S does A at T, then 5 does A freelyatT.

(This concerns immediate obligation. The corresponding claim concern-ing remote obligation, where Tis distinct from T', is this:

(2.25) if S ought at Tto do A at T' and 5 does A at T', then S does Afreely at T'.

This statement is not true; for, although S may be free at Tto do A at T',he may lose this freedom in the interim and hence, by the time he comesto do A, not do it freely.)

It might be thought that the reason why (2.23) is true, even though"can" (as used here) doesn't imply "free to," is simply this:

(2.26) S is free at Tto do A at T' (or to refrain from doing A at T') iff Scan at T do A at T' and also can at T refrain from doing A at T'.

In other words, "free to" is two-sided, as it were, implying optionality,whereas "can" is just one-sided. But I am not sure that this is so. For it seemsto me that a person can freely do something even though he cannot refrainfrom doing it (in other words, even if it is not personally optional; even ifhe is not in control of doing it).58 If this is so, and if, as I suspect, when some-one does something freely he is (or was) free to do it, then (2.26) is not true.

That someone can do something freely without being able to refrainfrom doing it seems to me to be one of the main lessons to be learned fromHarry Frankfurt's ground-breaking discussion of the relation betweenmoral responsibility and alternate possibilities.59 Frankfurt argues — success-fully, I think — that a person can be morally responsible for what he hasdone, even though he couldn't have done otherwise. (One example mightbe this: Larry cannot leave the room, because the door is locked; he isnonetheless morally responsible for staying in the room, in part because heis unaware of the door's being locked.) What I think this shows is not thata person can be morally responsible for what he has done even though hedidn't do it freely, but rather that a person can have done something freelyeven though he couldn't have done otherwise. (Thus Larry freely remainsin the room, even though he cannot leave it.) It is for this reason that moral

58 At least in "standard control" of doing it. Perhaps he is in some sort of "curtailed con-trol" of doing it. See Zimmerman (1988a), pp. 32-3.

59 Frankfurt (1969).

52

responsibility does not require alternate possibilities. The link betweenresponsibility and freedom remains intact.60

Let us suppose that freely doing something entails being, or having been,free to do it. Then, as I have said, (2.26) is false, given the fact that freelydoing something does not require being (or having been) able to refrainfrom doing it. Why, then, is (2.23) true, given that "can" does not imply"free to"? My answer is simply this: doing something, while being able todo otherwise, is sufficient for doing it freely;6 doing something freely issufficient for being (or having been) free to do it; and being free to do (orto refrain from doing) something is sufficient for being able (where "able"expresses what the "can" of personal possibility expresses) to do (or torefrain from doing) it. I am not sure just how each of these claims mightitself be justified.

At this point, though, it is natural to ask: if moral responsibility doesn'trequire alternate possibilities, why think that moral obligation does? I shallconsider this question in Subsection 3.1.5.

2.2.5 The ontological status of actions

A distinction is commonly drawn between two different ways of conceiv-ing of actions: as types, or as tokens. Types are properties or relations;tokens are instances of the exemplification of properties or relations.Consider killing. Killing is an act-type, whereas a specific killing is a tokenof this type. It seems that obligatoriness, rightness, and wrongness may bepredicated of both tokens and types. Allie may say that a certain killing waswrong; Barbara may say that killing is wrong (period). When faced with aformula of the form

(2.27) S ought at Tto do A at T (in W),

then, we must ask this question: how is "do A" to be understood, as avariable ranging over act-types, or as a variable ranging over act-tokens?62

The short answer is: it doesn't really matter. The analyses (I)—(XII)

60 See Zimmerman (1988a), Section 4.10, for a fuller discussion of this issue.61 In some very important, but perhaps in some ways minimal, sense of "freely." Richer

senses can be obtained by adding certain conditions, such as lack of coercion (seeZimmerman (1988a), pp. 24—6, including n. 22, on the distinction between "strictfreedom" and "broad freedom"), or competence (crucial to certain conceptions ofautonomy), and so on.

62 It was said earlier that in fact not just actions but other events too may be obligatory. Themore general question, then, is whether we're concerned with event-types or event-tokens. But I shall continue to talk just of actions, for simplicity's sake. What I say can begeneralized to apply to all events.

53

provided earlier in this chapter apply for the most part equally well toactions conceived as types or as tokens. Consider (2.27). On one reading,this can be reexpressed in terms of a four-term relation (actually five-term,if worlds are included) as follows:

(2.28) doing A at T" is obligatory for S at T.

Here "doing A" may be understood to be a variable ranging over act-types.On another reading, however, (2.27) can be reexpressed in terms of a two-term relation (actually three-term, if worlds are included) as follows:

(2.29) 5 doing A at T' is obligatory at T.

Here "S doing A at T" " may be understood to be a variable ranging overact-tokens. As substitution instances for each of (2.28) and (2.29), respec-tively, consider these illustrations:

(2.30) helping that old lady on Monday is obligatory for Jones on Sunday;(2.31) Jones's helping that old lady on Monday is obligatory on Sunday.

As far as I can tell, the corresponding substitution-instance for (2.27),namely,

(2.32) Jones ought on Sunday to help that old lady on Monday,

is neutral with respect to the ontological presuppositions of (2.30) and(2.31).63

It may seem, however, that one should not remain neutral on this issue.When we're concerned about what we ought to do (when Jones asks:"Should I help that old lady or shouldn't I?"), aren't we concerned aboutwhether a particular action is obligatory? After all, Jones isn't inquiring intothe morality of helping old ladies in general, but simply into the moralityof this particular helping. And doesn't this indicate that what he's really ask-ing is whether or not an act-token is obligatory, rather than an act-type?

The question is confused. Of course, we are frequently concerned aboutwhat to do here and now (or then and there) rather than in general, but thisdoesn't favor one reading over the other. The particularity that is built intothe act-token o£jones's helping that old lady on Monday is attached external-ly to the act-type of helping that old lady by means of the relata of Jones andMonday. (2.31) enjoys no advantage over (2.30) on this score.

63 In fact, (2.32) is compatible with yet another reading, namely: "Jones helping that old ladyon Monday is obligatory on Sunday' where "Jones helping that old lady" refers to a cer-tain state of affairs. There are a number of reasons for thinking it preferable to conceive ofactions as states of affairs, rather than as types or tokens. I provide such a treatment ofaction in Zimmerman (1984). There is no need to enter this territory here, though.

54

In fact, there is good reason to think that our inquiries into what weought to do are frequently concerned with act-types. Whether or not thisis so depends on how act-tokens are to be individuated. There are twoquestions to be asked here. First, how are act-tokens to be individuatedacross worlds? Second, how are they to be individuated within worlds? Ishall consider each in turn.

Suppose that Jones ought (on Sunday) to help that old lady on Mondayand that he does what he ought. Thus, we may assume, a certain act-token,X, to which one might refer by the description "Jones's helping that oldlady on Monday," occurs in the actual world. Suppose that in fact whatJones does is to help the old lady at noon on Monday. It would have beenequally permissible for him to help her at six o'clock in the evening. If hehad done this instead, would the act-token involved have been X, or wouldit have been some distinct token, Y ? I'm not sure what we should say here,but if Xis distinct from Y (which is what I suspect most people would say),and if Y would have been equally permissible, then X wasn't obligatory.Thus "Jones ought to help that old lady on Monday" would not be prop-erly understood in terms of a certain act-token's being obligatory. Ratherit would be best understood in terms of there being an obligation withrespect to the disjunction X or Y or Z... (for an indefinite number oftokens) — in terms, that is, of analysis (X) above rather than analysis (I).64

This wouldn't mean that (I) is strictly inapplicable to tokens, but it wouldmean that, when we say "5 ought at T to do A at T'," we frequently, andperhaps always, mean something else, namely, that S ought at T to do A atT" or B at T"", or... (for an indefinite number of tokens and times, whereall of the tokens are of some common type and all of the times are containedin T'). It seems to me unlikely that we should systematically misexpressourselves in this way, and thus likely that, when we say: "S ought at Ttodo A at T"," we do not take "do ^4" to refer to tokens - at least, not if wetake Xto be distinct from Yin the case just given.

Of course, one way to try to avoid the foregoing problem is to insist thatXis identical with Y after all. But such coarse individuation of tokens acrossworlds gives rise to a further problem, at least when coupled with coarseindividuation within worlds (of the sort advocated by Elizabeth Anscombe,Donald Davidson, and others).65 Suppose, as before, that X occurred; thatis, Jones helped the old lady (at noon on Monday). Suppose further that, inhelping her, Jones startled her, and that this is not something he should have

64 Cf. Mellema (1991), p. 31, and Vallentyne (1989), p. 308.65 See Anscombe (1969) and Davidson (1980). Contrast, among others, Alvin I. Goldman

(1976) and Kim (1976).

55

done. Call Jones's startling her Z. Those who advocate coarse individua-tion within worlds claim that Xis identical with Z. But then how could Xbe obligatory when Z was not? One response would be to say that X wasn'tobligatory; if Jones had helped the old lady without startling her, as heshould have done, then it wouldn't have been Xbut some other token thatwould have occurred. But this response is not consistent with coarse indi-viduation across worlds. Another response is to say that X (i.e., Z) wasobligatory under the description "helping the old lady" but not under thedescription "startling the old lady." Such a response would echo whatcoarse individuators have said before in other contexts (such as the contextof intentional action: some actions are claimed to be intentional under onedescription but not under another),66 but the price of accepting thisresponse would be that of declaring such formulae as

(2.27) S ought at Tto do A at T ' (in W)

crucially incomplete. Their completion would require mention of descrip-tions. I doubt whether this accords with our common understanding ofobligation.

An additional problem with saying that it is act-tokens that are obliga-tory is this. If a token exists only when it occurs, then what is said to beobligatory need not, indeed need never, exist. It is not clear that it makessense to talk of the obligatoriness of something that does not exist.67

One reason for thinking that we nonetheless do commonly conceive ofobligation as concerning act-tokens is this. Many people want to say thatan act is obligatory only if its consequences satisfy a certain condition; buttypes don't have consequences, only tokens do. I think this view is mis-taken. Of course, we might say:

(2.33) Jones's helping that old lady at noon on Monday caused that oldlady's being pleased at six o'clock on Monday,

and this makes apparent reference to a causal relation between two event-tokens. But I think we could equally well say:

(2.34) helping that old lady caused (relative to Jones, noon on Monday,that old lady, and six o'clock on Monday) being pleased,

and here there is no apparent reference to tokens.

66 See Anscombe (1969), p. 11 and elsewhere; Davidson (1980), pp. 5, 46, and elsewhere.67 Prichard appears to doubt that it makes sense to talk this way (see Prichard (1949), p. 37),

whereas Broad does not (see what he says about what he calls an "agibile" in Broad, 1985,p. 226).

56

This tactic might prompt the following response. The move from (2.33)to (2.34) involves a rethinking of the relation of causation.68 It involves"complicating" causation (by thinking of it as a six-term, rather than two-term, relation) in order to keep obligation "simple." This is no less objec-tionable than "complicating" obligation (by introducing descriptions) andthereby keeping causation "simple." I suspect that this response is unsatis-factory, but I shall not pursue the issue here. For, to repeat, the account ofobligation that has been given is intended for the most part to be adequateboth to a conception of actions as types and to a conception of actions astokens.

Why only "for the most part"? One reason has already been given: thosewho individuate tokens coarsely both across and within worlds and whoclaim that it is tokens that are obligatory must supplement the formulaegiven with mention of descriptions (or employ some similar device).Another reason is this: statements (2.11)—(2.14) cannot be understood (asthey stand) to concern act-tokens; for, as mentioned in Subsection 2.1.4,they presuppose the repeatability of actions, and tokens are not repeatable.Consider (2.11) again, which was supposed to capture one sense of what itis to have a perfect duty:

(2.11) for all times T (at which S exists) not earlier than Tu S ought at Tt

to do A at T.

This presupposes that S can do A on more than one occasion, which is per-fectly intelligible where A refers to an act-type but not where it refers to anact-token. Still, even here the proponent of understanding obligation interms of tokens need not concede defeat; he need only insist on revisingsuch formulae as (2.11). Perhaps this would do:

(2.35) for all times T (at which S exists) not earlier than Tu S ought at Tt

to do either A at T or B at T or... (for an indefinite number oftokens, all of which are of some common type).

Similarly, an "absolute" injunction against performance of a certain act-type might be recast in terms of all tokens of that type being wrong.

2.2.6 Indefiniteness

Indefiniteness can infect obligation in a number of different ways. Here Ishall mention four.

68 I advocated a similar rethinking in Zimmerman (1984), Section 3.1, where, as remarkedin note 63 to this chapter, actions are treated not as tokens or types but as states of affairs.

57

First, it might be thought that certain deontically relevant values areincommensurate with one another.69 For example, it might be thought thatliberty and equality are both morally valuable, and that for that reason anyworld that contains one or the other of them is, to that extent, deonticallygood, but that they are nonetheless incommensurate, in that it is impossi-ble to say that one is more or less valuable than the other or that they areequally valuable.70 Of course, this rough statement might be taken to meana number of different things. It might be understood only to say that we arenever in a position to tell which is more valuable, liberty or equality. Suchepistemological difficulty need not point to any incommensurability inprinciple. Or the statement might be understood only to say that there isno general rule as to which is more valuable, liberty or equality, since theformer is sometimes more valuable, sometimes less valuable, than the other.This too would not point to any incommensurability in principle; on thecontrary, it presupposes commensurability. Or the statement might beunderstood to mean that on some but not all occasions liberty is neithermore nor less valuable than, nor equal in value to, equality. This would bean instance of what Walter Sinnott-Armstrong calls "limited ^compara-bility."71 Or the statement might be understood to mean that on all occa-sions liberty is neither more nor less valuable than, nor equal in value to,equality. This would be an instance of what Sinnott-Armstrong calls"extreme incomparability."72

I suspect that certain morally relevant values may indeed be incom-mensurate and that, as a result, it can happen that one is faced with acces-sible worlds that are deontically incomparable, that is, worlds of which itmust be said that one is, deontically, neither better nor worse than, norequal in value to, the other.73 What does the present account imply in suchcases? That depends on just where the deontic incomparability applies.Remember that, according to (I), S ought to do A if and only if S can doA, S can refrain from doing A, and any accessible world in which S doesnot do A is deontically inferior to some accessible world in which 5 does

69 This is a distinct matter from that mentioned in Section 1.1, where the (incommensu-rability of moral with nonmoral requirements was at issue.

70 Cf. Berlin (1979), p. 151.71 Sinnott-Armstrong (1988), p. 62ff.72 Sinnott-Armstrong (1988), p. 59ff.73 This is not to say that the incommensurateness of deontically relevant values must result

in the deontic incomparability of worlds that contain these values. After all, even ifliberty and equality are incommensurate, it may be that one world contains more of eachthan another, so that the former world can be compared (favorably) with the latter.Cf. Seung and Bonevac (1992), pp. 800-1.

58

A. (I) does not require that there be some particular A-world that is deon-tically superior to every ~^4-world. Rather, it requires that, for every ~A-world, there be some y4-world that is deontically superior to it. This allowsfor deontic incomparability between worlds. For instance, it may be that Sis faced with two deontically incomparable sets of worlds, but as long as,within each set, every ~^4-world is deontically inferior to some ^4-world,5 ought to do A. (This even allows for deontic incomparability within setsbetween A-worlds and yl-worlds and between ~y4-worlds and ~v4-worlds.)But if the deontic incomparability at issue is between all A-worlds and some~^4-world and all ~v4-worlds and some ^4-world, then clearly (I) and (III)imply that it is neither the case that S ought to do A nor the case that 5ought not to do A. What's left? Obviously this: 5 may do A and S may notdo A (in the sense of (IV)); that is, A is morally optional.74 In other, rougherwords: if you have a choice to make between two deontically incompara-ble actions, you may take your pick; neither is obligatory, each is permis-sible. This is what my account implies, and it seems to me correct.

The second type of indefmiteness that I want to acknowledge is that ofvagueness. It may be that certain moral norms are vague, in that there areborderline cases where it is not clear whether or not they apply or whetheror not they have been satisfied. Suppose, for instance, that Gertrude oughtto be grateful to Lucy but that just what constitutes an adequate display ofgratitude is unclear. Perhaps it is clear that simply sending her a thank-younote would be to do too little, while giving her a brand-new BMW wouldbe to go far beyond what gratitude alone demands. But just where the min-imum and maximum fall may not be clear. Would sending her a bunch ofdaisies suffice, or would it have to be lilies? It may be tempting to say thatthe difficulty here is merely epistemological, that gratitude's boundaries arewell defined in every case, even if we sometimes cannot tell what they are.But this won't do, I think. For there certainly appear to be cases wherewhat's required of someone is in fact indeterminate, not just hard for us todetermine. One such case is this: I promise Gail that I will help her withher gardening around noon tomorrow. Suppose that this gives rise to anoverall moral obligation. Since the content of the promise is irremediablyvague, it seems that the content of the obligation is also. Clearly, if I turnup by noon and help Gail as much as I can until all the gardening is done,I will have done what I ought (and perhaps even more than I was requiredto do). But what if I don't turn up until 12:15, or 12:30, or 1:00, or 2:00.

74 Here the phrase "no greater than" in (II) and (IV) is important; "at least as great as" wouldnot do, since it implies comparability. Cf. Vallentyne (1989), p. 304; McConnell(1993b), p. 250; and Pietroski (1993), p. 508.

59

and what if I don't help Gail as much as I can but merely lend a hand, or afinger, every once in a while? Will I have failed to do what I ought? Whereis the line to be drawn?

I think that we must accept that there is no precise point at which todraw the line, and hence that obligations can be vague in the way indicat-ed. My account is not intended to imply otherwise. It is intended, rather,to be compatible with the claim that there are borderline cases with respectto whether or not a certain action, A, is featured in the best accessibleworlds. Obviously this is a somewhat disturbing thought, but it is not aproblem for my account in particular. Vagueness is disturbing generally;how to account for it satisfactorily is a general problem. Whatever the prop-er account of vagueness, I see no reason to think that it would not be con-sistent with what I have to say here about obligation.

The third type of indefiniteness concerns the indeterminacy of options,in the following sense. I have said, in (2.18), that S can do A only if a cer-tain counterfactual is true, one that concerns what would be the case if Swere to make a certain decision. But can't it happen that such a counter-factual lacks a determinate truth-value? Perhaps it can.76 If it can, then insuch cases the statement that S can do A also lacks a determinate truth-valueand so, too, each of the statements that S ought to do A, S may do A, Sought not to do A, and S may not do A lacks a determinate truth-value.

I think that we may well have to accept that obligation can be indeter-minate in this way. Again, my account is not intended to imply otherwise.But again, if the thought is disturbing, it doesn't point to a special problemwith my account. Rather, it is an issue concerning counterfactuals in gen-eral. Of course, one could try to avoid the problem in the present contextby denying that obligation is tied in this way to counterfactuals, but I thinkthat this would be incorrect.77

The final type of indefiniteness concerns the possibility (which I shallaccept for argument's sake)78 that an agent be faced with an infinite num-ber of accessible worlds and that each world is such that there is a deonti-cally better one. The account that I have given is adequate to this possibility.I have said that S ought to do A as long as (roughly) every accessible ~A-world is deontically inferior to some ^4-world. This is quite compatiblewith there being no deontically best world; it just means that, if S ought to

75 For a limited discussion of vagueness and obligation, see Sorensen (1991).76 Cf. Lewis (1973), pp. 66-7, 91-4; Mendola (1987), p. 127ff.77 Feldman tries to do just this, in Feldman (1986), p. 17, by denying that accessibility is to

be understood in terms of counterfactuals.78 I am in fact quite skeptical of this possibility, however. See Zimmerman (1990a).

60

do A, then, at a certain point, all worlds that are higher in the ranking mustbe A -worlds. Here, by the way, is where the final subclauses of (I) and (III)are required. It is conceivable, I suppose, that 5 be faced with an infinitenumber of accessible worlds so ordered that, although every ~y4-world isworse than some ^4-world, it is also true that every ^4-world is worse thansome ~A-world. Without their respective subclauses (c4), (I) and (III)would imply, in such a case, both that 5 ought to do A and that S oughtnot to do A. This is unacceptable. With their respective subclauses (c4), (I)and (III) imply that it is not the case that 5 ought to do A and that it is notthe case that S ought not to do A.79 Notice that (II) and (IV) imply, evenin the absence of any such subclause, both that S may and that S may not(in the sense that S is permitted not to) do A. This seems to me preciselywhat should be said in such a case.

Feldman calls the sort of situation where 5 is faced with an infinite num-ber of increasingly good accessible worlds "delightful."80 But he overlooksan implication that is somewhat disturbing, the implication that, no matterwhat S does, he cannot avoid doing something wrong.81 For if an accessi-ble world, Wu is distinct from another, W2i it must be because S acts dif-ferently in W1 than in W2;

S2 and if W1 is deontically inferior to W2, then Smust act wrongly in Wt.

One way to try to avoid this result would be to amend the account ofwrongdoing provided above. As (III) now stands, it implies that it is possi-ble that S ought not to do A (that is, it would be wrong for 5 to do A), eventhough there is no other act B that S ought to do instead; for it may be that5 ought not to do B, either. Indeed, this is precisely the case in the sort ofsituation presently envisaged. Such an implication may well seem odd, andit could be obviated by adding to (III) the condition that there is an acces-

79 Feldman's account of obligation contains no clause corresponding to subclause (c4) of(I) and is nonetheless able to accommodate the possibility that S ought to do A eventhough S is faced with an infinite number of increasingly good accessible worlds. (SeeFeldman, 1986, pp. 37—8.) This is because his account says (roughly) that 5 ought to doA just in case some ^4-world is better than every -~v4-world. In order to accommodatedeontic incomparability, however, my account (as noted above) says only that every ~A-world is worse than some A -world. Such an account requires supplementation with sub-clause (c4), given the possibility in question.

80 Feldman (1986), p. 37.81 Cf. Slote (1989), p. 110.82 This is because, on my view (presented in Subsection 2.2.1), worlds are accessed through

choices. Given that choosing is a (minimal) way of acting, if one accessible world differsfrom another, it will be (at least) because the agent makes a choice in one (namely, thechoice by which he accesses that world) that he does not in the other.

61

sible world, or set of accessible worlds, deontically superior to all otheraccessible worlds. Thus, we could say, one can do wrong only if there is abest that one can do. To avoid the appearance of being ad hoc, this sug-gestion could be coupled with the observation that the account alreadyimplies a parallel thesis, namely, that one can be obligated only if there is aless than best that one can do. (For, in the limiting case where all one'saccessible worlds are such that none is deontically superior to any other,everything is permissible, nothing is obligatory; there is not even the dis-junctive obligation to take one's pick, since — as (X) declares — such anobligation implies that some choices would be wrong.)

But I am disinclined to adopt this amendment, for three reasons. First,it destroys the symmetry of the treatment of O(A) in (I) with that of O(~A)in (III), a symmetry that seems natural to me. Second, there is a pair of par-allel theses to which appeal can be made other than the pair just mentionedin the last paragraph. These are the theses that one can do wrong only if onecan do better, and that one can be obligated only if one can do worse. Thesetwo theses are implied by my account as it already stands, and it is arguablethat whatever intuitive plausibility the first pair of theses has is itself to bederived from the plausibility of this second pair. And third, the suggestionof course has the implication that, in the sort of case under discussion, onecan do no wrong. This seems unacceptable to me, for it gives one licenseto actualize some deontically abysmal accessible world, even though thereis a far (indeed, infinitely) superior accessible world that one could chooseto actualize.

I think, then, that it is best to accept that, in the very special sort of sit-uation envisaged, it is indeed the case that S cannot avoid doing somethingwrong. Thus the present account (under the assumption that an agent maybe faced with an infinite number of accessible worlds, each of which isdeontically inferior to some other) does not rule out moral dilemmas of acertain sort. Notice, however, that the sort of dilemma at issue may be called(following Peter Vallentyne) a. prohibition dilemma: all of S's options involvewrongdoing. This is not the same as saying that obligation dilemmas are pos-sible, that is, that it can happen that 5 ought to do some act A and ought todo some act B but cannot do both.83 The present account implies that such

83 See Vallentyne (1987a), p. 113, and (1989), pp. 302 and 305. I am here extendingVallentyne's own use of "prohibition dilemma" somewhat. According to him, a prohi-bition dilemma has the form ~P(A) & ~P(~A). The account given in (II) and (IV) rulesout the possibility that ~P{A) & ~P(~A) be true. But, as just noted, it does not rule outthe possibility that all of an agent's options involve wrongdoing, and the term "prohibi-tion dilemma" seems to me aptly applied to such cases.

62

dilemmas are impossible (barring certain complications). This is an issuethat I shall address more fully in Chapter 7.

2.3 IMPLICATIONS OF THE ANALYSIS

2.3.1 Certain deontic implications

Anyone who has some familiarity with the work of deontic logicians willbe aware of the vast number of formulae which have been touted as eitheraxioms or theorems of one or another system of deontic logic. These for-mulae are usually alleged to express certain truths about obligation.Consider the following (which have been selected both because they arerepresentative of the sorts of formulae in question and because it will beespecially useful to comment on them):

(2.36)(2.37)(2.38)(2.39)(2.40)(2.41)(2.42)(2.43)(2.44)(2.45)

Op D Pp;Op = ~P ~p;OpZ>~O~p;O(pv~p);O(p&q) = (Op &P(p v q) = (Ppv .O(p Dq)D(Op(pD q) D (Op DOp D OOp;Pp D OPp.

Oq);Pq);DOq);

Oq);

Each of these formulae seems to suggest a proposition that looks reason-able, but, if you're at all like me, you'll be wondering just what propositionis supposed to be expressed and just how reasonable it is.

I think it's fair to say that in many cases the efforts of deontic logicianshave been expended prematurely. In many cases they have attempted toconstruct formal systems on the basis of considerations too weak to bear theweight imposed upon them. More work needs to be done on the basic con-cept of obligation before such systems can provide genuine philosophicalinsight. Nevertheless, such systems are certainly not without their philo-sophic uses. If formulae such as those just presented are at all plausible, thenthey can serve as a gauge of the adequacy of an account of obligation suchas the one that I have proposed. If the account departs from the formulae,it ought to be able to explain why.

First, though, the formulae must be interpreted, and this is no easy task.Standardly, the O and P are intended to express obligatoriness and permis-

63

sibility, respectively, but what are the variables p and q supposed to rangeover? Presumably propositions. This is so for two reasons. First, the con-nectives ~, D, =, &, and v have a familiar interpretation when applied topropositions and propositional variables, but not otherwise. Second, andless importantly, certain parallels with alethic modal logic (the logic ofnecessity and possibility, taken by many to be analogous in various ways toobligatoriness and permissibility, respectively) may then be drawn. Thetrouble is that, although such treatment of the variables lends itself nicelyto the expression of such claims as that it ought to be that little children nothave to suffer, it is hard to see how it can adequately handle such claims asthat Tim ought to do what he's been told, given that "It ought to be thatTim do what he's been told" does not express the same claim. (See Section1.2 on binding versus nonbinding senses of "ought.") One way around thiswould be to read Op as "5 ought to see to the occurrence ofp," but thishas difficulties of its own. Suppose that Roger ought to raise his hand. Justwhat event or proposition is it that Roger ought to see to the occurrenceof, his raising his hand or his hand's rising? In neither case is it clear that weget an equivalent claim; in one case too much seems to be said, in the othertoo little. To say that Roger ought to see to it that he raises his hand seemsto say too much, in that it suggests that he ought to do something over andabove just raising his hand. To say that Roger ought to see to it that hishand rises seems to say too little, in that it appears compatible with his ful-filling his obligation by way of ensuring that his hand rises in such a waythat it is not he but someone or something else that raises it.

In light of these considerations, I think it best to rewrite the formulaeunder consideration so that they conform with the account of obligationthat I have proposed. Thus first of all we should consider this claim:

(2.36') if O(A), then P(A); that is,if S ought [at T] to do A [at T in W], then S may [at T] do A [atT i n W].

It is clear that my account - specifically, (I) and (II) above - implies that thisis true. If all ~vl-worlds [accessible to 5 from Wat T] are worse than (thatis, deontically inferior, for 5 at T, to) some ^4-world [accessible to S fromWat T], then certainly the former are no better than the latter.

The next claim to be considered is this:

(2.37') O(A) iff ~P(~A).

84 Feldman employs this reading in Feldman (1986), p. 37.

64

This breaks down into two further claims, the first of which is this:

(2.37'a) ifO(,4),then~P(~,4).

It is clear that my account also implies this to be true. If all ~A-worlds areworse than some /1-world, then it is not the case that all v4-worlds are nobetter than some ~v4-world. The other claim at issue here is this:

(2.37'b) ]f~P(~A), then O(A).

This turns out not to be true on my account, for ~P(~^4) may be true dueto a lack of personal optionality, that is, true in virtue of there being eitherno /1-world accessible to S or no ~A-world accessible to S (that is, in virtueof S's not both being able to do A and being able to refrain from doing A),in which case O(A) will be false. For example, no one can jump over themoon; hence no one may refrain from doing so; but it doesn't follow thateveryone ought to do so. Still, the following is true on my account (whereC expresses the "can" of personal possibility):

(2.37'c) i£C(A)&C(~A) &~P(~A), then O(A).

If there are both an ^4-world and a ~v4-world, and if it is not the case thatall ^4-worlds are no better than some ~/l-world, then all ~y4-worlds areworse than some /1-world. Furthermore, in light of the facts that obliga-tion requires personal optionality and that, where there is personal option-ality, acting in a manner that is not permissible constitutes acting wrongly,we may indeed accept the following:

(2.37") O(A) iff W(~A).

Next is this claim:

(2.38') if O(A), then ~O(~,4).

My account implies that this is true. (Indeed, it follows from (2.36') and(2.37'a).) If 5 ought [at T] to do A [at T' in W], then it is not the case thatS ought [at T] not to do A [at T' in W]. This is because, if all ~y4-worldsare worse than some ^4-world (than which there is no better ~^4-world),then it is not the case that all yl-worlds are worse than some ~^4-world. Ofcourse, (2.38') is controversial (as are, perhaps, some of the other claimsalready considered). I shall discuss it and related claims in detail in Chapter 7.

It is not clear just how the next formula, (2.39), is to be rewritten so asto conform with the present account. It suggests the following:

(2.39') S ought to do either A or -A,

65

but just what this is supposed to mean is not clear. In accordance with (X),we might try interpreting it as follows:

(2.39") 0(Aor~A).

This is false, on my account, since it implies that, for any act A, S may doA and S may not do (that is, may refrain from doing) A. But I think it isplain that proponents of (2.39) have had something else in mind. Just asp v q is standardly taken to be equivalent to ~(~p&~q), so too, I think, wecan understand (2.39) to be rewritable as follows:

(2.39'") O(,4v~,4);thatis,O~(~A&A).

Even if this does capture what is intended, though, it turns out to be falseon my account. This is because its truth would imply the truth ofC(~A&A), and this is true of no agent.

The next claim is this:

(2.40') O(A&B) iff (O(A) & O(B)),

which breaks down into these two:

(2.40'a) if O(A&B), then O(A) &O(B);(2.40'b) if O(A) & O(B), then O(A&B).

The first of these is not true on my account, for it may happen that, although(A&B) is personally optional, A (or B) is not; if so, then although O(A&B)may be true, O(A) (or O(B)) won't be.

To this observation there are a number of possible responses. First, onemight claim that every genuine action is indeed personally optional; it can-not be, for instance, that S acts in moving his arm unless he could have doneotherwise. (Certainly, it can happen that one's arm moves irresistibly, butthen it is not oneself who is moving it.) Thus, where O(A&B) is true, eachof O(A) and O(B) will be too. But this is problematic, for two reasons. First,it seems false that action, to be genuine action, must be optional, for threereasons: hard determinism may rule out the possibility of free action, but itdoesn't rule out the possibility of all action whatsoever;85 certain cases ofactual action are plausibly understood to involve irresistible impulses, casessuch as, perhaps, certain compulsive or obsessive actions, actions due toaddiction or hypnosis, and sleepwalking; and Frankfurt-type cases, where

85 Unger (1977) notwithstanding, where the argument founders on an equivocation on"cause."

66

one does something "on one's own" but where one would have beenforced to do it anyway, seem possible.86 Secondly, and more importantly,it must be remembered that, in the formula O(A) and its ilk, the variablesare not restricted to ranging over actions. (See the discussion in Subsection2.2.1 above.) And certainly some sorts of events that can be optional for anagent don't have to be. For instance, O(A) might on occasion express "it isobligatory for me that I be indoors," and yet there can clearly be times whenmy being indoors is not personally optional for me (either because I can-not get indoors or because I cannot get outdoors).

A second response might be to acknowledge the second point just madeand yet to insist that, if O(A&B) is true, then each of O(A) and O(B) mustbe too, whether or not what is obligatory is said to be an action. Since myaccount doesn't imply this, it ought to be revised. Specifically, (V) shouldbe revised so that it implies, not just that (A&B) is optional, but that eachof A and B is too. But this response is problematic, too. First, it seems thata revision would be called for not just for (V) (the analysis of O(A&B)) butalso for (I) (the analysis of O(A)). Consider, for instance, the claim thatChuck ought to open a checking account. (I) implies that opening a check-ing account is optional for Chuck. But, on the present suggestion, it oughtto say more. For opening a checking account is surely a complex action.Suppose that it can be analyzed in terms of actions X, Y, and Z, so that wemay now say that Chuck ought to do (X&Y&Z). Then (V), revised accord-ing to the present suggestion, would imply that each of X and Y and Z isoptional for Chuck, whereas (I), left unrevised, does not have this implica-tion; (I) implies only that (X&Y&Z) is optional for Chuck. Thus (I) wouldneed to be revised to say that each component part of A is optional for S;but a precise expression of this would surely be very complicated and dif-ficult to accomplish. A second problem with this second response is simplythis: it's just not clear that, whenever O(A&B) is true, so too is each of O(A)and (OB). Consider, for example, the claim "You ought not to stand idlyby while there is poverty." Why wouldn't it be correct to analyze this inthe terms set out in (V), namely: (a) there is a (not-standing-idly-by-and-poverty)-world accessible to you; (b) there is a not-(not-standing-idly-by-and-poverty)-world accessible to you; and (c) all worlds of the latter sortare worse than some world of the former sort? Surely this can be true, eventhough every world accessible to you is a poverty-world; that is, nothingyou can do will eradicate poverty. Similarly, someone might claim,"Whatever you do, you ought to do it without complaining." Some of the

86 See Frankfurt (1969).

67

things one does may not be avoidable, although whether or not one com-plains while doing them is. This is enough to give the claim some plausi-bility. Whether or not the claim turns out to be true, we should not dismissit simply on the grounds that it implies statements of the form O(A&B)where A is conceded not to be optional.

Thus (2.40'a) is indeed to be rejected. But a third response is to note thatthree very closely related claims are nonetheless true on my account. Thefirst two are these:

(2.40'c) i£C(~A) & O(A&B), then O(A);(2.40'd) if C(~B) & O(A&B), then O(B).

These are true because, if all ~ (A &B) -worlds are worse than some (A&B)-world, then (since any ~^4-world is a ~(A&B)-world and any (v4&B)-worldis an ^4-world) all ~^4-worlds are worse than some yl-world; similarly, all~B-worlds are worse than some B-world.87 The third implication is this:

(2.40'e) i£O(A&B), then either O(A) or O(B).

This is true, because it cannot happen that (A&B) is optional without atleast one of A and B being optional.

Let us now turn to the second "half" of (2.40'), namely:

(2.40'b) if O(A) & O(B), then O(A&B).

This claim, sometimes called the "principle of agglomeration" (and to bediscussed further in Chapters 6 and 7), is true on my account. Suppose thatall ~^4-worlds are worse than some /1-world. Such an A-world must beeither an (A&B)-wov\d or an (A&~B) -world. But if we suppose also thatall ~B-worlds are worse than some B-world, then all ~y4-worlds are worsethan some (A&B)-world. Similarly, all ~B-worlds are worse than some(A&B)-world. Now, a ~(A&B)-wov\d is either a ~^4-world or a ~B-world.Hence all ~ (A &B)-worlds are worse than some (A &B) -world.

The next claim is presumably to be rewritten as follows (interpreting thev as before):

(2.41') P(A v B) iff (P(A) v P(B)); that is,P~ (-A&-B) iff (P(A) v P(B)).

This breaks down into two claims. The first is this:

(2.41'a) if P~(~A&~B), then P(A) v P(B).

87 It was thus on the implicit understanding of the personal optionality for Sam of each ofthe actions of sticking out his tongue and saying "Aah" that I claimed at the outset that(2.1) implies (2.2).

68

My account implies that this is true. If P~ (~A&~B) is true, then all(~A&~B)-worlds are no better than some ~(~A&~B)-world; but if nei-ther P(A) nor P(B) is true, then the best worlds will all be (-A&-B)-worlds. The second claim is this:

(2.41'b) if P(A) v P(B), then P~ (-A&-B).

My account implies that this is false. Suppose that P(A) is true, and hencethat P(A) v P(B) is true, but suppose that P(B) is false because C(~B) is false(that is, because S cannot refrain from doing B). If C(~B) is false, then sotoo is C(~A&~B); but from this it follows that P~ (~A&~B) is also false.

Inasmuch as p D q is equivalent to ~(p&~q), I think we may rewrite(2.42) as follows:

(2.42') if O(A D B), then, if O(A), then O(B); that is,if O~(A&~B), then, if O(A), then O(B).

My account implies that this is true. Suppose that all ~vl-worlds are worsethan some ^4-world. Either B is performed in the y4-world or it is not. If itis not, then, given the truth of O~ (A&~B), there is a still better world inwhich both A and B are performed, and all ~B-worlds will be worse thanthis world.

It is not clear just how (2.43) should be rewritten. Perhaps we can putit as follows:

(2.43') if~(,4&~B), then, if 0(^4), then O(B),

where we understand this to express the following: if S doesn't do A with-out doing B, then, if he ought to do A, then he ought to do B. My accountimplies that this is false. So what if S doesn't do A without doing B ? Perhapshe ought to. That is, it may be that the best ^4-worlds are ~B-worlds; and,if the best worlds are ^4-worlds, then O(B) will not be true. As an illustra-tion, let A be the action of helping Hilda and B be the action of humiliat-ing Hilda. Perhaps Laura ought to lend Hilda a hand and refrain from herusual caustic comments; if she helps Hilda but also humiliates her, it hard-ly follows that, in humiliating her, she has done what she ought.

A strengthened version of (2.43') may appear more plausible:

(2.43") if ~C(A&~B), then, i£O(A), then O(B).

That is: if S cannot do A without doing B (whether this is out of logicalnecessity, physical necessity, or something weaker), then, if he ought to doA, then he ought to do B. The foregoing example about humiliation isinapplicable here. If S cannot do A without doing B and the best accessi-

69

ble worlds are A- worlds, then a fortiori they are B- worlds as well. But in fact(2.43") isn't quite right, either; for it may be that 5 cannot do A withoutdoing B because he simply cannot refrain from doing B, in which case O(B)will not be true. So B must be said to be optional, and what is needed is this:

(2.43'") if ~C(A&~B) & C(~B), then, if O(A), then O(B).SS

This is implied by my account and will be featured in discussion later in thischapter and in Chapters 6 and 7.

The final two formulae are difficult to rewrite to conform with the pre-sent account.89 Perhaps the most suitable renditions are these:

(2.44') if S ought to do A, then S ought so to act that S ought to do A;(2.45') if 5 may do A, then S ought so to act that S may do A.

If these are to make sense, I think that it must be recognized that the timeof S's acting so that S ought to or may do A is prior to the time of S's actu-ally doing A. Thus these propositions apply at best when there is an implic-it clause in their consequents to this effect: if S has time so to act. On thisunderstanding, (2.44') is true, on my account, but (2.45') is not. As to theformer: if all ~/l-worlds are worse than some ^4-world and then S so actsthat this is no longer true, then S clearly departs from the high road andgoes wrong. Consider Figure 2.1 again, now under the assumption that W\is deontically superior to all the other worlds. Given this assumption, Sought at T1 to do C at T3, and he ought at T} so to act that he ought at T2

to do C at T3. If S does ~A at Tu or if he does ~B at T2, then it will nolonger be true that he ought (prior to T3) to do C at T3. But this will bebecause he went wrong at either Tt or T2.1 shall address this issue in greaterdetail in the next chapter. As to (2.45'): this is not true, even given theimplicit extra clause, and that is simply because it is permissible to make cer-tain preparations to do what is itself permissible, even if these preparationsrule out doing something else that is permissible. For instance, perhaps Imay either go to the theater or go to the opera this evening, although I can-not do both. If, as I may, I spend my money on tickets to the theater, thenI have rendered myself unable to go to the opera; hence it is not the caseany longer that I may now go to the opera; hence it is not the case that Iought so to have acted that I may now go to the opera.

It is interesting to note that the truth of (2.44') implies that there is asense in which we have an obligation to be morally perfect. This sense may

88 It was thus on the implicit understanding of the optionality of B that I claimed at the out-set of this chapter that we may assume the truth of (2.10).

89 Problems with the interpretation of such formulae are explored in Marcus (1966).

70

be called that of "perfection in action," in that (in this sense) an agent ismorally perfect if and only if he never acts wrongly.90 Note that to state thatwe have an obligation to be perfect in this sense is not to set an impossibly(or even unreasonably) high standard. It does not imply that we have anobligation to be perfect in some other sense (for instance, in the sense thatour characters, or motives, or whatever, be beyond all moral criticism)where it might be impossible for us to achieve such perfection (or unrea-sonable to claim that we should do so). Nor does it imply an impossiblyhigh ideal for action. It can of course happen that an agent can perform eachof a number of very good deeds and yet cannot, for one reason or another,perform all of them. Stating that this agent, like all agents, has an obligationto be perfect (in the present sense) does not commit me to stating that he isobligated to perform all of these very good deeds.91 On the contrary, sincehe cannot perform them all, my account commits me to saying that it is notthe case that he ought to perform them all. What he ought to do, of course,is to avoid wrongdoing. This might well involve his taking his pick amongthe various good deeds in question, performing some and not performingothers. As long as he does avoid wrongdoing, he will have satisfied theobligation to be perfect in the present sense.

There are of course many other formulae that we could examine here,but I don't think that there is any need to do so. The ten that have beendiscussed are representative of how my account handles such matters. I haveexplained why some of the formulae are not implied by the account, and Ihope that the explanation serves to confirm the acceptability of the account.Some of the formulae that are implied by the account are admittedly con-troversial; as noted, these will be discussed in later chapters. In addition,there are of course implications of my account that are not purely deontic.Foremost among these is surely the following:

(2.46) if O(A), then C(A); that is,if S ought [at T in W] to do A [at T'], then S can [at T in W] doA [at r ' ] .

This controversial claim will be examined in Chapter 3. When conjoinedwith other implications, (2.46) of course yields yet further implications. Forexample, the principle of agglomeration — (2.40'b), above — conjoined with(2.46) gives us this:

(2.47) if O(A) & O(B), then C(A&B),

90 See McGinn (1992), p. 33.91 Contrast Clark (1993), pp. 54-5.

71

and this is true, on my account, even though (as just noted) the followingis of course to be rejected:

(2.48) if C(A) & C(B), then C(A&B).

But this, too, will be discussed in later chapters.

2.3.2 Certain deontic paradoxes

Formal systems of deontic logic have been plagued by certain so-calledparadoxes. These paradoxes are not formal inconsistencies; rather, they arecases where the system's representations of certain "ought"-statementsseem separately plausible but jointly implausible, since when conjoinedthey have, or appear to have, counterintuitive implications. Some suchparadoxes will be treated in Chapter 3 (when self-imposed impossibility isdiscussed) and in Section 4.3 (when conditional obligation is discussed).But two (or three, depending on how you count) may be discussed now.

The first paradox is known as Ross's Paradox.92 In standard form, it maybe put as follows. Given

(2.49)

and

(2.43)

we may infer

(2.50) OpDO(pvq).

It has been thought that the premises are true but the conclusion false, sincethe latter seems to warrant saying (for instance) that, if Matt ought to maila letter, then he ought to either mail or burn it, and this seems to grant Mattlicense to burn the letter. It might be tempting to try to undermine the para-dox simply by pointing out that (2.43) is false (as argued above), but in factthis won't do; for a strengthened version of (2.43) has been acknowledgedas true, and the argument can be reformulated around it. So let me refor-mulate it in this way, now rewriting it so as to conform with the presentaccount. Accordingly, we may say that, given

(2.49') ~C(A&~(AvB))

and

(2.43"') if~C(,4&~B) & C(~B), then, if O(A), then O(B),

92 See Ross (1941), where the subject is imperative rather than declarative sentences.

72

we may infer

(2.50') if O(A) & C(~(A v £)), then O(A v B).

Now, (2.49') is necessarily true: one cannot do some act A and yet do nei-ther A nor B (for any act B). And I have said that my account implies that(2.43'") is true. Hence this account implies that (2.50') is true: if 5 oughtto do A and can refrain from doing either A or B, then 5 ought to do eitherA or B. But isn't this to be rejected? As noted, it would appear to giveemployees of the postal service license to burn rather than mail the lettersentrusted to them.

The resolution of this minor puzzle is of course to be found in the dis-tinction between O(A v B) and 0(AorB). Briefly, it is only the latter thatgives S license to do B (as spelled out in (X) above), but it is only the for-mer that is at issue in (2.50'). O(A v B) means simply that 5 ought notto do neither A nor B, and clearly this is true if 5 ought to do A. End ofproblem.93

Formula (2.43) gives rise, though, to another, much more interestingparadox, commonly known as the Good Samaritan Paradox.94 Let p be theproposition that Harry helps Vera, who has been robbed by Rita. Thisentails q, the proposition that Vera has been robbed by Rita. Suppose thatHarry ought to help Vera. (2.43) would seem to imply that it follows thatRita ought to have robbed Vera - a disturbing conclusion. Now, some sys-tems of deontic logic do imply this conclusion and so seem properly criti-cizable on this score. The present account, however, bars this conclusion,for two reasons. First, (2.43"') is to be read as follows: if S cannot do Awithout doing B and S can refrain from doing B, then, if S ought to do A,then S ought to do B. Obviously, this thesis is restricted to inferring what5 ought to do from what S ought to do; it will not in general permit infer-ring an obligation on the part of Rita from an obligation on the part ofHarry. Second, if Harry ought now to help Vera, and Rita's robbery of Veralies in the past, then once again no inference with respect to an obligationto rob is forthcoming; for "ought" does not apply to the past but only tothat which is not presently inevitable (as the insertion of C(~B) in (2.43'")guarantees).

But although this response is accurate, it does not put the GoodSamaritan Paradox to rest. For the paradox can be recast so as to concernonly a single agent's obligations and only the future.95 Suppose, then, that

93 See Follesdal and Hilpinen (1970), pp. 21-3; Castafieda (1981), pp. 63-5.94 See Aqvist (1967).95 Cf. Castafieda (1981), pp. 52-3.

73

A is the action of Harry's helping the woman (Vera) whom he (Harry) willrob at T3, and B is the action of Harry's robbing Vera. Suppose that Harryought at Tj to do A at T2. Given (2.43'"), it seems to follow that Harryought at Tt to do B at T3; for, it seems, Harry cannot at Tt do A at T2 with-out doing B at T3, and Harry can at Tt refrain from doing B at T3. Clearlythis is an unwelcome result.

The proper response is this. The claims that Harry ought to do A andthat Harry cannot do A without doing B are jointly unacceptable. Whichis to be rejected depends on just what A is taken to be. Perhaps this is bestexplained by explicitly invoking worlds in the description of actions. If Ais the action of Harry's helping in world H^the woman whom he robs atT3 in the actual world, then it is evident enough that Harry can do A withoutdoing B. For certainly there are worlds accessible to Harry at T1 in whichHarry helps Vera at T2 without robbing Vera at T3, even though it is truethat Harry will in fact (that is, in the actual world) rob Vera at T3. If A isthe action of Harry's helping in world Wthe woman whom he robs at T3

in W, then it is evident that Harry cannot do A without doing B. But thenwe should not say that Harry ought at Tt to do A at T2. For B does notoccur in the best worlds accessible to Harry, and if Harry cannot do A with-out doing B, then A doesn't occur in the best accessible worlds either.

This last point may be a bit hard to grasp, and I think that this is in partbecause the example lends itself so readily to an understanding of A and Baccording to which doing A does not entail doing B. Let me turn, then, toa version of the Good Samaritan Paradox where it is clear that doing a cer-tain thing 4 does entail doing another thing B. This is the version commonlyknown as the Paradox of the Knower. Suppose that Greg, a security guard,ought to provide truthful reports of any wrongdoing done on his watch andhence (to keep the discussion to a single-agent case) ought to provide atruthful report of any wrongdoing committed by himself. (The paradoxwas originally put in terms of knowing rather than truthful reporting,96 butthe latter is more suitable to single-agent cases.) Suppose that Greg oughtnot at Tj to do A at T2 but in fact will do A at T2. Is it the case that Gregought at Tj to provide at T3 a truthful report of his doing A at T2? Clearlynot, and this is because his doing so entails his doing A at T2. The best worldsaccessible to Greg at Tt are worlds in which he does not do A at T2 andhence does not provide at T3 a truthful report of his doing A at T2.97 Thisis of course perfectly compatible with saying that, once he has done A

96 See Aqvist (1967), p. 366ff.97 Cf. Feldman (1990), pp. 327-8.

74

at T2, Greg ought then to provide at T3 a truthful report of his doing so.Such shifts in obligation will be the topic of Chapter 3.

2.3.3 Analysis and neutrality

I take it to be a point in my account's favor that it satisfactorily handles thevarious deontic paradoxes. (Again, some of these paradoxes are to be dis-cussed in later chapters.) On the other hand, certain implications of theaccount have been acknowledged to be controversial. How, then, can Ipropose the account as an analysis of the concept of obligation? Aren'tanalyses supposed to be neutral between various conflicting conceptions ofa concept? Shouldn't, for example, an analysis of "ought" neither implythat "ought" implies "can" nor imply that it doesn't?98

This is a tricky issue, which could be circumvented simply by declaringthat (I)-(XII) are intended to be understood as necessary truths and leavingit at that, without declaring in addition that they are intended to be takenas analyses. But this would be misleading, for I do intend them to be takenas analyses.

It is a mistake to think that analyses must be neutral in the sense at issue.Consider what is probably the most celebrated philosophical analysis, thatof the concept of knowledge as true belief that is (in some way) justified.This is clearly not neutral in the sense at issue. It implies that knowledgeentails belief, that knowledge entails truth, and so on, and each of theseimplications has been contested.

But consider what Feldman says. Although he conceives of his projectas one of "laying out the logical structure of the concept of moral obliga-tion,"99 which matches a description that I have given of my project, hedenies that his account is intended as an analysis of this concept. He says:

I propose that we understand the principle that we ought to do the best we canin the following way:

MO: MOs, t, p is true at w iff (3u/)[As, t, w' &p is true at w' & ~(3u>"){As, t, w" &~p is true at w" & W(w") > IV(u/)}]

On this formulation,.. .s morally ought, as of a time, t, to see to the occurrenceof a state of affairs, p, iff p occurs in some world accessible to s at t, and it is notthe case that ~p occurs in any accessible world as [intrinsically] good as (orbetter than) that one.100

98 For discussion of a closely related issue, see Sayre-McCord (1986).99 Feldman (1986), pp. x-xi.

100 Feldman (1986), p. 38. Some typographical changes have been made.

75

This, of course, is structurally very similar to the account that I have pro-posed. Later, though, Feldman considers this principle:

(2.51) (s)(t)(p)(s morally ought, as of t, to see to p if and only if MOs, t, p),

and he has this to say of it:

Some readers...may have expected that I would take (2.51) to be an emptytautology. If my definition, MO, had been intended as an "analysis" of the con-cept of moral obligation, then this expectation would perhaps be justified. Butthe definition was not intended to be an analysis of some concept we alreadyhave. It was intended to be a stipulative definition of a concept to which Iwanted to draw attention.

If we compare (2.51) with the following principle, I think the differencebetween moral obligation and MO will emerge:

(2.52) (s)(t)(p)(MOs,t,p if and only if MOs,t,p)

If we grant (as I think we must) that MOs,t,p has a meaning, then we shouldgrant that (2.52) is an empty tautology. It tells us nothing of interest aboutmorality. Even those who think that my approach is hopelessly wronghead-ed.. .will accept (2.52). On the other hand, (2.51) is intended to tell us how weought to behave. It tells us that we morally ought to bring about all and onlythose states of affairs that fall under the concept MO... So even if (2.51) is anecessary truth, it is not an empty tautology. It is a fundamental moral doc-trine...

Since some will disbelieve the proposition expressed by (2.51) while not dis-believing the one expressed by (2.52), these propositions are distinct...

This "open question" argument serves to emphasize the fact that my defin-ition, MO, is not an analysis of the concept of moral obligation.101

I think these remarks are erroneous. I think we must grant that (2.51) and(2.52) are distinct propositions, but we should not infer from this that (2.51)is not an analysis of the concept of moral obligation. If we did, then wewould have to say that the following proposition is not an analysis of theconcept of knowledge:

(2.53) 5 knows thatp iff S has a justified, true belief that p ,

since it is distinct from this proposition:

(2.54) S has a justified, true belief that p iff S has a justified, true beliefthat p .

101 Feldman (1986), pp. 207-8. The numbering of propositions has been changed to con-form with the numbering that has been used in this chapter.

76

Of course no one will reject either (2.52) or (2.54), while some willreject either or both of (2.51) and (2.53). But analyses need not be uncon-troversial; they need not be neutral with respect to various conceptions ofthe concept that is analyzed. Perhaps some people102 believe that analysesmust be neutral and uncontroversial because they believe that analyses areanalytic and that analytic statements are trivial, and hence neutral anduncontroversial. But this would be a mistake. If (2.53) is a correct analysis,then the following statement is analytic:

(2.55) if S knows that]?, then S believes that p.

This statement derives its analyticity from (2.53). All analyticity is derived inthis way from analyses;103 thus analyses are not themselves analytic. Is (2.55)uncontroversial? Hardly. It is "trivial," in the sense that, given (2.53) (orsomething like it), belief is in a way conceptually "contained" in knowl-edge. But since (2.53) is not itself uncontroversial, there is no reason toexpect (2.55) to be, even if it is analytic.

Similarly, if my proposed analysis of obligation is correct, then so too isthe following claim:

(2.46) if 5 ought to do A, then S can do A.

Indeed, (2.46) will be analytic; but its being so is no reason to expect it tobe uncontroversial or in any way neutral with respect to all possible con-ceptions of obligation. In saying this, I am of course committed to sayingthat anyone who denies (2.46) is making some sort of conceptual error; butthis is not to say that the denial of (2.46) involves a logical error, at least notin the sense that it involves an inconsistency in and of itself, as would, forexample, the denial of this claim:

(2.56) if S can do A, then S can do A.

But it is conceptual truths, and not logical truths, that I am after.Two huge questions loom here. I can only point to them; I cannot do

them justice. The first is this. If analyses are not themselves analytic but syn-thetic, how are they to be distinguished from other synthetic necessaryequivalences that are not analyses? I wish I knew the answer to this ques-tion, for it would mean that I could clearly distinguish conceptual errorsfrom errors of some other sort (and thereby, perhaps, clearly distinguishmetaethics from normative ethics). Consider, for instance, this claim:

(2.57) S ought to do A iff A best pleases S.

102 I am not saying that Feldman is among these people.103 See Chisholm (1989), pp. 33-4.

77

Someone might propose this as a necessary truth. If so, I am sure that hewould be mistaken. I wouldn't accuse him of a conceptual error, though, butrather of a substantive error. Perhaps it is easy to see why this is so; for (2.57)in no way attempts to "break down" the concept of obligation into com-ponent concepts, while this is what an analysis attempts to do. But otherclaims in this vein are not so easy to judge. Consider Feldman's ownaccount again, with MO and (2.51) merged thus:

(2.58) (s)(t)(p)(s morally ought, as of t, to see to p iff @w')[Asy t, w' &pis true at w' &~(3w"){As, t, w" &~p is true at w" & W(w") >

Surely there is some sort of "breaking down" involved here. Nevertheless,I think that Feldman is correct to deny that this constitutes (pure) analysis.The explicit invocation of intrinsic value seems to me to render the claimsubstantive, and not purely conceptual. Even if Feldman is correct aboutthe link between obligation and intrinsic value, those who deny that heis are making a substantive error; if he is incorrect, then he has made asubstantive error — he hasn't made an error concerning the "structure ofthe concept" of obligation. Or so it seems to me, but I confess I don't knowhow to argue for this or against the claim that the errors mentioned areindeed conceptual rather than substantive.

The second question is this. I say that someone who denies (2.46), forexample, is making a conceptual error. But might it not be that someonewhom I take to be denying (2.46) is not in fact denying it but rather some-thing else? For might not this person be using "ought" (or "can") to expressa concept different from that which I take it to express, so that the proposi-tion that he takes to be expressed by "if S ought to do A, then S can do A"is indeed false? The answer, of course, is that this may indeed be so. Perhapsthere are ways of trying to rule this out — I can test him, for example, onjust when he takes it to be appropriate to use "ought" (or "can"), skillful-ly varying the contexts so as to get a fix on his meaning — but the disturb-ing possibility remains that those who affirm what they take (2.46) toexpress and those who deny what they take (2.46) to express are not inreality disagreeing with one another but rather talking at cross-purposes.

78

The dynamics of obligation

3.1 "OUGHT" AND "CAN"

There are few propositions more widely discussed in the literature onphilosophical ethics than the proposition that "ought" implies "can." I havenoted that my account of obligation implies this proposition to be true,indeed to be analytic, when understood as follows:

(2.46) if 5 ought [at Tin W] to do A [at T'], then 5 can [at T in W] doA [at T'].

That my account implies the truth of this proposition will no doubt inclinemany readers to reject the account.

In this section I shall consider a number of reasons for denying (2.46),proceeding from objections that seem to me quite weak to objections thatseem to me stronger (but still unsuccessful). In Subsection 2.2.4 I distin-guished two senses of "can" such that "ought" implies "can." The weakersense was just that of personal possibility or accessibility; it is this sense of"can" that I have employed in this work. The stronger sense was that ofbeing free to perform an action (stronger in that a person's being free toperform an action implies, but is not implied by, that action's being per-sonally possible for that person). In what follows, it is "can" merely in thesense of personal possibility that is explicitly at issue - except when alter-nate possibilities are discussed in Subsection 3.1.5, where it is "can" in thesense of "free to" that is at issue.

3.1.1 An internal criticism

First, it might seem that (2.46) should be deemed unacceptable by my ownlights, in that the account that I have proposed is avowedly intended toaccommodate such views as that what is obligatory is what is probably best

79

(in some substantive sense of "best"), or what is apparently best, and so on.And might it not easily be that what is probably or apparently best is in factnot open to the agent, even when what is best is open to the agent?

The answer is that there are two understandings of the claim

(3.1) 4 is probably (or apparently) the best (course of) action open to theagent.

On one reading what this means is the following:

(3.1/) of all actions (or courses of action), whether open to the agent ornot, A is that which is probably (or apparently) the best one opento the agent.

Here, it is true, there is no guarantee that A is indeed open to the agent.2

The second reading is this:

(3.1") of all actions (or courses of action) that are open to the agent, A isthat which is probably (or apparently) the best.

Here, of course, it is implied that A is open to the agent. (By way of anal-ogy, suppose that someone, looking down from a balcony at a room full ofpeople, proclaims

(3.2) Rex is probably (apparently) the richest Republican here.

This may mean one of two things. First, it may mean that

(3.2') of all the people in this room, Rex is probably (apparently) the rich-est Republican.

Now this may be true and yet Rex not even be a Republican. But the claimin question might instead be this:

(3.2") of all the Republicans present, Rex is probably (apparently) therichest.

This cannot be true if Rex is not a Republican.) The fact is that, while myaccount accommodates theories according to which (3.1") corresponds towhat is obligatory, it does not accommodate theories according to which(3.1') corresponds to what is obligatory. On my view, anyone who thinksthat (3.1') corresponds to what is obligatory is making a conceptual error(just as, for instance, anyone who thinks of obligation along the lines men-tioned in the second paragraph of Chapter 2 is making a conceptual error).

1 See the discussion of (A)-(C) in Section 1.4.2 Cf. Frankena (1963), p. 159.

80

The error has to do with the fact that, unlike (3.1"), (3.1') is consistent withthe denial of

(2.46) if 5 ought to do A, then S can do A.

Of course, my insistence that the denial of (2.46) constitutes a conceptualerror is not itself proof of (2.46). But so far I have only tried to show why(2.46) is not to be deemed unacceptable by my own lights.

3.1.2 "Can" and "can avoid"

A second reason for denying (2.46) is this.3 We shouldn't say that it is moral-ly permissible for Trudy not to travel faster than light; for what is physical-ly necessary is not the sort of thing to which moral permissibility pertains.But, as we all know, an act is obligatory if and only if it is not permissiblenot to do it, and so it turns out that Trudy is obligated to travel faster thanlight, even though she cannot.

This is a distinctly odd argument. Let A be Trudy's traveling faster thanlight. The argument may then be represented as follows:4

(3.3) ~P(~,4);(3.4) O(A) iff ~P(~A);

hence,

(3.5) O(A);

but

(3.6) ~C(A);

hence,

(3.7) it is not the case that, if O(A), then C(A).

In assessing this argument, we should ask why anyone would want toassert its first premise, (3.3). The answer, of course, is that we should notsay that something (such as not traveling faster than light) is permissibleif it is impossible to avoid. In other words, (3.3) is itself based on thefollowing claim:

(3.8) if P(~A)9 then C(A).

3 What follows is inspired by Shaw (1965).4 Here I use symbolism introduced in Subsections 2.1.3 and 2.3.1.

81

How odd that an argument against the claim that "ought" implies "can"should rest on the claim that "may" implies "can avoid"! Clearly, some-thing has gone wrong here. Some might diagnose (3.3) as the culprit, butI do not. I think it's quite right to say that "may" implies "can avoid." Myaccount is committed to this. Where the argument goes wrong, I believe,is in asserting (3.4). Far from being something that we all know to be true,(3.4) is, on my account, false. I pointed this out in Subsection 2.3.1. (3.4)is the same proposition as (2.37'), which itself implies both

(2.37'a) if O(A), then ~P(~A)

and

(2.37'b) if ~P(~A), then O(A).

The culprit is (2.37'b). This is false. What's true is this:

(2.37'c) if C(A) & C(~A) &~P(~A), then O(A).

And so I do affirm the following:

(3.4') O(A) iff C(A) & C(~A) & ~P(~A).

But of course this cannot be used in an argument against the claim that"ought" implies "can," since it implies that "ought" does imply "can."

3.1.3 An alleged redundancy

A quick argument against (2.46) has been provided by Alan White. He says:

It is...because what we ought to do is not necessarily something that we cando that we have occasion to use such an expression as 'We ought to deal withthis matter now, if we can'. The qualification 'if we can', like the qualification 'ifwe have time', would be simply redundant if we ought to do only what we can.5

The allegation of redundancy is inaccurate. Sometimes we are unsurewhether we can do something and whether, if we can, we ought to do it.The answer "We ought to do it, if we can" is partially helpful. While notspeaking to the first question, it does speak to the second.

3.1.4 Past and present obligations

Walter Sinnott-Armstrong has proposed rejecting the claim that "ought"implies "can" on the following grounds.6 He asks us to consider thestatement

5 White (1975), p. 148.6 Sinnott-Armstrong (1985).

82

(3.9) I ought to have mowed your grass.

He says that this admits of two interpretations:

(3.9') I now ought to have mowed your grass in the past;

and

(3.9") it was true in the past that I ought to mow your grass.

Sinnott-Armstrong favors the first interpretation, saying:

[S]uppose I promise on Monday to mow your grass by Saturday. (3.9") is stilltrue on Thursday and on all future days, because it was true on Tuesday that Iought to mow your grass by Saturday. In contrast, (3.9') is not true onThursday, because it is not yet true that I ought to have mowed your grass inthe past (before Thursday), since I promised only to mow it by Saturday.7

Sinnott-Armstrong then proceeds to provide two analogous interpreta-tions of

(3.10) I can have mowed your grass.

They are:

(3.10') I now can bring it about that I did mow your grass;

and

(3.10") it was true in the past that I could mow your grass.

He then argues as follows. If "ought" implies "can," then (3.9) implies(3.10). The proper interpretation of (3.9) is (3.9'); analogously, that of(3.10) is (3.10'). But (3.90 doesn't imply (3.10'), since (3.9) will be trueeven when the promise in question is not fulfilled by Saturday and thuseven when (3.10) is false. Hence "ought" does not imply "can."

This is intriguing but, I fear, quite muddled. We should note, first of all,that Sinnott-Armstrong's argument that (3.9') is preferable to (3.9") as aninterpretation of (3.9) is unpersuasive, in that it overlooks a third interpre-tation of (3.9) which I believe is clearly preferable to both, and that is:

(3.9'") it was true in the past that, for some time T, I ought to mow yourgrass at T, and Tis past.

In the example given, Tis Saturday, and it is for this reason that (3.9) is nottrue when uttered prior to Saturday. Now, if "ought" implies "can," then

7 Sinnott-Armstrong (1985), p. 46. The numbering of propositions has been changed toconform with the numbering that has been used in this chapter.

83

we should expect that (3.9"') implies this:

(3.10"') it was true in the past that, for some time T, I could mow yourgrass at T, and Tis past.

And, I believe, this implication does indeed hold. The fact that, as Sinnott-Armstrong correctly says, (3.9) can be true whereas (3.10') is false, is irrel-evant; for (3.10') is not the proper interpretation of (3.10).

To bolster his claim that (3.9') is the proper interpretation of (3.9),Sinnott-Armstrong says this:

[Someone might respond] that judgments like (3.9') are never true. However,such a universal claim cannot be justified without begging the question.Furthermore, it often seems natural to say that someone ought to have donesomething, even if he or she did not do it. We need to be able to make presentjudgments about what ought or ought not to have been done in the past inorder to justify present punishment for past acts or failures to act. The correctinterpretation of such judgments cannot have the form of (3.9") — [or, I amsure Sinnott-Armstrong would add, (3.9'")] - because [such] judgments.. .aretrue in many cases where punishment is not justified, such as when it was nolonger true at the time at which the agent did the act that he or she ought notto do it. Thus, it must be judgments like (3.9') that seem natural and needed.8

But this, too, is muddled. I don't know whether it's begging the questionfor me to insist, as I do,9 that, if S ought at T to do A at T', then T" is notearlier than T, but I can certainly give an account of the propriety of "pre-sent judgments" that differs from Sinnott-Armstrong's. Sinnott-Armstrongseems to suggest that punishment is justifiable only if judgments like (3.9')can be made, for judgments like (3.9") (and (3.9'")) can be true even whenpunishment is not justifiable. It is of course true that punishment may notbe justifiable even when judgments like (3.9'") are true, but this does notwarrant the inference that it is justifiable only when judgments like (3.9')are true. On the contrary, judgments like (3.9') are never true. It is judg-ments ofresponsibility, not (simply) ofobligation, that justify punishment, andit is certainly possible that someone should now be responsible for a pastaction, even though it is not possible that someone should now be oblig-ated to perform a past action.10

8 Sinnott-Armstrong (1985), p. 47. The numbering of propositions has been changed asbefore.

9 See the discussion of "ought to have" in Subsection 2.1.3, including note 25.10 See Zimmerman (1988a), Ch. 3, on what is there called "retrospective" responsibility.

84

3.1.5 Alternate possibilities

The question was raised in Chapter 2: if moral responsibility doesn't requirealternate possibilities (as Frankfurt has argued, and as I have concurred),why think that moral obligation does? I shall now address this question,doing so by way of a response to an argument made by David Widerker,according to which the claim that moral responsibility doesn't requirealternate possibilities in fact implies the claim that "ought" does not imply"can."11

One cautionary note: up to now, "can" has been understood in termsmerely of personal possibility or accessibility, but in the present discussionof alternate possibilities it is necessary that it be understood in terms offreedom. (On this matter, see the second paragraph of this chapter.) Tomake this explicit, let us use C' to symbolize this stronger sense of "can."The principle that is presently at issue, then, is this revised version of (2.46):

(2.46/) if O(A), then C(A).

The principle concerning moral responsibility and alternate possibilitiesthat Frankfurt and I and possibly Widerker12 reject is this:

(3.11) a person is morally responsible for what he has done only if hecould have done otherwise; that is (ignoring matters of tense),if R(A), then C'(~A)U

Frankfurt argues that the rejection of this principle does not require rejec-tion of (2.46').14 But Widerker argues that Frankfurt is mistaken, since(2.460 entails (3.11), or at least that part of (3.11) that deals with moralblame:

(3.11a) S is morally blameworthy for doing A only if 5 could have doneother than A; that is,if B{A), then C'{~A).

Widerker's argument is simple. He claims that (3.11a) follows from (2.46'),given the following proposition, which he claims to be a necessary truth:

(3.12) 5 is morally blameworthy for doing A only if 5 ought to have doneother than A; that is,if B(A), then O(~A).

11 Widerker (1991).12 Widerker is not fully explicit on this issue.13 Frankfurt (1969), p. 829ff.14 Frankfurt (1988), pp. 95-6.

85

Of course, (3.11a) does follow from (2.46') conjoined with (3.12).(Widerker is relying on that version of (2.46') which says that, if S oughtto do other than A, then S can do other than A.) But (3.12), far from beingnecessarily true, is false. This is because moral responsibility is a function ofwhat one believes about one's moral obligations and not of what one's moralobligations actually are.15 One can be blameworthy for what one has done,even if one was justified in doing it, as long as one believed at the time thatone was not justified in doing it. Thus blame worthiness does not entailwrongdoing, and to justify an action is not to exculpate its agent. (Failureto recognize this fact too often prematurely terminates our moral inquiries.)It is also the case (although this is not strictly germane to the present dis-cussion) that wrongdoing does not entail blame worthiness; this is widelyaccepted. Wrongdoing and blameworthiness are thus logically indepen-dent of one another (though not conceptually so; the latter is to be analyzedin terms of the former), and Widerker's argument against Frankfurt fails.

It could, of course, still be that Widerker's conclusion is true. But Ibelieve otherwise. I think that (2.46') doesn't imply (3.11a), because thelatter is false but the former true. The reason is this. (3.11a) is a contractionof two other principles:

(3.13) S is morally blameworthy for doing A only if S could have doneA; that is,if E(.4), then C(.4);

and

(3.14) S could have done A only if 5 could have done other than A; that is,ifC'(,4),thenC'(~,4).

What Frankfurt's argument against (3.11a) shows is that (3.14) is false;(3.13) is left untouched.16 (Here, by the way, is where it is most obviousthat we are dealing with the stronger "can" of freedom — C' — and notmerely that of personal possibility — C. For it would hardly require theinsightful efforts of Frankfurt to show that the following proposition is false:

(3.14') ifC(,4),thenC(~,4).

Similarly, there would hardly be any need to belabor the truth of thisproposition:

(3.13') if B(A), then C(A).)

15 I have argued for this in Zimmerman (1988a), Sections 3.1 and 3.6.16 See Zimmerman (1988a), Section 4.10, for discussion of this.

86

In other words, Frankfurt has not severed the link between moral respon-sibility and freedom, a link that, I believe, holds of conceptual necessity. Inthis regard, the concept of moral blameworthiness is just like many otherfundamental moral concepts. For example, in addition to (3.13), we mayalso accept:

(3.15) S is morally praiseworthy for doing A only if S could have doneA; that is,i£Pr(A), then C'(A).

More importantly, (2.46') is just one more instance of this link between morali-ty and freedom, as is the following principle:

(3.16) it is morally wrong for 5 to do A only if S can do A; that is,if W(A), then C(A).

Now, none of these principles themselves directly concerns alternatepossibilities. But when we note that

(2.37") O(A) iff W(~A),

we can immediately derive from (3.16) and (2.46') these further principles(with appropriate substitutions of ~A for A):

(3.17) S ought to do A only if 5 can do other than A; that is,ifO(,4),thenC'(~,4);

and

(3.18) it is morally wrong for S to do A only if 5 can do other than A;that is,if J/F(A),thenC'(~,4).

Thus the link between obligation and alternate possibilities is preserved byan appeal to (2.37"), not (3.14), and this leaves room for the cogency ofFrankfurt-type cases against the link between responsibility and alternatepossibilities by means of an attack on (3.14).

Some questions may remain, though. One is this. Even if it is grantedthat (3.14) is false, might not the link between responsibility and alternatepossibilities be restored by means of an argument similar to the one that Ihave given that aims to preserve the link between obligation and alternatepossibilities? In particular, might there not be a responsibility analogue to(2.37") that functions as (2.37") does in my argument? The answer is thatI know of no such analogue. In particular, it is clearly not the case that

87

(3.19) 5 is morally praiseworthy for doing A iff S is morally blame-worthy for doing other than A; that is,Pr(A) iff B{~A).

For S cannot both do A and fail to do A. Indeed, it is not even the case that,where 5 has not yet acted, S would be praiseworthy for doing A if and onlyif 5 would be blameworthy for failing to do A. For more is required forpraiseworthiness than the avoidance of blameworthiness, and more isrequired for blameworthiness than the failure to act in a praiseworthyfashion.17

Another question is this. Doesn't it simply seem that Frankfurt-stylecounterexamples work just as well when applied to obligation as whenapplied to responsibility? Consider, for example, this case. Murphy freelycommits murder but, for Frankfurt-type reasons, couldn't have done otherthan commit murder. Surely Murphy does something wrong. But, given(2.37"), it follows that (2.460 *s felse after all; tna t is> it follows that "ought"does not imply "can." A similar point can be made concerning (3.17), theclaim that "ought" implies "can avoid." Suppose that Ralph freely refrainsfrom committing murder, although again, for Frankfurt-type reasons, hecouldn't have done other than so refrain. Surely he does what he ought;but then it follows that (3.17) is false after all.

The situation here may be put as follows. I have argued:

(2.37") O(A) iff W(~A);(2.460 ifO(,4),thenC'(,4);

hence,

(3.18) if W(A), then C'(~A)\

and also:

(2.37") O(A) iff W(~A);

(3.16) if W(A), then C'{A)\

hence,

(3.17) if O(A), then C'(~A).The critic accepts the first premise of each argument but, because he rejectstheir conclusions, rejects their second premises. One alternative would ofcourse be to reject their first premise instead, thus leaving the link between"ought" and "can" and between "wrong" and "can" intact. But of course

17 See Zimmerman (1988a), Ch. 3.

88

I reject this alternative, for I take the arguments to be sound. Yet hasn't thecritic given good, Frankfurt-type reasons for rejecting their conclusions? Idon't think so. I would readily grant that Murphy does something that is insome way morally bad, and that Ralph avoids doing this; but, because nei-ther agent can act otherwise, it is incorrect to say that Murphy does some-thing wrong or that Ralph acts as he ought. (Similarly, it is incorrect to saythat it is or would be wrong for Ralph to commit murder, where this isunderstood to include the rider: under the circumstances. I can of course agreethat it would be wrong for Ralph to commit murder if he had the option ofdoing so.)

This is not a particularly satisfactory situation. It may seem that, whenthe critic makes a point, I simply dig in my heels; when I make a point, hedigs in his heels. Is there a way to resolve this? That is unclear to me. I canpoint to what is surely the initial intuitive plausibility of (2.46') and (3.16);I can point to the general kinship between (2.46'), (3.16), (3.15), and(3.13); I can (as I do in other parts of this chapter) defend (2.46) (and, byimplication, (2.46') and (3.16)) against other attacks. All of this seems to meto favor my position rather than the critic's, but I don't suppose it will winover everyone.

3.1.6 Psychological incapacitation

Having mentioned "can avoid" as well as "can," I shall turn now to an argu-ment against the claim that "ought" implies "can avoid," reverting to theweaker sense of "can" that merely involves personal possibility (for thereis no longer any need to stick with the stronger sense). The claim now atissue, then, is this:

(3.17') ifO(^4),thenC(M).

And the argument is simply this. It can happen that a person's moral train-ing during childhood is very rigorous, so rigorous that, in later years, he ispsychologically incapable of acting otherwise than in accordance with hismoral obligations. This implies that (3.17') is false.

While I agree that moral training can be very rigorous, I am doubtfulthat it can lead to such incapacity. Still, perhaps it can. George Washingtondeclared himself (while still a child) unable to tell a lie (although he wasapparently quite capable of destroying someone else's personal property).Suppose that this was literally true, and true as a result of his moral training.

18 Adapted from van Rijen (1993), pp. 265-6.

89

In avoiding mendacity, did he act as he ought to have done? Well, maybehe did; for perhaps he was obligated, at a time prior to his becoming unableto tell a lie, not to tell a lie at that later date when he had become unable todo so. But this of course is quite consistent with (3.17') and is, presumably,not what the critic has in mind. No, the critic means that, even after he hadbecome unable to tell a lie, Washington was obligated then not to do so.But I deny this. I would grant, as I did with Ralph, that Washington avoid-ed doing something that was in some way morally bad, but he did not avoiddoing something wrong; he did not act as he ought to have done.

It may again seem that I am just digging in my heels. But I am remind-ed here of a lovely remark attributed to Mark Twain: "I am morally supe-rior to George Washington. He couldn't tell a lie. I can and I don't."There is a moral insight here. One who can tell a lie but refrains from doingso is to be distinguished, morally, from one who "refrains" from doing sobecause he is unable to do so.20 Similarly, one who feels fear but nonethe-less acts courageously is to be distinguished, morally, from one who actsfearlessly. And so on. Of course, it might be argued that the moral distinc-tions in question have to do not with obligation but with something else(character, perhaps). I would agree that they have to do with somethingelse, but I don't see it as digging in my heels to say that they have also to dowith obligation.

3.1.7 The obligatory and the ideal

One of the stronger arguments against the claim that "ought" implies "can"comes from considerations of such common claims as the following:

(3.20) I ought not to have felt so covetous;(3.21) you ought to feel gratitude;(3.22) he ought to be ashamed;23

(3.23) she ought to stop laughing at her friend's new haircut.24

19 As quoted in van Inwagen (1983), pp. 63-4.20 It is arguable that it is possible on occasion to act freely on a certain impulse even when

the impulse is irresistible. (I explore this issue in Zimmerman, 1995b.) Even if this is pos-sible, I would say that, given the irresistibility of the impulse, such action cannot be eitherobligatory or wrong, although it can of course be morally significant in some other way.

21 See Moore (1922), p. 319.22 See Frankena (1963), p. 150.23 See White (1975), p. 198.24 See Sinnott-Armstrong (1988), p. 116.

90

In these cases, and in indefinitely many others like them, it seems that whatthe agent ought to do he cannot do. A standard response to cases of this sortis to say that there are two senses of "ought," one of which implies "can"and the other of which doesn't. The claims just given involve a sense of"ought" that does not imply "can," while the sense of "ought" at issue inthe account of obligation provided in Chapter 2 does imply "can."25

This response is frequently ridiculed. White, for example, says that "itoffers no reason, other than its alleged ability to explain failing to feel whatone ought, for supposing that there is a second sense of 'ought'."26 Sinnott-Armstrong similarly complains that the response is question-begging.27 Butwhile I grant that, if no independent reason for distinguishing two sensesof "ought" were given other than preservation of (2.46) in face of theexamples given in (3.20)—(3.23), then this would constitute a begging ofthe question against opponents of (2.46), the fact is that, despite what Whiteand Sinnott-Armstrong say, such independent reason is forthcoming.Moore himself distinguishes that sense of "ought" having to do with whatis ideal (or desirable) from that which has to do with duty (or obligation).There is surely reason to acknowledge this distinction; indeed, it is the verydistinction that I introduced in Section 1.2, when contrasting what I calledthe "binding" and "nonbinding" senses of "ought." I made no mentionthere of the issue of whether "ought" implies "can"; nevertheless, I submitthat the distinction had intuitive appeal. If any further evidence of the dis-tinction is needed, simply recall the fact, mentioned in Section 1.2, that cer-tain "ought"-statements (those which express the nonbinding sense of"ought") are what I called "passive-transformable," whereas others (thosewhich express the binding sense of "ought," that is, obligation) are not. Inthe nonbinding sense of "ought,"

(1.1) someone ought to help that old lady

expresses exactly the same proposition as

(1.1') that old lady ought to be helped by someone,

while, in the binding sense of "ought,"

(1.3) Jones ought to help that old lady

expresses a quite different proposition from that expressed by

(1.3') that old lady ought to be helped by Jones.

25 See Moore (1922), p. 318fF. Cf. Frankena (1963), pp. 150-1.26 White (1975), p. 150.27 Sinnott-Armstrong (1988), p. 119.

91

The charge that this response begs the question cannot, therefore, besustained.

It may of course be true on occasion that someone is in (indirect) con-trol of his emotions; perhaps he can suppress his covetousness, or summonup gratitude, or induce shame, and so on.28 If so, then he may well be oblig-ated to exhibit such control. The point being made here is simply this:(3.20)-(3.23) can all be true, even if the agent cannot control whether ornot he has the emotions in question; but if they are true in this way, thenthe "ought" that they contain does not express obligation.

It is worth noting also that the distinction between the two senses of"ought" is not confined to matters having to do with the emotions (suchas covetousness, gratitude, shame, and amusement). On the contrary, thecase just given of Jones and the old lady has to do with an action, that ofhelping. Other cases are easily imagined. For instance, we might say that

(3.24) Smith ought to give up smoking,

or that

(3.25) White ought to leave that wound alone,

even though neither Smith nor White can do what he ought.29 Here, too,it is the nonbinding, ideal sense of "ought," rather than the binding, oblig-ation sense, that is operative.

A final point is this. What has been said about uncontrollable emotionsapplies also to uncontrollable motives. Ross is well known for making thispoint:

That action from a good motive is never morally obligatory follows.. .from theKantian principle...that 'I ought' implies 'I can'... I can act from a certainmotive only if I have the motive; if not, the most I can do is to cultivate it bysuitably directing my attention or by acting in certain appropriate ways so thaton some future occasion it will be present in me, and I shall be able to act fromit. My present duty, therefore, cannot be to act here and now from it.30

Others have made a similar point.31 We must take care, however, not tomake unwarranted inferences from this observation. For example,although it may be that, imbued with malice, I cannot summon up anysympathy and hence that it is not true that I ought (am obligated) to act

28 Cf. Moore (1922), p. 316.29 Both cases are borrowed from White (1975), p. 148.30 Ross (1930), pp. 4-5.31 E.g., Prichard in Prichard (1949), pp. 129-30.

92

from sympathy, still it might be true that I ought not to act maliciously. Forwhile having a certain motive may not be in our present control, it is, I think,often possible to exercise control over whether or not we act/rom a motivethat we do have. In saying this I am, like Ross, clearly not using "motive"to refer simply to that which does in fact "move" one to action; rather, I useit to refer to that which has the capacity to do so. Motives, in my sense, maythus compete with one another for the role of "actual motivator," and, Iam claiming, which one plays that role is frequently up to the agent.

3.1.8 Excuses

It may be argued that, even if the distinction between what is obligatoryand what is ideal is accepted, still there are actions which are obligatory, andnot ideal, and yet which cannot be performed. Alan Montefiore gives thisexample: "I ought to be at the meeting tonight, but unfortunately I cannotmanage it."32 Of such a case White says: "We don't blame him because weaccept his inability as an excuse. But it wouldn't be an excuse for not doingwhat he ought to do if it could make that cease to be what he ought todo."33 Similarly, Sinnott-Armstrong says:

[I]f a bride cannot get to her wedding, then she ought at least to let someoneknow and to excuse her absence... [I]f it were not first true that she ought togo to her wedding, she would have nothing to excuse when she failed to go toher wedding. Excuses are distinguished from justification in that to justify anact is to show that it is right, whereas to excuse an act is to admit that it is wrongbut to deny responsibility.34

Here, I think, the opponents of the claim that "ought" implies "can"are in the grip of a dogma, one attributable to J. L. Austin, who declaredthat excuses are to be distinguished from justifications in just the way thatSinnott-Armstrong mentions.35 But what Austin said seems wrong.Excuses aren't offered only where wrongdoing (that is, the failure to satis-fy an obligation) is admitted. Consider this case, provided by TerranceMcConnell:

Suppose I am at a lake when a man drowns. Another accuses me of wrongdo-ing for failing to save the man. But I reply that I am not able to swim (or that

32 Montefiore (1958), p. 24.33 White (1975), p. 152.34 Sinnott-Armstrong (1988), p. 113, emphasis added.35 See Austin (1968), pp. 19-20.

93

the man was on the other side of the lake and I was not able to get there intime).36

It is surely clear in such a case that the appeal to inability does not presup-pose any admission of wrongdoing. And yet it would be quite natural tosay that the agent in the example offered his inability as an excuse.

Now, there are two ways to go at this point. One might simply insistthat excuses do presuppose wrongdoing, just as Austin claims. Indeed,McConnell himself says this. But then, as McConnell states, it would beinaccurate to say of his example that in it an excuse is being offered. Thiswould mean that what many find natural to say about his example is in factincorrect. I have no real quarrel with this. The moral would be that it is alsoincorrect, even if natural, to say of Montefiore's meeting case and Sinnott-Armstrong's wedding case that an excuse is being offered. Hence there isno grist after all for the mill of the opponent of the claim that "ought"implies "can."

Another way to go here — and one that I prefer — is simply to rejectAustin's claim that excuses presuppose wrongdoing and to assert that whatit seems natural to say in Montefiore's and Sinnott-Armstrong's andMcConnell's cases is in fact correct. For what is an excuse? An excuse isoffered in response to an accusation.37 It is a defense to a charge. It is a denialof responsibility when responsibility has been imputed. This can occur ina variety of settings, including those in which wrongdoing is admitted, butalso (as I mentioned when responding to Widerker's argument inSubsection 3.1.5) including those in which it is not. Given this under-standing of what an excuse is, we may admit that excuses have been offeredin the three cases just mentioned without thereby being committed todenying that "ought" implies "can."

A slight complication here is the following. In the meeting case, don'twe have an admission of wrongdoing? After all, the protagonist has said: "Iought to be at the meeting tonight." But this statement is open to variousinterpretations. First, he may mean simply that he ought in the ideal, non-binding sense to be at the meeting. Or he may mean that he has beencharged with being at the meeting. Or he may mean (and I think this is verycommon) that, although he can (literally) manage to go (and, perhaps, hasa prima facie obligation to go), he cannot do so without serious inconve-nience, to himself or others (and therefore, perhaps, without violating somemore stringent prima facie obligation). If the protagonist means any of these

36 McConnell (1989), p. 438.37 Austin says just this on p. 19 of Austin (1968).

94

things, he says nothing inconsistent with (2.46), the claim that "ought(overall)" implies "can." And if he means none of these, if he means thathe ought (overall) to be at the meeting although he literally cannot be there,then, I say (barring a complication to be discussed next), he is simply mis-taken; for it cannot be that he ought (overall) to do what he literallycannot do.

3.1.9 Self-imposed impossibility

Even if the foregoing discussion of excuses is plausible, though, it may seeminsufficient to salvage the claim that "ought" implies "can." For it seemsthat there just are cases where a person ought (is obligated) to do somethingbut cannot (literally) do it, and these are cases where the person appears tohave no excuse for not performing the action. Here I have in mind cases ofwhat may be called "self-imposed impossibility." Many philosophers haveappealed to such cases in an effort to prove (2.46) false.38 A typical case isthis. Bruce borrows some money on Friday, promising to repay it onSunday. Since nothing else of moral significance conflicts with such repay-ment, he is therefore overall obligated to repay the money on Sunday. Buthe spends all the money on Saturday night, thereby rendering himselfunable to repay it on Sunday. Surely, even though Bruce cannot repay themoney then, he is obligated to do so. To declare otherwise smacks of para-dox,39 for what an easy escape Bruce would then have from the bonds ofobligation — what an easy escape we would all have! We need only disableourselves in the relevant way and, voila, we're off the hook.

This seems to me to constitute the strongest challenge to the claim that"ought" implies "can." A similar argument can also be made against theclaim that "ought" implies "can avoid." It has to do with what may becalled "self-imposed necessity." Suppose that Betty, like Bruce, borrowssome money on Friday, promising to repay it on Sunday. Unlike Bruce,Betty is conscientious. To ensure that she keeps her promise, she takes themoney on Saturday to the room in which she is due to repay it on Sunday,enters the room, locks the door, and throws away the key, thereby thwart-ing, in the manner of Ulysses resisting the Sirens, the temptations ofSaturday-night spending sprees. Let us suppose that this action rendersBetty incapable of not repaying the money on Sunday. Must we thenconclude that it is not true after all that she ought to repay it then? Surely

38 See, e.g., Henderson (1966), p. 106; Robinson (1971), p. 252; Stocker (1971), p. 314;White (1975), p. 149; Richman (1983), p. 85; Sinnott-Armstrong (1988), pp. 116-17.

39 Cf. the discussion of "Suzy Mae" in Powers (1967), pp. 385-6.

95

not. Hence "ought" does not imply "can avoid" any more than it implies"can."

The proper response to this latest challenge requires careful develop-ment.

3.2 IMMEDIATE AND REMOTE OBLIGATION

A crucial component in the account of obligation that I have proposed isthe use of two time indices. The canonical phrase form at issue is this:

(3.26) S ought, at Tin W, to do A at T".

This reflects the fact, discussed in Subsection 2.2.1, that control may beeither direct or indirect. The example given there had to do with mythrowing a brick through a store window. I had control over whether ornot the window broke. Such control was indirect, however, being para-sitic on the direct control that I had over my decision to throw the brick.The control that I had over my breaking the window was hybrid, con-joining the direct control over the decision with the indirect control overthe window's breaking, which was a consequence of that decision.

Let's embellish the story a little. Suppose that breaking the window wasa necessary preliminary to my snatching the jewels, and I snatched them.Then the decision to snatch was something over which I had direct con-trol, the disappearance of the jewels was something over which I had indi-rect control, and the snatching was something over which I had hybridcontrol. But that's not all. At the time I made the decision to throw thebrick I had indirect control over each of the items just mentioned - thedecision to snatch, the disappearance of the jewels, and the snatching. Thus,although (on the account that I favor) one can have direct control only overone sort of thing (decisions) and one can have hybrid control only over onesort of thing (nonminimal actions), one can have indirect control over allsorts of things, including decisions and nonminimal actions, since these canthemselves be among the consequences of decisions.

Now, as I stated in Subsection 2.2.1, this decision-plus-consequencepicture is not essential to my account of obligation; what is essential is thedistinction between direct and indirect control. And this distinction givesrise to a closely related distinction between immediate and remote control.This has to do with the times involved. The canonical phrase form here isthis:

(3.27) S can, at Tin W, do AatV.

96

If T is distinct from (that is, earlier than) T', then S's control over A isremote; otherwise it is immediate.40 Thus we should understand the claimthat "ought" implies "can" as the claim that (3.26) implies (3.27). That is,to put it as before, but now with the times emphasized:

(2.46) if S ought, at Tin W, to do A at Y, then 5 can, at Tin W, do AatV.

This is what my account of obligation implies. And just as control may beimmediate or remote, so too may obligation. If S ought at T to do A at T',and Tis distinct from (that is, earlier than) T', then S's obligation is remote;otherwise it is immediate.

Some philosophers have rejected the distinction between immediateand remote obligation,41 others have accepted it,42 and still many othershave of course been silent on the matter.4 The distinction seems to me offirst importance. Not only does it solve the problems of self-imposedimpossibility and necessity, but it provides the wherewithal for a detailedaccount of the dynamics of obligation - the shifting of obligation and itssupersession. It is to this that I now turn.

3.2.1 Self-imposed impossibility

Consider again the puzzle concerning Bruce, who borrows money onFriday, spends it all on Saturday night, and thus fails to repay it on Sunday.Not only does the distinction between immediate and remote obligationsolve the problem here, it helps explain what we want to say about the case,and that is the following. Prior to the Saturday-night spending, Bruceought to repay the money on Sunday (in part because he can, prior toSaturday night, repay the money on Sunday). After he has spent all themoney, though, and so no longer can repay the money on Sunday, it is nolonger true that Bruce ought to repay the money on Sunday. On the con-trary, he is faced with a new (more restricted) set of options, and his oblig-ations have shifted accordingly (he ought to apologize, perhaps, and so on).

40 On the relation between immediate and remote control, on the one hand, and directand indirect control, on the other, see Subsection 2.2.2, including notes 51 and 52 toCh.2.

41 See, e.g., Prichard (1949), pp. 182-3.42 See, e.g., McKinsey (1975), Castafieda (1977), Goldman (1978), Feldman (1986),

Ch.2.43 Even when they recognize the distinction between immediate and remote control. Cf.

Alvin I. Goldman (1976), pp. 208-9.

97

7",

Figure 3.1

The situation may be pictured as in Figure 3.1 (where Tt is Friday, T2 isSaturday night, T3 is Sunday, and where B, S, R, and A stand for "borrow,""spend," "repay," and "apologize," respectively).44 This picture capturesthe following (among other things): Bruce ought on Friday, all the way upto the spending on Saturday night, to repay the money on Sunday; Bruceought (by default, as it were) at those same times not to apologize on Sundayfor not repaying the money;45 but, after the spending, Bruce ought thento apologize on Sunday for not repaying the money; it is not true thatBruce ought, at these post-spending times, to repay the money onSunday (because he cannot, at these times, do this); nor is it true that heought, at these times, not to repay it (because he cannot, at these times,avoid this).

Some philosophers resist this treatment of cases of self-imposed impos-sibility — even though they accept that the temporal indexing of "ought"and "can" is legitimate — because they wish to insist that (in my example)Bruce still ought on Sunday to repay the money.46 It is not clear to me whythey wish to insist on this. I suspect that it is because they wish to insist thatBruce does wrong in not repaying the money and that they believe that mysort of treatment of "ought" denies this.47 But, if so, they are mistaken, not

44 As in Figure 2.3, the numbers 1-3 on the right denote the value-rankings of the respec-tive worlds. Worlds 2a and 2b are thus of equal value, as are worlds 3a and 3b. That theseworlds be of equal value is stipulated merely for purposes of illustration.

45 By default, because he ought at those times not to act in such a manner that there's some-thing to apologize for.

46 See in particular Sinnott-Armstrong (1988), p. 117.47 Cf. Stocker (1971), pp. 314-15.

98

in insisting that Bruce does wrong in not repaying the money, but in think-ing that my account implies otherwise. On the contrary, when an obliga-tion is not satisfied, a wrong is done, whether or not the obligation isremote. If Bruce ought on Friday to repay the money on Sunday, andBruce doesn't repay the money on Sunday, then Bruce does wrong. Andthe wrong isn't just that he spends the money on Saturday night. True, heought on Friday, and up to Saturday night, not to spend the money onSaturday night, and so his spending is wrong — an immediate wrong, sincethe final time at which he ought not to spend coincides with (or is imme-diately prior to) the time of spending. But it is also true that Bruce oughton Friday, and up to Saturday night, to repay the money on Sunday, andso his not repaying is wrong, too - a remote wrong, since the final time atwhich he ought to repay does not coincide with the (pertinent) time ofnonrepayment. Indeed, in this case it seems that the remote wrong is in away more significant than the immediate wrong. Even though there is asense in which the remote wrong of nonrepayment is committed by virtueof the immediate wrong of spending (the remote wrong would not havebeen committed, in the manner in which it was committed, had not theimmediate wrong been committed), nonetheless the moral primacy restswith the remote wrong, in that (let us assume)48 the spending is wrongbecause and only because the nonrepayment that it necessitates is wrong (thespending would not have been wrong had not the nonrepayment beenwrong).

Although this moral primacy of a remote wrong over the immediatewrong by virtue of which it is committed helps to dramatize the fact thatremote wrongs are genuine wrongs, it is in fact not crucial that the moralprimacy rest with the remote wrong. Suppose, for instance, that Bruce'sspending were to take the specific form of gambling. It might then be thatthe gambling is wrong not just because it necessitates the nonrepaymentbut for some other reason also; it might even be that the gambling is moreseriously wrong than the nonrepayment, quite independently of the factthat it necessitates the nonrepayment. Still the nonrepayment would itselfbe wrong. Indeed, a remote wrong might be wrong only because some-thing else is wrong, and yet it would of course still be a wrong. Perhaps,after spending the money, Bruce hitches a ride to Boston, thereby render-ing himself unable to be in New York, the site of the repayment scheduledfor Sunday. Bruce ought (on Friday) to be in New York (on Sunday); his

48 It is on this assumption that the value-rankings in Figure 3.1 are based. See note 44 tothis chapter.

99

not being there constitutes a remote wrong, even though the wrongness ofthis is wholly parasitic on the wrongness of not repaying the money.49

In sum, then, it will be perfectly true to say on Monday that Bruce oughtto have repaid the money on Sunday and so did wrong in not repaying it,for what this amounts to saying is that there was a time T such that Brucewas obligated at Tto repay the money on Sunday. It will also be true to sayon Monday that Bruce could not have repaid the money on Sunday, if thisis taken to mean that there was a time T (not later than Sunday) such thatBruce could not at T repay the money on Sunday, but not if this is taken tomean that there was no time Tsuch that Bruce could at Trepay the moneyon Sunday. (2.46) is therefore preserved.

3.2.2 The time at which a wrong is done

Some might feel uneasy with the foregoing because it isn't clear just whenremote wrongs are supposed to be done. If Bruce ought on Friday to repaythe money on Sunday, then it is wrong on Friday for Bruce not to repaythe money on Sunday; but if Bruce doesn't repay the money on Sunday,it isn't on Friday that he does wrong in this regard. The (pertinent) nonre-payment occurs on Sunday, after all.50 But the trouble is, it also seems prob-lematic to say that Bruce does wrong on Sunday in this regard, since by thattime it is (according to the proposed account) no longer wrong for him notto repay the money then. So, if he does a wrong — a remote wrong —justwhen is it that this wrong is committed?

It seems to me useful to draw a distinction here between the time atwhich an agent goes wrong and the time at which the wrong occurs. "Dowrong" may be used in either case. The wrong that is done is the (perti-nent) nonrepayment. This occurs on Sunday.51 Thus the wrong is, in thissense, done on Sunday, even though by then it is no longer wrong for Bruceto fail to repay the money. In general, then, we may say this:

(3.28) the wrong constituted by S's not doing A at T' occurs at Tin Wiff

49 I have no precise account to offer of what I have called the "moral primacy" of one oblig-ation or wrongdoing over another, nor of the corresponding concept of one obligationor wrongdoing being "parasitic" on another. I hope that what I say here is intuitivelyplausible.

50 Clearly, nonrepayment occurs at other times, too; the point is that repayment is sched-uled for Sunday.

51 See the last note.

100

(a) Tis identical with T';(b) for some time T", S ought, at T " in W, to do A at T'; and(c) 5 does not, in W, do 4 at T'.

This is intended to capture the sense in which the wrong done by Bruceitself occurs on Sunday. (Here, talk of "doing wrong" can sometimes seema bit forced, for one can conceive of cases where a wrong is "done" in thissense even though the agent is no longer alive. For example, perhaps Sought at Tj to do A at T3 but commits suicide at T2.)

But even if this is, in one sense, when the wrong is done, it is still truethat, in some other sense, Bruce doesn't do wrong on Sunday with respectto the nonrepayment. For the point at which he goes wrong is Saturdaynight, when he spends all the money and thereby ensures the nonrepay-ment on Sunday. To capture this, and to accommodate straightforwardimmediate wrongdoings (where the times at which an agent goes wrongand at which the wrong occurs coincide), it might therefore seem that whatwe should in general say is this: an agent goes wrong with respect to failingto perform some obligatory action either at the time at which the action is"scheduled" to be performed or at the time at which the action is renderedimpossible. That is:

(3.29) S goes wrong, at Tin W, in not doing A at T iff(a) 5 ought, at Tin W, to do A at T\(b) S does not, in W, do A at T; and(c) it is not the case that S can, after Tin W, do A at T'.

This would yield the desired result in cases of immediate wrongdoing(where Tand T' are identical), since no one can change what has alreadytaken place. It would also yield the desired result in the spending case. If welook at Figure 3.1, it seems clear that Bruce goes wrong on Saturday night(this is where he takes a turn for the worse), not just with respect to spend-ing the money on Saturday but with respect to not repaying the money onSunday, since such repayment is rendered impossible at the moment ofspending.

But (3.29) is in fact not adequate to other cases of remote wrongdoing.Consider this case. On Friday, Mandy tells Phil that she will be able to helphim move furniture on Sunday, thereby making a sort of informal com-mitment to do so; on Saturday, however, Mandy promises Bill that she willbabysit his child on Sunday, thereby making a formal commitment to doso. Note that, although Mandy can still on Saturday afternoon help Phil onSunday, she cannot both do that and babysit the child. Let us assume that,

101

Figure 3.2

because of the informal commitment, Mandy ought on Friday to help Philon Sunday, but also that, because of the formal commitment, Mandy oughton Saturday to babysit for Bill on Sunday. Suppose that what Mandy in factdoes is to babysit on Sunday. Where does she go wrong with respect to nothelping Phil move furniture?

The situation may be pictured as in Figure 3.2 (where Tt is Friday, T2 isSaturday, T3 is Sunday, and where T, P, H, and B stand for "tell,""promise," "help," and "babysit," respectively).52 On this picture, Mandyought on Friday to help Phil on Sunday. By taking Track 2b, Mandy doeswrong in not helping Phil. This wrong occurs on Sunday. But Mandy doesnot go wrong on Sunday in not helping Phil, for Track 2b is preferable toTrack 3. Rather, it seems clear, it is at noon on Saturday that Mandy goeswrong in this respect. But this is not what (3.29) implies. Rather, (3.29)implies that Mandy goes wrong on Sunday in this respect, since at no timeprior to Sunday does it become impossible for Mandy to help Phil then.(This case, like the spending case, is supposed to be a case where the imme-diate wrong is wrong only because the remote wrong is wrong. That is, theformal commitment to Bill would not be wrong were it not for the infor-mal commitment to Phil. This is intended to dramatize the fact that remotewrongs are genuine wrongs. But, as before, there is in fact no necessity thatremote wrongs have such primacy in order to be genuine wrongs. If, ratherthan promise Bill to babysit his child, Mandy were instead to kidnap thechild, thereby rendering it obligatory for her to return the child to Bill on

52 Again, the value-rankings of the worlds are provided for purposes of illustration only.

102

Sunday and so not help Phil move furniture then, her not helping Philwould still constitute a remote wrongdoing.)

It is at noon on Saturday that Mandy goes wrong, for that is where shetakes a turn for the worse. Helping Phil is not then rendered impossible,but it is rendered relatively undesirable, so that it is no longer the case thatMandy ought to help Phil. It therefore seems plausible to replace (3.29) bythe following:

(3.29') 5 goes wrong, at Tin W, in not doing A&tT' iff(a) S ought, at Tin W, to do A at T' ;(b) S does not, in W, do A at T' ; and(c) it is not the case that S ought, after Tin W, to do A at T'.

The suggestion here is this. To find out where an agent goes wrong withrespect to not performing an action, retrace his steps back down the trackwhich he in fact took until you reach a juncture, a point of choice, whereit is true that he ought then to perform the action - where, that is, there isa top-level track leading out from that point and on which the act is fea-tured. In Figure 3.2, we need only go back along Track 2b to the point ofchoice at T2 to find a point where it is true that Mandy ought then to helpPhil move furniture. Hence that is where Mandy goes wrong with respectto not helping Phil. Similarly, if we suppose that Bruce takes Track 2b inFigure 3.1, we need only retrace his steps to T2 to find a point where it istrue that Bruce ought then to repay the money. Hence that is where Brucegoes wrong with respect to the nonrepayment.

While I think that what has just been said is very close to the truth, it isstill not quite adequate. And that is because, even if, when a wrong is done,it occurs only once, nonetheless the agent may have^one wrong with respectto it on several occasions. Consider this case. Victor is in a position at To

to visit each of, but not both of, Alf and Bert at T4. He ought to visit Alf.But at T1 he calls Bert on the phone, making Bert so eager to see him thathe now ought to visit Bert. However, at T2 he calls up Alf, making Alf soeager to see him that he now ought to visit Alf once again. And at T3

he calls up Bert again, so that he now ought, once again, to visit Bert.Nonetheless, he visits Alf. The situation may be pictured as in Figure 3.3(where it is assumed, for simplicity's sake, that the third call would not havebeen possible had not the second occurred and the second would nothave been possible had not the first occurred, and where Cu C2, C3t A, andB, stand for "make the first call," "make the second call," "make thethird call," "visit Alf," and "visit Bert," respectively).53 Victor takes

53 As before, the value-rankings of worlds are provided for purposes of illustration only.

103

Figure 3.3

Track 5. The question at issue is this: where does he go wrong in notvisiting Bert at T4? (3.29') implies that he goes wrong at T4 in this respect,and I don't wish to deny that this is so. After all, his immediate obligationis to visit Bert. But what (3.29') doesn't reveal is that Victor also goes wrongearlier in this respect. He goes wrong at T2. For from Tt to T2 he ought tovisit Bert, and yet at T2, by calling up Alf, he deviates from this obligation.This fact is not accounted for by (3.29').

For another example, suppose that, instead of Track 5, Victor takesTrack 4a. Where does he go wrong in not visiting Alf at T4? At T4, butalso at Tt\ for at both these points there is a top-level track on which visitingAlf at T4 is featured and which Victor chooses not to take.

To capture this possibility of multiple lapses, of multiple deviations fromthe strait and narrow with respect to one and the same action, I proposethat we accept this revision of (3.29'):

(3.29") S goes wrong, at T in W, in not doing A at T' iff(a) S ought, at Tin W, to do A at T' ;(b) S does not, in W, do A at T ' ; and(c) it is not the case that S ought, just after Tin W, to do A at T'.54

54 By clause (c) I mean that, for any time T* after T such that S ought, at T* in W, to doA at T\ there is an earlier time T** after T such that it is not the case that 5 ought, atT** in IV, to do A at T.

104

In general, then, one goes wrong whenever one takes a turn for theworse. Whenever one does this one of course commits an immediatewrong, but one may also be committing a remote wrong. Thus no remotewrong can be done without an immediate wrong being done (Bruce'swrongly not repaying the money can be traced to his wrongly spending;Mandy's wrongly not helping Phil can be traced to her wrongly promisingto babysit for Bill), although it sometimes happens that an immediate wrongis wrong only because some remote wrong is wrong (as in the casesdepicted in Figures 3.1, 3.2, and 3.3).

3.2.3 The time at which an obligation is satisfied

Just as commissions of wrongs are sometimes remote, so too are satisfac-tions of obligations. Recall the case of self-imposed necessity where Bettylocks herself in a room on Saturday, thereby ensuring repayment onSunday. After she throws away the key, she is no longer obligated to repaythe money, for she cannot avoid doing so. Nonetheless, in repaying themoney, she does what she ought. The situation may be pictured as in Figure3.4 (where Ti is Friday, T2 is Saturday, T3 is Sunday, and where B, T, andR stand for "borrow," "throw away," and "repay," respectively).

I propose that we treat this matter in a manner parallel to that in whichremote wrongdoing has been treated. Just as "do wrong" does double duty,so too does "satisfy an obligation." Just as the time of going wrong may bedistinguished from the time at which the wrong occurs, so too the time ofmaintaining an obligation may be distinguished from the time of its fulfill-ment. In Figure 3.4, the time at which Betty fulfills her obligation to repaythe money is Sunday (T3), but she maintains this obligation only up untilSaturday (T2). In general, then, I propose the following:

7",

Figure 3.4

105

(3.30) S's obligation to do A at T is fulfilled at Tin W iff(a) Tis identical with T';(b) for some time T", S ought, at T " in W, to do A at T'; and(c) 5 does, in W, do 4 at T';

(3.31) 5 maintains, from Tuntil T' in H , his obligation to do A at T " iff(a) S ought, from T until T" in W, to do 4 at T";(b) S does, in W, do A at T"; and(c) it is not the case that S ought, just after T' in W, to do A at T".

Let us apply these formulae to Figure 3.2. Suppose now that Mandy takesTrack 3; that is, that she doesn't babysit the child but does help Phil movefurniture. According to the formulae, Mandy thereby fulfills (on Sunday)and maintains (up until Saturday) an obligation to help Phil move furni-ture. Or again, with respect to Figure 3.3: suppose now that Victor takesTrack 4b. In this case he fulfills at T4 his obligation to visit Bert, an oblig-ation which he maintains from T1 until T2 and from T3 until T4.

Return to Figure 3.2. It would be accurate to observe that Mandy couldhave done worse - she could have taken Track 5 - and that she would thenhave satisfied no obligation at all. But we should not be misled by this obser-vation. It would be a mistake to infer that one satisfies an obligation as longas one avoids the worst track open to one or that one satisfies no obligationif one takes the worst track open to one. Neither of these is necessarily so.Consider the sort of case represented in Chart 3.1, where the numberingreflects the ranking of the options, or tracks, open to an agent, S. In such acase, if S takes Track 2, he satisfies no obligation, even though he couldhave done worse. (This claim will be qualified in Chapter 4.) On the otherhand, if S takes Track 4, he does satisfy an obligation, namely, the obliga-tion to do A. (This claim will be discussed further in Chapter 6.) The moralis that we should not necessarily be impressed if someone does somethingthat he ought, as long as he also does not do something else that he ought;for some ways of failing to do what one ought are superior to some ways ofdoing what one ought. Thus, even though it is accurate to say of Mandy that,in taking Track 3 in Figure 3.2, she satisfies her obligation to help Phil, thisneed not constitute praise for her action; for it does not tell the whole story.

1:2:3:4:

A&~A&~A&A&

B&...-B&...B&...

~B&...

Chart 3.1

106

3.2.4 The cancellation of obligation

I have said that, when an obligation is not satisfied, a wrong is done. Thisis clearly true of immediate obligations, but some may balk at accepting thisclaim when it is remote obligations that are at issue. Perhaps it will be agreedthat Bruce does wrong in not repaying the money and that Mandy doeswrong in not helping Phil (or in not babysitting the child, depending onwhich version of that story we're working with), but aren't there othercases where it is accurate to say that a remote obligation is not satisfied andyet no wrong is done? What if the reason for Bruce's not repaying themoney had not been that he spent it all but that he had been mugged (orstruck by lightning) on his way to repaying it? What if the reason forMandy's not helping Phil had not been that she was fulfilling a wronglymade promise to babysit for Bill but that Phil had released her on Saturdayfrom her informal commitment to help him (or that Phil had died from aheart attack on Saturday)? Isn't the pertinent obligation canceled in suchcases, and doesn't this show that not all failures to satisfy an obligationconstitute wrongdoing?

I grant the intuitive plausibility of this objection. It is interesting to notethat talk of such cancellation of obligation, if taken literally, would appearto presuppose the possibility of remote obligation; for how in general couldan obligation to do A at T2 be canceled at T1 unless it were already in exis-tence at Tj? Nonetheless, I reject the possibility of such cancellation.Although it may be, for instance, that Bruce would not have been obligat-ed to repay the money if he had been mugged, or that Mandy would nothave been obligated to help Phil if Phil had released her from her commit-ment, this doesn't imply that their obligations would thereby have beencanceled. For consider Bruce. Either he could on Friday have avoided themugging (or lightning) on Saturday or he couldn't. (It is extremely likelythat he could; he could have walked down Elm Street rather than OakStreet, and then he wouldn't have been mugged or struck by lightning.)Similarly, either Mandy could on Friday have avoided Phil's releasing herfrom her commitment (for example, by pleading with him not to be so gen-erous, or condescending, or whatever) or she couldn't; or again, eitherMandy could have avoided (prevented) Phil's having a heart attack or shecouldn't (and perhaps here it is likely that she couldn't). In every case whatwe should say is this. If the agent could not have avoided the event thatallegedly canceled the original obligation, then he (or she) did not have theoriginal obligation after all, although it may have been quite reasonable (forhim or someone else) to believe that he did. (If Bruce could not on Friday

107

have avoided the mugging or the lightning, then there was no world acces-sible to him on Friday in which he repaid the money on Sunday, and so hewas not obligated on Friday to repay the money on Sunday, although itmay have been quite reasonable to believe that he was. So too for the othercases.) On the other hand, if the agent could have avoided the "canceling"event, then, if he was already obligated to perform the pertinent action, hemay well have been obligated to avoid the "canceling" event itself. (IfBruce could on Friday have avoided the mugging, and if Bruce was obli-gated on Friday to repay the money on Sunday, then Bruce was obligatedon Friday to avoid the mugging on Saturday. If Mandy could on Fridayhave avoided Phil's releasing her from her commitment,55 and if Mandywas obligated on Friday to help Phil on Sunday, then she may well havebeen obligated on Friday to avoid Phil's releasing her from her commit-ment on Saturday - although not necessarily: if the deontically best thatMandy could on Friday do on Sunday was to help Phil, whether or not Philreleased her from her commitment, then Mandy was not obligated onFriday to avoid this release.) Thus every failure to satisfy an obligation doesindeed constitute wrongdoing.56

But, it may be objected, even if Bruce could literally have avoided themugging, he surely could not have known he would be mugged, or evenhad reason to believe that he would be. Or, even if he did have reason tobelieve this, surely it is not he but his muggers who did wrong. To sayotherwise is to blame the victim rather than the perpetrators of a wrong-doing — a reprehensible practice.

The proper response is this. I am not blaming Bruce in this version of thecase; I am merely saying that he did wrong. Doing wrong is not sufficient

55 Recall from Subsection 2.2.1 that others' actions enjoy no special status among eventswhen it comes to their accessibility.

56 Cf. Thomason (1981), pp. 178-9. Contrast, perhaps, McConnell, who says that debtscan sometimes be annulled, and gives this illustration (McConnell 1993a, p. 80): "If youhave a financial debt to a person and she steals from you the amount of money owed(or more), it seems that you no longer owe her any money." I would say here that, whilethere may be a sense in which, prior to the theft, you owed your creditor the money,it is not the case that you had an overall moral obligation to pay it to her, unless you hadan overall moral obligation to avoid her stealing it from you. Did you have an obliga-tion to reimburse her before she stole it? If so, you had, ipso facto, an obligation to avoidher reimbursing herself in this way; if not, then, unless for some other (and presumablyrather odd) reason you had an obligation to avoid the theft anyway, it was permissiblefor you so to act that she reimbursed herself in this way (even if not permissible for herso to act in this way), and so permissible for you not to see to her reimbursement in morecustomary fashion by paying her yourself.

108

(nor is it necessary) for being to blame. The two must not be confused.57

And anyway I am not insisting that Bruce did wrong in not avoiding themugging. I am saying that, if he was obligated to repay the money, then hedid wrong in not avoiding the mugging. Recall the distinction, discussedin Section 1.4, between theories according to which what is obligatory iswhat is best (in some substantive sense) and theories according to whichwhat is obligatory is what is probably best. It was best (let us assume) forBruce to walk down Elm Street rather than Oak Street; hence, accordingto the first sort of theory, that's what he ought to have done. But it mightwell be that it was (at the time) probably better for Bruce to walk downOak Street, in which case, according to the second sort of theory, he didwhat he ought, even though that involved being mugged and notrepaying the money. But, if so, then Bruce was not obligated to repay themoney after all (although it might be true that he was obligated to repay themoney if he didn't walk down Oak Street - a matter of conditional obliga-tion, to be discussed in Chapter 4).58

Another, similar case might help to drive home this point. W. D. Rossdiscusses returning a borrowed book through the mail. At one point59 hedeclares that the obligation in question is satisfied only if the book in factreaches the hands of the person from whom it was borrowed, although theagent may well not be to blame if the obligation is not satisfied (perhapsblame will rest with some employee of the postal service). Later Rossdeclares that the obligation in question is satisfied if the agent "exerts him-self" to return the book, even if the book gets lost in the mail. I prefer thefirst view, but that is not the present point. The present point is simply this.If the agent is obligated so to act that the book arrives at its final destina-tion, then he is obligated not so to act that the book gets lost in the mail; ifthe agent is not obligated not to use the mail in this way, then he is notobligated so to act that the book arrives at its final destination. I am notinsisting (although I believe) that the agent is obligated so to act (given thathe can so act) that the book arrives at its final destination, but only that, ifhe is so obligated, then he ought to avoid getting the book lost in the mail.

57 See Zimmerman (1988a), pp. 40ff. and 65.58 It might be thought that it was both probably best for Bruce to walk down Oak Street

and probably best for him to repay the money, so that each is something that he oughtto have done. But this is either to take an act-based, rather than world-based, approachto obligation, or to take a world-based approach where the worlds in question are onlyprobably rather than actually accessible. Both approaches have already been rejected; seeSubsections 2.1.2 and 3.1.1 above.

59 Ross (1930), p. 42ff.60 Ross (1939), p. 153ff.

109

Similarly, if Bruce is obligated to repay the money, then he is obligated toavoid getting mugged. This is not to blame Bruce if he gets mugged, justas it is not to blame the book-borrower if the book gets lost in the mail.Nor is it to let the mugger, or the postal worker, off the hook. These latteragents may well have done wrong themselves and may well be to blame forwhat they have done. But, if either of these judgments about the muggerand the postal worker is correct, their being so is wholly independent of thejudgments to be made about Bruce and the book-borrower.

In denying that obligation can be canceled, I have been arguing thatagents can sometimes be obligated in cases where many may think that theyare not. But uneasiness with my treatment of remote obligation and wrong-doing may come, as it were, from the opposite direction. There are caseswhere I am committed to saying that the agent is not obligated to performa certain action but where it might seem that we should say that he is obli-gated to do so. For instance, I have said that Bruce's spending on Saturdaynight renders it impossible for Bruce to repay the money on Sunday, so thatit is not true that he ought, from Saturday night on, to repay the moneyon Sunday. But what if Bruce were luckily to stumble across someextra money on Sunday morning? What if he were to win the lottery?Surely then he ought, as of that moment, to repay the borrowed moneyafter all.

Here the correct response is simply this. Either Bruce can so act onSaturday night that, despite spending the money then, he will stumbleacross some extra money on Sunday morning, or he cannot. (See Sub-section 2.2.3 above.) If he can, then his spending on Saturday night doesnot after all render it impossible for him to repay the money on Sunday,and so he still ought, after spending, to repay the money. If he cannot, thenthis is a case of self-imposed impossibility, just as originally described andintended.61 (Notice that it might be true in some sense that Bruce might winthe lottery even though he cannot, where the lottery is of the pre-imprint-ed ticket variety. This just means that, for all that we or anyone else knows,Bruce has the winning ticket.)

A final case is this. Suppose that Pru makes a promise that she cannot,and never could, keep. (A variation on this: suppose that Pru misrepresentsherself - as a competent lifeguard, say.) My account implies that Pruwas never obligated to keep her promise, and hence that Pru committedno wrong with respect to not keeping her promise. Surely this is counter-intuitive.

61 Cf. a variation on this sort of case in Sinnott-Armstrong (1988), p. 119.

110

Here I would reply that where Pru goes wrong is with respect to the ini-tial act of misrepresenting herself (her intentions, her abilities). This isavoidable and ought to be avoided. Sinnott-Armstrong agrees that Pruought not to make the promise but argues that it is also true that, havingmade it, she ought to keep it, on the grounds that, if it were not true thatPru ought to keep the promise, then there would be no reason why sheought not to make it in the first place.62 Not so. There is very good reasonwhy she ought not to make the promise, even though it is not the case thatshe ought to keep it, and that is that she cannot keep it. She is misrepresent-ing herself; she is misleading others; she is creating false beliefs and certain-to-be-frustrated expectations. Under the circumstances, this is (we mayassume) wrong.

3.2.5 The supersession of obligation

That obligation cannot be canceled does not imply that it cannot be super-seded. Obviously it can be. Every obligation is eventually superseded, ifonly because, if 5 ought at T to do A at T', T' will eventually arrive. Theinteresting cases of supersession, however, are those involving remoteobligation, where the obligation at T to do A at T' is superseded prior to

r.One thing that the previous discussion reveals is that obligation, even

when overall and not merely prima facie, can be overridden. Such overrid-ing is dynamic and to be distinguished from the sort of static overridinginvolved in one conditional obligation's overriding another and in oneprima facie obligation's overriding another (issues to be discussed in thenext two chapters). For example, Mandy's obligation to help Phil movefurniture is overridden by her obligation to babysit. In general we may say:(3.32) S's (overall) obligation to do A at T" is overridden, at Tin W, by

his (overall) obligation to do B at T' iff(a) 5 ought, just prior to Tin W, to do A at T';63

(b) it is not the case that S ought, just prior to Tin W, to do B at T';(c) S ought, at Tin W, to do B at T;(d) it is not the case that 5 ought, at Tin W, to do A at T'; and(e) S can, at Tin W, do A at T.

62 Sinnott-Armstrong (1988), pp. 118-19.63 By this I mean that, for any time T* prior to T such that it is not true that S ought, at

T* in W, to do A at T', there is a later time T** prior to T such that S ought, at T**in W, to do A at T.

I l l

T4

Figure 3.5

Notice four things. First, if S's obligation to do A at T' is overridden, itnonetheless survives, in the sense that, if he does A at T", he at least satis-fies that obligation. (As pointed out in the last section, however, this maybe no cause for celebration.)

Notice also that it is possible that an obligation that has been overriddenbe restored, in the sense that the agent may once again become such that heought overall to satisfy it. For example, if Mandy were to murder Bill's childon Saturday night, it might be that she ought then once again to help Philmove furniture on Sunday. The situation might be pictured as in Figure 3.5(where Tt is Friday, T2 is now Saturday noon, T3 is now Saturday night,and T4 is Sunday, and where T, P, H, and B are to be understood as beforein Figure 3.2, with Mstanding for "murder"). It is in principle possible foran obligation to be overridden and restored an indefinite number of times.If this happens, though, each moment of overriding and restoration marksa fresh turn for the worse by the agent.

Notice, thirdly, that, although in many cases of the overriding of anobligation to do A by an obligation to do B, the agent 5 cannot do both Aand B, this is not necessarily the case and is not implied by (3.32). Althoughit is true of Mandy (in that she cannot both help Phil move furniture andbabysit for Bill), it is not true of Gertrude in the following case. At Tt

Gertrude ought, from gratitude, to send Lucy lilies at T3; it is not the casethat she ought to send her daisies - they would be a poor second; and it is

112

emphatically not the case that she ought to send both lilies and daisies, sincethis would greatly disturb Lucy (bringing back certain horrible childhoodmemories). Suppose that, at T2, Gertrude spills weed killer on the lilies thatshe had intended to deliver to Lucy at T3. This changes things. Now sheought to send Lucy daisies, for the lilies have shriveled. The obligation tosend lilies has been overridden by the obligation to send daisies. Gertrudecan, of course, still send both lilies and daisies, but this remains somethingthat she clearly ought not to do.64

Notice, finally, that not all supersession of a remote obligation consistsin its being overridden. When Bruce's obligation to repay the money onSunday is superseded by his spending it on Saturday, the obligation is notoverridden; for it does not survive but is rather extinguished, in that hecannot any longer satisfy it. So too with Betty's obligation to repay themoney; it is superseded by her throwing away the key on Saturday, but itis extinguished rather than overridden. We may in general say:

(3.33) S's obligation to do A at T ' is extinguished at Tin W iff(a) S ought, just prior to Tin W, to do A at T'; and(b) either 5 cannot, at T in W, do A at T', or 5 cannot, at T in W,

refrain from doing A at T'.

Once extinguished, an obligation cannot be restored. Notice that the ful-fillment or nonfulfillment of an immediate obligation trivially counts as itsextinction. But extinction can occur remotely, too, as in Bruce's case(where his spending renders him unable to repay the money) and in Betty'scase (where her throwing away the key renders her unable to refrain fromrepaying the money).

64 It might be objected that Gertrude cannot any longer send lilies that are in a suitable con-dition, and that it is this obligation that has been overridden; thus it is true after all that,when an obligation has been overridden by another, not both can be satisfied. This isconfused. I grant that Gertrude had an obligation at Tls one that she no longer has at T2,to send unshriveled lilies; but an obligation to send unshriveled lilies entails an obligationto send lilies (see (2.43'"))' a nd so my remarks stand.

113

Conditional obligation

4.1 VARIETIES OF CONDITIONS ON OBLIGATION

So far the focus of inquiry has been the conceptual structure of statementsof the form "5 ought [at Tin W] to do A [at T ' ]" and various related state-ments. These are intended to be understood as statements of uncondition-al overall moral obligation. But statements of conditional overall moralobligation are just as common. These have the form "S ought to do A, if..."It is the purpose of this chapter to provide an account of the conceptualstructure of such statements.

The most complete account of conditional obligation that has yet beengiven is that provided by Fred Feldman.1 I shall not seek to reproducethe full details of this account here but wish rather to build on its insights.However, a brief overview of some of Feldman's findings is in order.

Feldman notes that statements of the form "S ought to do A, if..." areambiguous. He distinguishes four principal interpretations. The first inter-pretation treats the statement as one involving material implication. Usingthe symbolism introduced earlier, we may represent such statements ashaving this form:

(4.1) p D O(A).

Feldman remarks on a number of features of such statements. Here I willdraw attention to just one, that which he calls "factual detachment."2 Suchdetachment consists in the obligation's obtaining unconditionally when thecondition in question is satisfied. Statements of the form of (4.1) allow forsuch detachment, in that, given both (4.1) and

(4.2) p,

1 Feldman (1986), Ch. 4.2 He borrows this term from Greenspan (1975), Section 2.

114

we may of course infer

(4.3) O(A).

Feldman also remarks that he is doubtful that any sentence of ordinaryEnglish in fact represents the assertion of a statement of the form of (4.1).3

This may be, but it seems to me that such statements as that mentioned byAlan White and discussed in Subsection 3.1.3, namely,

(4.4) we ought to deal with this matter now, if we can,

are perhaps open to such interpretation. (Remember that (4.1) is equiva-lent to

(4.1') ~pvO(A).

It seems to me that (4.4) might plausibly be recast as

(4.4') either we ought to deal with this matter now or [it's not the case thatwe ought, because] we cannot.)

Nonetheless, it may be that such statements are really just subjunctive con-ditionals in disguise.

The second interpretation treats statements of the form "S ought to doA, if..." as involving strict implication. In symbols:

(4.5) p=>O(A)

or, equivalently,

(4.5') D ( ? D O(A)).

Here, too, factual detachment applies. And here there are, I think, manyclear uses of statements of this sort. For instance, each of the followinginvolves strict implication:

(2.37'c) if C(A) & C(~A) &~P(~A), then O(A);

(2.43'") if ~C{A & ~£) & C(~B), then, \£O{A), then O(£);

(2.44') if S ought to do A, then S ought so to act that S ought to do A.

In addition, as Feldman notes,4 a utilitarian will typically intend the fol-lowing statement to be so understood:

(4.6) if it would maximize utility, then he ought to do it.

3 Feldman (1986), pp. 78-9.4 Feldman (1986), p. 79.

115

Similarly, a Kantian will typically intend for the following statement to beso understood:

(4.7) if it would violate the Categorical Imperative, then he ought not todoit.

The third interpretation treats statements of the form "S ought to do A,if..." as involving a subjunctive conditional. This may be symbolized as

(4.8) p > O(A)

and understood to say: "if it were the case that/?, then 5 would be obligat-ed to do A " Clearly, factual detachment once again applies. Equally clear-ly, many English sentences typically represent assertions of such statements.Feldman gives this example:5

(4.9) if Smith were to win a million in the lottery, he would have an oblig-ation to give generously to charity.

Another example would be this:

(4.10) if Tom is (or were) a crook, then you ought to vote for Dick.

Let us embellish this last example, for it will prove instructive. Supposethat there are three candidates for election, Tom, Dick, and Harry, and thatthey are to be ranked in that order.6 Let us therefore suppose that youroptions are to be ranked as in Chart 4.1 (where T, D, and H stand for "votefor Tom," "vote for Dick," and "vote for Harry," respectively). In thatcase, what (4.10) amounts to is this. If Tom were a crook, then your optionswould not be ranked as in Chart 4.1 but in some other way, such as in Chart4.2. In other words, your options would be reordered, deontically.

1: T & ~D & ~H...2: ~T & D & ~H...3: ~T & ~D & H...4: ~T & ~D & ~H...

Chart 4.1

1: ~T & D & ~H...2: -7 (5 ~D & H...3:~T& ~D & ~H...4: T & ~D & ~H...

Chart 4.25 Feldman (1986), p. 85.6 The example is borrowed from Jackson and Pargetter (1987), pp. 77—8.

116

Contrast this case with another. Suppose that

(4.11) if you don't vote for Tom, then you ought to vote for Dick.

It is clear that this is not to be analyzed in terms of a subjunctive conditional.For (4.11) does not amount to this: if you don't (or weren't to) vote forTom, then your options will (or would be) reordered as in Chart 4.2.What's at issue in (4.11) is quite different from what's at issue in (4.10);whereas (4.10) warrants factual detachment, (4.11) does not. That is, sup-pose that it turns out that Tom is a crook; then (4.10) implies that you oughtto vote for Dick. But suppose that you won't in fact vote for Tom; it wouldnot follow from (4.11) that you ought to vote for Dick. You ought to votefor Tom, not Dick. To say that you ought to vote for Dick just because youwon't vote for Tom is to make things either impossibly difficult for you orridiculously easy. It makes things impossibly difficult if the claim that youought to vote for Tom is retained; for, on the assumption that you cannotboth vote for Tom and vote for Dick, you would then be faced with a moraldilemma (and, whatever one thinks about the theoretical possibility of suchdilemmas, surely no one thinks that they are so easily generated - moreon this in Chapter 7). It makes things ridiculously easy if the claim thatyou ought to vote for Tom is abandoned; for you cannot change thefact that you ought to vote for Tom simply by not voting for him (com-pare the objection concerning self-imposed impossibility discussed inChapter 3).

Insofar as (4.11) does not warrant factual detachment, it is to be distin-guished not just from statements of the form of (4.8) but also, of course,from statements of the form of (4.1) and (4.5). (4.11) concerns what I shallcall conditional obligation proper, and it is to the analysis of this that I nowturn.

4.2 THE ANALYSIS

There is no doubt that statements of conditional obligation proper play alarge role in pronouncements about morality. Examples abound:

(4.12) if Pru is going to fail to keep her promise, then she ought not tomake it in the first place (although she ought both to make it andto keep it);

(4.13) if Charlie changes lanes, then he ought to accelerate (although heought neither to change lanes nor to accelerate);7

7 This example is borrowed from Goldman (1978), p. 186.

117

(4.14) if Audrey administers medicine B on Monday, then she ought toadminister it on Tuesday too (although she ought in fact to admin-ister medicine A instead on both Monday and Tuesday);8

(4.15) if Sally sins, then she ought to repent (although she ought not to sinin the first place);

(4.16) if Alan fails to attend the 10:00 A.M. meeting on the first floor, thenhe ought to attend the 10:00 A.M. meeting on the second floor(although what he ought to do is attend the meeting on the firstfloor and not the one on the second);

and so on. In every case, we have a statement that (if used typically) is to beunderstood as a statement of conditional obligation (proper).9

I propose (following the lead of others)10 that conditional obligation beanalyzed in a manner that parallels the analysis already given for uncondi-tional obligation. The main gist of the latter was this: S ought to do A if andonly if (a) 5 can do A, (b) 5 can refrain from doing A, and (c) any accessi-ble world in which S does not do A is deontically inferior to some accessi-ble world in which S does A. More precisely:

(I) S ought, at T in W, to do A at V iff(a) there is a world W' such that W' is accessible to 5 from Wr at T and 5

does A at T ' in W'\(b) there is a world W" such that W" is accessible to S from P'Fat Tand

S does not do A at T ' in W"; and

8 This example is borrowed from Feldman (1986), p. 52.9 Henceforth, I shall usually omit the designation "proper."

10 In particular that in Feldman (1986), Ch. 4. It is perhaps worth noting explicitly heresomething to which Feldman pays scant attention: that statements of conditional oblig-ation proper cannot be satisfactorily analyzed by means of a material conditional withinthe scope of the obligation operator. For example, (4.11) cannot be satisfactorilyanalyzed as follows (where T and D stand for "vote for Tom" and "vote for Dick,"respectively):

O{~T D D).

For this statement is to be understood (see Subsection 2.3.1) as equivalent to the fol-lowing:

O~(~T&~D);

and this follows directly from the truth of

O(T).

Yet it is clear that (4.11) doesn't follow simply from the fact that you ought to votefor Tom. Compare the discussion of Ross's paradox in Subsection 2.3.2.

118

(c) for all worlds W" such that W" is accessible to S from Wat Tand Sdoes not do A at T ' in H7", there is a world Pi7' such that(1) W' is accessible to S from Wat T,(2) S does 4 at T ' in W7',(3) the deontic value for S at T of H7' is greater than the deontic value

for S at T o W , and(4) there is no world W'" such that

(i) W'" is accessible to S from Wat T,(ii) S does not do A at T ' in W7'", and(iii) the deontic value for S at T of W"' is greater than the deon-

tic value for S at T of Wf.

Conditional obligation is to be analyzed in analogous fashion. I propose,roughly, that S ought to do A, on the condition that p is true, if and only if(a) S can do A while p is true, (b) S can refrain from doing A while p is true,and (c) any accessible j?-world in which S does not do A is deonticallyinferior to some accessible p-world in which S does A. More precisely:

(XIII) S ought, at T in W, to do A at T", on the condition that p is trueiff(a) there is a world W' such that W' is accessible to 5 from Wat T

and S does A at T ' in W' and^ is true in W'\(b) there is a world W" such that H/~" is accessible to S from WatT

and S does not do A at T ' in W " andj? is true in W"; and(c) for all worlds W" such that H7"" is accessible to 5 from WatT and

5 does not do A at T' in W" avAp is true in W", there is a worldWr such that(1) W7"' is accessible to S from W at T,(2) S does ,4 at T ' i n W7',(3) p is true in W\(4) the deontic value for S at T of W' is greater than the deontic

value for 5 at TofW", and(5) there is no world W'" such that

(i) W'" is accessible to S from W at T,(ii) 5 does not do A at T ' in H7"",(iii) p is true in H7"'", and(iv) the deontic value for S at T of W'" is greater than the

deontic value for S at TofW'.

Applied to (4.11) and Chart 4.1, what this analysis of conditional obliga-tion tells us is that, given the restriction to ~ T-worlds, Option 2 is deonticallybest. This is of course consistent with saying that Option 2 is not the best

119

option, that is, that the best world is not a ~T-world.11 Morever, it is ofcourse consistent with saying that, even if the actual world is a ~T-world,the best world is not a D-world; hence, factual detachment is not validat-ed. Thus we cannot validly infer an obligation to vote for Dick simply fromthe fact that (4.11) is true and that you will in fact not vote for Tom.

Similar remarks pertain to the other examples. Concerning (4.12): thebest world accessible to Pru in which she fails to (keep a promise to) do Ais one in which she doesn't promise to do A; nonetheless, the best worldaccessible to Pru is one in which she both promises to do A and does A.Concerning (4.13): the best world accessible to Charlie in which hechanges lanes is one in which he accelerates; nonetheless, the best worldaccessible to Charlie is one in which he neither changes lanes nor acceler-ates. Concerning (4.14): the best world accessible to Audrey in which sheadministers medicine B on Monday is one in which she administers it onTuesday also, although the best world accessible to her is one in which sheadministers medicine A on both Monday and Tuesday. Concerning (4.15):the best world accessible to Sally in which she sins is one in which sherepents, although the best world accessible to her is one in which shedoesn't sin. And concerning (4.16): the best world accessible to Alan inwhich he fails to attend the meeting on the first floor is one in which heattends the meeting on the second floor, although the best world accessi-ble to him is one in which he attends the meeting on the first floor and notthe meeting on the second floor.

Siblings of (XIII) may be generated in a manner analogous to that inwhich siblings of (I) were generated. For example, just as (XIII) is analo-gous to (I), so the following is analogous to (II):

(XIV) 5 may, at T in W, do A at T', on the condition that p is true iff[as in (XIII), except that there is no corresponding subclause (c5)and subclause (c4) goes as follows:the deontic value for 5 at TofW" is no greater than the deonticvalue for S at T of W'.]

Mercifully, I shall refrain from giving explicit statements of the analoguesto (III)—(XII). Nonetheless, one further sibling of (XIII) deserves explicitrecognition, for it has no direct analogue among (I)-(XII). Consider thisstatement:

(4.17) you ought to take that item with you only if you pay the shop-keeper.

11 Such talk of a "best world" is rough. (See note 15 to Ch. 2.)

120

As typically uttered, this seems to me equivalent to the following:

(4.17') you ought not to take that item with you if you don't pay theshopkeeper.

And so in general I propose the following:

(XV) 5 ought, at T in W, to do A at T", only on the condition that p istrue iff S ought, at T in W, not to do A at T", on the conditionthatp is not true (that is, on the condition that ~p is true).

Statements of biconditional obligation are of course possible. Consider thisexample:

(4.18) you ought to go to the party if and only if you accept theinvitation.

This is equivalent to the conjunction of the following two statements:

(4.18a) you ought to go to the party if you accept the invitation;(4.18b) you ought not to go to the party if you don't accept the

invitation;

and it leaves open whether you ought to go, ought not to go, or neitherought to go nor ought not to go (and so, too, with respect to accepting).

Three other features of conditional obligation proper should be noted.First, in addition to its failure to validate factual detachment, conditionalobligation also fails to validate what Feldman calls "augmentation."12 Thatis, from the truth of

(4.19) 5 ought to do A, on the condition that p is true

we may not in general infer

(4.20) S ought to do A, on the condition that p & q is true.

You ought to vote for Dick if you don't vote for Tom; but you ought tovote for Harry if you don't vote for Tom and you don't vote for Dick.

Secondly, however, conditional obligation does validate a restrictedform of detachment that may be called "necessity detachment."13 That is,from the truth of (4.19) and that of

(4.21) S cannot so act that p is not true,

12 Feldman (1986), p. 91.13 See Feldman (1986), pp. 90-1; Greenspan (1975), p. 265.

121

we may indeed infer the truth of

(4.3') 5 ought to do A.

For if S cannot so act thatp is not true, then all worlds accessible to S arep-worlds, and so the deontically best worlds accessible to S are p-worlds; thus,if the best accessible p- worlds are yl-worlds, then the best accessible worldsare y4-worlds. Suppose, for example, that Charlie cannot avoid changinglanes (perhaps because some other car is physically forcing him to do so).In that case, it is inaccurate to state that he ought neither to change lanesnor to accelerate. On the contrary, given that

(4.13') if Charlie changes lanes, then he ought to accelerate,

we may infer

(4.22) Charlie ought to accelerate.

Finally, conditional obligation also validates what Feldman calls"deontic detachment."14 That is, from the truth of (4.19) and

(4.23) 5 ought so to act that p is true,

we may indeed infer (4.3'). For if S ought so to act that p is true, then thedeontically best worlds accessible to 5 arep-worlds; and so, again, if the bestaccessible p-worlds are y4-worlds, then the best accessible worlds are A-worlds. Suppose, for example, that you ought to accept the invitation; thenfrom (4.18a) we may infer that you ought to go to the party. On the otherhand, if you ought not to accept the invitation, then from (4.18b) we mayinfer that you ought not to go to the party.

4.3 CERTAIN DEONTIC PARADOXES

Various so-called deontic paradoxes have already been discussed. InSubsection 2.3.2 there was discussion of those paradoxes known as Ross'sParadox, the Good Samaritan Paradox, and the Paradox of the Knower; inChapter 3 a paradox concerning self-imposed impossibility (namely, thatsuch impossibility can rid one of certain obligations) was addressed.15 Allthe other standard paradoxes have to do with conditional obligation, andso I propose now to address some prominent examples.16

14 Feldman (1986), p. 91. Again, he is here following Greenspan (1975), Section 2.15 Here the paradox known as the Suzy Mae Paradox is relevant. See note 39 to Ch. 3.16 Much of what follows is handled in a similar fashion by Feldman in Feldman (1986),

(1989a), (1989b), and (1990).

122

Let me begin, however, with a type of case that, as far as I know, has notreceived discussion before. Suppose it is true that

(4.24) you ought to pay the shopkeeper if you take that item with you.

According to certain accounts of the logic of "ought," 7 this is equivalentto its being the case that

(4.17') you ought not to take that item with you if you don't pay the shop-keeper.

Such contraposition may seem unobjectionable at first, and yet it seems oninspection that (4.24) can be true while (4.17') is false. Consider a casewhere the item in question is food which your child desperately needs andfor which you can afford to pay. We may assume that (4.24) is true. Mightit not be, though, that, because of your child's need, you ought to take thefood even if you don't pay for it? If so, (4.17r) is false (as an expression ofoverall obligation; it may be true as an expression of primafacie obligation).

The account of conditional obligation that I have given allows for thepossibility that (4.24) be true but (4.17') false. Suppose that your optionsare to be ranked as in Chart 4.3 (where T stands for "take" and P for"pay"). In this case, the following are all true:

(4.25) you ought to take that item with you;(4.26) you ought to pay the shopkeeper;(4.24) you ought to pay the shopkeeper if you take that item with you;(4.27) you ought to take that item with you if you pay the shopkeeper.

But true, too, is the following:

(4.28) you ought to take that item with you (even) if you don't pay theshopkeeper,

and, correspondingly, (4.17') is false.Since (4.17') is false, so too is its equivalent

(4.17) you ought to take that item with you only if you pay the shop-keeper.

1: T & P ...2: T&-P...3: ~T & P ...4: ~T & ~P ...

Chart 4.3

17 For example, those that would symbolize (4.24) in some such way as O(T D P), whereD signifies a material conditional. Compare note 10 to this chapter.

123

And since (4.27) and (4.28) are true, so too are their equivalents:

(4.27') you ought not to take that item with you only if you don't pay theshopkeeper;

(4.28') you ought not to take that item with you only if you do pay theshopkeeper.

But I confess that (4.27') and (4.28') would be a pretty odd way of makingone's point. Still, (4.27) and (4.28) are somewhat odd too, and they're oddbecause (4.25), a statement of straightforwardly unconditional obligation,is true. When such statements are true, then statements of conditionalobligation, such as (4.27) and (4.28) and their equivalents, are generallyotiose, even if true.

At any rate, what it is important to note here is that, although (4.24)is true, (4.17) is not. Hence it would have been a mistake to offer thefollowing analysis:

(4.29) S ought, at T in W, to do A at T", only on the condition thatp istrue iff S ought, at T in W, so to act that p is true, on the condi-tion that S does A at T'.

We might symbolize this proposal as follows: O (A only ifp) iff O(p if A).This is to be rejected. What is to be accepted is (XV), which states: O(Aonly if;?) iff O(~A i£~p). (4.29) and (XV) are not equivalent.

The remainder of the paradoxes to be addressed all have to do with thematter of factual detachment. Hector-Neri Castaneda has introduced whathe calls the Paradox of the Second Best Plan.18 In this context, considerAudrey again, of whom it was said:

(4.14) if Audrey administers medicine B on Monday, then she ought toadminister it on Tuesday too (although she ought in fact to admin-ister medicine A instead on both Monday and Tuesday).

Suppose it happens to be true that

(4.30) Audrey will administer medicine JB on Monday.

In that case, don't we wish to say that

(4.31) Audrey ought to administer medicine B on Tuesday?

But how can we say both this and something else that is implied19 by (4.14),namely,

(4.32) Audrey ought to administer medicine A on Tuesday?

18 Castaneda (1981), pp. 58-60.19 Bywayof(2.40/c).

124

The answer is that we may not infer (4.31) from (4.14) and (4.30); such fac-tual detachment is not warranted. Of course, once it has happened thatAudrey has administered medicine B on Monday, it may well then be truethat she ought to administer it on Tuesday also. This is just one moreinstance of the sort of shift in obligations that was the topic of discussion inthe last chapter. But prior to Monday it will not be true to say that Audreyought then to administer medicine B on Tuesday, even if it is true then tosay that she will administer it on Monday.

In the case just discussed, there is a sense in which (4.31), though nottrue on Sunday, becomes true on Monday (once medicine B has beenadministered). ° But cases of this sort are possible where there is no time forsuch a shift in obligations, and hence no time at which a statement such as(4.31) becomes true.21 Consider again

(4.13) if Charlie changes lanes, then he ought to accelerate (although heought neither to change lanes nor to accelerate).

Suppose that the changing of lanes and the acceleration, were either or bothto occur, would occur at the same moment in time. In that case, even if itis true that

(4.33) Charlie will change lanes,

it is not true, at any time, that

(4.22) Charlie ought to accelerate.

For at no time does the best world accessible to Charlie contain his accel-erating. Charlie's options may be ranked as in Chart 4.4 (where C standsfor "change" and A for "accelerate"). At no time is Charlie faced with a setof options where A is featured in the top-ranked course of action.

1:2:3:4:

~CCC

~C

&&&&

Chart

~AA

~AA

4.4

20 Actually, this seems to me strictly false. What I would prefer to say is that (4.31) is ambigu-ous (depending on what the time of obligation is that is at issue) and that on one reading(where the time in question is Sunday) it is and will always remain false, whereas onanother reading (where the time in question is Monday, after medicine B has beenadministered for the first time) it is and will always be true.

21 Cf. Feldman (1990), p. 338, n. 23.

125

Let me return now to the Paradox of the Knower, first discussed inSubsection 2.3.2. This paradox may now be presented more perspicuous-ly, in light of the analysis of conditional obligation provided in this chap-ter. Recall Greg, the security guard. We may now understand him to havethe following conditional obligation: to provide a truthful report of his per-forming certain (wrongful) actions, if he does perform them. Consider onesuch action, A. Thus:

(4.34) Greg ought to provide a truthful report of his doing A, if he doesA.

The problem here is this. Suppose that

(4.35) Greg will in fact do A.

Then, according to certain accounts of the logic of "ought,"

(4.36) Greg ought to provide a truthful report of his doing A.

But his providing a truthful report of his doing A entails his doing A and,given that he ought to provide such a report, certain accounts of the logicof "ought" (including mine: see (2.43"')) imply that

(4.37) Greg ought to do A.

But how can it be that he ought to do A when A is wrong?The resolution of this problem may now be stated simply. Since condi-

tional obligation does not warrant factual detachment, we cannot infer(4.36) from (4.34) conjoined with (4.35). For although it may be that thebest v4-worlds accessible to Greg are worlds in which he provides a truth-ful report of his doing A, still the best worlds accessible to him are worldsin which he does not do A (and so also does not provide a truthful reportof his doing so).

The so-called Paradox of Reparation may be handled in exactly analo-gous fashion. Recall that

(4.15) if Sally sins, then she ought to repent...

Suppose also that

(4.38) Sally will sin.

Some may infer that therefore

(4.39) Sally ought to repent.

126

But let us suppose that repentance for sinning entails having sinned (at leastwhen the repentance is sincere). Surely we don't wish to infer that

(4.40) Sally ought to sin.

But, of course, we are in no danger of being committed to (4.40), and thatis because the move from (4.15) and (4.38) to (4.39) is invalid. Such factu-al detachment is unwarranted.

The so-called Paradox of Gentle Murder may be treated in the samemanner.23 Suppose that

(4.41) if Murphy murders Vic, then he ought to do it gently,

and that

(4.42) Murphy will murder Vic.

Some may infer that therefore

(4.43) Murphy ought to murder Vic gently.

But Murphy's murdering Vic gently entails his murdering Vic. Surely wedon't wish to infer that

(4.44) Murphy ought to murder Vic.

And of course we shouldn't. The move from (4.41) and (4.42) to (4.43) isinvalid.

Finally, a much-discussed paradox, concerning so-called contrary-to-duty imperatives and first introduced by Roderick Chisholm, is subject tosimilar resolution.24 Suppose that:

(4.45) Aida ought to go to the aid of her neighbors;(4.46) if Aida goes to the aid of her neighbors, then she ought to tell them

that she's coming;(4.47) if Aida doesn't go to the aid of her neighbors, then she ought not

to tell them that she's coming;

and

(4.48) Aida won't go to the aid of her neighbors.

22 Actually, this entailment appears false. Sincere repentance would seem at most to entailthat one believe that one has sinned.

23 This paradox was introduced in Forrester (1984). The treatment that I give is adequatealso to variations on the paradox presented in Goble (1991).

24 Chisholm (1963). The issue is there couched in terms of "ought-to-be."

127

Some may infer from (4.45) and (4.46) that

(4.49) Aida ought to tell her neighbors that she's coming.

But some may also infer from (4.47) and (4.48) that

(4.50) Aida ought not to tell her neighbors that she's coming.

Yet how could (4.49) and (4.50) both be true? As before, though, there isno danger of being committed to the joint truth of (4.49) and (4.50).Whereas the inference of the former is indeed valid (being an instance ofdeontic detachment), the inference of the latter is not (being an instance offactual detachment).

Castaneda has expressed dissatisfaction with this treatment ofChisholm's Paradox, in that he thinks that the joint statement of (4.48) and(4.49) is itself paradoxical.25 Feldman has replied that, although the jointassertion of (4.48) and (4.49) might seem odd in the absence of the explic-it insertion of (4.45), it is nonetheless true.26 Castaneda has expressed dis-satisfaction with this reply, in that he contends that what makes the jointassertion of (4.48) and (4.49) paradoxical is this. Suppose that Aida is intenton not going to the aid of her neighbors but is otherwise keen to minimizeher wrongdoing. Thinking along Feldman-type lines (lines that I have fol-lowed in this chapter), she therefore decides to satisfy the obligation allegedin (4.49) and tells her neighbors that she's coming. But this, far from min-imizing her wrongdoing, compounds it.27 Feldman has responded to this crit-icism, too, in a manner that seems to me appropriate,28 but he overlookswhat I take to be the most effective response to it. This response makesreference to subsidiary obligation, a matter to which I now turn.

4.4 SUBSIDIARY OBLIGATION

4.4.1 Factual detachment

You may be uneasy with the solutions just proposed to the paradoxes con-sidered. Should we dismiss factual detachment so summarily in the case ofconditional obligation (proper)? Should we really deny that

(4.22) Charlie ought to accelerate,(4.36) Greg ought to provide a truthful report of his doing A,

25 Castaneda (1989a), p. 9ff.26 Feldman (1989b), p. 35.27 Castaneda (1989b), p. 167.28 Feldman (1990), p. 331ff.

128

(4.39) Sally ought to repent,

(4.50) Aida ought not to tell her neighbors that she's coming?

If we do deny these statements, what then really is the point of making thecorresponding statements of conditional obligation, namely,

(4.13) if Charlie changes lanes, then he ought to accelerate.(4.34) Greg ought to provide a truthful report of his doing A, if he does

A,(4.15) if Sally sins, then she ought to repent...,

and

(4.47) if Aida doesn't go to the aid of her neighbors, then she ought notto tell them that she's coming?

It is of course true that detachment of an unconditional obligation, whenthe condition of the corresponding conditional obligation is satisfied, issometimes warranted. We have already seen that both necessity detachmentand deontic detachment are valid. Other such restricted detachment rulesmay seem plausible. As a variation on necessity detachment, it may seemthat, in cases where the action that is conditionally obligatory is "sched-uled" to occur after the condition in question is "scheduled" for satisfac-tion, if and when the condition is satisfied, the action then becomesobligatory; that is, it may seem that, if 5 ought at T1 to do A at T3, on thecondition that some event E occurs at T2, and if E occurs at T2, then Sought at T2 to do A at T3. For example, if Audrey ought on Sunday toadminister medicine B on Tuesday, on the condition that she administersit on Monday, and if she does administer it on Monday, then she ought onMonday (once medicine B has been administered) to administer it onTuesday. But in fact this does not hold without exception. I said in Section4.3 that it "may well" be that Audrey ought on Monday to administer med-icine B on Tuesday, but I didn't say that this was certain. For suppose that,while administering the medicine on Monday, Audrey also gives herpatient a lethal injection; presumably it would not then be true that sheought on Monday to administer the medicine again on Tuesday. (This isanother illustration of the failure of augmentation for conditional obliga-tion.) The situation may be represented as in Chart 4.5 (where AM, ATi BM,BT, and LM stand for "administer medicine A on Monday," "administermedicine A on Tuesday," "administer medicine B on Monday," "admin-

129

1: AM& ~BM & ~LM & AT& ~BT ...2: ~AM & BM & ~LM & ~AT & BT ...

3a: AM & BM & ~LM & AT & BT ...3b: [all other combinations with ~/_M

4a: ~AM & BM & LM & ~AT & BT ...4b: [all other combinations with LM]

Chart 4.5

ister medicine B on Tuesday," and "give a lethal injection on Monday,"respectively). Suppose that these are Audrey's options as of Sunday. If shetakes Track 2, it will indeed be true that she ought on Monday (once shehas administered medicine B) to administer medicine B on Tuesday. Butif she takes Track 4a, it will not be true that she ought on Monday to admin-ister medicine B on Tuesday.

Mark Vorobej has offered the following restricted detachment rule (avariant on the rule of deontic detachment): when the condition in ques-tion is satisfied and its nonsatisfaction is not obligatory (rather than, asbefore, its satisfaction being obligatory), then the obligation in questionmay be detached.29 This, too, is unacceptable. Suppose again that

(4.18a) you ought to go to the party if you accept the invitation.

But suppose also that

(4.51) it is not the case that you ought to accept the invitation,

(4.52) it is not the case that you ought not to accept the invitation,

and

(4.53) you don't accept the invitation.

Given (4.51), we surely do not want to say that

(4.54) you ought to go to the party,

and yet Vorobej's proposal implies that (4.54) follows from (4.18a), (4.52),and (4.53).

Are there other restricted detachment rules that are acceptable? Quitepossibly. But the fact remains that the sort of unrestricted detachment thathas here been called factual detachment is not valid for conditional obliga-tion (proper), given my analysis of such obligation. And, I have argued, thisis just as it should be. After all, it simply is not true that you ought to vote

29 Vorobej (1986), p. 20.

130

for Dick; for Tom is the best candidate. Moreover, rejection of factualdetachment allows for a straightforward solution to many of the deonticparadoxes.

Nonetheless, there is still the undeniable inclination to say that, in somesense, you ought to vote for Dick because you won't vote for Tom; thatCharlie ought to accelerate because he will change lanes; that Sally ought torepent because she will sin; that Aida ought not to tell her neighbors thatshe's coming because she won't go to their aid; that Alan ought to attend themeeting on the second floor because he won't attend the meeting on the firstfloor; and so on. I think this inclination is correct. Although I have just reject-ed factual detachment for conditional obligation, I want now to suggest thatthere is a sense in which such detachment is acceptable after all. We canhave our cake and eat it too, as long as we recognize that the sort of oblig-ation that is detached is, while of course unconditional and overall,nonetheless subsidiary.

4.4.2 Levels of obligation

The phenomenon that I have in mind has been noted by others. In partic-ular, Michael McKinsey has attempted to account for it.30 He has talked of"levels of obligation" and said that obligations may be "primary," "sec-ondary," "tertiary," or indeed "«-ary" (where n > 1). McKinsey is con-cerned with just the sort of cases that we are concerned with. The meetingcase brings this out well. In this case, McKinsey wants to say, Alan oughtt

to attend the 10:00 A.M. meeting on the first floor; given that he fails to dothis, he ought2 to attend the 10:00 A.M. meeting on the second floor; andso on. Here "ought/' expresses a primary obligation and "ought2" a sec-ondary obligation. Concerning such levels of obligation, McKinsey says:

By saying that an obligation is secondary (or tertiary, or...), I do not mean thatit is any less of an obligation than a primary one. In my view it is just as incum-bent upon a person to fulfill his secondary obligations, as it is incumbent uponhim to fulfill his primary ones. For notice that if a person does not do his best,we will blame him also for not doing his second best, just as we blame him fornot doing his best.31

This is both puzzling and promising. It is puzzling, in that it is not clearwhat it means to say that a non—primary obligation is "just as incumbent"as a primary one. Such talk seems to make things either impossibly difficult

30 McKinsey (1975).31 McKinsey (1975), p. 391.

131

or ridiculously easy, in the manner indicated earlier in Section 4.1. YetMcKinsey's proposal is nevertheless promising, for he has latched onto animportant fact, and that is that, if one fails to do what one ought to do butdoes what is "second best," one does less wrong32 than if one also fails to dowhat is "second best"; and one who fails to do what is "third best" does stillmore wrong; and so on. Alan does wrong if he fails to attend the 10:00 A.M.meeting on the first floor; but he does more wrong still if he also failsto attend the 10:00 A.M. meeting on the second floor (and this despite thefact that it is never the case that he ought - in the sense analyzed in (I) -to attend the meeting on the second floor).33 It should further be noted,however, that, if Alan attends the meeting on the first floor, he does nowrong — he violates no obligation — in not attending the meeting on thesecond floor.

We have noted that, if Alan fails to attend both meetings, he does a dou-ble wrong, even though he violates only a single obligation (in the sense of(I)). And this is clearly connected with the fact that he has a conditionalobligation to attend the meeting on the second floor if he fails to attend themeeting on the first floor. I propose that we say that Alan violates a prima-ry obligation to attend the meeting on the first floor and a secondary oblig-ation to attend the meeting on the second floor. (I) captures only theconcept of primary obligation; it is only with this concept that we havehitherto been working. We need an account of non-primary obligation -that is, subsidiary obligation.

For this purpose, the simplest assumption that one can make is the fol-lowing: all worlds accessible to the agent are comparable in deontic valueand at least one of these worlds is such that no other of them is more valu-able than it. Given this assumption, which I shall call the SimpleAssumption, we may say that some world (or worlds) accessible to the agentat a certain time has (or have) deontic value-rank 1 (relative to the agent atthat time) and all other accessible worlds have a corresponding subordinatedeontic value-rank n (where n is some positive integer greater than 1). It ison the basis of this assumption that the various diagrams and charts so farprovided have been drawn up.34

Working with this assumption, we may modify the accounts of uncon-ditional and conditional obligation that have been given. First, we should

32 I would prefer to put it in this way rather than, as McKinsey does, in terms of blame.See Zimmerman (1988a), pp. 41-3.

33 This is assured by the fact that both meetings occur at 10:00 A.M.; hence there is no timefor a shift in obligations of the sort discussed in Ch. 3

34 Again, see note 15 to Ch. 2.

132

explicitly acknowledge that the account of unconditional obligation thathas been provided is an account of unconditional primary obligation bymodifying (I) as follows:

(I') S ought!, at T in W, to do A at T iff[as in (I), but with the following addition to clause (c):and (5) W* has deontic value-rank I.]35

We may then assign levels to conditional obligation by modifying (XIII) asfollows:

(XIII') S oughtM, at T in W, to do A at T', on the condition that p is trueiff[as in (XIII), but with the following addition to clause (c):and (6) W' has deontic value-rank n.]

And then I propose that we allow for unrestricted (that is, factual) detach-ment of subsidiary obligation by providing this account of such obligation(where n is greater than 1):

(XVI) S oughtn, at T in W, to do A at T ' iff there is a proposition^ suchthat(a) S oughtn, at T in W, to do A at T', on the condition that p is true;

and(b) p is true in W.

Given (XVI), we may indeed say that you ought to vote for Dick becauseyou won't vote for Tom; that Alan ought to attend the meeting on the sec-ond floor because he won't attend the meeting on the first floor; and so on.These statements are true, given that they constitute statements of secondaryobligation; but they remain false if they are intended as statements of pri-mary obligation. This is how we can have our cake and eat it too. Youought! to vote for Tom, and so it is not the case that you ought! to vote forDick; otherwise things would be impossibly difficult or ridiculously easy.Nonetheless, inasmuch as you won't vote for Tom, you ought2 to vote forDick. If you don't satisfy this subsidiary obligation, you commit a subsidiarywrong. In this way, if you vote for neither Tom nor Dick, you do a doublewrong; you compound the primary wrongdoing involved in not voting forTom.

35 Given the Simple Assumption, this addition is strictly redundant, for every accessibleworld in which S does not do A at T" must be less than maximally valuable. But I sup-ply the addition for the sake of perspicuity.

133

It is clear that wrongdoing may be compounded still further. If you votefor neither Tom nor Dick nor Harry, you commit a primary and a sec-ondary and a tertiary wrong. In principle, one may pile subsidiary wrongupon subsidiary wrong ad indejinitum. Perhaps Alan oughtt to attend themeeting on the first floor, ought2 to attend the meeting on the second floorif he fails to attend the meeting on the first floor, ought3 to attend the meet-ing on the third floor if he fails to attend either meeting on the first and sec-ond floors, and so on for as many floors and meetings as you like. It is clearthat Alan has the opportunity to be a very naughty boy indeed. But notethat (XVI) also implies, quite correctly, that there is no naughtiness, nowrongdoing, if Alan fails to attend any of the higher-floor meetings but doesattend the first-floor meeting.

At this point we may make what I take to be the most effective responseto Castaneda's concern about the Chisholm case. He said that it seems tobe a mistake to claim that

(4.49) Aida ought to tell her neighbors that she's coming

when it is also true that

(4.48) Aida won't go to the aid of her neighbors;

for, if Aida were to try to minimize her wrongdoing (short of making (4.48)false) by satisfying the obligation at issue in (4.49), she would, paradoxical-ly, compound it instead. It is clear that I can now agree that Aida would becompounding her wrongdoing, for it is true that

(4.47) if Aida doesn't go to the aid of her neighbors, then she ought notto tell them that she's coming.

(4.48) and (4.47) jointly yield the following subsidiary obligation:

(4.50') Aida ought not to tell her neighbors that she's coming.

And Aida violates this obligation if she satisfies the (implicitly primary)obligation at issue in (4.49). It is worth repeating here a point made inSubsection 3.2.3: we should not necessarily be impressed if someone doessomething that he ought (that is, oughtj) to do; for some ways of failing todo what one ought to do are superior to some ways of doing what oneought to do.

All of this has been predicated on the Simple Assumption, though, andit may be that, on some substantive accounts of obligation, this assumptionis false. Such falsity will be due to one or other of two factors: either not allaccessible worlds are comparable in deontic value, or no accessible world

134

has a maximal deontic value.36 If either possibility were to arise, then fullyspecific judgments of subsidiary obligation would appear to be impossible.Although the accounts of unconditional and conditional obligation givenin (I) and (XIII) would still stand, either possibility would render a fully spe-cific deontic ranking of all accessible worlds unachievable, so that (XIII')and (XVI) would be inapplicable as originally intended. Nonetheless, evenif we therefore couldn't identify a certain wrong as secondary or tertiary,and so forth, there would still be sense to the claim that, in doing such-and-such, a certain agent has compounded his wrongdoing. One way to makesuch a judgment would be arbitrarily to assign the rank of 1 to some acces-sible world, then rank other deontically inferior worlds accordingly, andthen apply (XIII') and (XVI), thereby identifying certain wrongdoings asrelatively secondary, or tertiary, and so forth. This would still give somemeasure of the extent of wrongdoing committed by an agent, although thejudgment would be merely a relative one, either because some deontical-ly incomparable worlds will have been ignored or because, given that thereis no accessible world with a maximal deontic value, the agent will, in anabsolute sense, have done infinite wrong.

4.4.3 Overriding

In Subsection 3.2.5 a type of dynamic overriding of one (unconditional)overall obligation by another was explicated by means of (3.32). Now thatlevels of overall obligation have been distinguished, a different, thoughrelated, type of overriding - a static type - may be explicated. We may saythat one conditional overall obligation overrides another if and only if theformer is of a higher rank than the latter. Or more precisely:

(4.55) S's (overall) obligation^, at T in W, to do A at T", on the conditionthat p is true, is overridden by S's (overall) obligationm, at T in W,to do B at T', on the condition that q is true, iff(a) 5 oughttt, at T in W, to do A at T', on the condition that p is true;(b) S oughtm, at T in W, to do B at T', on the condition that q is true;

and(c) m is less than n.

A related account of one subsidiary obligation's overriding another may, ofcourse, also be given:

(4.56) S's (overall) obligation^, at T in W, to do A at T" is overridden byS's (overall) obligationw, at T in W, to do B at V iff

36 See the discussion of both possibilities in Subsection 2.2.6.

135

(a) 5 oughtM, at T in W, to do A at T';(b) S oughtw, at T in W, to do B at T'; and(c) m is less than n.

Note that, if a subsidiary obligation to do A is overridden by a subsidiaryobligation to do B, then the latter, overriding obligation is not satisfied. Forthe former, overridden, lower-level obligation "kicks in," as explainedabove, only if the higher-level obligation is violated.

It may be that the concept of overriding just accounted for is a rathertechnical concept, in that talk of "overriding" is more commonly appliedto the sort of dynamic overriding accounted for in (3.32) and to the sort ofstatic overriding of prima facie obligation to be accounted for in Subsection5.3.4. Nonetheless, I think that it may on occasion happen that the termi-nology is applied to what has been accounted for here, and so it is desirableto have such an account. Some might say, for instance, that, although thereis admittedly an obligation (of a sort) to vote for Harry, it is overridden byan obligation to vote for Dick, which is itself overridden by an obligationto vote for Tom. This may be understood in the manner just indicated.

It is worth noting that, as in the case of dynamic overriding accountedfor in (3.32), the sort of static overriding of an obligation to do A by anobligation to do B accounted for in (4.55) and (4.56) doesn't require thatthe agent be unable to do both A and B. The voting case is most naturallyunderstood as a case in which you cannot vote for more than one candi-date, but it needn't be understood in this way. It may be that you can votefor more than one, but any such vote would be deontically inferior to asingle vote or no vote at all. In that case, although the obligation to votefor Harry is overridden by the obligation to vote for Dick, it is nonethelesspossible for you to both vote for Harry and vote for Dick.

4.4.4 A rival account

The only other account of subsidiary obligation of which I am aware is theone provided by McKinsey. It too is explicitly predicated on the SimpleAssumption.37 His account is at odds with mine both with respect to whatcounts as a subsidiary obligation and with respect to what does not.

37 See McKinsey (1975), pp. 391-2. Greenspan (in Greenspan (1975), p. 266f., and (1978),p. 81) apparently thinks that subsidiary obligation is merely conditional obligation.Clearly, I believe this to be mistaken. Goldman (in Goldman, 1978, p. 211, n. 3)mentions this issue, but she reserves judgment and provides no account of levels ofobligation.

136

What McKinsey says is this:

(4.57) x oughtM at t to do Ax if and only if(1) A{ is contained in every <f>xt of rank n; and(2) for every (j>xt which has a rank m higher than n (i.e., where m < n),

there is an Aj such that cf>xt contains A- and x will not do Aj.3S

Here x is an agent, t a time, Ai and Aj are actions performed at times t{ andtj, respectively, and <j>xt is a "life-sequence" of actions open to x at t.

The case of Alan and the meetings can be used to show that McKinsey'saccount is more liberal than mine. The case may be partially represented asin Chart 4.6 (where M1 stands for "attend the meeting on the first floor"and M2 for "attend the meeting on the second floor"). Here it is assumedthat just one world has rank 1 and just one world has rank 2. We may ofcourse assume that there are lower-ranked accessible worlds in which M,occurs or in which M2 occurs, but there are none in which both occur. Myaccount of subsidiary obligation implies that Alan oughtj to attend themeeting on the first floor and, given that he does not do this, ought2 toattend the meeting on the second floor. My account also implies that Alanought j not to attend the meeting on the second floor. But it does not implythat Alan ought2 not to attend the meeting on the first floor, given thathe goes to the meeting on the second floor. (This would require that M^and M2 be jointly personally possible for Alan, and they're not.) ButMcKinsey's account does imply this. Such a claim strikes me as incorrect,indeed bizarre.

McKinsey's account is also more conservative than mine. Considera (rather strange) variation on the paying-the-shopkeeper case that maybe represented as in Chart 4.7. Here there is only one top-ranked world

1: M1 & ~M2 ...2: ~M1 & M2 ...

Chart 4.6

1: T& P...2a: T & ~P...2b: ~T & P...

3: ~T & ~P...

Chart 4.7

38 McKinsey (1975), p. 392.

137

but there are two second-ranked worlds. My account warrants sayingall of the following:

(4.25') you ought! to take that item with you;(4.26') you ough^ to pay the shopkeeper;(4.28") you ought2 to take that item with you if you don't pay the shop-

keeper;

and

(4.58) you ought2 to pay the shopkeeper if you don't take that item withyou.

In short, it is best for you to take and pay, but doing either one is prefer-able to doing neither, although (rather strangely) which one you do, if youdon't do both, is a matter of indifference. Further, my account implies that,given that the relevant condition is satisfied in either (4.28') or (4.58), oneor other of the following would be true:

(4.59) you ought2 to take that item with you;(4.60) you ought2 to pay the shopkeeper.

In this way, if you take Track 3, you will quite properly be said to havedone wrong on two levels. Yet McKinsey's account allows us to say neitherof (4.59) and (4.60), since in each case clause (1) of his account, (4.57), isviolated.

It is noteworthy that McKinsey seems not to be especially concernedwith conditional obligation and detachment when giving his account oflevels of obligation.39 Perhaps it is this lack of concern that has, at least as Isee it, led him astray. For the two are intimately related, as I have sought toshow.

4.4.5 Level of wrongdoing and seriousness of wrongdoing

It might be thought that my proposal is unnecessary. Why trouble ourselveswith levels of obligation when all we need do is note that, the further onedeparts from Track 1, the more serious the wrong that one does? Consideragain the voting case, represented in Chart 4.1. Is there really any need toinsist that you commit wrong on two levels if you take Track 3? Twowrongs, yes: ~T, and H. But why talk of two levels? Doesn't this multiplywrongdoing beyond necessity, in that, in addition to the primary wrongscommitted in the performance of ~T and H, it is being claimed that you

39 He does mention this issue in a footnote (McKinsey, 1975, p. 395, n. 10), but it is not amatter that occupies his attention in the rest of his paper.

138

commit a secondary wrong by performing ~D? Can't we express all thatwe need to express by noting the (primary) wrongs involved and also not-ing that Track 3 is a greater departure from Track 1 than Track 2 wouldhave been?

I think not. While I have no wish to deny that taking Track 3 is moreseriously wrong than taking Track 2, we must still contend with the incli-nation to infer that you really ought to vote for Dick, given that you won'tvote for Tom. My proposal accounts for this; the simpler counterproposaldoes not.

As to seriousness of wrongdoing, it should be observed that, on thosesubstantive accounts according to which the deontic ranking of optionsreflects some value that is specifiable independently of its relation to mat-ters of obligation,40 simply noting that taking Track 3 is more seriouslywrong than taking Track 2 does not tell the whole story. Suppose thatTrack 1 has a value of 100, Track 2 a value of 90, and Track 3 a value of 10(these numbers are of course purely ad hoc). Then taking Track 3 is verymuch more wrong than taking Track 2, a fact not revealed by their relativerankings alone. Nonetheless, it remains true that, in taking Track 3, you dowrong on two levels.

Level of wrongdoing is to be distinguished from seriousness of wrong-doing in that these constitute two different measures of the extent ofwrongdoing. Or better, there are two different extents. I said earlier that itis clear that Alan has the opportunity to be a very naughty boy indeed, butthis use of "very naughty" was merely an acknowledgment of all the dif-ferent levels of obligation applicable to Alan. Perhaps there was nothing thathe could do that would have been deontically very bad, and so nothing thatwould have been very seriously wrong, in which case he did not have theopportunity to be "very naughty" in another sense. Conversely, an agentmay have the opportunity to do very serious wrong and yet not have manylevels of obligation apply to him. Both senses must be allowed for, and thesimplifying counterproposal overlooks this.

4.4.6 Level of wrongdoing and level of obligation

It may be thought that, contrary to what I have intimated, the account thatI have proposed does not guarantee that level of wrongdoing correspondsstraightforwardly with level of obligation. What I have in mind here canperhaps be best put by means of an illustration. Consider Sally, the sinner.

40 On this issue, see Section 1.4.

139

1: ~S & ~R ...2: ~S & R ...3: S & R ...4: S & ~R ...

Chart 4.8

Her situation may be represented as in Chart 4.8 (where S stands for "sin"and R for "repent"). (By including Track 2, I am now allowing for thepossibility of repenting without having sinned.)41 Suppose that Sally takesTrack 3. In this case we may say each of the following:

(4.61) Sally oughtj not to sin;(4.62) Sally oughtj not to repent;(4.63) Sally ought2 not to sin (even) if she repents and, given that she does

indeed repent nonetheless, (still) she ought2 not to sin;

and

(4.64) Sally ought3 to repent if she sins and, given that she does sin, sheought3 to repent.

What may seem puzzling is this. Sally's obligation not to sin, mentioned in(4.61), is said to be primary, yet the obligation to repent, mentioned in(4.64), given that Sally sins, is said to be tertiary. But hasn't there been justone wrongdoing — namely, sinning — and so just one level of wrongdoing? Ifso, then there is no straightforward correspondence between level ofwrongdoing and level of obligation.

The answer is this. When Sally sins, she plummets to Level 3. Why?Because there is an intermediary level. As (4.63) shows, Sally not onlyoughtj not to sin, she also ought2 not to sin. Thus she does indeed do adouble wrong — wrong on two levels — when she sins, and that is why herobligation to repent is tertiary.

In general, the point can be put this way. If we keep our eye on the "big"picture, we can say that an agent has a primary obligation to take some first-level track, a secondary obligation to take some second-level track if he failsto take a first-level track, a tertiary obligation to take some third-level trackif he fails both to take a first-level track and to take a second-level track, andso on. It is through wrong being done on one level that an obligation onthe next level kicks in. A straightforward correspondence between level ofwrongdoing and level of obligation is thus assured.

41 Cf. note 22 to this chapter.42 Cf. note 45 to Ch. 3, concerning Bruce's obligation not to apologize when he does noth-

ing for which he owes an apology.

140

Prima facie obligation

5.1 A TALE OF TWO OBLIGATIONS

It was the best of acts, it was the worst of acts. Steve stopped to render firstaid to Carl, who had had a car accident. But he thereby failed to keep hisdinner date with Dave, who had badly wanted Steve to sample his latestculinary concoction. What a sorry business!

The story is a familiar one. W. D. Ross would say1 that Steve had a primafacie obligation (of beneficence) to tend to Carl and a conflicting primafacie obligation (of fidelity) to keep his date with Dave. Under the cir-cumstances, Steve was quite justified in satisfying the former obligation atthe expense of the latter; indeed, he was overall obligated to do so.2 Theterm "overall obligation" is not Ross's; he uses instead "absolute obliga-tion," "actual obligation," "duty proper," and "duty sans phrase," but noneof these seems to me as felicitous as "overall obligation." The term "primafacie obligation" is Ross's; I don't think that it is particularly felicitous either(nor did Ross),3 but, as noted in Chapter 1, since Ross's term is by now sowell entrenched, it seems on balance advisable to continue to use it.

The main question to be addressed in this chapter is simply this: how isthe concept of prima facie obligation to be understood? This of course pre-supposes that there is indeed such a concept in the first place. This has some-times been denied, and denied by philosophers from both ends of thenormative spectrum. Alan Donagan, for example, a dyed-in-the-woolKantian, takes Ross to task for claiming that there is a prima facie obliga-tion of fidelity that is overridden by a prima facie obligation of beneficencein the sort of case just given. Donagan says:

If a man accepts an invitation to dinner, it would be absurd for his host to under-stand him as having promised not to prevent a serious accident, or not to bring

1 See Ross (1930), p. 18ff.2 Broad would say that Steve had conflicting "component" obligations and that his "resul-

tant" obligation was to act as he did. See Broad (1985), p. 150.3 See Ross (1930), p. 20.

141

relief to the victims of one, if to do these things would prevent him from din-ing. It is a promiser's duty to express any condition to his promise which thepromisee might misunderstand; but there would be no misunderstanding insuch a case... To any relatively trivial promise there are a host of tacit condi-tions, all of which will normally be satisfied, which both promiser and promiseemust and do understand; and when, as occasionally happens, such a conditionis not satisfied, the promiser treats his obligation to the promisee as annulled.He has no need to consider himself as having a responsibility to fulfil it that isoutweighed by a heavier responsibility.4

I think Donagan is wrong. When Steve helped Carl, he did not keep hispromise to Dave. Of course, Steve didn't promise Dave to have dinner withhim even if he (Steve) could otherwise bring relief to the victim of an acci-dent; to that extent Donagan is right. Nonetheless, nor did Steve promiseto have dinner with Dave only if no such emergency arose; for then hewould not have failed to keep his promise to Dave when he helped Carl.The content of Steve's promise was not fully specific, even implicitly. Hesimply promised to have dinner with Dave, and this promise was silent,even implicitly, on the issue of conflicting emergencies. And it is at least inpart because he broke his promise, while having a (prima facie) obligationnot to break his promise, that Steve has the residual obligation to offer Davean explanation as to why he didn't show up - at least, that is what many,including Ross, would say, and I cannot see that Donagan has underminedthis view.5

It might be thought that, even if Donagan were correct about therebeing no prima facie obligation of fidelity in cases of the sort just discussed,still that would not provide sufficient reason to reject the concept of primafacie obligation altogether. I agree. But Donagan seems to think otherwise,for immediately after the passage just quoted he says the following:

It would be unjust to Ross not to record that, in his Foundations of Ethics, hetreated a number of casuistical problems about promising on these traditionallines [i.e., the lines just advocated by Donagan]. However, he did not see thatsuch a treatment deprives his theory in The Right and the Good of its raisond'etre.6

This seems to suggest that, if no prima facie obligation of fidelity is to befound in cases of the sort just discussed, then no prima facie obligations areto be found anywhere (for it must surely be agreed that the analysis andidentification of prima facie obligations is indeed the "raison d'etre" of

4 Donagan (1977), p. 93.5 Cf.McConnell(1995).6 Donagan (1977), p. 93.

142

Ross's theory of right in The Right and the Good). No argument for this boldclaim is given, but certainly Donagan does not concern himself with thenotion of prima facie obligation in the presentation of his own theory. Atany rate, both the claim and its antecedent strike me as false.

At the other end of the normative spectrum are utilitarians, also wellknown for their failure to employ the notion of prima facie obligation.Most are silent on the issue of what prima facie obligations there are or whatprima facie obligation itself might be, and their silence is eloquent: they dis-miss the notion. One who is not silent on this matter is Fred Feldman (whoproposes a modified version of traditional utilitarianism). Feldman explic-itly dismisses the concept of prima facie obligation, proclaiming it a"mirage,"7 by which he means that there really is no such concept, so thatterms that may appear to express it either are meaningless or express someother concept. Again, no argument is given for this (other than the con-cept's recalcitrance with respect to Feldman's efforts to analyze it).

I believe that the concept of prima facie obligation is genuine and, more-over, that it is amenable to analysis. Even if I am right, this of course doesnot mean that either Donagan's or Feldman's normative theory is false; forit is possible that the concept of prima facie obligation be a respectable con-cept and yet that there be no actual (or at least no very interesting) primafacie obligations. Nonetheless, in presenting my account of prima facieobligation, I shall frequently rely, for purposes of illustration, on Ross'sview that there are indeed certain significant prima facie obligations, such asthose of fidelity, beneficence, and the like. This is a natural view to hold, Ithink, and such reliance will render the analysis that I shall give easier tofollow.

5.2 DEFECTIVE ANALYSES

To many, prima facie obligation has seemed to be overall obligation that isessentially subject to some sort of condition. Ross himself suggests the term"conditional duty" as an alternative to the term "prima facie duty."8 Andthis may seem quite natural. Consider, for instance, the claim that oneought to keep one's promises, where this is taken to express a prima facieobligation. We may express this as follows:

(5.1) if 5 makes a promise to do A, then 5 ought to do A.

How is this is to be analyzed?

7 Feldman (1986), p. 144.8 Ross (1930), p. 19.

143

It is clear that none of the following will do as an analysis:

(5.2) that S makes a promise to do A materially implies that 5 ought over-all to do A;

(5.3) that S makes a promise to do A strictly implies that S ought overallto do A;

(5.4) that S makes a promise to do A subjunctively implies that S oughtoverall to do A.

None of these will do, for the simple reason that all such implication war-rants factual detachment (see the discussion of (4.1), (4.5), and (4.8), respec-tively, in the last chapter), whereas it is clear that the fact that one has madeor will make a promise to do something cannot by itself transform a primafacie obligation to keep it into an overall obligation to do so. An addition-al problem with (5.2) and (5.3), though not (5.4), is that neither materialnor strict conditionals are defeasible, where a conditional statement of theform "if p, then r" is understood to be defeasible just in case it is possiblethat it be true while "if p and q, then r" is false. (That is, to use terminolo-gy introduced in Chapter 4, both material and strict implication validate"augmentation.") Such indefeasibility would appear difficult to reconcilewith the overridability of prima facie obligation. For while one's primafacie obligation to keep a promise may on occasion also be an overall oblig-ation, it needn't be; and it won't be, even though one has made a promise,if in addition something more pressing requires one's attention.

We might try analyzing the prima facie obligation to keep a promise todo A as follows:

(5.5) 5 ought overall either not to make a promise to do A or to do A.

(The idea here is to have the overall obligation apply to the entire condi-tional, itself reformulated as a disjunctive statement, and not just to its con-sequent.)9 But again this fails to accommodate the overridability of primafacie obligation. Suppose that one makes a promise at Tt to do A at T3 andthat one therefore has a prima facie obligation at T2 to do A at T3. Accordingto the proposal, then, one ought overall at T2 either not to make thepromise at T1 or to do A at T3. But one cannot at T2 not make the promiseat TP Given the principle that, if one ought overall either to do Xor to doY and one cannot do X, then one ought overall to do Y, it follows that oneought overall at T2 to do A at T3.

10 But this may not be the case; for, even

9 Compare the analysis proposed in Hintikka (1970), p. 9Iff.10 This principle, whether read in terms of 0(XorY) or in terms of O(X v Y), is implied

by my account of overall obligation.

144

though one has a prima facie obligation at T2 to keep one's promise and doA at T3, it may be that one has no overall obligation at T2 to do so, becauseof some more pressing, contrary prima facie obligation.11

The fact that conditional overall obligation, as analyzed in (XIII) inSection 4.2, does not warrant factual detachment and is defeasible (that is,does not validate augmentation) may suggest that it is well suited to ananalysis of the concept of prima facie obligation. But, as Feldman has noted,since necessity detachment is valid for statements of conditional overallobligation, we cannot analyze (5.1) as follows:

(5.6) 5 ought overall to do A, on the condition that he promises to do A.

For, where it is also true that S cannot so act that he does not promise todo A (the most obvious type of case being one where the promise hasalready been made), we may detach from such a conditional overall oblig-ation the unconditional overall obligation to do A. But, as already noted,we don't wish to say that S has an overall obligation to do A just because hehas a prima facie obligation to keep his promise and has already promisedto do A.12

In light of the fact that the concept of prima facie obligation resists anyof the foregoing analyses, Feldman concludes, as noted earlier, that it ismerely a "mirage." But this conclusion appears premature, as I shall nowseek to demonstrate.

5.3 THE ANALYSIS

5.3.1 "All else being equal"

The basic idea underlying the analysis that I shall propose is this. When onehas a prima facie obligation to do A, doing A is of greater deontic value,taken in and of itself, than not doing A. Another, and common, way ofputting this is to say that doing A is deontically superior, all else being equal,to not doing A.13 (Strictly, this characterization applies only to direct, andnot to indirect, prima facie obligations, as will be explained shortly.) Thisway of putting matters is quite revealing, for it indicates that prima facieobligation is indeed essentially overall obligation conditioned in some way

11 Further problems with Hintikka's sort of analysis are discussed in Purtill (1973) andFeldman (1986), p. 182.

12 See Feldman (1986), pp. 130-1.13 Cf. Pietroski (1993) for an extended investigation into this characterization of prima facie

obligation.

145

(and thus that Ross's alternative term "conditional duty" is an apt one).Thus it is sometimes said: one ought (overall), all else being equal, to actbeneficently; one ought (overall), all else being equal, to keep one's promi-ses. Nonetheless, the fact that prima facie obligation is overall obligationconditioned in some way should not blind us to another fact that is fre-quently overlooked, and that is that prima facie obligation may itself boeither unconditional or conditional. The prima facie obligation of fidelityshould be construed as a conditional prima facie obligation: if one makes apromise (or some other sort of commitment) to do A, then one has a primafacie (and not necessarily overall) obligation to do A. For there is, it wouldseem, no call for fidelity tout court; it is fidelity to a commitment that isrequired, and, if there is no unconditional requirement to make a commit-ment (as in general there would seem not to be), there is no uncondition-al requirement of fidelity. That is, although commitment-plus-fidelity isdeontically superior, all else being equal, to commitment-plus-nonfideli-ty, we should not say that fidelity is deontically superior, all else being equal,to nonfidelity. For there are two ways to display nonfidelity - either by wayof commitment-plus-nonfidelity, or by way of noncommitment-plus-nonfidelity - and only the former is tantamount to infidelity; there seemsto be no reason to think that commitment-plus-fidelity is deontically supe-rior, all else being equal, to noncommitment-plus-nonfidelity.

Having taken note of the possibility that prima facie obligation be eitherconditional or unconditional, we should note also that the detachment ofan unconditional prima facie obligation from a conditional one is analo-gous to the detachment of an unconditional overall obligation from a con-ditional one. That is, although factual detachment is not valid, bothnecessity detachment and deontic detachment are valid. For example, evenif one is going to make a promise to do A, one does not yet have for thisreason alone an unconditional prima facie obligation to do A; rather, onesimply has a conditional prima facie obligation to do A if one promises todo so. But if the promise is inevitable (for example, one has already madeit), then one has an unconditional prima facie obligation to keep it; or ifone has an unconditional prima facie obligation to make the promise, thenone has an unconditional prima facie obligation to keep it.

Again, however, whether a prima facie obligation is itself uncondition-al or conditional, if it is properly characterized as an overall obligation "allelse being equal," then it is essentially an overall obligation that is condi-tioned in some way; for, whether or not it is itself conditional, it carrieswith it this extra condition of "all else being equal." How are we to under-stand this extra condition?

146

It would make matters very simple if this condition were some sort ofconstant (call it E for "equal"). For then the obligation at issue would bestraightforward conditional overall obligation of the sort analyzed in (XIII),and both necessity detachment and deontic detachment would be clearlyvalid. Consider the claim that Faith has a fidelity-based prima facie obliga-tion to pay Pam $10 (P), given her commitment (C) to do so. This would,on the present suggestion, be understood to mean that she has an overallobligation to do P on the condition that both C and E hold. Now considernecessity detachment, where this involves C's being inevitable. If Faithcannot so act that C does not hold, then all worlds accessible to her are C-worlds; and so, if the best (C&E)-worlds accessible to her are P-worlds,then the best E'-worlds accessible to her are P-worlds. As for deonticdetachment (where Faith has an unconditional prima facie obligation to doC): if the best (C&E)-worlds accessible to her are P-worlds and the bestE-worlds accessible to her are C-worlds, then, again, the best E-worldsaccessible to her are P-worlds.

What makes matters awkward is that it is clear that "all else being equal"is not to be interpreted as a constant. On the contrary, what it means variesfrom context to context; for it is a sort of indexical, in that "else" means"other than this." My suggestion is this: to say that 5 ought overall to do A,all else being equal, is to say, roughly, that any world, W, in which S doesA is deontically superior to any other world, W\ which is "just like" Wexcept that in W' S does not do A. (As before, this in fact applies only todirect, and not to indirect, prima facie obligation. Again, more on this in amoment.) Of course, this cannot be precisely correct, for no worlds can dif-fer from one another with respect to just one such fact. If S does A in Wbut not in W\ then any proposition which is distinct from but implies theproposition that S does A will also fail to be true in Wf. What we need tosay, then, is that W* is as "close" to W as possible, given that in it S doesnot do A; that is, Wf is "minimally different" from W with respect to Sdoing A.

This is not an unfamiliar notion. It, or something close to it, has beenput to good use by several philosophers in the analysis of counterfactuals.To adapt what John Pollock has said:14 the basic idea is that Wf is a worldthat is obtained from W by making minimal changes which suffice to makeit the case that S does not do A; truths in W that are in the appropriate sense"irrelevant" to or "independent" of 5's doing A must also be true in W'\gratuitous changes are disallowed. I would like to be able to provide a

14 Pollock (1976), pp. 70-1.

147

precise analysis of this crucial concept, but I do not know how to do so. Itmight seem that all that needs to be said is this: W' is minimally differentfrom W with respect to S doing A if and only if S does A'mW but not inW' and every proposition that is true in W but which does not imply thatS does A is also true in Wf. But this won't do. Consider any two proposi-tions^ and q, neither of which alone implies that S does A but which joint-ly imply this. Given the current proposal, each of p and q will be true inW'\ but then it will be true after all that S does A in W'. Pollock has him-self proposed an analysis of minimal difference that does not succumb tothis problem.15 It may be that this analysis is acceptable, but I am reluctantto commit myself to it; for it is very complex and, besides, rests on the con-troversial notion of a "simple" proposition. Moreover, the analysis (as istypical in the treatment of counterfactuals) involves a past—future asymme-try, so that similarity of the past is weighted more heavily than similarity ofthe future; and I do not wish to presuppose any such asymmetry here.Formally, then, I shall invoke the concept of the closeness of possible worldsunanalyzed. I hope that my use of this concept will be sufficiently intuitiveto render what I have to say about prima facie obligation both plausible andhelpful, just as its use (or the use of some closely related concept) has provenhelpful in the analysis of counterfactuals.

For reasons (having to do with detachment) that will become apparentshortly, I shall need to talk of those worlds that are minimally different forSatT in Wwkh respect to S doing A; that is, the focus of attention will beon those worlds that are accessible to S. The canonical phrase form in thiscontext is therefore this: W" is minimally different from W\ for S at TinW, with respect to 5 doing A at T". This may be understood in terms ofthe closeness of worlds as follows:

(5.7) W" is minimally different from W\ for S at Tin W, with respect to5 d o i n g s at T ' iff(a) W' is accessible to 5 from W at T and S does A at T ' in W'\(b) W" is accessible to S from W at T and 5 does not do A at T ' in

W"\ and(c) there is no world W'" such that

(1) W'" is accessible to S from Wat T,(2) S does not do A at T ' in W"\ and(3) W" is closer to W than W" is.

15 Pollock (1976), p. 70ff.

148

I shall also make use of the related concept ofconditional minimal difference,which may be understood as follows:

(5.8) W" is minimally different from W\ for S at T in W, with respectto S doing A at T ' , on the condition that p is true, iff(a) W' is accessible to S from W at T and S does A at T ' in W' and/?

is true in W'\(b) W7" is accessible to S from W 2X Tand S does not do A at T' in

W" and/? is true in W"\ and(c) there is no world W'" such that

(1) W'" is accessible to 5 from W2X. T,(2) S does not do A at T' in W'",(3) /? is true in W"\ and(4) W" is closer to W' than W " is.

Given this understanding of minimal difference, both unconditional andconditional, we may proceed to an analysis of prima facie obligation. I havesuggested: to say that S ought overall to do A, all else being equal, is to say,roughly, that all accessible worlds in which 5 does A are deontically supe-rior to all accessible worlds that are minimally different in this respect. Thisdouble use of "all" implies that there is something in general - "across theboard" - to be said in favor of doing A. It is this that indicates that doing Ais of greater deontic value, taken in and of itself, than not doing A. I pro-pose initially that we say this (revision will be called for later):

(5.9) S has, at T in W, a direct unconditional prima facie obligation todo A at T ' iff(a) there are a world W' and a world W" such that W" is minimally

different from W\ for S at T in W, with respect to S doing A atT'; and

(b) for all worlds W' and W",if (1) W" is minimally different from W\ for S at T in W, with

respect to S doing A at T' ,then (2) the deontic value for 5 at T of W' is greater than the

deontic value for S at T of W".Similarly, we may say:

(5.10) 5 has, at T in W, a. direct conditional prima facie obligation to doA at T ' , on the condition that/? is true, iff(a) there are a world Wf and a world W" such that W" is minimally

different from W\ for S at T in W, with respect to S doing A atT' , on the condition that/? is true; and

149

(b) for all worlds W' and W'\if (1) W" is minimally different from W\ for 5 at T in W, with

respect to S doing A at T', on the condition thatp is true,then (2) the deontic value for S at T of Wf is greater than the

deontic value for S at T of W".

Four points should be noted here.First, instead of talking of prima facie obligations, we could talk of deon-

tic reasons; for there is a sense of "reason" according to which one has a rea-son to do something just in case one has a prima facie obligation to do it.16

Note that reasons may be actual or merely potential; I take this to mean thatthey may be unconditional or merely conditional.

Second (5.9) and (5.10) imply that, if 5 has a direct prima facie obliga-tion to do A, then 5 can do A and S can refrain from doing A. A is, in otherwords, personally optional for S. This is in keeping with the analysis ofoverall obligation provided in Chapters 2 and 4.

Third, (5.9) and (5.10) are restricted to what I have called direct primafacie obligations. (Ross uses the term "direct" similarly.)17 If Ben has adirect prima facie obligation to act beneficently, this consists in every acces-sible beneficence-world being deontically superior to any accessible worldthat is minimally different with respect to beneficence. Similarly, if Faithhas a direct prima facie obligation to pay Pam $10, if she so promises, thisconsists in every accessible promise-plus-payment world being deontical-ly superior to any accessible promise-world that is minimally different withrespect to payment.

But not all prima facie obligations are direct. Some are indirect, andthere are two ways in which they may be so. First, some may be called (fol-lowing Ross again)18 incidental. For example, if Ben cannot act beneficent-ly without driving to Boston, then he has an incidental unconditional primafacie obligation to drive to Boston. Note that, since it would be highly sur-prising if all accessible worlds in which Ben drives to Boston were deonti-cally superior to all accessible worlds that are minimally different in thisrespect, we have no reason to think that Ben has a direct unconditionalobligation to drive to Boston. But, as long as he must drive to Boston todisplay beneficence, Ben does have an incidental, beneficence-based obliga-tion to drive to Boston. Similarly, if Faith cannot keep her promise with-

16 There is also another, nondeontic sense of "reason" that I shall not seek to characterizehere. Cf. Section 1.2 and also Ch. 2, n. 13.

17 Ross (1930), p. 46.18 Ross (1930), p. 46.

150

out going to the bank and withdrawing some cash, then she has an inci-dental conditional prima facie obligation to go to the bank and withdrawsome cash. The principle here is straightforward. It may be roughly statedas follows: if S has a direct prima facie obligation to do A and cannot do Awithout doing B, then S has an incidental prima facie obligation to do B.19

More precisely, let us say, first of all:

(5.11) p requires q, for S at T in W, iff for all worlds W' accessible to Sfrom W at T, if p is true in W\ then q is true in W'.

We may then account for incidental prima facie obligations as follows:

(5.12) S has, at T in W, an incidental unconditional prima facie obliga-tion to do B at T " iff for some act A and time T',(a) S has, at T in W, a direct unconditional prima facie obligation to

do A at T';(b) S doing A at T ' requires S doing B at T", for S at T in W\(c) 5 doing A at T ' is distinct from S doing B at T"; and(d) there is a world M/~' such that W' is accessible to S from Wat T

and S does not do B at T " in H/~';(5.13) 5 has, at T in W, an incidental conditional prima facie obligation

to do B at T" , on the condition that p is true, iff for some act Aand time T ' ,(a) 5 has, at T in H7, a direct conditional prima facie obligation to do

A at T', on the condition that p is true;(b) S doing AatT' requires S doing B at T", for S at T in H ;(c) S doing A at T ' is distinct from S doing B at T"; and(d) there is a world H7' such that W' is accessible to S from W at T

and 5 does not do B at T " in W'.

The third clause in each is required to ensure that not all direct prima facieobligations are also said to be incidental. The fourth clause is required topreserve the thesis that whatever one has a prima facie obligation to do ispersonally optional.

The second way in which prima facie obligations can be indirect maybe illustrated as follows. Suppose that Ben has an incidental prima facieobligation (grounded in beneficence) to drive to Boston and also has anincidental prima facie obligation (grounded in nonmaleficence) not to goto Boston without paying a visit to his mother, who lives there (in that,

19 This principle is no more vulnerable to the Good Samaritan Paradox than is the corre-sponding principle concerning overall obligation, namely, (2.43'")- See the discussion inSubsection 2.3.2.

151

although it would not be mean of him not to visit his mother, it would bemean of him to go to Boston and not visit her). Then Ben has what I shallcall a compound prima facie obligation (grounded in beneficence and non-maleficence) to visit his mother. The principle at issue here may be rough-ly stated as follows: if S has a prima facie obligation (whether direct orincidental or, indeed, compound) to do A and also has a prima facie oblig-ation not to do A without doing B, then S has a compound prima facieobligation to do B. More precisely:20

(5.14) S has, at T in W, a compound unconditional prima facie obliga-tion to do B at T" iff for some act A and time T",(a) S has, at T in IV, a prima facie obligation to do A at T';(b) 5 has, at T in W, a prima facie obligation not to both do A at T'

and not do B at T"; and(c) there is a world W' such that W' is accessible to S from W at T

and S does not do B at T ' in W'\(5.15) S has, at T in W, a compound conditional prima facie obligation

to do B at T", on the condition that p is true, iff for some act Aand time T",(a) S has, at T in W, a prima facie obligation to do A at T , on the

condition that p is true;(b) S has, at T in W, a prima facie obligation not to both do A at T'

and not do B at T"; and(c) there is a world Wr such that Wf is accessible to 5 from W at T

and 5 does not do B at T " in W'.21

Note finally that, just as the account of overall obligation presented inSection 2.1 included analyses of overall permissibility and wrongdoing, sotoo the present account of prima facie obligation may be extended toinclude analyses of prima facie permissibility and wrongdoing. Considerwrongdoing first. This may be accounted for simply as follows:

(5.16) it is prima facie wrong, for S at T in W, to do B at T " iff 5 has, atT in W, a prima facie obligation not to do B at T".

20 Compare the corresponding principle for overall obligation, (2.42'), discussed inSubsection 2.3.1.

21 A still more precise account of compound prima facie obligation, both unconditionaland conditional, would (in order to avoid any appearance of circularity) distinguishbetween level-1 compound obligations, which would themselves be generated by oblig-ations that are not themselves compound, and other level-n compound obligations,which would be generated by level-(« - 1) compound obligations.

152

This presupposes an account of what it is to have a prima facie obligationnot to do something. Such an account can be developed along lines exact-ly analogous to those along which the account of what it is to have a primafacie obligation to do something has been developed. That is, instead oftalking of a direct unconditional prima facie obligation to do A (see (5.9)),we can talk of a direct unconditional prima facie obligation not to do A;this would involve all accessible ~yl-worlds being deontically superior toall accessible worlds that are minimally different in this respect. (For exam-ple, it may be that Malcolm has a direct unconditional prima facie obliga-tion not to act maleficently; this would involve all accessible non-maleficence-worlds being better than all accessible worlds that are mini-mally different in this respect.) Instead of talking of an incidental uncondi-tional prima facie obligation to do B (see (5.12)), we can talk of an incidentalunconditional prima facie obligation not to do B; this would involve notdoing A requiring not doing B or, equivalently, doing B requiring doingA. (For example, it may be that Malcolm cannot play a certain practicaljokewithout acting maleficently; given his nonmaleficence-based reason not toact maleficently, he then has a nonmaleficence-based reason not to play thepracticaljoke.) And so on. An account of what it is to have a conditionalprima facie obligation not to do something may be likewise developed.

Developing an account of what it is for an act to be prima facie permis-sible is slightly more tricky. Evidently this must be grounded in the notionof one's not having a direct unconditional prima facie obligation not to doA, but how precisely is it to be accounted for? Should we say simply: someaccessible 4-world is at least as good as all worlds that are minimally differ-ent in this respect? Or should we say: all accessible ^4-worlds are at least asgood as all worlds that are minimally different in this respect? I think it isclear, on reflection, that we should say the latter; for the former is far tooweak. It may well be, for instance, that some walking-world accessible toWillis is at least as good as all worlds that are minimally different in thisrespect, but we shouldn't infer from this that it is prima facie permissiblefor Willis to walk. For it may be that every direct prima facie obligation thatWillis has requires that he not walk. On the other hand, if all beneficence-worlds accessible to Ben are at least as good as all worlds that are minimal-ly different in this respect, we may indeed infer that it is prima faciepermissible, as far as beneficence is concerned, for Ben to act beneficently.All of this could of course be laid out more formally, and it could be sup-plemented with an account of both indirect and conditional prima faciepermission, but I shall forbear doing this here.

153

5.3.2 Three objections

The foregoing account of prima facie obligation may appear objectionablefor at least three reasons.

First, it might be thought that the account implies that whatever isdirectly prima facie obligatory is also overall obligatory (or at least overallpermissible), for the following reason. According to the account, 5 has adirect prima facie obligation to do A just in case, for every two accessibleworlds that are just alike except in one S does A and in the other 5 doesn'tdo A, the former is deontically superior to the latter. But this appears toimply that the deontically best accessible world (or worlds), W, is one inwhich S does A; for otherwise that accessible world that is just like Wexcept in it 5 does A would be better than W.

Clearly such an implication would be unacceptable, but fortunately thepresent account doesn't have it. The term "just alike" is here not just roughbut inaccurate. (Note that, when giving my rough, intuitive account of "allelse being equal" above, I used the term "just like" and not the term "justalike.") In particular, one cannot assume that, if W" is minimally differentfrom W' (for 5 at T in W) with respect to S doing A at T", then W' is mini-mally different from W" with respect to S not doing A at T". Suppose, forexample, that Roy is faced with conflicting direct prima facie obligationsof reparation and beneficence. Let us say that the deontically best accessi-ble world is one in which he satisfies the former but not the latter (he can-not satisfy them both). Next best would be his satisfying the latter but notthe former. Worse still would be his satisfying neither. This may be repre-sented as in Chart 5.1 (where R stands for "reparation" and B for "benefi-cence") . Given a proper filling-in of "...," we can say that W2 is minimallydifferent from Wx with respect to Roy's not acting beneficently; and yetW^ (and not Wr) is minimally different from W2 with respect to Roy'sacting beneficently. Thus what is prima facie obligatory need not also beoverall obligatory or permissible.

A second objection stems from consideration of cases such as the fol-lowing. Suppose, as in Subsection 3.2.5, that Gertrude owes Lucy a debtof gratitude. The most appropriate way for her to display gratitude wouldbe to send Lucy lilies; daisies would be a poor second. But even though

W,: R & ~S & ...W2:~R & B & ...W3:~R & ~B & ...

Chart 5.1

154

sending daisies would be inferior, it might seem that, insofar as it wouldconstitute some display of gratitude, Gertrude has some reason, and hence aprima facie obligation, to send daisies; it's just that this obligation is over-ridden by the obligation to send lilies. But my account does not warrantour saying this. Presumably Gertrude has no direct prima facie obligation tosend daisies, and doing so is not necessary for a display of gratitude and henceis not even incidentally prima facie obligatory (barring its being necessaryfor the satisfaction of some other direct prima facie obligation). In fact, itmight appear that I cannot even say that Gertrude has an incidental primafacie obligation to send lilies, since even this appears not to be necessary fora display of gratitude (in that sending daisies would be an alternative, albeitinferior).

What I think we should say here is this. Granting that gratitude comesin degrees, let us say that "full" gratitude is gratitude to degree 10.Gertrude's sending lilies would thus constitute a display of gratitude todegree 10; her sending daisies would, let us say, only constitute a display ofgratitude to degree 5. Now, while it is plausible to say that a display of grat-itude is better, all else being equal, than a lack of a display of gratitude, itdoesn't seem plausible to say that a display of gratitude to degree x is bet-ter, all else being equal, than a lack of a display of gratitude to degree x; forwhat if the lack of a display of gratitude to degree x is due to a display ofgratitude to a greater degree y? Still, it is plausible to say that a display ofgratitude to at least degree x is better, all else being equal, than a lack of adisplay of gratitude to at least degree x. This observation allows us to saythat Gertrude does indeed have an incidental obligation to send Lucy lilies(given that there is no degree of gratitude greater than 10 and that sendinglilies is the only way to display gratitude to that degree); and, although itwon't allow us to say that she also has an incidental obligation to send Lucydaisies, it does allow us to say that she has an incidental obligation either tosend lilies or to send daisies (given that nothing else would constitute a dis-play of gratitude to at least degree 5). And this indeed seems to be whatshould be said.

In fact, the matter is even more complex than this. Suppose thatGertrude ought, from gratitude, to send Lucy lilies, and also ought, fromgratitude, to send Chrissie chrysanthemums, but cannot do both. Here, itseems, gratitude requires the impossible, and yet my account implies thatwhatever is prima facie obligatory is possible. The solution, of course, isthat gratitude, more fully characterized, is gratitude-to-a-benefactor (or,perhaps, would-be benefactor, or presumed benefactor), just as fidelity isfidelity-to-a-promisee (or someone to whom another sort of commitment

155

has been made). (If we talk of "what gratitude requires" without qualifica-tion, then, we must mean what gratitude to every benefactor requires; sotoo for fidelity, reparation, and any other obligation which is owed to a par-ticular person or persons.)22 My account implies, properly, that Gertrudecan satisfy each of her prima facie obligations of gratitude-to-Lucy and grat-itude-to-Chrissie; it does not imply, nor should it, that she can satisfy themboth.

A third objection is more troublesome. It is that the account implies thatcertain direct prima facie obligations that are intuitively thought of as beingmerely conditional turn out also to be unconditional. Consider fidelity.This has been assumed (for purposes of illustration) to constitute a directconditional prima facie obligation, in that all commitment-plus-fidelity-worlds have been assumed to be deontically superior to any worlds that areminimally different with respect to fidelity. But notice that one cannot (itseems) display fidelity in the absence of a commitment; that is, all accessi-ble fidelity-worlds are also commitment-worlds. But then, if all accessiblecommitment-plus-fidelity-worlds are deontically superior to their closestcommitment-plus-nonfidelity-worlds, it must also be true that all accessi-ble fidelity-worlds are deontically superior to their closest nonfidelity-worlds. In other words, the direct prima facie obligation of fidelity will benot just conditional but unconditional. The same applies to other (alleged)conditional direct prima facie obligations such as those of gratitude andreparation. For one cannot (it seems) display gratitude in the absence of thatcondition (namely, being a beneficiary) on which it is conditionally primafacie obligatory; likewise, one cannot make reparation for a nonexistentharm.

It is tempting at first simply to bite the bullet here but then go on to pointout that there are still certain intimately related direct prima facie obliga-tions that are merely conditional. For example, even if Faith cannot displayfidelity to her promise to pay Pam $10 without making that promise, stillshe can pay Pam $10 without promising to do so. Thus, while all accessi-ble promise-plus-payment-worlds may be better than their closest accessi-ble promise-plus-nonpayment-worlds, there is no need to think that allaccessible payment-worlds are better than their closest accessible nonpay-ment-worlds. And isn't that what really matters here?

Well, no, it isn't. There are two problems with this. First, on some occa-sions, it may not be possible to identify a conditionally obligatory act in such

22 Perhaps, for example, nonmaleficence is such an obligation, whereas beneficence is not.This has to do with whether persons have a right to the performance of an obligatory act.Rights are discussed further in Section 5.5.

156

a way that the condition is not included in the identification. AlthoughFaith can do what fidelity requires (namely, pay Pam $10) without promis-ing to do it, it may be that Roy cannot do what reparation requires (mend-ing a broken vase, say) without committing the relevant harm. Secondly,if it is admitted that Faith has a direct unconditional prima facie obligationto display fidelity (even if it is also pointed out that her direct prima facieobligation to pay Pam $10 is merely conditional), it must then also beadmitted, given the foregoing account, that she has an incidental primafacie obligation to make the relevant promise that is tmconditional.Similarly, if it is admitted that Roy has a direct unconditional prima facieobligation to make some reparation, then it must also be admitted that hehas an incidental prima facie obligation to commit the relevant harm thatis ^conditional. This is surely to admit too much. The original accountmust therefore be revised.

We must deny not only that Faith has an unconditional prima facieobligation to pay Pam $10 but also that she has an unconditional prima facieobligation to display fidelity to the commitment to pay Pam $10 (given thatshe has not already made this commitment and that there is no indepen-dent unconditional reason to make it). (Similar remarks pertain to the primafacie obligations of gratitude and reparation.) What I think warrants suchdenial is an observation already made above: although commitment-plus-fidelity is deontically superior, all else being equal, to commitment-plus-nonfidelity, it is not deontically superior, all else being equal, tononcommitment-plus-nonfidelity; for nonfidelity constitutes infidelityonly when there is an actual commitment. It is in virtue of this fact that theprima facie obligation of fidelity is merely conditional. But how can thisbe? After all, consider a situation where it is possible to display a great dealof maleficence to one person while also displaying minor beneficence toanother. We may suppose in this case that maleficence-plus-beneficence isdeontically superior, all else being equal, to maleficence-plus-nonbenefi-cence, and also that maleficence-plus-beneficence is not deontically superior,all else being equal, to nonmaleficence-plus-nonbeneficence. Yet thisdoesn't make the prima facie obligation of beneficence a merely condi-tional one. So why should the corresponding fact about commitment-plus-fidelity make the prima facie obligation of fidelity a merely conditionalone?

The answer is to be found in the fact that commitment is a prerequisite offidelity, whereas maleficence is not a prerequisite of beneficence. This isthe very feature about fidelity that has given rise to the present problem,and it points to the proper solution. Commitment is a prerequisite of

157

fidelity, not just in the sense that there cannot be fidelity without acommitment, but in the sense that fidelity requires that a commitment pre-cede it. Commitment and fidelity thus come in a well-ordered "package,"as it were, and the package must commend itself a6 initio if fidelity is to beunconditionally obligatory. (Similar remarks pertain to gratitude and repa-ration and their respective initial conditions.) To put this point moreprecisely, let us first say this:

(5.17) W" is minimally different from W', for 5 at T in W, with respectto S doing A at T", on the condition that p varies, iff(a) W' is accessible to S from W at T and S does A at T' in W' andp

is true \x\W'\(b) W" is accessible to S from W at T and S does not do A at T ' in

W" and |? is not true in W"; and(c) there is no world W'" such that

(1) W'" is accessible to S from W at T,(2) 5 does not do A at T ' in W"\(3) p is not true in W'", and(4) W'" is closer to W' than W" is.

With this in hand, we may now revise (5.9) by appending an extra clauseas follows:

(5.9') S has, at T in IV, a direct unconditional prima facie obligation to doA at T iff(a) there are a world W' and a world W" such that W" is minimally

different from W\ for S at T in W, with respect to S doing 4 at T';(b) for all worlds H/' and H/",

if (1) W" is minimally different from PF', for 5 at T in M , withrespect to S doing A at T' ,

then (2) the deontic value for S at T of W7' is greater than thedeontic value for S at T of W"; and

(c) for all propositions p,if ( 1 ) 5 has, at T in W, a direct conditional prima facie obliga-

tion to do A at T', on the condition that/? is true,(2) for some event E and time T " earlier than T', /? is the

proposition that E occurs at T",(3) S doing A2XT' requires p, for S at T in W, and(4) there is a world accessible to 5 from W at T in which p is

not true,then (5) for all worlds W' and W",

158

if (i) W" is minimally different from W\ for S at T in W, withrespect to 5 doing A at T', on the condition that p varies,

then (ii) the deontic value for S at T of W' is greater than thedeontic value for 5 at TofW".

The "official" account of prima facie obligation is thus as follows: replace(5.9) by (5.9') and leave (5.10) (which, indeed, is presupposed by (5.9')) and(5.11)-(5.15)astheyare.

Finally, it should be noted that this account accommodates the possibil-ity that deontic value be significantly agent- or time-relative, inasmuch asthe relevant value of the worlds is the value they have for S at T. Let menow elaborate on this with respect to Ross's theory. I said earlier that atleast some of the values that ground prima facie obligations are thought byRoss to be significantly agent-relative. Fidelity, for example, is clearly avalue of this sort. Ross says that each agent always has a prima facie oblig-ation to see to it that he keeps his promises.231 take this to imply that eachof us always has a (conditional) prima facie obligation to see to it that weourselves display fidelity. Now it might be that Ross also believes that eachof us always has a prima facie obligation to see to it that others display fideli-ty; for he might believe that such displays are intrinsically valuable, and hedoes say that we do always have a prima facie obligation to do that whichwill promote such value.24 But even if this is so, it does not detract from thefact that it seems pretty clear that, according to Ross, for each of us a spe-cial value attaches to our displaying fidelity. So too for gratitude and repa-ration, it seems.25 Whether or not this is so for other prima facie obligationsthat Ross lists is unclear to me. Does he believe, for instance, that there isa special value for each of us to be attached to our not harming innocent

23 Ross (1930), p. 17ff.24 On the obligation to promote intrinsic value, see Ross (1930), pp. 24 and 39. In fact,

however, the discussion in Ross (1930), Ch. 5, especially p. 140, suggests that Ross doesnot believe that each of us always has, in virtue of the prima facie obligation to promoteintrinsic value, a prima facie obligation to see to it that others display fidelity. For therehe lists four intrinsic goods - virtue, pleasure, the allocation of pleasure to the virtuous,and knowledge - and he explicitly says that he doubts the existence of any others (savecombinations of those mentioned). It might be thought that Ross takes fidelity to be avirtue, but this would be a mistake. For he takes virtue to concern motive, and he explic-itly divorces the duty of what he calls "fidelity" from motive. (See Ross, 1930, p. 22 andalso p. 5.) Still, Ross does say (1930, p. 36) that fidelity can be bonific - that is, instru-mentally good — and so he clearly thinks that, on those occasions on which it is bonific, thereis a prima facie obligation to promote it.

25 See Ross (1930), pp. 27, 30, and 41-2.

159

people, even if there is also a value to be attached to preventing others fromharming innocent people? I am not sure, although certain of his remarkssuggest that he does believe this.26

5.3.3 Detachment

The move from (5.9) to (5.9') blocks factual detachment. Even if Faith willpromise to pay Pam $10, the prima facie obligation to display fidelity to thiscommitment is merely conditional; for clause (c) of (5.9') is not satisfied.To be acceptable, though, the account must not only block factual detach-ment, it must also validate necessity detachment and deontic detachment.And so it does. Consider necessity detachment first. If Faith cannot avoidthe commitment to pay Pam $10, then clause (c) of (5.9') is satisfied afterall, and trivially so; for there will then be no accessible worlds that are min-imally different with respect to Faith's displaying fidelity to this commit-ment, on the condition that her making this commitment varies. Thus herconditional prima facie obligation to display fidelity is transformed into anunconditional prima facie obligation to do so, and her obligation to payPam $10 is likewise transformed.

Deontic detachment also holds, although the story here is more com-plicated. Let us look first at deontic detachment having to do with directprima facie obligations. Suppose that Faith has already made a promise(promise 1) to make another promise (promise 2) to pay Pam $10. Then,by necessity detachment, she has an unconditional prima facie obligationto make promise 2. Given the conditional prima facie obligation to pay Pam$10 if she makes promise 2, it should turn out that Faith has an ^condi-tional prima facie obligation to pay Pam $10. And it does. Consider theworld-schemata represented in Chart 5.2 (where P2 stands for "promise 2"and P for "payment"). We already know (given a proper filling in of "...")that, because of the direct unconditional prima facie obligation to makepromise 2, Wt is better than W3. We also know that Wt is better than W2,for this is what the direct conditional prima facie obligation to pay Pam $10if promise 2 is made amounts to. Thus (P2&P) is better, all else being equal,than ~(P2&P), and so Faith has a direct unconditional prima facie obliga-

W,: P2 & P & ...W2: P2 & ~P & ...W3:~P2 & P & ...

Chart 5.2

26 See Ross (1930), p. 22.

160

IV,:W2:

B &B &

W3:~B&

IV,: BW2: B

Chart

& P2

& P2

Chart

P~P

P

5.3

&

&&&

P& ~P

5.4

&&

tion to do (P2&P).27 Thus she also has an incidental prima facie obligationto do P.

Consider, next, incidental prima facie obligations. Suppose that Faith'sprima facie obligation to make promise 2 is not grounded in any already-made promise but rather in beneficence. Beneficence requires that shemake promise 2; her obligation to do so is thus incidental and not direct.Consider the world-schemata represented in Chart 5.3 (where B stands for"beneficence" and P for "payment"). Since beneficence is directly uncon-ditionally obligatory, Wt is better than W3. In addition, since B requires P2

(where P2 stands for "promise 2"), Wt and W2 must, more precisely, havethe forms represented in Chart 5.4. Since Pis directly prima facie obliga-tory on the condition that P2 occurs, Wi is better than W2. Hence (B&P)is unconditionally obligatory, and so P is as well.

Consider, finally, compound prima facie obligations. Suppose thatbeneficence requires that Faith visit her mother and that nonmaleficencerequires that she not visit her mother without making promise 2. Then shehas an unconditional compound obligation to make promise 2. Considerthe world-schemata represented in Chart 5.5 (where B stands for "benefi-

IV,:W2:IV3:W4:

BBB

~B

&&&&

NN

~NN

&&

a&

Chart 5.5

P~P

PP

&&&&

27 Clause (c) of (5.9') is satisfied, since the only relevant prerequisite of (P2 & P) is promise1. Note that there is no need in this context to compare Wx with

W4:~P2&~P&...For, although this world is a ~(P2 & P)-world, it is not as close to WA as either W2 orW3 is. (Still, I think that Wi is better than W4, and that this could be satisfactorilyproven.)

161

W,:W2:

BB

&&

NN

&&

Chart

11

5

3

O

.6

&&

P~P

cence," AT for "nonmaleficence," and Pfor "payment"). Given the directunconditional prima facie obligation of beneficence, we may say that Wi

is better than W4. Given the direct unconditional prima facie obligation ofnonmaleficence, we may say that Wt is better than W3. In addition, since(B&N) requires P2 (where P2 stands for "promise 2"), Wt and W2 must,more precisely, have the forms represented in Chart 5.6. Since P is primafacie obligatory on the condition that P2 occurs, Wt is better than W2.Hence (B&N&P) is unconditionally obligatory, and so Pis as well.

5.3.4 Overriding

A straightforward account of the overriding of one prima facie obligationby another emerges from the foregoing. Roughly, what we should say issimply this: S's prima facie obligation to do A is overridden by his primafacie obligation to do B if and only if every accessible world in which hedoes A is deontically inferior to some accessible world in which he does B.Or more precisely:

(5.18) S's prima facie obligation, at T in W, to do A at T' is overriddenby S's prima facie obligation, at T in W, to do B at T' iff(a) S has a prima facie obligation, at T in W, to do A at T';(b) S has a prima facie obligation, at T in W, to do B at T'\ and(c) for all worlds W" such that W" is accessible to S from W at T and

S does A at T ' in W"\ there is a world W' such that(1) W' is accessible to 5 from W at T,(2) S does B at T ' in W\ and(3) the deontic value for S at T of Wf is greater than the deontic

value for S at T of W".

This applies to all prima facie obligations, whether unconditional orconditional, direct or indirect.

Notice that it is possible that direct obligations override in a manner thatfails to match the manner in which their corresponding incidental obliga-tions override. Suppose that Ben has a direct prima facie obligation ofbeneficence and that on a certain occasion this overrides a direct prima facieobligation of fidelity. Suppose also that he has the corresponding inciden-tal obligations to drive to Boston and to drive to Baltimore, respectively. It

162

may be that his obligation to drive to Boston does not override his obliga-tion to drive to Baltimore; indeed, it may even be that the latter overridesthe former. This would happen if there is yet another direct prima facieobligation, say an obligation of reparation, that overrides the obligation ofbeneficence and the fulfillment of which itself requires that Ben drive toBaltimore.

5.4 ROSS ON PRIMA FACIE OBLIGATION

Since it was Ross who introduced the concept of prima facie obligation tous, I have relied fairly heavily on what he says in the presentation of myown analysis of the concept. I have invoked his distinction concerningdirect and incidental obligations, and I have made free use of some of hisexamples of prima facie obligation (fidelity, beneficence, gratitude, repara-tion, and so on). It is appropriate at this point to try to determine to whatextent my account squares with Ross's, at least with regard to certain keyissues.

5.4.1 Conditional and unconditional prima facie obligation

First, it might be thought that Ross does not distinguish between condi-tional and unconditional prima facie obligation, inasmuch as he sometimesrefers to all prima facie obligation by the term "conditional duty." But infact I think that Ross does make this distinction but simply uses differentterminology to do so. I have in mind his comments on what he calls "spe-cial" and "general" prima facie obligations. Special obligations, he says,rest on the performance of some prior act (such as the making of a promise,the infliction of an injury, or the acceptance of a benefit); the obligations(of fidelity, reparation, and gratitude, respectively) are therefore condi-tional on the performance of these acts. General obligations (among whichRoss counts the obligations of justice, beneficence, and nonmaleficence)do not rest on the performance of such prior acts and may therefore be saidto be unconditional.

At least, this is how I understand Ross's distinction between special andgeneral obligations and what he means by "rest on." Note that, in someother, more liberal sense of "rest on," all obligations, both conditional andunconditional, will rest on the satisfaction of some condition, for the per-formance of any act requires that certain conditions be satisfied. For exam-ple, just as no one can keep a promise unless he has made one and no one

28 Ross (1930), p. 27.

163

can make reparation for an injury unless one has been inflicted, so too noone can administer justice unless there is occasion for doing so, and no onecan act beneficently unless someone is available as a potential beneficiary.So what Ross means by "rest on" cannot be this. Rather it is, I suggest,what I have indicated, namely, that, with respect to fidelity, not only doesfidelity require commitment, but noncommitment-plus-nonfidelity is noworse than (that is, is not deontically inferior to) commitment-plus-fideli-ty, and so fidelity isn't unconditionally obligatory; similarly, noninfliction-of-harm-plus-nonreparation is no worse (indeed, it is better, given theprima facie duty of nonmaleficence) than infliction-of-harm-plus-repara-tion, and so reparation isn't unconditionally obligatory; and so on. But withrespect to beneficence, for example, there is no "prior act" like that of com-mitment (or the infliction of harm) which it is prima facie permissible toforgo, and so nothing that is morally on a par with noncommitment-plus-nonfidelity. All nonbeneficence is morally on a par with infidelity (that is,with commitment-plus-nonfidelity). Or so I understand Ross. Of course,someone might take the view that beneficence is in general only condi-tionally required, in that its obligatoriness rests on the performance of someprior act (such as that of becoming a neighbor of the potential beneficiary,in some sense of "neighbor" according to which not everyone else is auto-matically one's neighbor); but I am sure that this is not Ross's view.

5.4.2 The self-evidence of prima facie obligations

Ross says that the direct prima facie obligations he mentions are self-evi-dent. His remarks concerning this matter clearly indicate that he believesthat it is a necessary truth that certain acts are prima facie obligatory(whether conditionally or unconditionally).29 Such a claim of course goeswell beyond what a mere conceptual analysis could be thought to imply,but is it consistent with what I have said?

This is a complicated matter. The account so far provides an analysis ofwhat it is to claim that a particular agent, S, has a prima facie obligation ata particular time. Unless S exists necessarily, no such claim can be a neces-sary truth. Still, we can allow for necessary truths concerning prima facieobligation by expanding the account that has already been given. We can,in particular, say the following:30

29 Ross (1930), p. 29.30 Here, as in the presentation of (2.11)—(2.14) at the end of Subsection 2.1.4, acts are treat-

ed as being "repeatable." But see again the remarks made in Subsection 2.2.5 on thismatter.

164

(5.19) S has a permanent unconditional prima facie obligation in W to doA iff for all times T and T',if (a) there is a world accessible to S from W at T in which S does

A at T', and(b) there is a world accessible to S from H7 at T in which 5 does

not do A at T',then (c) S has an unconditional prima facie obligation, at T in W, to

do A at T'.

Similarly, we may also say:

(5.20) S has a permanent conditional prima facie obligation in W to do Aiff for all times T and T', there is a proposition p such that,if (a) there is a world accessible to S from JF at T in which 5 does

A at T' and p is true, and(b) there is a world accessible to S from W at T in which S does

not do A at T' and^ is true,then (c) S has a conditional prima facie obligation, at T in W7, to do A

at T', on the condition that j? is true.

A prima facie obligation is thus a permanent one if it applies on all occa-sions where it is applicable (rather than simply on all occasions, period,which would imply that it is always applicable, that is, can always be actedon; this would be an extremely restrictive, and hence not very useful,understanding of what it is for a prima facie obligation to be permanent).We can then say:

(5.21) there is a universal unconditional prima facie obligation in W to doA iff for all persons S,if (a) S exists in W,then (b) S has a permanent unconditional prima facie obligation in W

to do A;(5.22) there is a universal conditional prima facie obligation in W to do A

iff for all persons 5,if (a) 5 exists in W,then (b) S has a permanent conditional prima facie obligation in W to

do A

And, finally, we can say:

(5.23) there is a necessary unconditional prima facie obligation to do A ifffor all worlds W, there is a universal unconditional prima facieobligation in W to do A;

165

(5.24) there is a necessary conditional prima facie obligation to do A ifffor all worlds W, there is a universal conditional prima facie oblig-ation in Wto do A.

One can then go on to claim, as Ross appears to do, that, if ever an agenthas a prima facie obligation to do some act A, this is grounded in the factthat there is a necessary direct prima facie obligation to do either A itself orsome other act for which A is necessary.31 Although the analysis that I'veprovided does not imply this claim, it is certainly consistent with it.

Nonetheless, there are complications, one minor, one major, in trying toreconcile what Ross has to say about the self-evidence of certain particularprima facie obligations with the account that I have provided. The minorcomplication concerns the (alleged) prima facie obligation of fidelity.Although I have allowed myself to invoke this obligation for the sake ofillustration here, it is in fact doubtful that there is a necessary direct primafacie obligation of fidelity of the sort that Ross alleges. For what he appearsto say is that every commitment necessarily gives rise to a prima facie oblig-ation to fulfill that commitment. This cannot be right. For one thing, peo-ple sometimes promise to do something that they in fact cannot do. Theaccount that I have given implies that there is no prima facie obligation tokeep such promises.33 Of course, even if this is what Ross says, there is aneasy qualification available, and that is to say that every commitment nec-essarily gives rise to a prima facie obligation to fulfill it, if the agent can ful-fill it. But there is reason to think that this, too, is false. For it often happensthat people promise to do what is in fact overall wrong. What reason is therethen to think that all commitment-worlds accessible to the agent in whichhe displays fidelity are better than all accessible commitment-worlds thatare minimally different in this respect? For any world that is minimally dif-ferent in this respect will at least involve his not committing the overallwrong in question. Once again, though, a qualification is perhaps avail-able, although it is not so easy to see just how it should go.

31 Or, we might add, some other acts which imply that one has a compound prima facieobligation to do A. Ross does not appear explicitly to acknowledge the possibility ofcompound prima facie obligations.

32 This is most evident at Ross (1930), pp. 37 and 40, but see also pp. 21, 24, 27, and 28.33 Some philosophers would of course respond by denying that "ought" implies "can." I

shall not discuss this issue any further here. Others would respond by claiming that appar-ent promises to do what is impossible are in fact not genuine promises, and then claimthat every genuine promise does in fact ground a prima facie obligation to fulfill it. (See,e.g., Martinich, 1987.) This seems mistaken, in that it seems possible that a person shouldsincerely, and thus genuinely, promise to do something, in the mistaken belief that hecan do that thing.

34 Compare what Donagan says about promising, in Donagan (1977), pp. 92-3.

166

The major complication is this, and it applies to all the particular primafacie obligations that Ross alleges to be self-evident, except perhaps theprima facie obligation to promote intrinsic value.35 The account that I haveprovided relies on the notion of minimal difference, where this is explicit-ly tied to worlds that are accessible to the agent in question. But surely it is pos-sible that some agents be such that they cannot avoid some evil's occurringif they satisfy some Ross-alleged direct prima facie obligation; in such acase, given that the evil is sufficiently great (so that it greatly diminishes thedeontic value of any world in which it occurs and renders any world inwhich it does not occur very valuable, deontically, relative to any world inwhich it does occur), it will not be the case that all accessible worlds inwhich the alleged obligation is satisfied are deontically superior to all acces-sible worlds that are minimally different in this respect. Hence this allegedobligation does not hold for such people; hence it does not hold of neces-sity; hence it is not self-evident that it holds. As an illustration of this, con-sider the possibility that Martha will commit mass murder if and only ifMalcolm acts nonmaleficently, and that Malcolm can do nothing aboutthis. We may suppose (barring certain Kantian considerations) that not allworlds accessible to Malcolm in which he acts nonmaleficently are deon-tically superior to all worlds accessible to him that are minimally differentin this respect. In fact, it may well be that the very reverse is true, so thatMalcolm has (on my account) a prima facie obligation to act maleficently.

Thus it cannot be, on the account that I have provided, that there is anecessary unconditional prima facie obligation to act nonmaleficently. So,too, for beneficence and other obligations that Ross calls "general."Similarly, the conditional obligations of fidelity, gratitude, and so on thatRoss calls "special" cannot hold of necessity, on the account that I haveprovided.

If we are to accommodate what Ross says about the self-evidence ofsuch prima facie obligations, therefore, a different account must be given.Let me indicate briefly here how I think this account should go. First, wewould need to revise our understanding of minimal difference, so as not torestrict our focus to worlds accessible to the agent. We would simply revise(5.7) by saying:

(5.7') W' is minimally different from W with respect to S doing A at T 'iff

35 The example I give below concerns an act that is plausibly thought to involve a greatintrinsic evil. But if there are acts that do not involve great intrinsic evils but whichnonetheless greatly diminish the deontic values of worlds in which they occur, then eventhe prima facie obligation to promote intrinsic value is subject to this complication.

167

(a) S does ,4 at T ' i n W\(b) S does not do A at T ' in W'\ and(c) there is no world W" such that

(1) S does not do A at T ' in W", and(2) W" is closer to W than W' is.

(5.8) would be similarly revised. Then (5.10) would be revised as follows:

(5.10') S has, at T in W, a direct conditional prima facie obligation to doA at T ' , on the condition thatp is true, iff(a) there are a world W' and a world W " such that

(1) W' is accessible to S from W at T,(2) W " is accessible to 5 from W at T, and(3) H7" is minimally different from W' with respect to S doing A

at T' , on the condition thatp is true; and(b) for all worlds W' and W"\

if (1) W7" is minimally different from Wf with respect to 5doing A at T', on the condition that p is true,

then (2) the deontic value for 5 at T of W' is greater than thedeontic value for S at T of W .

Changes to (5.9') and (5.11)—(5.24) would then be made accordingly. Onthis approach, the personal optionality of what is prima facie obligatory ispreserved (see subclauses (al) and (a2) of (5.10')), but it could still be thatMalcolm has a prima facie obligation to act nonmaleficentry despite hisinability to avoid Martha's committing mass murder if he does so act.

Why, then, did I not simply proceed in this fashion in the first place?The answer has to do with detachment. It is not that I think such a proce-dure mistaken; on the contrary, it gives us an account of what might becalled prima facie obligations "in principle." We could then say, with Ross,that everyone necessarily has a reason, in principle, to act nonmaleficently;as noted, this would apply even to Malcolm. But the problem is that,although versions of necessity detachment and deontic detachment doindeed hold for conditional prima facie obligations in principle, they do notwarrant certain judgments that we wish intuitively to make. For example,the version of necessity detachment that concerns prima facie obligationsin principle involves inferences of this form: if all -worlds in which S doesA are better than all p-worlds that are minimally different in this respect,and all worlds are p-worlds, then all worlds in which S does A are betterthan all worlds that are minimally different in this respect. Note: "all

168

worlds," not just "all accessible worlds"; this is because minimal differenceis now being understood in terms of possible, and not just accessible,worlds. Now, with necessity detachment so understood, we could notemploy it to infer, as we surely wish to, that, for example, someone whohas made a promise has an unconditional prima facie obligation to keep it.For, although a promise is inevitable once made, it is not necessary, in thestronger sense of "necessary" now at issue. Similarly, for deontic detach-ment with respect to prima facie obligations in principle, the uncondition-al obligation that triggers the detachment must itself be an obligation inprinciple, and this fails to validate certain detachments that we intuitivelywish to accept (such as that of detaching an unconditional obligation to doA by appealing to an unconditional obligation to promise to do A, the lat-ter having itself been detached - by necessity detachment - owing to con-siderations having to do merely with inevitability and not with necessity inthe stronger sense).

The fact is that there are, I think, two types of prima facie obligations,and thus there are also two categories of detachment. There are prima facieobligations in principle, and there are prima facie obligations in practice. Thusthere are also (necessity and deontic) detachment in principle (where theobligations in question are prima facie obligations in principle) anddetachment in practice (where the obligations in question are prima facieobligations in practice). The detailed account that I have provided is ofprima facie obligations and detachment in practice; the account that I havesketched in the last two paragraphs is of prima facie obligations and detach-ment in principle. We need both. We need the latter to accommodate whatRoss says about the self-evidence and necessity of prima facie obligations;we need the former to accommodate common intuitions concerning thedetachment of prima facie obligations.

5.4.3 Prima facie obligation and overall obligation

Let me turn now to what Ross has to say about the relation between primafacie obligation and overall obligation. Here he makes two importantclaims. The first is this:

I suggest 'prima facie duty' or 'conditional duty' as a brief way of referring to thecharacteristic.. .of being an act which would be a duty proper if it were not atthe same time of another kind which is morally significant.36

36 Ross (1930), p. 19.

169

This might be interpreted as follows:

(5.25) S has a prima facie obligation to do A iff S would have an overallobligation to do A if any prima facie obligation that 5 has not to doA were overridden.

Does my account imply that this is true?It might be thought that any account of obligation that purports to cap-

ture Ross's concept of prima facie obligation must imply that (5.25) is true.But the fact is that this thesis is not implied by my account (although theyare consistent with one another). The implication fails because my accountof prima facie obligation incorporates, as noted earlier, a certain generality,and such generality, which I take to be essential to prima facie obligations,37

is not incorporated in my account of overall obligation. Does this show upa failing in either of these accounts?

I don't think so. Ross seems to think that there are several basic, inde-pendent values (such as those of fidelity, gratitude, and so on); that noth-ing else is valuable (at least with respect to the determination of obligation)unless it is a means to the promotion of these values; and that there is always,indeed necessarily, a deontic reason to promote these values. If we grant allthis, and we also grant (as seems natural) that the deontic value of a worldis a direct function of those deontically relevant values which are promot-ed at that world, then, given the foregoing accounts of prima facie andoverall obligation, it seems that we could argue cogently for (5.25); forthere will be no deontically relevant value that does not correspond to someprima facie obligation.38 But whether or not we should grant all this seemsto me to go well beyond questions of conceptual analysis; it is a normativequestion. (See Subsection 2.3.3.) This being the case, it is not a failure in(I) that it fails to imply (5.25). In fact, it would be a failure in (I) if it betrayedsuch a normative bias.

Nonetheless, (5.25) might seem promising as the basis of an analysis ofthe concept of prima facie obligation, an analysis that is quite different fromthe one that I have proposed. Of course, (5.25) could not itself be accept-ed as such an analysis, because it would be circular. But consider thisproposal by Frank Snare:

37 Indeed essential, in some form or other, to reasons of all sorts.38 This is not to say that it is easy to see just how this argument would go. There are a num-

ber of complications, having to do with the distinctions between conditional and uncon-ditional obligations, obligations in principle and obligations in practice, permanent versusuniversal versus necessary obligations, and with other matters.

170

(5.26) 5 has a prima facie obligation to do A iff for some set T of act-types(a) A is a token of some member of T, and(b) if the omission of A were not also a token of some member of T,

then S would have an overall obligation to do A.39

Intuitively, T is supposed to be that set of morally significant act-types invirtue of which an act-token may be prima facie obligatory. Circularity isavoided, however, since such identification of T is not necessary for thestatement of (5.26). In addition, the proposal would appear to avoid thenormatively biased implication that whatever is overall obligatory is primafacie obligatory, insofar as it requires that, to be prima facie obligatory, Abe a token of some member of T, whereas this is not said to be a require-ment for an act's being overall obligatory.

But Snare's proposal is nonetheless problematic. It would seem to fallafoul of the fact that subjunctive conditionals (of the sort employed in theproposal) warrant factual detachment. For might it not happen that,although the antecedent of clause (b) in (5.26) is satisfied, S ought so to actthat it is not satisfied? Consider this case. Ben is in a position at T1 to per-form some beneficent act A at T2. Let us assume that acting beneficently isa type of act that is a member of set T, and that Ben (therefore) has a primafacie obligation at Tt to do A at T2. Let us suppose also that, if he omits todo A at T2, he will not thereby perform an act that is of a type that is also amember of T. In this case, (5.26) implies that Ben has an overall obligationat Tt to do A at T2. But suppose that Ben ought overall at T1 to make andthen keep a certain promise, and that the keeping of it would require himto omit to do A at T2. He doesn't make this promise, and that is why hisomitting to do A is not otherwise morally significant. (The omission is notan instance of fidelity, for example.) But he ought overall both to make itand to keep it, and so (contrary to what (5.26) implies) he ought overall notto do A.

One might respond to this case by maintaining that Ben's omitting todo A must be morally significant: even though it is not an instance of fidelity(for he doesn't make the promise), it must nonetheless be an instance ofsome member of T, since after all he does not do what he overall ought todo. But this would be to revert to the normative bias of (5.25), for it wouldbe to insist that nothing is overall obligatory unless it exemplifies somemember of T, that is, unless it is prima facie obligatory.

39 Snare (1974), p. 237, somewhat revised. A similar proposal is made in Atwell (1978).

171

Another response would be to try to fix Snare's proposal by tinkeringwith its clause (b) so as to circumvent this objection,40 but I don't think thiswould do. For the proposal in fact does not escape the normative bias of(5.25) after all. This is because counterfactuals of the sort used in its clause(b) are true if both their antecedent and their consequent are true. Now,suppose that S has an overall obligation to do A; then (b)'s consequent istrue. Next, take any type of action of which A is a token and let T be theset which has that type as its only member; then (a) is true and (b)'santecedent is also (likely to be) true. Hence both (a) and (b) are true, andfrom this Snare's proposal allows us to infer that 5 has a prima facie oblig-ation to do A.

For these and other42 reasons, then, I think that Snare's proposal is to berejected, and in general the project of giving an analysis of the concept ofprima facie obligation on the basis of (5.25) seems to me unpromising.

I began this subsection by saying that Ross makes two important claimsabout the relation between prima facie obligation and overall obligation.The second claim is this:

Every act..., viewed in some aspects, will be prima facie right, and viewed inothers, prima facie wrong, and right acts can be distinguished from wrong actsonly as being those which, of all those possible for the agent in the circum-stances, have the greatest balance of prima facie tightness, in those respects inwhich they are prima facie right, over their prima facie wrongness, in thoserespects in which they are prima facie wrong.43

This might be interpreted as follows:

(5.27) S has an overall obligation to do A iff S has a prima facie obliga-tion to do A that overrides any prima facie obligation that S has notto do A.

Does my account imply that this is true?Again, the answer is that it does not. The reason for this is the same

as that given before when discussing (5.25): there is no guarantee that

40 For example, one might try modifying clause (b) so that it reads as follows: if the omis-sion of A were not also a token of some member of T, and if it were not the case that S ought[or, better, may] overall so act that the omission of A is a token of some member of T, then S

would have an overall obligation to do A.41 Cf. Lewis (1973), p. 26; Pollock (1976), p. 38.42 The presupposition that there is a single act-token corresponding to the omission of A

is problematic, as is the presupposition that obligatoriness applies to act-tokens. On thelatter point, see Subsection 2.2.5.

43 Ross (1930), p. 41.

172

the generality inherent in prima facie obligation is also present in overallobligation.

Even though (I) implies neither (5.25) nor (5.27), it is certainly consis-tent with them. With respect to (5.27), though, this claim might be dis-puted, on the grounds that Ross conceives of it as an analysis.44 If overallobligation is to be understood in terms of prima facie obligation, as thisinterpretation of (5.27) contends, then it is not to be analyzed indepen-dently of the concept of prima facie obligation, as (I) contends.

Now, I think it must be granted that, if (5.27) is accepted as an analysisof overall obligation, my proposed analysis of overall obligation must berejected.451 therefore reject (5.27) as an analysis (though I grant that it mayexpress a necessary truth). How does Ross himself view (5.27)? The evi-dence seems to me inconclusive: although some passages46 in The Right andthe Good suggest that he does accept (5.27) as an analysis, he never states thisexplicitly; moreover, other passages47 suggest that he eschews any suchanalysis. At any rate, if Ross does accept (5.27) as an analysis, I oppose himon this point. I leave it to you to judge whether this indicates an inadequa-cy in my account.

One reason to suspect that it does indicate an inadequacy might be this.As has been noted, Ross claims that (certain of) one's prima facie obliga-tions are self-evident, but he also claims that one's overall obligations arenot and that one may appeal to the former when trying to determine thelatter. This is surely plausible, but it may appear to indicate a priority ofprima facie obligation over overall obligation that is inconsistent with (I).Not so. As Snare has pointed out, the logical or conceptual independenceof overall obligation from prima facie obligation is not inconsistent withthe former being epistemologically dependent on the latter.48 Similarly,one might claim that sometimes it is appropriate to appeal to one's primafacie obligations in order to explain what one's overall obligations are.49 But,once again, the account given in (I) is not inconsistent with such priorityof prima facie obligation over overall obligation. For it may well be thatcertain matters are, and that we find them to be, of fundamental moral

44 Others have certainly proposed analyses along the lines of (5.27). See, in particular,Chisholm (1974), pp. 11-13. See also McCloskey (1963), pp. 336 and 343; Loewer andBelzer (1991), pp. 365-6.

45 This claim presupposes that there cannot be two acceptable analyses of a single concept,where one of these analyses itself employs concepts that the other does not.

46 Ross (1930), pp. 41 and 46-7.47 Ross (1930), pp. 10-12 and 16.48 Snare (1974), p. 244.49 Cf. again Ross (1930), p. 46. Cf. also Pietroski (1993), p. 513.

173

significance —well-being, perhaps, or respect, or integrity, or fidelity, or jus-tice, or freedom, or something else — such that our overall obligations turnon, and we believe them to turn on, whether these values are promoted.Far from wishing to deny this, I believe it to be true. No doubt, then, wewill agree that there are prima facie obligations, both in principle and inpractice, having to do with the promotion of these values, so that some-times, when we deliberate about what we ought overall to do in some sit-uation, we may bear in mind that these prima facie obligations apply in thissituation. But this is quite consistent with the analyses of overall and primafacie obligation that have been proposed here.

5.4.4 Residual obligation

Steve missed his dinner date with Dave. Although he was perfectly justi-fied in doing so - for Carl was in dire need of his ministrations - still therewas a cost involved, a moral cost. He broke his promise, and Dave was dis-appointed. For this reason, Steve's helping Carl appears to have left a cer-tain "moral residue," as it is sometimes put. Ross distinguishes two types ofmoral residue, one involving emotion, the other action. He says:

When we think ourselves justified in breaking, and indeed morally obliged tobreak, a promise in order to relieve some one's distress, we do not for a momentcease to recognize a prima facie duty to keep our promise, and this leads us tofeel, not indeed shame or repentance, but certainly compunction, for behav-ing as we do; we recognize, further, that it is our duty to make up somehow tothe promisee for the breaking of the promise.50

Few would quarrel with this, and yet it is a fact that my account of primafacie obligation does not imply that it is true. Is this a problem with theaccount?

I cannot see that it is. There might be a problem if the account impliedthat Ross's view is false, but it doesn't. It is neutral with respect to the view,and this seems perfectly acceptable, since the view is clearly not simply aconceptual one; it is normative.51 To say that it is appropriate to feel suchcompunction (as Ross obviously believes; it is not just that we are led to feelit, we ought to feel it),52 and to say that there is a (presumably prima facie)obligation to make amends — these are normative claims. However mis-

50 Ross (1930), p. 28.51 "Clearly" may be a bit bold, given the remarks made in the penultimate paragraph of

Subsection 2.3.3.52 This is a nonbinding sense of "ought," for reasons given in Subsection 3.1.7.

174

taken someone might be to deny them, the mistake would not be a concep-tual one.

Indeed, although the view may be natural, it is clearly not uncontro-versial. Some would claim that, as long as one does what is morally justifi-able, then there is no cause for compunction at all.53 (Donagan's position,cited in Section 5.1, would seem to suggest this.) Although I think that thisis false, nonetheless we would do well to note two additional points. First,even if compunction (whatever that is, exactly) is called for in cases likeSteve's, certain other "negative" moral emotions (such as guilt or shame)certainly are not; second, even if compunction is called for, still, as long asone has clearly done what one overall ought, an overall sense of moral easewould seem appropriate.54

Similarly, some would claim that, as long as one acts inculpably, theneven if, like Steve, one breaks a promise to someone, one does not owe thatperson anything. Indeed, /would claim this (where it is understood that tosay that S owes S' something is to say that 5 has an obligation to S' to dosomething. Here I am talking of what in Section 1.3 I called "obligation-to" as opposed to "mere obligation." I shall say more about obligation-toin Section 5.5). I acknowledge that this claim may well initially appear quiteimplausible; my reasons for making it are too complicated to go into here,however.55 Even if I am right, though, this still leaves untouched the close-ly related claim that, for example, Steve has a prima facie obligation (evenif not a prima facie obligation-to-Dave) to make amends to Dave. Thisseems to me likely to be true, although even here there are difficulties. First,it must be acknowledged that the case could be filled in so that Steve clear-ly doesn't have such a prima facie obligation; any scenario in which Steve isincapable of making amends to Dave would have this implication. (Stevemight die; Dave might die or move away; and so on.) Second, it seems clear,on reflection, that not every failure to satisfy a prima facie obligation givesrise to a residual obligation to make amends, even when one is not inca-pacitated in the relevant way. Suppose that Steve had a prima facie obliga-tion to visit Dave on Monday, grounded not in any commitment to Davebut simply in beneficence. Had he then helped Carl en route and therebyrendered himself unable to visit Dave on Monday, certainly no explanationor apology to Dave would seem called for.56 Perhaps a residual obligation

53 Gowans discusses this as an implication of what he calls the "Elimination Thesis" (asopposed to the "Remainders Thesis"), in Gowans (1994), pp. 91-2.

54 Cf. Zimmerman (1993c) on the possibility of the "resolution" without "dissolution" ofambivalence.

55 See Zimmerman (1994).56 Cf. McConnell (1995); Pietroski (1993), p. 510, n. 27.

175

to visit Dave on Tuesday emerges from this, but perhaps not; for perhapsthe condition that gave rise to the desirability of Steve's visiting Dave onMonday passes before Tuesday arrives.

Indeed, this seems to get to the heart of the matter. Is there a condition,arising out of the failure to satisfy the original prima facie obligation, thatgives rise to the desirability of some further action? If so, then it is likely thatsome residual prima facie obligation will have emerged. But not necessari-ly; the agent might not have the relevant capacity. Indeed, it may not beonly the original agent who incurs such an obligation. If Steve was walk-ing with Willis when he came upon Carl, then perhaps, while Steve wasbusy tending to Carl, Willis had the obligation to call Dave and let himknow what was happening.57 Here an obligation emerges from Steve'sbreaking his promise to Dave, but it is Willis's, not Steve's.

In general, then, we may say that, if residual obligations emerge fromthe failure to satisfy prima facie obligations, this is perfectly consistent withthe account of prima facie obligation presented here. Whether or not theydo, however, is a complicated business, and it is certainly no defect in theaccount that it fails to imply that they do.

Finally, it should be noted that residual obligations can arise not just fromthe failure to satisfy prima facie obligations. They can arise also where anoverall obligation is not satisfied (whether or not this involves the failure tosatisfy a prima facie obligation). As discussed in Section 3.2, often, when anoverall obligation is not satisfied, a shift in obligations occurs, and a newoverall obligation emerges. Thus Mandy became obligated to babysit Bill'schild, and Victor became obligated to visit Bert. Also, as discussed inSection 4.4, often, when an n-level overall obligation is not satisfied, antt+1-level overall obligation emerges. Thus you became obligated to votefor Dick, and Alan became obligated to attend the meeting on the secondfloor. These are all naturally classified as residual obligations; just likeSteve's residual obligation to apologize to Dave, they have emerged fromthe failure to satisfy some initial obligation.

5.5 RIGHTS

5.5.1 The correlativity of rights and obligations

The term "right" is ambiguous, even when used as a noun. WesleyNewcomb Hohfeld made a lasting contribution to our understandingof rights when he distinguished among four species of them: "claims,"

57 Cf. McConnell (1995).

176

"liberties," "powers," and "immunities."58 According to him, the first twoare related in this way (to put the matter somewhat roughly):

(5.28) S' has a claim against S that S do A iff S has no liberty [is not at lib-erty] as regards S' not to do A.

Hohfeld also held that a similar relationship binds powers and immunities.My interest here is with how moral rights are related to moral obliga-

tions. This issue has already been briefly addressed in Section 1.3. There thefollowing four propositions (as formulated by Ross) were distinguished:

(1.4) a right of A against B implies a duty of B to A;(1.5) a duty of B to A implies a right of A against B;(1.6) a right of A against B implies a duty of A to B;(1.7) a duty of A to B implies a right of A against B.

I gave reasons for rejecting (1.6) and (1.7) but said that I accept (1.4) and(1.5). I wish now to elaborate a little on this treatment of (1.4) and (1.5)

The rights at issue in the foregoing statements are claims. Hohfeld, too,accepts (1.4) and (1.5), which may now be formulated more fully as fol-lows:

(1.4') if S' has a claim at T in W against S that S do A at T', then S has anobligation at T in W to S' to do A at T';

(1.5') if S has an obligation at T in W to S' to do A at T\ then S' has aclaim at T in W against 5 that S do A at T'.

Combined, these statements yield the thesis that claims and obligations arein a certain way correlative to one another. The following points should benoted.

First, given the relationship between claims and liberties just noted in(5.28), (1.40 and (1.50 imply the following:

(5.29) S has an obligation at T in W to S' to do A at T ' iff 5 is not at lib-erty at T in W as regards S' not to do A at T'.

Second, the correlativity of claims and obligations is compatible withthe normative priority of claims over obligations. It seems plausible to saythat obligations derive from claims, rather than vice versa. For example, it

58 Hohfeld (1919), p. 35ff. Hohfeld was explicitly concerned with giving a taxonomy oflegal rights, but his account is plausibly extended to moral rights also. Also, Hohfeld tend-ed to prefer the term "privilege" to the term "liberty," although he acknowledged theappropriateness of the latter (at Hohfeld, 1919, pp. 42-3). Most commentators appear toprefer "liberty" to "privilege."

177

is a promisee's claim against a promiser that gives rise to the latter's duty;the duty doesn't give rise to the claim. If this is so, and whatever exactly itinvolves, it is perfectly consistent with the joint assertion of (1.4') and (1.5').

Third, (1.4') and (1.5') are intended to be consistent with the possibil-ity that there be posthumous rights and also with the possibility that therebe rights held by persons not yet born or even conceived. In such cases itwould have to be true that S' has a claim at T even though S' does not existat T. I doubt whether this is indeed possible, but I don't wish to rule it outhere.59 Note that allowing for this possibility does not require allowing forthe possibility that there be post- or prevital obligations. Clearly it is impos-sible for someone who has existed but will not exist to have obligations, forsuch a person no longer has any choice among accessible worlds. I wouldsay, too, that it is impossible for a not-yet-existing person to have obliga-tions (even remote ones), for I think that there being worlds accessible atsome time to someone requires that that person actually exist at that time.But these are all difficult metaphysical issues that need not be resolved here.

Fourth, (1.4') and (1.5') serve to emphasize the distinction, introducedin Section 1.3, between mere obligation and obligation-to. S's having anobligation-to-S' is not the same as S's having an obligation towards or con-cerning S\ Whether or not S has an obligation-to-5' to do A is a questionof whether or not S owes it to S' to do A; whether or not S has an obliga-tion concerning S' is a question of whether or not the obligatory action,A, itself somehow involves S'. Often, of course, the two will coincide. IfPru has promised Zoe to take her to the zoo, then (let us assume) Pru hasan obligation-to-Zoe to take Zoe to the zoo. Here, Pru's obligation-to-Zoe also concerns Zoe. But sometimes the one to whom an obligation isowed is not the one whom the obligation concerns. If Pru has promisedOllie to take Zoe to the zoo, then Pru's obligation still concerns Zoe, butit is an obligation she owes to Ollie, not Zoe.60

Fifth, (1.4') and (1.5') serve to emphasize a related distinction, thatbetween mere claims and claims-against. For all that I have said so far, itmay be that someone can have a claim concerning something that is not aclaim-against-someone (or -something)61 to that thing. Thus (1.4') and(1.5') do not assert that there is a correlativity between claims and obliga-

59 Cf. Feinberg (1980), pp. 173-6 and 180-2.60 Cf. Hart (1979), p. 18.61 It is sometimes asserted that we can, and do, have claims against entities that are not them-

selves individual persons. If there is to be correlativity in this sort of case, then it must bethat these entities can and do have obligations. This is an issue that I shall take up inChapter 9.

178

tions generally; they assert that there is a correlativity between claims-against and obligations-to in particular. Objections that have been raisedagainst the correlativity thesis would appear to succeed, if they succeed atall, only in impugning the general thesis; they leave the more particularthesis untouched.

For example, David Lyons notes:

It is often said...that one should be charitable and generous — that one has aduty to behave in that way. This appears to make good sense, even though noother person could have a right to one's charity or generosity (for if the otherhad a right, then what would be required of one would not be classifiable ascharity or generosity).

Of course, it is controversial whether or not one has an obligation to dis-play charity or generosity. As noted in Section 1.3, it is arguable that allobligation is obligation-to, and thus it might be argued that, if no one hasa claim to charity, then no one has an obligation to display charity. This,however, serves only to confirm the correlativity of claims-against andobligations-to. If there is an obligation to display charity but no claim tocharity, then the obligation simply isn't an obligation owed to anyone (oranything)63 in particular, and so the more particular correlativity thesis isnot shown false by these considerations.64

Similarly, Joel Feinberg states:

The manifesto writers...who seem to identify needs, or at least basic needs,with what they call "human rights," are more properly described...as urgingupon the world community the moral principle that all basic human needsought to be recognized as claims.. .worthy of sympathy and serious considera-tion right now, even though, in many cases, they cannot plausibly be treat-ed. . .as grounds of any other people's duties.65

Although Feinberg does not wholeheartedly endorse this way of talking,he is nonetheless sympathetic to it. The point to be made here, though, issimply that, even if we accept the propriety of talking of basic needs asconstituting claims in those cases where no one is in a position, and henceno one has an obligation, to satisfy those needs, all this would mean is that

62 Lyons (1979a), p. 11.63 Just as it might be (as stated in note 61 to this chapter) that entities that are not themselves

individual persons have obligations, so too it might be that such entities have claims.64 Lyons would appear to agree; see Lyons (1979b), pp. 60-1. So too would Feinberg;

see Feinberg (1979), p. 79.65 Feinberg (1979), p. 89.

179

it is possible for there to be claims that are not claims held against anyone(or anything) in particular; and so, once again, the more particular correl-ativity thesis is not threatened.

There is one sort of case, however, which I think poses a more seriouschallenge to this correlativity thesis, in particular to the latter half of it,(1.5'). It is commonly said that some moral duties are self-regarding. Forinstance, Kant is famous for asserting that there is a duty not to commit sui-cide and a duty to develop one's talents.66 Of course, this might simply beinterpreted as an assertion that we have duties concerning ourselves, in whichcase they would not pose any challenge to (1.5'). And, even when so weak-ly interpreted, some would dismiss the assertion on the grounds that moral-ity has to do only with how we treat other people. This seems mistaken,though. Considerations concerning self-respect would seem to indicatethat morality does indeed have in part to do with how we treat ourselves.67

If this is granted, then I don't think we can rest content simply with sayingthat some of our duties concern ourselves. It would seem, rather, that we oweit to ourselves to do certain things for ourselves. Yet it also seems odd to saythat we have claims against ourselves. This is the challenge to (1.5').68

Other than dismissing self-regarding duties altogether or simply rele-gating them to the category of duties that merely concern the agent, thereare two possible responses to this challenge, short of capitulation. One is,of course, to bite the bullet and insist that we do indeed have claims againstourselves; it is just that we are not likely to press or stand on these claims ifthey appear to be in danger of being flouted. I am not sure what to makeof this response. A second response is to try to plumb a middle groundbetween mere obligation concerning oneself and obligation-to-oneself.According to this response, self-regarding obligations, though not strictlyowed to oneself, nonetheless have an essential connection to oneself in thatthey are grounded in considerations of self-respect; thus they do not mere-ly concern oneself by accident, as it were.

I find the second response promising, although I concede that, unela-borated, it remains pretty obscure. It is not my intention to try to clarify ithere, though. If it should ultimately fail, then perhaps we must grant that(1.5') is false. Even then a closely related claim may be accepted, and thatis (1.5') conjoined with the stipulation that 5 is distinct from S'. ((1.4') willstill stand without qualification.)

66 Kant (1964), pp. 89-90.67 Cf. Hill (1979).68 Cf. Lyons (1979a), pp. 11-12.

180

5.5.2 Absolute and prima facie rights

Like Ross, I am inclined to say that those obligations to which rights(claims) are correlative are prima facie, and (1.4') and (1.5') are intended tobe so construed. This means that rights are themselves prima facie and so,like the obligations to which they are correlative, subject to being overrid-den.69 It means also that all those distinctions made earlier with respect toprima facie obligations — concerning whether they are direct or indirect, inprinciple or in practice, unconditional or conditional, and whether they arepermanent, universal, or necessary - apply also to rights. Consider the(alleged, and suitably qualified) right of a promisee against a promiser to thelatter's fidelity. Corresponding, as it supposedly does, to a direct prima facieobligation, it will be a direct right. As such, it can and will ground all sortsof incidental rights; for example, if the promiser must drive to Boston toensure that he manifests fidelity, then the promisee has the right that thepromiser drive to Boston. Also, as was noted earlier, the prima facie obli-gation of fidelity is a conditional one, the condition being that a commit-ment is made; if and when the commitment is made, then the obligationbecomes unconditional (at least in practice, if not in principle). So too,then, a (potential) promisee's right is conditional on the promise beingmade; if and when it is made, the right becomes unconditional. Finally, ifa (suitably qualified) prima facie obligation of fidelity is a necessary one,then the correlative right is necessary also.

Note that the failure to satisfy a right that is prima facie does not neces-sarily constitute overall wrongdoing. If the right is merely prima facie andone fails to satisfy it because one satisfies instead another right that overridesit, then one does no overall wrong (unless there is a yet more stringent rightthat one has failed to satisfy). Useful terminology to adopt here is JudithThomson's:70 we may say that, if a right (that is, a claim) is not satisfied, thenit is infringed; if such infringement is overall wrong, then the right is violat-ed. We may also say that, if a right has been violated, then the violator hasnot only done (overall) wrong, he has done a wrong; equivalently, he haswronged the right-holder.

Some philosophers, however, conceive of rights as correlative not tomerely prima facie obligations but to overall obligations. This is to conceiveof rights as absolute. Let us now investigate this conception of rights.

69 Roughly, a right to X overrides a right to Yjust in case the obligation that correspondsto X overrides the obligation that corresponds to Y. Cf. Thomson (1990), for a recent,thorough treatment of rights understood as prima facie.

70 See Thomson (1986), p. 40, and (1990), p. 122.

181

First, it will be helpful to distinguish degrees of the absoluteness of rights.There are at least six. We may say, roughly, that someone has an absoluteright of the lowest degree - call it a grade-1 right -just in case someoneowes him a certain overall obligation; he has a grade-2 right just in casesomeone always owes him such an obligation; he has a grade-3 right just incase everyone always owes him such an obligation; there is a grade-4 rightjust in case everyone of a certain sort is such that everyone always owes himsuch an obligation; there is a grade-5 right just in case everyone is such thateveryone owes him such an obligation; and there is a grade-6 right just incase it is necessarily the case that everyone is such that everyone always oweshim such an obligation. More precisely:

(5.30) S' has a grade-1 absolute claim at T in W against S that 5 do A atT iff 5 has an overall obligation at T in W to S' to do A at T';

(5.31) S' has a grade-2 unconditional absolute claim in W against S thatS do A iff for all times T and T',if (a) there is a world accessible to S from W at T in which S does

AztT', and(b) there is a world accessible to 5 from W at T in which 5 does

not do A at T',then (c) S' has a grade-1 absolute claim at T in W against S that S do

A at T\(5.32) S' has a grade-3 unconditional absolute claim in W that A be done

iff for all persons S,if (a) S exists in W,then (b) S' has a grade-2 absolute claim in W against S that S do A;

(5.33) there is a grade-4 unconditional absolute claim in W that A be doneiff for all persons S',if (a) S' exists in W and has a certain characteristic C,then (b) S' has a grade-3 absolute claim in W that 4 be done;

(5.34) there is a grade-5 unconditional absolute claim in W that A be doneiff for all persons S',if (a) S' exists in W,then (b) S' has a grade-3 absolute claim in W that A be done;

(5.35) there is a grade-6 unconditional absolute claim that A be done ifffor all worlds W, there is a grade-5 absolute claim in W that A bedone.

I think it is clear that other degrees of absoluteness might be distinguished,but the foregoing should suffice to give an idea as to how absoluteness cancome in degrees. (Note that (5.33) can itself be "subdivided," depending

182

on what the characteristic, C, is taken to be. It might be a relatively simplecharacteristic — innocence, for example — or a relatively complex charac-teristic — innocence-cum-parenthood, say.) In addition, each of the fore-going statements is supposed to be adequate to an account of so-callednegative rights (where, to be more explicit, reference to doing A would bereplaced by reference to not doing A).

The foregoing account of unconditional absolute rights can be straight-forwardly supplemented with an account of conditional absolute rights.We can say:

(5.36) S' has a grade-1 absolute claim at T in W against S that 5 do A atT', on the condition that p is true iff S has an overall obligation atT in W to S' to do A at T', on the condition thatp is true;

(5.37) S' has a grade-2 conditional absolute claim in W against S that 5 doA iff for all times T and T', there is a condition^ such that,if (a) there is a world accessible to 5 from W at T in which S does

A at T' andp is true, and(b) there is a world accessible to 5 from W at T in which S does

not do A at T' and p is true,then (c) S' has a grade-1 absolute claim at T in W against S that S do

A at T', on the condition that|> is true;

and so on through the various other grades. It might thus be claimed (truth-fully or not; presumably not) that there is a grade-6 conditional absoluteright to fidelity.

Now, certainly no proponent of rights will quarrel with the claim thatsome moral rights are grade-1 absolute. For even those who regard rightsas correlative to prima facie obligations will accept that sometimes theobligation in question is not merely prima facie but overall. But problemsarise when it is claimed either that some moral rights are grade-2 absolute(or grade-3, or higher) or that all moral rights are absolute (whether of grade1 or higher). Let us examine these issues.

To claim that a moral right is grade-2 absolute (or higher) is to claim thatits unoverridden status holds on more than one occasion. The greater thedegree, the greater the generality of this status. What is problematic is that,given the vast variety of considerations that may come into conflict withthat consideration which the right in question represents, the greater thegenerality of "unoverriddenness" that one seeks to establish in the right,the less likely it seems that one will succeed in establishing it. Still, somehave claimed success in such an endeavor. Alan Gewirth, for example, hasargued that a mother's right not to be tortured to death by her son is

183

absolute, 'where by "absolute right" he means a right that "cannot be over-ridden in any circumstances."71 Here he appears to be contending that theright in question is at least grade-4 absolute. (He claims that there are othersuch rights, too, grounded in the general principle that there is an absoluteprohibition against degrading persons.)721 shall not dispute what Gewirthsays here, although it is surely open to debate. What can be said here is this:if there are any such rights, there are surely not many of them. No one, forexample, is likely to wish to grant such an absolute status to the right to becompensated for injury, or to the right that others not damage one's prop-erty, and so on.

In general, then, it would appear that the greater the degree (that is, thehigher the grade) of absoluteness, the fewer absolute rights there are. Butthis claim must now be qualified; for it can be argued that there are just asmany absolute rights as there are prima facie rights. One can argue this byemploying a rather obvious technical move (one that Gewirth himselfrepudiates, though for reasons specific to his own particular agenda of iden-tifying certain rights as being especially significant),73 and that is to buildinto the description of the content of a right a list of conditions under whichit would be overridden if its content were not qualified in this way. Forexample, it is often said that the right to life (understood as the right not tobe killed) is, though particularly weighty, not absolute (in a sense of"absolute" that would seem to indicate at least grade 5, since the right isreferred to simply as "the" right to life, with no restriction to particular per-sons or times), inasmuch as it is sometimes overall justifiable to kill people.One response is to "downgrade" the right by restricting it to all innocentpeople (thus understanding it to be a grade-4 absolute right, with C inter-preted as innocence). But this response is frequently rejected on the basisthat it is sometimes overall justifiable to kill innocent people. To this onemight respond that it is never overall justifiable to kill innocent peopleunless a certain (probably disjunctive) set of conditions Q through Cn

obtains74 and then declare that there is a grade-4 absolute conditional rightthat innocent persons have not to be killed, where the condition in ques-tion is that this set does not obtain. (Or, indeed, one might "upgrade" theright back to its original status and claim that there is a grade-5 absolute con-ditional right that all persons have not to be killed, where the condition in

71 Gewirth (1981), p. 2; the right is asserted on p. 8 and defended in ensuing pages.72 Gewirth (1981), p. 16.73 Gewirth (1981), p. 5.74 Cf. Donagan (1977), p. 72.

184

question is that the set of Ct through Cn does not obtain and the persons inquestion are innocent.) Indeed, this tactic of redescribing the content ofrights could be used to argue that there are in fact no prima facie rights atall; rather, all rights are absolute, all (or certainly most) having this sort ofextensive, complicated, conditional content. But why think of rights in thisway?

The most prominent proponent of the view that all rights are absoluteis Joel Feinberg, at least in certain of his writings.75 According to Feinberg,rights (of the sort that are correlative to obligations) are not just claims butwhat he calls "valid claims." A claim is valid, he says, when "the balance ofreasons" favors "recognition" of the claim.76 Elsewhere, he says that a validclaim is "a decisive case, invulnerable or conclusive."77 It seems clear thatFeinberg here takes rights to be claims that are not overridden; for this rea-son the obligations that they impose are not overridden either. Feinbergreaches this conclusion through an examination of how legal rights func-tion. He says:

"Having a claim".. .is an expression very much like the legal phrase "having aprima facie case." A plaintiff establishes a prima facie case for the defendant's lia-bility when he establishes grounds that will be sufficient for liability unless out-weighed by reasons of a different sort that may be offered by the defendant...In a parallel "prima facie sense" of "claim," having a claim to Xis not (yet) thesame as having a right to X, but is rather having a case of at least minimal plau-sibility that one has a right to X, a case that does establish a right, not to X, butto a fair hearing and consideration. Claims, so conceived, differ in degree: someare stronger than others. Rights, on the other hand, do not differ in degree; noone right is more of a right than another.

Now, I think that Feinberg may well be correct in what he says here, aslong as we restrict his account to legal rights. Suppose that Paul and Paulahave parted; their divorce was bitter, and now each is laying claim to a cer-tain indivisible piece of property. A judge rules that the item belongs toPaula. In this case (barring an appeal and overrule) it is Paula who has the

75 The passages from Feinberg that I am about to quote and that appear clearly to commithim to the view that rights are not merely prima facie, would appear to be inconsistentwith what he says elsewhere (Feinberg, 1980, p. 225ff.) when he denies that all rights areabsolute and endorses Thomson's distinction between the violation of a right and themere infringement of a right.

76 Feinberg (1979), p. 88.77 Feinberg (1980), p. 187.78 Feinberg (1979), p. 88.

185

right to the item, not Paul; this is a right she holds against everyone, and soeveryone is obligated (overall, legally)79 not to steal it, damage it, and so on.

But even if Feinberg is correct in claiming that legal rights are not primafacie but absolute, we may not automatically infer from this (as he appearsto do)80 that moral rights are similarly not prima facie but absolute. Noticethe implication. It would mean that our everyday discussions about moralrights are to a very great extent ill formed. As noted earlier, we could notsay, without qualification, that someone who has been injured has a moralright to compensation; we could not say that a property-owner has a moralright that others not damage his property; and so on. We could not say thesethings because it can happen, as we all know, that it is sometimes overallmorally right for an injurer not to give compensation, for someone to dam-age another's property, and so on, due to overriding considerations. For thesame reason, many current debates would have to be reformulated. If a fetushas a right not to be killed, and this is a (presumably at least grade-4) absoluteright, then no abortion is ever overall permissible; appeals to the rights ofthe mother would be misguided, because either mistaken or irrelevant. Ifa psychiatric patient, who has just expressed to his therapist a clear inten-tion to commit murder, has a right to confidentiality, and this is an absoluteright, then no breach of such confidence is ever overall permissible; appealsto the rights of his intended victim would be misguided, because either mis-taken or irrelevant. But we don't normally find such appeals misguided, andthat is because we don't normally take moral rights to be absolute.

It is this implication of treating all moral rights as absolute — the impli-cation that our everyday discussions of rights would need to be radicallyrevised — that seems to me the best reason to reject such treatment; for thereappears to be no compensating advantage. Others have given other reasons.For example, it has been argued that the denial that certain rights are primafacie (such as the right of property-owners that others not make unautho-rized use of their property) cannot account for the obligation that othershave to compensate for certain actions (such as the overall justified, butnonetheless unauthorized, use of property).81 This argument fails, I believe,

79 It is in fact not at all clear that legal obligations can be merely prima facie. When Feinbergsays that 5's having a (legal) claim to X presents a "prima facie case" that S has a right toX, he is not saying that this imposes a prima facie obligation on others not to interferewith S's use of X. He is, rather, using "prima facie" in a purely evidentiary sense; he is say-ing, that is, that S's having a (legal) claim to X provides some reason (which is in princi-ple open to conclusive rebuttal) to think that S has a (legal) right to X

80 See Feinberg (1979), p. 90.81 See Thomson (1986), pp. 40-1, and Feinberg (1980), p. 230. (On Feinberg's use of this

argument, see note 75 to this chapter.)

186

in that a proponent of the view that all rights are absolute could certainlyseek to explain the (alleged)82 obligation of compensation without resort-ing to the notion of the justifiable infringement of a right.83 The difficultyin specifying the contents of absolute rights (in completing the "unless"-clauses) is also often cited as a reason to reject the view that all moral rightsare absolute. However the significance of this difficulty should not be exag-gerated, for there is a complementary difficulty that besets the view thatmoral rights are prima facie, and that is the difficulty of specifying exactlywhat the conditions are under which whatever moral rights there may beare overridden. (Exactly when, for example, does fidelity trump gratitude?Exactly when does gratitude trump fidelity?) Still, it may be that these dif-ficulties are not completely parallel, for this reason. Suppose that rights areprima facie and that the difficulties in specifying exactly, for all rights, theconditions under which they are overridden, are simply insurmountable;that is, no account of this sort can be given. Even if this is so, we can stillmake explicit reference to such rights; we can name them, discuss them,make certain determinations of their relative weights under certain cir-cumstances, and so on. But on the view that all moral rights are absolute, ifwe cannot fully specify their content, then we cannot make explicit refer-ence to them at all. It is hard to see how, under such circumstances, wecould hope to justify our moral judgments by an appeal to rights. Yet wecommonly attempt to do just this.

In sum, we simply do not take all moral rights to be absolute, and I knowof no good reason why we should do so.

5.5.3 Rights and overall obligation

If moral rights are prima facie, what is their relation to overall moral obli-gation? Here I have no precise account to offer. If I had given a preciseaccount of prima facie obligation-to, then, given the correlativity of primafacie obligations-to and rights, I could have derived a precise account ofrights (or claim-rights and liberty-rights, at least). But I have not given suchan account of obligation-to; the account that I have offered in this chapteris an account of mere prima facie obligation, and I confess myself unable tosay precisely what it is that transforms an obligation into an obligation-to.In addition, if, as I believe, there are obligations that aren't obligations-to,then it cannot be that a determination of what rights pertain in a given case

82 See Zimmerman (1994).83 See Rainbolt (1993).

187

is in general all that is required for a determination of what obligationspertain in that case.

Nonetheless, it is of course possible that an accounting of the pertinentrights is of great help in determining what ought to be done. Rights, wecommonly think, are (frequently) especially weighty considerations. I havealready acknowledged that overall obligation may (on occasion) be episte-mologically or explanatorily dependent on prima facie obligation; that is,that determining what prima facie obligations pertain is (on occasion) help-ful in, and perhaps even required for, determining what overall obligationsthere are. If certain rights are correlative to certain particularly weighty (thatis, difficult to override) prima facie obligations, then it clearly can help,when trying to determine what we ought overall to do, to determinewhether these rights are at issue. Still, for all that, the hope that a determi-nation of the pertinent rights will by itself provide for a quick, uncompli-cated determination of what we ought overall to do seems to me in manycases to be a vain hope.

188

Actualism and possibilism

6.1 THE DISPUTE

Rick has received an invitation to review a book. It would be deonticallybest if he accepted the invitation and submitted his review on time. Thetrouble is, if he accepted, he would not submit the review on time,although he could. Perhaps it is just because he is lazy, perhaps it is becausehe has taken on a great many other tasks, but the fact is that this task, thoughaccomplishable, would remain unaccomplished. It would be deonticallybetter for him to decline the invitation than to accept it but fail to submitthe review on time; for then, at least, someone else could be invited to doit, someone less qualified but someone who would get the job done. Whatought (overall) Rick to do, accept the invitation or decline it?1

The situation may be pictured as in Figure 6.1 (where A stands for"accept" and S for "submit"), and it may be represented as in Chart 6.1.With the case so understood, it is clear what the present account of overallobligation implies: Rick ought to accept the invitation. But a number ofphilosophers believe that this is the wrong answer; they believe that Rickought to decline the invitation. This is not because they are unsympathe-tic with the general idea that one ought to do one's deontic best; on thecontrary, they are impressed by the fact that Rick would do better if hedeclined the invitation than if he accepted it, and it is for this reason thatthey believe he ought to decline it.

This issue can be couched in terms of two broad doctrines, which havecome to be called "actualism" and "possibilism."2 What the actualist says isroughly this:

(6.1) S ought (overall) to do A iff what would happen if S did A is deon-tically superior to what would happen if S performed any alternativeaction;

1 I borrow this illustration (in modified form) from Jackson and Pargetter (1986), p. 235,who themselves borrow it (in modified form) from Goldman (1978), pp. 185—6.

2 These terms are introduced in Jackson and Pargetter (1986).

189

7",

Figure 6.1

1: A & S ...2:-A & ~S ...3: A & ~S ...

Chart 6.1

whereas what the possibilist says is roughly this:

(6.2) 5 ought (overall) to do A iff what could happen if 5 did A is deon-tically superior to what could happen if 5 performed any alternativeaction,

where "could" expresses what in Chapter 2 I have called "personal possi-bility."3 My account accords with possibilism; those who say that Rick

3 These formulations of actualism and possibilism are rough; for instance, there need not bejust one way that things would or could turn out if 5 did A. Still, they should suffice forpresent purposes.

Note the following, however. My use of "would" in the statement of actualism pre-supposes that, if 5 does A at T2 and an event E occurs at T3 as a consequence of this, thenit was true at T? that E would occur at T3 as a consequence of 5's doing A at T2. This isso even if the probability (whether objective or otherwise) at Tt of E's occurring at T3 asa consequence of 5's doing A at T2 was less than 1. In saying this, I don't mean to denythat there is a sense of "would" according to which it would not be appropriate to say this(in that "might" should be used instead — see Vallentyne, 1987b); I mean only that thereis a sense of "would" according to which it is appropriate to say this, and that it is thissense that I'm employing here. This fact about how "would" is to be construed here alsoaffects how "could" is to be construed, for the two are intimately related. (See Subsections2.2.2 and 2.2.3.) Suppose that S can at T, do A at T2, that there is a probability of .1 atTj that Ej will occur at T3 as a consequence of 5's doing A at T2, and that there is a prob-ability of .9 at 7\ that E2 will occur at T3 as a consequence of 5's doing A at T2. Supposethat, despite these probabilities, E1 would (in my sense) occur and E2 would not occur atT3 if 5 did A at T2. Then Et could (in my sense) happen and E2 could not happen (unless5 could make E2 happen by some other means). For more on the relevance of probabil-ity to the present issue, see Section 6.4.

190

ought to decline the invitation (for the reason given above) are actualists.The distinction between possibilism and actualism was brought to light, asfar as I know, only recently. At first actualism was favored;4 then possibil-ism was touted;5 but since then actualism has been vigorously defended.6

In this chapter I shall investigate this dispute.

6.2 OBJECTIONS TO ACTUALISM

6.2.1 Certain theses denied

It might seem that possibilism is obviously preferable to actualism. If Rickdeclines the invitation, then he takes Track 2. Clearly this is not the best hecan do, since Track 1 is better, and so anyone who is working within thetradition that one ought to do one's deontic best (as the actualist avowed-ly is) must surely accept that Rick does wrong in declining the invita-tion. And yet the actualist says that it is not wrong for Rick to declinethe invitation, indeed that Rick ought to decline it. Isn't this plainlyunacceptable?

The actualist has a response here. He will point out that, according tohim, Rick does do wrong in not accepting the invitation. He doesn't dowrong with respect to not-accepting as such, but he does do wrong withrespect to not-(accepting-and-submitting). Why? Because, even though itis true that accepting would have been deontically inferior to its alterna-tive, nonetheless accepting-and-submitting would have been deonticallysuperior to any of its alternatives.

But this response comes at a considerable cost, for it highlights the factthat the actualist is committed to a denial of several very plausible theses.

First, the actualist must deny the following:

(2.40'd) if C(~£) & O(A&B), then O{B).

This is demonstrated, in the present case, by the fact that, according to actu-alism, Rick is obligated to accept-and-submit but is not obligated to accept,even though he can of course not-accept. Some actualists have explicitlyand boldly embraced this implication. For example, Frank Jackson andRobert Pargetter, probably the most prominent proponents of actualism,say this: "Perhaps an overweight Smith ought to stop smoking and eat less,

4 See Holly S. Goldman (1976) and Sobel (1976).5 See Goldman (1978), Greenspan (1978), Thomason (1981), Humberstone (1983), and

Feldman (1986), Ch. 2.6 See Jackson and Pargetter (1986) and Goble (1993).

191

but it may not be true that he ought to stop smoking. For it may be thatwere he to stop smoking he would compensate by eating more."7

Notice that (2.40'd) is implied by the following (where <} signifies log-ical possibility):

(6.3) if- O(A&~B) & C(~B), then, if O(A), then O(B).

(This can be seen by substituting A & B for A in (6.3), which yields:

(6.3') if- O((A&B) &~B) & C(~£), then, i£O(A&B), then O(B).

Clearly, - O((A&B) &~B) is necessarily true.) Thus actualism implies that(6.3), also, is false. And (6.3) is itself implied by the following:

(2.43"') i£~C(A& ~B) & C(~B), then, if O(A), then O(£).

Thus actualism implies that (2.43'") is false, too.The actualist must also deny this thesis:

(2.40'b) if O(A) & O(B), then O(A&B).

Let A be accepting-and-submitting and B be not-accepting. Rick cannotboth accept-and-submit and not-submit; hence, given

(2.46) ifO(^4),thenC(i4)

(a thesis to which the actualist is standardly committed), it is not the casethat Rick ought both to accept-and-submit and not-accept. Yet each,according to the actualist, is such that Rick ought to do it. Similarly, theactualist must also deny this:

(2.47) if O(A) & O(B), then C(A&B).

For, he says, Rick ought to accept-and-submit and ought also to not-accept; yet Rick cannot do both. Now, some would deny (2.47) anyway,claiming that moral dilemmas show it to be false. Although I reject this lineof argument (the issue will be discussed further in the next chapter), thepoint to be made here is that the actualist does not reject (2.47) because heaccepts the possibility of moral dilemmas. Suppose that what would promptRick not to submit the review on time, were he to accept the invitation todo so, is sheer laziness; he just "couldn't" (more accurately: wouldnX) bebothered. Even in this case the actualist would say that Rick ought to not-accept. Yet he would also say that Rick ought to accept-and-submit. Now

7 Jackson and Pargetter (1986), p. 247. Cf. also Tannjo (1985), pp. 116-17.

192

Rick cannot do both, but it would surely be a gross distortion of his situa-tion to portray it as a moral dilemma, for the actualist must admit that Rickcould do all that he ought (namely, accept-and-submit, the doing of whichwould imply that it is not the case that Rick ought to not-accept).8

6.2.2 Sanctioning wrongdoing

This last point brings out what is perhaps oddest and, in my view, mostobjectionable about the actualist's view: it sanctions wholly avoidablewrongdoing. Let me not be misunderstood here. I do not mean that actu-alism sanctions doing what the possibilist calls wrong (although this is trueon occasion). I mean that the actualist sanctions doing what the actualist callswrong. True, there is no flat-out inconsistency to the effect that in oneand the same breath the actualist says W(A) & P(A) or W(A) & O(A);10

nonetheless, the actualist is committed to claiming on occasion "P(A) (infact, O(A)), even though S cannot avoid doing wrong if he does A but canavoid doing wrong if he doesn't do A" The case of Rick is just such anoccasion, for the actualist says that Rick ought to not-accept, even thoughRick cannot avoid doing wrong if he does not-accept — in that, if he doesnot-accept, he cannot avoid not-(accepting-and-submitting), which iswrong by the actualist's own lights — but can avoid doing wrong if he doesn'tnot-accept (but rather accepts-and-submits). Why should Rick's obliga-tions be thought to be thus tailored to his own easily avoidable moral fail-ings? Of course, Rick's failure to submit the review on time might (in someother version of the case) be due not to laziness but rather to his havingtaken on a great many other tasks about which he is understandably con-cerned, and here one might be inclined to accept what the actualist saysabout Rick's obligations. This would be a mistake, though, for reasons thatI shall discuss in Section 6.4.

I have just claimed that no acceptable account of obligation sanctionswholly avoidable wrongdoing, where this is understood to consist in theconjunction of these propositions: S may, or ought, to do A; S cannot doA without doing 23; it is wrong for S to do B; and S can act in a way thatincludes his not doing A and his not doing B and his doing no wrong. Since

8 If Rick were to accept-and-submit, then it would not be true that, if he were to accept,then he would not submit. Cf. Jackson and Pargetter (1986), pp. 242 and 249.

9 Goble apparently misunderstands me (in Goble, 1993, p. 161, n. 7), claiming that thiscriticism (first presented in Zimmerman, 1990b) "begs the question, since the issuebetween possibilism and actualism is what constitutes wrongdoing."

10 I say "inconsistency" rather than "contradiction," for reasons discussed in Section 7.1.

193

actualism allows for such a situation to arise, actualism must be rejected. Buthere one might be tempted to invoke an argument of Derek Parfit's to theeffect that it can sometimes be right to cause oneself to do wrong. Parfitgives a case of the following sort.11 Patty is a politician, and Enid is herenemy. Enid tells Patty, truthfully, that she will kill all Patty's children ifand only if Patty refuses to be filmed performing some obscene act, the filmto be used later by Enid to persuade Patty to commit several minor crimes(in that, if Patty refuses to commit these crimes, the film will be released tothe press, thereby ruining Patty's career). On the assumption that Pattycannot save her children except by submitting to being filmed performingthe obscene act in question, Parfit claims that Patty ought to do this, eventhough she knows that this will cause her to act wrongly later (by commit-ting the crimes), since she knows that she will choose saving her career overforgoing the crimes. Hence not only would it be right for Patty to causeherself to do wrong, she in fact ought to cause herself to do wrong.

There are two main ways to construe this situation, and in neither casedo we have a sanctioning of wrongdoing of the sort that I am claiming tobe unacceptable. First, we can understand "cause" in such a way that acause's effects cannot be avoided, once the cause has occurred. In this case,the situation may be represented as in Chart 6.2 (where 5 stands for "sub-mit to being filmed" and C for "committing the later crimes"). Here thesubmitting renders the later commission of the crimes inevitable. On thisunderstanding of the situation, the later commission is not wrong; hence,contrary to Parfit's description of the situation, there is no wrongdoingcaused.

The second way to construe the situation - which is presumably Parfit'sway - is to understand "cause" in such a way that a cause's effects are avoid-able, even when the cause has already occurred (but the effects haven't). Inthis case, the situation may be represented as in Chart 6.3. Here, indeed,the submitting is obligatory but the commission wrong. But this does notconstitute a sanctioning of wrongdoing in the foregoing sense. Foralthough it is being said that Patty ought to submit and that this would cause

1: S & C ...2: ~S & ~C ...3: ~S & C ...

Chart 6.2

11 Parfit (1984), p. 38.

194

1: S & ~C ...2: S & C ...3: ~S & ~C ...4: ~S & C ...

Chart 6.3

her to commit the crimes, which would be wrong, the fact is that it is opento her to do what it is said she ought (namely, submit) while refraining fromdoing the wrong in question.

6.2.3 A nondiachronic case

Recall from Section 4.3 the case of Charlie, captured in Chart 4.4, wherethe actions in question, whichever track Charlie chooses, would be simul-taneous. Suppose that, whether or not Charlie accelerates, he's going tochange lanes. Then actualism implies that it is not the case that he ought tonot-accelerate, even though it is the case that he ought to not-change-and-not-accelerate. (Possibilism, of course, implies that he ought to not-accel-erate.) Thus, as Jackson and Pargetter acknowledge,12 the diachronic natureof the reviewing case is not essential to the present discussion. Indeed, thecase of Charlie brings out equally well the fact that actualism implies thefalsity of the theses just discussed.

6.2.4 A slightly amended version of actualism

Jackson and Pargetter subscribe to a version of actualism slightly but signi-ficantly different from that with which we have been working so far. It isthis:13

(6.1') 5 ought to do A iff what would happen if S did A is deonticallysuperior to what would happen if S did not do A.

This version is no more successful than the first. Like (6.1), (6.1') alsoimplies that the theses just discussed are false, for it yields the same pres-criptions as (6.1) for both Rick and Charlie. And like (6.1), (6.1') also sanc-tions wholly avoidable wrongdoing. Indeed, (6.1') seems to me even moreproblematic than (6.1), in that the former yields some counterintuitive pres-criptions that the latter does not. For consider another case, that which was

12 Jackson and Pargetter (1986), p. 236.13 Jackson and Pargetter (1986), p. 247.

195

first presented in Section 4.1 and captured in Chart 4.1, and with respectto which we may suppose the following to be true: if you were to vote forTom, that would be better than if you were to perform any alternativeaction (that is, better than if you were to vote for Dick, or Harry, or noone). It follows that it is not the case that, if you were to vote for Dick, orHarry, or no one, that would be better than if you were to perform anyalternative action (for it wouldn't be better than voting for Tom). But nownote this. It is consistent with the foregoing to suppose not only that, if youwere to vote for Tom, that would be better than if you did not vote for Tom,but also that, if you were to vote for Dick, that would be better than if youdid not vote for Dick; for we may suppose that, if you didn't vote for Dick,you'd vote for Harry. Hence, while both (6.1) and (6.1') imply that youought to vote for Tom, the latter (but not the former) also implies that youought to vote for Dick. This implication is surely to be rejected.

6.3 OBJECTIONS TO POSSIBILISM

Just how successful the foregoing criticisms of actualism are is debatable. Iargue: ifp then q (if actualism is true, then such-and-such a thesis is false);not-q (the thesis is true);14 hence not-p (actualism is false). Adherents toactualism may well respond: if p then q; p; hence q. This is not an uncom-mon situation in philosophy, and in this case, as in others, an indisputableresolution of it is probably not forthcoming. But there is room for somemaneuvering. For instance, the actualist might try to bolster his view bypointing out that, although it implies the falsity of the theses just discussed,it is nonetheless consistent with certain closely related theses. (For exam-ple, Jackson and Pargetter, although they reject

(2.40'd) if C(~B) & O(A&B), then O(£),

in effect endorse this thesis:

(6.4) if C(~B) & O(A&B), and if S does (A&B), then O(£).15

Similarly, Lou Goble, although he rejects

(2.43'") if ~C(A&~B) & C(~£), then, if O(A), then O(B),

14 My account implies the truth of each of (2.40'd), (6.3), (2.43'"), (2.40'b), and (2.47). (SeeSubsection 2.3.1 for a discussion of such implications.) It also implies that wrongdoingis not to be sanctioned.

15 Jackson and Pargetter (1986), p. 248.

196

in effect endorses this thesis:

(6.5) if ~C(A&~B) & C(~B), then, if O(y4), then O(A&B))X6

While this might go some way toward mitigating the impact of the rejec-tion of the original theses, it is surely a tack one is tempted to take only ifsome good reason can be given for rejecting possibilism, which implies thetruth of the original theses without qualification. Here, though, the actu-alist thinks that he can provide such reason; at least, Jackson and Pargetterthink that they can. I turn now to a consideration of certain arguments thatthey have raised against possibilism.

6.3.1 Arbitrariness

The first objection that Jackson and Pargetter raise against possibilism is this.They claim that possibilism is "arbitrary" in asserting that Rick ought toaccept the invitation, for although it is true that accepting is part of thedeontically best course of action open to him, it is also true that it is part ofthe deontically worst course of action open to him. Surely, then, there mustbe more to say before the question of what Rick ought to do is settled. Inthemselves the fact that accepting is part of the best and the fact that it is partof the worst show nothing. The first points to the answer that Rick oughtto accept, but the second points equally to the answer that he ought not toaccept. Possibilism errs in allowing the first fact to settle the matter.17

This is a puzzling objection. It rests on the assumption that acceptance'sbeing part both of what's deontically best and of what's deontically worstpoints equally to its being obligatory and to its being wrong. No reason hasbeen given for this claim, and the possibilist of course rejects it. For whilethe possibilist accepts the claim that, where a conjunction of acts (A&B) isobligatory, then so is each conjunct (see (2.40'd) for a more precise rendi-tion of this), he of course rejects the claim that, where a conjunction of acts(B&C) is wrong, then so is each conjunct. There is no need, therefore, tothink that an act, B, cannot feature in each conjunction.

Consider an analogy. Suppose that a state of affairs, S, is part both of theactual world, W\, and of some nonactual world, W2. Despite S's being part

16 Goble (1993), p. 144.17 Jackson and Pargetter (1986), p. 237. Goble seems to have something similar in mind

when he says (in Goble, 1993, p. 137): "One problem with possibilism.. .is that when [itsays] that one ought to do the best one can..., [it disregards] the fact that the best one cando may also be the very worst one can do... When that happens, the possibilist mustadvocate that one ought to do the worst one can do."

197

of IV2, we have no hesitation in saying that S obtains. There is nothing per-suasive in the following objection: "It is arbitrary to claim that S obtains,for although it is true that S is part of Wu it is also true that it is part of W2.Surely, then, there must be more to say before the question of whether 5obtains is settled. In themselves the fact that S is part of W\ and the fact thatit is part of W2 show nothing. The first points to the answer that S obtains,but the second points equally to the answer that it does not obtain."18

6.3.2 Advice

The second objection is this. Rick might turn to Addie for advice.Knowing Rick's propensity for procrastination, Addie might well adviseRick to decline the invitation. Is she advising him to act immorally? Is shehoping for this? Clearly not. Her advice is the right advice, and this impliesthat what she advises him to do is right.19

This objection cannot be accepted. It would prove too much, for itcould just as easily be applied to the voting case. Knowing your propensi-ty for overlooking the best-qualified candidate, I might well advise you tovote for Dick (rather than Harry). This advice would be just as appropriatehere as Addie's advice to Rick to decline the invitation. But, we all agree,it is not the case that you ought to vote for Dick.

So where has the objection gone wrong? Presumably in its implicitreliance on these two theses:

(6.6) if S ought to advise S' to do A, then S' ought to do A;(6.7) if S advises S' to do A, and S' ought not to do A, then 5 advises S'

to act immorally.

With respect to the first thesis, all that needs to be said is this. Suppose that(and that Addie is aware that) Rick will follow Addie's advice as to whetheror not to accept the invitation, but that nothing she can do would persuadehim to submit the review on time. Thus, although Rick can of course so actthat Rick submits the review on time, Addie cannot so act that Rick sub-mits the review on time, and so the best that she can do does not match upwith the best that he can do. (Rick may be said to be intransigent relative to

18 Note, further, that the actualist, too, must acknowledge that, on occasion, "one oughtto do the worst one can do" (to use Goble's phrase - see the last note). After all, if Rickwould submit the review on time if he were to accept the invitation, then the actualistwould agree that he ought to accept, even though accepting is also (part of) the worst hecan do.

19 Jackson and Pargetter (1986), p. 237. Cf. Goble (1993), p. 139.

198

Addie concerning his submitting the review on time. This is an issue to betaken up in some detail in Chapter 9.) Under the circumstances, then,Addie has a choice between advising Rick to accept the invitation (andthereby seeing to it that he does accept the invitation but does not submitthe review on time) and advising Rick to decline the invitation (and there-by seeing to it that he does not accept the invitation and so, again, does notsubmit the review on time). We may assume that the latter course of actionis deontically better. Here possibilism implies that Addie ought to adviseRick to decline, while still implying that Rick ought to accept.20

As for the second thesis: there is a sense in which it is true and a sense inwhich it is false. The sense in which it is true is this:

(6.7') if 5 advises S ' to do A, and S' ought not to do A, then there is someaction such that:(a) S advises that S ' perform it; and(b) it is wrong.

The sense in which the second thesis is false is this:

(6.7") if S advises S' to do A, and S' ought not to do A, then S advises thatS' perform a wrong action,

where this is understood to imply that the wrongness of the action formspart of the content of the advice and thus part of the objective of the advi-sor. After all, it is not Addie's objective (or hope), when advising Rick todecline the invitation, that Rick should act immorally; that is somethingwhich she correctly takes as given, that is, as beyond her remedy (thoughnot his). What she is aiming at is the minimization of /mmorality on his part,consistent with the restrictions on her influence under which she is operating.

I have said that the possibilist rejects (6.6), but the actualist must rejectit too. For there is no necessity that what would happen if S' did A be deon-tically better than what would happen if S' acted otherwise, even thoughwhat would happen if S advised S' to do A would be deontically better thanwhat would happen if 5 acted otherwise. Suppose that Rick always actscontrary to Addie's advice and that Addie can do nothing about this. Then,on the actualist's view, Addie ought not to advise Rick to decline the invi-tation, even though Rick ought to decline it. Similarly, both the possibilistand the actualist must reject the converse thesis, namely:

(6.8) if 5 ' ought to do A, then 5 ought to advise S' to do A

20 Cf. Feldman (1986), pp. 55-7.

199

(where it is understood that 5 is in a position to give such advice). For boththe possibilist and the actualist agree that Rick ought to accept-and-sub-mit, but both must acknowledge that it might happen that Addie ought notto advise Rick so to act, in that such advice could easily backfire and thusbe deontically undesirable.

There is therefore no support for actualism to be found in considera-tions concerning advice.

6.3.3 Detachment

The third objection that Jackson and Pargetter raise is this. They claim thatfactual detachment for conditional obligation is valid; that is, that theunconditional obligation to perform an action may be inferred from theconditional obligation to perform it when the condition in question is sat-isfied. Thus, given the fact that Rick ought to decline the invitation if hewon't submit the review on time and the fact that he won't submit thereview on time, it follows that he ought to decline it (period).21

In response, we need only note that, as remarked in Chapter 4, factualdetachment for conditional obligation is not valid. Consider the voting caseonce again, as represented in Chart 4.1. You ought to vote for Dick if youdon't vote for Tom. Suppose that you won't vote for Tom. It doesn't fol-low that you ought to vote for Dick. On the contrary, you still ought tovote for Tom. Now, elsewhere, Jackson and Pargetter recognize this verypoint.22 But then they try to save their original thesis by drawing a distinc-tion between ways in which obligation may be subject to conditions. Firstthere is what they call "hypothetical obligation." A statement of hypothet-ical obligation has the form, "If it were the case thatp, then S would be(overall) obligated to do A " That is, what Jackson and Pargetter call hypo-thetical obligation is obligation that involves a subjunctive conditional andcan be symbolized (as in Section 4.1) in this way:

(4.8) p>O(A).

As Jackson and Pargetter note, and as was noted in Section 4.1, factualdetachment is valid for statements of hypothetical obligation. But thenthere is what Jackson and Pargetter call "restrictive obligation."23 A state-ment of restrictive obligation has the form, "Given that p, S is (overall)

21 Jackson and Pargetter (1986), p. 238.22 Jackson and Pargetter (1987), pp. 77-8. They also hint at it in note 20 on p. 253 of

Jackson and Pargetter (1986).23 Jackson and Pargetter (1987), p. 77.

200

obligated to do A." By this they mean to give expression to what I havecalled "conditional obligation proper." As Jackson and Pargetter note, andas I have noted repeatedly, factual detachment is not valid for statements ofsuch obligation.

Making this perfectly correct distinction helps Jackson and Pargetter,though, only if it can be shown that Rick's obligation to decline the invi-tation, if he won't submit the review on time, is a hypothetical rather thana restrictive one. This is of course what they contend, but their rationale isproblematic, for three reasons.

Consider, first, the voting case. There is an inconsistency in Jackson'sand Pargetter's position with respect to this case. They declare the obliga-tion that you ought to vote for Dick, if you don't vote for Tom, to be arestrictive one. This seems absolutely right. But then they say that, since(factual) detachment for restrictive obligation is not valid, it is not the casethat you ought to vote for Dick. Yet we have seen that, on their version ofactualism (captured in (6.1')), you ought to vote for Dick (since your doingso would be better than your not doing so).

Second, Jackson's and Pargetter's position overlooks the apparent iso-morphism between the voting case, where it is granted by all concernedthat restrictive obligation is at issue, and the reviewing case. The mark ofrestrictive obligation, as Jackson and Pargetter rightly note, is that it con-cerns obligation relative to a restricted set of options. You ought to votefor Tom, given all your options; but, with the option of voting for Tomdiscounted, you ought to vote for Dick. Similarly, Rick ought to acceptthe invitation, given all his options; but, with the option of submitting thereview on time discounted, he ought to decline the invitation. Or again:you ought to vote for Tom; if you won't do this, then you ought at least tovote for Dick. Similarly, Rick ought to submit the review on time; if hewon't do this, then he ought at least to (have the courtesy to) decline theinvitation. Jackson and Pargetter in fact admit that the propriety of the locu-tion "at least" signals restrictive obligation. 5 Since it is proper to use it inthe reviewing case, again we find a tension in their position.

Third, their position is undermined by their own characterization ofhypothetical obligation. They say that what follows "if," in cases of hypo-thetical obligation, involves not a restriction of options but a reordering ofthem. This again seems quite right. Recall this example:

(4.10) if Tom is (or were) a crook, then you ought to vote for Dick.

24 Jackson and Pargetter (1987), p. 78.25 Jackson and Pargetter (1986), p. 253, n. 20.

201

1:2:3:

~AAA

&&&

~S~S

S

Chart 6.4

As noted in Section 4.1, this is tantamount to saying that, if Tom were acrook, then your options would not be ranked as in Chart 4.1 but in someother way, such as in Chart 4.2. But notice how different this is from thereviewing case. In saying that Rick ought to decline the invitation if hewon't submit the review on time, we are surely not saying that, since it istrue that he won't submit the review on time, his options are to bereordered in some such way as in Chart 6.4.

I think, then, that we simply must accept that Rick's obligation todecline the invitation, if he won't submit the review on time, is a restric-tive one — that is, a conditional obligation proper. Since factual detach-ment is acknowledged by all concerned to be invalid for such obligation,we cannot argue that it is such detachment that warrants the claim thatthe actualist wants to make, namely, that Rick ought to decline theinvitation.

Recall, however, that in saying that factual detachment is not valid forconditional obligation proper, I am saying that we cannot derive a primaryunconditional obligation in this way. But we can, as noted in Section 4.4,derive a subsidiary unconditional obligation in this way. If you don't votefor Tom, you ought (at least) to vote for Dick; if you don't even do that,you compound your wrongdoing, in that you violate a secondary uncondi-tional obligation to vote for Dick. Similarly, if Rick won't submit thereview on time, then he ought to decline the invitation; if he doesn't evendo that, he compounds his wrongdoing by violating a secondary uncondi-tional obligation to decline the invitation. Jackson and Pargetter claim thatthere is a "positive advantage" to actualism.26 It can account for the fact that,intuitively, if Rick declines the invitation, he meets some but not all of hisobligations. For, they say, he ought to decline (even though he ought alsoto accept-and-submit). But my possibilist account also allows me to say thatRick meets some but not all of his obligations in declining the invitation.He meets the secondary obligation to decline; he fails to meet the primaryobligation to accept.27 Thus, contrary to Jackson's and Pargetter's claim,actualism enjoys no advantage here.

26 Jackson and Pargetter (1986), p. 241.27 This point is applicable to what Goble says in Goble (1993), p. 153.

202

6.3.4 Act-individuation

The final objection that Jackson and Pargetter raise against possibilism isthis. It is clear that Rick ought not to delay the reviewing of the book; but,by hypothesis, this is just what he would be doing if he accepted the invi-tation to review it, for then the invitation would not be extended to any-one else. That is, under the circumstances, his accepting the invitationwould be his delaying the review. Hence he ought not to accept.28

This objection rests on two assumptions. The first is that acts are to beindividuated coarsely and thus that Rick's accepting the invitation wouldbe identical with his delaying the review. The second is this:

(6.9) if S ought to do A, and his doing A is identical with his doing B,then S ought to do B.

These assumptions are jointly unacceptable. Either Rick's accepting theinvitation is not identical with his delaying the review but is related to it insome less intimate way (expressed by the term "by" in "Rick delays thereview by accepting the invitation"), or (6.9) is false, since, even if Rick'saccepting the invitation is his delaying the review, this identity holds onlycontingently, and he ought not to allow it to hold. That is, the possibilist, work-ing under the assumption that acts are to be individuated coarsely, wouldaccept that

(6.10) if S ought to do A, and he ought to allow his doing A also to be hisdoing B, then S ought to do B,

but he would have no reason to accept (6.9). (Recall here the discussion inSubsection 2.2.5 concerning Jones, who helped the old lady, as he oughtto have done, but who in doing so startled her, as he ought not to havedone. Someone who believes that Jones's helping the old lady and his star-tling her are identical must reject (6.9) and resort to some such device assaying that one and the same act can be obligatory under one descriptionbut not under another.)

6.4 WHAT IT IS BEST TO DO

While it seems to me that the foregoing should suffice to put actualism torest, I grant that this thesis may still seem appealing. Even if it is conceded,as surely it should be, that Rick ought to accept the invitation when it isonly laziness that would lead him to fail to submit the review on time, what

28 Jackson and Pargetter (1986), p. 239.

203

if instead he would fail to submit the review on time because he has takenon a great many other tasks or for some other reason would find it verydifficult to do so, even though, strictly, he could do so? Isn't it plausible tosay what the actualist says in this case and declare that Rick is not obligatedto accept the invitation?

Consider the medicine case first presented in Section 4.2. There are twomedicines, A and B, that Audrey can administer. Let us suppose that, if sheadministers A on Monday and A on Tuesday, then the patient will be fullycured; if she administers B on Monday and B on Tuesday, the patient willbe pretty well off, but not as well off as on the first course of action. Anyother course of action would be disastrous for the patient. It seems that pos-sibilism requires Audrey to administer A on Monday. But what if we nowsay that, if Audrey were to administer A on Monday, she would not admin-ister it on Tuesday, and this not because of laziness but for some other whol-ly understandable reason? For instance, the second dose of medicine A maybe very hard for Audrey to lay her hands on, although strictly she could doso (it's in a safe, and she doesn't know the combination).29 Or perhaps itwouldn't be difficult for Audrey to do this, but it's rather that she doesn'tknow that two doses of A will cure the patient; on the contrary, she hasvery good reason to believe that a second dose of A would be fatal, althoughshe knows that with two doses of B the patient will be pretty well off. Underthese circumstances, don't we want to agree with the actualist that Audreyought not to administer A on Monday but should give the patient Binstead?

Or consider, finally, this case.30 Gary is playing a game. If he wins, greatgood fortune will abound; if he loses, it will be disastrous; if the game endsin a draw, there will be no significant consequences. He has in his hand atoken, which he can place in one of two boxes. The game consists simplyin the following. If he places the token in the first box, then he either winsor loses; if he places it in the second box, it's a draw. What determineswhether he wins or loses is whether he places the token in the first box withhis left hand or with his right hand. Although Gary is aware that this is so,he has no idea which hand is correlated with great good fortune, whichwith disaster. Ought Gary to place the token in the first box? Possibilismappears to imply that he ought. But don't we want to say, with the actual-ist, that whether or not Gary ought to do this depends on which hand hewould use and what would happen as a result?

29 See Goble (1993), pp. 140-1. I accept that Audrey can, strictly speaking, lay her handson the medicine. See the discussion in Subsection 2.2.3.

30 I borrow this illustration (in modified form) from Goble (1993), p. 145.

204

It is not clear to me that it is correct to say that Rick ought to declinethe invitation (when he would fail to submit the review on time due to hishaving so many other tasks to attend to), that Audrey ought to administermedicine B on Monday (when she would have great difficulty in laying herhands on a second dose of A on Tuesday, or when she would fail to admin-ister A on Tuesday in the belief that it would be disastrous to do so), or thatGary ought not to place the token in the first box (if he would do so withhis right hand, say, and then disaster would ensue). After all, there is in prin-ciple a distinction between what it would be wrong for someone to do andwhat he would be blameworthy for doing, and while I readily grant thatthe agents in these examples would not be blameworthy for failing to acceptthe invitation, to administer A on Monday, and to place the token in thefirst box (with the left hand), respectively, it does not follow that theywould do no wrong in failing to do these things.31 Still, it is also not clearto me that it is incorrect to say that the agents ought to decline, administerB, and put the token in the second box, respectively. But does this pose anyproblem for possibilism?

I cannot see that it does. It was noted in Section 1.4 that the presenttheory is not committed to any particular account of what deontic valueconsists in. It is perfectly consistent with the present theory to say that, inthe relevant sense of "best," it would be best for Rick to decline, best forAudrey to administer B on Monday, and best for Gary to place the tokenin the second box. This would be because the relevant sense of "best"(namely, that of "deontically best") is the same as that of "probably best"or "most reasonably thought best" or something of the sort (where "best"in these latter two phrases is understood in some other, substantive sense).Thus the possibilist can accept the actualist's diagnosis of what the agentought to do in these cases, while of course rejecting the reason for thisdiagnosis that the actualist himself gives.

But here the actualist may well protest.32 His point is not just that Rickought to decline the invitation, that Audrey ought to administer medicineB on Monday, and that Gary ought to put the token in the second box(given that he'd use his right hand if he put it in the first box). Rather,it's also true that Rick ought to accept-and-submit, Audrey ought toadminister-A-on-Monday-and-administer-A-on-Tuesday, and Gary oughtto put-the-token-in-the-first-box-with-his-left-hand. Thus we do havean objection to possibilism after all, in that in each case we have a

31 Indeed, it might well be that the agent would be blameworthy for accepting the invitation,administering A on Monday, and so on. See Zimmerman (1988a), Ch. 3.

32 Goble would. See Goble (1993), p. 141.

205

counterexample to

(2.43'") if ~C(A&~B) & C(~B), then, i£O(A), then O(B).

But here I demur. If we grant that Rick ought to decline, why then alsogrant that he ought to accept-and-submit? (Of course, actualism may wellyield this judgment, but actualism is unacceptable. It implies that Rickought to decline even when he would fail to submit the review on time outof sheer laziness.) The reason why we are tempted to say that Rick oughtto decline, where his failure to submit the review on time would be due tohis having so many other tasks to attend to, is, I suggest, because of the greatdifficulty he would have in submitting the review on time. But if this is so,then there is equally good reason to say that it is not the case that he oughtto accept-and-submit. (This would be a matter of supererogation, an issueto be discussed further in Chapter 8.) Or again, the great risk involved inAudrey's case may lead us to say that it is not the case that she ought toadminister A on Monday; but this should lead us also to say that it is not thecase that she ought to administer-^4-on-Monday-and-administer-^4-on-Tuesday. So too with Gary's putting-the-token-in-the-first-box-with-his-left-hand. The risk, we may say, is too great — either (as in Audrey'scase) with respect to whether the complex action will be performed, or(as in Gary's case) with respect to what will be achieved by means of thecomplex action — and for that reason the complex action in question is notobligatory.33

In sum, actualism has unacceptable implications, whereas possibilismdoes not. It is clear which is preferable.

33 Goble's reason for saying that Rick ought to accept-and-submit, that Audrey ought toadminister-^4-on-Monday-and-administer-y4-on-Tuesday, and so on, is that it would be"appropriate to regret" the agents' not performing these complex actions, although not"appropriate to regret" their not performing the relevant component action. (See Goble,1993, p. 141.) Perhaps we can agree that regret would be appropriate, or not, as Gobleindicates, but this is not determinative of what is obligatory, or not. Regret is appropri-ate in many cases where no obligation is violated (I can quite appropriately regret some-one's having died in an earthquake); and (overall) regret may be inappropriate evenwhere an obligation is violated, if things luckily turn out for the better (in some sense)because of this violation.

206

7

Dilemmas

7.1 THE NATURE OF MORAL DILEMMAS

The aloofness of the metaethicist may seem particularly objectionablewhen the subject is moral dilemmas. A dispassionate inquiry into moraldilemmas may appear especially inappropriate — their very nature demandsa passionate response — and those metaethicists who deny that such dilem-mas can even arise may seem to have reserved for themselves a unique placeamong out-of-touch academics.

To which I reply (as politely as possible): the criticism is misdirected. Noone is denying that life can present people with choices that are excruciat-ing, both morally and personally. No one is denying that anyone who isconfronted with such a choice, whether as participant or simply as specta-tor, should be deeply moved by it. The question is: just what is at stake insuch choices? What is the nature of the situation in which such choicesarise? To shun such inquiry is to court confusion, and that is what isobjectionable.

Moral dilemmas, as typically construed, are conflicts of moral obliga-tion, but not all conflicts of obligation are dilemmas. A conflict of obliga-tion involves there being two or more obligations which cannot all besatisfied, although each can be. In earlier chapters I have distinguished twomain species of moral obligation- overall obligation and prima facie obliga-tion — and from this emerge three main types of conflict: overall versusoverall, overall versus prima facie, and prima facie versus prima facie. Letus consider the last type of conflict first.

Clearly, conflicts between prima facie obligations are possible. I beganChapter 5 with a story about one such conflict: Steve had a prima facieobligation to render first aid to Carl and also had a prima facie obligationto keep his dinner date with Dave, but he couldn't satisfy both obligations.Such conflicts are possible because the obligations are what Ross calls

207

"parti-resultant" rather than what he calls "toti-resultant":1 beneficencerequired rendering the aid, but it was fidelity that required keeping the date.The obligations stemmed from different considerations, and such consid-erations may conflict. Even when conflicts between prima facie obligationsarise within one and the same "category" of obligation, the considerationsthat conflict are distinct from one another. While gratitude required boththat Gertrude send Lucy lilies and that she send Chrissie chrysanthemums,it was, more particularly, gratitude-to-Lucy that required the first andgratitude-to-Chrissie that required the second.

There is nothing particularly mysterious about such conflicts of obliga-tion, nor about conflicts between prima facie and overall obligations. (Afterall, typically, and perhaps always,2 the most stringent prima facie obligationwill be an overall obligation.) Clearly not all such conflicts merit the label"dilemma," however. This label should surely be reserved for occasionswhen the choice between conflicting obligations is a hard one; and not allconflicts of prima facie obligations (whether or not some of these are alsooverall obligations) need involve hard choices.

Still, some clearly can. There are two principal ways in which a choicebetween conflicting prima facie obligations might be hard. It might be hardto tell which prima facie obligation is the most stringent, since more thanone might appear equally compelling. Or it might be hard to make thechoice for that prima facie obligation which is readily recognized as moststringent, since the moral cost involved in failing to satisfy some opposingprima facie obligation (or obligations) is so great.3 In both types of case, thechoice may be an especially wrenching one, and I have no quarrel withcalling such situations dilemmas.

Nonetheless, it is not such situations that are the focus of inquiry in thischapter. Rather, I shall be looking into cases of conflict between overallobligations. Such cases, too, appear to warrant the label "dilemma," sincethey imply that overall wrongdoing cannot be avoided. I shall hence-forth reserve the label "dilemma" for cases where overall wrongdoingis inevitable, and it is only into dilemmas in this sense that I shall beinquiring.

It is useful to distinguish between dilemmas and quandaries. A quandary(in the present sense)4 is a situation in which the agent does not know whatit is overall right or obligatory for him to do. Clearly this can happen

1 Ross (1930), p. 28.2 See the discussion of (5.25) and (5.27) in Subsection 5.4.3.3 See Subsection 5.4.4.4 Following Conee (1989), p. 133.

208

(despite what some philosophers say).5 Such a situation presents the agentwith a hard choice, but it doesn't involve inevitable wrongdoing; for theagent may luckily make the right choice, despite his ignorance. Similarly,dilemmas (in the present sense) don't arise simply in virtue of there beinga tie for what it is deontically best for the agent to do. Clearly, some casesof such ties can be wrenching (Ruth Barcan Marcus describes what seemsto be such a case, where there are identical twins whose lives are in jeop-ardy and only one can be saved),6 but they don't imply inevitable wrong-doing. On the contrary, in cases of ties for what is deontically best, as thepresent account of overall obligation implies (see (II) in Subsection 2.1.3),the agent may, but is not obligated to, opt for any of the top-ranked worlds;he would do wrong only if he opted for none of them. This point bearsrepeating: such cases do not constitute dilemmas, in the present sense. Thisbears repeating because at least one major recent work on dilemmas definesdilemmas in just this way. Walter Sinnott-Armstrong says that a dilemmaarises when there is a conflict of two or more requirements (for all intentsand purposes, a conflict of prima facie obligations), none of which is over-ridden.8 If there is an unoverridden prima facie obligation to do some actA, then A is, as (5.18) implies, featured in some world accessible to theagent to which there is none that is deontically superior. But this does notimply that A is overall obligatory, merely that it is overall permissible; henceit does not imply that refraining from doing A is overall wrong. Sinnott-Armstrong grants this.10 Hence what he calls a moral dilemma is not a moraldilemma in the present sense.

I noted in Subsection 2.2.6 that the present account of overall obliga-tion rules in the possibility of certain dilemmas that I there called prohibitiondilemmas, as long as it is possible that each world accessible to the agent besuch that some other accessible world is deontically superior to it. In sucha situation the agent cannot avoid doing something wrong. But I there dis-tinguished prohibition dilemmas from obligation dilemmas, which consist inconflicts of overall obligation. Like prohibition dilemmas, obligation

5 See note 55 to Ch. 2.6 Marcus (1987), p. 192.7 Sinnott-Armstrong distances himself somewhat from what Ross says about prima facie

obligations, but nothing he says about requirements is incompatible with 'what I havesaid in Chapter 5 about prima facie obligation. See Sinnott-Armstrong (1988),pp. 97-102.

8 Sinnott-Armstrong (1988), pp. 19-20, 29.9 As long as nothing is of deontic value unless it gives rise to a prima facie obligation; see

again the discussion of (5.25) and (5.27) in Subsection 5.4.3.10 Sinnott-Armstrong (1988), p. 20.

209

dilemmas involve the inevitability of overall wrongdoing, but the two dif-fer in that, whereas obligation dilemmas arise from a surfeit of obligations,prohibition dilemmas arise from a dearth of obligations. Prohibition dilem-mas involve inevitable overall wrongdoing because they involve the pos-sibility that it would be overall wrong for 5 to do A even though there isno act B that S ought overall to do instead; obligation dilemmas involveinevitable overall wrongdoing because they involve the possibility that itwould be overall wrong for S to do A even though 5 ought overall to doA. As I indicated in Subsection 2.2.6, the present account of overall oblig-ation rules out the possibility of obligation dilemmas, barring certain com-plications to be addressed in Section 7.4. It is this issue that is the focus ofthis chapter. Henceforth, when I write of "dilemmas" without qualifica-tion, I shall mean obligation dilemmas exclusively.

What renders (obligation) dilemmas especially problematic is not justthat they involve inevitable overall wrongdoing (after all, prohibitiondilemmas do too), but that they involve a certain moral inconsistency. Theinconsistency consists in the implication that one and the same act can beboth overall obligatory and overall wrong. This is an implication, at least,if the following two propositions are granted:

(2.46) ifO(,4), thenC(y4);(2.43'") if ~C(A&~B) & C(~£), then, if O{A), then O(B).

Suppose we have a dilemma, due to the fact that

(7.1) O(X) & O(Y) &~C(X&Y),

for some acts X and Y. Then, given (2.46), we may infer (substituting Yfor A)

(7.2) C(Y),

and hence also that

(7.3) ~C(X&Y) & C(Y) & O(X).

Then, given (2 A3'"), we may infer (substituting Xfor A and Yfor ~JB, andso too ~Y for B)

(7.4) O(~Y),

which is, as I have said, inconsistent (in a way) with the second conjunct of(7.1), namely,

(7.5) O(Y).

210

Of course, (7.4) and (7.5) do not flat-out contradict one another, as do (7.5)and

(7.6) ~O(Y).

Nonetheless, there is clearly a troubling tension of sorts between (7.4) and(7.5) to which anyone who wants to admit the possibility of moral dilem-mas must attend. I shall argue that such tension is in fact not acceptable andhence that moral dilemmas cannot arise.

More precisely, I shall argue that basic moral dilemmas are impossible,where these are understood to consist in the following:

(7.7) (a) S ought overall at Tt to do A at T3;(b) S ought overall at Tt to do B at T4; and(c) S cannot at Tt do both A at T3 and B at T4\

where Tt is not later than either T3 or T4, but where the relation betweenT3 and T4 is left open. ((7.7) of course concerns dilemmas that involve onlytwo actions that cannot both be performed, although basic dilemmas canbe envisaged that involve an indefinitely larger number of actions that can-not all be performed.) It is basic dilemmas that appear to have been at thecenter of much recent debate on moral dilemmas.

7.2 THE CASE AGAINST BASIC MORAL DILEMMAS

One argument against the possibility of basic moral dilemmas is simply this:my account of overall obligation implies that they are impossible. That thisis so is easily seen. If clauses (a) and (b) of (7.7) are both true, then the deon-tically best worlds accessible to S at Tt are, on my account, worlds in whichSboth does A at T3 and does B atT4. But this implies that clause (c) of (7.7)is false. Of course, this might simply be taken as reason to reject my account;but I don't think it should be, for there are independent reasons to rejectthe possibility of basic moral dilemmas.

One reason is this: our own moral experience indicates that we do nottake ourselves to be, or to have been, in dilemmas in situations where manypeople tend to diagnose a dilemma. I have in mind such cases as those thatproponents of dilemmas have (or at least appear to have - more on this inthe next section) themselves diagnosed as dilemmatic. A case presented byJean-Paul Sartre is probably the most celebrated of all.11 It is the case of astudent in France, during the Second World War, who must choose

11 Sartre (1973), pp. 35-7.

211

between joining the Free French Forces and taking care of his mother.There seem to him to be equally good reasons to say of each option that heought overall to choose it, and yet he cannot choose both. Now certainlysuch a situation can arise, for, as so far described, it amounts to "no more"than a particularly excruciating quandary. What would you and I do in sucha case? Speaking for both of us, I am sure that we would seek advice as towhat to do, if we had someone to whom to turn for such advice. Indeed,this is just what the student does; he consults Sartre (an unwise choice,under the circumstances). But this indicates that we, and the student,believe that we are not (or, at least, may well not be) in a genuine dilemmaafter all; for if the advice we seek is moral advice (that is, advice as to whatwe morally ought overall to do), then we believe that there is (or, at least,may well be) a solution to our predicament which our advisor can pointout to us; and this indicates that we believe that at least one of the optionsthat we're faced with is not (or, at least, may well not be) overall obligato-ry after all. In similar fashion, those of us who have been faced with suchchoices frequently look back at the situation in question and wonderwhether we made the right choice; such wondering again indicates that webelieve that we were not (or, at least, may well not have been) in a genuinedilemma after all. Of course, one reply to this is simply that our beliefs insuch cases are mistaken; another reply is that, while our beliefs in such casesare not mistaken, there are other cases where we do not, or would not, holdsuch beliefs. Perhaps so, but the present considerations surely indicate thatbasic dilemmas aren't as common or as widespread as many of their propo-nents appear to maintain.

There are other, more forceful reasons for rejecting the possibility ofbasic moral dilemmas. One argument concerns these two theses, firstpresented in Subsection 2.3.1:

(2.40'b) if O(A) & O(B), then O(A&B);(2.46)

The first of these theses, often called the "principle of agglomeration,"yields the following proposition (when times are explicitly stated):

(7.8) if (a) S ought overall at T, to do A at T3, and(b) S ought overall at T? to do B at T4,

then (c) 5 ought overall at Tt to do both A at T3 and B at T4.

The second thesis yields the following proposition (when times are expli-citly stated):

212

(7.9) if 5 ought overall at T, to do both A at T3 and B at T4, then 5 canat T1 do both A at T3 and B at T4.

Jointly, (7.8) and (7.9) contradict (7.7), in that they imply that the con-junction of clauses (a) and (b) of (7.7) is inconsistent with clause (c) of (7.7).To preserve the possibility of dilemmas, then, one must reject either theprinciple of agglomeration or the principle that "ought (overall)" implies"can." I have defended the latter of these in Section 3.1, and I have no moreto say on the matter. But the principle of agglomeration deserves furtherconsideration.

Some proponents of dilemmas have explicitly rejected the principle ofagglomeration. In every case that I know of, the point seems to be this: itmay be reasonable for S to do A and reasonable for S to do B, but it doesn'tfollow that it is reasonable for 5 to do both; and this applies to moral andnonmoral reasons alike.12 For example, it may be reasonable (morally ornonmorally) for Alex to accept Elvis's invitation to dinner and reasonablefor her to accept Irving's, but not reasonable for her to accept both. But wemust ask what "reasonable" means here. There are two principal senses: (a)what reason requires, and (b) what reason permits;13 and there are two otherprincipal senses: (1) what is overall reasonable, and (2) what is prima faciereasonable. We thus get four agglomeration principles for reasonableness,namely, those corresponding to categories (al), (bl), (a2), and (b2).

Consider (a2) first. It is clear that agglomeration fails for prima facieobligation; this explains how conflicts of prima facie obligation are possi-ble. Steve has a beneficence-based reason to tend to Carl and a fidelity-based reason to dine with Dave, but there is no X such that Steve has anX-based reason both to tend to Carl and to dine with Dave (for he cannotdo both these things).14 Similar considerations pertain to (b2).

Consider (bl) next, where we now take this to concern what is overallmorally permissible. Again, it is clear that agglomeration fails here, that is,that we must reject

(7.10) if P(A) &P(B), then P(A&B)9

and my account of overall obligation certainly doesn't imply that (7.10) istrue. It may well be overall permissible for Alex to accept Elvis's invitationand overall permissible for her to accept Irving's, without its being overall

12 See Williams (1987), pp. 132-3; van Fraassen (1987), p. 146; Hudson (1986), pp. 16-18.13 Cf. Gert (1988), pp. 23-4.14 Cf. Swank (1985), pp. 113-14; Sinnott-Armstrong (1988), pp. 129-30; Pietroski (1993),

pp. 490, 504.

213

permissible for her to accept both. For it may be that the deontically bestworlds accessible to Alex are worlds in which she accepts one or other invi-tation but not both.15

But the failure of agglomeration for categories (bl), (a2), and (b2) givesus no reason to think that it fails for category (al); that is, we have beengiven no reason by proponents of dilemmas to think that (2.40'b)/(7.8) isfalse. Indeed, there seems to be very good reason to think that it is true,independently of the fact that my account of obligation implies that it istrue. For, as Paul Pietroski notes,16 nondilemmatic cases seem clearly (bar-ring the actualist's concerns, which I addressed in Section 6.2) to conformto the principle of agglomeration, and it is difficult to explain this other thanby appealing to the principle itself.17 Of course, a proponent of dilemmascould reply that the principle at issue is simply this:

(7.11) if O(A) & O(B) & C(A&B), then O(A&B);

but the addition of C(A&B) seems purely ad hoc.A second argument against the possibility of basic dilemmas concerns

these four theses, first presented in Subsection 2.3.1:

(2.46) ifO(y4),thenC(/l);(2.43///) \£~C(A&~B) & C(~B), then, if O(A), then O(B);(2.36') ifO(A),thenP(/l);(2.37'a) if O{A)y then ~P(~A).

As noted in the last section, the first two of these jointly imply that, in casesof dilemmas, there is an act A such that the following is true of it:

(7.12) O(A)&O(~A).

Let us now focus our attention on (2.36'), (2.37'a), and (7.12). The first ofthese yields the following proposition (when times are explicitly stated):

(7.13) if S ought overall at Tt to do A at T3, then S may overall at Tt doA at T3.

The contrapositive of the second yields this proposition (when A is substi-tuted for ~A, and vice versa):

(7.14) if S may overall at Ti do A at T3, then it is not the case that S oughtoverall at T7 not to do A at T3.

15 Cf. Sinnott-Armstrong (1988), pp. 133-4.16 Pietroski (1993), pp. 503-4.17 Cf. also Sinnott-Armstrong (1988), pp. 134-5.

214

And it is clear that (7.13) and (7.14) jointly contradict (7.12). To preservethe possibility of dilemmas, then, one must reject at least one of the fourtheses that function as premises in this argument.

Again, I shall not seek to defend (2.46) here; I have said all that I wantto say in Chapter 3. As to (2.43'"): the strong intuitive plausibility of thisproposition was noted in both Chapter 2 and Chapter 6; I have no more toadd here. And (2.36') and (2.37'a) are surely extremely plausible. A propo-nent of basic dilemmas might nonetheless seek to undermine this secondargument by noting the following: I have already acknowledged that it ispossible for there to be the analogue of a basic dilemma when it is primafacie, and not overall, obligation that is at issue and also when it is permis-sibility (whether prima facie or overall), and not obligation, that is at issue;hence I must admit that the analogues of this second argument break downsomewhere in these cases, and it may be that this would give us reason tothink that the second argument itself likewise breaks down. Let us nowexamine this matter.

Consider prima facie obligation first. I of course grant that the analogueof (7.12) is possible, that is, that it can happen that

(7.15) S ought prima facie at Tt to do A at T3 and S ought prima facie atTt not to do A at T3.

Moreover, the account of prima facie obligation that I have given inChapter 5 commits me to accepting this analogue of (2.46):

(7.16) if S ought prima facie at Tt to do A at T3, then S can at T1 do A at

And I grant also the truth of the relevant analogues of (2.43"') and (2.36').But the analogue of (2.37'a)/(7.14) is to be rejected. That is, we should notaccept:

(7.17) if S may prima facie at Tt do A at T3, then it is not the case that Sought prima facie at Tt not to do A at T3.

This is because prima facie obligations (and prima facie permissions) are, asnoted above, parti-resultant rather than toti-resultant. Thus, for example,although Steve has a beneficence-based prima facie obligation (and primafacie permission) to tend to Carl, he also has a fidelity-based prima facieobligation not to tend to Carl but to have dinner with Dave; again, thisexplains how conflicts of prima facie obligations are possible. But I cannotsee how this provides any support whatsoever for the proponent of basicdilemmas.

215

As to permissibility (whether prima facie or overall - let's focus on over-all): it is clear that my account commits me to the analogue of (2.46),namely,

(7.18) if P(A), then C(A);

and it also commits me to the analogue of (2A3"'). In addition, theanalogue of (2.36') is trivially true. But, once again, the analogue of(2.37'a)/(7.14) is to be rejected. That is, we should not accept:

(7.19) if 5 may overall at T1 do A at T3, then it is not the case that 5 mayoverall at Tt not do A at T3,

and this is for the obvious reason that ties for what is deontically best areperfectly possible. Once again, though, I cannot see how this provides anysupport whatsoever for the proponent of basic dilemmas.

I am thus insisting on the truth of (2.37'a)/(7.14), on the truth, that is,of the claim that what it is overall obligatory to do it is not overall permis-sible not to do, or, contrapositively, that what it is overall permissible to doit is not overall obligatory not to do. I am not sure how to defend this claim,beyond pointing to its being implied by the account of overall obligationthat I have provided. Perhaps one way to defend it stems from considera-tions of supervenience. Many philosophers will agree that, if an act is oblig-atory (whether prima facie or overall), as of a certain time, then this moralfact supervenes on a certain set of nonmoral facts that obtains at thattime. This set of nonmoral facts grounds the obligatoriness of the act; itrenders the act obligatory. Similarly, if refraining from an act is permissible(whether prima facie or overall), as of a certain time, then this moral factalso supervenes on a certain set of nonmoral facts that obtains at that time.This set of nonmoral facts grounds the permissibility of refraining from theact; it does not, therefore, render the act obligatory. Now, as has beenrepeatedly observed, it is of course quite possible that an act should be primafacie obligatory and yet refraining from it be prima fade permissible. This isbecause such moral facts are, as just noted, pdrfr'-resultant; they are, that is,grounded in different sets of nonmoral facts, one set rendering the act oblig-atory, the other not. But an act cannot be overall obligatory while refrain-ing from it is overall permissible, for such moral facts are fofr'-resultant; theyare grounded in one and the same set of nonmoral facts (namely, the set ofall the nonmoral facts18 that obtain at the time in question), and it cannotbe that this set both renders the act obligatory and fails to render it so.

18 Or, at least, all the nonmoral facts that can serve as a supervenience base for moral facts.

216

I find this argument persuasive. No doubt some will not find it so, eitherbecause they are skeptical about the supervenience of moral on nonmoralfacts or because they believe that it begs the question to say that, if a set ofnonmoral facts grounds the permissibility of refraining from an act, thenthat same set cannot ground the obligatoriness of the act. If so, I do notknow what else to say in defense of (2.37'a)/(7.14). Further defense of thisclaim seems to me unnecessary, though, for it is the kind of proposition thatshould serve as the touchstone of an ethical theory. If my account of over-all obligation had contradicted it, then it is the account, not the claim, thatwould have had to be rejected.

7.3 THE CASE FOR BASIC MORAL DILEMMAS

The ground covered in the last section is, for the most part, extremely welltrodden. Almost all the considerations that I have raised against the possi-bility of basic dilemmas have been raised, many times, by others.19 Butthey deserve to be stated again, because, despite what I take to be thegreat strength of these considerations, it seems that nowadays proponentsof moral dilemmas far outnumber their opponents. This is a puzzling phe-nomenon. Of course, it may be — may well be — that the appearance of dis-agreement is merely an appearance, that what "proponents of dilemmas"are proposing does not in fact conflict with what "opponents of dilemmas"are opposing. This indeed seems frequently to be the case. Opponents ofdilemmas may, I think, in every case be taken to be opponents of what Ihave called "basic moral dilemmas." But what is it that proponents ofdilemmas are proposing? Well, consider the article by E. J. Lemmon thatinstigated the recent debate on dilemmas. There Lemmon distinguishesvarious types of situations that he calls moral dilemmas, but none of themseems to be a basic moral dilemma. In particular, what he calls "thesimplest variety of moral dilemma in the full sense" seems best understoodas simply a conflict of prima facie obligations, for the example he gives isthis:

A friend leaves me with his gun, saying that he will be back for it in the evening,and I promise to return it when he calls. He arrives in a distraught condition,demands his gun, and announces that he is going to shoot his wife because shehas been unfaithful. I ought to return the gun, since I promised to do so... Andyet I ought not to do so, since to do so would be to be indirectly responsible

19 See especially McConnell (1987), Conee (1987), and Feldman (1986), Section 9.1. Cf.also Donagan (1987).

217

for a murder... I am in an extremely straightforward moral dilemma, evidentlyresolved by not returning the gun.20

It could be that Lemmon regards this situation as what I have called a basicmoral dilemma, but his talk of resolution indicates otherwise. Conflicts ofprima facie obligation are resolved when the most stringent prima facieobligation is acted on; no such resolution is available for conflicts of overallobligation. Of course, conflicts of overall obligation might be subject tosome sort of nonmoral resolution (perhaps satisfying one of the obligationswould serve the agent better than satisfying the other), but it is apparentthat Lemmon is concerned with a moral resolution to the conflict that heillustrates.

It is clear that other prominent proponents of dilemmas are also at timesconcerned only with conflicts of prima facie obligation, so that there is inreality no opposition on this score between them and opponents of dilem-mas. For example, Bernard Williams allows for the (moral) solubility of atleast some of the conflicts of obligation that he is concerned with, even ifthese are not "soluble without remainder,"21 that is, even if their resolutionleaves a certain sort of moral residue. Bas van Fraassen talks of "what oughtto be for one reason and what ought to be for another reason,"22 and this isstrongly reminiscent of the/wfr'-resultant nature of prima facie obligation.Marcus declares that "it is a better fit with the moral facts that all dilemmasare real, even where the reasons for doing x outweigh, and in whateverdegree, the reasons for doing y."23 And Christopher Gowans endorses anapproach to dilemmas according to which "even though S ought to do Aand S ought to do B, there is nonetheless some sense in which, in the finalanalysis, (say) S ought to do B [and not ^4]."24 All such talk is most readilyunderstood in terms of conflicts of prima facie obligation and, as such,seems wholly unobjectionable (except that Marcus's claim is surely too lib-eral; as we saw earlier, not all conflicts of prima facie obligation warrant thelabel "dilemma").

Nonetheless, it is also clear that proponents of dilemmas are at times con-cerned with unresolvable conflicts of obligation, and this may seem to indi-cate a concern with basic moral dilemmas. But let us not be hasty. Therecan be situations that are (morally) unresolvable that don't constitute basicdilemmas. For example, if more than one course of action is overall

20 Lemmon (1987), p. 105.21 Williams (1987), p. 129.22 Van Fraassen (1987), p. 141.23 Marcus (1987), p. 193.24 Gowans (1987a), p. 25.

218

permissible, then there is no moral resolution to this conflict between alter-natives.25 Even if one chooses to call such a situation "dilemmatic," it iscertainly not a basic dilemma; for wrongdoing is not inevitable. Indeed, asnoted earlier, it is just such a situation that Sinnott-Armstrong calls "dilem-matic," and so, although I am an opponent of dilemmas, I do not opposehis view. And it may be that others have adopted his view. For example,Williams at one point talks of "extreme cases of moral conflict, tragiccases," where the following may be said:

The agonies that a man will experience after acting in full consciousness of sucha situation are not to be traced to a persistent doubt that he may not have cho-sen the better thing; but, for instance, to a clear conviction that he has not donethe better thing because there was no better thing to be done.26

So described, a tragic case appears to be "no more" than a particularlyexcruciating conflict between equally stringent prima facie obligations thanwhich there is none more stringent. Of course, I accept the possibility thatsuch cases may arise. Similarly, van Fraassen talks of conflicts "which can-not be resolved in terms of one reason overriding another."27 This, too,merely suggests a tie for top place.

Further evidence that proponents of dilemmas are often not concernedwith proposing the possibility of basic moral dilemmas may be adducedsimply from the fact that they frequently don't even bother to address theformidable arguments presented in the last section against this possibility.Such an attitude betrays a total lack of concern with the possibility of basicdilemmas. And even when attention is apparently paid to these arguments,we need to be careful what to infer from this. For example, Lemmon rejectsthe principle that "ought" implies "can," and yet, as I have said, it is clearthat he is concerned with "prima facie ought." Of course, I believe that heis mistaken to reject the principle even when so understood, but the pointhere is that his apparent concern with the arguments presented in the lastsection is, again, merely apparent. Others also may appear to have addressedsome of these arguments (Williams and Marcus both reject the principle ofagglomeration;29 van Fraassen rejects the principle that "ought" implies"may"),30 but, again, it may be that their concern is not with these argu-

25 At least, none of the alternatives is preferable in terms of deontic value. There are othermorally relevant values, as noted in Ch. 2, n. 13.

26 Williams (1987), p. 123.27 Van Fraassen (1987), p. 141.28 Lemmon (1987), p. 114, n. 2.29 Williams (1987), pp. 132-3; Marcus (1987), p. 200.30 Van Fraassen (1987), pp. 145-6.

219

ments but with some closely related arguments. If so, there may not afterall be any conflict between their views and that of opponents of dilemmas.

Still, proponents of dilemmas do sometimes make remarks which sug-gest that they accept the possibility of basic moral dilemmas, the possibili-ty that an agent will inevitably do overall wrong. Williams talks of tragiccases as cases where the agent accepts that "he has done something that heought not to have done, or not done something that he ought to havedone;"31 Marcus talks of being "damned if you do or ... damned if youdon't";32 and although such talk can be construed as advocating merely thepossibility that an agent will do prima facie wrong, it seems wise to grantthat it may be intended as an endorsement of the possibility that an agentwill inevitably do overall wrong and to examine it as such.

What, then, are the reasons that have (perhaps) been advanced by pro-ponents of dilemmas for accepting the possibility of basic dilemmas? Anumber of reasons appear to have been given. I shall now look into someof these, in what I take roughly to be ascending order of plausibility.

First, it may be that some proponents of dilemmas believe that it is justobvious that basic dilemmas can occur. Cases such as Sartre's, that ofAgamemnon at Aulis (cited by Williams),33 that of Sophie's choice34 (citedby many), and others are often presented as being clearly dilemmatic. Well,it is clear that a conflict of some sort is at issue in these cases, but is it clearthat what's at issue is a basic dilemma? How could this be, even from theproponents' point of view? For they of course grant that there can be con-flicts that are morally resolvable, and some of these may be very hard cases(in the sense that it is very difficult to figure out just what the solution is).If so, it would seem that it can never be obvious that a particular conflict con-stitutes a basic dilemma rather than merely a very hard but resolvable case.

Another reason for accepting the possibility of basic dilemmas that mayhave been proposed is that cases of symmetry imply that such dilemmas arepossible. One such case is that provided by Marcus of the twins whose livesare in jeopardy. It is clear, however, that such cases do not imply that basicdilemmas are possible; they simply involve equally overall permissible, notequally overall obligatory, alternatives.

Dilemmas have often been diagnosed in cases not of symmetry but ofincommensurability.35 Such cases involve one course of action's being

31 Williams (1987), p. 123.32 Marcus (1987), p. 194.33 Williams (1987), p. 123.34 Styron (1980).35 See, e.g., Sinnott-Armstrong (1988), p. 58fF.; Lebus (1990).

220

neither superior nor inferior nor equal in deontic value to another. But, aswe saw in Subsection 2.2.6, what such cases, just like cases of symmetry,imply is that each such alternative is equally overall permissible, not thateach is equally overall obligatory. It may be that such cases are or can beespecially wrenching, in that something of incomparable value is inevitablyforgone, but this doesn't give them the status of a basic dilemma. (Actually,I don't see why they should be especially wrenching; Marcus's case of thetwins, which would appear to involve two comparable alternatives, seemsto me just about as wrenching as any situation can be.)

Some have diagnosed dilemmas when the agent cannot avoid having"dirty hands." This may simply be an acknowledgment that some situa-tions are, as Williams says, not "soluble without remainder," that is, that nomatter what one does, there will be a moral cost involved. But a moral costcan be incurred without an overall obligation's being violated; when Stevebroke his dinner date with Dave, he incurred such a cost. Michael Stocker,however, suggests that in certain cases, cases where one is wrongly coercedby another to help implement some evil plan, there is a special reason totalk of "dirty hands" and that, in these cases, wrongdoing is inevitable.36 Anexample would be Williams's case of Jim and the Indians, cited in Section2.1.37 But once again, while such cases may be especially wrenching, thereis no need to regard them as involving inevitable overall wrongdoing.Obviously we may agree that it is usually wrong, overall, to implement evilplans, but the question is what one should do under the actual circumstances.After all, Jim's circumstances hardly merit being termed "usual." Perhaps,under the circumstances, Jim ought overall to shoot the Indian, as the con-sequentialist would presumably say. Perhaps he ought overall not to shootthe Indian, as the Kantian would presumably say. Perhaps there is both astrong reason to shoot and a strong reason not to shoot, as the Rossianwould presumably say. It may not be clear to Jim just what he oughtoverall to do, in which case he is in a quandary. It may be that no option isdeontically better than the other, in which case he is in a situation involv-ing either symmetry or incommensurability. But none of this impliesboth that Jim ought overall to shoot and that he ought overall not toshoot.

Remarks made by Marcus suggest the following argument:38

(7.20) one ought overall always to avoid conflicts of obligation;

36 Stocker (1990), p. 25.37 Williams (1973), p. 98.38 Marcus (1987), p. 199.

221

(7.21) unless such conflicts were possible and were basic dilemmas, itwould not be the case that one ought overall always to avoid them;

hence,

(7.22) basic dilemmas can occur.

Both premises are false. Sometimes it is not the case that one ought to avoidconflicts of obligation of the sort that Marcus cites. After all, perhaps theonly way to have avoided being in a position to save one but not both twinswould have been to avoid the scene of the tragedy altogether, in which caseboth lives would have been lost. And even when one ought to avoid suchconflicts of obligation, this can be explained without invoking basic dilem-mas. For example, one ought (usually) to avoid making incompatiblepromises. Why? Not because they put one in a basic dilemma, but simplybecause they (usually) involve incurring unnecessary moral cost.

It might be argued, however, that such cost can give rise to further over-all obligations and that this is, on these occasions, a sign that one was in abasic dilemma. Perhaps one would have an overall obligation to explain tothe twins' next of kin why not both were saved; perhaps one would havean overall obligation to apologize to, or even to compensate, the disap-pointed promisee. But it is clear, as the discussion of residual obligationin Subsection 5.4.4 attests, that such overall obligations can arise outof the failure to meet prima facie obligations that are not themselvesoverall obligations. There is no need to diagnose a basic dilemma in suchcases.

There is one type of moral residue in particular that is frequently citedas a sign of moral dilemmas, and that concerns emotions. As noted inSubsection 5.4.4, Ross believes that a certain feeling of compunction iscalled for when a prima facie obligation is not satisfied. But in some casesmore than mere compunction is called for. Williams notes that it is some-times appropriate to regret whatever it is that one has done, and he takes thisas evidence of dilemmas.39 If it is to be taken as evidence of basic dilemmas,the argument would seem to be this:

(7.23) it is appropriate for S to regret doing A [or not doing A] only if itwas overall obligatory for 5 not to do A [or to do A];

(7.24) it is possible both that it would be appropriate for 5 to regret doingA and that it would be appropriate for S to regret not doing A;

39 Williams (1987), p. 122ff.

222

hence,

(7.22') it is possible both that it was overall obligatory for 5 not to do Aand that it was overall obligatory for S to do A.

We must surely grant the second premise. Sartre's student had cause forregret no matter which choice he made; Agamemnon likewise; Sophie too.But the first premise is false. Regret is appropriate whenever a moral costis incurred, and this can happen without an overall moral obligation's beingviolated.

This last point is widely accepted, but then proponents of dilemmas sug-gest this modification to the argument: replace talk of regret with talk ofremorse or, equivalently, guilt feelings.40 Instead of (7.23) and (7.24) we wouldthen have, respectively:

(7.25) it is appropriate for 5 to feel remorse for (guilty about) doing A [ornot doing A] only if it was overall obligatory for S not to do A [orto do A];

(7.26) it is possible both that it would be appropriate for 5 to feel remorsefor (guilty about) doing A and that it would be appropriate for S tofeel remorse for (guilty about) not doing A;

and we could then infer (7.22') afresh.This is an interesting argument. I think we should agree that there is a

conceptual tie between remorse and overall wrongdoing that does notobtain between regret and overall wrongdoing. Certainly being guilty isintimately related to overall wrongdoing; moreover, it seems plausible tosay (although I shall express reservations about this shortly) that one feelsguilty about, or remorse for, doing something only if one believes that oneis guilty with respect to doing it. Given this, it seems plausible to say, alongwith van Fraassen, that "it is appropriate to feel guilt if and only if one isguilty";41 for then the belief that is part of the emotion will correspond withthe facts. Let us therefore (provisionally) accept the following:

(7.27) it is appropriate for S to feel remorse for (guilty about) doing A[or not doing A] iff S is guilty with respect to doing A [or notdoing A].

The question is, when is it that one is guilty with respect to doing [or notdoing] something? One answer is simply this:

40 See, e.g., van Fraassen (1987), p. 147; Marcus (1987), pp. 196-7.41 Van Fraassen (1987), p. 147.

223

(7.28) S is guilty with respect to doing A [or not doing A] iff 5 failed tosatisfy an overall obligation not to do A [or to do A].

That is, one is guiltyjust in case one does, or has done, overall moral wrong.(7.27) and (7.28) jointly imply (7.25), the first premise of the argumentpresently under consideration. But why, then, accept the second premise,(7.26) (once we are clear that it is remorse in particular, rather than someother negative emotion such as regret, that is at issue)? I know of no reasonhaving been advanced for accepting it, other than to insist that basic dilem-mas are possible, and to cite this as a reason would of course be to beg thequestion. (Actually, I'm not being quite honest here; I shall qualify thisremark in the next section.)

There is another answer to the question of when it is that one is guiltywith respect to doing [or not doing] something. It is this:

(7.29) 5 is guilty with respect to doing A [or not doing A] iff 5 believedthat he was failing to satisfy an overall obligation not to do A [or todo^].4 2

I think we must agree that, on this understanding of what it is to be guilty,the second premise of the argument, (7.26), is true. For it is, obviously,quite possible for someone to believe that he ought overall to do some actionand also that he ought overall not to do it. But now it is the first premise,(7.25), that is to be rejected. (7.27), conjoined with (7.29), does not war-rant the claim that actual overall wrongdoing is necessary for actual guiltand thus necessary for appropriate remorse; on the contrary, presumedoverall wrongdoing will suffice.43

Thus this, the strongest argument for basic dilemmas of which I amaware, fails. It is worth noting, finally, that the account of the appropriate-ness of remorse, (7.27), which I have provisionally conceded to the pro-ponent of dilemmas, is in fact open to question. Suppose that Tracy, a truckdriver, has just run over Reggie, a small boy who darted out into the streetin front of Tracy's truck. Tracy did no wrong, let us assume; nor did she ordoes she believe that she did wrong. She is therefore not guilty, accordingto either (7.28) or (7.29), with respect to running over Reggie. Still, I think

42 This is roughly the answer given in Zimmerman (1988a), Chapter 3, with "guilty"understood to mean the same as "culpable." The account given there actually supportsonly the "only if," and not the "if," of (7.29), since more than such a belief is requiredfor culpability; freedom is also required.

43 Qualification: suffice along with other conditions, not among which is actual wrongdo-ing. See the last note.

224

we would all be a little put out if she showed no remorse for what she haddone. True, there is, in a sense, nothing for her to feel guilty about, and yetwouldn't such a reaction bespeak a coolness of character that would likelyprompt an absence of remorse in cases where it was clearly called for? I thinkthat what we should say here is that Tracy's reaction is intrinsically appro-priate but (probably) extrinsically inappropriate, inasmuch as it (probably)manifests a tendency to react in an intrinsically inappropriate way on otheroccasions.44 It seems, then, that (J.21) is at best acceptable only as anaccount of the conditions of the intrinsic appropriateness of remorse. But,even when it is so understood, there is reason to question it. This is becausethe claim concerning the connection between feeling guilty and beingguilty, on which I originally based (7.27), is itself dubious. Robert Robertshas contended, with considerable plausibility, that feeling guilty doesn'tpresuppose believing oneself to be guilty so much as construing oneself asguilty.45 One example might be that of a lapsed Catholic, who believes thelapse to be justifiable and yet feels guilty about not attending Mass. Robertslikens such cases to optical illusions which are understood to be illusory andyet persist. Given such cases, it is arguable that it is sometimes appropriate,and not merely extrinsically so, for agents to feel guilty, even though theyhave not done overall wrong and did not believe themselves to have doneoverall wrong. If this is correct (and here I shall not seek to settle the ques-tion), then (7.27) is to be rejected.

7.4 NONBASIC MORAL DILEMMAS

Once again, the ground covered in the last section is, for the most part,extremely well trodden.46 But once again, because proponents of dilemmasare so numerous and so vocal, it deserves to be covered anew. Let me nowturn to some relatively fresh considerations, several of which are designedto undermine the proponents' case even further.

I have so far been concerned with what I have called "basic moral dilem-mas," that is, with situations where the following is true:

(7.7) (a) 5 ought overall at Tt to do A at T3;(b) S ought overall at T1 to do B at T4; and(c) S cannot at T1 do both A at T3 and B at T4.

44 Cf. Zimmerman (1988a), pp. 134-5.45 Roberts (1988), pp. 183, 186-9, 195-6, and 201.46 See especially McConnell (1987), (1993b), (1995); Conee (1987); Feldman (1986),

Section 9.1.

225

I have argued that such situations cannot arise. Notice that, inasmuch as(2.46) (the claim that "ought" implies "can") is true, basic dilemmas mayin fact be more fully characterized as follows:

(7.7') (a) S ought overall at T, to do A at T3;(b) 5 can at Tt do A at T3\(c) S ought overall at T, to do B at T4;(d) 5 can at T1 do B at T4; and(e) S cannot at Tt do both A at T3 and B at T4.

Notice, also, that the term "basic moral dilemma" suggests that some moraldilemmas are not basic. Consider, for example, this sort of situation (whereTj is earlier than T2, and T2 is not later than either T3 or T4):

(7.30) (a) S ought overall at Tt to do A at T3;(b) S can at T1 do 4 at T3;(c) 5 ought overall at T2 to do B at T4;(d) S can at T2 do B at T4\ and(e) S cannot at T2 do both A at T3 and B at T4.

Can such situations arise?Indeed they can, and that they can seems to me of great significance. Let

us call the type of nonbasic dilemma captured in (7.30) a bind. Examples ofbinds have in fact already been given. For instance, recall from Subsection3.2.2 the case of Mandy, who on Friday tells Phil that she will be able tohelp him move furniture on Sunday, thereby making a sort of informalcommitment to do so, but who on Saturday promises Bill that she willbabysit his child on Sunday, thereby making a formal commitment to doso. Her situation may be pictured as in Figure 3.2. When Mandy makes herpromise to Bill on Saturday, she puts herself in a bind, for she can no longeravoid doing overall wrong on Sunday. If she keeps her promise as sheought, then she'll do the remote wrong of not helping Phil; if she helps Philas she (once) ought, then she'll do the immediate wrong of not babysittingBill's child. Here it is most important to note that the remote wrong, if itwere committed, would be a genuine wrong, as I argued in Subsection3.2.2.

It seems to me perfectly appropriate to call binds a type of moral dilem-ma, for they involve wrongdoing that is, in a way, inevitable. It is true thatthe wrongdoing was not always inevitable, but it has become so (inMandy's case, it became so on Saturday). It is also true that there is no sin-gle time at which the conflicting "oughts" apply (for then we would havea basic dilemma); one applies at one time (from Friday to Saturday, in

226

Mandy's case), the other at another (from Saturday to Sunday). Nonethelessoverall wrong will have been done no matter what the agent chooses at thetime of action to do, for by then it is too late to extricate himself from thedilemma.

Consider now once again the argument concerning remorse, but withtime indices explicitly provided as follows:

(7.25') it is appropriate for 5 to feel at T4 remorse for (guilty about) doingA [or not doing A] at T3 only if there is a time Tsuch that it wasoverall obligatory at Tfor 5 not to do A [or to do A] at T3;

(7.26') it is possible both that it would be appropriate for S to feel at T4

remorse for (guilty about) doing A at T3 and that it would beappropriate for S to feel at T4 remorse for (guilty about) not doingA at T3;

hence,

(7.22") it is possible both that there be a time Tsuch that it was overallobligatory at T for S not to do A at T3 and that there be a time Tsuch that it was overall obligatory at Tfor S to do A at T3.

This argument may be sound. We should accept the first premise, if (7.27)survives the doubts raised at the end of the last section and guilt is construedas consisting in actual wrongdoing (see (7.28)). I think that we should alsoaccept the second premise. Now I said earlier that I knew of no reason hav-ing been advanced to accept (7.26) other than to insist that basic dilemmasare possible, but I also said that I wasn't being quite honest in making thisclaim. I can now explain, for / now wish to advance a different reason foraccepting (7.26'). Look at (7.22"). It doesn't say that basic dilemmas are pos-sible. That basic dilemmas are possible is expressed by the followingproposition:

(7.22'") it is possible that there be a time Tsuch that it was overall oblig-atory at Tfor S not to do A at T3 and it was overall obligatory atTfor S to do A at T3.

Here the quantifier over times has broad scope, so that the time of "ought"is said to be identical with the time of "ought not." But in (7.22") no suchidentity is posited, and this is why it is acceptable. As long as binds are pos-sible — and they are — we must accept (7.22"); for the following, which con-stitutes a bind, implies (7.22"): it was overall obligatory at Tt for S not todo A at T3 and it was overall obligatory at T2 for 5 to do A at T3. But thisdoesn't commit us to accepting the possibility of basic dilemmas, and now

227

we can see that (7.26') is acceptable after all. Think of Mandy on Monday.If she kept her promise to Bill on Sunday, then it would be appropriate forher on Monday to feel remorse for not having helped Phil; and if she helpedPhil on Sunday, then it would be appropriate for her on Monday to feelremorse for doing so, since she would thereby have failed to keep herpromise to Bill. In this way I think we can concede to proponents ofdilemmas their strongest argument and yet deny that basic dilemmas canoccur.

One objection to what I have just said is this.47 If it is to be appropriatefor Mandy to feel remorse on Monday for what she did (whatever it was)on Sunday, then her belief concerning her guilt that renders her remorseappropriate must itself be accurate. Thus she must either believe onMonday that she was obligated from Friday to Saturday (but not beyond)to help Phil on Sunday, or believe that she was obligated from Saturday toSunday to babysit Bill's child on Sunday, or both. But such time-specificbeliefs concerning remote obligations are surely extremely rare, and henceit is extremely unlikely that Mandy's remorse will be appropriate after all.

While I agree (barring the doubts raised at the end of the last section)that an agent's belief concerning his guilt must be accurate if his remorse isto be appropriate, I don't think we need agree that it must be fully accurate.If Mandy believes on Monday both that she ought to have helped Phil andthat she ought to have babysat Bill's child, then her remorse for havingfailed to satisfy one of these obligations will be appropriate (given (7.28)),whether or not the content of her belief was so specific as to include explic-it time indices with respect to her obligations; at least, this is so as long asshe doesn't attach incorrect time indices to her obligations. Thus it need notbe at all rare that someone, who feels guilty about an action performedwhen he was in a moral bind, does so in an appropriate way.

Binds are not the only type of nonbasic dilemma. Consider this situa-tion (where "ought/' expresses primary overall obligation and "ought2"secondary overall obligation, as discussed in Section 4.4):

(7.31) (a) S oughtj overall at T, to do A at T3;(b) 5 can at T1 do A at T3;(c) 5 ought2 overall at T1 to do B at T4;(d) S can at Tt do B at T4; and(e) S cannot at Tt do both A at T3 and B at T4.

47 What follows is borrowed from comments made by Earl Conee on an early version ofZimmerman (1987) on the occasion of its presentation at the 1984 meeting of the EasternDivision of the American Philosophical Association.

228

We may call this sort of situation a subsidiary dilemma. Such situations arepossible; indeed, examples have already been given. Recall Alan, forinstance, who ought t to attend the 10:00 A.M. meeting on the first floor butfails to do so and, as a result, ought2 to attend another 10:00 A.M. meetingon the second floor. He can do each of these things, but he cannot do themboth.

Consider, finally, this situation (where S1 is one person and S2 another):

(7.32) (a) Sj ought overall at T, to do A at T3;(b) Sj can at T, do A at T3;(c) S2 ought overall at Tt to do B at T4;(d) S2 can at Tt do B at T4; and(e) S1 and S2 cannot jointly at Tt do both A at T3 and B at T4.

We may call this sort of situation an interpersonal dilemma. Whether or notsuch situations can arise is an issue that I shall address in Section 9.4.

We find, then, that at least three types of nonbasic moral dilemma canbe distinguished. There are also variations on these types ("subsidiarybinds," "interpersonal binds," "interpersonal subsidiary dilemmas," and soon). Moreover, the account of overall obligation that I have proposed,while repudiating the possibility of basic dilemmas, accommodates the pos-sibility of at least some nonbasic dilemmas. There is an aspect to this accom-modation at which I have only hinted but which I wish now toacknowledge explicitly. With respect to binds and subsidiary dilemmas, myaccount implies that such dilemmas can arise only by way of some wrong-doing. It thus gives detailed expression to the traditional Thomist distinc-tion between a perplexity simpliciter and a perplexity secundum quid.4S Thedistinction may be broadly characterized in this way: a perplexity simpliciterarises if, through no wrongdoing on anyone's part, it happens that what-ever can be done will be wrong; a perplexity secundum quid arises, on theother hand, if it happens that whatever can be done will be wrong, but thispredicament has been induced by wrongdoing on someone's part. AlanDonagan, one of the few philosophers to have entertained this distinctionin recent years, claims, as does Aquinas, that morality would be inconsis-tent if it admitted perplexities simpliciter but not if it admitted perplexitiessecundum quid; he claims further that morality is not inconsistent, and so doesnot admit perplexities simpliciter, although it does admit perplexities secun-dum quid.49 In this he echoes Georg von Wright, who puts the point pic-

48 Aquinas (1964-75), Ia2ae. 19,6 and 3a. 64,6.49 Donagan (1977), pp. 144-5, and (1987), pp. 285-6.

229

turesquely as follows: "[P]redicament, though logically possible, can onlyarise through antecedent sin... It is only as a consequence of a fall that aman can come to be in a predicament."50 Von Wright talks of the agent'ssituation before the "fall" as "prelapsarian." Let us say that (7.30) and (7.31)represent two different types of predicament into which agents may enterby way of a moral lapse.

It is no accident that the illustrations that have been given of nonbasicdilemmas involve such lapses. (No illustration has been given of an inter-personal dilemma, and you have no way yet to tell whether my accountaccommodates such dilemmas or whether, if it does, it does so only byvirtue of some lapse. Again, I shall discuss this later, in Section 9.4.) Theillustration of a bind has it that Mandy wrongly promises Bill to babysit hischild. Such a lapse is required on my view, for only by way of such a lapse isit possible that what formerly was deontically best (helping Phil move fur-niture) no longer is so. Had Mandy not deviated from the strait and narrow(departed from the path of righteousness, made a turn for the worse) bymaking her promise to Bill, babysitting Bill's child would not have becomethe best thing for her to do; it would have remained second-best.51 Butgiven the deviation onto a different path, helping Phil drops from first- tosecond-best. Hence she becomes obligated to babysit Bill's child; buthence, also, she is no longer obligated to help Phil. Her lapse does notgenerate a basic dilemma.52

50 Von Wright (1968), p. 79.51 It might be thought that Mandy's helping Phil move furniture could have been rendered

second-best not by Mandy but by something else (e.g., by Phil's not really caring verymuch whether she helped him, or by Bill's child desperately needing to be babysat byMandy in particular), so that its being obligatory from Saturday on for Mandy to babysitthe child is not due to any wrongdoing on her part (and perhaps not on anyone's part).But this is in fact not possible. Either Mandy could have prevented this other event - therendering second-best of her helping Phil - or she could not. If she could not, then herhelping Phil simply never was the best that she could achieve, contrary to the initialhypothesis. If she could, then her failure to prevent the event either constituted a lapseon her part or it didn't. If it did not, then the best that she could achieve involved allow-ing her helping Phil to be rendered second-best; but in this case her helping Phil simplynever was the best that she could achieve, again contrary to the initial hypothesis.Compare the remarks made in Subsection 3.2.4 concerning the cancellation of obliga-tion.

52 Opponents of dilemmas have sometimes muddied the waters by claiming that perplex-ities secundum quid are possible but without acknowledging that such perplexities cannottake the form of basic dilemmas. (See von Wright, 1968, pp. 80—1; Donagan, 1977, pp.144-5, and 1987, pp. 285-6; McConnell, 1987, p. 160.) Basic dilemmas cannot arise byway of a lapse, because basic dilemmas are impossible, period, for the reasons given inSection 7.2. Cf. Feldman (1986), pp. 204-5.

230

It is clear, too, that subsidiary dilemmas can arise only by way of wrong-doing. Alan's obligation to attend the meeting on the second floor arisesonly out of his failure to attend the meeting on the first floor. Were he todo what he ought! to do, it never would be the case that he ought2 to dodifferently. I confess that here talk of "dilemma" is quite possibly mislead-ing, for it seems to be at best an exaggeration to say that it is inevitable thatAlan do overall wrong. While it is true that he cannot avoid doing wrongif he does what he ought2 to do, the fact is that he can avoid its being thecase that he ought2 to go to the meeting on the second floor; he can avoidthis by going to the meeting on the first floor, and in so doing he would dono overall wrong at all. In this respect, subsidiary "dilemmas" are both likeand unlike binds. They are like binds in that the latter also do not involvewrongdoing that can in no way be avoided by the agent; if Mandy had notacted wrongly on Saturday by making her promise to Bill, she would haveavoided her bind altogether. But subsidiary "dilemmas" are unlike binds inthat the latter do involve there being a certain time (Saturday, in Mandy'scase) at which wrongdoing has become inevitable. So perhaps conflicts ofsubsidiary obligations really should not be called a species of dilemma at all.I include discussion of them here for the simple reason that their formresembles that of other conflicts that are properly called dilemmatic (com-pare (7.31) with (7.7), (7.30), and (7.32)) and because they can arise onlyby way of a lapse.

231

8

Supererogation

Steve, introduced in Chapter 5, stopped to render first aid to Carl, who hadhad a car accident, thereby failing to keep his dinner date with Dave. Thiscase was used, following W. D. Ross, as an illustration of one prima facieobligation overriding another and thereby constituting an overall obliga-tion. But was Steve really overall obligated to do what he did? I am surethat many would say that he was not. On the contrary, they would say thathe went well beyond the call of duty; his action was supererogatory.

8.1 THE NATURE OF SUPEREROGATION

8.1.1 Going beyond prima facie obligation

It may be that some writers consider supererogation simply to consist ingoing beyond some prima facie duty or obligation. This is one way to inter-pret, for instance, the following passage by Joel Feinberg:

A janitor has a duty to spend eight hours cleaning his employer's floors. Heworks ten hours for eight hours' pay. Duty required eight hours; he gave dutyplus two, and thus in a perfectly intelligible sense, did "more than" his duty.1

This passage may seem to suggest that the janitor has a prima facie obliga-tion (of fidelity, perhaps) to work eight hours; in working ten hours, hegoes beyond this duty and thus performs a supererogatory action. But evenif that is what this passage suggests, it is not, I am sure, what most peoplehave meant by the term "supererogatory," and it is not what I shall meanby it. For it is clearly possible that, in going beyond some prima facie duty,one is merely doing what one is overall obligated to do or, worse, failingto do what one is overall obligated to do, and in neither case is one actingsupererogatorily.

1 Feinberg (1968), p. 397.

232

8.1.2 Going beyond perfect obligation

Some writers conceive of supererogation in terms of going beyond perfectoverall duty or obligation. This leaves open the possibility that one never-theless has an imperfect duty to do the act in question. In Subsection 2.1.4several different senses of "(im)perfect duty" were distinguished, and someof these have been proposed as appropriate for the definition of supereroga-tion.

Kant claims that we have an imperfect duty of charity. One suggestedinterpretation of this was that, while we have an overall obligation to actcharitably, we enjoy a certain leeway with respect to how to satisfy it, inthat we may choose between displaying charity to A and displaying it to Band displaying it to C and so on. The duty is thus, in a certain sense, dis-junctive (see (X) in Subsection 2.1.3). In this way, it might be said, when1 act charitably towards a particular person, I am acting supererogatorily,since I have no overall obligation to do that particular act; I could have sat-isfied my overall obligation to display charity by acting charitably towardssomeone else entirely.

This won't do. Suppose that I have an overall obligation either to actcharitably towards A or to act charitably towards B or to act charitablytowards C (there are no other candidate recipients of my charity), and thattherefore none of these individual acts is such that I have an overall oblig-ation to perform it. Surely I do not "go beyond" duty, in the requisite sense,simply by acting charitably towards A while bypassing B and C; I simplyfulfill my duty of charity. Of course, if I act charitably towards all three, Imight then do something supererogatory; but if so, this does not consistsimply in the performance of a duty that is imperfect, in the present sense.

Another suggested interpretation of "imperfect duty" was that the dutyin question is a duty not so much to act in a certain way as to adopt a cer-tain maxim for action, and it has been proposed that a supererogatory actionis one that conforms with a maxim that satisfies such a duty.3 But while thismay be the best that can be done by way of squaring Kant's theory withsupererogation, it falls short of what is required for supererogation as this isnormally understood; for it does not accord the supererogatory act thestatus of lying beyond duty altogether.4

2 Cf. Heyd (1982), pp. 121-2.3 See Hill (1971), p. 71. Hill is speaking for Kant, and he suggests other conditions as well,

but not among these conditions is the agent's having no duty of any sort that applies to theaction in question.

4 Cf. Baron (1987), p. 244.

233

Another suggested interpretation of "imperfect duty" was that of anobligation that is not an obligation-to, that is, of an obligation that is notcorrelative to some right. (See Subsection 5.5.1 for a discussion of this cor-relativity.) I think many writers have understood supererogation in termsof acts that are not owed to anyone in the way that an act to the performanceof which someone has a claim-right is owed to that person. Feinberg'sexample of the janitor can certainly be construed in this way, as can hisexample (cited in Section 1.3) of giving a match to a stranger.5 Similarly,concerning cases such as Steve's Frances Myrna Kamm writes that only oneduty is involved (namely, Steve's duty-to-Dave to have dinner with him),and she says that the alternative action (namely, Steve's helping Carl) thussupersedes duty.6 But again this won't do, unless all overall obligations areobligations-to. For otherwise one may have an overall obligation to dosomething that goes beyond an obligation-to, and this can hardly rank assupererogatory. For example, even if a lifeguard who is "off duty" goesbeyond any specific duty-to when saving a swimmer, whereas a lifeguardwho is "on duty" does not, if the former nonetheless has an overall obliga-tion to save the swimmer — if, that is, he would do overall wrong in notsaving the swimmer — then he can hardly be said to be acting supereroga-torily in saving the swimmer.

8.1.3 Going beyond overall obligation

No, for an act to be supererogatory, it must be such that the agent has nooverall obligation to perform it, period. That is how I shall understand"supererogatory," and that is, I believe, how most people understand it.8

Of course, that is not all that's required. The act must also be such that theagent has no overall obligation not to perform it. Thus it must be neitherobligatory nor wrong; it must be (not just personally optional but) morallyoptional. And it must also be especially valuable, in some way. After all, notall optional acts are supererogatory; some merely serve to fulfill a disjunc-tive obligation rather than to "go beyond" obligation.

5 Feinberg (1968), pp. 392-3. Actually, Feinberg would not call giving the stranger a matchsupererogatory, since it does not involve any great sacrifice or risk. More on this later.

6 Kamm (1985), pp. 119-20.7 See New (1974), p. 179ff.; Heyd (1982), pp. 119-20.8 The two most extensive treatments of supererogation to date concur. See Heyd (1982),

pp. 120, 125; Mellema (1991), p. 11 and passim.9 On the distinction between personal and moral optionality, see the discussion in

Subsection 2.1.2 immediately following the presentation of (I).

234

There is a flip-side to supererogation that some writers have recentlyremarked on. We may call it "suberogation."10 Whereas a supererogatoryact goes beyond duty, a suberogatory act, we may say, falls short of decen-cy. Perhaps Steve acted in a supererogatory manner when he stopped andhelped Carl; let us now assume so. Let us assume, that is, that he would havedone no overall wrong if he had passed Carl by and kept his date with Dave.But what if, in passing Carl by, he had taunted Carl, mocking his misfor-tune? This would have been a nasty thing to do, but perhaps it wouldn'thave been wrong. If it wouldn't have been wrong, it would have beensuberogatory: optional but especially disvaluable, in some way. (Here andhenceforth in this chapter, when I use the term "optional," I mean thatwhich is morally optional.)

How to render more precise the sense of "valuable" that pertains tosupererogation (and that of "disvaluable" that pertains to suberogation) isa very tricky question that I shall address later. (Be forewarned, though: Iwill not attempt to resolve this question.) Nonetheless, this much shouldbe said now: a supererogatory act is more valuable than any merely option-al — that is, optional but not supererogatory — alternative; for a supereroga-tory act "goes beyond" what duty requires and, where there are moraloptions, duty requires only that one not do what is not optional. (Similarly,a suberogatory act is less valuable than some merely optional - that is,optional but not suberogatory - alternative.) This condition must not bemisunderstood. In saying that supererogatory acts go beyond duty by beingmore valuable than any merely optional alternative, I am not saying that allsupererogatory acts are more valuable than all nonsupererogatory (whetherobligatory or merely optional or wrong) acts. Suppose that in Situation 1John, a janitor, has two alternatives, that of working ten hours (act A) andthat of working eight (act B); and let us suppose that A would besupererogatory and B not. In Situation 2 John has two other alternatives,that of working twelve hours (act C), as he has promised to do, and that ofworking eleven hours (act D); and let us suppose that C would be obliga-tory and D therefore wrong. Whatever it is that makes A more valuablethan B, it seems plausible to say that both C and D are therefore more valu-able still, even though A is supererogatory while neither C nor D issupererogatory (indeed, the latter is wrong). The point is, however, that Ais more valuable than any nonsupererogatory alternative, that is, more valu-able than any nonsupererogatory option that John has in Situation 1. This

10 Chisholm (1968) was perhaps the first to emphasize this phenomenon. It is there called"offence." The term "suberogatory" is used in Driver (1992) and Stump (1992).

235

point has sometimes been missed. For example, Roderick Chisholm saysthe following:

[I]f we contrast the man who performs an heroic act of duty with the man whoperforms a trifling act of supererogation,.. .we can say that the obligatory act ofthe former is better.. .than.. .the supererogatory act of the latter... [Hence we]cannot say that the supererogatory must be better than the obligatory.11

This observation leads Chisholm to give an account of supererogation thatdoes not acknowledge any sense in which the supererogatory "goesbeyond" duty.12 In doing so, while he allows for the superiority of C to A,he neglects the superiority of A to B.

To repeat, then: a supererogatory act is one that is both optional andmore valuable, in some way, than any nonsupererogatory alternative.Insofar as it involves optionality, supererogation is of course agent- andtime-relative, in that an act's being optional for one agent at one time is noguarantee that it is optional for that same agent at another time, let alonefor another agent at that same or another time. It can happen, therefore,13

that it is supererogatory for S at Tt to do A at T3 but not supererogatory for5 at T2 to do A at T3; and it can also happen that it is not supererogatory forS at T1 to do A at T3 but supererogatory for S at T2 to do A at T3. As to thefirst possibility: perhaps at T2 S makes a morally optional promise to do Aat T3, thereby rendering the act obligatory; or perhaps at T2 S makes anoptional promise not to do A at T3, thereby rendering the act wrong. As tothe second possibility: perhaps at T2 S commits an act that converts hisdoing A at T3 from obligatory or wrong to optional. (Such an act woulditself have to be wrong. See the discussion of lapses in Section 7.4.)

It is worth noting that talk of the optionality of a supererogatory act is,if left unqualified, most naturally associated with acts not being construedas act-tokens. As noted in Subsection 2.2.5, the claim that Jones ought tohelp that old lady on Monday is probably not to be understood as imply-ing that there is a particular act-token that is obligatory; rather, it is bestunderstood either as implying that a certain act-type (or some other abstractentity)14 is obligatory or as implying that a certain disjunction of act-tokensis obligatory. Similarly, the claim that it is optional, indeed supererogato-ry, for Jones to help that old lady on Monday is probably not to be under-

11 Chisholm (1968), pp. 415-16 and 421. Cf. Chisholm and Sosa (1966), p. 331.12 Chisholm (1968), p. 424ff.13 For all that has been said so far; but see the discussion of the possibility of supererogation

in the next section.14 See notes 63 and 68 to Ch. 2.

236

stood as implying that there is a particular act-token that is optional(supererogatory); rather, it is best understood either as implying that acertain act-type is optional (supererogatory), or as implying that a certainconjunction of act-tokens is such that each token is optional (supererogatory).

Similar remarks pertain to the suberogatory. Insofar as the suberogato-ry is optional, it can happen that an act is suberogatory for an agent at onetime but not at another. Furthermore, if "suberogatory" is to be applied toact-tokens, it is probably best applied to conjunctions of them.

8.2 THE POSSIBILITY OF S U P E R E R O G A T I O N

The foregoing is intended to provide a rough account of the nature ofsupererogation and suberogation. Even if it is accurate as far as it goes, how-ever, this of course does not imply that there are, or even can be, anysupererogatory or suberogatory acts; still less does it imply that those actsthat we normally call supererogatory are in fact so. The philosophical liter-ature is full of reasons for thinking that supererogation is impossible. Manyof the most influential normative theories seem to rule it out. It appears, forinstance, to be inconsistent with certain strains of Christianity,16 withKantianism,17 with utilitarianism and, more generally, consequentialism,18

and with Ross's theory;19 in every case the problem seems (roughly) to beeither that, according to the theory, no action that is nonobligatory can bevaluable in the requisite way, or that, according to the theory, no actionthat is valuable in the requisite way can be nonobligatory.

In addition, there may be good reason to deny the possibility of subero-gation even if that of supererogation is accepted. This is because there is acertain asymmetry between these concepts that might not at first be appar-ent.21 In particular, the greater the extent to which an act goes beyond duty,the more supererogatory it is; its status as optional is not thereby threatened.But the same cannot be said of suberogatory acts. That is, it is not the casethat the greater the extent to which an act falls short of decency, the moresuberogatory it is; on the contrary, its status as optional is thereby threat-

15 Cf. Mellema (1991), p. 29ff.16 Cf. Heyd (1982), p. 26ff.; Mellema (1991), Ch. 3.17 Pace Hill (1971).18 Cf. Feldman (1978), pp. 48-50; Heyd (1982), p. 76ff.; Mellema on Feldman's theory

in Mellema (1991), p. 79ff.; New (1974); Kagan (1989).19 See the discussion of (5.25) and (5.27) in Subsection 5.4.3.20 Cf. Heyd (1982), p. 73.21 See especially Heyd (1982), p. 128ff., and Mellema (1991), p. 185ff.

237

ened, in that at some point it presumably becomes wrong. And it may bethat this point is reached immediately, as it were, so that the status of subero-gation is never attained. It seems to me not accidental that the most plau-sible of the relatively few examples of allegedly suberogatory acts that existin the literature (including, if it is plausible, the example given earlier in thischapter of Steve's taunting Carl) are most readily understood as exampleswhere the agent has managed, in some morally repugnant way, not to vio-late an obligation- to. For instance, Chisholm and Ernest Sosa talk of minoracts of discourtesy (such as taking too long in a restaurant when others arewaiting) and of not-so-minor acts committed while competing withothers (such as advertising in a way calculated to drive a competitor out ofbusiness), and in such cases they explicitly say that the act is permissiblebecause the agent is acting "within his rights."22 Similarly, Julia Driver givesseveral illustrations of what she takes to be suberogatory actions (a singleperson takes a seat on a train, thereby preventing two other people fromsitting together, although another seat is available; a man refuses to donatea kidney to his brother, although the latter will die without it), and she tooexplicitly talks of the agent as acting "within his rights."23 But if the realmof obligation is not exhausted by what there is an obligation-to to do, thenthese examples lose some of their force. Perhaps we can agree that, typi-cally, one acts within one's (liberty-)rights, that is, that one violates no oneelse's (claim-)rights, when one is discourteous to another or "plays hard-ball" in competitive contexts or refuses to donate a bodily part, but this doesnot mean that we should agree that in such cases one does no overall wrong.And if one does do overall wrong in such cases, then clearly one's action isnot optional, and hence not suberogatory.

There would seem to be powerful reasons, therefore, for thinking thatsupererogation is impossible, and further, independent reasons for think-ing that suberogation is impossible. However, determination of whether ornot these moral phenomena are in fact impossible would require a detailedinvestigation of the theories to which I have just alluded (something that Ishall not attempt) as well as a detailed account of the pertinent senses of"valuable" and "disvaluable" that apply to supererogation and suberoga-tion (something that I have already said I shall not attempt). In the absenceof such investigation, it would seem premature to rule out the possibilityof supererogation and suberogation. Yet it may seem that this is just whatmy account of obligation does. For it is a maximizing account, and certainly

22 Chisholm and Sosa (1966), p. 326.23 Driver (1992), p. 287.

238

many philosophers appear to believe that maximization itself precludessupererogation.24 If this belief is correct, then, unless cogent arguments canbe given, independently of considerations about maximization, to supportsuch preclusion (as arguments were given in Chapter 7 against the possi-bility of basic moral dilemmas), such preclusion may be taken as a sign ofthe inadequacy of the proposed account of obligation. But is the beliefcorrect?

8.2.1 Supererogation and discretion

Supererogatory acts, it has been said, are at once both optional and espe-cially valuable. But a maximizing theory links value with obligation, andthereby hangs the problem. As Michael Slote puts it, when discussing con-sequentialism in particular: "[C]onsequentialism, precisely because it insiststhat one do the best one can, leaves no room for moral supererogation: bar-ring ties for first place, everything it morally permits it morally requires."25

And yet such ties are possible, and thus acts can be optional. So why say thatmaximization precludes supererogation altogether?

It might be replied that, while a maximizing theory does admittedlyimply that some acts can be optional, it doesn't allow for the possibility thatacts be fully optional, whereas supererogatory acts are not just optional butfully optional. An act is fully optional only if it is not just not obligatory butalso such that its performance does not constitute a way to fulfill someobligation; it is, as it were, wholly beyond the reaches of obligation. Nowit is true that a maximizing theory is hardly likely to classify an optional actas fully optional. Consider a case that has the oversimplified form depictedin Figure 8.1. Here the agent has four and only four alternative courses ofaction, the first two equally valuable, the others less so. In this case, acts Aand B are optional, whereas C is wrong. But neither A nor B is fully option-al, since there is still an obligation lurking not too far in the background,namely, the obligation to do either A or B (see (X) in Subsection 2.1.3).Thus, although the agent is not obligated to do A and is not obligated todo B, performance of either will fulfill an obligation.

24 Included in this number are some who accept and also some who reject a maximizingtheory. See, for example, the following, all of whom appear essentially to hold the opin-ion in question: Moore (1965), p. 78; Baier (1958), pp. 203-4; Smart (1973), p. 55; New(1974), p. 179ff.; Heyd (1982), p. 78; Schemer (1984), pp. 21-2; Feldman (1986), pp.48-50; Slote (1989), pp. 3, 171; Kagan (1989), pp. xi-xii, 1-2, 7-8, and elsewhere;Stocker(1990),p.311.

25 Slote (1989), p. 3.

239

Figure 8.1

But here two points are to be noted, one minor, the other not. Theminor point is this: a maximizing theory can in principle allow for the pos-sibility of fully optional acts, as is evident in the case depicted in Figure 8.2.The much more important point is this: it is not true that supererogatoryacts must be fully optional. Consider the case in Figure 8.1. Why is it thatit is obligatory to do either A or B? The answer is, of course, that any courseof action involving neither A nor B is wrong. In doing either A or B, theagent is thus fulfilling the obligation not to do something else. It cannot beaccepted that this alone renders both A and B nonsupererogatory. After all,it is surely possible that a supererogatory act have some alternative that is

Figure 8.2

240

wrong (even if not every alternative can be wrong — for then it would beobligatory), so that, if the supererogatory act is performed, that very per-formance, though of course not itself obligatory, will nonetheless consti-tute the fulfillment of an obligation, to wit, the obligation not to performthe alternative act.

Why is it, then, that maximizing theories have been thought not toaccommodate supererogation? Well, consider the classic maximizingtheory of act utilitarianism, which is traditionally seen to be an amalgama-tion of hedonism (the view that all and only states of pleasure are intrinsi-cally good and all and only states of pain are intrinsically bad) andconsequentialism (the view that one ought to produce the most favorablebalance of intrinsic goodness over intrinsic badness that one can). It is oftensaid that act utilitarianism rules out supererogatory action,26 but this is ques-tionable. What it certainly does do is imply that many of the acts that wewould normally call supererogatory are in fact not so, since it implies thatthese acts are either obligatory or wrong. But this is not the same as imply-ing that there are no supererogatory acts, let alone the same as implying thatthere can be no such acts. Indeed, the consequentialist aspect of act utilitar-ianism (which is what makes the latter a maximizing theory) rules in the pos-sibility of optional acts (as long as ties in sum intrinsic value are in principlepossible, and there seems to be no reason to deny this). This fact might sug-gest to some that it is act utilitarianism's hedonistic, rather than conse-quentialist, aspect that renders it inimical to supererogation, and thatanother version of consequentialism would accommodate supererogationmore readily.

But actually it is not that easy. If we turn to the nonhedonistic conse-quentialism of G. E. Moore, we find that Moore's theory, too, has a hardtime accounting for supererogation. While this theory (which allies conse-quentialism with a theory of value according to which not just states of plea-sure and pain but also other states — knowledge, moral qualities, beauty27 —have non-neutral intrinsic value) of course allows for the possibility of acts'being optional — as does any consequentialist or, more broadly, maximiz-ing theory - it seems not to square with our common intuitions concern-ing supererogation any better than act utilitarianism does. For example, justas act utilitarianism implies that many instances of self-sacrifice that wewould normally call supererogatory are in fact either obligatory or wrong,so too does Moore's theory imply this. If one can maximize intrinsic value

26 See, for example: Baier (1958), pp. 203-4; Smart (1973), p. 55; New (1974), p. 179ff.;Feldman (1978), pp. 49-50; McConnell (1980), p. 36ff.

27 Moore (1965), p. 102.

241

only by way of extreme self-sacrifice, then Moore's theory implies that onemust, one is obligated to, undertake such sacrifice.28 This observation appliesequally well to the neoconsequentialist view of Fred Feldman, couched interms of possible worlds rather than in terms of consequences,29 and it is thisview that inspired the account of obligation that I have proposed.30 Must wetherefore conclude that maximization does indeed preclude supererogation?

Not yet. It could be that the problem lies with (neo)consequentialismin particular rather than maximizing theories in general. Although(neo)consequentialism allows for acts' being optional, its focus on the pro-motion of intrinsic value neglects the fact that it is in the area of personalinterest that we commonly look for a moral theory to give the agent some"slack," some latitude as to how to proceed. What, then, if a maximizingtheory were to admit into the final determination of right and wrong cer-tain agent-relative values? What if, in the final computation of what is bestand what not, an agent's own projects and commitments were given aweight disproportionate to that which their intrinsic value alone supplies?This possibility has been explicitly admitted in the account of obligationthat I have proposed; recall that in this account S's being obligated at T toperform some act is analyzed in terms of the value for S at T of accessibleworlds in which he does and does not perform the act. Does this featureof the account provide the means of reconciling maximization withsupererogation?

Again, it is not that easy. Once more, there are two points to be made,one minor, the other not. The minor point is that it would be a mistake tothink that supererogation essentially concerns agent-relative values. Somepeople might claim, for instance, that, although it is obligatory to refrainfrom killing babies whenever possible, it is supererogatory to refrain fromkilling fetuses. Whether or not this is correct, it is certainly intelligible, andno tie to agent-relative values is, or need be, involved. The much moreimportant point is this: even though a maximizing theory that incorporatesagent-relative values in the way indicated allows, as do all maximizingtheories, for certain acts to be optional, it does not appear to allow for cer-tain acts to be discretionary in the way that supererogation requires. Considera case where A and B are each optional, according to the maximizing the-ory in question, and where their values (and for this reason the values of thetop-ranked worlds that include them) are, respectively, predominantlyagent-neutral and predominantly agent-relative. For example, A might be

28 Cf. Heyd (1982), pp. 77-8.29 Cf.Mellema(1991),p. 87.30 See Ch. 2, n. 7.

242

a charitable act of some sort (in which the agent has no vested personalinterest), while B is the purchase of some high-priced compact-disk play-er (in which the agent has considerable vested personal interest). Now con-sider a very slightly modified case, where B's value marginally overrides A's(and for this reason no accessible v4-world is quite as valuable as any acces-sible B-world), because the agent's vested personal interest is marginallyincreased. The maximizing theory would imply that in this modified caseneither A nor B is optional; on the contrary, A is wrong and B obligatory!Similarly, if 4's value were marginally to override B's, then the maximiz-ing theory would once again imply that neither A nor B is optional; on thecontrary, A would be obligatory and B wrong. In such cases, engaging inthe sort of self-sacrifice that A would entail is clearly not left to the agent'sdiscretion.31

It is probably at this point that many will diagnose a fatal flaw in all max-imizing theories. It is not, as it may at first have seemed, that such theoriescannot admit that certain acts are optional; clearly they can. It is rather thatsuch theories cannot admit that certain acts are discretionary, in the sensethat it is left to the agent's discretion whether or not to perform the act whenits performance and nonperformance are of different values. This, it mightbe claimed, is something that is essential to supererogation, but somethingthat maximizing theories cannot accommodate.

Somewhat more precisely, and applied specifically to the account ofobligation that I have proposed, the argument is this:

(8.1) necessarily, for any act A, if the performance of A is supererogatoryand its nonperformance merely optional, then all accessible ~A-worlds are less valuable than some accessible /1-world;

(8.2) if my account of obligation is correct, then, necessarily, for any actA, if all accessible ~/l-worlds are less valuable than some accessible^4-world, then the performance of A is obligatory;33

(8.3) necessarily, for any act A, if the performance of A is obligatory, thenit is not supererogatory;

hence,

(8.4) if my account of obligation is correct, then, necessarily, there is noact A such that the performance of A is supererogatory and its non-performance merely optional.

31 Cf. Bennett (1989), pp. 56-7.32 Cf. Raz (1975), p. 164. Cf. also Stocker (1990), p. 311. See also Heyd (1982), pp. 167-8.33 This is a bit rough. It ignores subclause (c4) of (I), which is rendered necessary by con-

siderations laid out in Subsection 2.2.6.

243

An analogous argument can be made concerning suberogation:

(8.5) necessarily, for any act A, if the performance of A is suberogatory andits nonperformance merely optional, then all accessible 4-worlds areless valuable than some accessible ~;4-world;

(8.6) if my account of obligation is correct, then, necessarily, for any actA, if all accessible ^4-worlds are less valuable than some accessible~y4-world, then the performance of A is wrong;

(8.7) necessarily, for any act A, if the performance of A is wrong, then itis not suberogatory;

hence,

(8.8) if my account of obligation is correct, then, necessarily, there is noact A such that the performance of A is suberogatory and its non-performance merely optional.

8.2.2 The division of values

There is but one possible way for me to resist the conclusions of these argu-ments while hanging on to maximization. Like any maximizer, I mustendeavor both to have my cake and to eat it - that is, endeavor both to saythat, where A is supererogatory, its performance is superior to its nonper-formance and, where A is suberogatory, its performance is inferior to itsnonperformance, and yet also to say that, since A is optional, its perfor-mance and its nonperformance are of equal value. This can be achieved by,and only by, declaring that there is more than one set of values pertinent tothe moral evaluation of an act.34 One set of values must be said to be perti-nent to the determination of right and wrong and obligation, the other not.Where A is supererogatory, and so it is left to the agent's discretion whetheror not to perform it, a maximizing theory must declare its performance tobe nondeontically superior but deontically equivalent to its nonperfor-mance; and where A is suberogatory, a maximizing theory must declare itsperformance to be nondeontically inferior but deontically equivalent to itsnonperformance. In this way, the first two premises of each argument willbe seen to be true, but only because of equivocation on the term "less valu-able," and so, despite the evident truth of each of the third premises, the

34 Note: moral evaluation. It is in some morally relevant way that supererogatory acts aresuperior to their nonsupererogatory alternatives. Simon seems to miss this in Simon(forthcoming).

244

arguments are to be declared unsound and their conclusions therebyresisted.

This solution should come as no surprise. In previous chapters I stressedthat my concern, when giving my account of obligation, was with the deon-tic superiority of one action, or world, over another. This was intended toleave open the possibility that one act, or world, be superior, even moral-ly, to another in a nondeontic way.35 Let me now elaborate.

Suppose that consequentialists are correct and that 'what determineswhether or not an act is morally obligatory has to do with and only withhow it and its alternatives actually promote intrinsic value. In this case,deontic superiority consists in superiority concerning such promotion.Nonetheless, it is surely possible that a consequentialist should accept thatacts are open to a sort of moral evaluation that has nothing to do with theirdeontic status. For example, some consequentialist might say that an act ismorally meritorious if and only if its agent intends by means of it to achievewhat is intrinsically best. Since intended results need not match actualresults, it would follow that an act can be morally obligatory without beingmorally meritorious, and vice versa. A morally meritorious act would besuperior in a morally relevant way to a nonmeritorious one, but this supe-riority would be nondeontic.

Note that, on this consequentialist view, there is a common core to anact's deontic and its nondeontic moral status. Both have to do with the pro-motion of intrinsic value. What distinguishes them is that the one has todo with the actual promotion of such value, the other with its intendedpromotion.

A nonconsequentialist can similarly distinguish between an act's deon-tic and its nondeontic moral status. For example, someone might say thatwhether or not an act is morally obligatory has to do with and only withwhether it actually respects people's rights. In this case, deontic superiori-ty consists in superiority concerning such respect. Nonetheless, it is surelypossible that a nonconsequentialist of this sort should accept that an act ismorally meritorious if and only if its agent intends by means of it to respectpeople's rights. Since, as before, intended results need not match actualresults, it would follow that an act can be morally obligatory without beingmorally meritorious, and vice versa. Also, as before, there would in this casebe a common core to an act's deontic and its nondeontic moral status.

There would seem, then, to be good reason to accept the possibility thatan act be subject to a variety of moral evaluations, only some of which have

35 See note 13 to Ch. 2.

245

to do with its deontic status. A maximizing theory of obligation can exploitthis possibility in order to accommodate supererogation. Unless there arefurther restrictions on supererogation that I have so far overlooked (an issueto which I shall return in the next subsection), I therefore conclude thatmaximization does not preclude supererogation.

Even if this conclusion is accepted, though, it may be thought that itcomes at too high a price; for some may find such a division of values to bea deus ex machina with no rationale other than to save maximization. Butthis is not so. First, it seems to me that appeal to such division is routine.When someone is told that he should or ought to do some act, although hedoesn't have to,36 what's apparently meant is that it is valuable in a way thatdoesn't render it obligatory. (Thus "ought" is here used in a nonbindingsense. See Section 1.2.) Why should this common appeal to a double stan-dard37 be deemed unavailable to a maximizing theory? The problem withmost maximizing theories is that they don't make such an appeal, not thatthey can't, and this is why they cannot (as they stand) account forsupererogation. Second, and more importantly, there is no such thing as anonmaximizing theory of obligation, as I pointed out in Chapters 1 and 2,unless such a theory does not imply that an obligatory act is "unique inrespect of value" (as Moore puts it)38 - that is, it does not in some way rankan obligatory act higher than its nonobligatory alternatives - in which caseI would say that it completely misconstrues the notion of obligation.39 If so,then the strategy of dividing values is the strategy that any theory of obliga-tion must adopt if it is to admit supererogation and suberogation.

The argument underlying all this is very simple. It is as follows:

(8.9) if the performance of A is not deontically equivalent to its nonper-formance, then one is obligatory and the other wrong;

(8.10) if one is obligatory and the other wrong, then neither is optional;(8.11) if neither is optional, then neither is supererogatory or suberoga-

tory;

hence,

(8.12) if either the performance of A or its nonperformance is eithersupererogatory or suberogatory, then they are deontically equi-valent;

36 Cf. Feinberg (1968), pp. 392-3. Cf. also Chisholm (1968), p. 417.37 Cf. Attfield (1979). Cf. also Raz (1975).38 See note 37 to Ch. 1.39 This is not to say that every theory of obligation exactly fits the form of (I), as I acknowl-

edged in Subsection 2.1.2.

246

hence,

(8.13) any superiority (due to being supererogatory) or inferiority (due tobeing suberogatory) of the performance of A to its nonperfor-mance, or vice versa, is nondeontic.

And so, if the strategy of dividing values comes at too high a price, thensupererogation and suberogation must simply be declared impossible.

But is the price too high? I don't see why. Surely it is plausible to saythat there is more to morality than the deontic. An act, or its agent, may bepraiseworthy or blameworthy, virtuous or vicious, and so on,40 and only ifsuch matters are directly reducible to the act's being obligatory, permissi-ble, or wrong - which is surely not the case - would the act's deontic sta-tus exhaust its moral status. If this is so, then supererogation andsuberogation can involve nondeontic matters such as these. The questionis which.

Well, this is the question that I said I would not try to resolve: precise-ly what senses of "valuable" and "disvaluable" are at issue in supereroga-tion and suberogation. Many different answers have been proposed.Almost all accounts of supererogation claim that, if an act is supererogato-ry, then either it or its agent is in some way praiseworthy, but just what goesinto praiseworthiness is controversial. Some claim that, if an act or agent isto be praiseworthy, the act must be unusually valuable, and hence that thereare no trifling acts of supererogation;41 others allow for trifling acts ofsupererogation that are, because trifling, only minimally praiseworthy.42

Some philosophers claim that, for an act to be supererogatory, its omissionmust be such that either it or its agent is not blameworthy.43 Some philoso-phers claim that a supererogatory act, even though it is not obligatory, issuch that it ought (in a nonbinding sense) to be done.44 Some philosophers

40 I would prefer to say that agents, rather than acts, are praiseworthy or blameworthy, vir-tuous or vicious, and so on, but this might seem to put some strain on my contentionthat there is a variety of ways to evaluate acts, morally. Recall, however, from Section 1.5that, even when it comes to obligation, I am in fact not at all sure that it is best to cate-gorize the evaluation in question as drt-evaluation. Perhaps it is better categorized as amatter of a^enf-evaluation, in which case the strategy of dividing values is under no strainat all.

41 E.g., Feinberg (1968), p. 400.42 E.g., Chisholm (1968), p. 417, although he resists talk of praiseworthiness as such; Heyd

(1982), p. 127; Montague (1989), p. 107; Mellema (1991), p. 111.43 E.g., Chisholm (1968), p. 424, although he resists talk of blameworthiness as such; Heyd

(1982), p. 115; Mellema (1991), p. 5. If this condition is not met, Mellema says that theact can at best be "quasi-supererogatory." See Mellema (1991), Ch. 5.

44 E.g., Feinberg (1968), p. 392; Chisholm (1968), p. 417; Forrester (1975), pp. 225-6.

247

say that a supererogatory act, though not obligatory, is nonetheless "rec-ommended," in a quasi-obligatory way, by morality.45 Some philosophersclaim that an act is supererogatory only if it has good consequences, or isintended to have good consequences.46 Some claim that an act issupererogatory only if it is performed for someone else's sake. And so on.A similar variety of conditions may be proposed for suberogation. The mat-ter is clearly both complex and controversial.

It is perhaps easier to envisage a division of morally relevant values whenthe nondeontic values are not said to supply a reason for acting, harderwhen they are said to supply such a reason. Suppose we agree that an act issupererogatory only if it or its agent is praiseworthy. On some accounts ofpraiseworthiness, an act or agent is rendered praiseworthy by the motiveunderlying the act; on some accounts of motivation, the motive underly-ing an act is beyond the agent's control; and, on some accounts of reasons,what is beyond the agent's control cannot feature in a reason to perform anact. Thus some accounts can declare the performance of a supererogatoryact to be wholly on a par with its nonperformance with respect to reasonsfor action; the nondeontic superiority of the performance will be explainedwithout appeal to reasons for action.

On the other hand, some accounts of supererogation will declare thesuperiority of the performance of a supererogatory act to constitute a spe-cial reason for performing it. If this is essential to supererogation, then, ifsupererogation is possible, there must be not only two sets of morally rel-evant values but also two sets of moral reasons for acting. But, while it maybe difficult to see just how such an account should go, it is clearly not with-out precedent. On the contrary, the common invocation of what is idealrather than obligatory seems, often at least, to constitute an appeal to a non-deontic set of reasons for acting.

8.2.3 Continuity

It may seem that something has gone wrong in the foregoing defense of thecompatibility of maximization with supererogation and suberogation, inthat I have overlooked a condition of supererogation and suberogationwhich blocks the strategy of dividing values. This is the condition that

45 E.g., Velazco y Trianosky (forthcoming).46 E.g., Heyd(1982),p. 115.47 E.g., Heyd(1982),p. 115.48 Cf. Heyd (1982), p. 181.

248

David Heyd calls the condition of "continuity"49 and which GregoryMellema characterizes as follows: "There is a common and continuous scaleof value shared by supererogation and duty, and the difference betweensupererogation and duty is reflected in the degree to which the value is real-ized in each. The value is realized to a greater degree in supererogation thanin obligation."50 How can the strategy of dividing values be reconciled withthis talk of a common and continuous scale of value?

Whether or not such reconciliation is possible depends on just what isinvolved in this condition of continuity that (allegedly) applies tosupererogation. It may seem that it is simply being claimed that the valueexhibited by supererogatory action, like that exhibited by obligatoryaction, is a morally relevant value. Clearly I agree with this. But this can-not be all that is meant; for, although this gives sense to the talk of therebeing a common scale of value, it does not give sense to the talk of there beinga continuous scale of value.

It may seem that it is simply being claimed that all supererogationinvolves the sort of "oversubscription" (to use Feinberg's term) that is dis-played by John when he works a supererogatory ten hours instead of themerely optional eight, for clearly there is a sort of continuity operative here.But this, too, cannot be what is meant, for two reasons. First, like Feinberg,both Heyd and Mellema acknowledge that there are cases of supereroga-tion that do not involve such oversubscription, as when, for instance, onegives a stranger a match.51 Second, the sort of continuity that cases ofoversubscription display and cases of nonoversubscription do not concerns,not a type of value, but a type of action (in John's case, the type of actionbeing that of working).

Perhaps what is being claimed is that, if A is supererogatory and somealternative B merely optional, then both A and B must exhibit a commontype of morally relevant value, A exhibiting more of this value than B. Thisis compatible with what I have said (although I doubt that it is true); foreven if B is merely optional, it may nonetheless exhibit a certain nondeon-tic value, just not as much of this value as is needed to render it, like A,supererogatory. For example, if no act of supererogation can be trifling, assome claim, then a trifling favor will not be supererogatory; nonetheless, asa favor it may be good in a certain way that is irrelevant to its deontic sta-tus and thus share a type of nondeontic value with its less trifling,supererogatory alternatives.

49 Heyd (1982), p. 5.50 Mellema (1991), p. 26.51 Feinberg (1968), p. 397; Heyd (1982), p. 135; Mellema (1991), p. 27.

249

Perhaps, though, what is being claimed is that, if A is supererogatory andB merely optional, then both A and B owe their respective statuses whollyto a certain single type of morally relevant value that they both exhibit, theirstatuses being different solely due to the fact that the former exhibits moreof this value than the latter. Given (8.12) and (8.13), I cannot accept this;but I know of no reason to accept it.

Now, some may respond to this dismissal of the condition of continu-ity (understood according to the latest interpretation) as follows: certainaccounts of supererogation clearly include it, and thus, even if the condi-tion hasn't been established as necessary for supererogation, still it is per-fectly intelligible, and hence something must have gone wrong with theargument for (8.12) and (8.13).

Perhaps the sort of account of obligation and supererogation that mightbe thought most obviously to include the condition of continuity (under-stood according to the latest interpretation) is a "satisficing" account. Slote,who is favorably disposed towards such an account, says the following:

From the standpoint of common-sense it can be acceptable, permissible, to beless than optimifically benevolent or beneficent, to give others less than one isin a position to give them, so long as one is sufficiently beneficent, that is, givesothers a good deal or enough.52

If one does well enough, one's act is "satisficing," even if not maximizing,with respect to the good that one can achieve; and an act need only be sat-isficing to avoid being wrong. Then, with respect to supererogation, Slotesays this:

[S]atisficing forms of.. .consequentialism.. .are consequentialistic because theytreat overall consequences as the only thing relevant to act-evaluation andbecause, more particularly, they regard any act with overall consequences lessgood than those of some alternative as morally less good than that alternative.But they allow for supererogation and for the moral permissibility of less thanoptimal behavior.53

Note: "the only thing relevant." Even if such a satisficing theory is to berejected for some reason, doesn't its form demonstrate that a theory can sat-isfy the condition of continuity (understood according to the latest inter-pretation), and hence that my contention, that supererogation mwsdnvolvea division of values, is false?

Well, recall that my contention that supererogation requires a divisionof values rests on the dual claims that maximizing theories of obligation can

52 Slote (1989), p. 26, italics deleted.53 Slote (1989), p. 142.

250

accommodate supererogation only in this way and that there is no suchthing as a nonmaximizing theory of obligation. Perhaps the former claimwill be granted, but what about the latter? What, in particular, about satis-ficing theories of obligation?

There is of course a tension between the claim that obligation involvesmaximization and the claim that it involves satisficing; but it is only a sur-face tension. To resolve it, one need only recognize that I am saying thatobligation involves maximization of some sort, and this is clearly compati-ble with its involving satisficing ofanother sort. In particular, if two acts, oneof which optimizes a certain value and the other of which merely satisficesthat value, are both equally permissible, then clearly they are to be ranked asdeontic equals; their deontic status is not determined solely by the degree towhich they realize the value in question. For instance, a theory might saythat an act is permissible if and only if it displays beneficence to at leastdegree 5, supererogatory if and only if it displays beneficence to a degreegreater than 5. This is intelligible, but what it implies is that it is not simplybeneficence that is pertinent to the determination of right, wrong, andobligation. Rather, it is beneficence-to-at-least-degree-5 that is at issue;any acts that display this characteristic are deontically equivalent, whetheror not one displays a greater degree of beneficence than another. And it isnot beneficence-to-at-least-degree-5 that determines whether and to whatextent an act goes beyond duty; on the contrary, it is beneficence-to-a-degree-greater-than-5. Thus the condition of continuity (understoodaccording to the latest interpretation) is not satisfied by this theory. Despitethere being a common core (namely, beneficence) to that characteristicwhich determines an act's deontic status and that which determines its non-deontic status, the characteristics are distinct.

(One would, of course, like a rationale for making the cutoff pointbetween mere optionality and supererogation be beneficence to degree 5.One such rationale might be that other people's needs are satisfied whenand only when they are treated with beneficence to at least degree 5, andthat one is obligated to satisfy other people's needs but not obligated to domore than this. Another rationale might give a different story. Once therationale, whatever it is, is given, it will be easier, I think, to see that on anytheory that accommodates supererogation, even a satisficing theory, thereis indeed more than one consideration relevant to the evaluation of an act,contrary to what Slote says. For instance, if the rationale were the one justillustrated, then an act's deontic value would turn on whether it meets otherpeople's needs, whereas its nondeontic value would turn on something elseentirely.)

251

Perhaps this will have served to show how modest the strategy of divid-ing values really is. Drawing a distinction between deontic and nondeon-tic values merely reflects the fact that supererogatory acts and their merelyoptional alternatives are both alike (with respect to optionality) and unalike(with respect to being beyond duty), and, as such, it seems to me whollybeyond reproach.

It should be noted, finally, that I have not been arguing that supereroga-tion and suberogation are possible. For all that I have said, they may yet beimpossible. Certainly not every theory of obligation can, as it stands,accommodate them. Those theories that deny any division of moral valuescannot, and others that do allow for such division may nonetheless under-mine themselves. For example, any theory that says that something's beinggood from the ideal point of view is itself a deontic reason for trying toachieve it would appear to have undercut any advantage that this divisionof values initially afforded it with respect to the accommodation ofsupererogation. Ross's theory seems to have just this difficulty. In distanc-ing himself from consequentialism by divorcing the right from the good,Ross may at first seem to have proposed a theory that can account forsupererogation in a way that Moore's theory (unaccompanied by any divi-sion of values) cannot. For the special prima facie duties of which Rossspeaks that vie for the status of duty proper do not owe their deontic statusto the fact (if it is a fact) that the acts in question, or what they bring about,are in themselves desirable. Their deontic ground lies elsewhere, and thiswould appear to allow Ross to declare that the desirability of an act, or ofwhat the act brings about, provides at most a nondeoritic reason for its per-formance. But he does not say this. Rather, he declares the desirability ofan act, or of what it brings about, itself to be a deontic reason for perform-ing the act; there is, he says, a general duty to maximize intrinsic value.54

Thus, as long as it is always possible for an agent to promote intrinsic value,Ross seems no better able than Moore to accommodate supererogation.55

Even if a proponent of supererogation avoids such pitfalls, however,there may be other reasons to find fault with what he says. Perhaps an act'sbeing supererogatory does (or would) constitute a moral reason to performit, and perhaps no such reason can be nondeontic after all.56 Or perhaps,once it is clearly recognized that what is especially valuable aboutsupererogation is nondeontic, any inclination to say that what is (appar-ently) supererogatory is in fact neither obligatory nor wrong is to that extent

54 Ross (1930), p. 27.55 Contrast this assessment of Ross's theory with that given in Mellema (1991), pp. 72-3.56 Cf.Kagan(1989),p.381.

252

undermined. (For example, some people may be moved to call an actsupererogatory because they wish to ensure that the agent is not blamed forits nonperformance. But if an agent can avoid blame and yet do wrong,then there may be reason to doubt that no "supererogatory" act is morallywrong.)57 Whether or not any of this is so, I shall not seek here to deter-mine. My aim has been merely to argue that maximization does not, assuch, preclude either supererogation or suberogation.

57 Cf. Moore (1965), pp. 79-80, with respect to consequentialism. Cf. also Baron (1987),p. 262, with respect to Kant's theory.

253

Cooperation

You and I never act in isolation. Like it or not, we are all members ofgroups. Groups can achieve things that individuals alone cannot. This is anobvious fact that has so far been neglected in this book, but it is a fact thatposes a serious problem for the account of obligation that I have proposed.

9.1 THE PROBLEM

The account that I have proposed is presented in Chapter 2. Its main claimis captured in (I). For the time being, let us work with the following sim-pler paraphrase of (I):

(9.1) for any person 5 and times T and T", S ought at T to do at T" thebest that he can at T do at T'.

The problem that besets this account may be illustrated as follows.1 Supposethat there are two voters, Vincent and Virgil, and two candidates for elec-tion, Smith and Jones, and suppose that the voters' options may be rankedaccording to the matrix given in Chart 9.1. That is, it would be (deonti-cally) best if both Vincent and Virgil voted for Smith; next best if one votedfor Smith and the other abstained; next best if both voted for Jones; nextbest if one voted for Jones and the other abstained; and worst if either bothabstained or one voted for Smith and the other for Jones. The question is:given (9.1), what should Vincent and Virgil do?

Before we attempt to answer this question, we should make note ofwhat the illustration presupposes. In particular, we must recognize that thenumbers represent the values of certain outcomes achievable by the groupof Vincent and Virgil, that is, achievable by them, jointly. It is thus beingpresupposed that there are certain worlds accessible to this group, and also

1 The problem has received extensive treatment in Regan (1980). It is anticipated inGibbard (1965), Sobel (1968), and Barnes (1971). My illustration borrows from Regan(1980), p. 18, and Feldman (1986), p. 155.

254

VincentVote for SmithVote for JonesAbstain

Vote forSmith

1007

VirgilVote forJones

053

Abstain730

Chart 9.1

that these worlds may be ranked according to their relative values. Let uslook at each presupposition in turn.

There can be no doubt that groups can sometimes achieve things thatsubgroups or individuals cannot. Perhaps a group of four people can lift aheavy piece of furniture but no fewer than four can do so. In general, then,we must reject the claim that, whatever a group can do, some proper sub-group or individual member can do also.

What about the converse? Should we accept or reject the claim that,whatever a proper subgroup or individual member of a group can do, thegroup itself can do? It may seem that we should reject this too. Perhaps one-quarter of the assembled company can squeeze into the back of aVolkswagen, but the whole group cannot; perhaps Paul can marry Paula,whereas the group of Peter and Paul cannot; and so on. But these factsshould not blind us to the truth that whatever is accessible to a proper sub-group or individual member of a group is also accessible to the group. Ifone-quarter of the assembled company can squeeze into the car, then cer-tainly the group as a whole can so act that one-quarter of its numbersqueezes into it; if Paul can marry Paula, then certainly the group of Peterand Paul can so act that Paul marries Paula.

In general, then, we should reject the first and accept the second of thefollowing two claims (where "{S^,..., SH}" signifies the group of personsSj through SM, and where an individual member S( of a group is, for the sakeof convenience, treated as the subgroup {5,}):

(9.2) if Wf is accessible to {Su ..., Sn} from W at T, then W is accessibleto some proper subgroup, if any, of {Su ..., Sn} from W at T;

(9.3) if W' is accessible to some proper subgroup of {Su ..., Sn} from Wat T, then W* is accessible to {Sl9 ..., Sn} from W at T.3

It is the falsity of (9.2) that underlies the problem with which I am present-ly concerned.

2 This illustration is borrowed from Humberstone (1983), p. 29.3 Cf. Humberstone (1983), p. 29; Feldman (1986), pp. 158-9.

255

The second presupposition of the illustration is that worlds that areaccessible to a group have values (and may be ranked accordingly); that is,they have deontic values relative to the group. Some philosophers will rejectthis presupposition. If a group (as opposed to individuals in the group) can-not comply with the Categorical Imperative, then presumably Kant wouldreject the presupposition, since according to him deontic value relative toan agent is rooted entirely in that agent's ability to comply with theCategorical Imperative. Similarly, if a group cannot display gratitude, thenRoss would presumably find the presupposition suspect, at least in thisregard. But other philosophers would not hesitate to accept the presuppo-sition. Consequentialists, for example, will accept that worlds that areaccessible to groups have, like all worlds, an intrinsic value. And Ross, inso-far as he accepts that intrinsic value is relevant to the determination ofobligation and hence counts as a source of deontic value, will also acceptthe presupposition, at least in this regard.

There is a twist to this second presupposition. If Chart 9.1 is to be usedin the determination of Vincent's and Virgil's individual obligations, it mustbe that the numbers that serve to indicate the relative desirability of out-comes do so in an agent-neutral way. That is, the unanimous election ofSmith must be seen to be superior relative to all perspectives — to Vincent's,to Virgil's, and to {Vincent, Virgil}'s - and thus the agent-relativizationthat is built into (I) must be seen to be insignificant.

By conjoining (9.3) with the second presupposition, we may assert thefollowing:

(9.4) if a world of deontic value Vis accessible to some proper subgroupof { St, ..., Sn} from W at T, then a world of deontic value Fis acces-sible to {Su ..., Sn} from W at T.

That is, whatever deontic good (or bad) a subgroup can achieve, the groupcan achieve also. But, as with (9.2), the converse does not hold. That is, weshould reject this:

(9.5) if a world of deontic value Vis accessible to {S2, ..., Sn} from W atT, then a world of deontic value Vis accessible to some proper sub-group, if any, of {Su ...,£„} from H^at T.

The problem with (9.1) that I am about to expose has, therefore, onlylimited application; it applies only to those theories that satisfy the form of(I) and subscribe to both the presuppositions noted. This is not to say thattheories that avoid this problem are superior. First, if they avoid it by fail-ing to satisfy the form of (I), they are inferior; for (I), as I have argued

256

throughout, would (so far) appear to be the form of any acceptable theoryof obligation. Second, if they avoid it by avoiding the presuppositions inquestion, they face a problem of their own. The first presupposition is sure-ly undeniable, and denial of the second comes at considerable cost.4 Third,the problem is soluble, as I shall seek to show.

Having made note of these presuppositions, let us ask again: given (9.1),what should Vincent and Virgil do? We cannot yet say; more informationis needed. Can Vincent act in such a way that Virgil will vote for Smith? Ifso, (9.1) directs Vincent to act in that way and to vote for Smith too.Similarly, mutatis mutandis, for Virgil. Thus there is a scenario in which (9.1)is satisfied and in which the best possible outcome (ten units of deonticvalue) is achieved.5 This of course poses no problem for (9.1). Nor does thisscenario: Vincent can act in such a way that Virgil will vote for Smith (sothat, we may say, Virgil is not intransigent relative to Vincent concerningSmith's being unanimously elected), but Virgil cannot act in such a way thatVincent will vote for Smith (because Vincent is intransigent relative toVirgil concerning Smith's being unanimously elected; Vincent insists onvoting for Jones). Here (9.1) directs Vincent to act in such a way that Virgilwill vote for Smith and also directs Vincent himself to vote for Smith; butit directs Virgil to vote for Jones (since this is the best he can do, givenVincent's intransigence). It may at first seem odd that (9.1) should direct anagent to perform an action which, if performed, guarantees that less thanthe best possible outcome will be achieved (in the sense that ten units ofdeontic value will be forgone in favor of five), but on reflection this is sure-ly right. After all, it's not Virgil's fault that voting for Jones is the best hecan do, it's Vincent's; it is Vincent who does not do the best he can.

But now notice this. There is a sense in which (9.1) seems to allow twowrongs to make a right. Suppose that Virgil is just as intransigently in favorof Jones as Vincent is. Then not only does (9.1) direct Virgil to vote forJones (since this is the best that he can do under the circumstances), it directsVincent to do this too (since this is now also the best that he can do). Thus,given (9.1), neither Vincent nor Virgil does wrong in voting for Jones, eventhough this yields only five units of deontic value rather than ten; eventhough, that is, the best that both could do is not done. Somehow, Vincent'sintransigently voting for Jones is transformed from being wrong to beingright when it is balanced by a matching intransigence on Virgil's part. This

4 The cost in question is this: they are subject to analogues of the Prisoner's Dilemma andare thus "directly collectively self-defeating," as Parfit puts it. See Parfit (1984), pp. 54 and95ff. Cf. Regan (1983), p. 108ff.

5 Cf. Regan (1980), pp. 52-5.

257

implication of (9.1) should surely trouble anybody inclined to the view thatone ought to do the best one can, inasmuch as it demonstrates that there isa sense in which universal satisfaction of (9.1) is compatible with the bestthat can be done not being done.

9.2 NONSOLUTIONS

9.2.1 Unrequited cooperation

It may seem that the problem raised in the last section has brought to lighta limitation to (9.1) that is easily remedied. (9.1) was drawn up with theindividual alone in mind; it ignored the possibility of cooperation. Theproblem brings this out starkly: through a mutual failure to cooperate,Vincent and Virgil, though doing the best they can individually,6 fail to dothe best they can jointly. To ensure that they do the best they can jointly,each need only do his part in the mutually cooperative action open to both.Each, that is, need only vote for Smith and the best possible outcome willbe assured.

This suggests the following view:

(9.6) for any person S and times T and T\ S ought at T to do at T" whathe would do at T" if his group did at T ' the best that it could at Tdo at T ' (i.e., S ought to "do his part").7

Apart from the fact that (9.6) implies that the best is done if everyone actsaccording to it, it has what may seem to be another attractive feature.Notwithstanding the discussion of supererogation in Chapter 8, it mightstill be thought that theories of the sort captured by (9.1) are too demand-ing, that they require of each of us that we exhaust ourselves in the pursuitof the good (for example, by giving and giving to charity), since manyothers will in fact not pursue the good and we cannot get them to do so(that is, they are intransigent relative to us concerning the pursuit of thegood) — although, if everyone were to pursue the good, none of us wouldhave to extend himself all that much (if everyone were to give to charity,none of us would have to give very much). (9.6), on the other hand, seems

6 Here and henceforth, when I say that two (or more) people do the best they can individ-ually, I mean simply that each does the best he can in fact do under the circumstances. Idon't mean that each does the best he can do "all on his own," i.e., exclusive of any coop-eration from others. And so, if full cooperation is in fact forthcoming, the best that two(or more) people can do individually will in fact be to do whatever it is that they woulddo if they were to do the best they could jointly. At least, this is so given (9.4) above.

7 Cf. Regan (1980), p. 85, on what Regan calls "COP."

258

to imply that we each ought to pursue the good only to that extent whichwould be required of us under the counterfactual condition where every-one does his part. And this might appear to be just what should be said.Why, after all, should one person be obligated to make up for the moralfailures of another?

But this won't do. (9.6) faces a number of problems.First, a relatively minor problem: (9.6) overlooks the fact that a person

may belong to several groups at once.8

Second, a far more serious problem: (9.6) violates the thesis that "ought"implies "can." This is illustrated by any case in which one person's doinghis part depends on another's doing his, and where the latter is intransigentrelative to the former concerning the group's generation of the best possi-ble outcome. For example, my part in generating the best possible outcomemay be to enter a room whose door only you can unlock. If you adamant-ly refuse to do your part, then I cannot do mine.

Third, whether or not the discussion of supererogation in Chapter 8 isacknowledged to have dealt successfully with the apparent overdemand-ingness of (9.1), certainly (9.6) doesn't deal with it. The claim, unqualified,that no one should be said to be obligated to make up for the moral failuresof another is surely far too sweeping. Just consider a case where all that isneeded to prevent a child from dying is the donation of one dollar. Thereare one hundred people present who can each give one cent; you can giveone dollar without any hardship whatsoever; none of the others givesmoney to the child; and you give one cent, smugly assuring yourself that,in doing your part, you've done your duty. The child dies, of course.Similarly, with reference to the voting case, surely it is in (9.1)'s favor thatit directs Virgil to vote for Jones, given Vincent's intransigence; for had hevoted for Smith, zero rather than five units of deontic value would havebeen achieved. Such unrequited cooperation does no good at all.10

9.2.2 Procedural cooperation

What it seems we want, then, is a theory that will direct Virgil to vote forJones, given Vincent's intransigence, but which will not permit a similarintransigence on Virgil's part. Donald Regan has produced a very elabo-

8 Cf. Feldman (1986), p. 153.9 Cf. Feldman (1986), p. 165.

10 Cf. Humberstone (1983), p. 29. "Cooperation" ought perhaps to be put in quotes here,since it must fail. (Cf. Regan, 1980, p. 126ff., on what goes into complete cooperation.)

259

rate theory designed to meet this goal. Cutting through the elaboration, wemay give this rough version of Regan's proposal:

(9.7) for any person 5 and times T and T", S ought at T to:(a) hold himself ready at T (and up to T') to cooperate with others at T';(b) identify prior to T ' other would-be cooperators at T ' ; and(c) do his part at T ' in the best action that the group of actual cooper-

ators can at T do at T'.n

This is an attractive proposal. If everyone follows the prescribed procedure,the best that can be done by all will be done; anyone not prepared to coop-erate does wrong; anyone prepared to cooperate but frustrated by othernoncooperators, who nonetheless does his best under these less than idealcircumstances, does what he ought. Doesn't this capture just what we wantto capture?

Unfortunately not. There are two principal objections to be made, basedon the fact that the proposal prescribes a certain procedure.12 First, Regan'stheory violates the thesis that "ought" implies "can": S may be unable todo what (9.7) requires. (In particular, S may be unable to identify otherwould-be cooperators.) Second, Regan's theory ignores the possible deon-tic cost involved in undertaking the specified procedure.13 It could be thatidentifying other would-be cooperators is itself a very costly activity, andto overlook this is surely inconsistent with the view that one ought to dothe best one can.

9.2.3 Group obligation

A proposal made prior to Regan's has been resurrected by several others inresponse to the difficulties that Regan's faces.14 It is, in effect, that (9.1) begeneralized thus:

(9.8) for any persons St...Sn and times T and T", {Su ..., Sn} ought at Tto do at T ' the best that {Si9 ..., Sn} can at T do at T'.15

11 Regan (1980), pp. 135-6.12 See Barley (1984), pp. 157-9; Conee (1983), pp. 422-3; and Feldman (1986), pp. 171-

3. Regan responds to such criticism (in Regan, 1980, p. 169ff.), but the response seemsunsatisfactory.

13 Regan deliberately ignores this (Regan, 1980, p. 176ff.).14 Originally: Postow (1977), p. 51 (and see also McKinsey, 1981, p. 317). Later: Conee

(1983), pp. 422-3; Parfit (1984), p. 73; Feldman (1986), p. 161; Jackson (1987), p. lOOff.15 As noted in Subsection 2.1.3, "at T'" is supposed to cover "throughout T'," "during

T'," and so on; for it is clear that the individual components of a group action may betemporally scattered.

260

(Where n equals 1, (9.8) is equivalent to (9.1), insofar as we are treating eachindividual S{ as {S,}.) This proposal takes advantage of the presuppositionsthat give rise to the problem in the first place. If we acknowledge thatgroups can sometimes achieve outcomes that individuals alone cannot, andif we acknowledge that these outcomes are themselves of deontic value,then why not accept that groups can be the agents of right and wrong andcan thus have obligations of their own? If we do accept this and accept (9.8)as the correct account of it, we can then point to a certain lack of corre-spondence between individual and group obligation. For example, giventhe falsity of (9.5), we must reject the following:

(9.9) if{S? , . . . , Sn} ought at T to do A at T", then some proper subgroup,if any, of { Si9 ..., Sn} ought at T to do A at T.

For the best achievable by the group may not be achievable by any propersubgroup, as in the case of Vincent and Virgil. And we must also reject thefollowing for the same reason:

(9.10) if some proper subgroup of {Sl9 ..., Sn} ought at T to do A at T",then {Sl9 ..., Sn} ought at T to do A at T'.

In addition, while we must of course accept that, where there is full coop-eration - that is, everyone does his part in the best action performable by thegroup - then neither the group nor any subgroup does wrong, we shouldnot accept that, if all its members act as they ought, then neither the groupnor any subgroup does wrong. That is, we must accept the following:

(9.11) i f S j d o e s ^ a t T\ and...and Sn does An at Ttt, and if {Si9 ..., Sn}thereby does at T' the best that {Su ..., Sn} can at T do at T', thenit is not the case that {Sl9 ..., Sn} or any subgroup of {Su ..., Sn}goes wrong at or after T in so acting at T';

but, as the present case of Vincent and Virgil shows, we must reject the fol-lowing:

(9.12) if 5 ought at T to do At at T\ and.. .and Sn ought at T to do An atT'n9 and if St does At at T\ and.. .and Sn does An at T'n9 then it isnot the case that {Sl9 ..., Sn} or any subgroup of {Sh ..., Sn} goeswrong at or after T in so acting at T".

It is on the falsity of (9.12) in particular that proponents of (9.8) have wishedto capitalize in the attempt to solve the problem before us. Their sugges-tion is this:16 it is not that, where Vincent's intransigence is matched by

16 See Feldman (1986), p. 177; Jackson (1987), p. 100.

261

Virgil's, two wrongs have made a right. On the contrary, although (9.8) ofcourse yields the same story about obligation and wrongdoing in the indi-vidual case as (9.1) does — each of Vincent and Virgil ought to vote for Jonesand hence does no wrong in doing so — unlike (9.1), (9.8) still diagnoses awrongdoing in this case: Vincent and Virgil jointly do wrong in not elect-ing Smith. Thus we need not accept the disturbing and unsatisfactory con-tention that nothing goes wrong in the election of Jones.

This is an intriguing and elegant proposal, but I think it must be reject-ed. First, it is not clear that groups of more than one member can haveobligations. I have acknowledged that, if the problem with which we'reconcerned is genuine, then wre must grant the presupposition that groupscan achieve outcomes that have a certain deontic value. But this does notmean that we must accept that groups can have obligations. When I pre-sented (I) in Chapter 2,1 noted17 that it might be that we should say that anagent, to be the subject of obligations, must be a moral agent; if not, we mayhave to accept that nonhuman animals have obligations (for they, too, maybe able to achieve outcomes that have a certain deontic value). What isinvolved in an agent's being a moral agent is unclear, but arguably it musthave the capacity to conceive of its action in moral terms,18 and it seemssafe to say that no group with more than one member (as opposed to indi-viduals in the group) has this capacity. Furthermore, some have contend-ed that, for something to be obligatory, it must be something about whichdecisions can be made, and that joint outcomes achievable by groups withmore than one member (as opposed to the individual component outcomesachievable by individual members of the group) are not entities of thissort.19 If either of these points is correct, then, despite the fact (if it is a fact)that groups with more than one member can achieve an outcome that hasdeontic value and, often, has a degree of deontic value that no outcomeachievable by some proper subgroup has, they cannot be obligated toachieve this outcome.

My rejection of the proposal is not based on these points, however.Indeed, I am prepared to accept (9.8) as an interesting and fruitful exten-sion of (9.1) and to accept, therefore, that Vincent and Virgil jointly dowrong when they unanimously elect Jones. What I wish to deny is that thisis all that goes wrong in the election of Jones. The problem is simply this:on the present proposal, the group wrongdoing is not attributable to any ofthe members of the group; neither Vincent nor Virgil can be faulted for his

17 In note 9 to Ch. 2.18 Cf. Thomson (1990), p. 215.19 Cf. Smith (1986), p. 342; McConnell (1988a), p. 27.

262

intransigence.20 Suppose that Vincent and Virgil are friends. Virgil sees thatVincent is intransigently in favor of Jones; not wanting to have his frienddo wrong, Virgil decides to be intransigently in favor of Jones too, believ-ing that Vincent will thereby be absolved. (He resolves not to listen toentreaties even from Vincent himself; he believes that Vincent, being hisfriend but subscribing to an account of obligation other than (9.8), mayattempt to dissuade him from acting in this way.) Now, if Virgil acts in thisway, (9.8) - as Virgil recognizes - implies only that Vincent and Virgiljointly do wrong and that neither does wrong singly. But this is too muchto swallow; the scenario still smacks of two (individual) wrongs making aright.

The proponent of (9.8) might reply in this way. What's wrong withintransigence, when it's met with matching intransigence? Given Vincent'sintransigence, Virgil's intransigence makes no difference; Vincent's intran-sigence alone rules out attainment of the group's best. So too, mutatis mutan-dis, for Vincent's intransigence making no difference. Thus, under thecircumstances, there is nothing wrong with the attitude of either.21

But there is something wrong with this attitude. True, there is a sort ofoverdetermination here, such that each agent's being a party to achievingless than the best outcome in a way renders the other's being a party to itredundant. Nonetheless, each is a. party to it, by virtue of his intransigence;each contributes to the failure to do what is best; each does wrong. Neither(9.1) nor (9.8) appears adequate to this fact, and the problem remains.

9.2.4 Counterfactual cooperation

The solution, I believe, is in outline simply (and unsurprisingly) this: one'smoral obligation is to do the best one can while avoiding intransigence; thatis (to coin a term), one ought transigently to do the best one can. Doing thebest one can must be accompanied by the adoption of a certain attitude. Thisis in essence Regan's proposal shorn of its objectionable aspect: one mustdo the best one can (see clause (c) of (9.7)) while holding oneself ready tocooperate (clause (a)), but there is no obligation to undertake something

20 Here "faulting" someone means ascribing wrongdoing to that person. Thus "attribut-ing" a group wrongdoing to an individual involves finding that that individual has donewrong with respect to the group outcome in question. It is natural to use the phrase "Sis responsible for X" instead of the phrase "Xis attributable to S," but one should be waryof this. At least, one should not take wrongdoing itself to suffice for culpability. Cf.Zimmerman (1988a), pp. 41-3.

21 Cf. Feldman (1986), p. 177.

263

(identification of cooperators — clause (b)) that may prove either impossi-ble or overly costly.

But what is it to "hold oneself ready to cooperate"? One answer is this:to hold oneself ready (that is, to be prepared) to cooperate is simply to besuch that one would do one's part if all others did theirs. When conjoinedwith the injunction to do one's best, this yields the following thesis:

(9.13) for any person S and times T and T', S ought at T to:(a) do at T' the best that he can at T do at T'; and(b) be at T such that he would do his part at T' if all others did theirs.

One problem with this proposal is that it is too narrow. It overlooks thehypothetical case where not all others cooperate but where a sufficientnumber do so and their doing so would generate an outcome superior tothat generated by 5 if S were to do the best he could in the actual situation.Surely, if we wish to insist on transigence, we wish to have 5 prepared tocooperate in the generation of any superior outcome, not just the best pos-sible, so long as this outcome would in fact be the best available (given thebehavior of any noncooperators). Suppose, for example, that the best thatBob, Carol, and Ted can jointly do is to raise and push aside a tree that hasfallen on Alice. Suppose also that, if Bob refuses to cooperate, the best thatCarol and Ted can jointly do is to raise the tree so that Alice can crawl outfrom under it. Suppose finally that, if both Bob and Carol refuse to coop-erate, the best that Ted can do alone is go for help. What we want is forTed to be prepared to help raise and push aside the tree if both Bob andCarol are prepared to cooperate, and (9.13) gives us that. But what we alsowant, and what (9.13) does not give us, is for Ted to be prepared to helpraise the tree if Carol is prepared to cooperate but Bob is not.

One way to try to provide what's missing is to amend (9.13) as follows:

(9.13') for any person S and times T and T' , 5 ought at T to:(a) do at T' the best that he can at T do at T'; and(b) be at T such that he would do at T" the best that he could at T do

at T' no matter what the circumstances might be at T'.22

There are of course difficulties of interpretation here, but notice how(9.13') would deal with the problem afflicting (9.13). Suppose that neitherBob nor Carol is prepared to cooperate with Ted. Then (9.13') is designedto direct Ted to (a) go for help but also (b) be prepared to cooperate withboth Bob and Carol if they are prepared to cooperate (something that (9.13)

22 Cf. Schwartz (1985), p. 85, for a similar proposal concerning compliance with a policy.

264

also says) and to be prepared to cooperate just with Carol if Bob is not pre-pared to cooperate (something that (9.13) doesn't say).

Nonetheless (9.13') is problematic. The requirement of counterfactualcooperation stipulated in clause (b) is too strong. Suppose that Fred wereforce-fed some mind-altering drug. Would he do the best he could underthese circumstances? Quite conceivably not (especially if the drug alteredhis moral perception). Would this show that he was in fact unprepared tocooperate? It would not.

A second problem with (9.13') is that it appears not to conform with thethesis that "ought" implies "can." While it is of course always in one'spower to do the best one can, is it always in one's power to be such thatone would do the best one could if circumstances were different? Fredappears to provide a counterexample. Let us assume that, having ingestedthe drug, he would not (although of course he could) do the best that hecould do. There seems to be no reason to assume that whether or not thisis true of him is presently up to him.

9.2.5 Mere openness

One attractive suggestion that does conform with the thesis that "ought"implies "can" is to view transigence as openness, where this is understoodas follows:

(9.14) S is open at T to G concerningp at T ' iff(a) {S, G} can at T so act at T ' thatp is true; and(b) G can at T so act at T' that p is true.

(Here 5 is distinct from G, and G is a group of persons, possibly with justone member.) We might then say:

(9.15) for any person 5 and times T and T\ S ought at T to:(a) do at T' the best that he can at T do at T'; and(b) for any group of persons G and proposition p, be open at T to G

concerning p at T" iff {5, G} so acting at T ' that p is true is thebest that {5, G} can at T do at T'.

The reason why this conforms with the thesis that "ought" implies "can"is this. If {S, G} can so act thatp is true, then all that S has to do to see to itthat G can so act that p is true is do his part in the group action in question.Now it may be (as noted before in the discussion of you, me, and the lockeddoor) that S cannot do his part because G prevents him from doing so;nonetheless G needn't do this and so can act in such a way that S does his

265

part - as long, that is, as S doesn't prevent G from so acting. Each of 5 andG can, then, perform his or its respective part if the other does; otherwiseit would not be the case that they could jointly perform the action in ques-tion. Thus each of S and G can be open to the other concerning p; each ofthem can so act that the other can so act that p is true. If either S or Gcannot so act that p is true, it is because the other has avoidably preventedthis.

I think that we should accept that transigence is openness, as this isdefined in (9.14). (9.15) might therefore appear to give expression to thesolution to which I have alluded. Indeed, (9.15) is notably superior to anyof the other proposals discussed. For example, it avoids the severity of(9.13') in that it requires simply that others can so act that 5 cooperates andnot that S would cooperate no matter what the circumstances. Like (9.13'),though, (9.15) aims to satisfy, as (9.13) did not, the requirement that S'spreparedness to cooperate be general.

The fact is, however, that (9.15) faces serious problems; the claim thatone ought transigently to do the best one can is only roughly accurate.

First, (9.15) on occasion issues in a basic moral dilemma, namely, that 5remain transigent while performing an action that precludes such transi-gence. Suppose, for example, that it is time for Vincent to vote, that Virgilhas not yet voted, but that Virgil remains intransigently in favor of Jones.Then clause (a) of (9.15) requires that Vincent now vote for Jones. In doingthis, Vincent cannot of course remain transigent concerning Smith's unan-imous election; yet clause (b) requires that he still be transigent in thisrespect, since it is still possible for {Vincent, Virgil} to so act that Smith isunanimously elected, and this can be achieved only by Vincent's voting forSmith.

It may seem that the obvious solution to this first problem is simply tostipulate in clause (b) that T is earlier than T", to stipulate, that is, that tran-sigence concerning the best outcome achievable by the group is requiredonly up until the time at which 5 must act in order to do what is in fact thebest that he can do. But a second problem is that, even with this amend-ment, (9.15) would still on occasion issue in a basic moral dilemma. This isbecause it can sometimes happen that the only way in which 5 can be opento G concerning some group-achievable outcome is by doing his part inthe achievement, or what would be the achievement, of that outcome.Suppose, for example, that it is not yet time to vote, that both Vincent andVirgil are going to vote for Jones, and that each is closed (that is, not open)to the other concerning Smith's being unanimously elected; but supposenow that this mutual intransigence is due simply to the fact that Vincent

266

and Virgil live at different ends of the country and cannot interact. Here itis the mere fact that Vincent is going to vote for Jones that makes it the casethat he is closed to Virgil concerning Smith's being unanimously elected;so too, mutatis mutandis, for Virgil's being closed to Vincent. Their beingmutually intransigent in this regard is due, therefore, not to any ornerinesson anyone's part, but simply to their mutual disconnectedness. Theirinability to interact renders each inflexible to, unmanipulable by, the other;neither can affect the way the other votes, and hence, however the onevotes, it is perforce afait accompli for the other. Now, (9.15) condemns evensuch closedness, but it ought not to. For, under the circumstances, even ifT is earlier than T", the only way in which Vincent can at T be open toVirgil concerning Smith's being unanimously elected at T" is by actuallyvoting for Smith at T". Thus clause (b) of (9.15) implies that Vincent oughtat T to vote for Smith at T', even though, given the fact that Virgil willvote for Jones, clause (a) implies that Vincent ought at T not to vote forSmith at T" but to vote for Jones instead.

Perhaps this result should not surprise us. If Vincent and Virgil are dis-connected in this way, then, even though there is a sense in which they canjointly, or as a group, achieve Smith's unanimous election, such actionhardly seems paradigmatic of what we normally think of as joint, or group,action. It is certainly not concerted action, action that is undertaken by agroup of interacting individuals. For this reason, even if we accept (in thespirit of (9.8)) that Vincent and Virgil have, as a group, done wrong, therewould seem to be no reason to think of such wrongdoing as attributableeven in part to any member(s) of the group. If anything, it is "attributableto nature."23

A third problem with (9.15) is that it requires that S be open to anygroup's best group-achievable outcome. This is surely too sweeping; onceagain, it can issue in a basic dilemma. For it can happen that what S mustdo to remain open to one group's best achievable outcome conflicts withwhat he must do to remain open to another group's best achievable out-come. For example, it could happen that Bob's and Carol's and Ted's joint-ly doing their best involves Ted's taking ten paces to the right, while Carol'sand Ted's jointly doing their best involves Ted's taking ten paces to the left.

A final problem with (9.15) is that, like (9.7), it ignores the possible cost-liness of doing what it requires. Even openness can be costly, as when (per-haps - it depends on just what deontic value consists in) S's openness to G,having become public knowledge, prompts someone to perform a harm-ful act that would otherwise not have been performed.

23 This is an ironic use of "attributable." Nature can do no wrong.

267

9.3 THE S O L U T I O N : U N I N T R U S I V E T R A N S I G E N C E

All four problems with (9.15) can, I believe, be handled by a simple stipu-lation. The requirement that S be open to G must be conditional on suchopenness not preventing S from doing what is in fact the best that he cando under the circumstances. One ought transigently to do one's best, butthe transigence must be unintrusive.

I propose, therefore, this revision of (9.1) as a way of resolving the prob-lem concerning cooperation that besets that proposition.

(9.1/) for any person 5 and times T and T", S ought at T to:(a) do at T7 the best that he can at T do at T7; and(b) for any group of persons G and proposition p, be open at T to G

concerning p at T ' iff(1) {S, G} so acting at T ' thatp is true is the best that {5, G} can

at T do at T', and(2) for any act A' and time T " such that T 7/ is no earlier than T

and no later than T7, \£A' is the best that S can at T do at T",then S can at T both be open at T to G concerning p at T ' anddo ,4'at T "

Roughly: clause (a) requires that 5 do the best he can; and subclause (bl)requires that he do it transigently, while subclause (b2) requires that thistransigence be unintrusive.

Just as (9.1) only gives rough expression to (I), so (9.1') only gives roughexpression to what I take to be the proper revision of (I). But before I setout this revision more precisely, let me address certain issues. In particular,let me explain how it is that the requirement that transigence be unintru-sive is designed to solve the four difficulties that (9.15) faces, and then letme indicate certain other important features of (9.17).

In stipulating that transigence is required only when it is unintrusive,(9.1') clearly avoids the possibility that 5 can meet his requirement to dowhat is best only if he fails to meet his requirement to be transigent. Hencethe sort of basic dilemma outlined in the first problem with (9.15) cannotarise on (9.17). Nor can the second sort of basic dilemma outlined in thesecond problem. If the best that Vincent can in fact do is vote for Jones, andthe only way in which he can be transigent concerning Smith's unanimouselection is to vote for Smith, then he is not required to be transigent in thisrespect. Hence (9.17) requires transigence only where S is connected to G,that is, only where whether or not S does his part in contributing to thebest achievable by { S, G} is not of necessity zfait accompli for G. Where there

268

is connectedness, then, S can (although he might, of course, refuse to doso) leave it up to G whether or not S does his part, and it is in virtue of thisfact that S ought to be open to G concerning what is the best achievable by{ S, G}. As an illustration, consider Vincent and Virgil once again, but nowthink of them as mutually connected. Vincent's being connected to Virgilconcerning Smith's being unanimously elected consists in this: Vincent canact in such a way that it is left up to Virgil whether both of them vote forSmith. (How? This will of course depend on other details of the case. (9.1')itself includes no stipulation as to just how this might be achieved. PerhapsVincent can simply inform Virgil — truthfully — that he will vote as Virgilwishes.) That Vincent can act in this way is, of course, no guarantee that hewill. While disconnectedness ensures lack of influence, connectednessdoesn't ensure influence. If Vincent opts for intransigence concerningSmith's being unanimously elected, then, even though he is connected toVirgil in this respect, Virgil cannot influence Vincent in this regard, and soVirgil's access to the best that is jointly achievable is effectively obstructedby Vincent. This is Vincent's fault; because of his being connected to Virgil,his intransigence is not "attributable to nature."

As to the third problem with (9.15): the best that anyone can ever do(given, as (9.4) implies, that an individual can never achieve a good that isnot achievable by some group of which he is a member, and also, as the fal-sity of (9.5) implies, that a group can sometimes achieve a good that nomember can) is to do his part in the best group action that those who canin fact be moved to cooperate can perform. If S's holding himself ready tocooperate in another group action in fact conflicts with his holding himselfready to cooperate in this group action, then it conflicts with his doing whatis in fact best, and (9.17) therefore implies that he ought not to hold himselfready to cooperate in the other group action.

Finally, if S's holding himself ready to cooperate is unduly costly, thenit prevents his doing the best he can, and, if that is the case, then once again(9.1') implies that he ought not to hold himself ready to cooperate.

Let me now indicate certain other important features of (9.1').First, (9.17) of course conforms with the thesis that "ought" implies

"can." One can always do the best one can; one can (as noted in the dis-cussion of (9.15)) always be transigent; when one cannot achieve both,clause (b2) kicks in and says that the requirement of transigence is nullified.Still, it is important to note that (9.1') differs from (9.1) in this respect,namely, that, whereas the latter implies that it is only what one does thatcounts (with respect to acting as one ought), the former implies that howone does it counts as well. In this way, (9.1') appears to accord with Regan's

269

claim that a satisfactory solution to our problem must involve a theory thatis not "exclusively act-oriented," where (it seems) a theory is exclusivelyact-oriented just in case one's satisfying it depends solely on whether (andnot on how) one performs some act.24 Sometimes this difference between(9.1) and (9.1') will issue in a difference in what action is required. This isso in the case where at least one of Vincent and Virgil can both do his bestand be transigent concerning Smith's unanimous election. In this case,whereas (9.1) implies that both Vincent and Virgil do as they ought if theyboth intransigently vote for Jones, (9.1') implies that both do as they oughtonly if they both transigently vote for Smith. For, suppose that Vincent canboth do his best and be transigent concerning Smith's being unanimouslyelected. Then (9.1') implies that he ought to be transigent in this respect,and, if he is, then Virgil can so act that Smith is unanimously elected, andso Virgil would not do his best by voting for Jones; and the same holds,mutatis mutandis, for Virgil's being transigent in this respect.

Nonetheless, and this is the second important feature of (9.1'), if neitherVincent nor Virgil can both do his best and be transigent concerningSmith's being unanimously elected, then (9.1') does not differ from (9.1) inits implication that each of them ought to vote for Jones. This is so eventhough they would and could jointly do their best by unanimously electingSmith. Consider, for instance, the case where Vincent and Virgil are mutu-ally disconnected with respect to Smith's being unanimously elected. Herethe group does not do its best (and, if groups of more than one member canbe obligated, the group does wrong), but no individual does wrong. Buthere, I submit, this result is not anomalous, precisely because of the dis-connectedness. It is when transigence can be unintrusive that a group's fail-ure to achieve its best is attributable to some member's doing wrong;otherwise it is "attributable to nature" and must be accepted as such.

Thirdly, if everyone other than S does his part in producing the best thatis jointly achievable, then (9.1') certainly implies that S does no wrong ifhe does his part too, for he can do no better (given (9.4)). But does (9.1')imply that 5 ought to do his part? If we construe "do his part" liberally, sothat an agent can on occasion do his part simply by letting others get onwith what they're doing, the answer is yes. But it is not the case that S must

24 Regan (1980), 113-15. This description of (9.1') may be misleading, however. The atti-tude of (in) transigence is always within the agent's control — it is an optional attitude — andclearly its adoption will sometimes involve action (as when Vincent tells Virgil that he willvote as Virgil wishes); and if, as I suspect (and as noted in Subsection 2.2.1), nothing is inone's control except by way of a decision, and if a decision can be said to constitute a min-imal type of action, then in fact adopting (in) transigence will always involve action.

270

always actively contribute to the achievement of the best group-achievableoutcome. There are three sorts of cases where his remaining idle in thisregard is permissible. First, it may be that he cannot actively contribute tothe outcome. Second, it may be that, although he can actively contributeto some outcome, he ought nonetheless to remain idle, because only in thisway can the group achieve its best. (If Rachel is a superior swimmer toRoberta and Roberta's joining in the rescue of a drowning person wouldonly mess things up, then Roberta ought to stand by while Rachel carriesout the rescue.) Third, it may be that S is permitted a choice as to whetherto remain idle, because there are more than enough people actively con-tributing to the group-best and his active contribution, though not coun-terproductive, would not help either. In this last sort of case, if S does notactively contribute to the outcome, he gets a kind of free ride; the best isachieved without any effort on his part. Such free riding seems to me unob-jectionable, though, if accompanied by the transigence that (9.1/) requires.

Fourthly, the illustrations used so far (of Vincent and Virgil; of Bob andCarol and Ted and Alice; of Rachel and Roberta) might be most natural-ly understood to concern group actions whose individual componentsoccur simultaneously, but such simultaneity is not essential to the illustra-tions. Indeed, I have already mentioned the possibility that Vincent's voteprecede Virgil's, and it may be that Bob and Carol and Ted must performactions that occur at (slightly) different times in order to save Alice. Now,it might seem that the temporal order of the individual component actionsin a group action can make a difference as to just what the members' indi-vidual obligations are - that whether or not to participate is optional for theearlier links in the chain in a way in which it is not for the last link; for bythe time the last link has a chance to contribute, all the others have con-tributed, their participation is settled, and so it is now clearly up to himalone whether the outcome is achieved. But in fact to single out the lastlink in this way would be a mistake. For each link there is the question ofwhether he ought to participate, and the answer is contingent on others'actions, whether or not these actions have already taken place. Thus, whereten are needed to deliver some vital message, what the first link, Archie,ought to do is to be determined in just the same way as what the last link,Jack, ought to do. If all nine of Archie, Barry,..., and Ian have cooperated,then Jack ought to pass the message along too. Similarly, if all nine of Barry,Chip,..., and Jack will cooperate, then Archie ought to pass the messagealong too. On the other hand, if at least one of Archie, Barry,..., and Ianhas not cooperated, then it is not the case that Jack ought to cooperate (forhe cannot; he has no message to pass on). Similarly, if one of Barry,

271

Chip,.. . , and Jack will not cooperate, then (presumably) it is not the casethat Archie ought to cooperate (for his effort will be wasted; he could pre-sumably do something better — unless, of course, he could prevail upon thenoncooperators to cooperate). In each instance, then, the timing of others'actions is irrelevant to the determination of an agent's individual obliga-tion, except insofar as such timing renders the others insusceptible to theagent's influence. 5

In summary, then, (9.1') condemns, as (9.1) does not, one's obstructinganother's access to the best possible outcome, when avoidance of suchobstruction is compatible with one's doing one's best. In this way, theproblem facing (9.1) is solved. However, (9.17) only gives rough expres-sion to the revision of (I) that I think is called for. At this point I offer thefollowing more precise account.

Recall (I). It says this:

(I) 5 ought, at T in W, to do A at time T ' iff(a) there is a world W' such that W' is accessible to 5 from W atT and 5

does v4 at T ' i n Wf;(b) there is a world W" such that W" is accessible to S from W at T and

S does not do A at T ' in W"; and(c) for all worlds W" such that W" is accessible to S from W' at T and S

does not do A at T' in W'\ there is a world W' such that(1) W' is accessible to S from W at T,(2) S does A at T ' in W\(3) the deontic value for 5 at T of W* is greater than the deontic value

for S at T of W", and(4) there is no world W'" such that

(i) W is accessible to S from W at T,(ii) 5 does not do A at T' in W"\ and

(iii) the deontic value for 5 at T of W"' is greater than the deon-tic value for S at T of W'.

Let us say that any act and time that meet the conditions laid out in (I) forA and T ' satisfy (I). Then, I think, the revision of (I) that is adumbrated in(9.1') may be put this way:

(I") 5 ought, at T in W, to act as follows:(a) for any act A and time T', do A at T ' iff A and T ' satisfy (I); and(b) for any act A and time T' that satisfy (I), and for any group of persons

G and proposition^?, be open at T to G concerning p at T' iff

25 Cf. Lyons (1965), pp. 79-80; Parfit (1984), p. 30.

272

(1) for all worlds W" such that W" is accessible to {S, G} from Wat T and S does A at T' in W", there is a world W' such that

(i) W' is accessible to {S, G} from W at T,(ii) {5, G} so acts at T ' in W' that p is true,

(iii) the deontic value26 of W' is greater than the deontic valueof W\ and

(iv) there is no world W'" such that(a) W'" is accessible to {S, G} from W at T,(j8) 5 does ,4 at T ' in W'", and(y) the deontic value of W'" is greater than the deontic

value of W'\ and(2) for any act A' and time T " such that T " is no earlier than T and

no later than T', if 4' and T " satisfy (I), then there is a world W'such that

(i) W' is accessible to S from ^ at T,(ii) S does ,4' at T " in P^', and

(iii) S is open at T in W' to G concerning p at T'.

Similar revision is also required for (II)—(XII), but I shall show mercy andforgo the details. Furthermore, if a group of more than one individual canbe obligated to do something, as (9.8) contends, then (I") can be extendedto account for this, in just the same manner as (9.1) is transformed into (9.8);but, again, I shall forgo the details.

I shall end with two observations. The first concerns the puzzle posedby Bernard Williams and first presented in Section 1.2. It concerns thesethree statements:

(1.1) someone ought to help that old lady:(1.2) Jones is the only person who can help her;(1.3) Jones ought to help her.

Williams contends that (1.1) does not express an obligation but that (1.3)does, and yet that (1.3) follows from (1.1) and (1.2). I disputed this. I wishnow to point to a possible reading of (1.1) which Williams doesn't consi-der and which may account for the inclination, if there is one, to accept theinference in question. It is this: as Williams says, (1.1) does not express anindividual obligation, but it does express a group obligation. That is, onemight take (1.1) to mean this:

(1.1") that group ought to help that old lady (by virtue of some member (s)of it doing so).

26 Recall the presupposition in Section 9.1 that deontic value is agent-neutral.

273

Taken by itself, this does not ascribe an obligation to any particular indi-vidual in the group to help the old lady. But if we now say, with (1.2), thatonly Jones is so situated that he can carry out the group's obligation, thenwe may indeed infer, with (1.3), that he, as an individual, is obligated tohelp the old lady. For the best that he can achieve can be no better than thebest that the group can achieve, and if the group can achieve its best onlyby his acting in this way, then he can achieve his best only by acting in thisway. This of course does not help Williams's cause, which is to establishsome important link between binding and nonbinding senses of "ought,"but it does at least provide a way to grant him his claim that the inferenceis to be accepted even though the first premise does not express an indi-vidual obligation while the conclusion does.

Finally, it may well have occurred to you that the move from (I) to (I")is unnecessary. For, you may say, (I) can itself accommodate the require-ment concerning transigence, as long as it is acknowledged that, in any sit-uation, the best that one can do is to perform some act transigently. In otherwords, transigence is just one of the features of (and thus just one of the fea-tures that contributes to the deontic value of) what one does and is simplyto be included in the calculus of values along with all the other value-con-tributory properties of actions (whatever they may be).

I have no real quarrel with this. As long as transigence is acknowledgedas being deontically significant, my main point in this chapter has beenaccepted. Just how this acknowledgment is to be made is not so important.If we grant that transigence itself has deontic value, then we can revert from(I") to (I). This may seem simpler and hence preferable, but we should notbe misled. We can revert to (I) only if we stipulate that transigence, asspelled out in clause (b) of (I"), is a necessary feature of the best that anagent can do. There is, therefore, no gain in simplicity. In fact, this way ofproceeding strikes me as pretty messy. We would need to be able to talk ofwhat would otherwise be best if transigence were bracketed, in order to beable to identify just what it is concerning which one ought to be transigent.I don't know quite how this should go. But if it can be done, so be it. Theresult would be an account of obligation extensionally equivalent to theone that I have proposed.

9.4 INTERPERSONAL MORAL DILEMMAS

If groups of more than one individual can be obligated, as (9.8) contends,then, as I have noted, (I") can be extended to account for this. Given thisextension, the question of group moral dilemmas could be treated in a man-

274

ner analogous to that employed in Chapter 7 when individual moral dilem-mas were discussed. That is, basic group moral dilemmas would be impos-sible - being ruled out by analogues to (2.46) (the thesis that "ought(singly)" implies "can (singly)"), (2.40'b) (the principle of agglomerationfor "ought (singly)"), and so on27 - but certain nonbasic group moraldilemmas (group binds, group subsidiary dilemmas) would be possible.However, the fact that individuals never act in isolation gives rise to thepossibility of another sort of dilemma that I have called an "interpersonalmoral dilemma." As stated in Section 7.4, the form of a (two-person) inter-personal moral dilemma is this:

(7.32) (a) St ought overall at T1 to do A at T3;(b) Sj can at Ti do A at T3;(c) S2 ought overall at T1 to do B at T4;(d) S2 can at T1 do B at T4; and(e) St and S2 cannot jointly (that is, {Su S2} cannot) at Ti do both

A at T3 and B at T4.

I postponed comment on such dilemmas to this chapter.One instance of an interpersonal dilemma is this.28 Donna ought to buy

her daughter a Cabbage Patch doll for her birthday tomorrow; Donaldought to buy his daughter a Cabbage Patch doll for her birthday tomor-row; and the only store in town that carries these dolls has just one left (and,I suppose we should add, Donna's daughter is distinct from Donald's). Howmight this case arise?

If Donna and Donald are mutually disconnected, then this case couldarise only by way of some lapse; hence the dilemma would constitute a per-plexity secundum quid, as explained in Section 7.4. This is so for the follow-ing reason. Suppose that Donna cannot see to it whether Donald buys thedoll, and suppose that Donald does as he ought and buys the doll. Thenevery world accessible to Donna is one in which Donald buys the doll; henceno world accessible to her is one in which she buys it; hence it is not the caseafter all that she ought to buy it. Thus it can happen that Donna ought tobuy the doll only if Donald fails to act as he ought; her obligation presup-poses his lapse. The same applies, mutatis mutandis, where Donald cannotsee to it whether Donna buys the doll.

The point here is not simply the obvious one that a conflict of obliga-tions implies that some wrong is done; it's that the sort of conflict present-

27 Cf. McConnell (1988a), p. 27.28 This illustration is borrowed, in modified form, from McConnell (1988a), p. 28.

275

ly at issue, in the circumstances presently at issue, presupposes that somewrong is done. It is not simply that, if one person satisfies his obligation,then the other fails to satisfy his; it's that the one has an obligation only ifthe other fails to satisfy his obligation.

If Donna and Donald are connected, however, the story is different. Butif we make the presuppositions operative throughout this chapter - in par-ticular, the presupposition that deontic value is agent-neutral - the result isstill that interpersonal moral dilemmas can arise only by way of some lapse.This result is obvious if we admit the possibility of group obligation; for agroup can always achieve its best and, if it does so, then, as (9.11) says, nomember does wrong. If a situation arises, then, in which some member can-not avoid some member's doing wrong, this can only be because the groupdoes wrong; the members' having conflicting obligations presupposes thegroup's lapse. But even if we do not admit the possibility of group obliga-tion, still such conflicting obligations presuppose a lapse on someone's part.There are three possibilities. Either Donna's buying the doll is deonticallysuperior to Donald's doing so (more precisely, every world accessible to{Donna, Donald} in which Donald buys the doll is deontically inferior tosome world accessible to {Donna, Donald} in which Donna buys the doll),or Donald's buying the doll is deontically superior, or they are of equaldeontic value. Suppose that the first is the case. Given the agent-neutrali-ty of deontic value, it must be that, if Donna is transigent concerning herbuying the doll, then Donald can and ought to ensure that she buys the dolland that he does not buy it himself; hence, if in fact he has the obligation tobuy it, this presupposes both that Donna does not do her best and that sheis intransigent concerning her (and thus the group's) best, and so it presup-poses a double lapse on her part. So too, mutatis mutandis, for the case whereDonald's buying the doll is deontically superior to Donna's doing so. As tothe final possibility: if the group-best is for one or the other of Donna andDonald to buy the doll, then it could not be individually better for Donnaor Donald to buy it unless either Donna or Donald acts so as to precludeachievement of what is group-best; this, too, requires a lapse on someone'spart.

Finally, if Donna and Donald are connected but deontic value is notagent-neutral, it appears that interpersonal moral dilemmas can arise with-out any lapse on someone's part. Their dilemma would be rooted simplyin the agent-relativity of deontic value. If it is thought that no moral dilem-ma of any sort can arise unless it constitutes a perplexity secundum quid, thiswould constitute a reason for rejecting the possibility that deontic value besignificantly agent-relative.

276

Postscript

I have endeavored in this book to develop and defend a certain analysis ofthe concept of moral obligation, an analysis that can be put to good use inthe resolution of a number of philosophical issues. In the Preface, I mademention of several questions and claimed that they would receive answersin the ensuing pages. And so they have, although some of the answers havebeen somewhat tentative. My hope is that these answers will have servedcollectively to set the concept of moral obligation in high relief Such clar-ification is valuable not just for its own sake. Although the account of oblig-ation that I have presented is, appropriately, neutral regarding competingsubstantive theories of obligation, it should provide a useful frameworkwithin which these theories can be developed and assessed. Thus the exer-cise in metaethics undertaken in the foregoing pages has significant impli-cations for normative ethics. Of course, it has not been my purpose toexpound on these implications. Undoubtedly, too, my account has impli-cations in the area of applied ethics, but tracing these is a very complex mat-ter, and I offer no suggestions on how best to carry out this task. To repeatwhat was said in the Preface, then, anyone seeking solutions to such prob-lems as those besetting the unfortunate Indianapolis would be ill advised tolook for them in this book. I trust that by now that much is clear.Nonetheless, if it can be agreed that some progress has been made hereregarding the questions that have been addressed, I shall be content.

277

Appendix: List of propositions

(A) 11(B)ll(C)ll(D)12(E)12(MO) 75

(1)26(I') 133(I") 272

(II) 32(III) 33(IV) 33(V)34(VI) 35(VII) 35(VIII) 35(IX) 35(X)36(XI) 36(XII) 36(XIII) 119

(XIII') 133(XIV) 120(XV) 121(XVI) 133

(1.1)4(1.1')5(1.1") 273

(1.2)4(1.3)4

(1.30 5(1.4)6

(1.40 177(1.5)6

(1.50 177(1.6)6(1.7)6(1.8) 11(1.9)11(1.10) 11(1.11)11

(2.1)21(2.2) 21(2.3) 21(2.4) 22(2.5) 22

(2.50 24(2.6) 22(2.7) 22

(2.70 24(2.8) 22

(2.80 24(2.9) 23(2.10) 23(2.11)39(2.12) 40(2.13) 40

(2.14) 40(2.15) 45(2.16) 45(2.17) 47(2.18) 47(2.19) 48(2.20) 48(2.21) 51(2.22) 51(2.23) 51(2.24) 52(2.25) 52(2.26) 52(2.27) 53(2.28) 54(2.29) 54(2.30) 54(2.31) 54(2.32) 54(2.33) 56(2.34) 56(2.35) 57(2.36) 63

(2.360 64(2.37) 63

(2.370 64(2.37'a) 65(2.37'b) 65(2.37'c) 65

(2.37") 65

278

(2.38) 63(2.38') 65

(2.39) 63(2.39') 65(2.39") 66(2.39'") 66

(2.40) 63(2.40') 66

(2.40'a) 66(2.40'b) 66(2.40'c) 68(2.40'd) 68(2.40'e) 68

(2.41) 63(2.41') 68

(2.41'a) 68(2.41'b) 69

(2.42) 63(2.42') 69

(2.43) 63(2.43') 69(2.43") 69(2.43'") 70

(2.44) 63(2.44') 70

(2.45) 63(2.45') 70

(2.46) 71(2.46') 85

(2.47) 71(2.48) 72(2.49) 72

(2.49') 72(2.50) 72

(2.50') 73(2.51) 76(2.52) 76(2.53) 76(2.54) 76(2.55) 77

(2.56) 77(2.57) 77(2.58) 78

(3.1) 80(3.1') 80(3.1") 80

(3.2) 80(3.2') 80(3.2") 80

(3.3) 81(3.4) 81

(3.4') 82(3.5) 81(3.6) 81(3.7) 81(3.8) 81(3.9) 83

(3.9') 83(3.9") 83(3.9'") 83

(3.10) 83(3.10') 83(3.10") 83(3.10'") 84

(3.11)85(3.11a) 85

(3.12) 85(3.13) 86

(3.13') 86(3.14) 86

(3.14') 86(3.15) 87(3.16) 87(3.17) 87

(3.17') 89(3.18) 87(3.19) 88(3.20) 90(3.21) 90

(3.22) 90(3.23) 90(3.24) 92(3.25) 92(3.26) 96(3.27) 96(3.28) 100(3.29) 101

(3.29') 103(3.29") 104

(3.30) 106(3.31) 106(3.32) 111(3.33) 113

(4.1) 114(4.1') 115

(4.2) 114(4.3) 115

(4.3') 122(4.4) 115

(4.4') 115(4.5) 115

(4.5') 115(4.6) 115(4.7) 116(4.8) 116(4.9) 116(4.10) 116(4.11) 117(4.12) 117(4.13) 117

(4.13') 122(4.14) 118(4.15) 118(4.16) 118(4.17) 120

(4.17') 121(4.18) 121

(4.18a) 121

279

(4.18b) 121(4.19) 121(4.20) 121(4.21) 121(4.22) 122(4.23) 122(4.24) 123(4.25) 123

(4.25') 138(4.26) 123

(4.26') 138(4.27) 123

(4.27') 124(4.28) 123

(4.28') 124(4.28") 138

(4.29) 124(4.30) 124(4.31) 124(4.32) 124(4.33) 125(4.34) 126(4.35) 126(4.36) 126(4.37) 126(4.38) 126(4.39) 126(4.40) 127(4.41) 127(4.42) 127(4.43) 127(4.44) 127(4.45) 127(4.46) 127(4.47) 127(4.48) 127(4.49) 128(4.50) 128

(4.50') 134(4.51) 130

(4.52) 130(4.53) 130(4.54) 130(4.55) 135(4.56) 135(4.57) 137(4.58) 138(4.59) 138(4.60) 138(4.61) 140(4.62) 140(4.63) 140(4.64) 140

(5.1) 143(5.2) 144(5.3) 144(5.4) 144(5.5) 144(5.6) 145(5.7) 148

(5.7') 167(5.8) 149(5.9) 149

(5.9') 158(5.10) 149

(5.10') 168(5.11) 151(5.12) 151(5.13) 151(5.14) 152(5.15) 152(5.16) 152(5.17) 158(5.18) 162(5.19) 165(5.20) 165(5.20) 165(5.21) 165(5.22) 165

(5.23) 165(5.24) 166(5.25) 170(5.26) 171(5.27) 172(5.28) 177(5.29) 177(5.30) 182(5.31) 182(5.32) 182(5.33) 182(5.34) 182(5.35) 182(5.36) 183(5.37) 183

(6.1) 189(6.1') 195

(6.2) 190(6.3) 192

(6.3') 192(6.4) 196(6.5) 197(6.6) 198(6.7) 198

(6.7') 199(6.7") 199

(6.8) 199(6.9) 203(6.10) 203

(7.1)210(7.2) 210(7.3) 210(7.4) 210(7.5) 210(7.6)211(7.7)211

(7.7') 226(7.8) 212

280

(7.9) 213(7.10)213(7.11)214(7.12)214(7.13)214(7.14)214(7.15)215(7.16)215(7.17) 215(7.18)216(7.19) 216(7.20) 221(7.21) 222(7.22) 222

(7.22') 223(7.22") 227(7.22'") 227

(7.23) 222(7.24) 222(7.25) 223

(7.25') 227

(7.26) 223(7.26') 227

(7.27) 223(7.28) 224(7.29) 224(7.30) 226(7.31) 228(7.32) 229

(8.1)243(8.2) 243(8.3) 243(8.4) 243(8.5) 244(8.6) 244(8.7) 244(8.8) 244(8.9) 246(8.10) 246(8.11)246(8.12)246

(8.13)247

(9.1) 254(9.1') 268

(9.2) 255(9.3) 255(9.4) 256(9.5) 256(9.6) 258(9.7) 260(9.8) 260(9.9) 261(9.10)261(9.11)261(9.12)261(9.13)264

(9.13') 264(9.14)265(9.15)265

281

List of works cited

Anscombe, G. E. M. (1969). Intention. Ithaca: Cornell University Press.Aquinas, St. Thomas (1964—75). Summa Theologiae, translated by Thomas

Gilby et al. New York: McGraw-HillAqvist, Lennart (1967). "Good Samaritans, Contrary-to-Duty Imperatives,

andEpistemic Obligations." Nous, 1: 361—79.Attfield, Robin (1979). "Supererogation and Double Standards." Mind,

88: 481-99.Atwell, John (1978). "Ross and Prima Facie Duties." Ethics, 88: 240-9.Austin, J. L. (1968). "A Plea for Excuses." In The Philosophy of Action,

edited by Alan R. White. Oxford: Oxford University Press, 1968,19-42.

Baier, Kurt (1958). The Moral Point of View. Ithaca: Cornell UniversityPress.

Barley, Blake (1984). Review of Regan (1980). Nous, 18: 152-9.Barnes, Gerald (1971). "Utilitarianisms." Ethics, 82: 56-64.Baron, Marcia (1987). "Kantian Ethics and Supererogation." Journal of

Philosophy, 84: 237-62.Bennett, Jonathan (1989). "Two Departures from Consequentialism."

Ethics, 100: 54-66.Berlin, Isaiah (1979). Concepts and Categories. New York: Viking Press.Brand, Myles (1971). "The Language of Not Doing." American Philosophical

Quarterly, 8: 45-53.Broad, C. D. (1946). "Some of the Main Problems in Ethics." Philosophy,

21:99-117.Broad, C. D. (1985). Ethics. Dordrecht: Martinus Nijhoff.Castafieda, Hector-Neri (1968). "A Problem for Utilitarianism." Analysis,

28: 141-2.Castafieda, Hector-Neri (1975). Thinking and Doing. Dordrecht: D.

Reidel.

282

Castaneda, Hector-Neri (1977). "Ought, Time, andDeontic Paradoxes."Journal of Philosophy, 74: 775-91.

Castaneda, Hector-Neri (1981). "The Paradoxes of Deontic Logic: TheSimplest Solution to All of Them in One Fell Swoop." In Hilpinen(1981): 37-85.

Castaneda, Hector-Neri (1989a). "Paradoxes of Moral Reparation:Deontic Foci vs. Circumstances." Philosophical Studies, 57: 1—21.

Castaneda, Hector-Neri (1989b). "Moral Obligation, Circumstances, andDeontic Foci (A Rejoinder to Fred Feldman)." Philosophical Studies,57: 157-74.

Chisholm, Roderick M. (1963). "Contrary-to-Duty Imperatives andDeontic Logic." Analysis, 24: 33—6.

Chisholm, Roderick M. (1968). "Supererogation and Offence." InThomson and Dworkin (1968): 412-29.

Chisholm, Roderick M. (1974). "Practical Reason and the Logic ofRequirement." In Practical Reason, edited by Stephan Korner. NewHaven: Yale University Press, 1974, 1-17.

Chisholm, Roderick M. (1976). Person and Object. La Salle, 111.: OpenCourt.

Chisholm, Roderick M. (1989). Theory of Knowledge, 3rd edition.Englewood Cliffs: Prentice-Hall.

Chisholm, Roderick M. and Sosa, Ernest (1966). "Intrinsic Preferabilityand the Problem of Supererogation." Synthese, 16: 321-31.

Clark, Michael (1993). "On Wanting to Be Morally Perfect." Analysis, 53:54-6.

Conee, Earl (1983). Review of Regan (1980). Journal of Philosophy, 80:415-24.

Conee, Earl (1987). "Against Moral Dilemmas." In Gowans (1987b):239-49.

Conee, Earl (1989). "Why Moral Dilemmas Are Impossible." AmericanPhilosophical Quarterly, 26: 133-41.

Davidson, Donald (1980). Essays on Actions and Events. Oxford: ClarendonPress.

Davis, Lawrence H. (1979). Theory of Action. Englewood Cliffs: Prentice-Hall.

Donagan, Alan (1977). The Theory of Morality. Chicago: University ofChicago Press.

Donagan, Alan (1987). "Consistency in Rationalist Moral Systems." InGowans (1987b): 271-90.

283

Dreier, James (1993). "Structures of Normative Theories." Monist, 76:22-40.

Driver, Julia (1992). "The Suberogatory." Australasian Journal ofPhilosophy,70: 286-95.

Feinberg, Joel (1968). "Supererogation and Rules." In Thomson andDworkin (1968): 391-411.

Feinberg, Joel (1973). Social Philosophy. Englewood Cliffs: Prentice-Hall.Feinberg, Joel (1979). "The Nature and Value of Rights." In Lyons

(1979c): 78-91.Feinberg, Joel (1980). Rights, Justice, and the Bounds of Liberty. Princeton:

Princeton University Press.Feldman, Fred (1978). Introductory Ethics. Englewood Cliffs: Prentice-Hall.Feldman, Fred (1986). Doing the Best We Can. Dordrecht: D. Reidel.Feldman, Fred (1989a). "The Paradox of the Knower." Philosophical

Studies, 55: 93-100.Feldman, Fred (1989b). "Concerning the Paradox of Moral Reparation

and Other Matters." Philosophical Studies, 57: 23-39.Feldman, Fred (1990). "A Simpler Solution to the Paradoxes of Deontic

Logic ." Philosophical Perspectives, 4: 309—41.

Feldman, Fred (1993). "On the Consistency of Act- and Motive-Utilitarianism." Philosophical Studies, 70: 201-12.

Follesdal, Dagfinn and Hilpinen, Risto (1970). "Deontic Logic: AnIntroduction." In Hilpinen (1970): 1-35.

Foot, Philippa (1980). "The Problem of Abortion and the Doctrine ofDouble Effect." In Killing and Letting Die, edited by Bonnie Stein-bock. Englewood Cliffs: Prentice-Hall, 1980, 156-65.

Forrester, James (1984). "Gentle Murder, or the Adverbial Samaritan."Journal of Philosophy, 86: 193-7.

Forrester, Mary (1975). "Some Remarks on Obligation, Permission, andSupererogation." Ethics, 85: 219—26.

Frankena, William K. (1963). "Obligation and Ability." In PhilosophicalAnalysis, edited by Max Black. Englewood Cliffs: Prentice-Hall, 1963,148-65.

Frankfurt, Harry G. (1969). "Alternate Possibilities and MoralResponsibility." Jowratf/o/P/n/osop/*)/, 66: 829-39.

Frankfurt, Harry G. (1988). The Importance of What We Care About. Cam-bridge: Cambridge University Press.

Fried, Charles (1978). Right and Wrong. Cambridge, Mass.: HarvardUniversity Press.

284

Gert, Bernard (1988). Morality: A New Justification of the Moral Rules. NewYork: Oxford University Press.

Gewirth, Alan (1981). "Are There Any Absolute Rights?" PhilosophicalQuarterly, 31: 1-16.

Gibbard, Allan (1965). "Rule-Utilitarianism: Merely an Illusory Alter-native?" Australasian Journal of Philosophy, 43: 211-20.

Goble, Lou (1991). "Murder Most Gentle: The Paradox Deepens."Philosophical Studies, 64: 217-27.

Goble, Lou (1993). "The Logic of Obligation, 'Better' and 'Worse.'"Philosophical Studies, 70: 133-63.

Goldman, Alvin I. (1976). A Theory of Human Action. Princeton: PrincetonUniversity Press.

Goldman, Holly S. (1976). "Dated Rightness and Moral Imperfection."Philosophical Review, 85: 449-87.

Goldman, Holly S. (1978). "Doing the Best One Can." In Values andMorals, edited by Alvin I. Goldman and Jaegwon Kim. Dordrecht:D. Reidel, 1978, 185-14.

Gorr, Michael (1979). "Omissions." Tulane Studies in Philosophy, 28:93—102.Gowans, Christopher (1987a). "Introduction: The Debate on Moral

Dilemmas." In Gowans (1987b): 3-33.Gowans, Christopher (1994). Innocence Lost. New York: Oxford

University Press.Gowans, Christopher, ed. (1987b). Moral Dilemmas. New York: Oxford

University Press.Greenspan, Patricia S. (1975). "Conditional Oughts and Hypothetical

Imperatives." Journal of Philosophy, 72: 259-76.Greenspan, Patricia S. (1978). "Oughts and Determinism: A Response to

Goldman." Philosophical Review, 87: 77—83.Hart, H. L. A. (1979). "Are There Any Natural Rights?" In Lyons (1979c):

14-25.Henderson, G. P. (1966). "'Ought' Implies 'Can.'" Philosophy, 41:101-12.Heyd, David (1982). Supererogation. Cambridge: Cambridge University

Press.Hill, Thomas E., Jr. (1971). "Kant on Imperfect Duty and Super-

erogation." Kant-Studien, 72: 55-76.Hill, ThomasE.Jr. (1979). "Servility and Self-Respect." InLyons (1979c):

111-24.Hilpinen, Risto, ed. (1970). Deontic Logic: Introductory and Systematic Readings.

Dordrecht: D. Reidel.

285

Hintikka, Jaakko (1970). "Some Main Problems in Deontic Logic." InHilpinen (1970), 59-104.

Hohfeld, Wesley Newcomb (1919). Fundamental Legal Conceptions. NewHaven: Yale University Press.

Hudson, Stephen D. (1986). Human Character and Morality. London:Routledge and Kegan Paul.

Humberstone, I. L. (1983). "The Background of Circumstances." PacificPhilosophical Quarterly, 64: 19-34.

Humberstone, I. L. (1991). "Two Kinds of Agent-Relativity." PhilosophicalQuarterly, 41: 144-66.

Hume, David (1888). Treatise of Human Nature, edited by L. A. Selby-Bigge. Oxford: Clarendon Press.

Jackson, Frank (1985). "On the Semantics and Logic of Obligation." Mind,94: 177-95.

Jackson, Frank (1987). "Group Morality." In Metaphysics and Morality, edit-ed by Philip Pettit, Richard Sylvan, and Jean Norman. Oxford: BasilBlackwell, 1987, 91-110.

Jackson, Frank (1991). "Decision-theoretic Consequentialism and theNearest and Dearest Objection." Ethics, 101: 461-82.

Jackson, Frank and Pargetter, Robert (1986). "Oughts, Options, andActualism." Philosophical Review, 95: 233—55.

Jackson, Frank and Pargetter, Robert (1987). "Two Puzzles aboutConditional Obligation." Philosophical Papers, 16: 75—83.

Kagan, Shelly (1989). The Limits of Morality. Oxford: Clarendon Press.Kamm, Frances Myrna (1985). "Supererogation and Obligation." Journal

of Philosophy, 82: 118-38.Kant, Immanuel (1964). Groundwork of the Metaphysic of Morals, translated

by H. J. Paton. New York: Harper and Row.Kim, Jaegwon (1976). "Events as Property Exemplifications." In Action

Theory, edited by Myles Brand and Douglas Walton. Dordrecht: D.Reidel, 1976, 159-77.

Lebus, Bruce (1990). "Moral Dilemmas: Why They are Hard to Solve."Philosophical Investigations, 13: 110—25.

Lemmon, E. J. (1987). "Moral Dilemmas." In Gowans (1987b): 101-14.Lemos, Ramon M. (1980). "Duty and Ignorance." Southern Journal of

Philosophy, 18: 301-12.Lewis, David (1973). Counterfactuals. Cambridge, Mass.: Harvard University

Press.Loewer, Barry and Belzer, Marvin (1991). "Prima Facie Obligation: Its

Deconstruction and Reconstruction." In John Searle and His Critics,

286

edited by Ernest Lepore and Robert van Gulick. Cambridge, Mass.:Basil Blackwell, 1991, 359-70.

Lyons, David (1965). Forms and Limits of Utilitarianism. Oxford: ClarendonPress.

Lyons, David (1979a). "Introduction." In Lyons (1979c): 1-13.Lyons, David (1979b). "Rights, Claimants, and Beneficiaries." In Lyons

(1979c): 58-77.Lyons, David, ed. (1979c). Rights. Belmont, Calif: Wadsworth.Marcus, Ruth Barcan (1966). "Iterated Deontic Modalities." Mind, 75:

580-2.Marcus, Ruth Barcan (1987). "Moral Dilemmas and Consistency." In

Gowans (1987b): 188-204.Martinich, A. P. (1987). "Obligation, Ability and Prima Facie Promising."

Philosophia, 17: 323-30.McCloskey, H. J. (1963). "Ross and the Concept of Prima Facie Duty."

Australasian Journal of Philosophy, 41: 336-45.McCloskey, H. J. (1973). "Utilitarianism: Two Difficulties." Philosophical

Studies, 24: 62-3.McConnell, Terrance C. (1980). "Utilitarianism and Supererogatory

Acts," Ratio, 22: 36-8.McConnell, Terrance C. (1987). "Moral Dilemmas and Consistency in

Ethics." In Gowans (1987b): 154-73.McConnell, Terrance C. (1988a). "Interpersonal Moral Conflicts."

American Philosophical Quarterly, 25: 25-35.McConnell, Terrance C. (1988b). "Ross on Duty and Ignorance." History

of Philosophy Quarterly, 5: 79-95.McConnell, Terrance C. (1989). "'Ought' Implies 'Can' and the Scope of

Moral Requirements." Philosophia, 19: 437-54.McConnell, Terrance C. (1993a). Gratitude. Philadelphia: Temple

University Press.McConnell, Terrance C. (1993b). "Dilemmas and Incommensurateness."

Journal of Value Inquiry, 27: 247—52.McConnell, Terrance C. (1995). "Moral Residue and Dilemmas." In

Moral Dilemmas and Moral Theory, edited by H. E. Mason. New York:Oxford University Press, forthcoming.

McGinn, Colin (1992). "Must I Be Morally Perfect?" Analysis, 52: 32-4.McKinsey, Michael (1975). "Levels of Obligation." Philosophical Studies,

35: 385-95.McKinsey, Michael (1981). "Obligations to the Starving." Nous, 15:

309-23.

287

Mellema, Gregory (1991). Beyond the Call of Duty. Albany: State Universityof New York Press.

Mellema, Gregory (1993). "Quasi-Obligation and the Failure to BeVirtuous." Journal of Social Philosophy, 24: 176-85.

Mendola, Joseph (1987). "The Indeterminacy of Options." AmericanPhilosophical Quarterly, 24: 125-36.

Mill, John Stuart (1957). Utilitarianism. Indianapolis: Bobbs-Merrill.Montague, Phillip (1989). "Acts, Agents, and Supererogation." American

Philosophical Quarterly, 26: 101-11.Montefiore, Alan (1958). "'Ought' and 'Can.'" Philosophical Quarterly, 8:

24-40.Moore, G. E. (1903). Principia Ethica. Cambridge: Cambridge University

Press.Moore, G. E. (1922). Philosophical Studies. London: Routledge and Kegan

Paul.Moore, G. E. (1965). Ethics. Oxford: Oxford University Press.Moore, Robert E. (1979). "Refraining." Philosophical Studies, 36: 407-24.Nagel, Thomas (1986). The View from Nowhere. New York: Oxford

University Press.New, Christopher (1974). "Saints, Heroes, and Utilitarians." Philosophy,

49: 179-89.Nozick, Robert (1974). Anarchy, State, and Utopia. New York: Basic

Books.Parfit, Derek (1984). Reasons and Persons. Oxford: Clarendon Press.Pietroski, Paul M. (1993). "Prima Facie Obligations, Ceteris Paribus Laws

in Moral Theory." Ethics, 103: 489-515.Pollock, John L. (1976). Subjunctive Reasoning. Dordrecht: D. Reidel.Postow, B. C. (1977). "Generalized Act Utilitarianism." Analysis, 37:

49-52.Powers, Lawrence (1967). "Some Deontic Logicians." Nous, 1: 381-400.Prichard, H. A. (1949). Moral Obligation. Oxford: Clarendon Press.Purtill, Richard L. (1973). "Deontically Perfect Worlds and Prima Facie

Obligations." Philosophia, 3: 429-38.Rainbolt, George W. (1993). "Moral Rights Conflict and Compensation:

An Objection to Thomson." Presented at the 1993 Central DivisionMeeting of the American Philosophical Association.

Raz, Joseph (1975). "Permissions and Supererogation," AmericanPhilosophical Quarterly, 12: 161-8.

Regan, Donald H. (1980). Utilitarianism and Co-operation. Oxford:Clarendon Press.

288

Regan, Donald H. (1983). "Against Evaluator Relativity: A Response toSen." Philosophy and Public Affairs, 12: 93-112.

Richman, Robert J. (1983). God, Free Will, and Morality. Dordrecht: D.Reidel.

Roberts, Robert C. (1988). "What an Emotion Is: A Sketch." PhilosophicalReview, 97: 183-209.

Robinson, Richard (1971). "Ought and Ought Not." Philosophy, 46:193-202.

Ross, Alf (1941). "Imperatives and Logic." Theoria, 7: 53-71.Ross, W. D. (1930). The Right and the Good. Oxford: Clarendon Press.Ross, W. D. (1939). Foundations of Ethics. Oxford: Clarendon Press.Rousseau, Jean Jacques (1992). Discourse on the Origins of Inequality, edited

by Roger W. Masters and Christopher Kelly. Hanover, N.H.:University Press of New England.

Russell, Bertrand (1970). "The Elements of Ethics." In Readings in EthicalTheory, edited by Wilfrid Sellars and John Hospers, 2nd edition.Englewood Cliffs: Prentice-Hall, 1970, 3-28.

Sartre, Jean-Paul (1973). "Existentialism Is a Humanism." In ExistentialistPhilosophy, edited by James A. Gould and Willis H. Truitt. Encino,Calif: Dickenson, 1973, 27-40.

Sayre-McCord, Geoffrey (1986). "Deontic Logic and the Priority of MoralTheory." Nous, 20: 179-97.

Scheffler, Samuel (1984). The Rejection of Consequentialism. Oxford:Clarendon Press.

Schwartz, Thomas (1985). "Dr. Krankheit and the Concept of Comp-liance." American Philosophical Quarterly, 22: 81—7.

Sen, Amartya (1983). "Evaluator Relativity and ConsequentialistEvaluation." Philosophy and Public Affairs, 12: 113-32.

Seung, T. K. and Bonevac, Daniel (1992). "Plural Values andIndeterminate Rankings." Ethics, 102: 799-813.

Shaw, P. D. (1965). "'Ought' and 'Can.'" Analysis, 23: 118-20.Simon, Caroline J. (forthcoming). "Supererogation: Puzzle or Pseudo-

Problem?"Sinnott-Armstrong, Walter (1985). "'Ought to Have' and 'Could Have.'"

Analysis, 45: 44-8.Sinnott-Armstrong, Walter (1988). Moral Dilemmas. Oxford: Basil Blackwell.Slote, Michael (1989). Beyond Optimizing. Cambridge, Mass.: Harvard

University Press.Smart, J. J. C. (1973). "An Outline of a System of Utilitarian Ethics." In

Smart and Williams (1973): 3-74.

289

Smart, J. J. C. and Williams, Bernard (1973). Utilitarianism: For and Against.Cambridge: Cambridge University Press.

Smith, Holly M. (1986). "Moral Realism, Moral Conflict, and CompoundActs." Journal ofPhilosophy, 83: 341-5.

Snare, Frank (1974). "The Definition of Prima Facie Duties." PhilosophicalQuarterly, 24: 235-44.

Sobel, Jordan Howard (1968). "Rule-Utilitarianism." Australasian Journalof Philosophy, 46: 146-65.

Sobel, Jordan Howard (1976). "Utilitarianism and Past and FutureMistakes." Nous, 10: 195-219.

Sorensen, Roy A. (1991). "Moral Dilemmas, Thought Experiments, andConflict Vagueness." Philosophical Studies, 63: 291-308.

Stocker, Michael (1971). "'Ought' and 'Can.'" Australasian Journal ofPhilosophy, 49: 303-16.

Stocker, Michael (1990). Plural and Conflicting Values. Oxford: ClarendonPress.

Strawson, Galen (1986). Freedom and Belief. Oxford: Clarendon Press.Stump, Eleonore (1992). "God's Obligations." Philosophical Perspectives, 6:

475-91.Styron, William (1980). Sophie's Choice. New York: Bantam.Swank, Casey (1985). "Reasons, Dilemmas and the Logic of'Ought.'"

Analysis, 45: 111-16.Tannsjo, Torbjorn (1985). "Moral Conflict and Moral Realism." Journal of

Philosophy, 82: 113-17.Thomason, Richmond H. (1981). "Deontic Logic and the Role of

Freedom in Moral Deliberation." In New Studies in Deontic Logic,edited by Risto Hilpinen. Dordrecht: D. Reidel, 1981, 177-86.

Thomson, Judith Jarvis (1986). Rights, Restitution, and Risk. Cambridge,Mass.: Harvard University Press.

Thomson, Judith Jarvis (1990). The Realm of Rights. Cambridge, Mass.:Harvard University Press.

Thomson, Judith Jarvis and Dworkin, Gerald, eds. (1968). Ethics. NewYork: Harper and Row.

Unger, Peter (1977). "Impotence and Causal Determinism." PhilosophicalStudies, 31:289-305.

Vallentyne, Peter (1987a). "Prohibition Dilemmas and Deontic Logic."Logique et Analyse, 117-18: 113-21.

Vallentyne, Peter (1987b). "Utilitarianism and the Outcomes of Actions."Pacific Philosophical Quarterly, 68: 57-70.

290

Vallentyne, Peter (1988). "Gimmicky Representations and MoralTheories." Metaphilosophy, 19: 253—63.

Vallentyne, Peter (1989). "Two Types of Moral Dilemmas." Erkenntnis,30: 301-18.

Vallentyne, Peter (1992). "Moral Dilemmas and Comparative Concep-tions of Morality." Southern Journal of Philosophy, 30: 117—24.

van Fraassen, Bas (1987). "Values and the Heart's Command." In Gowans(1987b): 138-53.

van Inwagen, Peter (1983). An Essay on Free Will. Oxford: Clarendon Press.van Rijen, Jeroen (1993). Review of Feldman (1986). Nous, 27: 264-6.Velazco y Trianosky, Gregory (forthcoming). "On Defining the

Supererogatory."von Wright, Georg Henrik (1968). "An Essay in Deontic Logic and the

General Theory of Action." Ada Philosophica Fennica, 21.Vorobej, Mark (1986). "Conditional Obligation and Detachment."

Canadian Journal of Philosophy, 16: 11—26.Walton, Douglas N. (1980). "Omitting, Refraining and Letting Happen."

American Philosophical Quarterly, 17: 319-26.White, Alan R. (1975). Modal Thinking. Oxford: Basil Blackwell.Widerker, David (1991). "Frankfurt on 'Ought Implies Can' and

Alternative Possibilities." Analysis, 51: 222—4.Williams, Bernard (1973). "A Critique of Utilitarianism." In Smart and

Williams (1973), 77-150.Williams, Bernard (1981). Moral Luck. Cambridge: Cambridge University

Press.Williams, Bernard (1987). "Ethical Consistency." In Gowans (1987b):

115-37.Zimmerman, Michael J. (1984). An Essay on Human Action. New York:

Peter Lang.Zimmerman, Michael J. (1986). "Subsidiary Obligation." Philosophical Studies,

50: 65-75.Zimmerman, Michael J. (1987). "Remote Obligation." American Philoso-

phical Quarterly, 24: 199-205.Zimmerman, Michael J. (1988a). An Essay on Moral Responsibility. Totowa,

N.J.: Rowman and Littlefield.Zimmerman, Michael J. (1988b). "Lapses and Dilemmas." Philosophical

Papers, 17: 103-12.Zimmerman, Michael J. (1990a). "The Range of Options." American Philoso-

phical Quarterly, 27: 345-55.

291

Zimmerman, MichaelJ. (1990b). "Where Did I Go Wrong?" PhilosophicalStudies, 58: 83-106.

Zimmerman, Michael J. (1992). "Cooperation and Doing the Best OneCan." Philosophical Studies, 65: 283-304.

Zimmerman, Michael J. (1993a). "Obligation, Responsibility, andAlternate Possibilities." Analysis, 53: 51-3.

Zimmerman, Michael J. (1993b). "Supererogation and Doing the BestOne Can." American Philosophical Quarterly, 30: 373-80.

Zimmerman, Michael J. (1993c). "A Plea for Ambivalence." Meta-philosophy, 24: 382-9.

Zimmerman, MichaelJ. (1994). "Rights, Compensation, and Culpability."Law and Philosophy, 13: 1—32.

Zimmerman, MichaelJ. (1995a). "Prima Facie Obligation and Doing theBest One Can." Philosophical Studies, 78: 87-123.

Zimmerman, Michael J. (1995b). "Responsibility Regarding theUnthinkable." Midwest Studies in Philosophy, 20: forthcoming.

292

Index of names

Anscombe, G. E. M , 55, 55n., 56, 278Aquinas, St. Thomas, 229, 229n., 278Aqvist, Lennart, 73n., 74n., 278Attfield, Robin, 246n., 278Atwelljohn, 171n., 278Austin, J. L., 93, 93n., 94, 94n., 278

Badhwar, Neera, xiiiBaier, Kurt, 239n., 241n., 278Barley, Blake, 260n., 278Barnes, Gerald, 254n., 278Baron, Marcia, 233n., 253n., 278Belzer, Marvin, 173n., 282Bennett, Jonathan, 243n., 278Berlin, Isaiah, 58n., 278Bonevac, Daniel, 58n., 285Brand, Myles, 38n., 278Broad, C. D., 2n., 3n., 6n., 12, 12n.,

13n., 37n., 44n., 56n., 141n.,278

Castaiieda, Hector-Neri, 2n., 3n., 21,21n., 36n., 37n., 44, 45n., 73n.,97n., 124, 124n., 128, 128n.,134, 278, 279

Chisholm, Roderick M., xiv, 2n., 3,3n., 4n., 39n., 47n., 77n., 127,127n., 128, 134, 173n., 235n.,236, 236n., 238, 238n., 246n.,247n., 279

Clark, Michael, 71n., 279Clarke, Randolph, xiiiClinton, Bill, 45Coley, James, xiiiConee, Earl, xiii, 208n., 217n., 225n.,

228n., 260n., 279Copp, David, xiii

Davidson, Donald, 43n., 55, 55n.,56n., 279

Davis, Lawrence H., 4In., 279Donagan, Alan, 17n., 40n., 141, 142,

142n., 143, 166n., 175, 184n.,217n., 229, 229n., 230n., 279

Dreier James, 18n., 280Driver, Julia, 235n., 238, 238n., 280Dworkin, Gerald, 286

Feinberg, Joel, 6n., 7n., 8, 8n., 9, 9n.,178n., 179, 179n., 185, 185n.,186, 186n., 232, 232n., 234n.,246n., 247n., 249, 249n., 280

Feldman, Fred, xiv, In., 2n., 4n., 12,12n., 18n., 21n., 24n., 31, 31n.,37n., 42n., 45n., 46n., 49n.,50n., 51n., 60n., 61, 61n., 64n.,74n., 75, 75n., 76, 76n., 77n.,78,97n., 114, 114n., 115,

n., 116, 116n., 118n., 121,n., 122, 122n., 125n., 128,

128n., 143, 143n., 145, 145n.,199n., 217n., 225n., 230n.,237n., 239n., 241n., 242, 254n.,255n., 259n., 260n., 261n.,263n., 279, 280, 287

Follesdal, Dagfinn, 73n., 280Foot, Philippa, 38n., 280Forrester, James, 127n., 280Forrester, Mary, 247n., 280Frankena, William K., 80n., 90n.,

91n., 280Frankfurt, Harry G., 52, 52n., 66, 67n.,

85, 85n., 86, 87, 88, 89, 280,287

Fried, Charles, 17n., 40n., 280

293

Gert, Bernard, 213n., 281Gewirth, Alan, 183, 184, 184n., 281Gibbard, Allan, 254n., 281Goble, Lou, 127n., 191n., 193n., 196,

197n., 198n., 202n., 204n.,205n., 206n., 281

Goldman, Alvin I., 55n., 97n., 281Goldman, Holly S., 31, 31n., 97n.,

117n., 136n., 189n., 191n., 281Gorr, Michael, 38n., 281Gowans, Christopher, 175n., 218,

218n.,281Greenspan, Patricia S., 114n., 121n.,

122n., 136n., 191n., 281

Haji, Ishtiyaque, xiiiHart, H. L. A., 7n., 178n., 281Henderson, G. P., 95n., 281Heyd, David, xiii, 233n., 234n.,

237n., 239n., 242n., 243n.,247n., 248n., 249, 249n., 281

Hill, Thomas E., Jr., xiii, 7n., 39n.,180n., 233n., 237n., 281

Hill, Thomas E., Sr., xiiiHilpinen, Risto, 73n., 280, 281Hintikka, Jaakko, 31, 31n., 144n.,

145n., 282Hoffman, Joshua, xiiiHohfeld, Wesley Newcomb, 176, 177,

177n., 282Hudson, Stephen D., 213n., 282Humberstone, I. L., 3n., 191n., 255n.,

259n., 282Hume, David, 17n., 282

Jackson, Frank, 4n., 16, 16n., 17, 17n.,116n., 189n., 191, 191n.,192n., 193n., 195, 195n., 196,196n., 197, 197n., 198n., 200,200n.,201,201n.,202,202n.,203, 203n., 260n., 261n., 282

Kagan, SheUy, 5n., 237n., 239n.,252n., 282

Kamm,. Frances Myrna, 234, 234n.,282

Kant, Immanuel, 17, 17n., 18, 19, 28,29, 30, 31, 38, 38n., 39, 40, 42,180, 180n., 233, 233n., 253n.,256,281,282

Kim, Jaegwon, 55n., 281, 282King, John, xiii

Landers, Ann, ix .Lebus, Bruce, 220n., 282Lemmon, E. J., 217, 218, 218n., 219,

219n., 282Lemos, Ramon M., 17n., 50n., 282Leplin, Jarrett, xiiiLewis, Carl, 46Lewis, David, 18n., 60n., 172n., 282Loewer, Barry, 173n., 282Lyons, David, 7n., 45n., 179, 179n.,

180n., 272n., 283

Marcus, Ruth Barcan, 2n., 70n., 209,209n., 218, 218n., 219, 219n.,220,220n.,221,221n.,222,222n., 223n., 283

Martinich, A. P., 166n., 283McCloskey, H. J., 50n., 173n., 283McConnell, Terrance C , xiv, 16n.,

44n., 59n., 93, 94, 94n., 108n.,142n., 175n., 176n., 217n.,225n., 230n., 241n., 262n.,275n., 283

McGinn, Colin, 7In., 283McKinsey, Michael, 31, 3In., 97n.,

131, 131n., 132, 132n., 136,136n., 137, 137n., 138, 138n.,260n., 283

Mele, Alfred, xiiiMellema, Gregory, xiii, 3n., 55n.,

234n., 237n., 242n., 247n.,249, 249n., 252n., 284

Mendola, Joseph, 60n., 284Mill, John Stuart, 39n., 284Montague, Phillip, 247n., 284Montefiore, Alan, 93, 93n., 94, 284Moore, G. E., 3n., 15n., 90n., 91,

91n.,92n.,239n.,241,241n.,242, 246, 252, 253n., 284

Moore, Robert E., 38n., 284

294

Nagel, Thomas, 18n., 284New, Christopher, 234n., 237n.,

239n., 241n., 284Nozick, Robert, 18n., 19n., 284

Parfit, Derek, 194, 194n., 257n.,260n., 272n., 284

Pargetter, Robert, 116n., 189n., 191,191n., 192n., 193n., 195,195n., 196, 196n., 197, 197n.,198n.,200, 200n.,201,201n.,202, 202n., 203, 203n., 282

Pietroski, Paul M., 59n., 145n., 173n.,175n.,213n., 214, 214n., 284

Pollock, John L., xiv, 147, 147n., 148,148n., 172n., 284

Postema, Gerald, xivPostow,B. C.,260n.,284Powers, Lawrence, 95n., 284Prichard, H. A., 3n., 11, lln., 12, 13,

13n., 14, 16, 16n., 17n., 20n.,41n., 43, 43n., 44, 44n., 50n.,56n., 92n., 97n., 284

Purtill, Richard L., 145n., 284

Quinn, Philip, xiv

Rainbolt, George W., 187n., 284Raz, Joseph, 243n., 246n., 284Reath, Andrews, xivRegan, Donald H., 254n., 257n., 258n.,

259, 259n., 260, 260n., 263,269, 270n., 278, 279, 284, 285

Richman, Robert J., 95n., 285Roberts, Robert C , 225, 225n., 285Robinson, Richard, 95n., 285Rosenkrantz, Gary, xivRoss, Alf, 72, 72n., 118n., 122, 285Ross, W. D., 3n., 5, 5n., 6, 6n., 7, 7n.,

11, lln., 12, 12n., 13, 16, 16n.,17n., 18, 20n., 43, 43n., 44,44n., 50n., 92, 92n., 93, 109,109n., 141, 141n., 142, 143,143n., 146, 150, 150n., 159,159n., 160n., 163, 163n., 164,164n., 166, 166n., 167, 168,169, 169n., 170, 172, 172n.,

173, 173n., 174, 174n., 177,181,207,208n.,209n.,220,232, 237, 252, 252n., 256, 285

Rousseau, Jean Jacques, 17n., 285Russell, Bertrand, 11, lln., 12n., 285

Sartre, Jean-Paul, 211, 211n., 212,220, 223, 285

Sayre-McCord, Geoffrey, xiv, 75n.,285

Scheffler, Samuel, 29n., 239n., 285Schwartz, Thomas, 264n., 285Sen, Amartya, 18n., 285Seung, T. K., 58n., 285Shaw, P. D., 81n., 285Sider, Ted, xivSimon, Caroline J., 244n., 285Sinnott-Armstrong, Walter, 58, 58n.,

82, 82n., 83, 83n., 84, 84n.,90n.,91,91n.,93,93n.,94,95n.,98n., 110, 111, l l ln. ,209, 209n., 213n., 214n., 219,220n., 285

Slote, Michael, 18n., 61n., 239, 239n.,250, 250n.,251,285

Smart, J. J. C , 239n., 241n., 285, 286Smith, Holly M., 262n., 286Snare, Frank, 170, 171, 171n., 172,

173, 173n., 286Sobel, Jordan Howard, 191 n., 254n.,

286Sorensen, Roy A., 60n., 286Sosa, Ernest, xiv, 236n., 238, 238n.,

279Stocker, Michael, 19, 19n., 95n., 98n.,

221,221n.,239n.,243n.,286Strawson, Galen, 14n., 286Stump, Eleonore, 235n., 286Styron, William, 220n., 286Swank, Casey, 213n., 286

Tannsjo, Torbjorn, 192n., 286Thomason, Richmond H., 108n.,

191n.,286Thomson, Judith Jarvis, xi, 13, 13n.,

17n.,23n.,27n., 181,186n., 262n., 284, 286

295

Tirrell, Lynne, xivTwain, Mark, 90

Unger, Peter, 66n., 286

Vallentyne, Peter, xiv, 19n., 28n.,55n., 59n., 62, 62n., 190n.,286, 287

van Fraassen, Bas, 213n., 218, 218n.,219, 219n., 223, 223n., 287

van Inwagen, Peter, 90n., 287van Rijen, Jeroen, 89n., 287Velazco y Trianosky, Gregory, xiv,

248n., 287

von Wright, Georg Henrik, 229, 230,230n., 287

Vorobej, Mark, 130, 130n., 287

Walton, Douglas N., 38n., 287Washington, George, 89, 90White, Alan R., 82, 82n., 90n., 91,

91n.,92n.,93,93n.,95n., 115,287

Widerker, David, 85, 85n., 86, 94, 287Williams, Bernard, 4, 4n., 5, 29, 29n.,

213n.,218,218n.,219,219n.,220,220n.,221,221n.,222,222n., 273, 274, 286, 287

296

Index of subjects

absolutism, 40accessibility, 45-50, 60-2, 79, 85, 86,

89, 108n., 109n., 167, 254-7,269

and choice, 25-6, 27n., 41-2,46-8, 59, 61n., 103, 178, 199,208-9, 212, 223, 271

of possible worlds, 25-6, 48, 276see also "can"; options

act-categories, 11-18act-evaluation, 20, 244-6, 247n., 250-1act-individuation, 203

see also actions, types and tokensact-oriented theories, 270actions

compound, 22, 23n., 34-5, 37, 63,66-8, 237

courses of, 24, 80, 240group, Ch. 9 passimtypes and tokens, 53-7, 164n.,

171-2, 236-7actualism, Ch. 6 passim

objections to, 191-6advice, 1,3, 13, 198-200,212agent-evaluation, 20, 247n.agents, moral, 27n., 261-2agglomeration, 68, 71, 212-14, 219,

275alternatives, 14-15, 18-19, 21-2,

85-9, 189-91, 195-6, 218-21,235-6, 239-41, 244n., 245-6,249-50, 252

see also options, personalanalysis, 75-8, 170, 172-3attitudes, 40, 263, 270n.augmentation, 121, 129, 144-5

beneficence, 7, 141, 143, 146, 150-7,161-4, 167, 171, 175, 208, 213,215, 217-8, 250-1

"best," see value, maximization ofbinds, 226-31, 275blame, 13, 16-17, 44n., 85-8, 93,

108-10, 131, 132n., 205, 224n.,247, 253, 263n.

and "can avoid," 81-2, 86-90,95-6see also options, personal

and control, 40-6, 48-53, 92-3,96-7, 248, 270n.

and freedom, 41, 46, 50-3, 66, 79,85-9, 90n., 224n.

and knowledge, 49-50, 204see also accessibility; "ought,"

and "can"Categorical Imperative, 18, 28-31,

116,256character, 20, 71, 90, 225charity, 38-9, 116, 179, 233, 243,

258-9choice, see accessibility, and choicecommensurability

intramoral, 6, 58-9, 61n., 132,134-5, 220-1

moral and nonmoral, 1—2, 6comparability, see commensurabilitycompunction, 174-5, 222conflicts of obligation, see dilemmas;

obligation, conflicts ofconjunction, see actions, compoundconnectedness, see disconnectedness

297

conscientiousness, 16-17, 95consequentialism, 18-19, 27, 30-1,

221, 237, 239, 241-2, 245, 250,252, 253n., 256

continuity, see value, continuity ofcontrol, see, "can," and controlcooperation, 42-5, Ch. 9 passim

counterfactual, 263—5procedural, 259-60unrequited, 258-9

culpability, see blame

deontic status, 14-15, 221, 233,237-8, 245-7, 249-52

deontic value, see value, deonticdeontology, 17n., 18-19, 27, 30-1detachment

deontic, 122, 128-30, 146-8,160-2, 168-9

factual, 114-17, 120-1, 124-31,133, 138, 144-6, 148, 160, 171,200-2

necessity, 121-2, 129, 145-8, 160,168-9

determinism, 45n., 46, 51, 66dilemmas, 117, 192-3, Ch. 7 passim

avoidance of, 221—2basic, 211-27, 229-30, 239, 266-8,

275group,275interpersonal, 229-30, 274-6nonbasic, 225-31, 275obligation-, 62-3, 209-10prohibition-, 62, 209-10subsidiary, 228-9, 231, 275

disconnectedness, 267—70, 275—6discretion, 239-44disjunction, 35-7, 62-6, 68-9, 72-3,

144n., 233, 236, 239-40divine command theory, 18—19division of values, see value, division ofduty

absolute, 15n.ceteris paribus, 5

conditional, 5, 143, 146, 163, 169correlativity to rights, 6—9, 39

see also rights, correlativity toobligations

perfect and imperfect, 38-40, 57,233-4

positive and negative, 38proper, 5, 141sans phrase, 5, 141

self-regarding, 180see also obligation

excuses, see justification, and excuseexperience, 211—12

fault, 257, 262, 263n., 269fidelity, 141-3, 146-7, 150-1, 155-64,

166-7, 170-1, 174, 181, 183,187, 208, 213, 215, 232

free-riding, 271freedom, 58, 174

see also "can," and freedom

gratitude, 59-60, 90, 92, 112-13,154-9, 163, 167, 170, 187, 208,256

guilt, 175, 223-5, 227-8

hedonism, 241

ideals, 2-4, 6n., 8-9, 71, 90-4, 248,252

see also "ought," binding and non-binding

intransigence, 198-9, 257-9, 261,266-7, 269-70, 276

justice, 18, 30, 163-4, 174justification

and excuse, 86, 93-5

lapses, 104, 225, 230-1, 236, 275-6see also wrongness and wrong-

doinglaziness, 189, 192-3, 203-4, 206liberty, see freedom; rights, libertiesluck, 44, 49, 110, 206n., 209lying, 89-90

making amends, see reparationmaximization, see value, maximization

ofmaxims, 19, 23, 31, 39, 42, 233

298

minimal difference, 147-50, 153-4,156, 158-60, 166-9

motives, 20, 42, 71, 92-3, 159n., 248

nonmaleficence, 7, 151—3, 156n., 157,161-4, 167-8

obligationabsolute, 5, 57, 141actual, 5, 141all else being equal, 145-60all things considered, 5-6cancellation of, 107-11, 230n.component, 141n.compound, 152, 161-2, 166n.conditional, 20, 72, 109, Ch. 4

passim, 145-7, 149-53, 156-7,159-60, 162-9, 170n., 181, 183,200-2

conflicts of, 141-4see also dilemmas

defeasibility of, see augmentationand descriptions, 56—7, 203

direct and indirect, 145, 147,149-64, 166-8, 181

dynamics of, 20, 75, Ch. 3 passim,125, 227-8, 236

extinction of, 113formal, 12group, 178n., 179n., 260-3, 270,

273-6hypothetical, 200-1immediate, 52, 96-113, 226in practice, 169, 170n., 174, 181in principle, 168-9, 170n., 174, 181incidental, 150-3, 155, 157, 161-3and indefiniteness, 57—63and intention, 17and knowledge, 17, 50, 108-10,

164-9, 173-4levels of, 131-40, 152n., 176material, 12mere, 10-11, 13, 20-1, 39, 175,

178, 180, 187moral and nonmoral, 1-2, 10necessary, 165-70, 181-2objective and subjective, 10—20overall, 5-10, 27, 39, 59, 95,

111-12,114,123,131,135,141, 143-7, 149-50, 152, 154,169-76, 181-4, 186-90, 200-1,207-37, 275

overriding of, 111-13, 135-6, 141,144, 155, 162-3, 170, 172, 181,188, 209, 219, 232

permanent, 165, 170n., 181prima facie, 5-10, 18, 20, 94, 111,

123, 136, Ch. 5 passim, 207-9,213, 215-20, 222, 232, 252

primacy of, 99-100, 102remote, 52, 96-113, 178, 226, 228residual, 142, 174-6, 222restoration of, 112-13restrictive, 200-2resultant, 141n.subsidiary, 20, 128-40, 202, 228-9,

231supersession of, 97, 111—13and timing, 105-6, 271-2

see also obligation, dynamics of-to, 7-10, 39, 175, 178-80, 187,

234, 238universal, 165-6, 170n., 181see also "ought"; paradoxes of

obligationopenness, 265—7

see also transigenceoptions

indeterminacy of, 60infinity of, 60-2moral, 27, 33, 59, 234-44, 246,

249-52, 271personal, 1, 15, 26-7, 31, 33, 40,

52, 60, 62, 65-8, 70, 89, 106,119-20,123,125,130,139,150-1,168,212,221,234-5,254

reordering of, 116-17, 201-2restriction of, 97, 119-20, 201

"ought"binding and nonbinding, 2—5, 9—10,

64, 91-2, 94, 174n., 246-7, 274and "can," 1, 26-7, 32, 40, 46, 71,

75, 77-100, 111-13, 136, 155-6,166n., 167, 179-80, 210-12,214-16, 219, 225-6, 259-60,265-6, 269-70, 275

299

"ought" (Contd.)ideal, see "ought," binding and

nonbindingiteration of, 63, 70-1and knowledge, 49-50and material implication, 63,

69-70, 118n.,123n., 144and "may," see permissibilitypassive-transformable statements of,

5,91and tense, 37, 82-4see also obligation

oversubscription, 249

paradoxes of obligation, 72-5, 191,122-8, 131

Chisholm's, 127-8, 134Gentle Murder, 127Good Samaritan, 73-5, 122, 151n.Knower, 74-5, 122, 126Reparation, 126-7Ross's, 72-3, 118n., 122Second Best Plan, 124-5SuzyMae, 95n., 122n.

perfection, 70-1permissibility, 32-7, 55, 59, 63-5,

68-70, 81-2, 152-4, 164, 186,209,213-21,238-9,247,250-1,271

perplexitiessecundum quid, 229, 230n., 275-6simplidter, 229

personal possibility, see accessibilitypossibilism, Ch. 6 passim

objections to, 196-203possible worlds, 31, 45, 147, 242

see also accessibility, of possibleworlds

praise, 87-8, 247-8preparedness to act, 42, 260, 264-6,

269probability, 190n.

see also value, probablepromises, 18, 38, 59-60, 110-11,

141-5, 166n., 174-6, 178, 222see also fidelity

punishment, 84

quandaries, 208-9, 212, 221

reasons, 150, 153, 155, 157, 168, 170,185, 213-14, 218-19, 221, 248,252

regret, 16, 206n., 222-4remorse, see guiltreparation, 126-7, 154, 156-9, 163-4,

184, 186-7respect, 17-18, 174, 180,245responsibility, 13-14, 16, 44n., 52-3,

84-8, 93, 217-18, 263n.tightness and doing right, 12, 15, 16n.,

29, 33-4, 53, 143, 172, 194,198, 208, 242, 244, 251-2,257-8, 261, 263

rights, 17-18, 23, 33n., 50, 156n.,176-88, 245

absolute and prima facie, 6-7,181-8

claims, 6-9, 176-88, 234, 238conditional, 183-5correlativity to obligations, 6—9,

176-80, 183, 187-8, 234see also duty, correlativity torights

incidental, 181infringement of, 181, 187legal, 177n., 185-6liberties, 177, 187, 238to life, 184-6overriding of, 181, 183-7violation of, 181

risk, 206, 234n.

satisficing, 250-1self-imposed impossibility, 72,

95-101,110,113,117,122,175

self-imposed necessity, 95-7, 105,113

Simple Assumption, 132, 133n., 134-8suberogation, 235, 237-8, 244, 246-8,

252-3supererogation, 3, 206, Ch. 8 passim,

258-9supervenience, 216-17

300

transigence, 263—6, 276unintrusive, 268-74

utilitarianism, 18, 50, 115, 143, 237,241

valueactual, 11,15-17, 109agent-relative, 18, 27-30, 159-60,

242-3, 256,276continuity of, 248-53deontic, 14-15, 19-20, Ch. 2

passim, 98n., 108, 118-20, 122,132-6, 139, 145-7, 149-50,153-4, 156-9, 162, 164, 167-8,170, 189-91, 195, 197,199-200,205,209,211,214,216, 219n., 221, 230, 244-9,251-2, 254, 256-7, 259-62,267-8, 272-4,276

division of, 244-53instrumental, 15, 159n.intrinsic, 18, 23, 27-8, 31, 75, 78,

159, 167, 241-2, 245, 252,256

limits of, 3In.maximization of, 18-19, Ch. 2

passim, 238-44, 246, 248,250-3, Ch. 9 passim

moral, 14-15

morally relevant, 15, 18, 30, 58-9,248-51see also value, division of

perceived, 11, 15-17, 31, 79-81probable, 11, 15-17, 79-81, 109,

205substantive views of, 15-19, 23,

27-8,30-1,50,139,277time-relative, 27-30, 159

vice, 30n., 247virtue, 30n., 159n., 247

wrongness and wrongdoing, 9, 12,14n., 15-16, 28-9, 33-6, 53,57, 61-2, 65, 70-1, 74, 86-90,93-4,98-105,107-11,126,152-3, 166, 172, 181, 191, 197,199,205,208-10,219-21,223-7, 229-31, 234-6, 238-44,246-7, 250-3, 260-3, 267, 270,275-6

causing of, 194—5compounding of, 128, 132-5, 202levels of, 132, 134, 138-40minimization of, 128, 134, 199sanctioning of, 193—5, 196n.seriousness of, 17, 29, 99, 138-40timing of, 100-5

wrongs, 33n., 181two making a right, 257-8, 262

301