The Concept of Moral Obligation (Cambridge Studies in Philosophy)

The principal aim of this book is to develop and defend an analysis of the concept of moral obligation. The analysis is neutral regarding competing substantive theories of obligation, whether consequentialist or deontological in character. It seeks to generate new solutions to a range of philosophical problems concerning the concept of obligation and its application. Among the topics treated are deontic paradoxes, the supersession of obligation, conditional obligation, prima facie obligation, actualism and possibilism, dilemmas, supererogation, and cooperation. By virtue of its normative neutrality, the analysis provides a theoretical framework within which competing substantive theories of obligation can be developed and assessed.

CAMBRIDGE STUDIES IN PHILOSOPHY

The Concept of Moral Obligation

CAMBRIDGE STUDIES IN PHILOSOPHY General editor Ernest Sosa Advisory editors Jonathan Dancy University of Keele Gilbert Harman Princeton University Frank Jackson Australian National University William G. Lycan University of North Carolina, Chapel Hill Sidney Shoemaker Cornell University Judith J. Thomson Massachusetts Institute of Technology

RECENT TITLES Andrew Newman The Physical Basis of Prediction David Owens Causes and Coincidences Jaegwon Kim Supervenience and Mind Warren Quinn Morality and Action Michael Jubien Ontology, Modality, and the Fallacy of Reference John W. Carroll Laws of Nature Henry S. Richardson Practical Reasoning about Final Ends Joshua Hoffman and Gary S. Rosencrantz Substance among Other Categories M. J. Cresswell Language in the World Noah Lemos Intrinsic Value Lynne Rudder Baker Explaining Attitudes Robert A. Wilson Cartesian Psychology and Physical Minds Michael Devitt Coming to Our Senses Barry Maund Colours

The Concept of Moral Obligation Michael]. Zimmerman University of North Carolina, Greensboro

CAMBRIDGE UNIVERSITY PRESS

CAMBRIDGE UNIVERSITY PRESS Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521497060 © Cambridge University Press 1996 This publication is in copyright. Subject to statutory exception and 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 1996 This 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 ISBN 978-0-521-49706-0 hardback ISBN 978-0-521-03874-4 paperback

CIP

Contents Preface Acknowledgments

page ix xiii

1

Groundwork: Some distinctions 1.1 Moral and nonmoral obligation 1.2 Binding and nonbinding senses of "ought" 1.3 Overall and prima facie obligation 1.4 Objective and subjective obligation 1.5 Act-evaluation and other types of evaluation

1 1 2 5 10 20

2

Moral obligation: An analysis 2.1 Value maximization 2.1.1 Two problems 2.1.2 The analysis 2.1.3 Extension of the analysis 2.1.4 Perfect and imperfect duties 2.2 Underlying metaphysical and conceptual issues 2.2.1 "Can" and control 2.2.2 "Can" and accessibility 2.2.3 "Can" and knowledge 2.2.4 "Can" and control again 2.2.5 The ontological status of actions 2.2.6 Indefiniteness 2.3 Implications of the analysis 2.3.1 Certain deontic implications 2.3.2 Certain deontic paradoxes 2.3.3 Analysis and neutrality

21 21 21 25 31 38 40 40 45 49 50 53 57 63 63 72 75

3

The dynamics of obligation 3.1 "Ought" and "can" 3.1.1 An internal criticism 3.1.2 "Can" and "can avoid"

79 79 79 81

3.1.3 An alleged redundancy 3.1.4 Past and present obligations 3.1.5 Alternate possibilities 3.1.6 Psychological incapacitation 3.1.7 The obligatory and the ideal 3.1.8 Excuses 3.1.9 Self-imposed impossibility 3.2 Immediate and remote obligation 3.2.1 Self-imposed impossibility 3.2.2 The time at which a wrong is done 3.2.3 The time at which an obligation is satisfied 3.2.4 The cancellation of obligation 3.2.5 The supersession of obligation Conditional obligation 4.1 Varieties of conditions on obligation 4.2 The analysis 4.3 Certain deontic paradoxes 4.4 Subsidiary obligation 4.4.1 Factual detachment 4.4.2 Levels of obligation 4.4.3 Overriding 4.4.4 A rival account 4.4.5 Level of wrongdoing and seriousness of wrongdoing 4.4.6 Level of wrongdoing and level of obligation

82 82 85 89 90 93 95 96 97 100 105 107 111

Prima facie obligation 5.1 A tale of two obligations 5.2 Defective analyses 5.3 The analysis 5.3.1 "All else being equal" 5.3.2 Three objections 5.3.3 Detachment 5.3.4 Overriding 5.4 Ross on prima facie obligation 5.4.1 Conditional and unconditional prima facie obligation 5.4.2 The self-evidence of prima facie obligations 5.4.3 Prima facie obligation and overall obligation 5.4.4 Residual obligation 5.5 Rights 5.5.1 The correlativity of rights and obligations

141 141 143 145 145 154 160 162 163

VI

114 114 117 122 128 128 131 135 136 138 139

163 164 169 174 176 176

5.5.2 Absolute and prima facie rights 5.5.3 Rights and overall obligation

181 187

6 Actualism and possibilism 6.1 The dispute 6.2 Objections to actualism 6.2.1 Certain theses denied 6.2.2 Sanctioning wrongdoing 6.2.3 A nondiachronic case 6.2.4 A slightly amended version of actualism 6.3 Objections to possibilism 6.3.1 Arbitrariness 6.3.2 Advice 6.3.3 Detachment 6.3.4 Act-individuation 6.4 What it is best to do

189 189 191 191 193 195 195 196 197 198 200 203 203

7 Dilemmas 7.1 The nature of moral dilemmas 7.2 The case against basic moral dilemmas 7.3 The case for basic moral dilemmas 7.4 Nonbasic moral dilemmas

207 207 211 217 225

8 Supererogation 8.1 The nature of supererogation 8.1.1 Going beyond prima facie obligation 8.1.2 Going beyond perfect obligation 8.1.3 Going beyond overall obligation 8.2 The possibility of supererogation 8.2.1 Supererogation and discretion 8.2.2 The division of values 8.2.3 Continuity

232 232 232 233 234 237 239 244 248

9 Cooperation 9.1 The problem 9.2 Nonsolutions 9.2.1 Unrequited cooperation 9.2.2 Procedural cooperation 9.2.3 Group obligation 9.2.4 Counterfactual cooperation 9.2.5 Mere openness

254 254 258 258 259 260 263 265

vii

9.3 The solution: Unintrusive transigence 9.4 Interpersonal moral dilemmas Postscript Appendix: List of propositions List of works cited Index of names Index of subjects

268 274 277 278 282 293 297

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-yearold divorced man for five years. "Jack" and I were very much in love. I wore his engagement ring for a year, and we talked of marriage, but we bickered a lot. 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 we were made for each other. I postponed the wedding because Jack wouldn't continue with the counseling. Also, he wanted to elope, and I wanted a church wedding. 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. Jack insists that he doesn't love her, but she cooks and cleans for him, which I never did. Ann, we've been seeing each other behind the secretary's back. We agree that we love each other, but he's afraid he hasn't the will-power or self-discipline 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 were reading it, a slight smile played over your lips. I hope so; the rest of this book is dead serious and will afford very little opportunity for amusement. The second reason is this. A preface should let the reader know what to expect from a book, and I want you to know that you should not expect an answer to Indianapolis's question, even if the "should" in her "What should I do?" is a moral one. Nor should you expect answers to any questions remotely like hers. Some people seem to think that the task of moral philosophy 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 no further. Once, many years ago, I wrote a paper on the relation between intrinsic value and the appropriateness of pleasure, and I submitted it to a leading philosophy journal. One referee recommended rejection, saying: Actual cases of moral questions and issues are hardly mentioned; yet what else is morality about except Joan wondering whether to leave her husband or have an abortion, Albert wondering whether to tell his grandmother what her doctor just told him, or whether a new bike would spoil his son? That Albert and Joan would find this paper totally irrelevant to their dilemmas, even if they were philosophers and understood it, seems to me to sit very differently than the idea of the ordinary man who knows his coffee-cup is real not being interested in The Theory of Knowledge or Science, Perception, and Reality... If I were advising

Joan, I would tell her to go and read The French Lieutenant's Woman, and not this paper, and I consider that a wholly legitimate criticism of a piece of philosophical ethics. I hope you are as appalled by this as I still am. (If not, then, again, please read no further.) I have found that this referee's attitude is by no means an isolated one, but I find it quite baffling. Joan's situation, whatever it is, may be a difficult one, and certainly moral philosophers may legitimately try their hands at resolving it; but it had never been my intention in the paper even to pretend to be concerned with the resolution of such matters, and to claim that such resolution is the only legitimate task of moral philosophy is utterly preposterous. At any rate, this book attempts something quite different and much less ambitious. A great deal of contemporary moral philosophy is concerned with finding answers to such general substantive questions as what it is that makes right acts right, what the various virtues and their interrelations are, what constitutes a proper excuse for vile behavior, and so on. These are all very important questions, and moral philosophers (some of them, anyway) are especially well equipped to handle them. But, again, it is not with such ambitious questions as these that this book has to do. No, the focus of this inquiry is a concept: the concept of moral obligation. The main question that is addressed is not what it is that Joan (or Albert, or Indianapolis) ought to do in particular, nor what it is that makes an act obligatory in general; it is, rather, what it is for an act to be obligatory. The task that I have set myself is thus one of conceptual cartography, of mapping the contours of the concept of moral obligation. The goal is simply 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. Although you won't find in it an answer to Indianapolis's predicament, you will find answers to these questions (and many more besides): What is the relation between "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 someone 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 one to be perfect? Does "ought" imply "can"? If so, can someone divest himself of an obligation simply by rendering himself unable to fulfill it? Can someone succeed in sloughing off an obligation through sheer laziness? Can someone be obligated to do something he cannot avoid doing? Can someone have an excuse for doing something that it was not wrong for him to do? Can something that it was obligatory to do become wrong? Can something that it was wrong to do become obligatory? Are there gradations of obligation and wrongdoing? Can there be obligations without rights? Can someone be obligated to do, or to cause himself to do, wrong? If one person ought to advise another to do something, ought the latter to do that thing? Can someone (Indianapolis, say) be in a genuine moral dilemma? What is the relation between wrongdoing and guilt? Can one go beyond the call of duty? Can it be right to prevent someone from doing what's right? 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 customary eloquence, exactly what I would want to say: On the other hand, "she" is no better. Indeed, it is in two ways worse. In the first place, prose should be transparent, like a pane of glass through which one sees the thoughts behind it; the use of "she" for these purposes is like a smudge on 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 think entirely right; but it is annoying to have that moral point introduced (with the back of the hand, as it were) into matters it has no connection with — one feels nagged. 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 what a 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 system of numbering that is adopted in the text. Numbers will be used to refer to chapters, sections, subsections, charts, figures, and tables. The practice that I shall adopt is illustrated in the following examples: Chapter 1, Section 1.1, Subsection 2.1.1, Chart 1.1, Figure 2.1, Table 1.1.1 hope this practice is self-explanatory. In no case will such numbers be used alone; in every case it will be explicitly stated whether reference is being made to a chapter, section, chart, and so on. I shall also use numbers to refer to propositions (and portions thereof). Here the numbers will stand alone, but they will always appear within parentheses. (Thus, for example, "(1.1)" refers to the first proposition so treated in Chapter 1.) When, in referring to propositions, I make use of a prime or primes, this indicates a variation on some other proposition. (Thus (2.41') is a variation on (2.41).) When I make use of a lower-case letter, this indicates a variation on part of some other proposition. (Thus (2.41'a) is a variation on part of (2.41').) In addition I have employed the capital Roman numerals (I) — (XVI) to refer to certain core analyses of concepts having to do with obligation; (I) — (XII) are introduced in Chapter 2, (XIII) - (XVI) in Chapter 4. Finally, in Chapter 1 I also use the capital letters (A) - (E) to refer to certain portions of propositions that I there call "act-categories." All of this is, I confess, rather cumbersome, but I 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 from others in the writing of this book. I thank the students in my graduate seminars on obligation given at Brown University in 1984 and 1987. They helped me deal with the material in its early stages of development. I was awarded a research leave by the University of North Carolina at Greensboro for the fall semester of 1993. This enabled me to complete a first draft of the book. I believe I would still be engaged in writing the book had I not received this leave, and I am very grateful for it. Several anonymous referees, both for Cambridge University Press and for certain journals, provided penetrating criticism and useful advice concerning portions of the manuscript. I am very thankful for this help. I am thankful, too, for permission to reproduce (often with modification) portions of previously published works. These works are listed in the List of Works Cited as: Zimmerman (1986) - copyright 1986 by D. Reidel Publishing Company, reprinted by permission of Kluwer Academic Publishers; (1987) - reproduced by permission of the editor of the American Philosophical Quarterly; (1988b) - reproduced by permission of the editor of Philosophical Papers; (1990b) - copyright 1990 Kluwer Academic Publishers, reprinted by permission of Kluwer Academic Publishers; (1992) - copyright 1992 Kluwer Academic Publishers, reprinted by permission of Kluwer Academic Publishers; (1993b) - reproduced by permission of the editor of the American Philosophical Quarterly; and (1995a) - copyright 1995 Kluwer Academic Publishers, reprinted by permission of Kluwer Academic Publishers. I received a great deal of help from a number of other people. It is a distinct 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., Joshua Hoffman, John King, Jarrett Leplin, Alfred Mele, Gregory Mellema,

John Pollock, Gerald Postema, Philip Quinn, Andrews Reath, Gary Rosenkrantz, Geoffrey Sayre-McCord, Ted Sider, Ernest Sosa, Lynne Tirrell, and Gregory Velazco y Trianosky. Finally, there are four individuals whom I wish to set aside for special mention, since their contribution to the present work has been so extensive. Peter Vallentyne read a version of the entire manuscript and provided me with numerous detailed comments. He stopped me from making a number of errors and pointed the way to a number of improvements, for all of which I am most thankful. Terrance McConnell read various versions of the entire manuscript, some portions several times, and always provided helpful and judicious comments. He has set himself a dangerous precedent. As an expert on gratitude, he knows how grateful to him I should be. I think that in this instance, if not in others, I have met the demands of morality. Roderick Chisholm has been a source of tremendous inspiration to me throughout, and indeed prior to, my professional career. He gave me sound advice and staunch encouragement when I needed them most. His support has been invaluable, and I extend to him my deepest thanks. Fred Feldman has shaped this work more than any other person. His work on obligation is the source of a great many of the ideas expressed in this book; anyone acquainted with his work will immediately see how much I have lifted practically wholesale from it. In addition, he is a wonderful teacher and an extremely incisive critic ("cruel but fair," as Monty Python might say). Even though he will doubtless find much still to criticize in this book, it is to him that I dedicate it with great gratitude and respect.

xiv

1 Groundwork: Some distinctions

I believe that one ought to do the best one can. You may be inclined to disagree. If you are, I think it's because you don't yet know what I mean. Let me try to explain. 1.1 M O R A L AND N O N M O R A L O B L I G A T I O N I am using the term "ought" to express moral obligation. It is of course sometimes used to express what is required not by morality but by prudence, or law, or aesthetics, or the rules of chess, or whatever. "You ought not to have used so much red," for example, is likely to constitute aesthetic rather than moral criticism. "You ought to watch your pennies" is likely to constitute prudential rather than moral advice. This well-recognized fact gives rise to a profound puzzle that I shall mention only to put aside. Suppose that you have investigated your options and reached the disconcerting conclusion that, although morality requires you to do one thing, prudence requires another, and the law yet a third (and aesthetics 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. What ought I to do?" But what do you mean? "What ought I to do from the moral point of view?" Surely not, for you've already settled on an answer to that question. Similarly, you've already settled on answers to the corresponding questions having to do with the various nonmoral points of view. Presumably, then, you mean, "What ought I to do, period?" This is very puzzling. It appears to presuppose that the various points of view, moral and nonmoral, are in some manner commensurate with one another, so that they may be weighed one against the other and your question receive a 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 various points 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 understand how it could be that the various points of view, moral and nonmoral, are commensurate with one another. But perhaps this is just a deficiency in me. At any rate, as I said, I shall put the matter aside, for even if it makes sense to say on occasion that one ought, period, not to do what one ought morally to do, this would of course not alter the fact that one still ought morally to do what one ought morally 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 do the best one can (in some sense), but only to assert that morally one ought to do the best one can (in some sense to be elaborated on later). It may well be that nonmoral obligation is, in many instances, to be analyzed in terms similar to those in which moral obligation is to be analyzed.2 Indeed, it would be surprising if this were not so. But whether or not it is so is not my concern here. 1.2 BINDING AND NONBINDING SENSES OF "OUGHT" "Ought" doesn't always express requirement or obligation, even within morality. Sometimes it expresses an ideal, a desideratum rather than a moral necessity. Here we might use "should" instead of "ought"; we would probably not use "must." Examples of this use of "ought" are: "Everyone ought to be happy," and "Little children ought not to have to suffer." Sometimes the ideal, nonbinding sense of "ought" is called the "oughtto-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 discussing what 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 be happy," and "It ought to be that little children not have to surfer"; whereas, when discussing what is obligatory, we can express our thoughts in the following way (to employ fresh illustrations): "Tim ought to do what he's been told," and "Pru ought to do what she promised." The trouble is that the "It ought to be that" locution can be applied to what is done, too (as in: "It ought to be that Helen do what she can to help out"), and anyway it is surely arguable that certain actions are sometimes morally desirable but 2 3

Cf. Feldman (1986), Chs. 5 and 7. 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 be something that is done (as evidenced in: "Mary ought to be with her ailing mother").5 It is better, therefore, to talk of what is binding or obligatory and 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 derivative from the binding sense (as in, perhaps: "God ought to prevent the suffering 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 has to say about supererogation: "The status of supererogation might be summarized in this way. If I seek advice, concerning an act which would be supererogatory, 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, there are 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 are morally required to do so, but no one has any moral claim on you that you do so." Fair enough. (I shall have more to say about the latter possibility in a moment.) What I don't see is that someone cannot also quite properly mean, "It would be morally desirable if you were to do so, but you are not morally required to do so (or even to try to do so, or to do something related to doing so, or...)." Here, though, someone might object: "What can it mean to say that I ought morally to do something if I am not morally obligated to do it?" The idea underlying this objection is not that it is improper to use "ought" in an ideal, nonbinding sense, but that such an "ought" cannot be said to be a moral "ought." If "Little children ought not to have to suffer" implies no moral obligation, how is its "ought" a moral one? The objection concerns the proper extension of the term "moral." This is a huge issue, which, once again, 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 5

Cf. Mellema (1993). 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 there are certain goals which, from a moral point of view, are ideal or desirable but which do not imply any obligation on anyone's part. Whether or not this is legitimate, however, is not germane to the present inquiry. However the ideal "ought" is to be analyzed9 and whether or not it is properly said to be a moral "ought," I shall confine my attention to the "ought" of moral obligation. But ought the ideal "ought" to be so casually set aside? Might it not be that, 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 moral obligation were to be so analyzed, the notion of an ideal "ought" could not be 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 any reference to the ideal "ought." Might there nonetheless not be some interesting connection between these two "oughts," so that a full account of the "ought" of obligation would still require that greater attention be given to the ideal "ought" than I 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 is some 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 well seem rash. But I don't think that he is right. An indication that something has gone awry in the derivation of (1.3) from (1.1) is that, as Williams intends 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 fact but 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 equivalent as follows: (1.1') that old lady ought to be helped by someone; but (1.3) cannot be similarly transformed without loss of meaning as follows: (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-transformable or as both being (more on the former possibility in Chapter 9), and the argument's validity might thereby be salvaged, but this will not provide Williams with the link that he wants between the binding and nonbinding senses 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 moral obligation. Ross himself used a variety of terms to refer to these species. On the one hand there is what he called "absolute obligation," "actual obligation," "duty proper," or "duty sans phrase" \ on the other hand there is what he called "prima facie obligation," "prima facie duty," "conditional duty," "duty ceteris paribus," or "what tends to be duty."12 None of these terms is particularly felicitous, as Ross himself acknowledged. More appropriate, perhaps, is to talk of "overall obligation" or "obligation all things considered" for what Ross meant by "absolute obligation," and this is what I shall do. It would also be more appropriate to talk of "obligation some things considered" (or "in some respects") or of "pending" or "pro tanto"13 obligation for what Ross meant by "prima facie obligation." Nevertheless, since the term "prima facie obligation" is by now so well entrenched, it will be 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 where a 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 as others of them are concerned. What one ought on balance to do will, presumably, be some function of these various "insofar as" "oughts"; to put it differently, what one's overall obligation is will be some function of what one's various prima facie obligations are (if any). It is worth repeating that what is presently at issue is two species of moral obligation. 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 that it has to do with what one ought to do all morally relevant things considered.14 The question that was briefly raised above concerning the commensurability of various points of view, both moral and nonmoral, is not at 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 of different moral considerations; there is. It is one that I shall address in the next chapter.) My primary concern in this work is to provide an account of overall moral obligation, although I shall also provide an account of prima facie obligation in Chapter 5, an account according to which it is closer to the truth to say that prima facie obligation may be understood in terms of overall obligation rather than the reverse.15 Sometimes moral obligations or duties are said to correspond to moral rights or claims. Ross distinguishes four different theses having to do with such 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 prima facie obligation (and hence, presumably, "right" expresses a prima facie right), the other where "duty" expresses an overall obligation. Ross him14 Indeed, it is to say that it has to do with what one ought to do all moral-obligationrelevant things considered, inasmuch as some moral considerations, as noted in the last 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 true but that, since we can have duties to (nonhuman) animals but no rights against them, (1.7) is false.17 Our having duties to animals but their having no duties to us also leads Ross to note that not both (1.5) and (1.6) can be true. He is uncertain which to reject but is inclined to say that animals, not being 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 that we have duties towards or concerning animals, but do we have duties to them in the requisite sense? Ross thinks so, saying: "I suppose that to say we have a duty to so-and-so is the same thing as to say that we have a duty, grounded 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 or something is too broad. It would appear to imply that certain potential beneficiaries (such as the homeless) have a right not just to nonmaleficence but to beneficence, if it is agreed that we have a duty to act not just nonmaleficently but also beneficently towards them. Although some may not balk at this (either because they reject the view that we have a duty to act beneficently towards such potential beneficiaries, or because, like Ross, they accept the view that such persons have a right to our beneficence), others will balk at it and claim that it is possible to have a duty towards or concerning somebody or something without owing that duty to that person or thing.21 In this way one might try to support not just (1.4) and (1.6) but also to defend (1.5) and (1.7) against Ross's argument. For one could argue that all and only moral agents are such that duties are owed to them (in the requisite sense), and all and only moral agents are such that they owe duties to others. 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 believe that there is indeed a difference between having a duty towards or concerning someone or something and having a duty to someone or something. The latter, I believe, is correlative to a right held by the someone or something in question, whereas the former is not. Hence I accept both (1.4) 17 18 19 20 21

Ross (1930), pp. 48-9. Ross (1930), pp. 49-50, 53-4. Ross (1930), p. 49. Ross (1930), pp. 52-3. 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 such contexts.

and (1.5). But I reject both (1.6) and (1.7), because, contrary to Ross and to the view just mentioned, I'm inclined to think that, although one must be a moral agent to have duties to others, one need not be a moral agent to have rights against others. In particular, certain animals appear to me to have rights 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 obligation to someone (or something) to do that thing, it is not possible to be obligated to do something without one's owing that obligation to someone to do that thing. Consider this example, provided by Joel Feinberg: [CJonsider how the word "ought" differs from the word "duty." Suppose a stranger 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 I should... Now suppose that Jones is on the street corner and another stranger politely 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 for human conduct under such circumstances; ... that it was not what Jones ought to have done. If we reproach Jones, however, ... he may present us with a vigorous selfdefense. "Perhaps I was not civil," he might admit. "But surely I was under no obligation to give a match to that man. Who is he to me? He had no claim on me...; I don't owe him anything. It may be nice to do favors for people; but a favor, 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] verything Jones said ... was true. The moral I draw from this tale is that there are some actions which it would be desirable for a person to do, and which, indeed, he ought to do, 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 X is 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 positions to adopt concerning Jones's behavior (let alone Jones himself) in this story. Even if we disregard the distinction between prima facie and overall obligation, still eight different positions suggest themselves. Using the terminology 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 23

Feinberg (1968), pp. 392-3. Feinberg (1968), p. 393.

Table 1.1

Position 1 2 3 4 5 6 7 8

Jones ought, in the nonbinding sense, to do A

Jones ought, in the binding sense, to do A

Jones has an obligation to the stranger to do A

yes yes yes yes no no no no

yes yes no no yes yes no no

yes no yes no yes no yes no

may be immediately dismissed, since one cannot have an obligation-to to do 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 not adopt any position according to which Jones has an obligation to the stranger to give him a match (where this is understood to mean that Jones owes it to the stranger to do so), and so he does not accept any of positions 1, 3, 5, or 7. It seems clear, too, that he does not accept any of positions 5-8, 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, and I don't know which of these it is that Feinberg accepts. His remarks seem to 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 give him a match, but that something has gone awry, morally, when Jones fails to give him a match, still it is hard to know which of positions 2 and 4 to accept. (Position 2, with its binding "ought," implies that Jones did wrong in not giving the stranger a match, whereas position 4 does not.) Of course, if one rejects the possibility that there be a binding "ought" without there being an obligation-to, then one will reject position 2. As before, however, I raise all this only to put it aside. Feinberg wishes to reserve the term "obligation" for what I have called "obligation-to"; if he 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 the stranger a match, due to considerations of health. It is worth noting that Feinberg first presented his example in 1960, long before health-related condemnation of smoking became common.

that Jones has an obligation, but not an obligation-to, to give the stranger a match. That is fine by me. I choose, however, to use the term "obligation" simply as a nominalization of the binding "ought" and to say that, even

on position 2, Jones has an obligation to give the stranger a match. One must adopt a certain terminology, and this is the terminology that I, like many others, adopt. What matters in the end is not what terminology one adopts but that one get clear as to what the concepts at issue are. That is what I am trying to do. My purpose in this work is to clarify the concept expressed by the binding "ought." It is important to recognize that, for present purposes, I am neither presupposing nor denying the view that position 2 (or position 6) is possible, that is, that it is possible that there be an obligation (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 possible). My concern here is to provide an account of obligation without venturing into this territory of "to," that is, to provide an account of what may be called "mere obligation." If it is possible to have an obligation that is not an 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 it is not, still every obligation-to is of course an obligation, and my account can then be construed as being designed to say at least to that extent what obligation-to amounts to. 1.4 OBJECTIVE AND SUBJECTIVE OBLIGATION So far I have, in effect, been noting the ambiguity of "ought" and trying to narrow down that sense of it that is my main concern here. This process of specification may be represented as in Chart 1.1.25 We have made our way moral binding overall mere obligation

nonmoral nonbinding

prima facie obligation-to

Chart 1.1 25 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, nonmoral obligation may be subdivided into binding and nonbinding obligation, nonbinding obligation may be subdivided into overall and prima facie obligation, and prima facie obligation may be subdivided into mere obligation and obligation-to.

10

to mere obligation. Does this suffice? According to some, it does not; there is, they allege, ambiguity even here. T o assess this claim of ambiguity, consider a distinction that Bertrand Russell draws between different categories of acts. W e may distinguish, he says, 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 fortunate," while the second is the "wisest." 26 Ross and H. A. Prichard draw a similar distinction between (1.10) that act which best suits the objective features of the agent's situation (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 situation (that is, which best suits the agent's situation as the agent conceives his situation to be). 27 If we ignore the question as to whether what makes an act "best" has to do with 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 claimed that "ought" is ambiguous, even when it has been narrowed down to mere obligation. 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), and what 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 differently. For example, Fred Feldman says that it is (A) and not (B) that corresponds to objective obligation and that what corresponds to subjective obligation is neither (C) nor (D) but this: (E) that act which it is most reasonable for the agent to believe is objectively obligatory.29 Others have used somewhat different terminology. C. D. Broad, for instance, distinguishes between what he calls "material obligation," "formal obligation," and "subjective obligation," saying that (A) corresponds to what is materially obligatory, whereas (C), read in one way, corresponds to what is formally obligatory and, read in another way, corresponds to what is subjectively obligatory.30 And it might seem that Ross and Prichard would 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 to deny 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 the objective features of the situation, and sometimes that it suits the subjective features. And when people express different opinions about the rightness or wrongness of an act, the difference is often due to the fact that one of them is thinking of objective and the other of subjective rightness. The recognition of the difference between the two is therefore in itself important as tending to reconcile what might otherwise seem irreconcilable differences of opinion. But the question remains, which of the characteristics — objective or subjective rightness — is ethically the more important, which of the two acts is that which we ought to do.31 Here Ross seems to grant the distinction between objective and subjective obligation (or, as he puts it in the quoted passage, "rightness"; he draws no distinction between rightness and obligatoriness) only to snatch it away again by asking what the agent (really) ought to do. Prichard is more force28 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 mistaken and the latter correct.32 He thus denies that (A) captures a (proper) sense of "obligation," insisting that only (C) does. More recently, Judith Thomson has also insisted on unambiguity here, although her view is that there is no subjective sense of "ought" (which in her view would, ifit existed, correspond to a variation on (D)).33 But others have insisted just as strongly 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 perhaps other act-categories) corresponds to some distinct sense of "ought." Accepting that (A)-(E) are distinct (as one ought to do) does not require accepting that each expresses some sense of "ought." Still, I grant that more than one sense of "ought" may be at issue here. Consider the claim, which is 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 he ought 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 the advisor takes the advisee, despite the latter's uncertainty, to be a fine judge as to what he ought to do and is simply telling him that, in this case as in others, his (the advisee's) judgment can be trusted. At other times, however, the claim seems to be offered as an account of what it is to be such that one ought (in some sense) to do something. Here the claim seems to be this: one ought (in some sense) to do what one thinks one ought (in some other 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 that there is any sense of "ought" such that one ought (in that sense) to do what one thinks one ought (in some other sense) to do? Is it appropriate ever to speak in this way? I'm inclined to think not. It seems to me that anyone who puts matters this way would do far better to put them as follows: one is to blame if one fails to do what one thinks one ought (in the only 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 is only one sense of "ought" (at this level of inquiry — the level of "mere obligation") 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 that there are two senses of "ought" here, we need only say in reply that we are presently concerned just with one of them, namely, that in terms of which 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 an account 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 inclined to deny the ambiguity of "ought" at this level. I don't like it because it presupposes that "ought" (in the sense that we are concerned with) is in some important way "objective" (whatever that may mean, exactly), and, although I do believe this, it is not something that I wish to presuppose for present 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 understood in 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 that Prichard is right and that it is to be understood in terms of (C); and it would be 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 by saying that one ought to do the best one can. Surely, you may think, this corresponds to (A) and not (C) (and not (B) either), and so the "ought" that I'm after is properly called objective. But if you think this, you probably still 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 do the best one can, I am contending merely that there is some way in which what one ought to do is superior to any other of one's alternatives. I shall call such superiority deontic. My claim, then, is that, to have the deontic status of being obligatory, an act's deontic value must be greater than that of any other alternative action. (Note that I am here restricting the use of "deontic" to the moral realm, whereas others sometimes use the term more widely. An act's deontic value, then, although not necessarily itself a species of moral value, is that which, in conjunction with the deontic values of its 35 36

Zimmerman (1988a) contains a detailed elaboration of this view. 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, one cannot 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 incoherent, 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 will emerge shortly, I shall steadfastly decline to specify what it is that deontic value consists in; I wish merely to commit myself to the formal view that one 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 still very 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 liberal, construal of "best." To see this, suppose that we adopt, merely for the purpose of illustration, the position that "best" in (A), (B), and (C) refers to the act's instrumental value. Then, according to the view that "ought" corresponds to (A), what is deontically superior is what is instrumentally best, while according to the view that "ought" corresponds to (B) what is deontically superior is what is probably instrumentally best, and according to the view that "ought" corresponds to (C) what is deontically superior is what is believed instrumentally best by the agent. On this understanding of "best," then, the view involving (A) incorporates the substantive moral view that deontic value just is actual instrumental value, while the view involving (B) incorporates the substantive moral view that deontic value just is probable instrumental value, and so on. Insofar as what is actually instrumentally best need not be what is probably instrumentally best, which itself need not be what is believed instrumentally best by the agent, these three views will on occasion issue in three different deontic rankings of the agent's options. Nonetheless, each of these substantive views is a species of the purely formal view that an act is obligatory if and only if its deontic value is greater than that of (that is, the act is deontically superior to) any of its alternatives. Note how modest this formal view is. It merely states that what is obligatory is unique relative to its alternatives and that this uniqueness constitutes some sort of morally pertinent superiority; there is no commitment to any substantive view as to what this superiority consists in.37 In this book I shall 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, something else) that "ought" corresponds. 37

Cf. Moore (1903), p. 147: "It is plain that when we assert that a certain action is our absolute duty, we are asserting that the performance of that action at that time is unique in respect of value." Of course, Moore is here talking about the instrumental value of the act in particular, whereas, as I have just tried to explain, I am not committing myself to the 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 as I have noted, and Ross, persuaded by Prichard, abandoned his earlier view (which was that (A) captured what "ought" consists in)39 in favor of Prichard's view.40 The reasons given by both Prichard and Ross seem quite inadequate, but there is no need to go into this here, since the present account does not presuppose that "ought" does not correspond to (C).4 I suspect that, frequently at least, someone who is attracted to (C) in this regard confuses it with (D) and also fails to distinguish between blame worthiness (and other concepts having to do with moral responsibility) on the one hand and wrongdoing (and other concepts having to do with moral obligation) 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 a conscientious person, who is trying to determine what he ought to do, seeks to discover: that act which is best (in some substantive sense) or that act which is probably best? There is reason to think the former. Suppose that he believes that he has a duty to help those in distress and that he comes upon the victim of a car accident. He must choose whether to leave the victim where he is or to pull him free from the wreck. Suppose that he pulls the victim free, believing that this is probably best. Does he believe that he is doing what he ought to do, or does he believe only that he is probably doing what he ought to do? Quite possibly the latter. After all, he may later discover that it would have been better to leave the victim where he was and thus come to believe, regretfully, that (however innocently, and however much the evidence pointed the other way) he in fact did the wrong thing. 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 proposed 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 but not cure the ailment, while Y will either cure it or kill the patient and Z will either cure it or kill the patient; but Jill doesn't know which of Y and Z will cure and which kill.42 If "ought" corresponds to (A), then presum38 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 ought to 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 asking just one question when she says: "What is it that I ought to do in this case?" But if there is ambiguity, then I shall simply say this: the "ought" that I'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 believes he 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 compatible with each of them. In fact, it is compatible with a wide variety of other views 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 perhaps some other view is more plausible still. Certainly other views have been proposed. According to some philosophers, what one ought to do is a function 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 someone of whose plight one is nonnegligently ignorant).46 According to other philosophers, what one ought to do is a function of what one intends, and this may or may not have anything to do with what is, or is probably, or is believed to be, best (in some less-than-fully-liberal sense). Kant is, of course, the obvious figure to cite here,47 but one needn't be a Kantian to agree that intended harm constitutes a more serious wrong than unintended harm - a matter of adding insult to injury48 - and it is difficult to subsume this satisfactorily under either (A) or (B) or (C), at least as these would ordinarily be understood.49 Again, many philosophers have contended that what one ought to do is a function of respecting persons' rights;5 others have 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 principle which, as commonly interpreted, concerns the (im)permissibility of acting in certain ways. 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 try to subsume these under (A) or (B) or (C), at least as these would ordinarily be understood. But my account, I claim, in abstracting from all that is substantive 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 what is obligatory is superior in some morally relevant way to its alternatives. For present purposes, then, the principle by which alternatives are to be ranked can be left wholly unspecified, as long as what is morally obligatory is acknowledged as morally unique (in some positive way) relative to its alternatives. It is important to note that I am not seeking here to effect some sort of reconciliation between consequentialist and deontological views, as might be attempted, for example, by a consequentialist who ascribes intrinsic value to justice, or promise keeping, or the respect of rights.53 That would be a substantive moral enterprise. Rather, I am simply claiming that, on a more abstract level, there is something common to (almost) all substantive theories of obligation, whether consequentialist or deontological (or something else, if "deontological" is narrower than "nonconsequentialist"), and that is that they share this view about the superiority of what is obligatory to what is nonobligatory. That which is superior maximizes some value and, I contend, (almost) all substantive theories of obligation can be appropriately cast in a maximizing mold. Thus, besides consequentialism ("It is obligatory to do that which maximizes intrinsic value"), each of rule utilitarianism ("It is obligatory to do that which maximizes rule utility"), Ross's early theory ("It is obligatory to do that which maximizes prima facie stringency"), Kant's theory ("It is obligatory to do that which maximizes compliance with the Categorical Imperative"), the divine command theory ("It is obligatory to do that which maximizes compliance with God's commands"), and rights theory ("It is obligatory to do that which maximizes nonviolation of persons' rights"), and so on, can be put in terms of maximization - 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 peculiarly unhelpful to represent (almost) all substantive theories of obligation in terms of maximization. Why do it? Because, as I hope to show in ensuing chapters, it is extraordinarily fruitful; we can come to a finer understand51 52 53 54 55

E.g., Nagel (1986), Ch. 9. The reason for the qualifier "almost" will be given in Chapter 2. Cf. Feldman (1986), Section 3.3; contrast Slote (1989), pp. 91-2. Cf. Lewis (1973), p. 76. 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 maximizing terms. Michael Stocker claims that, if putting a theory of obligation in terms of maximization is to be informative, then the value that it is said should be maximized must be specifiable independently of the determination of what one ought to do. This condition appears to be satisfied by certain theories (such as consequentialism, where in principle what is intrinsically valuable is determined first, and then it is said that what ought to be done is that which maximizes intrinsic value) but not by others (such as Kant's theory, which presumably accords no value to a certain maxim's being universalizable independently of its being such that one ought to act on it; or such as the divine command theory, which presumably accords no value 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 making this move [of viewing all substantive theories of obligation in maximizing terms] 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 maximizers are now open to the charge of shirking the needed honest toil... After all, one important task in ethics.. .is finding which objects of care are such that more 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 that it amounts to "trivialization." For much work can be accomplished by this move. 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 move does not do much to illuminate certain substantive theories with respect to how they are to be contrasted with others (does it help us to understand what 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 obligation itself. 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 maximization, that that which is obligatory is infinitely deontically superior to that which is wrong (so that, we might say, that which is obligatory has a value of 1 and all other alternatives a value of 0), then nothing is really gained in this 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 are in fact extremely impoverished, conceptually. We shall see that, where finer deontic rankings of alternatives are permitted by a theory, we can provide an account not just of mere obligation but also of conditional obligation, 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 other aspects of the concept of obligation.

1.5 ACT-EVALUATION AND O T H E R TYPES OF 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 of acts. An inquiry, such as the present one, into the concept of obligation might be classified as an inquiry into the evaluation of acts. It is certainly quite distinct from an inquiry into some types of evaluation of agents, their characters, their motives, and so on. Still, the lines are hard to draw with precision. We might describe the present work as an inquiry into what it is for an act to be obligatory, and then the classification of it as an inquiry into the evaluation of acts might seemjustified. But, although I shall often write, and have already written, as if such classification is appropriate, we might equally well describe the present work as an inquiry into what it is for an agent to be obligated to do some act, and then such classification might seem less appropriate.58 The fact is that I am not really concerned with what the proper classification in these terms is; I hope simply that I have done enough by now to give you a fair idea of what the subject of inquiry is. I propose now 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 order to vote, Vincent drives his car to the polling station. Puzzles lurk within these mundane examples. The claim that one ought to do the best one can (where, as explained in the last chapter, "ought" expresses what I have called "mere obligation," and "best" refers to deontic value) is perhaps most naturally, indeed standardly, understood to mean the following: of all those actions which one can perform (one's "alternatives," as it is often put), one ought to perform that action (if any) whose deontic value would be greater, if it were performed, than the deontic value of any other, if it were performed. But there are at least two problems with so understanding it.1 The first problem was exposed by Hector-Neri Castaneda.2 Suppose that 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 the best 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 it were 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 were performed, than the deontic value of his sticking out his tongue, if it 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 sticking out his tongue and saying "Aah") as actions or, at least, as the sort of thing one ought to do. But this is surely a revisionary proposal. Second, one might deny that (2.1) implies (2.2). But, although some do deny this, it is surely counterintuitive to do so.3 Third, one might deny that Sam's sticking out his tongue is a genuine alternative to his saying "Aah." Although I have some sympathy with this position, it is of no help to the thesis that one ought to do the best one can, on its present interpretation. For, on this interpretation, the thesis concerns simply all those actions that one can perform, 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 strictly 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 one ought to do the best one can requires a somewhat different interpretation from the standard one. This, I believe, is the correct response to the puzzle. The second problem reinforces this response. Suppose that the deontic value of Vincent's voting would be greater, if he were to perform it, than the deontic value of any other action that he can perform, if he were to perform 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 must drive 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 not driving 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 without 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 general principle embodied in (2.10), but this is surely counterintuitive.4 It is worth 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 his driving worse than his not driving, when he cannot vote without driving? The answer must, of course, rest on just what deontic value consists in, but certainly this is perfectly possible on some substantive views about such value. Consider, just by way of example, the view that an action's deontic value is a function of the sum intrinsic value of its consequences. It could well be that the consequences of Vincent's driving (including those consequences which are the action of voting and its consequences) would be intrinsically worse than those of his not driving, whereas the consequences of his voting (now that the driving, for good or ill, has occurred) would be intrinsically better than the consequences of his not voting. Or consider, by way of another example, the view that an action is deontically superior to its omission if and only if the action's maxim is universalizable whereas the omission's is not. As far as I can tell, there is no guarantee that, if the maxim for Vincent's voting were universalizable and its omission's not, then the maxim for his driving would be universalizable and its omission's not. Or, finally, consider the view that an action is deontically superior to its omission if and only if the action does not involve the infringement of anyone's rights whereas the omission does. Again, one might wonder whether there is any guarantee that, if Vincent's voting involved no such infringement, 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 performing 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 "Means Principle for Claims" in Thomson (1990), pp. 156—7. This is in contradistinction to her earlier 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 of the passage of time involved and contend that it is a mistake to claim that both (2.6) and (2.9) are true. At the time at which it is true to say that it is not the case that Vincent ought to drive his car, it is not yet, and may never be, true that Vincent ought to vote. Whether or not he ought to vote will depend on whether he drives. Only if he drives will it be the case that he ought to vote. But while this response may seem plausible, there are reasons to think it mistaken; I shall go into these in Chapter 3. More to the point here, the response is irrelevant, since the problem at issue does not rest on the passage of time. Consider a parallel case in which (2.5') Vincent's voting for Smith would be deontically superior to his voting for no one, which itself would be deontically superior to his voting 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 not doing so are all true, and where the reason that (2.8') is true is that, if Vincent were to pull the lever, he'd vote for Jones. A fourth response is this. We should not talk of individual acts being obligatory; we should talk only of entire courses of action being obligatory. Hence it is not the case that Vincent ought to vote. Rather, either he ought to drive-and-vote-and-X or he ought to not-drive-and-not-voteand-Y (presumably we can rule out its being the case that he ought to driveand-not-vote-and-Z), where X and Y (and Z) represent whatever the remainders are in the respective courses of action. But this is surely much too revisionary a proposal. It would imply that almost all, and probably all, of our everyday assertions concerning what we ought to do are simply false (since they do not concern entire courses of action but only segments thereof). A final response is this: we should revise the standard interpretation of the thesis that one ought to do the best one can. Once again, it is this response 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 deontically best course of action (and concomitant events) open to one. We should note 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 combination thereof), but that we can affect how the world will be. This may sound grand, even melodramatic, but it's the truth. For example, if I can break 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 to me whether or not the window breaks — then it's also true that I can make the world be such that it contains a certain broken window (or not) — I have a choice as to whether or not it contains this; whether or not it contains this is up to me. Of course, much - most - of what happens in the world is not up to me, but that does not alter the fact that how the world as a whole will be is (often) up to me. Another, and by now fairly well-established way of putting this point is as follows: at any moment of choice a number of different possible worlds are accessible to me; which of these will be the actual world is up to me. The situation may be very roughly illustrated as in Figure 2.1. Here the agent, 5, has a choice at time T1 between doing A and not doing A. (Already this is oversimple, in that there would presumably be a number of ways of not 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. Doing something else, E, may not be sufficient for not doing A, since it may be that S can do both A and E; thus, if not doing A is effected by doing E, E must be done instead of A.) If S does A at Tu he will have a choice at T2 as to 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 whether or not to do C, but his doing D will be a foregone conclusion. If S doesn't do B at T2, his not doing C will be a foregone conclusion, but he will have a choice at T3 whether or not to do D. The nodes in the figure thus represent points of choice and the paths W-W6 represent different possible worlds (each sharing the same history up to T?) accessible to S as of Tx by means of his choosing one way or the other. Of course, huge amounts of information are left out of the picture: what happens before T?, what happens after T3, what happens between T, and T3 apart from the small number of actions to which reference is made. Nonetheless the picture is helpful, I think, in bringing home two important points. First, as noted, at any moment of choice our choice is, in effect, a choice between worlds. Second, as time progresses, our options never increase but in fact grow narrower, in that with each choice we cut ourselves off from worlds which had hitherto been accessible to us; what is presently accessible to us always was accessible to us (during our lifetime), but it may not remain so for long. Clearly, what has just been said is far from precise and needs further work. I shall attend to that shortly. But at this point I want to lay out my interpretation of the thesis that one ought to do the best one can; for then the refinements to what has just been said will, I think, be more readily appreciated. The main gist of the interpretation is simply this: an agent 5 ought to do an act A if and only if (a) 5 can do A, (b) 5 can refrain from doing A9 and (c) any accessible world in which S does not do A is deontically inferior to some accessible world in which S does A. More precisely: (I) agent 5 ought, at time T i n 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 S does A at T' in W'\ (b) there is a world W" such that W" is accessible to S from Wat Tand 5 does not do A at T ' in W"\ and (c) for all worlds W" such that W" is accessible to S from Wat Tand S does 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 deontic value for S at Tof W'.9 Clauses (a) and (b) combined imply that A is optional for S, in the sense that S can do A and am also refrain from doing it. We may call this the personal sense of "optional." Of course, these clauses do not imply that A is morally optional for S, in the sense that 5 may, morally, do A and may also refrain from doing it. (On the contrary, if S ought to do A, then it is not the case that 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). I shall explain why this is so in Subsection 2.2.6 below. Perhaps it is as well to repeat here that "ought" is intended to express overall moral obligation and that no substantive theory is being presupposed 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 substantive theories of obligation. Thus, once again, not only is (I) intended to be compatible with consequentialist theories, but it is intended also to be compatible with a wide range of deontological theories. It is for this reason that the values of the worlds are said to be the values that they have^br 5 at T. The values are thus both agent- and time-relative; whether or not they are 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 latter is redundant. Nonetheless I include clause (a) in order to emphasize the fact that one ought 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 (whatever exactly that amounts to), for I can imagine that some people might want to say the following: some (nonhuman) animals in fact have a choice between various alternative actions, and these actions are in fact such that some are better (even deontically) than others, but it is false that such animals morally ought to choose the best alternatives, because such animals are not moral agents. (See, e.g., Thomson, 1990, p. 215.) There may be something 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 insignificant and can be ignored, insofar as the values for S at T of these worlds are the values that they have for any possible agents at any possible times. But on other theories the relativization may be significant. Consider Kant's theory, for example. It might at first be thought that such a theory resists accommodation by (I), but this is in fact not so. If the substantive mark of obligation is compliance with the Categorical Imperative, then the deontically best world open to an agent is simply one in which such compliance is maximized. This statement is perfectly legitimate and should not be thought to be at odds with what a Kantian would want to say. To see this clearly, consider the (oversimplified) situation depicted in Figure 2.2. Suppose that each of A, B, and C would constitute compliance with the Categorical Imperative by the agent, whereas none of ~^4, ~B, and ~C would. A "minimal" Kantianism might declare any failure by the agent to comply with the Categorical Imperative categorically wrong and thus deem W t infinitely deontically superior to any of the other worlds, so that (as we might put it) it has a deontic value of 1 whereas all the others have a deontic value of 0.10 A somewhat richer Kantianism might declare every noncompliance as deontically bad as any other but allow for them nonetheless to "add up," in which case the deontic ranking would be this: first, Wt (no noncompliances); second, W2, W'3, and W5 (one noncompliance each); third, W4, W6, and W7 (two noncompliances each); and last, 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 allow for a still finer deontic ranking of the worlds. This way of picturing matters is also compatible with saying (as I think almost all Kantians would) that it can sometimes happen that a person can avoid someone else's noncompliance only by way of his own noncompliance. Think, for instance, of Bernard Williams's celebrated case ofJim and the Indians: Jim cannot get the captain not to kill the Indians except by killing one himself.11 A Kantian would say: either Jim will fail to comply with the Categorical Imperative or the captain will. In such a case, a Kantian would presumably say that it is right for one to comply and wrong for one not to comply (even at the cost of the other's noncompliance). This means that the relativization of the values of worlds to agents is significant, in that S's noncompliance has a special disvalue relative to S that it does not have relative to someone else. This point perhaps requires elaboration. I am envisaging the Kantian as saying the following. Jim's killing the Indian would constitute a violatingby-Jim of the Categorical Imperative, and hence it would be morally wrong for Jim to do it. Admittedly, Jim's not killing the Indian would guarantee (in the sense that Jim could not avoid them, even if the captain could) several violatings-by-the-captain of the Categorical Imperative, and thus would guarantee that (in some sense) more serious wrong is done, but it would not itself be a wrongdoing by Jim. It is for this reason that the agentrelative valuation of worlds is significant. In some nonrelative sense we might well want to say that some Jim-accessible world in which Jim kills the Indian is better than any such world in which he doesn't, since the former involves only one violation of the Categorical Imperative whereas the latter involves several. But for the Kantian such a valuation of worlds, while perhaps accurate and informative for certain purposes, is not the relevant valuation for determining what it is obligatory for Jim to do. When it comes to determining Jim's obligations, some Jim-accessible world in which Jim does not kill the Indian is better than any such world in which he does.12 Thus all, or almost all, Kantians will say that deontic value is significantly agent-relative. It may be that they would also say that it is significantly timerelative; this will depend on whether they accept that it is possible that a person can avoid his own noncompliance with the Categorical Imperative only 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 minor injustice and B a major injustice, and that, for this reason, each would constitute a violation of the Categorical Imperative. I take it that the Kantian who accepts that this is possible would say that Joe may not do A, even if only thereby can he avoid the greater injustice, B. This means that any accessible world containing his not doing B at T3 is deontically worse, relative to Joe at Tj, than some accessible world containing his not doing A at T2; and this is so even if, by some other measure, every ~^4-world accessible to Joe at Tj is worse than some ~B-world accessible to him then. In such a case, then, the time-relativization of deontic value is significant; agentrelativization alone would not suffice to yield the deontic inferiority of every ~B-world to some ~^4-world, since the actions in question are both Joe's. This last case also serves to reinforce the need to talk in terms of the valuation of worlds rather than that of acts. An art-based theory might well imply that Joe ought at Tt to avoid A at T2, since avoiding A would be better than doing A, and that he ought at Tt to avoid B at T3, since avoiding B would be better than doing B. But, as with the case of Vincent presented 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 in this chapter implies that it cannot be both that Joe ought at Tt to avoid A at T2 and that he ought at Tt to avoid B at T3, since {ex hypothesi) he cannot at Tj avoid both. Thus Kantianism appears perfectly compatible with (I). So too do all sorts of other deontological theories. Still, at this point, certain limitations to the analysis must be acknowledged. First, it is at best only a partial analysis. I have said that the value at issue in (I) is deontic value; I have not tried to give an account of this notion in turn. Nor shall I. Clearly, deontic value is a type of morally relevant value, but there are other types of morally relevant value.13 A fuller analysis would give an account both of how morally relevant value is to be distinguished from morally irrelevant value and of how deontic value in particular is to be distinguished from other types of morally relevant value. I have no such account to offer. Second, it must be recognized that (I) does not accommodate all substantive theories of obligation. After all, consequentialism has traditionally been put in terms similar 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 thefirstis superior to the second with respect to deontic value, whereas the second is superior to the first with respect to some other morally relevant value - for example, with respect to the incidence of virtue and vice.

30

terms are unacceptable, involving a misunderstanding of the very concept of moral obligation. Perhaps, too, a similar problem besets Kant's emphasis on the universalizability of maxims, as this is traditionally conceived. At any rate, any substantive theory based on such a misunderstanding will of course not be compatible with (I). Nonetheless, (I) is surely compatible with the basic view underlying traditional consequentialism (namely, that obligation is some rather straightforward function of intrinsic value), and I see no reason to think that the basic view underlying Kantianism and other deontological theories should not be thought to be similarly compatible. Still, there is this limitation. Consider the sort of theory that says that one ought to do what one believes to be best — in some less than fully liberal sense of "best." This can be accommodated by (I) only if the theory is understood 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 that is concerned with what, among those actions the agent believes he can perform, is believed by the agent to be best, unless the agent's beliefs concerning his options are accurate.

2.1.3 Extension of the analysis The thesis that moral obligation is to be understood in terms of the comparative values of possible worlds is by no means new. Jaakko Hintikka, among others, has made such a proposal; and Holly Goldman, Michael McKinsey, and, especially, Fred Feldman have provided accounts of obligation that are very closely related to the account given in (I). Each is somewhat 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 respective worlds.15 (Worlds la and lb are thus of equal deontic value, relative to the 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 topranked world containing ~B). 14 See Hintikka (1970), pp. 70-1; Goldman (1978), p. 202; McKinsey (1975), p. 385; and Feldman (1986), p. 38. There is no space here to make a detailed comparison between each 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 of accessible 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 us express 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 does not do A is not deontically superior to some accessible world in which 5 does 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 ' i n 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 their respective subclauses (c3) and, second, in the fact that (II) lacks a corresponding subclause (c4). As to the former point: this concerns the issue, just noted, 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 as well. Moreover, he may do B, too, although it's not the case that he ought to 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 and he 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 poses certain problems in this context. "May not" is ambiguous, as is illustrated in the statements of (III) and (IV); sometimes it is used to express the obligation not to do something, sometimes merely the permission not to do something. Further problems of this sort lie ahead. To help us deal with them, 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 an account of what is expressed by O(A); (II) gives an account of what is expressed 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 wrong for 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 the doing of something other than A (which may be compatible with the doing of A). Thus O(~A) expresses the obligation to refrain from doing A, not merely the obligation to do something other than A. 17 Here I am concerned exclusively with the adjectival form of these terms. The nominal forms of "right" and "wrong" (as in "Rachel has a right to be left alone" and "You have done 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 people talk of "the right" thing to do under the circumstances; here what is meant, I think, is that which one ought to do. Sometimes people talk ofwhat it 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 middle ground 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 to accommodate it explicitly in the present account. The account may be extended further. First: (V) O(A&B): S ought, at T i n 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 a world 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 f o r S a t T o f 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 deontic 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 them both, (b) S can refrain from doing them both (whether by doing neither of them or by doing one but not the other), and (c) any accessible world in which S does not do them both is deontically inferior to some accessible world in which S does them both. (Clearly this can be extended even further to accommodate not just two actions but any number of actions.) O n Figure 2.3, then, not only ought S to do each of A and B and C (that is, not only 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 difference really 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 two are 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, so consider first of all this: (VI) P(A&B): S may, at T i n W, do both A at T ' and B at T " iff [as in (II), with changes as from (I) to (V)] O n Figure 2.3, S may of course do (A&B&C); and it is also the case that he may do A, he may do B, and he may do C. But notice, as mentioned earlier, that he may also do ~B. Clearly, though, it's not the case that he may do (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, and I hope now that it will be clear how it is that there is a difference between saying that O(A) & O(B) and saying that O(A&B), even though they are nearly equivalent. W e may of course also say the following: (VII) O(~(A&B)) [W(A&B)]i S ought, at T i n W, not to do both A at T ' and B at T " [i.e., S may not, at T i n W, do both A at T ' and B at T"] [i.e., it is wrong, at T i n 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 i n 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 he ought not to do D (although it remains true that he may do ~B). Also, S may not do both A and B (in the sense of (VIII); he is permitted not to do them both); this is because he may do (A&~B) (although it remains true that he ought to do A). A final set of four statements concerns the tricky connective "or." As far as I can tell, the following best captures prevailing English usage: (IX) P(AorB): S may, at T i n 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"] 1 9 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 T i n 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 Tin W, 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 other things): S may do either A or B; S ought to do either (A&B) or (^&~B>); 5 ought 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 do either A or B" and "S ought to do either A or B" are to be standardly understood as interpreted in (IX) and (X), respectively, and not, as some logicians would 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 logical symbol for disjunction, v. The point here is this. For any propositions p and q, \ip is true, then so is p or q (p V q). Thus, where p and q concern actions (as where p is the proposition that S does A and q is the proposition that 5 does B), then, if S does A, it is also true that either 5 does A or S does B. Some have gone on to infer from this that, if S does A, then S does either A or B. Now, it may be that there is a sense of "or" that permits this inference, 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) do either 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 not necessarily, 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 claims as "Robin ought to return the book by next Wednesday," "Robin ought to return the book between Tuesday and Wednesday," and so on. Let A be the act of returning the book and Tx... Tn be a (perhaps discontinuous) series of 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 either A at T2 or A at T3 or.. .or A at Tn. Similarly, (V) can be used to capture what is at issue in such common claims as "Marsha ought to maintain silence throughout (or during) the exam." Let A be the act of maintaining silence and T2... Tn be a series of times, with T2 being the beginning of the exam and Tn being the end. Then what's at issue in this example is this: Marsha ought at 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 can simply use "will be obligated" to express future obligations.22 Second, "ought" appears to have no past tense.23 But, again, this is of no philosophical significance; one can use "was obligated" or "ought to have" to express past obligations.24 At least one philosopher has been misled by these grammatical 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 to do A, always thereafter 5 ought to have done A. Although this may seem correct (indeed is correct, given that "ought to have" expresses a past obligation), what in fact is meant here is that, if 5 ought at Tt to do A at T2, then 5 ought at all times subsequent to Tt to do A at T2.25 This, I submit, is patent nonsense, and it is ruled out in (I) by clause (b); after A is done at T2, there is no longer an accessible world in which A is not done at T2. 22 23

See Broad (1985), p. 225. 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 committed to something like this, too, but not on the basis of a misunderstanding of what "ought to have" means. See Feldman (1986), p. 43.

37

2.1.4 Perfect and imperfect duties Sometimes a distinction is drawn, following Kant, between "perfect duties" and "imperfect duties." Kant's famous four examples of duty concern the duty not to commit suicide, the duty not to deceive others by virtue of making a false promise, the duty to develop one's talents, and the duty to act charitably.26 The first two are said to be perfect duties (the first to oneself, the second to others) and the last two are said to be imperfect duties (the first to oneself, the second to others). Just what the distinction is supposed to consist in is controversial. In a note Kant says that a perfect duty is one that admits of no exception in the interests of inclination, whereas an imperfect duty admits of such exception, but just what this means is unclear. The matter seems to me very complex. First, it is worth remarking that it is natural to express the perfect duties negatively (as the duties not to do such-and-such), as I have done, while the imperfect duties are naturally expressed positively. Is this significant? It's hard to tell. Certainly, some have thought that there is a significance to the distinction between positive and 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 (in this case, as the duties to preserve one's life and to behave honestly towards others) and imperfect duties negatively (in this case, as the duties not to let one's talents go to waste and not to behave meanly towards others), although it must be acknowledged that these alternative renderings might not quite hit the mark: isn't there, after all, a difference between (positive) action and (negative) omission or refraining?28 Whatever the truth of the matter, the present account is designed to accommodate both positive duties (see (I)) and negative duties (see (HI)).29 Something else that Kant might have had in mind is this. A perfect duty is fully specific, whereas an imperfect duty leaves it to the agent to determine just how the duty is to be fulfilled. For example, a duty not to kill oneself is just that, whereas a duty to be charitable can be satisfied either by being 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 doing A (or omit to do A, etc.)" can be understood in terms of (I), with "refrain from doing A" being substituted for "do A."

38

This distinction, I think, is problematic. Hardly any duty is fully specific in the 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 At any rate, the present account is adequate to the expression of what is at issue here. If S ought not to commit suicide, then every accessible world in which 5 commits suicide is one that is deontically inferior to some world in which he does not. If 5 ought to act charitably, then every accessible world in which S fails to act charitably is deontically inferior to some world in which he acts charitably. If there is more than one top-ranked accessible world in which S acts charitably, then S may enjoy a certain sort of latitude with respect to how to satisfy the duty to act charitably. Perhaps in one such world he is charitable towards A but not towards B, in another towards B but not towards A, and so on. In this case S ought either to act charitably towards A or to act charitably towards B (or...) (see (X)), but it's not the case that he ought to act charitably towards A and it's not the case that he ought to act charitably towards B. Thus it is not (morally) up to S whether to act charitably, but how to do so is up to him. Another thing that Kant might have had in mind is this: a perfect duty is a duty-to (with a correlative right), but an imperfect duty is merely a duty. 31 The present account is not designed to accommodate this distinction explicitly. As noted in Section 1.3,1 seek here only to account for what I 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 account of what it is to have a duty-to (or obligation-to) is something that I shall leave aside. Sometimes it is said that what Kant had in mind is this: a perfect duty is a duty to perform some act (or to refrain from doing so), whereas an imperfect duty is a duty to adopt a certain maxim for action. 32 It may seem that the present account cannot accommodate this, since it has been put in terms of " 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 perfect duty ought always to be satisfied, but an imperfect duty only ought sometimes to be satisfied.33 What is meant by this? Presumably the following. To say 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 31 32 33

Cf. Chisholm (1968), p. 416, n. 6. This is how Mill understands the distinction in Mill (1957), Ch. 5. Cf. Hill (1971), p. 57. 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 one occasion - that is, that A is "repeatable." This view of action is controversial; I shall address it in Subsection 2.2.5 below.) Of course, Kant presumably meant that the duties of which he spoke are duties had by everyone (or at least by everyone capable of having duties of the sort in question). In this sense, a perfect duty to do A would be understood to consist in the truth of the following: (2.13) for all agents S and times T' and T (such that Tis not earlier than T' 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 duty is 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 of the following: (2.14) necessarily, for all agents 5 and times T' and T (such that Tis not earlier than T" and S exists at these times), 5 ought at T' to do A at T. The present account shows how these statements are to be understood.

2.2 UNDERLYING METAPHYSICAL AND CONCEPTUAL ISSUES 2.2.1 "Can" and control According to the analysis of obligation just presented, S ought to do A only if 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 personally optional for S. That "ought" implies "can" is of course controversial; I shall address the matter in Section 3.1. Here I want to attend to a preliminary question. The present account may seem to restrict itself to obligations concerning actions ("5 ought to do A"; "5 can do A"). Is such restriction warranted? 34

Or Donagan (1977) or Fried (1978).

40

Strictly, it is not. The sort of control, or freedom, expressed by "can" in this context may be either direct or indirect. It is controversial just what sort or sorts of item are such that an agent may have direct control over them, but it is clear that there must be some such sort or sorts for there to be any control at all. Consider the example given earlier: that of breaking a certain store window by throwing a certain brick through it. Let us say that this is both something that I can do on a particular occasion and something that I can refrain from doing. I thus have control over whether or not the window breaks. This must surely be only indirect control, though; I can achieve the window's being broken only by way of achieving something else. What is this something else? Moving my arm, for one thing; and releasing the brick; and so on. On my view (although the truth of this claim is 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 to throw the brick (or some suitably related decision — one that will result in the window's being broken). I have direct control over the decision and thereby indirect control over its consequences. These consequences include my arm's moving, the brick's being released, the brick'sflyingthrough the air, the brick's hitting the window, the window's breaking, the alarm's sounding, and so on. It is worth noting that none of the items on the list given in the last two sentences is an action - unless a decision is itself an action, which I suppose we may grant in a minimal sense of "action."35 Yet I have control over all of them. Where do nonminimal actions themselves fit in? As follows, I believe.36 The action that is my moving my arm just is (roughly) my deciding 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 control over the decision involved and (thereby) indirect control over its consequence, and my action just is this decision plus this consequence. Similarly, my releasing the brick is this same decision coupled with the consequence of the brick's being released; my breaking the window is this same decision coupled with the consequence of the window's breaking; and so on; and I have a sort of hybrid control over all these actions. This applies to omissions, too. Perhaps I have made a promise to stay indoors and water the plants; when I go outside and throw bricks, I omit to water the plants. I have control over this omission. How? Once again, I believe, 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 watering the plants. My not remaining indoors and watering the plants may thus be said to be a consequence37 of this decision. It may be that certain items over which we have control don't fit easily into this decision-plus-consequence picture. I am thinking primarily of certain mental phenomena, such as desires, motives, intentions, and other such attitudes, as I shall call them. (Obviously, there are important distinctions to be drawn among these phenomena, but doing so is not necessary for my purposes here.) Sometimes, it is clear, the adoption of an attitude is the result of acting in a certain way (I may have had to school myself, at length, in order to gain my newly acquired tolerance towards others who are in certain ways different from myself; I may have had to work hard to rid myself of my desire for cigarettes; and so on), and this fits in well enough with the picture in question. But it may seem that at other times I can adopt a certain attitude directly, that is, without doing so by way of deciding or acting in a certain way. Perhaps, when Kant urges the adoption of a certain maxim, he means (in part) that we ought to adopt a certain attitude - an attitude of preparedness to act in certain ways. (I shall say something similar concerning cooperation in Chapter 9.) And perhaps it is possible to do this directly. I suspect not, but I'm not sure. At any rate, to repeat, it is not essential for present purposes that the foregoing decision-plus-consequence picture be accepted. What is essential is only that the distinction between direct and indirect control be accepted, and this distinction seems to me unassailable. Surely one cannot have control over anything without having direct control over something; and surely not everything one has control over is such that one has direct control over it. Moreover, it is surely 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 indirect control. We thus have three categories of control: direct, indirect, and hybrid. And so we have three candidates for what is obligatory: those items over which we have direct control, those items over which we have indirect control, and those items over which we have hybrid control. I see no reason in principle to reject any of these candidates.38 That is, it seems to me perfectly proper to say, for example (and now I'll resort to the account of action 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 claimed to be wider than the class of actions.

42

and water the plants; it is obligatory for me that I stay indoors and that I water the plants; and it is obligatory for me that I be indoors and that the plants receive water. Similarly, it is obligatory for me that I not decide to break a certain window; it is obligatory for me that I not break the window; and it is obligatory for me that the window not break. Part of the reason why all of these things are obligatory for me is that I have control over them, although only some of these things are actions — in the minimal sense, those over which I have direct control, and, in the nonminimal sense, those over which I have hybrid control; those over which I have only indirect control are not actions in any normal sense of the term. Nonetheless, I shall continue to talk in the abstract in such terms as "S ought to do A," "S can do A" and so on, where A ranges over things we do. The reasons for this are first that doing so simplifies exposition, and second that, even though we cannot properly be said to "do" the consequences of our decisions, they are nonetheless something that we bring about — and it cannot ever be the case that a certain consequence be obligatory for S without its being the case that it is obligatory for S to bring it about, that is, so to act (whether nonminimally or only minimally) that the consequence occurs. At this point, though, some would demur. In particular, H. A. Prichard and W. D. Ross would claim that only those items (in their view, what they call "volitions" or "self-exertions") over which an agent has direct control are such that they may be said to be obligatory. Prichard and Ross at times talk 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' receiving water) are not, strictly, actions39 and not, strictly, obligatory.40 The reason that they say that only volitions can be obligatory is apparently that they believe that only volitions are, strictly, in an agent's control.41 But this cannot 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 think that what Prichard and Ross have in mind is that anything that is only in an agent's indirect control is notfully in the agent's control. For example, after I make the decision to throw the brick, what happens is not up to me but rather, as it is sometimes put, "up to nature."42 And nature may not cooperate. 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 hand and window, or the window may be made ofplexiglass, and so on, in which case the window will not break despite my decision to throw the brick; and whether or not any of these things is so is not (at the present time) in my control. Now, it is surely correct to say that whether or not the window breaks is, 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; whether or not it breaks is, to some extent, a matter of luck. But so what? Under the circumstances (which include the fact that nature is cooperative), whether or not the window breaks is up to me. Why not, then, say that I ought not to break it? Prichard and Ross appear to think that what's obligatory must be fully in one's control, and they seem to think that that and only that which is 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 a matter of luck, just as whether or not my exertions bear fruit is to some extent a matter of luck; for such self-exertion requires the cooperation of all 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 so on).44 If'full control were required for obligation, nothing would be obligatory; but some things are obligatory, and so full control is not required. Some measure of control is still required, however; but both direct and indirect control constitute some measure of control, and so I see no need to say that, strictly, only those items over which we have direct control are obligatory.45 The obvious fact that our control is limited (not simply in the sense that we are never fully in control of anything, but in the sense that some things are such that we are not in control of them at all) means that we may divide the 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 over which 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 or degree to which one is to blame — or so I argue in Zimmerman (1988a), p. 54ff. and Sections 4.3 and 4.11. There is no analogous claim to be made about obligation, for it does 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 Future Framework is that over which we have no control; it is fixed; it is featured in every world accessible to us. That which lies in the Future Zone of Indeterminacy is that over which we have (some measure of) control; it is not fixed; it is featured in some but not every world accessible to us. Of course, 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 fixed for him what he will do tomorrow.47 This is an important point: what others do may not be fixed for them, but it may be fixed for me, from which it 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 Chapter 9.) Of course, sometimes what others do may not be fixed for me; I may be able to influence their actions. In general, then, the actions of others may be a part of my Future Framework or be a part of my Future Zone of Indeterminacy, just as many other things may be (such as a certain window's being broken).48 2.2.2 "Can" and accessibility What exactly is the relation between "can" and accessibility? So far, I have hardly mentioned the latter when discussing the former, and yet it is in terms 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 from Wat Tin which S does A at T'. Or more generally (to cover those items over which one may have control but which aren't strictly actions): (2.16) it is personally possible for S at Tin Wthatp be true iff there is a world accessible to S from Wat Tin whichp is true. (Here p may itself be an entire world, that is, a possible proposition that entails every proposition or its negation.) Although true, however, these statements 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 shall not discuss) and the debate concerning whether "ought" implies "can" (which I shall discuss in Section 3.1). "Can" can, of course, be used to express a number of different concepts of possibility, and so it would be desirable if some further account of the particular concept of personal possibility were given here. Unfortunately, I haven't much to offer. I've already indicated that, when it is personally possible for S to do A and it is also personally possible for 5 to refrain from doing A, then S is in control of his doing A. Still, doubtless a variety of concepts of control might be distinguished. Perhaps an illustration will help. There is a sense in which Carl Lewis can run 100 meters in under 11 seconds but in which I cannot. But it would be surprising if, in the sense of personal possibility, Lewis can do this right now; for that to be true, there must be no obstacle in the way of his doing so right now. Still, it is true, even in the sense of personal possibility, that he can do so later (or, more accurately, that he can now do so later); 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 no world now accessible to me in which at some time (now or later) I run 100 meters in under 11 seconds, there is a world now accessible to Carl Lewis in 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. Although some come close to advocating this,491 do not. If anything, the notion of accessibility seems to me a philosopher's artifice that is itself to be understood in terms of the notion of personal possibility. I suggest the following, which is explicitly based on the view (already acknowledged to be controversial, and certainly not to be defended by me here) that direct control concerns decisions.

To say that someone can (in the sense of personal possibility) do something 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 something else that he can directly do such that, if he were to do that, then he'd be one step closer to doing this. That is, taking the first step would enable him directly to take a second step, and doing this would enable him directly to take a third step, and so on, for however many steps are needed. This is in keeping with the branching pattern of Figures 2.1-2.3: a move by the agent from one point of choice takes him to the next, where he is able to make 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 in terms 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 and decisions 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 being now able to do anything now. After all, there are worlds accessible to me even as I sleep; for I will wake up, and then, at some point, I will be directly able to do something that will enable me to do something else, and so on. This point is, I think, easily handled by revising clause (a) so that T is said not to be identical with Tt but simply not later than it (as long as 5 exists atT). 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 so on. But suppose that Chris is an inveterate daredevil and in fact would make the decision to cross the road only if there were a lot of traffic about, in and out of which he could (he thinks) dart; and suppose that, if Chris were to make this decision under those circumstances, he would in fact get run over and 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 conditions (that in fact obtain) such that, if Chris were to make the decision to cross under these conditions, then that would enable him to do something else, and so on. For these reasons, I think that (2.17) should be replaced by the following: (2.18) 5 can at T d o A at T" in W iff there are times Tv..Tn9 decisions Dv..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 similar approaches. 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 at

Ti+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 A atTn. This statement concerns only actions, of course, whereas it has been noted that other sorts of things may be under an agent's control. To accommodate 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 some time 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: p is accessible to 5 from Wat T. Thus personal possibility just is accessibility, 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 the world contains his doing A, and vice versa. I have said that nothing ever becomes accessible that wasn't already accessible. In a sense, then, whatever one can do one could always do. But this might seem to fly in the face of an obvious fact: one can learn to do things one couldn't do before; one can gain an ability. But here we need only note that personal possibility is to be relativized to times; in particular, "can" has a double time index. What is directly in one's control is, temporally, immediately so;51 but what is indirectly in one's control is, temporally, only remotely so.52 Given this, the acquisition of a "new abili51 At least, this is typically so with respect to its initiation. Whether it is necessarily so is unclear to me. Perhaps even the exercise of direct control is or can be a noninstantaneous process, such that one can have direct control at some time over an event that begins to occur only at a later time. 52 At least, this is typically so. Again, whether it is necessarily so is unclear to me. Perhaps it is possible that one have immediate direct control over some event E, that E have F as a consequence, and that E and F occur simultaneously, so that one's indirect control over F is not remote but immediate.

48

ty" may be understood in terms of something's becoming immediately possible that was hitherto only remotely so (or, perhaps, in terms of something's becoming close to immediately possible that was hitherto a long way from being so). If I can now speak French now, having learned to do so, I could always speak French now. Of course, this doesn't mean that, at all earlier 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 because I don't know the combination to the lock; I can't get my VCR to work because 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 unless one can do it intentionally. There is certainly such a sense of "can." The illustrations just given confirm that. But it is not this sense of "can" that I'm working with here. On the contrary, I can open the safe, and I can operate the VCR, in the present sense of "can," and this is simply because I can (easily) perform those physical movements which would result in these things being done.53 Of course, there is a sense in which I'd be very lucky if I managed to do these things; but in this sense of "luck" it is not that their accomplishment is strictly out of my control, but rather that, among the very many things that I can apparently do, I have no reason to choose to make this particular set of movements (that would result in the safe's being open or in the VCR's working) 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 win it; that depends on how it works. If I would win it by inscribing a certain number on a card, and I can inscribe the number, then I can win it. But if to win the lottery I have to purchase a certain ticket with the winning number already imprinted, then it is highly unlikely that I can win it. It is highly unlikely, that is, that there is a world accessible to me in which I win the lottery; for there are only a relatively few tickets that I can purchase (given that my purchasing power is limited and that there are only a relatively few stores that sell the tickets and that I can reach in time), and the likelihood that one of these tickets has the winning number is very small. There is a distinction to be drawn, then, between its being unlikely that I will do one of the many things that I can do (as in the safe, VCR, and card-inscribing cases) and its being unlikely that I can do a certain thing (as in the preimprinted ticket case). 53 Cf. Feldman (1986), pp. 24-5. 49

Many, indeed most, of the things that I can (in the present sense) do are things I know nothing about. Those worlds that are presently accessible to me 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 my immediate and close-to-immediate personal possibilities; but the more remote the possibilities, the less I will tend to know about them. Unless I die within a year, I can now do many things a year hence; but, except for the most mundane among them, I have little idea what these things may be.54 In saying that the "can" that "ought" implies has no interesting tie to knowledge, I am attempting to accommodate those substantive theories of obligation (act utilitarianism, as traditionally understood, is one; certain rights-theories are others) that allow for the possibility that an act be obligatory even though the agent doesn't know that he can perform it or how to 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 he can perform it or how to perform it. Still, it is consistent to make a claim that comes close to this, namely, that unknown personal possibilities confer no deontic value on the accessible worlds that include them and thus do not affect the relative rankings of these worlds. If this were true, then, although some such possibilities could nonetheless be obligatory, their obligatoriness would be wholly parasitic on the obligatoriness of known possibilities. 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 do something 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 that time. Hence, since the actual world is accessible to everyone, whatever a person must do he does do, and whatever he does do he can do. No surprises so far: whatever is necessary is actual, and whatever is actual is possible. But there is a puzzle lurking here. Sometimes, when talking of what is personally possible, people use the term "free to" instead of "can." "S can 54 55

See Feldman (1986), p. 24. 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 this appropriate? The answer must be no, I think, at least if the following two points are conceded: first, that "can" is related to "muse" and "does" in the way 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 points together, and if one uses "can" and "free to" interchangeably, then one obtains the result that whatever anyone must or does do he does freely. And this is surely unacceptable in any standard use of "does freely" — and here I mean to be invoking that common, if not very well understood, sense of the term that is at issue in the debate concerning freedom and determinism.56 And yet I declared earlier, in Subsection 2.2.1, that the "can" implied by "ought" expresses control or freedom. How can this be so, given what has just been said? The fact is that what I said earlier was said too loosely, and I must now put matters right. I have already noted (and this turns out to be crucial) that there 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 are a world accessible to S in which S does A and a world accessible to S in which 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 at T' and is also free at Tto refrain from doing A at T'. Thus, perhaps surprisingly, even though "can" in the present sense does not imply "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 from doing 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 relevant senses. 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 leads him to say that whatever is inevitable is obligatory - including all that lies in the past. This is very odd. Still, he does also note (p. 43) that a second, related sense of "obligation" can be accounted for simply by including the point about "can refrain."

51

accessibility, but in a new, stronger sense of "can" that might be put in terms of "free to." And, given the relation between "free to" and "does freely" 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 freely atT. (This concerns immediate obligation. The corresponding claim concerning 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 A freely 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 comes to 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 S can 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 seems to me that a person can freely do something even though he cannot refrain from doing it (in other words, even if it is not personally optional; even if he is not in control of doing it).58 If this is so, and if, as I suspect, when someone 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 refrain from doing it seems to me to be one of the main lessons to be learned from Harry Frankfurt's ground-breaking discussion of the relation between moral responsibility and alternate possibilities.59 Frankfurt argues — successfully, I think — that a person can be morally responsible for what he has done, even though he couldn't have done otherwise. (One example might be this: Larry cannot leave the room, because the door is locked; he is nonetheless morally responsible for staying in the room, in part because he is unaware of the door's being locked.) What I think this shows is not that a person can be morally responsible for what he has done even though he didn't do it freely, but rather that a person can have done something freely even though he couldn't have done otherwise. (Thus Larry freely remains in 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 control" of doing it. See Zimmerman (1988a), pp. 32-3. 59 Frankfurt (1969).

52

responsibility does not require alternate possibilities. The link between responsibility 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 freely doing something does not require being (or having been) able to refrain from 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 to do otherwise, is sufficient for doing it freely;6 doing something freely is sufficient for being (or having been) free to do it; and being free to do (or to refrain from doing) something is sufficient for being able (where "able" expresses what the "can" of personal possibility expresses) to do (or to refrain from doing) it. I am not sure just how each of these claims might itself be justified. At this point, though, it is natural to ask: if moral responsibility doesn't require alternate possibilities, why think that moral obligation does? I shall consider 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 conceiving 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 token of this type. It seems that obligatoriness, rightness, and wrongness may be predicated of both tokens and types. Allie may say that a certain killing was wrong; Barbara may say that killing is wrong (period). When faced with a formula 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 a variable 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 61

See Zimmerman (1988a), Section 4.10, for a fuller discussion of this issue. 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 (see Zimmerman (1988a), pp. 24—6, including n. 22, on the distinction between "strict freedom" and "broad freedom"), or competence (crucial to certain conceptions of autonomy), and so on. 62 It was said earlier that in fact not just actions but other events too may be obligatory. The more general question, then, is whether we're concerned with event-types or eventtokens. But I shall continue to talk just of actions, for simplicity's sake. What I say can be generalized to apply to all events.

53

provided earlier in this chapter apply for the most part equally well to actions 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 twoterm 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 over act-tokens. As substitution instances for each of (2.28) and (2.29), respectively, 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 about whether a particular action is obligatory? After all, Jones isn't inquiring into the morality of helping old ladies in general, but simply into the morality of this particular helping. And doesn't this indicate that what he's really asking is whether or not an act-token is obligatory, rather than an act-type? The question is confused. Of course, we are frequently concerned about what to do here and now (or then and there) rather than in general, but this doesn't favor one reading over the other. The particularity that is built into the act-token o£jones's helping that old lady on Monday is attached externally to the act-type of helping that old lady by means of the relata of Jones and Monday. (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 lady on Monday is obligatory on Sunday' where "Jones helping that old lady" refers to a certain state ofaffairs. There are a number of reasons for thinking it preferable to conceive of actions as states of affairs, rather than as types or tokens. I provide such a treatment of action 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 we ought to do are frequently concerned with act-types. Whether or not this is so depends on how act-tokens are to be individuated. There are two questions to be asked here. First, how are act-tokens to be individuated across worlds? Second, how are they to be individuated within worlds? I shall consider each in turn. Suppose that Jones ought (on Sunday) to help that old lady on Monday and 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 old lady on Monday," occurs in the actual world. Suppose that in fact what Jones does is to help the old lady at noon on Monday. It would have been equally permissible for him to help her at six o'clock in the evening. If he had done this instead, would the act-token involved have been X, or would it 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 properly understood in terms of a certain act-token's being obligatory. Rather it would be best understood in terms of there being an obligation with respect to the disjunction X or Y or Z... (for an indefinite number of tokens) — 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 would mean that, when we say "5 ought at T to do A at T'," we frequently, and perhaps always, mean something else, namely, that S ought at T to do A at T" or B at T"", or... (for an indefinite number of tokens and times, where all of the tokens are of some common type and all of the times are contained in T'). It seems to me unlikely that we should systematically misexpress ourselves in this way, and thus likely that, when we say: "S ought at Tto do A at T"," we do not take "do ^4" to refer to tokens - at least, not if we take Xto be distinct from Yin the case just given. Of course, one way to try to avoid the foregoing problem is to insist that Xis identical with Y after all. But such coarse individuation of tokens across worlds gives rise to a further problem, at least when coupled with coarse individuation within worlds (of the sort advocated by Elizabeth Anscombe, Donald Davidson, and others).65 Suppose, as before, that X occurred; that is, Jones helped the old lady (at noon on Monday). Suppose further that, in helping her, Jones startled her, and that this is not something he should have 64 65

Cf. Mellema (1991), p. 31, and Vallentyne (1989), p. 308. 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 individuation within worlds claim that Xis identical with Z. But then how could X be obligatory when Z was not? One response would be to say that X wasn't obligatory; if Jones had helped the old lady without startling her, as he should have done, then it wouldn't have been Xbut some other token that would have occurred. But this response is not consistent with coarse individuation across worlds. Another response is to say that X (i.e., Z) was obligatory under the description "helping the old lady" but not under the description "startling the old lady." Such a response would echo what coarse individuators have said before in other contexts (such as the context of intentional action: some actions are claimed to be intentional under one description but not under another),66 but the price of accepting this response 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 descriptions. I doubt whether this accords with our common understanding of obligation. An additional problem with saying that it is act-tokens that are obligatory is this. If a token exists only when it occurs, then what is said to be obligatory need not, indeed need never, exist. It is not clear that it makes sense to talk of the obligatoriness of something that does not exist.67 One reason for thinking that we nonetheless do commonly conceive of obligation as concerning act-tokens is this. Many people want to say that an act is obligatory only if its consequences satisfy a certain condition; but types don't have consequences, only tokens do. I think this view is mistaken. Of course, we might say: (2.33) Jones's helping that old lady at noon on Monday caused that old lady's being pleased at six o'clock on Monday, and this makes apparent reference to a causal relation between two eventtokens. 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 67

See Anscombe (1969), p. 11 and elsewhere; Davidson (1980), pp. 5, 46, and elsewhere. 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 twoterm, relation) in order to keep obligation "simple." This is no less objectionable than "complicating" obligation (by introducing descriptions) and thereby keeping causation "simple." I suspect that this response is unsatisfactory, but I shall not pursue the issue here. For, to repeat, the account of obligation that has been given is intended for the most part to be adequate both to a conception of actions as types and to a conception of actions as tokens. Why only "for the most part"? One reason has already been given: those who individuate tokens coarsely both across and within worlds and who claim that it is tokens that are obligatory must supplement the formulae given with mention of descriptions (or employ some similar device). Another reason is this: statements (2.11)—(2.14) cannot be understood (as they 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 it is 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 perfectly intelligible where A refers to an act-type but not where it refers to an act-token. Still, even here the proponent of understanding obligation in terms of tokens need not concede defeat; he need only insist on revising such 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 of tokens, all of which are of some common type). Similarly, an "absolute" injunction against performance of a certain acttype 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 I shall mention four. 68

I advocated a similar rethinking in Zimmerman (1984), Section 3.1, where, as remarked in 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 are incommensurate with one another.69 For example, it might be thought that liberty and equality are both morally valuable, and that for that reason any world that contains one or the other of them is, to that extent, deontically good, but that they are nonetheless incommensurate, in that it is impossible to say that one is more or less valuable than the other or that they are equally valuable.70 Of course, this rough statement might be taken to mean a number of different things. It might be understood only to say that we are never in a position to tell which is more valuable, liberty or equality. Such epistemological difficulty need not point to any incommensurability in principle. Or the statement might be understood only to say that there is no general rule as to which is more valuable, liberty or equality, since the former is sometimes more valuable, sometimes less valuable, than the other. This too would not point to any incommensurability in principle; on the contrary, it presupposes commensurability. Or the statement might be understood to mean that on some but not all occasions liberty is neither more nor less valuable than, nor equal in value to, equality. This would be an instance of what Walter Sinnott-Armstrong calls "limited ^comparability."71 Or the statement might be understood to mean that on all occasions 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 incommensurate and that, as a result, it can happen that one is faced with accessible worlds that are deontically incomparable, that is, worlds of which it must be said that one is, deontically, neither better nor worse than, nor equal in value to, the other.73 What does the present account imply in such cases? 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 do A, S can refrain from doing A, and any accessible world in which S does not 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 (incommensurability 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 if liberty and equality are incommensurate, it may be that one world contains more of each than 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 deontically superior to every ~^4-world. Rather, it requires that, for every ~Aworld, there be some y4-world that is deontically superior to it. This allows for deontic incomparability between worlds. For instance, it may be that S is 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 sets between 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 5 ought not to do A. What's left? Obviously this: 5 may do A and S may not do A (in the sense of (IV)); that is, A is morally optional.74 In other, rougher words: if you have a choice to make between two deontically incomparable actions, you may take your pick; neither is obligatory, each is permissible. This is what my account implies, and it seems to me correct. The second type of indefmiteness that I want to acknowledge is that of vagueness. It may be that certain moral norms are vague, in that there are borderline cases where it is not clear whether or not they apply or whether or not they have been satisfied. Suppose, for instance, that Gertrude ought to be grateful to Lucy but that just what constitutes an adequate display of gratitude is unclear. Perhaps it is clear that simply sending her a thank-you note would be to do too little, while giving her a brand-new BMW would be to go far beyond what gratitude alone demands. But just where the minimum and maximum fall may not be clear. Would sending her a bunch of daisies suffice, or would it have to be lilies? It may be tempting to say that the difficulty here is merely epistemological, that gratitude's boundaries are well 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 where what's required of someone is in fact indeterminate, not just hard for us to determine. One such case is this: I promise Gail that I will help her with her gardening around noon tomorrow. Suppose that this gives rise to an overall moral obligation. Since the content of the promise is irremediably vague, it seems that the content of the obligation is also. Clearly, if I turn up 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 required to 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" would not 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 a finger, every once in a while? Will I have failed to do what I ought? Where is the line to be drawn? I think that we must accept that there is no precise point at which to draw the line, and hence that obligations can be vague in the way indicated. My account is not intended to imply otherwise. It is intended, rather, to be compatible with the claim that there are borderline cases with respect to whether or not a certain action, A, is featured in the best accessible worlds. Obviously this is a somewhat disturbing thought, but it is not a problem for my account in particular. Vagueness is disturbing generally; how to account for it satisfactorily is a general problem. Whatever the proper account of vagueness, I see no reason to think that it would not be consistent 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 certain counterfactual is true, one that concerns what would be the case if S were to make a certain decision. But can't it happen that such a counterfactual lacks a determinate truth-value? Perhaps it can.76 If it can, then in such cases the statement that S can do A also lacks a determinate truth-value and so, too, each of the statements that S ought to do A, S may do A, S ought 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 indeterminate 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 problem with my account. Rather, it is an issue concerning counterfactuals in general. Of course, one could try to avoid the problem in the present context by denying that obligation is tied in this way to counterfactuals, but I think that this would be incorrect.77 The final type of indefiniteness concerns the possibility (which I shall accept for argument's sake)78 that an agent be faced with an infinite number of accessible worlds and that each world is such that there is a deontically 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 ~Aworld is deontically inferior to some ^4-world. This is quite compatible with there being no deontically best world; it just means that, if S ought to 75 76 77

For a limited discussion of vagueness and obligation, see Sorensen (1991). Cf. Lewis (1973), pp. 66-7, 91-4; Mendola (1987), p. 127ff. 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 must be 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 infinite number of accessible worlds so ordered that, although every ~y4-world is worse than some ^4-world, it is also true that every ^4-world is worse than some ~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 ought not 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 not the case that S ought not to do A.79 Notice that (II) and (IV) imply, even in 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 precisely what should be said in such a case. Feldman calls the sort of situation where 5 is faced with an infinite number of increasingly good accessible worlds "delightful."80 But he overlooks an implication that is somewhat disturbing, the implication that, no matter what S does, he cannot avoid doing something wrong.81 For if an accessible world, Wu is distinct from another, W2i it must be because S acts differently in W1 than in W2;S2 and if W1 is deontically inferior to W2, then S must act wrongly in Wt. One way to try to avoid this result would be to amend the account of wrongdoing provided above. As (III) now stands, it implies that it is possible that S ought not to do A (that is, it would be wrong for 5 to do A), even though there is no other act B that S ought to do instead; for it may be that 5 ought not to do B, either. Indeed, this is precisely the case in the sort of situation presently envisaged. Such an implication may well seem odd, and it could be obviated by adding to (III) the condition that there is an acces79

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 even though S is faced with an infinite number of increasingly good accessible worlds. (See Feldman, 1986, pp. 37—8.) This is because his account says (roughly) that 5 ought to do A just in case some ^4-world is better than every -~v4-world. In order to accommodate deontic incomparability, however, my account (as noted above) says only that every ~Aworld is worse than some A -world. Such an account requires supplementation with subclause (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 differs from another, it will be (at least) because the agent makes a choice in one (namely, the choice 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 other accessible worlds. Thus, we could say, one can do wrong only if there is a best that one can do. To avoid the appearance of being ad hoc, this suggestion could be coupled with the observation that the account already implies a parallel thesis, namely, that one can be obligated only if there is a less than best that one can do. (For, in the limiting case where all one's accessible worlds are such that none is deontically superior to any other, everything is permissible, nothing is obligatory; there is not even the disjunctive obligation to take one's pick, since — as (X) declares — such an obligation 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 parallel theses to which appeal can be made other than the pair just mentioned in the last paragraph. These are the theses that one can do wrong only if one can do better, and that one can be obligated only if one can do worse. These two theses are implied by my account as it already stands, and it is arguable that whatever intuitive plausibility the first pair of theses has is itself to be derived from the plausibility of this second pair. And third, the suggestion of course has the implication that, in the sort of case under discussion, one can do no wrong. This seems unacceptable to me, for it gives one license to actualize some deontically abysmal accessible world, even though there is a far (indeed, infinitely) superior accessible world that one could choose to actualize. I think, then, that it is best to accept that, in the very special sort of situation envisaged, it is indeed the case that S cannot avoid doing something wrong. Thus the present account (under the assumption that an agent may be faced with an infinite number of accessible worlds, each of which is deontically inferior to some other) does not rule out moral dilemmas of a certain sort. Notice, however, that the sort of dilemma at issue may be called (following Peter Vallentyne) a. prohibition dilemma: all of S's options involve wrongdoing. This is not the same as saying that obligation dilemmas are possible, that is, that it can happen that 5 ought to do some act A and ought to do 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 extending Vallentyne's own use of "prohibition dilemma" somewhat. According to him, a prohibition dilemma has the form ~P(A) & ~P(~A). The account given in (II) and (IV) rules out the possibility that ~P{A) & ~P(~A) be true. But, as just noted, it does not rule out the possibility that all of an agent's options involve wrongdoing, and the term "prohibition dilemma" seems to me aptly applied to such cases.

62

dilemmas are impossible (barring certain complications). This is an issue that 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 will be aware of the vast number of formulae which have been touted as either axioms or theorems of one or another system of deontic logic. These formulae are usually alleged to express certain truths about obligation. Consider the following (which have been selected both because they are representative of the sorts of formulae in question and because it will be especially 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 & Oq); P(p v q) = (Ppv Pq); . O(p Dq)D(Op DOq); (pD q) D (Op D Oq); Op D OOp; Pp D OPp.

Each of these formulae seems to suggest a proposition that looks reasonable, but, if you're at all like me, you'll be wondering just what proposition is 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 logicians have been expended prematurely. In many cases they have attempted to construct formal systems on the basis of considerations too weak to bear the weight imposed upon them. More work needs to be done on the basic concept of obligation before such systems can provide genuine philosophical insight. Nevertheless, such systems are certainly not without their philosophic uses. If formulae such as those just presented are at all plausible, then they can serve as a gauge of the adequacy of an account of obligation such as 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 permis63

sibility, respectively, but what are the variables p and q supposed to range over? Presumably propositions. This is so for two reasons. First, the connectives ~, D, =, &, and v have a familiar interpretation when applied to propositions and propositional variables, but not otherwise. Second, and less importantly, certain parallels with alethic modal logic (the logic of necessity and possibility, taken by many to be analogous in various ways to obligatoriness and permissibility, respectively) may then be drawn. The trouble is that, although such treatment of the variables lends itself nicely to the expression of such claims as that it ought to be that little children not have to suffer, it is hard to see how it can adequately handle such claims as that Tim ought to do what he's been told, given that "It ought to be that Tim do what he's been told" does not express the same claim. (See Section 1.2 on binding versus nonbinding senses of "ought.") One way around this would be to read Op as "5 ought to see to the occurrence ofp," but this has difficulties of its own. Suppose that Roger ought to raise his hand. Just what event or proposition is it that Roger ought to see to the occurrence of, his raising his hand or his hand's rising? In neither case is it clear that we get an equivalent claim; in one case too much seems to be said, in the other too little. To say that Roger ought to see to it that he raises his hand seems to say too much, in that it suggests that he ought to do something over and above just raising his hand. To say that Roger ought to see to it that his hand rises seems to say too little, in that it appears compatible with his fulfilling his obligation by way of ensuring that his hand rises in such a way that it is not he but someone or something else that raises it. In light of these considerations, I think it best to rewrite the formulae under consideration so that they conform with the account of obligation that 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 [at T i n W]. It is clear that my account - specifically, (I) and (II) above - implies that this is true. If all ~vl-worlds [accessible to 5 from Wat T] are worse than (that is, deontically inferior, for 5 at T, to) some ^4-world [accessible to S from Wat 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 are worse than some /1-world, then it is not the case that all v4-worlds are no better 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 due to a lack of personal optionality, that is, true in virtue of there being either no /1-world accessible to S or no ~A-world accessible to S (that is, in virtue of 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 the moon; hence no one may refrain from doing so; but it doesn't follow that everyone ought to do so. Still, the following is true on my account (where C 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 that all ^4-worlds are no better than some ~/l-world, then all ~y4-worlds are worse than some /1-world. Furthermore, in light of the facts that obligation requires personal optionality and that, where there is personal optionality, 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 that S ought [at T] not to do A [at T' in W]. This is because, if all ~y4-worlds are 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. Of course, (2.38') is controversial (as are, perhaps, some of the other claims already 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 as to 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 do A and S may not do (that is, may refrain from doing) A. But I think it is plain that proponents of (2.39) have had something else in mind. Just as p v q is standardly taken to be equivalent to ~(~p&~q), so too, I think, we can 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 false on my account. This is because its truth would imply the truth of C(~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). Thefirstof 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, one might claim that every genuine action is indeed personally optional; it cannot be, for instance, that S acts in moving his arm unless he could have done otherwise. (Certainly, it can happen that one's arm moves irresistibly, but then it is not oneself who is moving it.) Thus, where O(A&B) is true, each of 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 three reasons: hard determinism may rule out the possibility of free action, but it doesn't rule out the possibility of all action whatsoever;85 certain cases of actual action are plausibly understood to involve irresistible impulses, cases such as, perhaps, certain compulsive or obsessive actions, actions due to addiction 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 been forced 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 variables are not restricted to ranging over actions. (See the discussion in Subsection 2.2.1 above.) And certainly some sorts of events that can be optional for an agent don't have to be. For instance, O(A) might on occasion express "it is obligatory for me that I be indoors," and yet there can clearly be times when my being indoors is not personally optional for me (either because I cannot get indoors or because I cannot get outdoors). A second response might be to acknowledge the second point just made and yet to insist that, if O(A&B) is true, then each of O(A) and O(B) must be too, whether or not what is obligatory is said to be an action. Since my account doesn't imply this, it ought to be revised. Specifically, (V) should be revised so that it implies, not just that (A&B) is optional, but that each of A and B is too. But this response is problematic, too. First, it seems that a revision would be called for not just for (V) (the analysis of O(A&B)) but also for (I) (the analysis of O(A)). Consider, for instance, the claim that Chuck ought to open a checking account. (I) implies that opening a checking account is optional for Chuck. But, on the present suggestion, it ought to 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 we may now say that Chuck ought to do (X&Y&Z). Then (V), revised according to the present suggestion, would imply that each of X and Y and Z is optional for Chuck, whereas (I), left unrevised, does not have this implication; (I) implies only that (X&Y&Z) is optional for Chuck. Thus (I) would need 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 difficult to accomplish. A second problem with this second response is simply this: 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 idly by while there is poverty." Why wouldn't it be correct to analyze this in the terms set out in (V), namely: (a) there is a (not-standing-idly-by-andpoverty)-world accessible to you; (b) there is a not-(not-standing-idly-byand-poverty)-world accessible to you; and (c) all worlds of the latter sort are worse than some world of the former sort? Surely this can be true, even though every world accessible to you is a poverty-world; that is, nothing you 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 complains while doing them is. This is enough to give the claim some plausibility. Whether or not the claim turns out to be true, we should not dismiss it 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 that three very closely related claims are nonetheless true on my account. The first 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)-world is 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 at least 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 be discussed further in Chapters 6 and 7), is true on my account. Suppose that all ~^4-worlds are worse than some /1-world. Such an A-world must be either an (A&B)-wov\d or an (A&~B) -world. But if we suppose also that all ~B-worlds are worse than some B-world, then all ~y4-worlds are worse than 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 the v 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 of the 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 neither 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 hence that 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 so too 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 worse than some ^4-world. Either B is performed in the y4-world or it is not. If it is not, then, given the truth of O~ (A&~B), there is a still better world in which both A and B are performed, and all ~B-worlds will be worse than this world. It is not clear just how (2.43) should be rewritten. Perhaps we can put it 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 without doing B, then, if he ought to do A, then he ought to do B. My account implies that this is false. So what if S doesn't do A without doing B ? Perhaps he 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 illustration, let A be the action of helping Hilda and B be the action of humiliating Hilda. Perhaps Laura ought to lend Hilda a hand and refrain from her usual caustic comments; if she helps Hilda but also humiliates her, it hardly 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 logical necessity, physical necessity, or something weaker), then, if he ought to do A, then he ought to do B. The foregoing example about humiliation is inapplicable here. If S cannot do A without doing B and the best accessi69

ble worlds are A- worlds, then afortiori 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 without doing 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 this chapter and in Chapters 6 and 7. The final two formulae are difficult to rewrite to conform with the present 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 time of S's acting so that S ought to or may do A is prior to the time of S's actually doing A. Thus these propositions apply at best when there is an implicit clause in their consequents to this effect: if S has time so to act. On this understanding, (2.44') is true, on my account, but (2.45') is not. As to the former: if all ~/l-worlds are worse than some ^4-world and then S so acts that this is no longer true, then S clearly departs from the high road and goes wrong. Consider Figure 2.1 again, now under the assumption that W\ is deontically superior to all the other worlds. Given this assumption, S ought 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 no longer be true that he ought (prior to T3) to do C at T3. But this will be because he went wrong at either Tt or T2.1 shall address this issue in greater detail in the next chapter. As to (2.45'): this is not true, even given the implicit extra clause, and that is simply because it is permissible to make certain preparations to do what is itself permissible, even if these preparations rule out doing something else that is permissible. For instance, perhaps I may either go to the theater or go to the opera this evening, although I cannot do both. If, as I may, I spend my money on tickets to the theater, then I have rendered myself unable to go to the opera; hence it is not the case any longer that I may now go to the opera; hence it is not the case that I ought 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 a sense 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 outset 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 is morally perfect if and only if he never acts wrongly.90 Note that to state that we 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 an obligation to be perfect in some other sense (for instance, in the sense that our characters, or motives, or whatever, be beyond all moral criticism) where it might be impossible for us to achieve such perfection (or unreasonable to claim that we should do so). Nor does it imply an impossibly high ideal for action. It can of course happen that an agent can perform each of 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 obligation to be perfect (in the present sense) does not commit me to stating that he is obligated to perform all of these very good deeds.91 On the contrary, since he cannot perform them all, my account commits me to saying that it is not the 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 among the various good deeds in question, performing some and not performing others. As long as he does avoid wrongdoing, he will have satisfied the obligation 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 been discussed are representative of how my account handles such matters. I have explained why some of the formulae are not implied by the account, and I hope that the explanation serves to confirm the acceptability of the account. Some of the formulae that are implied by the account are admittedly controversial; 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] do A [at r ' ] . This controversial claim will be examined in Chapter 3. When conjoined with other implications, (2.46) of course yields yet further implications. For example, 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 91

See McGinn (1992), p. 33. Contrast Clark (1993), pp. 54-5.

71

and this is true, on my account, even though (as just noted) the following is 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-called paradoxes. These paradoxes are not formal inconsistencies; rather, they are cases where the system's representations of certain "ought"-statements seem separately plausible but jointly implausible, since when conjoined they have, or appear to have, counterintuitive implications. Some such paradoxes will be treated in Chapter 3 (when self-imposed impossibility is discussed) 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 may be 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, since the latter seems to warrant saying (for instance) that, if Matt ought to mail a letter, then he ought to either mail or burn it, and this seems to grant Matt license to burn the letter. It might be tempting to try to undermine the paradox simply by pointing out that (2.43) is false (as argued above), but in fact this won't do; for a strengthened version of (2.43) has been acknowledged as true, and the argument can be reformulated around it. So let me reformulate it in this way, now rewriting it so as to conform with the present account. 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 neither 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 ought to do A and can refrain from doing either A or B, then 5 ought to do either A or B. But isn't this to be rejected? As noted, it would appear to give employees of the postal service license to burn rather than mail the letters entrusted to them. The resolution of this minor puzzle is of course to be found in the distinction between O(A v B) and 0(AorB). Briefly, it is only the latter that gives S license to do B (as spelled out in (X) above), but it is only the former that is at issue in (2.50'). O(A v B) means simply that 5 ought not to do neither A nor B, and clearly this is true if 5 ought to do A. End of problem.93 Formula (2.43) gives rise, though, to another, much more interesting paradox, commonly known as the Good Samaritan Paradox.94 Let p be the proposition that Harry helps Vera, who has been robbed by Rita. This entails q, the proposition that Vera has been robbed by Rita. Suppose that Harry ought to help Vera. (2.43) would seem to imply that it follows that Rita ought to have robbed Vera - a disturbing conclusion. Now, some systems of deontic logic do imply this conclusion and so seem properly criticizable 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 A without 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 what 5 ought to do from what S ought to do; it will not in general permit inferring an obligation on the part of Rita from an obligation on the part of Harry. Second, if Harry ought now to help Vera, and Rita's robbery ofVera lies in the past, then once again no inference with respect to an obligation to rob is forthcoming; for "ought" does not apply to the past but only to that 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 Good Samaritan Paradox to rest. For the paradox can be recast so as to concern only 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) will rob at T3, and B is the action of Harry's robbing Vera. Suppose that Harry ought at Tj to do A at T 2 . Given (2.43'"), it seems to follow that Harry ought at Tt to do B at T3; for, it seems, Harry cannot at Tt do A at T 2 without doing B at T 3 , and Harry can at Tt refrain from doing B at T3. Clearly this is an unwelcome result. The proper response is this. The claims that Harry ought to do A and that Harry cannot do A without doing B are jointly unacceptable. Which is to be rejected depends on just what A is taken to be. Perhaps this is best explained by explicitly invoking worlds in the description of actions. If A is the action of Harry's helping in world H^the woman whom he robs at T3 in the actual world, then it is evident enough that Harry can do A without doing B. For certainly there are worlds accessible to Harry at T1 in which Harry helps Vera at T 2 without robbing Vera at T3, even though it is true that Harry will in fact (that is, in the actual world) rob Vera at T3. If A is the 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 then we should not say that Harry ought at Tt to do A at T2. For B does not occur in the best worlds accessible to Harry, and if Harry cannot do A without 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 part because the example lends itself so readily to an understanding of A and B according to which doing A does not entail doing B. Let me turn, then, to a version of the Good Samaritan Paradox where it is clear that doing a certain thing 4 does entail doing another thing B. This is the version commonly known as the Paradox of the Knower. Suppose that Greg, a security guard, ought to provide truthful reports of any wrongdoing done on his watch and hence (to keep the discussion to a single-agent case) ought to provide a truthful report of any wrongdoing committed by himself. (The paradox was originally put in terms of knowing rather than truthful reporting, 96 but the latter is more suitable to single-agent cases.) Suppose that Greg ought not at Tj to do A at T2 but in fact will do A at T 2 . Is it the case that Greg ought at Tj to provide at T3 a truthful report of his doing A at T2? Clearly not, and this is because his doing so entails his doing A at T 2 . The best worlds accessible to Greg at Tt are worlds in which he does not do A at T 2 and hence does not provide at T3 a truthful report of his doing A at T2.97 This is 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 the various deontic paradoxes. (Again, some of these paradoxes are to be discussed in later chapters.) On the other hand, certain implications of the account have been acknowledged to be controversial. How, then, can I propose the account as an analysis of the concept of obligation? Aren't analyses supposed to be neutral between various conflicting conceptions of a concept? Shouldn't, for example, an analysis of "ought" neither imply that "ought" implies "can" nor imply that it doesn't?98 This is a tricky issue, which could be circumvented simply by declaring that (I)-(XII) are intended to be understood as necessary truths and leaving it at that, without declaring in addition that they are intended to be taken as analyses. But this would be misleading, for I do intend them to be taken as 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, that of 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 knowledge entails belief, that knowledge entails truth, and so on, and each of these implications has been contested. But consider what Feldman says. Although he conceives of his project as one of "laying out the logical structure of the concept of moral obligation,"99 which matches a description that I have given of my project, he denies 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 can in 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 occurrence of a state of affairs, p, iff p occurs in some world accessible to s at t, and it is not the case that ~p occurs in any accessible world as [intrinsically] good as (or better 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 proposed. 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 empty tautology. If my definition, MO, had been intended as an "analysis" of the concept of moral obligation, then this expectation would perhaps be justified. But the definition was not intended to be an analysis of some concept we already have. It was intended to be a stipulative definition of a concept to which I wanted to draw attention. If we compare (2.51) with the following principle, I think the difference between 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 should grant that (2.52) is an empty tautology. It tells us nothing of interest about morality. Even those who think that my approach is hopelessly wrongheaded.. .will accept (2.52). On the other hand, (2.51) is intended to tell us how we ought to behave. It tells us that we morally ought to bring about all and only those states of affairs that fall under the concept MO... So even if (2.51) is a necessary truth, it is not an empty tautology. It is a fundamental moral doctrine... Since some will disbelieve the proposition expressed by (2.51) while not disbelieving the one expressed by (2.52), these propositions are distinct... This "open question" argument serves to emphasize the fact that my definition, 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 we would have to say that the following proposition is not an analysis of the concept 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 belief that p . 101 Feldman (1986), pp. 207-8. The numbering of propositions has been changed to conform 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 will reject either or both of (2.51) and (2.53). But analyses need not be uncontroversial; they need not be neutral with respect to various conceptions of the concept that is analyzed. Perhaps some people102 believe that analyses must be neutral and uncontroversial because they believe that analyses are analytic and that analytic statements are trivial, and hence neutral and uncontroversial. 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 in this 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) (or something like it), belief is in a way conceptually "contained" in knowledge. But since (2.53) is not itself uncontroversial, there is no reason to expect (2.55) to be, even if it is analytic. Similarly, if my proposed analysis of obligation is correct, then so too is the 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 to be uncontroversial or in any way neutral with respect to all possible conceptions of obligation. In saying this, I am of course committed to saying that anyone who denies (2.46) is making some sort of conceptual error; but this is not to say that the denial of (2.46) involves a logical error, at least not in the sense that it involves an inconsistency in and of itself, as would, for example, 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 synthetic, how are they to be distinguished from other synthetic necessary equivalences that are not analyses? I wish I knew the answer to this question, for it would mean that I could clearly distinguish conceptual errors from errors of some other sort (and thereby, perhaps, clearly distinguish metaethics 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 he would be mistaken. I wouldn't accuse him of a conceptual error, though, but rather 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 component concepts, while this is what an analysis attempts to do. But other claims in this vein are not so easy to judge. Consider Feldman's own account 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' &p is 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 claim substantive, and not purely conceptual. Even if Feldman is correct about the link between obligation and intrinsic value, those who deny that he is are making a substantive error; if he is incorrect, then he has made a substantive error — he hasn't made an error concerning the "structure of the concept" of obligation. Or so it seems to me, but I confess I don't know how to argue for this or against the claim that the errors mentioned are indeed conceptual rather than substantive. The second question is this. I say that someone who denies (2.46), for example, is making a conceptual error. But might it not be that someone whom I take to be denying (2.46) is not in fact denying it but rather something else? For might not this person be using "ought" (or "can") to express a concept different from that which I take it to express, so that the proposition 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. Perhaps there are ways of trying to rule this out — I can test him, for example, on just when he takes it to be appropriate to use "ought" (or "can"), skillfully varying the contexts so as to get a fix on his meaning — but the disturbing possibility remains that those who affirm what they take (2.46) to express and those who deny what they take (2.46) to express are not in reality 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 on philosophical ethics than the proposition that "ought" implies "can." I have noted 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] do A [at T']. That my account implies the truth of this proposition will no doubt incline many 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 that seem to me stronger (but still unsuccessful). In Subsection 2.2.4 I distinguished two senses of "can" such that "ought" implies "can." The weaker sense 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 of being free to perform an action (stronger in that a person's being free to perform an action implies, but is not implied by, that action's being personally possible for that person). In what follows, it is "can" merely in the sense of personal possibility that is explicitly at issue - except when alternate possibilities are discussed in Subsection 3.1.5, where it is "can" in the sense 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 own lights, in that the account that I have proposed is avowedly intended to accommodate 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 fact not 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 the agent.

On one reading what this means is the following: (3.1/) of all actions (or courses of action), whether open to the agent or not, A is that which is probably (or apparently) the best one open to 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 is that which is probably (or apparently) the best. Here, of course, it is implied that A is open to the agent. (By way of analogy, suppose that someone, looking down from a balcony at a room full of people, 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 richest Republican. Now this may be true and yet Rex not even be a Republican. But the claim in question might instead be this: (3.2") of all the Republicans present, Rex is probably (apparently) the richest. This cannot be true if Rex is not a Republican.) The fact is that, while my account accommodates theories according to which (3.1") corresponds to what is obligatory, it does not accommodate theories according to which (3.1') corresponds to what is obligatory. On my view, anyone who thinks that (3.1') corresponds to what is obligatory is making a conceptual error (just as, for instance, anyone who thinks of obligation along the lines mentioned 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 with the 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 conceptual error 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 morally permissible for Trudy not to travel faster than light; for what is physically 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 permissible not to do it, and so it turns out that Trudy is obligated to travel faster than light, even though she cannot. This is a distinctly odd argument. Let A be Trudy's traveling faster than light. 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 to assert its first premise, (3.3). The answer, of course, is that we should not say that something (such as not traveling faster than light) is permissible if it is impossible to avoid. In other words, (3.3) is itself based on the following 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, something has gone wrong here. Some might diagnose (3.3) as the culprit, but I do not. I think it's quite right to say that "may" implies "can avoid." My account 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 can do that we have occasion to use such an expression as 'We ought to deal with this matter now, if we can'. The qualification 'if we can', like the qualification 'if we 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 unsure whether 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 not speaking 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 the statement 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 still true on Thursday and on all future days, because it was true on Tuesday that I ought to mow your grass by Saturday. In contrast, (3.9') is not true on Thursday, because it is not yet true that I ought to have mowed your grass in the past (before Thursday), since I promised only to mow it by Saturday.7 Sinnott-Armstrong then proceeds to provide two analogous interpretations 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 true even when the promise in question is not fulfilled by Saturday and thus even 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 an interpretation of (3.9) is unpersuasive, in that it overlooks a third interpretation 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 your grass at T, and Tis past. In the example given, Tis Saturday, and it is for this reason that (3.9) is not true when uttered prior to Saturday. Now, if "ought" implies "can," then 7

Sinnott-Armstrong (1985), p. 46. The numbering of propositions has been changed to conform 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 your grass at T, and Tis past. And, I believe, this implication does indeed hold. The fact that, as SinnottArmstrong correctly says, (3.9) can be true whereas (3.10') is false, is irrelevant; 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 done something, even if he or she did not do it. We need to be able to make present judgments about what ought or ought not to have been done in the past in order to justify present punishment for past acts or failures to act. The correct interpretation of such judgments cannot have the form of (3.9") — [or, I am sure Sinnott-Armstrong would add, (3.9'")] - because [such] judgments.. .are true in many cases where punishment is not justified, such as when it was no longer true at the time at which the agent did the act that he or she ought not to 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 question for me to insist, as I do,9 that, if S ought at T to do A at T', then T" is not earlier than T, but I can certainly give an account of the propriety of "present judgments" that differs from Sinnott-Armstrong's. Sinnott-Armstrong seems to suggest that punishment is justifiable only ifjudgments like (3.9') can be made, for judgments like (3.9") (and (3.9'")) can be true even when punishment is not justifiable. It is of course true that punishment may not be justifiable even when judgments like (3.9'") are true, but this does not warrant 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 judgments ofresponsibility, not (simply) ofobligation, that justify punishment, and it is certainly possible that someone should now be responsible for a past action, even though it is not possible that someone should now be obligated to perform a past action.10

8

Sinnott-Armstrong (1985), p. 47. The numbering of propositions has been changed as before. 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 require alternate 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 require alternate possibilities in fact implies the claim that "ought" does not imply "can."11 One cautionary note: up to now, "can" has been understood in terms merely of personal possibility or accessibility, but in the present discussion of alternate possibilities it is necessary that it be understood in terms of freedom. (On this matter, see the second paragraph of this chapter.) To make 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 possibilities that Frankfurt and I and possibly Widerker12 reject is this: (3.11)

a person is morally responsible for what he has done only if he could 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 rejection 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 moral blame: (3.11a) S is morally blameworthy for doing A only if 5 could have done other 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)

11 12 13 14

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

Widerker (1991). Widerker is not fully explicit on this issue. Frankfurt (1969), p. 829ff. 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 ought to do other than A, then S can do other than A.) But (3.12), far from being necessarily true, is false. This is because moral responsibility is a function of what one believes about one's moral obligations and not of what one's moral obligations 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 that one was not justified in doing it. Thus blame worthiness does not entail wrongdoing, and to justify an action is not to exculpate its agent. (Failure to recognize this fact too often prematurely terminates our moral inquiries.) It is also the case (although this is not strictly germane to the present discussion) that wrongdoing does not entail blame worthiness; this is widely accepted. Wrongdoing and blameworthiness are thus logically independent of one another (though not conceptually so; the latter is to be analyzed in terms of the former), and Widerker's argument against Frankfurt fails. It could, of course, still be that Widerker's conclusion is true. But I believe otherwise. I think that (2.46') doesn't imply (3.11a), because the latter is false but the former true. The reason is this. (3.11a) is a contraction of two other principles: (3.13) S is morally blameworthy for doing A only if S could have done A; 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 obvious that we are dealing with the stronger "can" of freedom — C' — and not merely that of personal possibility — C. For it would hardly require the insightful 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 this proposition: (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 responsibility and freedom, a link that, I believe, holds of conceptual necessity. In this regard, the concept of moral blameworthiness is just like many other fundamental moral concepts. For example, in addition to (3.13), we may also accept: (3.15)

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

More importantly, (2.46') is just one more instance of this link between morality andfreedom, 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 alternate possibilities. 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 by an appeal to (2.37"), not (3.14), and this leaves room for the cogency of Frankfurt-type cases against the link between responsibility and alternate possibilities by means of an attack on (3.14). Some questions may remain, though. One is this. Even if it is granted that (3.14) is false, might not the link between responsibility and alternate possibilities be restored by means of an argument similar to the one that I have given that aims to preserve the link between obligation and alternate possibilities? In particular, might there not be a responsibility analogue to (2.37") that functions as (2.37") does in my argument? The answer is that I 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 blameworthy 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 only if 5 would be blameworthy for failing to do A. For more is required for praiseworthiness than the avoidance of blameworthiness, and more is required for blameworthiness than the failure to act in a praiseworthy fashion.17 Another question is this. Doesn't it simply seem that Frankfurt-style counterexamples work just as well when applied to obligation as when applied to responsibility? Consider, for example, this case. Murphy freely commits murder but, for Frankfurt-type reasons, couldn't have done other than commit murder. Surely Murphy does something wrong. But, given (2.37"), it follows that (2.460 *sfelseafter all; t n a t is> it follows that "ought" does not imply "can." A similar point can be made concerning (3.17), the claim that "ought" implies "can avoid." Suppose that Ralph freely refrains from committing murder, although again, for Frankfurt-type reasons, he couldn'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 thefirstpremise of each argument but, because he rejects their conclusions, rejects their second premises. One alternative would of course be to reject theirfirstpremise 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 the critic given good, Frankfurt-type reasons for rejecting their conclusions? I don't think so. I would readily grant that Murphy does something that is in some way morally bad, and that Ralph avoids doing this; but, because neither agent can act otherwise, it is incorrect to say that Murphy does something wrong or that Ralph acts as he ought. (Similarly, it is incorrect to say that it is or would be wrong for Ralph to commit murder, where this is understood to include the rider: under the circumstances. I can of course agree that it would be wrong for Ralph to commit murder if he had the option of doing so.)

This is not a particularly satisfactory situation. It may seem that, when the critic makes a point, I simply dig in my heels; when I make a point, he digs in his heels. Is there a way to resolve this? That is unclear to me. I can point 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, by implication, (2.46') and (3.16)) against other attacks. All of this seems to me to favor my position rather than the critic's, but I don't suppose it will win over everyone. 3.1.6 Psychological incapacitation Having mentioned "can avoid" as well as "can," I shall turn now to an argument against the claim that "ought" implies "can avoid," reverting to the weaker sense of "can" that merely involves personal possibility (for there is no longer any need to stick with the stronger sense). The claim now at issue, then, is this: (3.17') ifO(^4),thenC(M). And the argument is simply this. It can happen that a person's moral training during childhood is very rigorous, so rigorous that, in later years, he is psychologically incapable of acting otherwise than in accordance with his moral obligations. This implies that (3.17') is false. While I agree that moral training can be very rigorous, I am doubtful that it can lead to such incapacity. Still, perhaps it can. George Washington declared himself (while still a child) unable to tell a lie (although he was apparently 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, maybe he did; for perhaps he was obligated, at a time prior to his becoming unable to tell a lie, not to tell a lie at that later date when he had become unable to do 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 had become 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 avoided doing something that was in some way morally bad, but he did not avoid doing 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 reminded here of a lovely remark attributed to Mark Twain: "I am morally superior 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 doing so is to be distinguished, morally, from one who "refrains" from doing so because he is unable to do so.20 Similarly, one who feels fear but nonetheless acts courageously is to be distinguished, morally, from one who acts fearlessly. And so on. Of course, it might be argued that the moral distinctions in question have to do not with obligation but with something else (character, perhaps). I would agree that they have to do with something else, but I don't see it as digging in my heels to say that they have also to do with 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) (3.21) (3.22) (3.23)

I ought not to have felt so covetous; you ought to feel gratitude; he ought to be ashamed;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 possible, I would say that, given the irresistibility of the impulse, such action cannot be either obligatory 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 what the agent ought to do he cannot do. A standard response to cases of this sort is 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 in the account of obligation provided in Chapter 2 does imply "can."25 This response is frequently ridiculed. White, for example, says that "it offers no reason, other than its alleged ability to explain failing to feel what one ought, for supposing that there is a second sense of 'ought'."26 SinnottArmstrong similarly complains that the response is question-begging.27 But while I grant that, if no independent reason for distinguishing two senses of "ought" were given other than preservation of (2.46) in face of the examples given in (3.20)—(3.23), then this would constitute a begging of the question against opponents of (2.46), the fact is that, despite what White and Sinnott-Armstrong say, such independent reason is forthcoming. Moore himself distinguishes that sense of "ought" having to do with what is 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 very distinction that I introduced in Section 1.2, when contrasting what I called the "binding" and "nonbinding" senses of "ought." I made no mention there of the issue of whether "ought" implies "can"; nevertheless, I submit that the distinction had intuitive appeal. If any further evidence of the distinction is needed, simply recall the fact, mentioned in Section 1.2, that certain "ought"-statements (those which express the nonbinding sense of "ought") are what I called "passive-transformable," whereas others (those which express the binding sense of "ought," that is, obligation) are not. In the 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, be sustained. It may of course be true on occasion that someone is in (indirect) control of his emotions; perhaps he can suppress his covetousness, or summon up gratitude, or induce shame, and so on.28 If so, then he may well be obligated 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 or not he has the emotions in question; but if they are true in this way, then the "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 (such as covetousness, gratitude, shame, and amusement). On the contrary, the case just given of Jones and the old lady has to do with an action, that of helping. 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, obligation sense, that is operative. A final point is this. What has been said about uncontrollable emotions applies also to uncontrollable motives. Ross is well known for making this point: That action from a good motive is never morally obligatory follows.. .from the Kantian principle...that 'I ought' implies 'I can'... I can act from a certain motive only if I have the motive; if not, the most I can do is to cultivate it by suitably directing my attention or by acting in certain appropriate ways so that on some future occasion it will be present in me, and I shall be able to act from it. 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 to make unwarranted inferences from this observation. For example, although it may be that, imbued with malice, I cannot summon up any sympathy 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. For while 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 motive that we do have. In saying this I am, like Ross, clearly not using "motive" to refer simply to that which does infact "move" one to action; rather, I use it to refer to that which has the capacity to do so. Motives, in my sense, may thus compete with one another for the role of "actual motivator," and, I am 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 obligatory and what is ideal is accepted, still there are actions which are obligatory, and not ideal, and yet which cannot be performed. Alan Montefiore gives this example: "I ought to be at the meeting tonight, but unfortunately I cannot manage it."32 Of such a case White says: "We don't blame him because we accept his inability as an excuse. But it wouldn't be an excuse for not doing what he ought to do if it could make that cease to be what he ought to do."33 Similarly, Sinnott-Armstrong says: [I]f a bride cannot get to her wedding, then she ought at least to let someone know and to excuse her absence... [I]f it were not first true that she ought to go to her wedding, she would have nothing to excuse when she failed to go to her wedding. Excuses are distinguished from justification in that to justify an act is to show that it is right, whereas to excuse an act is to admit that it is wrong but 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 declared that excuses are to be distinguished from justifications in just the way that Sinnott-Armstrong mentions.35 But what Austin said seems wrong. Excuses aren't offered only where wrongdoing (that is, the failure to satisfy an obligation) is admitted. Consider this case, provided by Terrance McConnell: Suppose I am at a lake when a man drowns. Another accuses me of wrongdoing 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 in time).36 It is surely clear in such a case that the appeal to inability does not presuppose any admission of wrongdoing. And yet it would be quite natural to say that the agent in the example offered his inability as an excuse. N o w , there are two ways to go at this point. One might simply insist that excuses do presuppose wrongdoing, just as Austin claims. Indeed, McConnell himself says this. But then, as McConnell states, it would be inaccurate to say of his example that in it an excuse is being offered. This would mean that what many find natural to say about his example is in fact incorrect. I have no real quarrel with this. The moral would be that it is also incorrect, even if natural, to say of Montefiore's meeting case and SinnottArmstrong's wedding case that an excuse is being offered. Hence there is no 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 reject Austin's claim that excuses presuppose wrongdoing and to assert that what it seems natural to say in Montefiore's and Sinnott-Armstrong's and McConnell's cases is in fact correct. For what is an excuse? An excuse is offered in response to an accusation.37 It is a defense to a charge. It is a denial of responsibility when responsibility has been imputed. This can occur in a variety of settings, including those in which wrongdoing is admitted, but also (as I mentioned when responding to Widerker's argument in Subsection 3.1.5) including those in which it is not. Given this understanding of what an excuse is, we may admit that excuses have been offered in the three cases just mentioned without thereby being committed to denying that "ought" implies "can." A slight complication here is the following. In the meeting case, don't we have an admission of wrongdoing? After all, the protagonist has said: "I ought to be at the meeting tonight." But this statement is open to various interpretations. First, he may mean simply that he ought in the ideal, nonbinding sense to be at the meeting. Or he may mean that he has been charged with being at the meeting. O r he may mean (and I think this is very common) that, although he can (literally) manage to go (and, perhaps, has a prima facie obligation to go), he cannot do so without serious inconvenience, to himself or others (and therefore, perhaps, without violating some more 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 that he 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 mistaken; for it cannot be that he ought (overall) to do what he literally cannot do. 3.1.9 Self-imposed impossibility Even if the foregoing discussion of excuses is plausible, though, it may seem insufficient to salvage the claim that "ought" implies "can." For it seems that there just are cases where a person ought (is obligated) to do something but cannot (literally) do it, and these are cases where the person appears to have no excuse for not performing the action. Here I have in mind cases of what may be called "self-imposed impossibility." Many philosophers have appealed to such cases in an effort to prove (2.46) false.38 A typical case is this. Bruce borrows some money on Friday, promising to repay it on Sunday. Since nothing else of moral significance conflicts with such repayment, he is therefore overall obligated to repay the money on Sunday. But he spends all the money on Saturday night, thereby rendering himself unable to repay it on Sunday. Surely, even though Bruce cannot repay the money then, he is obligated to do so. To declare otherwise smacks of paradox,39 for what an easy escape Bruce would then have from the bonds of obligation — what an easy escape we would all have! We need only disable ourselves 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 the claim that "ought" implies "can avoid." It has to do with what may be called "self-imposed necessity." Suppose that Betty, like Bruce, borrows some money on Friday, promising to repay it on Sunday. Unlike Bruce, Betty is conscientious. To ensure that she keeps her promise, she takes the money 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 thwarting, in the manner of Ulysses resisting the Sirens, the temptations of Saturday-night spending sprees. Let us suppose that this action renders Betty incapable of not repaying the money on Sunday. Must we then conclude 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 development. 3.2 IMMEDIATE AND REMOTE OBLIGATION A crucial component in the account of obligation that I have proposed is the 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 be either direct or indirect. The example given there had to do with my throwing a brick through a store window. I had control over whether or not the window broke. Such control was indirect, however, being parasitic 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, conjoining the direct control over the decision with the indirect control over the window's breaking, which was a consequence of that decision. Let's embellish the story a little. Suppose that breaking the window was a necessary preliminary to my snatching the jewels, and I snatched them. Then the decision to snatch was something over which I had direct control, the disappearance of the jewels was something over which I had indirect control, and the snatching was something over which I had hybrid control. But that's not all. At the time I made the decision to throw the brick I had indirect control over each of the items just mentioned - the decision to snatch, the disappearance of the jewels, and the snatching. Thus, although (on the account that I favor) one can have direct control only over one sort of thing (decisions) and one can have hybrid control only over one sort of thing (nonminimal actions), one can have indirect control over all sorts of things, including decisions and nonminimal actions, since these can themselves be among the consequences of decisions. Now, as I stated in Subsection 2.2.1, this decision-plus-consequence picture is not essential to my account of obligation; what is essential is the distinction between direct and indirect control. And this distinction gives rise to a closely related distinction between immediate and remote control. This has to do with the times involved. The canonical phrase form here is this: (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 is remote; otherwise it is immediate.40 Thus we should understand the claim that "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 A atV. This is what my account of obligation implies. And just as control may be immediate 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 immediate and remote obligation,41 others have accepted it,42 and still many others have of course been silent on the matter.4 The distinction seems to me of first importance. Not only does it solve the problems of self-imposed impossibility and necessity, but it provides the wherewithal for a detailed account of the dynamics of obligation - the shifting of obligation and its supersession. It is to this that I now turn. 3.2.1 Self-imposed impossibility Consider again the puzzle concerning Bruce, who borrows money on Friday, spends it all on Saturday night, and thus fails to repay it on Sunday. Not only does the distinction between immediate and remote obligation solve 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, Bruce ought to repay the money on Sunday (in part because he can, prior to Saturday night, repay the money on Sunday). After he has spent all the money, though, and so no longer can repay the money on Sunday, it is no longer true that Bruce ought to repay the money on Sunday. On the contrary, he is faced with a new (more restricted) set of options, and his obligations 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 direct and indirect control, on the other, see Subsection 2.2.2, including notes 51 and 52 to Ch.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 is Saturday night, T3 is Sunday, and where B, S, R, and A stand for "borrow," "spend," "repay," and "apologize," respectively).44 This picture captures the following (among other things): Bruce ought on Friday, all the way up to the spending on Saturday night, to repay the money on Sunday; Bruce ought (by default, as it were) at those same times not to apologize on Sunday for not repaying the money;45 but, after the spending, Bruce ought then to apologize on Sunday for not repaying the money; it is not true that Bruce ought, at these post-spending times, to repay the money on Sunday (because he cannot, at these times, do this); nor is it true that he ought, 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 impossibility — 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 why they wish to insist on this. I suspect that it is because they wish to insist that Bruce does wrong in not repaying the money and that they believe that my sort 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 respective worlds. Worlds 2a and 2b are thus of equal value, as are worlds 3a and 3b. That these worlds 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 something 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 thinking that my account implies otherwise. On the contrary, when an obligation is not satisfied, a wrong is done, whether or not the obligation is remote. If Bruce ought on Friday to repay the money on Sunday, and Bruce doesn't repay the money on Sunday, then Bruce does wrong. And the wrong isn't just that he spends the money on Saturday night. True, he ought on Friday, and up to Saturday night, not to spend the money on Saturday night, and so his spending is wrong — an immediate wrong, since the final time at which he ought not to spend coincides with (or is immediately prior to) the time of spending. But it is also true that Bruce ought on Friday, and up to Saturday night, to repay the money on Sunday, and so his not repaying is wrong, too - a remote wrong, since the final time at which he ought to repay does not coincide with the (pertinent) time of nonrepayment. Indeed, in this case it seems that the remote wrong is in a way more significant than the immediate wrong. Even though there is a sense in which the remote wrong of nonrepayment is committed by virtue of the immediate wrong of spending (the remote wrong would not have been committed, in the manner in which it was committed, had not the immediate wrong been committed), nonetheless the moral primacy rests with the remote wrong, in that (let us assume)48 the spending is wrong because and only because the nonrepayment that it necessitates is wrong (the spending would not have been wrong had not the nonrepayment been wrong). Although this moral primacy of a remote wrong over the immediate wrong by virtue of which it is committed helps to dramatize the fact that remote wrongs are genuine wrongs, it is in fact not crucial that the moral primacy rest with the remote wrong. Suppose, for instance, that Bruce's spending were to take the specific form of gambling. It might then be that the gambling is wrong not just because it necessitates the nonrepayment but for some other reason also; it might even be that the gambling is more seriously wrong than the nonrepayment, quite independently of the fact that it necessitates the nonrepayment. Still the nonrepayment would itself be wrong. Indeed, a remote wrong might be wrong only because something 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 rendering himself unable to be in New York, the site of the repayment scheduled for 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 to this chapter.

99

not being there constitutes a remote wrong, even though the wrongness of this 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 ought to 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 Bruce was obligated at Tto repay the money on Sunday. It will also be true to say on Monday that Bruce could not have repaid the money on Sunday, if this is taken to mean that there was a time T (not later than Sunday) such that Bruce could not at T repay the money on Sunday, but not if this is taken to mean that there was no time Tsuch that Bruce could at Trepay the money on 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 when remote wrongs are supposed to be done. If Bruce ought on Friday to repay the money on Sunday, then it is wrong on Friday for Bruce not to repay the 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) nonrepayment occurs on Sunday, after all.50 But the trouble is, it also seems problematic to say that Bruce does wrong on Sunday in this regard, since by that time it is (according to the proposed account) no longer wrong for him not to repay the money then. So, if he does a wrong — a remote wrong —just when is it that this wrong is committed? It seems to me useful to draw a distinction here between the time at which an agent goes wrong and the time at which the wrong occurs. "Do wrong" may be used in either case. The wrong that is done is the (pertinent) nonrepayment. This occurs on Sunday.51 Thus the wrong is, in this sense, done on Sunday, even though by then it is no longer wrong for Bruce to 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 W iff 49

I have no precise account to offer ofwhat I have called the "moral primacy" of one obligation or wrongdoing over another, nor of the corresponding concept of one obligation or wrongdoing being "parasitic" on another. I hope that what I say here is intuitively plausible. 50 Clearly, nonrepayment occurs at other times, too; the point is that repayment is scheduled 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 Bruce itself occurs on Sunday. (Here, talk of "doing wrong" can sometimes seem a bit forced, for one can conceive of cases where a wrong is "done" in this sense even though the agent is no longer alive. For example, perhaps S ought 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 true that, in some other sense, Bruce doesn't do wrong on Sunday with respect to the nonrepayment. For the point at which he goes wrong is Saturday night, when he spends all the money and thereby ensures the nonrepayment on Sunday. To capture this, and to accommodate straightforward immediate wrongdoings (where the times at which an agent goes wrong and at which the wrong occurs coincide), it might therefore seem that what we should in general say is this: an agent goes wrong with respect to failing to 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 rendered impossible. That is: (3.29) S goes wrong, at T i n 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 T a n d T' are identical), since no one can change what has already taken place. It would also yield the desired result in the spending case. If we look 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 spending the money on Saturday but with respect to not repaying the money on Sunday, since such repayment is rendered impossible at the moment of spending. But (3.29) is in fact not adequate to other cases of remote wrongdoing. Consider this case. O n Friday, Mandy tells Phil that she will be able to help him move furniture on Sunday, thereby making a sort of informal commitment to do so; on Saturday, however, Mandy promises Bill that she will babysit his child on Sunday, thereby making a formal commitment to do so. Note that, although Mandy can still on Saturday afternoon help Phil on Sunday, 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 Phil on Sunday, but also that, because of the formal commitment, Mandy ought on Saturday to babysit for Bill on Sunday. Suppose that what Mandy in fact does is to babysit on Sunday. Where does she go wrong with respect to not helping Phil move furniture? The situation may be pictured as in Figure 3.2 (where Tt is Friday, T2 is Saturday, T3 is Sunday, and where T, P, H, and B stand for "tell," "promise," "help," and "babysit," respectively).52 On this picture, Mandy ought on Friday to help Phil on Sunday. By taking Track 2b, Mandy does wrong in not helping Phil. This wrong occurs on Sunday. But Mandy does not go wrong on Sunday in not helping Phil, for Track 2b is preferable to Track 3. Rather, it seems clear, it is at noon on Saturday that Mandy goes wrong 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 time prior 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 immediate wrong is wrong only because the remote wrong is wrong. That is, the formal commitment to Bill would not be wrong were it not for the informal commitment to Phil. This is intended to dramatize the fact that remote wrongs are genuine wrongs. But, as before, there is in fact no necessity that remote wrongs have such primacy in order to be genuine wrongs. If, rather than promise Bill to babysit his child, Mandy were instead to kidnap the child, 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 Phil would still constitute a remote wrongdoing.) It is at noon on Saturday that Mandy goes wrong, for that is where she takes 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 that Mandy ought to help Phil. It therefore seems plausible to replace (3.29) by the following: (3.29') 5 goes wrong, at T i n 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 with respect to not performing an action, retrace his steps back down the track which he in fact took until you reach a juncture, a point of choice, where it is true that he ought then to perform the action - where, that is, there is a top-level track leading out from that point and on which the act is featured. In Figure 3.2, we need only go back along Track 2b to the point of choice at T2 to find a point where it is true that Mandy ought then to help Phil move furniture. Hence that is where Mandy goes wrong with respect to not helping Phil. Similarly, if we suppose that Bruce takes Track 2b in Figure 3.1, we need only retrace his steps to T2 to find a point where it is true that Bruce ought then to repay the money. Hence that is where Bruce goes wrong with respect to the nonrepayment. While I think that what has just been said is very close to the truth, it is still 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 respect to 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 that he now ought to visit Bert. However, at T2 he calls up Alf, making Alf so eager 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 have been possible had not the second occurred and the second would not have been possible had not the first occurred, and where Cu C2, C3t A, and B, stand for "make the first call," "make the second call," "make the third 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 not visiting 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 obligation is to visit Bert. But what (3.29') doesn't reveal is that Victor also goes wrong earlier in this respect. He goes wrong at T2. For from Tt to T2 he ought to visit 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 takes Track 4a. Where does he go wrong in not visiting Alf at T4? At T4, but also at Tt\ for at both these points there is a top-level track on which visiting Alf at T4 is featured and which Victor chooses not to take. To capture this possibility of multiple lapses, of multiple deviations from the strait and narrow with respect to one and the same action, I propose that we accept this revision of (3.29'): (3.29") S goes wrong, at T i n 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 do A at T\ there is an earlier time T** after T such that it is not the case that 5 ought, at T** in IV, to do A at T.

104

In general, then, one goes wrong whenever one takes a turn for the worse. Whenever one does this one of course commits an immediate wrong, but one may also be committing a remote wrong. Thus no remote wrong can be done without an immediate wrong being done (Bruce's wrongly not repaying the money can be traced to his wrongly spending; Mandy's wrongly not helping Phil can be traced to her wrongly promising to babysit for Bill), although it sometimes happens that an immediate wrong is wrong only because some remote wrong is wrong (as in the cases depicted 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 satisfactions of obligations. Recall the case of self-imposed necessity where Betty locks herself in a room on Saturday, thereby ensuring repayment on Sunday. After she throws away the key, she is no longer obligated to repay the money, for she cannot avoid doing so. Nonetheless, in repaying the money, she does what she ought. The situation may be pictured as in Figure 3.4 (where Ti is Friday, T2 is Saturday, T3 is Sunday, and where B, T, and R stand for "borrow," "throw away," and "repay," respectively). I propose that we treat this matter in a manner parallel to that in which remote 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 be distinguished from the time at which the wrong occurs, so too the time of maintaining an obligation may be distinguished from the time of its fulfillment. In Figure 3.4, the time at which Betty fulfills her obligation to repay the money is Sunday (T3), but she maintains this obligation only up until Saturday (T2). In general, then, I propose the following: 7",

Figure 3.4 105

(3.30) S's (a) (b) (c)

obligation to do A at T is fulfilled at T i n W iff Tis identical with T'; for some time T", S ought, at T " in W, to do A at T'; and 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 takes Track 3; that is, that she doesn't babysit the child but does help Phil move furniture. According to the formulae, Mandy thereby fulfills (on Sunday) and maintains (up until Saturday) an obligation to help Phil move furniture. Or again, with respect to Figure 3.3: suppose now that Victor takes Track 4b. In this case he fulfills at T4 his obligation to visit Bert, an obligation 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 could have done worse - she could have taken Track 5 - and that she would then have satisfied no obligation at all. But we should not be misled by this observation. It would be a mistake to infer that one satisfies an obligation as long as one avoids the worst track open to one or that one satisfies no obligation if 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 numbering reflects the ranking of the options, or tracks, open to an agent, S. In such a case, if S takes Track 2, he satisfies no obligation, even though he could have done worse. (This claim will be qualified in Chapter 4.) On the other hand, if S takes Track 4, he does satisfy an obligation, namely, the obligation to do A. (This claim will be discussed further in Chapter 6.) The moral is that we should not necessarily be impressed if someone does something that 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 of doing 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, this need not constitute praise for her action; for it does not tell the whole story.

1: A& B&... 2: ~A& -B&... 3: ~A& B&... 4: A& ~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. This is clearly true of immediate obligations, but some may balk at accepting this claim when it is remote obligations that are at issue. Perhaps it will be agreed that Bruce does wrong in not repaying the money and that Mandy does wrong in not helping Phil (or in not babysitting the child, depending on which version of that story we're working with), but aren't there other cases where it is accurate to say that a remote obligation is not satisfied and yet no wrong is done? What if the reason for Bruce's not repaying the money had not been that he spent it all but that he had been mugged (or struck by lightning) on his way to repaying it? What if the reason for Mandy's not helping Phil had not been that she was fulfilling a wrongly made promise to babysit for Bill but that Phil had released her on Saturday from her informal commitment to help him (or that Phil had died from a heart attack on Saturday)? Isn't the pertinent obligation canceled in such cases, and doesn't this show that not all failures to satisfy an obligation constitute wrongdoing? I grant the intuitive plausibility of this objection. It is interesting to note that talk of such cancellation of obligation, if taken literally, would appear to presuppose the possibility of remote obligation; for how in general could an obligation to do A at T2 be canceled at T1 unless it were already in existence at Tj? Nonetheless, I reject the possibility of such cancellation. Although it may be, for instance, that Bruce would not have been obligated to repay the money if he had been mugged, or that Mandy would not have been obligated to help Phil if Phil had released her from her commitment, this doesn't imply that their obligations would thereby have been canceled. For consider Bruce. Either he could on Friday have avoided the mugging (or lightning) on Saturday or he couldn't. (It is extremely likely that he could; he could have walked down Elm Street rather than Oak Street, and then he wouldn't have been mugged or struck by lightning.) Similarly, either Mandy could on Friday have avoided Phil's releasing her from her commitment (for example, by pleading with him not to be so generous, or condescending, or whatever) or she couldn't; or again, either Mandy could have avoided (prevented) Phil's having a heart attack or she couldn't (and perhaps here it is likely that she couldn't). In every case what we should say is this. If the agent could not have avoided the event that allegedly canceled the original obligation, then he (or she) did not have the original obligation after all, although it may have been quite reasonable (for him 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 accessible to him on Friday in which he repaid the money on Sunday, and so he was not obligated on Friday to repay the money on Sunday, although it may have been quite reasonable to believe that he was. So too for the other cases.) On the other hand, if the agent could have avoided the "canceling" event, then, if he was already obligated to perform the pertinent action, he may well have been obligated to avoid the "canceling" event itself. (If Bruce could on Friday have avoided the mugging, and if Bruce was obligated on Friday to repay the money on Sunday, then Bruce was obligated on Friday to avoid the mugging on Saturday. If Mandy could on Friday have avoided Phil's releasing her from her commitment,55 and if Mandy was obligated on Friday to help Phil on Sunday, then she may well have been obligated on Friday to avoid Phil's releasing her from her commitment on Saturday - although not necessarily: if the deontically best that Mandy could on Friday do on Sunday was to help Phil, whether or not Phil released her from her commitment, then Mandy was not obligated on Friday to avoid this release.) Thus every failure to satisfy an obligation does indeed constitute wrongdoing.56 But, it may be objected, even if Bruce could literally have avoided the mugging, he surely could not have known he would be mugged, or even had reason to believe that he would be. Or, even if he did have reason to believe this, surely it is not he but his muggers who did wrong. To say otherwise is to blame the victim rather than the perpetrators of a wrongdoing — a reprehensible practice. The proper response is this. I am not blaming Bruce in this version of the case; 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 events when it comes to their accessibility. 56 Cf. Thomason (1981), pp. 178-9. Contrast, perhaps, McConnell, who says that debts can sometimes be annulled, and gives this illustration (McConnell 1993a, p. 80): "Ifyou have 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, while there 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 had an overall moral obligation to avoid her stealing it from you. Did you have an obligation to reimburse her before she stole it? If so, you had, ipso facto, an obligation to avoid her reimbursing herself in this way; if not, then, unless for some other (and presumably rather odd) reason you had an obligation to avoid the theft anyway, it was permissible for you so to act that she reimbursed herself in this way (even if not permissible for her so to act in this way), and so permissible for you not to see to her reimbursement in more customary 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 the mugging. I am saying that, if he was obligated to repay the money, then he did wrong in not avoiding the mugging. Recall the distinction, discussed in Section 1.4, between theories according to which what is obligatory is what is best (in some substantive sense) and theories according to which what is obligatory is what is probably best. It was best (let us assume) for Bruce to walk down Elm Street rather than Oak Street; hence, according to the first sort of theory, that's what he ought to have done. But it might well be that it was (at the time) probably better for Bruce to walk down Oak Street, in which case, according to the second sort of theory, he did what he ought, even though that involved being mugged and not repaying the money. But, if so, then Bruce was not obligated to repay the money after all (although it might be true that he was obligated to repay the money if he didn't walk down Oak Street - a matter of conditional obligation, to be discussed in Chapter 4).58 Another, similar case might help to drive home this point. W. D. Ross discusses returning a borrowed book through the mail. At one point59 he declares that the obligation in question is satisfied only if the book in fact reaches the hands of the person from whom it was borrowed, although the agent may well not be to blame if the obligation is not satisfied (perhaps blame will rest with some employee of the postal service). Later Ross declares that the obligation in question is satisfied if the agent "exerts himself" to return the book, even if the book gets lost in the mail. I prefer the first 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 destination, then he is obligated not so to act that the book gets lost in the mail; if the agent is not obligated not to use the mail in this way, then he is not obligated so to act that the book arrives at its final destination. I am not insisting (although I believe) that the agent is obligated so to act (given that he can so act) that the book arrives at its final destination, but only that, if he 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 ought to have done. But this is either to take an act-based, rather than world-based, approach to obligation, or to take a world-based approach where the worlds in question are only probably rather than actually accessible. Both approaches have already been rejected; see Subsections 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 to avoid getting mugged. This is not to blame Bruce if he gets mugged, just as 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 latter agents may well have done wrong themselves and may well be to blame for what they have done. But, if either of these judgments about the mugger and the postal worker is correct, their being so is wholly independent of the judgments to be made about Bruce and the book-borrower. In denying that obligation can be canceled, I have been arguing that agents can sometimes be obligated in cases where many may think that they are not. But uneasiness with my treatment of remote obligation and wrongdoing may come, as it were, from the opposite direction. There are cases where I am committed to saying that the agent is not obligated to perform a certain action but where it might seem that we should say that he is obligated to do so. For instance, I have said that Bruce's spending on Saturday night renders it impossible for Bruce to repay the money on Sunday, so that it is not true that he ought, from Saturday night on, to repay the money on Sunday. But what if Bruce were luckily to stumble across some extra money on Sunday morning? What if he were to win the lottery? Surely then he ought, as of that moment, to repay the borrowed money after all. Here the correct response is simply this. Either Bruce can so act on Saturday night that, despite spending the money then, he will stumble across some extra money on Sunday morning, or he cannot. (See Subsection 2.2.3 above.) If he can, then his spending on Saturday night does not 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, then this is a case of self-imposed impossibility, just as originally described and intended.61 (Notice that it might be true in some sense that Bruce might win the lottery even though he cannot, where the lottery is of the pre-imprinted 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 misrepresents herself - as a competent lifeguard, say.) My account implies that Pru was never obligated to keep her promise, and hence that Pru committed no wrong with respect to not keeping her promise. Surely this is counterintuitive. 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 initial act of misrepresenting herself (her intentions, her abilities). This is avoidable and ought to be avoided. Sinnott-Armstrong agrees that Pru ought not to make the promise but argues that it is also true that, having made it, she ought to keep it, on the grounds that, if it were not true that Pru ought to keep the promise, then there would be no reason why she ought not to make it in the first place.62 Not so. There is very good reason why she ought not to make the promise, even though it is not the case that she ought to keep it, and that is that she cannot keep it. She is misrepresenting herself; she is misleading others; she is creating false beliefs and certainto-be-frustrated expectations. Under the circumstances, this is (we may assume) wrong. 3.2.5 The supersession of obligation That obligation cannot be canceled does not imply that it cannot be superseded. Obviously it can be. Every obligation is eventually superseded, if only because, if 5 ought at T to do A at T', T' will eventually arrive. The interesting cases of supersession, however, are those involving remote obligation, 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 overriding is dynamic and to be distinguished from the sort of static overriding involved in one conditional obligation's overriding another and in one prima facie obligation's overriding another (issues to be discussed in the next two chapters). For example, Mandy's obligation to help Phil move furniture 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 T i n 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.

Ill

T4

Figure 3.5 Notice four things. First, if S's obligation to do A at T' is overridden, it nonetheless survives, in the sense that, if he does A at T", he at least satisfies that obligation. (As pointed out in the last section, however, this may be no cause for celebration.) Notice also that it is possible that an obligation that has been overridden be restored, in the sense that the agent may once again become such that he ought overall to satisfy it. For example, ifMandy were to murder Bill's child on Saturday night, it might be that she ought then once again to help Phil move 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 before in Figure 3.2, with Mstanding for "murder"). It is in principle possible for an obligation to be overridden and restored an indefinite number of times. If this happens, though, each moment of overriding and restoration marks a fresh turn for the worse by the agent. Notice, thirdly, that, although in many cases of the overriding of an obligation to do A by an obligation to do B, the agent 5 cannot do both A and B, this is not necessarily the case and is not implied by (3.32). Although it is true of Mandy (in that she cannot both help Phil move furniture and babysit 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 case that 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, since this would greatly disturb Lucy (bringing back certain horrible childhood memories). Suppose that, at T2, Gertrude spills weed killer on the lilies that she had intended to deliver to Lucy at T3. This changes things. N o w she ought to send Lucy daisies, for the lilies have shriveled. The obligation to send lilies has been overridden by the obligation to send daisies. Gertrude can, of course, still send both lilies and daisies, but this remains something that she clearly ought not to do. 64 Notice, finally, that not all supersession of a remote obligation consists in its being overridden. When Bruce's obligation to repay the money on Sunday is superseded by his spending it on Saturday, the obligation is not overridden; for it does not survive but is rather extinguished, in that he cannot any longer satisfy it. So too with Betty's obligation to repay the money; it is superseded by her throwing away the key on Saturday, but it is extinguished rather than overridden. W e may in general say: (3.33) S's obligation to do A at T' is extinguished at T i n 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 fulfillment or nonfulfillment of an immediate obligation trivially counts as its extinction. But extinction can occur remotely, too, as in Bruce's case (where his spending renders him unable to repay the money) and in Betty's case (where her throwing away the key renders her unable to refrain from repaying the money). 64

It might be objected that Gertrude cannot any longer send lilies that are in a suitable condition, 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 is confused. 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 obligation to send lilies (see (2.43'"))' a n d so my remarks stand.

113

Conditional obligation

4.1 VARIETIES OF C O N D I T I O N S ON OBLIGATION So far the focus of inquiry has been the conceptual structure of statements of the form "5 ought [at Tin W] to do A [at T ' ] " and various related statements. These are intended to be understood as statements of unconditional overall moral obligation. But statements of conditional overall moral obligation 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 conceptual structure of such statements. The most complete account of conditional obligation that has yet been given is that provided by Fred Feldman.1 I shall not seek to reproduce the 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..." are ambiguous. He distinguishes four principal interpretations. The first interpretation treats the statement as one involving material implication. Using the symbolism introduced earlier, we may represent such statements as having this form: (4.1) p D O(A). Feldman remarks on a number of features of such statements. Here I will draw attention to just one, that which he calls "factual detachment."2 Such detachment consists in the obligation's obtaining unconditionally when the condition in question is satisfied. Statements of the form of (4.1) allow for such 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 ordinary English 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 by Alan 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 equivalent 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 that we ought, because] we cannot.) Nonetheless, it may be that such statements are really just subjunctive conditionals in disguise. The second interpretation treats statements of the form "S ought to do A, 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, many clear uses of statements of this sort. For instance, each of the following involves 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 following statement to be so understood: (4.6) if it would maximize utility, then he ought to do it. 3 4

Feldman (1986), pp. 78-9. Feldman (1986), p. 79.

115

Similarly, a Kantian will typically intend for the following statement to be so understood: (4.7) if it would violate the Categorical Imperative, then he ought not to doit. 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 obligated to do A " Clearly, factual detachment once again applies. Equally clearly, 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 obligation 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. Suppose that there are three candidates for election, Tom, Dick, and Harry, and that they are to be ranked in that order.6 Let us therefore suppose that your options are to be ranked as in Chart 4.1 (where T, D, and H stand for "vote for Tom," "vote for Dick," and "vote for Harry," respectively). In that case, what (4.10) amounts to is this. If Tom were a crook, then your options would not be ranked as in Chart 4.1 but in some other way, such as in Chart 4.2. In other words, your options would be reordered, deontically.

1: T & ~D 2: ~T & D 3: ~T & ~D 4: ~T & ~D

& ~H... & ~H... & H... & ~H...

Chart 4.1

1: ~T & 2:-7(5 3:~T& 4: T &

D & ~H... ~D & H... ~D & ~H... ~D & ~H...

Chart 4.2 5 6

Feldman (1986), p. 85. The example is borrowedfromJackson 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 for Tom, 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, suppose that it turns out that Tom is a crook; then (4.10) implies that you ought to vote for Dick. But suppose that you won't in fact vote for Tom; it would not follow from (4.11) that you ought to vote for Dick. You ought to vote for Tom, not Dick. To say that you ought to vote for Dick just because you won't vote for Tom is to make things either impossibly difficult for you or ridiculously easy. It makes things impossibly difficult if the claim that you ought to vote for Tom is retained; for, on the assumption that you cannot both vote for Tom and vote for Dick, you would then be faced with a moral dilemma (and, whatever one thinks about the theoretical possibility of such dilemmas, surely no one thinks that they are so easily generated - more on this in Chapter 7). It makes things ridiculously easy if the claim that you ought to vote for Tom is abandoned; for you cannot change the fact that you ought to vote for Tom simply by not voting for him (compare the objection concerning self-imposed impossibility discussed in Chapter 3). Insofar as (4.11) does not warrant factual detachment, it is to be distinguished 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 shall call conditional obligation proper, and it is to the analysis of this that I now turn. 4.2 THE ANALYSIS There is no doubt that statements of conditional obligation proper play a large role in pronouncements about morality. Examples abound: (4.12) if Pru is going to fail to keep her promise, then she ought not to make it in the first place (although she ought both to make it and to keep it); (4.13) if Charlie changes lanes, then he ought to accelerate (although he ought 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 to administer it on Tuesday too (although she ought in fact to administer medicine A instead on both Monday and Tuesday);8 (4.15) if Sally sins, then she ought to repent (although she ought not to sin in the first place); (4.16) if Alan fails to attend the 10:00 A.M. meeting on the first floor, then he 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 first floor and not the one on the second); and so on. In every case, we have a statement that (if used typically) is to be understood as a statement of conditional obligation (proper).9 I propose (following the lead of others)10 that conditional obligation be analyzed in a manner that parallels the analysis already given for unconditional obligation. The main gist of the latter was this: S ought to do A if and only if (a) 5 can do A, (b) 5 can refrain from doing A, and (c) any accessible world in which S does not do A is deontically inferior to some accessible 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 here something to which Feldman pays scant attention: that statements of conditional obligation proper cannot be satisfactorily analyzed by means of a material conditional within the scope of the obligation operator. For example, (4.11) cannot be satisfactorily analyzed 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 following: 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 vote for 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 S does 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 H 7 ' 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 deontic 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 deontically inferior 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 true iff (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 world Wr 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 obligation tells us is that, given the restriction to ~ T-worlds, Option 2 is deontically best. 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 of course 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 validated. Thus we cannot validly infer an obligation to vote for Dick simply from the 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): the best world accessible to Pru in which she fails to (keep a promise to) do A is one in which she doesn't promise to do A; nonetheless, the best world accessible 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 he changes lanes is one in which he accelerates; nonetheless, the best world accessible to Charlie is one in which he neither changes lanes nor accelerates. Concerning (4.14): the best world accessible to Audrey in which she administers medicine B on Monday is one in which she administers it on Tuesday also, although the best world accessible to her is one in which she administers medicine A on both Monday and Tuesday. Concerning (4.15): the best world accessible to Sally in which she sins is one in which she repents, although the best world accessible to her is one in which she doesn't sin. And concerning (4.16): the best world accessible to Alan in which he fails to attend the meeting on the first floor is one in which he attends the meeting on the second floor, although the best world accessible to him is one in which he attends the meeting on the first floor and not the meeting on the second floor. Siblings of (XIII) may be generated in a manner analogous to that in which siblings of (I) were generated. For example, just as (XIII) is analogous 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 deontic value for S at T of W'.] Mercifully, I shall refrain from giving explicit statements of the analogues to (III)—(XII). Nonetheless, one further sibling of (XIII) deserves explicit recognition, for it has no direct analogue among (I)-(XII). Consider this statement: (4.17) you ought to take that item with you only if you pay the shopkeeper. 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 the shopkeeper. 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 is true iff S ought, at T in W, not to do A at T", on the condition thatp is not true (that is, on the condition that ~p is true).

Statements of biconditional obligation are of course possible. Consider this example: (4.18)

you ought to go to the party if and only if you accept the invitation.

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 neither ought 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, conditional obligation also fails to validate what Feldman calls "augmentation."12 That is, 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 to vote for Harry if you don't vote for Tom and you don't vote for Dick. Secondly, however, conditional obligation does validate a restricted form 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 arepworlds, 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 worlds are y4-worlds. Suppose, for example, that Charlie cannot avoid changing lanes (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 lanes nor 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 the deontically best worlds accessible to 5 arep-worlds; and so, again, if the best accessible p-worlds are y4-worlds, then the best accessible worlds are Aworlds. Suppose, for example, that you ought to accept the invitation; then from (4.18a) we may infer that you ought to go to the party. On the other hand, if you ought not to accept the invitation, then from (4.18b) we may infer that you ought not to go to the party. 4.3 CERTAIN DEONTIC PARADOXES Various so-called deontic paradoxes have already been discussed. In Subsection 2.3.2 there was discussion of those paradoxes known as Ross's Paradox, the Good Samaritan Paradox, and the Paradox of the Knower; in Chapter 3 a paradox concerning self-imposed impossibility (namely, that such impossibility can rid one of certain obligations) was addressed.15 All the other standard paradoxes have to do with conditional obligation, and so 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 not received 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 equivalent to its being the case that (4.17') you ought not to take that item with you ifyou don't pay the shopkeeper. Such contraposition may seem unobjectionable at first, and yet it seems on inspection that (4.24) can be true while (4.17') is false. Consider a case where the item in question is food which your child desperately needs and for which you can afford to pay. We may assume that (4.24) is true. Might it not be, though, that, because of your child's need, you ought to take the food even if you don't pay for it? If so, (4.17r) is false (as an expression of overall obligation; it may be true as an expression of primafacie obligation). The account of conditional obligation that I have given allows for the possibility that (4.24) be true but (4.17') false. Suppose that your options are 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) (4.26) (4.24) (4.27)

you ought to take that item with you; you ought to pay the shopkeeper; you ought to pay the shopkeeper if you take that item with you; 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 the shopkeeper, 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 shopkeeper. 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), where D 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 the shopkeeper; (4.28') you ought not to take that item with you only if you do pay the shopkeeper. But I confess that (4.27') and (4.28') would be a pretty odd way of making one's point. Still, (4.27) and (4.28) are somewhat odd too, and they're odd because (4.25), a statement of straightforwardly unconditional obligation, is true. When such statements are true, then statements of conditional obligation, such as (4.27) and (4.28) and their equivalents, are generally otiose, 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 the following analysis: (4.29) S ought, at T in W, to do A at T", only on the condition thatp is true iff S ought, at T in W, so to act that p is true, on the condition 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(A only 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 the matter of factual detachment. Hector-Neri Castaneda has introduced what he calls the Paradox of the Second Best Plan.18 In this context, consider Audrey again, of whom it was said: (4.14) if Audrey administers medicine B on Monday, then she ought to administer it on Tuesday too (although she ought in fact to administer 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 factual detachment is not warranted. Of course, once it has happened that Audrey has administered medicine B on Monday, it may well then be true that she ought to administer it on Tuesday also. This is just one more instance of the sort of shift in obligations that was the topic of discussion in the last chapter. But prior to Monday it will not be true to say that Audrey ought then to administer medicine B on Tuesday, even if it is true then to say that she will administer it on Monday. In the case just discussed, there is a sense in which (4.31), though not true on Sunday, becomes true on Monday (once medicine B has been administered). ° But cases of this sort are possible where there is no time for such 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 he ought neither to change lanes nor to accelerate). Suppose that the changing of lanes and the acceleration, were either or both to occur, would occur at the same moment in time. In that case, even if it is 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 accelerating. Charlie's options may be ranked as in Chart 4.4 (where C stands for "change" and A for "accelerate"). At no time is Charlie faced with a set of options where A is featured in the top-ranked course of action. 1: ~C & ~A 2: C & A 3: C & ~A 4: ~C & A Chart 4.4 20 Actually, this seems to me strictly false. What I would prefer to say is that (4.31) is ambiguous (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 on another reading (where the time in question is Monday, after medicine B has been administered 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 in Subsection 2.3.2. This paradox may now be presented more perspicuously, in light of the analysis of conditional obligation provided in this chapter. Recall Greg, the security guard. We may now understand him to have the following conditional obligation: to provide a truthful report of his performing certain (wrongful) actions, if he does perform them. Consider one such action, A. Thus: (4.34) Greg ought to provide a truthful report of his doing A, if he does A. 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 logic of "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 conditional obligation does not warrant factual detachment, we cannot infer (4.36) from (4.34) conjoined with (4.35). For although it may be that the best v4-worlds accessible to Greg are worlds in which he provides a truthful report of his doing A, still the best worlds accessible to him are worlds in which he does not do A (and so also does not provide a truthful report of his doing so). The so-called Paradox of Reparation may be handled in exactly analogous 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 least when 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 that is because the move from (4.15) and (4.38) to (4.39) is invalid. Such factual detachment is unwarranted. The so-called Paradox of Gentle Murder may be treated in the same manner.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 we don'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) is invalid. Finally, a much-discussed paradox, concerning so-called contrary-toduty imperatives and first introduced by Roderick Chisholm, is subject to similar 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 entail that one believe that one has sinned. 23 This paradox was introduced in Forrester (1984). The treatment that I give is adequate also 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 is no 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 of deontic detachment), the inference of the latter is not (being an instance of factual detachment). Castaneda has expressed dissatisfaction with this treatment of Chisholm'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 joint assertion of (4.48) and (4.49) might seem odd in the absence of the explicit insertion of (4.45), it is nonetheless true.26 Castaneda has expressed dissatisfaction with this reply, in that he contends that what makes the joint assertion of (4.48) and (4.49) paradoxical is this. Suppose that Aida is intent on not going to the aid of her neighbors but is otherwise keen to minimize her wrongdoing. Thinking along Feldman-type lines (lines that I have followed in this chapter), she therefore decides to satisfy the obligation alleged in (4.49) and tells her neighbors that she's coming. But this, far from minimizing her wrongdoing, compounds it.27 Feldman has responded to this criticism, too, in a manner that seems to me appropriate,28 but he overlooks what I take to be the most effective response to it. This response makes reference to subsidiary obligation, a matter to which I now turn.

4.4 SUBSIDIARY O B L I G A T I O N 4.4.1 Factual detachment You may be uneasy with the solutions just proposed to the paradoxes considered. Should we dismiss factual detachment so summarily in the case of conditional 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 26 27 28

Castaneda (1989a), p. 9ff. Feldman (1989b), p. 35. Castaneda (1989b), p. 167. 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 the corresponding 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 not to tell them that she's coming? It is of course true that detachment of an unconditional obligation, when the condition of the corresponding conditional obligation is satisfied, is sometimes warranted. We have already seen that both necessity detachment and deontic detachment are valid. Other such restricted detachment rules may seem plausible. As a variation on necessity detachment, it may seem that, in cases where the action that is conditionally obligatory is "scheduled" to occur after the condition in question is "scheduled" for satisfaction, if and when the condition is satisfied, the action then becomes obligatory; that is, it may seem that, if 5 ought at T1 to do A at T3, on the condition that some event E occurs at T2, and if E occurs at T2, then S ought at T2 to do A at T3. For example, if Audrey ought on Sunday to administer medicine B on Tuesday, on the condition that she administers it on Monday, and if she does administer it on Monday, then she ought on Monday (once medicine B has been administered) to administer it on Tuesday. But in fact this does not hold without exception. I said in Section 4.3 that it "may well" be that Audrey ought on Monday to administer medicine B on Tuesday, but I didn't say that this was certain. For suppose that, while administering the medicine on Monday, Audrey also gives her patient a lethal injection; presumably it would not then be true that she ought on Monday to administer the medicine again on Tuesday. (This is another illustration of the failure of augmentation for conditional obligation.) The situation may be represented as in Chart 4.5 (where AM, ATi BM, BT, and LM stand for "administer medicine A on Monday," "administer medicine A on Tuesday," "administer medicine B on Monday," "admin129

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 she takes Track 2, it will indeed be true that she ought on Monday (once she has administered medicine B) to administer medicine B on Tuesday. But if she takes Track 4a, it will not be true that she ought on Monday to administer medicine B on Tuesday. Mark Vorobej has offered the following restricted detachment rule (a variant on the rule of deontic detachment): when the condition in question is satisfied and its nonsatisfaction is not obligatory (rather than, as before, its satisfaction being obligatory), then the obligation in question may 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? Quite possibly. But the fact remains that the sort of unrestricted detachment that has here been called factual detachment is not valid for conditional obligation (proper), given my analysis of such obligation. And, I have argued, this is 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 factual detachment allows for a straightforward solution to many of the deontic paradoxes. Nonetheless, there is still the undeniable inclination to say that, in some sense, you ought to vote for Dick because you won't vote for Tom; that Charlie ought to accelerate because he will change lanes; that Sally ought to repent because she will sin; that Aida ought not to tell her neighbors that she's coming because she won't go to their aid; that Alan ought to attend the meeting on the second floor because he won't attend the meeting on the first floor; and so on. I think this inclination is correct. Although I have just rejected factual detachment for conditional obligation, I want now to suggest that there is a sense in which such detachment is acceptable after all. We can have our cake and eat it too, as long as we recognize that the sort of obligation 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 particular, Michael McKinsey has attempted to account for it.30 He has talked of "levels of obligation" and said that obligations may be "primary," "secondary," "tertiary," or indeed "«-ary" (where n > 1). McKinsey is concerned with just the sort of cases that we are concerned with. The meeting case 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 do this, he ought2 to attend the 10:00 A.M. meeting on the second floor; and so on. Here "ought/' expresses a primary obligation and "ought2" a secondary obligation. Concerning such levels of obligation, McKinsey says: By saying that an obligation is secondary (or tertiary, or...), I do not mean that it is any less of an obligation than a primary one. In my view it is just as incumbent upon a person to fulfill his secondary obligations, as it is incumbent upon him 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 for not doing his best.31 This is both puzzling and promising. It is puzzling, in that it is not clear what 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. Yet McKinsey's proposal is nevertheless promising, for he has latched onto an important fact, and that is that, if one fails to do what one ought to do but does what is "second best," one does less wrong32 than if one also fails to do what is "second best"; and one who fails to do what is "third best" does still more 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 fails to attend the 10:00 A.M. meeting on the second floor (and this despite the fact 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 no wrong — he violates no obligation — in not attending the meeting on the second floor. We have noted that, if Alan fails to attend both meetings, he does a double 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 conditional obligation to attend the meeting on the second floor if he fails to attend the meeting on the first floor. I propose that we say that Alan violates a primary obligation to attend the meeting on the first floor and a secondary obligation to attend the meeting on the second floor. (I) captures only the concept of primary obligation; it is only with this concept that we have hitherto 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 following: all worlds accessible to the agent are comparable in deontic value and at least one of these worlds is such that no other of them is more valuable than it. Given this assumption, which I shall call the Simple Assumption, we may say that some world (or worlds) accessible to the agent at a certain time has (or have) deontic value-rank 1 (relative to the agent at that time) and all other accessible worlds have a corresponding subordinate deontic value-rank n (where n is some positive integer greater than 1). It is on the basis of this assumption that the various diagrams and charts so far provided have been drawn up.34 Working with this assumption, we may modify the accounts of unconditional 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 time for 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 that has been provided is an account of unconditional primary obligation by modifying (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 W e may then assign levels to conditional obligation by modifying (XIII) as follows: (XIII') S oughtM, at T in W, to do A at T', on the condition that p is true iff [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) detachment of subsidiary obligation by providing this account of such obligation (where n is greater than 1): (XVI)

S ought n , at T in W, to do A at T' iff there is a proposition^ such that (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 because you won't vote for Tom; that Alan ought to attend the meeting on the second 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 secondary obligation; but they remain false if they are intended as statements of primary obligation. This is how we can have our cake and eat it too. You ought! to vote for Tom, and so it is not the case that you ought! to vote for Dick; otherwise things would be impossibly difficult or ridiculously easy. Nonetheless, inasmuch as you won't vote for Tom, you ought 2 to vote for Dick. Ifyou don't satisfy this subsidiary obligation, you commit a subsidiary wrong. In this way, if you vote for neither Tom nor Dick, you do a double wrong; you compound the primary wrongdoing involved in not voting for Tom. 35

Given the Simple Assumption, this addition is strictly redundant, for every accessible world in which S does not do A at T" must be less than maximally valuable. But I supply the addition for the sake of perspicuity.

133

It is clear that wrongdoing may be compounded still further. If you vote for neither Tom nor Dick nor Harry, you commit a primary and a secondary and a tertiary wrong. In principle, one may pile subsidiary wrong upon subsidiary wrong ad indejinitum. Perhaps Alan oughtt to attend the meeting on the first floor, ought2 to attend the meeting on the second floor if he fails to attend the meeting on the first floor, ought3 to attend the meeting on the third floor if he fails to attend either meeting on the first and second floors, and so on for as many floors and meetings as you like. It is clear that Alan has the opportunity to be a very naughty boy indeed. But note that (XVI) also implies, quite correctly, that there is no naughtiness, no wrongdoing, if Alan fails to attend any of the higher-floor meetings but does attend the first-floor meeting. At this point we may make what I take to be the most effective response to Castaneda's concern about the Chisholm case. He said that it seems to be 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, ifAida were to try to minimize her wrongdoing (short of making (4.48) false) by satisfying the obligation at issue in (4.49), she would, paradoxically, compound it instead. It is clear that I can now agree that Aida would be compounding her wrongdoing, for it is true that (4.47) if Aida doesn't go to the aid of her neighbors, then she ought not to 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 in Subsection 3.2.3: we should not necessarily be impressed if someone does something that he ought (that is, oughtj) to do; for some ways of failing to do what one ought to do are superior to some ways of doing what one ought to do. All of this has been predicated on the Simple Assumption, though, and it may be that, on some substantive accounts of obligation, this assumption is false. Such falsity will be due to one or other of two factors: either not all accessible worlds are comparable in deontic value, or no accessible world 134

has a maximal deontic value.36 If either possibility were to arise, then fully specific judgments of subsidiary obligation would appear to be impossible. Although the accounts of unconditional and conditional obligation given in (I) and (XIII) would still stand, either possibility would render a fully specific deontic ranking of all accessible worlds unachievable, so that (XIII') and (XVI) would be inapplicable as originally intended. Nonetheless, even if 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-andsuch, a certain agent has compounded his wrongdoing. One way to make such a judgment would be arbitrarily to assign the rank of 1 to some accessible world, then rank other deontically inferior worlds accordingly, and then apply (XIII') and (XVI), thereby identifying certain wrongdoings as relatively secondary, or tertiary, and so forth. This would still give some measure of the extent of wrongdoing committed by an agent, although the judgment would be merely a relative one, either because some deontically incomparable worlds will have been ignored or because, given that there is no accessible world with a maximal deontic value, the agent will, in an absolute 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). N o w that levels of overall obligation have been distinguished, a different, though related, type of overriding - a static type - may be explicated. We may say that one conditional overall obligation overrides another if and only if the former 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 condition that 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, of course, also be given: (4.56) S's (overall) obligation^, at T in W, to do A at T" is overridden by S'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 subsidiary obligation to do B, then the latter, overriding obligation is not satisfied. For the former, overridden, lower-level obligation "kicks in," as explained above, only if the higher-level obligation is violated. It may be that the concept of overriding just accounted for is a rather technical concept, in that talk of "overriding" is more commonly applied to the sort of dynamic overriding accounted for in (3.32) and to the sort of static overriding of prima facie obligation to be accounted for in Subsection 5.3.4. Nonetheless, I think that it may on occasion happen that the terminology is applied to what has been accounted for here, and so it is desirable to have such an account. Some might say, for instance, that, although there is admittedly an obligation (of a sort) to vote for Harry, it is overridden by an obligation to vote for Dick, which is itself overridden by an obligation to vote for Tom. This may be understood in the manner just indicated. It is worth noting that, as in the case of dynamic overriding accounted for in (3.32), the sort of static overriding of an obligation to do A by an obligation to do B accounted for in (4.55) and (4.56) doesn't require that the agent be unable to do both A and B. The voting case is most naturally understood as a case in which you cannot vote for more than one candidate, but it needn't be understood in this way. It may be that you can vote for more than one, but any such vote would be deontically inferior to a single vote or no vote at all. In that case, although the obligation to vote for Harry is overridden by the obligation to vote for Dick, it is nonetheless possible 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 the one provided by McKinsey. It too is explicitly predicated on the Simple Assumption. 37 His account is at odds with mine both with respect to what counts 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 of obligation.

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 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{ and tj, 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's account is more liberal than mine. The case may be partially represented as in 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 assumed that just one world has rank 1 and just one world has rank 2. We may of course 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. My account of subsidiary obligation implies that Alan oughtj to attend the meeting on the first floor and, given that he does not do this, ought2 to attend the meeting on the second floor. My account also implies that Alan ought j not to attend the meeting on the second floor. But it does not imply that Alan ought2 not to attend the meeting on the first floor, given that he 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.) But McKinsey's account does imply this. Such a claim strikes me as incorrect, indeed bizarre. McKinsey's account is also more conservative than mine. Consider a (rather strange) variation on the paying-the-shopkeeper case that may be 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 saying all 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 shopkeeper; and (4.58)

you ought2 to pay the shopkeeper if you don't take that item with you.

In short, it is best for you to take and pay, but doing either one is preferable to doing neither, although (rather strangely) which one you do, if you don'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), one or other of the following would be true: (4.59) (4.60)

you ought2 to take that item with you; you ought2 to pay the shopkeeper.

In this way, if you take Track 3, you will quite properly be said to have done wrong on two levels. Yet McKinsey's account allows us to say neither of (4.59) and (4.60), since in each case clause (1) of his account, (4.57), is violated. It is noteworthy that McKinsey seems not to be especially concerned with conditional obligation and detachment when giving his account of levels of obligation.39 Perhaps it is this lack of concern that has, at least as I see it, led him astray. For the two are intimately related, as I have sought to show. 4.4.5 Level of wrongdoing and seriousness of wrongdoing It might be thought that my proposal is unnecessary. Why trouble ourselves with levels of obligation when all we need do is note that, the further one departs from Track 1, the more serious the wrong that one does? Consider again the voting case, represented in Chart 4.1. Is there really any need to insist that you commit wrong on two levels if you take Track 3? Two wrongs, yes: ~T, and H. But why talk of two levels? Doesn't this multiply wrongdoing beyond necessity, in that, in addition to the primary wrongs committed 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 a matter that occupies his attention in the rest of his paper.

138

commit a secondary wrong by performing ~D? Can't we express all that we need to express by noting the (primary) wrongs involved and also noting that Track 3 is a greater departure from Track 1 than Track 2 would have been? I think not. While I have no wish to deny that taking Track 3 is more seriously wrong than taking Track 2, we must still contend with the inclination to infer that you really ought to vote for Dick, given that you won't vote for Tom. My proposal accounts for this; the simpler counterproposal does not. As to seriousness of wrongdoing, it should be observed that, on those substantive accounts according to which the deontic ranking of options reflects some value that is specifiable independently of its relation to matters of obligation,40 simply noting that taking Track 3 is more seriously wrong than taking Track 2 does not tell the whole story. Suppose that Track 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 very much more wrong than taking Track 2, a fact not revealed by their relative rankings alone. Nonetheless, it remains true that, in taking Track 3, you do wrong on two levels. Level of wrongdoing is to be distinguished from seriousness of wrongdoing in that these constitute two different measures of the extent of wrongdoing. Or better, there are two different extents. I said earlier that it is clear that Alan has the opportunity to be a very naughty boy indeed, but this use of "very naughty" was merely an acknowledgment of all the different levels of obligation applicable to Alan. Perhaps there was nothing that he could do that would have been deontically very bad, and so nothing that would have been very seriously wrong, in which case he did not have the opportunity to be "very naughty" in another sense. Conversely, an agent may have the opportunity to do very serious wrong and yet not have many levels of obligation apply to him. Both senses must be allowed for, and the simplifying 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 that I have proposed does not guarantee that level of wrongdoing corresponds straightforwardly with level of obligation. What I have in mind here can perhaps 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 the possibility of repenting without having sinned.)41 Suppose that Sally takes Track 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, she ought3 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 just one wrongdoing — namely, sinning — and so just one level of wrongdoing? If so, then there is no straightforward correspondence between level of wrongdoing 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 only oughtj not to sin, she also ought2 not to sin. Thus she does indeed do a double wrong — wrong on two levels — when she sins, and that is why her obligation to repent is tertiary. In general, the point can be put this way. Ifwe keep our eye on the "big" picture, we can say that an agent has a primary obligation to take some firstlevel track, a secondary obligation to take some second-level track if he fails to take a first-level track, a tertiary obligation to take some third-level track if he fails both to take a first-level track and to take a second-level track, and so on. It is through wrong being done on one level that an obligation on the next level kicks in. A straightforward correspondence between level of wrongdoing and level of obligation is thus assured. 41 42

Cf. note 22 to this chapter. Cf. note 45 to Ch. 3, concerning Bruce's obligation not to apologize when he does nothing 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 first aid to Carl, who had had a car accident. But he thereby failed to keep his dinner date with Dave, who had badly wanted Steve to sample his latest culinary concoction. What a sorry business! The story is a familiar one. W. D. Ross would say1 that Steve had a prima facie obligation (of beneficence) to tend to Carl and a conflicting prima facie obligation (of fidelity) to keep his date with Dave. Under the circumstances, Steve was quite justified in satisfying the former obligation at the expense of the latter; indeed, he was overall obligated to do so.2 The term "overall obligation" is not Ross's; he uses instead "absolute obligation," "actual obligation," "duty proper," and "duty sans phrase," but none of these seems to me as felicitous as "overall obligation." The term "prima facie 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 so well entrenched, it seems on balance advisable to continue to use it. The main question to be addressed in this chapter is simply this: how is the concept of prima facie obligation to be understood? This of course presupposes that there is indeed such a concept in the first place. This has sometimes been denied, and denied by philosophers from both ends of the normative spectrum. Alan Donagan, for example, a dyed-in-the-wool Kantian, takes Ross to task for claiming that there is a prima facie obligation of fidelity that is overridden by a prima facie obligation of beneficence in the sort of case just given. Donagan says: If a man accepts an invitation to dinner, it would be absurd for his host to understand 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 "resultant" 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 dining. It is a promiser's duty to express any condition to his promise which the promisee might misunderstand; but there would be no misunderstanding in such a case... To any relatively trivial promise there are a host of tacit conditions, all of which will normally be satisfied, which both promiser and promisee must and do understand; and when, as occasionally happens, such a condition is 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 is outweighed by a heavier responsibility.4 I think Donagan is wrong. When Steve helped Carl, he did not keep his promise to Dave. Of course, Steve didn't promise Dave to have dinner with him even if he (Steve) could otherwise bring relief to the victim of an accident; to that extent Donagan is right. Nonetheless, nor did Steve promise to have dinner with Dave only if no such emergency arose; for then he would 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. He simply promised to have dinner with Dave, and this promise was silent, even implicitly, on the issue of conflicting emergencies. And it is at least in part because he broke his promise, while having a (prima facie) obligation not to break his promise, that Steve has the residual obligation to offer Dave an 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 undermined this view.5 It might be thought that, even if Donagan were correct about there being 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 prima facie 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, he treated a number of casuistical problems about promising on these traditional lines [i.e., the lines just advocated by Donagan]. However, he did not see that such a treatment deprives his theory in The Right and the Good of its raison d'etre.6 This seems to suggest that, if no prima facie obligation of fidelity is to be found in cases of the sort just discussed, then no prima facie obligations are to be found anywhere (for it must surely be agreed that the analysis and identification 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 bold claim is given, but certainly Donagan does not concern himself with the notion of prima facie obligation in the presentation of his own theory. At any rate, both the claim and its antecedent strike me as false. At the other end of the normative spectrum are utilitarians, also well known 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 what prima facie obligation itself might be, and their silence is eloquent: they dismiss the notion. One who is not silent on this matter is Fred Feldman (who proposes a modified version of traditional utilitarianism). Feldman explicitly dismisses the concept of prima facie obligation, proclaiming it a "mirage,"7 by which he means that there really is no such concept, so that terms that may appear to express it either are meaningless or express some other concept. Again, no argument is given for this (other than the concept's recalcitrance with respect to Feldman's efforts to analyze it). I believe that the concept ofprima facie obligation is genuine and, moreover, that it is amenable to analysis. Even if I am right, this of course does not mean that either Donagan's or Feldman's normative theory is false; for it is possible that the concept of prima facie obligation be a respectable concept and yet that there be no actual (or at least no very interesting) prima facie obligations. Nonetheless, in presenting my account of prima facie obligation, I shall frequently rely, for purposes of illustration, on Ross's view that there are indeed certain significant prima facie obligations, such as those of fidelity, beneficence, and the like. This is a natural view to hold, I think, and such reliance will render the analysis that I shall give easier to follow. 5.2 DEFECTIVE ANALYSES To many, prima facie obligation has seemed to be overall obligation that is essentially subject to some sort of condition. Ross himself suggests the term "conditional duty" as an alternative to the term "prima facie duty."8 And this may seem quite natural. Consider, for instance, the claim that one ought to keep one's promises, where this is taken to express a prima facie obligation. 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 8

Feldman (1986), p. 144. 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 overall to do A; (5.3) that S makes a promise to do A strictly implies that S ought overall to do A; (5.4) that S makes a promise to do A subjunctively implies that S ought overall to do A. None of these will do, for the simple reason that all such implication warrants factual detachment (see the discussion of (4.1), (4.5), and (4.8), respectively, in the last chapter), whereas it is clear that the fact that one has made or will make a promise to do something cannot by itself transform a prima facie obligation to keep it into an overall obligation to do so. An additional problem with (5.2) and (5.3), though not (5.4), is that neither material nor strict conditionals are defeasible, where a conditional statement of the form "if p, then r" is understood to be defeasible just in case it is possible that it be true while "ifp and q, then r" is false. (That is, to use terminology introduced in Chapter 4, both material and strict implication validate "augmentation.") Such indefeasibility would appear difficult to reconcile with the overridability of prima facie obligation. For while one's prima facie obligation to keep a promise may on occasion also be an overall obligation, 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 to do 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 conditional, itself reformulated as a disjunctive statement, and not just to its consequent.)9 But again this fails to accommodate the overridability of prima facie obligation. Suppose that one makes a promise at Tt to do A at T3 and that one therefore has a prima facie obligation at T2 to do A at T3. According to the proposal, then, one ought overall at T2 either not to make the promise at T1 or to do A at T3. But one cannot at T2 not make the promise at TP Given the principle that, if one ought overall either to do Xor to do Y and one cannot do X, then one ought overall to do Y, it follows that one ought 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 do A at T3, it may be that one has no overall obligation at T2 to do so, because of some more pressing, contrary prima facie obligation.11 The fact that conditional overall obligation, as analyzed in (XIII) in Section 4.2, does not warrant factual detachment and is defeasible (that is, does not validate augmentation) may suggest that it is well suited to an analysis of the concept of prima facie obligation. But, as Feldman has noted, since necessity detachment is valid for statements of conditional overall obligation, 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 to do A (the most obvious type of case being one where the promise has already been made), we may detach from such a conditional overall obligation 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 he has a prima facie obligation to keep his promise and has already promised to do A.12 In light of the fact that the concept of prima facie obligation resists any of the foregoing analyses, Feldman concludes, as noted earlier, that it is merely a "mirage." But this conclusion appears premature, as I shall now seek 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 one has 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 of putting 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, and not to indirect, prima facie obligations, as will be explained shortly.) This way of putting matters is quite revealing, for it indicates that prima facie obligation is indeed essentially overall obligation conditioned in some way 11 Further problems with Hintikka's sort of analysis are discussed in Purtill (1973) and Feldman (1986), p. 182. 12 See Feldman (1986), pp. 130-1. 13 Cf. Pietroski (1993) for an extended investigation into this characterization ofprima 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 act beneficently; one ought (overall), all else being equal, to keep one's promises. Nonetheless, the fact that prima facie obligation is overall obligation conditioned in some way should not blind us to another fact that is frequently overlooked, and that is that prima facie obligation may itself bo either unconditional or conditional. The prima facie obligation of fidelity should be construed as a conditional prima facie obligation: if one makes a promise (or some other sort of commitment) to do A, then one has a prima facie (and not necessarily overall) obligation to do A. For there is, it would seem, no call for fidelity tout court; it is fidelity to a commitment that is required, and, if there is no unconditional requirement to make a commitment (as in general there would seem not to be), there is no unconditional requirement of fidelity. That is, although commitment-plus-fidelity is deontically superior, all else being equal, to commitment-plus-nonfidelity, 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 way of commitment-plus-nonfidelity, or by way of noncommitment-plusnonfidelity - and only the former is tantamount to infidelity; there seems to be no reason to think that commitment-plus-fidelity is deontically superior, all else being equal, to noncommitment-plus-nonfidelity. Having taken note of the possibility that prima facie obligation be either conditional or unconditional, we should note also that the detachment of an unconditional prima facie obligation from a conditional one is analogous to the detachment of an unconditional overall obligation from a conditional one. That is, although factual detachment is not valid, both necessity detachment and deontic detachment are valid. For example, even if one is going to make a promise to do A, one does not yet have for this reason alone an unconditional prima facie obligation to do A; rather, one simply has a conditional prima facie obligation to do A if one promises to do so. But if the promise is inevitable (for example, one has already made it), then one has an unconditional prima facie obligation to keep it; or if one has an unconditional prima facie obligation to make the promise, then one has an unconditional prima facie obligation to keep it. Again, however, whether a prima facie obligation is itself unconditional or conditional, if it is properly characterized as an overall obligation "all else being equal," then it is essentially an overall obligation that is conditioned in some way; for, whether or not it is itself conditional, it carries with it this extra condition of "all else being equal." How are we to understand this extra condition? 146

It would make matters very simple if this condition were some sort of constant (call it E for "equal"). For then the obligation at issue would be straightforward conditional overall obligation of the sort analyzed in (XIII), and both necessity detachment and deontic detachment would be clearly valid. Consider the claim that Faith has a fidelity-based prima facie obligation 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 overall obligation to do P on the condition that both C and E hold. Now consider necessity detachment, where this involves C's being inevitable. If Faith cannot so act that C does not hold, then all worlds accessible to her are Cworlds; 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 deontic detachment (where Faith has an unconditional prima facie obligation to do C): if the best (C&E)-worlds accessible to her are P-worlds and the best E-worlds accessible to her are C-worlds, then, again, the best E-worlds accessible 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 varies from 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 does A is deontically superior to any other world, W\ which is "just like" W except that in W' S does not do A. (As before, this in fact applies only to direct, and not to indirect, prima facie obligation. Again, more on this in a moment.) Of course, this cannot be precisely correct, for no worlds can differ from one another with respect to just one such fact. If S does A in W but not in W\ then any proposition which is distinct from but implies the proposition that S does A will also fail to be true in Wf. What we need to say, then, is that W* is as "close" to W as possible, given that in it S does not do A; that is, Wf is "minimally different" from W with respect to S doing A. This is not an unfamiliar notion. It, or something close to it, has been put 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 world that is obtained from W by making minimal changes which suffice to make it 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. It might seem that all that needs to be said is this: W' is minimally different from W with respect to S doing A if and only if S does A'mW but not in W' and every proposition that is true in W but which does not imply that S does A is also true in Wf. But this won't do. Consider any two propositions^ and q, neither of which alone implies that S does A but which jointly imply this. Given the current proposal, each of p and q will be true in W'\ but then it will be true after all that S does A in W'. Pollock has himself proposed an analysis of minimal difference that does not succumb to this problem.15 It may be that this analysis is acceptable, but I am reluctant to commit myself to it; for it is very complex and, besides, rests on the controversial notion of a "simple" proposition. Moreover, the analysis (as is typical in the treatment of counterfactuals) involves a past—future asymmetry, so that similarity of the past is weighted more heavily than similarity of the future; and I do not wish to presuppose any such asymmetry here. Formally, then, I shall invoke the concept of the closeness ofpossible worlds unanalyzed. I hope that my use of this concept will be sufficiently intuitive to render what I have to say about prima facie obligation both plausible and helpful, just as its use (or the use of some closely related concept) has proven helpful in the analysis of counterfactuals. For reasons (having to do with detachment) that will become apparent shortly, I shall need to talk of those worlds that are minimally different for SatT in Wwkh respect to S doing A; that is, the focus of attention will be on those worlds that are accessible to S. The canonical phrase form in this context is therefore this: W" is minimally different from W\ for S at Tin W, with respect to 5 doing A at T". This may be understood in terms of the closeness of worlds as follows: (5.7) W" is minimally different from W\ for S at Tin W, with respect to 5 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 i n W, with respect to 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 and conditional, we may proceed to an analysis of prima facie obligation. I have suggested: 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 superior to all accessible worlds that are minimally different in this respect. This double use of "all" implies that there is something in general - "across the board" - to be said in favor of doing A. It is this that indicates that doing A is of greater deontic value, taken in and of itself, than not doing A. I propose 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 to do 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 at T'; 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 do A 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 at T', 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 deontic reasons; for there is a sense of "reason" according to which one has a reason 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 that they may be unconditional or merely conditional. Second (5.9) and (5.10) imply that, if 5 has a direct prima facie obligation to do A, then 5 can do A and S can refrain from doing A. A is, in other words, personally optional for S. This is in keeping with the analysis of overall obligation provided in Chapters 2 and 4. Third, (5.9) and (5.10) are restricted to what I have called direct prima facie obligations. (Ross uses the term "direct" similarly.)17 If Ben has a direct prima facie obligation to act beneficently, this consists in every accessible beneficence-world being deontically superior to any accessible world that is minimally different with respect to beneficence. Similarly, if Faith has a direct prima facie obligation to pay Pam $10, if she so promises, this consists in every accessible promise-plus-payment world being deontically superior to any accessible promise-world that is minimally different with respect to payment. But not all prima facie obligations are direct. Some are indirect, and there are two ways in which they may be so. First, some may be called (following Ross again)18 incidental. For example, if Ben cannot act beneficently without driving to Boston, then he has an incidental unconditional prima facie obligation to drive to Boston. Note that, since it would be highly surprising if all accessible worlds in which Ben drives to Boston were deontically superior to all accessible worlds that are minimally different in this respect, we have no reason to think that Ben has a direct unconditional obligation to drive to Boston. But, as long as he must drive to Boston to display beneficence, Ben does have an incidental, beneficence-based obligation to drive to Boston. Similarly, if Faith cannot keep her promise with16 There is also another, nondeontic sense of "reason" that I shall not seek to characterize here. 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 incidental conditional prima facie obligation to go to the bank and withdraw some cash. The principle here is straightforward. It may be roughly stated as follows: if S has a direct prima facie obligation to do A and cannot do A without 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 S from W at T, if p is true in W\ then q is true in W'. W e may then account for incidental prima facie obligations as follows: (5.12) S has, at T in W, an incidental unconditional prima facie obligation 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 A and 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 H 7 ' 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 facie obligations are also said to be incidental. The fourth clause is required to preserve the thesis that whatever one has a prima facie obligation to do is personally optional. The second way in which prima facie obligations can be indirect may be illustrated as follows. Suppose that Ben has an incidental prima facie obligation (grounded in beneficence) to drive to Boston and also has an incidental prima facie obligation (grounded in nonmaleficence) not to go to 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 corresponding principle concerning overall obligation, namely, (2.43'")- See the discussion in Subsection 2.3.2.

151

although it would not be mean of him not to visit his mother, it would be mean of him to go to Boston and not visit her). Then Ben has what I shall call a compound prima facie obligation (grounded in beneficence and nonmaleficence) to visit his mother. The principle at issue here may be roughly stated as follows: if S has a prima facie obligation (whether direct or incidental or, indeed, compound) to do A and also has a prima facie obligation not to do A without doing B, then S has a compound prima facie obligation to do B. More precisely:20 (5.14) S has, at T in W, a compound unconditional prima facie obligation 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 A and 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 in Section 2.1 included analyses of overall permissibility and wrongdoing, so too the present account of prima facie obligation may be extended to include analyses of prima facie permissibility and wrongdoing. Consider wrongdoing 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, at T in W, a prima facie obligation not to do B at T". 20

Compare the corresponding principle for overall obligation, (2.42'), discussed in Subsection 2.3.1. 21 A still more precise account of compound prima facie obligation, both unconditional and conditional, would (in order to avoid any appearance of circularity) distinguish between level-1 compound obligations, which would themselves be generated by obligations 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 obligation not to do something. Such an account can be developed along lines exactly analogous to those along which the account of what it is to have a prima facie obligation to do something has been developed. That is, instead of talking 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 to all accessible worlds that are minimally different in this respect. (For example, it may be that Malcolm has a direct unconditional prima facie obligation not to act maleficently; this would involve all accessible nonmaleficence-worlds being better than all accessible worlds that are minimally different in this respect.) Instead of talking of an incidental unconditional prima facie obligation to do B (see (5.12)), we can talk of an incidental unconditional prima facie obligation not to do B; this would involve not doing A requiring not doing B or, equivalently, doing B requiring doing A. (For example, it may be that Malcolm cannot play a certain practicaljoke without acting maleficently; given his nonmaleficence-based reason not to act maleficently, he then has a nonmaleficence-based reason not to play the practicaljoke.) And so on. An account of what it is to have a conditional prima facie obligation not to do something may be likewise developed. Developing an account of what it is for an act to be prima facie permissible is slightly more tricky. Evidently this must be grounded in the notion of one's not having a direct unconditional prima facie obligation not to do A, but how precisely is it to be accounted for? Should we say simply: some accessible ^4-world is at least as good as all worlds that are minimally different in this respect? Or should we say: all accessible ^4-worlds are at least as good as all worlds that are minimally different in this respect? I think it is clear, on reflection, that we should say the latter; for the former is far too weak. It may well be, for instance, that some walking-world accessible to Willis is at least as good as all worlds that are minimally different in this respect, but we shouldn't infer from this that it is prima facie permissible for Willis to walk. For it may be that every direct prima facie obligation that Willis has requires that he not walk. On the other hand, if all beneficenceworlds accessible to Ben are at least as good as all worlds that are minimally different in this respect, we may indeed infer that it is prima facie permissible, 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 supplemented with an account of both indirect and conditional prima facie permission, but I shall forbear doing this here. 153

5.3.2 Three objections The foregoing account of prima facie obligation may appear objectionable for at least three reasons. First, it might be thought that the account implies that whatever is directly prima facie obligatory is also overall obligatory (or at least overall permissible), for the following reason. According to the account, 5 has a direct prima facie obligation to do A just in case, for every two accessible worlds that are just alike except in one S does A and in the other 5 doesn't do A, the former is deontically superior to the latter. But this appears to imply that the deontically best accessible world (or worlds), W, is one in which S does A; for otherwise that accessible world that is just like W except in it 5 does A would be better than W. Clearly such an implication would be unacceptable, but fortunately the present account doesn't have it. The term "just alike" is here not just rough but inaccurate. (Note that, when giving my rough, intuitive account of "all else being equal" above, I used the term "just like" and not the term "just alike.") In particular, one cannot assume that, if W" is minimally different from W' (for 5 at T in W) with respect to S doing A at T", then W' is minimally different from W" with respect to S not doing A at T". Suppose, for example, that Roy is faced with conflicting direct prima facie obligations of reparation and beneficence. Let us say that the deontically best accessible world is one in which he satisfies the former but not the latter (he cannot satisfy them both). Next best would be his satisfying the latter but not the former. Worse still would be his satisfying neither. This may be represented as in Chart 5.1 (where R stands for "reparation" and B for "beneficence") . Given a proper filling-in of "...," we can say that W2 is minimally different from Wx with respect to Roy's not acting beneficently; and yet W^ (and not Wr) is minimally different from W2 with respect to Roy's acting beneficently. Thus what is prima facie obligatory need not also be overall obligatory or permissible. A second objection stems from consideration of cases such as the following. Suppose, as in Subsection 3.2.5, that Gertrude owes Lucy a debt of gratitude. The most appropriate way for her to display gratitude would be 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 would constitute some display of gratitude, Gertrude has some reason, and hence a prima facie obligation, to send daisies; it's just that this obligation is overridden by the obligation to send lilies. But my account does not warrant our saying this. Presumably Gertrude has no direct prima facie obligation to send daisies, and doing so is not necessary for a display of gratitude and hence is not even incidentally prima facie obligatory (barring its being necessary for the satisfaction of some other direct prima facie obligation). In fact, it might appear that I cannot even say that Gertrude has an incidental prima facie obligation to send lilies, since even this appears not to be necessary for a display of gratitude (in that sending daisies would be an alternative, albeit inferior). What I think we should say here is this. Granting that gratitude comes in degrees, let us say that "full" gratitude is gratitude to degree 10. Gertrude's sending lilies would thus constitute a display of gratitude to degree 10; her sending daisies would, let us say, only constitute a display of gratitude to degree 5. Now, while it is plausible to say that a display of gratitude is better, all else being equal, than a lack of a display of gratitude, it doesn't seem plausible to say that a display of gratitude to degree x is better, all else being equal, than a lack of a display of gratitude to degree x; for what if the lack of a display of gratitude to degree x is due to a display of gratitude to a greater degree y? Still, it is plausible to say that a display of gratitude to at least degree x is better, all else being equal, than a lack of a display of gratitude to at least degree x. This observation allows us to say that Gertrude does indeed have an incidental obligation to send Lucy lilies (given that there is no degree of gratitude greater than 10 and that sending lilies is the only way to display gratitude to that degree); and, although it won't allow us to say that she also has an incidental obligation to send Lucy daisies, it does allow us to say that she has an incidental obligation either to send lilies or to send daisies (given that nothing else would constitute a display of gratitude to at least degree 5). And this indeed seems to be what should be said. In fact, the matter is even more complex than this. Suppose that Gertrude ought, from gratitude, to send Lucy lilies, and also ought, from gratitude, to send Chrissie chrysanthemums, but cannot do both. Here, it seems, gratitude requires the impossible, and yet my account implies that whatever is prima facie obligatory is possible. The solution, of course, is that gratitude, more fully characterized, is gratitude-to-a-benefactor (or, perhaps, would-be benefactor, or presumed benefactor), just as fidelity is fidelity-to-a-promisee (or someone to whom another sort of commitment 155

has been made). (If we talk of "what gratitude requires" without qualification, then, we must mean what gratitude to every benefactor requires; so too for fidelity, reparation, and any other obligation which is owed to a particular person or persons.)22 My account implies, properly, that Gertrude can satisfy each ofher prima facie obligations of gratitude-to-Lucy and gratitude-to-Chrissie; it does not imply, nor should it, that she can satisfy them both. A third objection is more troublesome. It is that the account implies that certain direct prima facie obligations that are intuitively thought of as being merely conditional turn out also to be unconditional. Consider fidelity. This has been assumed (for purposes of illustration) to constitute a direct conditional prima facie obligation, in that all commitment-plus-fidelityworlds have been assumed to be deontically superior to any worlds that are minimally different with respect to fidelity. But notice that one cannot (it seems) display fidelity in the absence of a commitment; that is, all accessible fidelity-worlds are also commitment-worlds. But then, if all accessible commitment-plus-fidelity-worlds are deontically superior to their closest commitment-plus-nonfidelity-worlds, it must also be true that all accessible fidelity-worlds are deontically superior to their closest nonfidelityworlds. In other words, the direct prima facie obligation of fidelity will be not just conditional but unconditional. The same applies to other (alleged) conditional direct prima facie obligations such as those of gratitude and reparation. For one cannot (it seems) display gratitude in the absence of that condition (namely, being a beneficiary) on which it is conditionally prima facie obligatory; likewise, one cannot make reparation for a nonexistent harm. It is tempting at first simply to bite the bullet here but then go on to point out that there are still certain intimately related direct prima facie obligations that are merely conditional. For example, even if Faith cannot display fidelity to her promise to pay Pam $10 without making that promise, still she can pay Pam $10 without promising to do so. Thus, while all accessible promise-plus-payment-worlds may be better than their closest accessible promise-plus-nonpayment-worlds, there is no need to think that all accessible payment-worlds are better than their closest accessible nonpayment-worlds. And isn't that what really matters here? Well, no, it isn't. There are two problems with this. First, on some occasions, 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. Although Faith can do what fidelity requires (namely, pay Pam $10) without promising to do it, it may be that Roy cannot do what reparation requires (mending a broken vase, say) without committing the relevant harm. Secondly, if it is admitted that Faith has a direct unconditional prima facie obligation to display fidelity (even if it is also pointed out that her direct prima facie obligation to pay Pam $10 is merely conditional), it must then also be admitted, given the foregoing account, that she has an incidental prima facie obligation to make the relevant promise that is tmconditional. Similarly, if it is admitted that Roy has a direct unconditional prima facie obligation to make some reparation, then it must also be admitted that he has an incidental prima facie obligation to commit the relevant harm that is ^conditional. This is surely to admit too much. The original account must therefore be revised. We must deny not only that Faith has an unconditional prima facie obligation to pay Pam $10 but also that she has an unconditional prima facie obligation to display fidelity to the commitment to pay Pam $10 (given that she has not already made this commitment and that there is no independent unconditional reason to make it). (Similar remarks pertain to the prima facie obligations of gratitude and reparation.) What I think warrants such denial is an observation already made above: although commitment-plusfidelity is deontically superior, all else being equal, to commitmentplus-nonfidelity, it is not deontically superior, all else being equal, to noncommitment-plus-nonfidelity; for nonfidelity constitutes infidelity only when there is an actual commitment. It is in virtue of this fact that the prima facie obligation of fidelity is merely conditional. But how can this be? After all, consider a situation where it is possible to display a great deal of maleficence to one person while also displaying minor beneficence to another. We may suppose in this case that maleficence-plus-beneficence is deontically superior, all else being equal, to maleficence-plus-nonbeneficence, and also that maleficence-plus-beneficence is not deontically superior, all else being equal, to nonmaleficence-plus-nonbeneficence. Yet this doesn't make the prima facie obligation of beneficence a merely conditional one. So why should the corresponding fact about commitment-plusfidelity make the prima facie obligation of fidelity a merely conditional one? The answer is to be found in the fact that commitment is a prerequisite of fidelity, whereas maleficence is not a prerequisite of beneficence. This is the 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 a commitment, but in the sense that fidelity requires that a commitment precede 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 be unconditionally obligatory. (Similar remarks pertain to gratitude and reparation and their respective initial conditions.) To put this point more precisely, let us first say this: (5.17) W" is minimally different from W', for 5 at T in W, with respect to 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 clause as follows: (5.9') S has, at T in IV, a direct unconditional prima facie obligation to do 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 ^4 at T'; (b) for all worlds H/' and H/", if (1) W" is minimally different from PF', for 5 at T in M^, with respect to S doing A at T', then (2) the deontic value for S at T of W7' is greater than the deontic 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 obligation 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, with respect 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 the deontic 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 possibility that deontic value be significantly agent- or time-relative, inasmuch as the relevant value of the worlds is the value they have for S at T. Let me now elaborate on this with respect to Ross's theory. I said earlier that at least some of the values that ground prima facie obligations are thought by Ross to be significantly agent-relative. Fidelity, for example, is clearly a value of this sort. Ross says that each agent always has a prima facie obligation to see to it that he keeps his promises.231 take this to imply that each of us always has a (conditional) prima facie obligation to see to it that we ourselves display fidelity. Now it might be that Ross also believes that each of us always has a prima facie obligation to see to it that others display fidelity; for he might believe that such displays are intrinsically valuable, and he does say that we do always have a prima facie obligation to do that which will promote such value.24 But even if this is so, it does not detract from the fact that it seems pretty clear that, according to Ross, for each of us a special value attaches to our displaying fidelity. So too for gratitude and reparation, it seems.25 Whether or not this is so for other prima facie obligations that Ross lists is unclear to me. Does he believe, for instance, that there is a 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 does not believe that each of us always has, in virtue of the prima facie obligation to promote intrinsic value, a prima facie obligation to see to it that others display fidelity. For there he 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 (save combinations of those mentioned). It might be thought that Ross takes fidelity to be a virtue, but this would be a mistake. For he takes virtue to concern motive, and he explicitly divorces the duty of what he calls "fidelity" from motive. (See Ross, 1930, p. 22 and also p. 5.) Still, Ross does say (1930, p. 36) that fidelity can be bonific - that is, instrumentally good — and so he clearly thinks that, on those occasions on which it is bonific, there is 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 from harming innocent people? I am not sure, although certain of his remarks suggest that he does believe this.26 5.3.3 Detachment The move from (5.9) to (5.9') blocks factual detachment. Even if Faith will promise to pay Pam $10, the prima facie obligation to display fidelity to this commitment is merely conditional; for clause (c) of (5.9') is not satisfied. To be acceptable, though, the account must not only block factual detachment, it must also validate necessity detachment and deontic detachment. And so it does. Consider necessity detachment first. If Faith cannot avoid the commitment to pay Pam $10, then clause (c) of (5.9') is satisfied after all, and trivially so; for there will then be no accessible worlds that are minimally different with respect to Faith's displaying fidelity to this commitment, on the condition that her making this commitment varies. Thus her conditional prima facie obligation to display fidelity is transformed into an unconditional prima facie obligation to do so, and her obligation to pay Pam $10 is likewise transformed. Deontic detachment also holds, although the story here is more complicated. Let us look first at deontic detachment having to do with direct prima 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 obligation to 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 ^conditional prima facie obligation to pay Pam $10. And it does. Consider the world-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 make promise 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 $10 if 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 obligaW,: P2 & P & ... W2: P2 & ~P & ... W3:~P2 & P & ...

Chart 5.2 26

See Ross (1930), p. 22.

160

IV,: B & P & W2: B & ~P & W3:~B& P & Chart 5.3

IV,: B & P2 & P & W2: B & P2 & ~P & Chart 5.4

tion to do (P2&P).27 Thus she also has an incidental prima facie obligation to do P. Consider, next, incidental prima facie obligations. Suppose that Faith's prima facie obligation to make promise 2 is not grounded in any alreadymade promise but rather in beneficence. Beneficence requires that she make 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 unconditionally obligatory, Wt is better than W3. In addition, since B requires P2 (where P2 stands for "promise 2"), Wt and W2 must, more precisely, have the forms represented in Chart 5.4. Since Pis directly prima facie obligatory 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 that beneficence requires that Faith visit her mother and that nonmaleficence requires that she not visit her mother without making promise 2. Then she has an unconditional compound obligation to make promise 2. Consider the world-schemata represented in Chart 5.5 (where B stands for "benefiIV,: B W2: B IV3: B W4: ~B

& N & P & & N & ~P & & ~N a P & & N & P & Chart 5.5

27

Clause (c) of (5.9') is satisfied, since the only relevant prerequisite of (P2 & P) is promise 1. 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 or W3 is. (Still, I think that Wi is better than W4, and that this could be satisfactorily proven.)

161

W,: B & N & W2 : B & N &

13 & P 1O & ~P

Chart 5.6 cence," AT for "nonmaleficence," and Pfor "payment"). Given the direct unconditional prima facie obligation of beneficence, we may say that Wi is better than W4. Given the direct unconditional prima facie obligation of nonmaleficence, 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 prima facie 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 obligation by another emerges from the foregoing. Roughly, what we should say is simply this: S's prima facie obligation to do A is overridden by his prima facie obligation to do B if and only if every accessible world in which he does 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 overridden by 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 or conditional, direct or indirect. Notice that it is possible that direct obligations override in a manner that fails to match the manner in which their corresponding incidental obligations override. Suppose that Ben has a direct prima facie obligation of beneficence and that on a certain occasion this overrides a direct prima facie obligation of fidelity. Suppose also that he has the corresponding incidental 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 obligation to drive to Baltimore; indeed, it may even be that the latter overrides the former. This would happen if there is yet another direct prima facie obligation, say an obligation of reparation, that overrides the obligation of beneficence and the fulfillment of which itself requires that Ben drive to Baltimore. 5.4 R O S S O N P R I M A FACIE O B L I G A T I O N Since it was Ross who introduced the concept of prima facie obligation to us, I have relied fairly heavily on what he says in the presentation of my own analysis of the concept. I have invoked his distinction concerning direct and incidental obligations, and I have made free use of some of his examples of prima facie obligation (fidelity, beneficence, gratitude, reparation, and so on). It is appropriate at this point to try to determine to what extent my account squares with Ross's, at least with regard to certain key issues. 5.4.1 Conditional and unconditional prima facie obligation First, it might be thought that Ross does not distinguish between conditional and unconditional prima facie obligation, inasmuch as he sometimes refers to all prima facie obligation by the term "conditional duty." But in fact I think that Ross does make this distinction but simply uses different terminology to do so. I have in mind his comments on what he calls "special" 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 conditional on the performance of these acts. General obligations (among which Ross counts the obligations of justice, beneficence, and nonmaleficence) do not rest on the performance of such prior acts and may therefore be said to be unconditional. At least, this is how I understand Ross's distinction between special and general obligations and what he means by "rest on." Note that, in some other, more liberal sense of "rest on," all obligations, both conditional and unconditional, will rest on the satisfaction of some condition, for the performance of any act requires that certain conditions be satisfied. For example, 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 no one can administer justice unless there is occasion for doing so, and no one can 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 does fidelity require commitment, but noncommitment-plus-nonfidelity is no worse than (that is, is not deontically inferior to) commitment-plus-fidelity, and so fidelity isn't unconditionally obligatory; similarly, noninflictionof-harm-plus-nonreparation is no worse (indeed, it is better, given the prima facie duty of nonmaleficence) than infliction-of-harm-plus-reparation, and so reparation isn't unconditionally obligatory; and so on. But with respect to beneficence, for example, there is no "prior act" like that of commitment (or the infliction of harm) which it is prima facie permissible to forgo, and so nothing that is morally on a par with noncommitment-plusnonfidelity. 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 conditionally required, in that its obligatoriness rests on the performance of some prior act (such as that of becoming a neighbor of the potential beneficiary, in some sense of "neighbor" according to which not everyone else is automatically 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-evident. His remarks concerning this matter clearly indicate that he believes that it is a necessary truth that certain acts are prima facie obligatory (whether conditionally or unconditionally).29 Such a claim of course goes well 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 of what it is to claim that a particular agent, S, has a prima facie obligation at a particular time. Unless S exists necessarily, no such claim can be a necessary truth. Still, we can allow for necessary truths concerning prima facie obligation by expanding the account that has already been given. We can, in particular, say the following:30 29 30

Ross (1930), p. 29. Here, as in the presentation of (2.11)—(2.14) at the end of Subsection 2.1.4, acts are treated as being "repeatable." But see again the remarks made in Subsection 2.2.5 on this matter.

164

(5.19) S has a permanent unconditional prima facie obligation in W to 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 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 A iff for all times T and T', there is a proposition p such that, if (a) there is a world accessible to S from J^F 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 occasions where it is applicable (rather than simply on all occasions, period, which would imply that it is always applicable, that is, can always be acted on; 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 do A 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 iff for all worlds W, there is a universal unconditional prima facie obligation in W to do A; 165

(5.24) there is a necessary conditional prima facie obligation to do A iff for all worlds W, there is a universal conditional prima facie obligation in Wto do A. One can then go on to claim, as Ross appears to do, that, if ever an agent has a prima facie obligation to do some act A, this is grounded in the fact that there is a necessary direct prima facie obligation to do either A itself or some other act for which A is necessary.31 Although the analysis that I've provided does not imply this claim, it is certainly consistent with it. Nonetheless, there are complications, one minor, one major, in trying to reconcile what Ross has to say about the self-evidence of certain particular prima facie obligations with the account that I have provided. The minor complication concerns the (alleged) prima facie obligation of fidelity. Although I have allowed myself to invoke this obligation for the sake of illustration here, it is in fact doubtful that there is a necessary direct prima facie obligation of fidelity of the sort that Ross alleges. For what he appears to say is that every commitment necessarily givesriseto a prima facie obligation to fulfill that commitment. This cannot be right. For one thing, people sometimes promise to do something that they in fact cannot do. The account that I have given implies that there is no prima facie obligation to keep such promises.33 Of course, even if this is what Ross says, there is an easy qualification available, and that is to say that every commitment necessarily givesriseto a prima facie obligation to fulfill it, if the agent can fulfill it. But there is reason to think that this, too, is false. For it often happens that people promise to do what is in fact overall wrong. What reason is there then to think that all commitment-worlds accessible to the agent in which he displays fidelity are better than all accessible commitment-worlds that are minimally different in this respect? For any world that is minimally different in this respect will at least involve his not committing the overall wrong in question. Once again, though, a qualification is perhaps available, 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 facie obligation to do A. Ross does not appear explicitly to acknowledge the possibility of compound 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 apparent promises to do what is impossible are in fact not genuine promises, and then claim that 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 should sincerely, and thus genuinely, promise to do something, in the mistaken belief that he can 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 prima facie obligations that Ross alleges to be self-evident, except perhaps the prima facie obligation to promote intrinsic value.35 The account that I have provided relies on the notion of minimal difference, where this is explicitly tied to worlds that are accessible to the agent in question. But surely it is possible that some agents be such that they cannot avoid some evil's occurring if they satisfy some Ross-alleged direct prima facie obligation; in such a case, given that the evil is sufficiently great (so that it greatly diminishes the deontic value of any world in which it occurs and renders any world in which it does not occur very valuable, deontically, relative to any world in which it does occur), it will not be the case that all accessible worlds in which the alleged obligation is satisfied are deontically superior to all accessible worlds that are minimally different in this respect. Hence this alleged obligation does not hold for such people; hence it does not hold of necessity; hence it is not self-evident that it holds. As an illustration of this, consider the possibility that Martha will commit mass murder if and only if Malcolm acts nonmaleficently, and that Malcolm can do nothing about this. We may suppose (barring certain Kantian considerations) that not all worlds accessible to Malcolm in which he acts nonmaleficently are deontically superior to all worlds accessible to him that are minimally different in this respect. In fact, it may well be that the very reverse is true, so that Malcolm 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 a necessary 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 that Ross calls "special" cannot hold of necessity, on the account that I have provided. If we are to accommodate what Ross says about the self-evidence of such prima facie obligations, therefore, a different account must be given. Let me indicate briefly here how I think this account should go. First, we would need to revise our understanding of minimal difference, so as not to restrict our focus to worlds accessible to the agent. We would simply revise (5.7) by saying: (5.7')

35

W' is minimally different from W with respect to S doing A at T' iff

The example I give below concerns an act that is plausibly thought to involve a great intrinsic evil. But if there are acts that do not involve great intrinsic evils but which nonetheless greatly diminish the deontic values of worlds in which they occur, then even the 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 do A 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) H 7 " 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 5 doing A at T', on the condition that p is true, then (2) the deontic value for 5 at T of W' is greater than the deontic value for S at T of W . Changes to (5.9') and (5.11)—(5.24) would then be made accordingly. On this approach, the personal optionality of what is prima facie obligatory is preserved (see subclauses (al) and (a2) of (5.10')), but it could still be that Malcolm has a prima facie obligation to act nonmaleficentry despite his inability 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 procedure mistaken; on the contrary, it gives us an account of what might be called prima facie obligations "in principle." W e 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 do indeed hold for conditional prima facie obligations in principle, they do not warrant certain judgments that we wish intuitively to make. For example, the version of necessity detachment that concerns prima facie obligations in principle involves inferences of this form: if all ^-worlds in which S does A 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 better than all worlds that are minimally different in this respect. Note: "all 168

worlds," not just "all accessible worlds"; this is because minimal difference is now being understood in terms of possible, and not just accessible, worlds. Now, with necessity detachment so understood, we could not employ it to infer, as we surely wish to, that, for example, someone who has 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 the stronger sense of "necessary" now at issue. Similarly, for deontic detachment with respect to prima facie obligations in principle, the unconditional obligation that triggers the detachment must itself be an obligation in principle, and this fails to validate certain detachments that we intuitively wish to accept (such as that of detaching an unconditional obligation to do A by appealing to an unconditional obligation to promise to do A, the latter having itself been detached - by necessity detachment - owing to considerations having to do merely with inevitability and not with necessity in the 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 facie obligations in principle, and there are prima facie obligations in practice. Thus there are also (necessity and deontic) detachment in principle (where the obligations in question are prima facie obligations in principle) and detachment in practice (where the obligations in question are prima facie obligations in practice). The detailed account that I have provided is of prima facie obligations and detachment in practice; the account that I have sketched in the last two paragraphs is of prima facie obligations and detachment in principle. We need both. We need the latter to accommodate what Ross says about the self-evidence and necessity of prima facie obligations; we need the former to accommodate common intuitions concerning the detachment 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 prima facie obligation and overall obligation. Here he makes two important claims. The first is this: I suggest 'primafacie duty' or 'conditional duty' as a brief way of referring to the characteristic.. .of being an act which would be a duty proper if it were not at the 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 overall obligation to do A if any prima facie obligation that 5 has not to do A were overridden. Does my account imply that this is true? It might be thought that any account of obligation that purports to capture 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 they are consistent with one another). The implication fails because my account of 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 up a failing in either of these accounts? I don't think so. Ross seems to think that there are several basic, independent values (such as those of fidelity, gratitude, and so on); that nothing 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 all this, and we also grant (as seems natural) that the deontic value of a world is a direct function of those deontically relevant values which are promoted at that world, then, given the foregoing accounts of prima facie and overall obligation, it seems that we could argue cogently for (5.25); for there will be no deontically relevant value that does not correspond to some prima facie obligation.38 But whether or not we should grant all this seems to me to go well beyond questions of conceptual analysis; it is a normative question. (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 betrayed such a normative bias. Nonetheless, (5.25) might seem promising as the basis of an analysis of the concept of prima facie obligation, an analysis that is quite different from the one that I have proposed. Of course, (5.25) could not itself be accepted as such an analysis, because it would be circular. But consider this proposal 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 number of complications, having to do with the distinctions between conditional and unconditional obligations, obligations in principle and obligations in practice, permanent versus universal 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 acttypes (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 in virtue of which an act-token may be prima facie obligatory. Circularity is avoided, however, since such identification of T is not necessary for the statement of (5.26). In addition, the proposal would appear to avoid the normatively biased implication that whatever is overall obligatory is prima facie obligatory, insofar as it requires that, to be prima facie obligatory, A be a token of some member of T, whereas this is not said to be a requirement for an act's being overall obligatory. But Snare's proposal is nonetheless problematic. It would seem to fall afoul of the fact that subjunctive conditionals (of the sort employed in the proposal) warrant factual detachment. For might it not happen that, although the antecedent of clause (b) in (5.26) is satisfied, S ought so to act that it is not satisfied? Consider this case. Ben is in a position at T1 to perform some beneficent act A at T 2 . Let us assume that acting beneficently is a type of act that is a member of set T, and that Ben (therefore) has a prima facie obligation at Tt to do A at T2. Let us suppose also that, if he omits to do A at T 2 , he will not thereby perform an act that is of a type that is also a member of T. In this case, (5.26) implies that Ben has an overall obligation at Tt to do A at T2. But suppose that Ben ought overall at T1 to make and then keep a certain promise, and that the keeping of it would require him to omit to do A at T2. He doesn't make this promise, and that is why his omitting to do A is not otherwise morally significant. (The omission is not an instance of fidelity, for example.) But he ought overall both to make it and to keep it, and so (contrary to what (5.26) implies) he ought overall not to do A. One might respond to this case by maintaining that Ben's omitting to do 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 of some member of T, since after all he does not do what he overall ought to do. But this would be to revert to the normative bias of (5.25), for it would be to insist that nothing is overall obligatory unless it exemplifies some member 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 tinkering with its clause (b) so as to circumvent this objection,40 but I don't think this would 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 is true. Next, take any type of action of which A is a token and let T be the set which has that type as its only member; then (a) is true and (b)'s antecedent is also (likely to be) true. Hence both (a) and (b) are true, and from this Snare's proposal allows us to infer that 5 has a prima facie obligation to do A. For these and other42 reasons, then, I think that Snare's proposal is to be rejected, and in general the project of giving an analysis of the concept of prima facie obligation on the basis of (5.25) seems to me unpromising. I began this subsection by saying that Ross makes two important claims about the relation between prima facie obligation and overall obligation. The second claim is this: Every act..., viewed in some aspects, will be prima facieright,and viewed in others, prima facie wrong, and right acts can be distinguished from wrong acts only as being those which, of all those possible for the agent in the circumstances, have the greatest balance of prima facietightness,in those respects in which they are prima facie right, over their prima facie wrongness, in those respects 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 obligation to do A that overrides any prima facie obligation that S has not to 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 omission of A were not also a token of some member of T, andifit 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 the latter point, see Subsection 2.2.5. 43 Ross (1930), p. 41.

172

the generality inherent in prima facie obligation is also present in overall obligation. Even though (I) implies neither (5.25) nor (5.27), it is certainly consistent with them. With respect to (5.27), though, this claim might be disputed, on the grounds that Ross conceives of it as an analysis.44 If overall obligation is to be understood in terms of prima facie obligation, as this interpretation of (5.27) contends, then it is not to be analyzed independently 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 analysis of overall obligation, my proposed analysis of overall obligation must be rejected.451 therefore reject (5.27) as an analysis (though I grant that it may express a necessary truth). How does Ross himself view (5.27)? The evidence seems to me inconclusive: although some passages46 in The Right and the Good suggest that he does accept (5.27) as an analysis, he never states this explicitly; moreover, other passages47 suggest that he eschews any such analysis. At any rate, if Ross does accept (5.27) as an analysis, I oppose him on this point. I leave it to you to judge whether this indicates an inadequacy 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 obligations are self-evident, but he also claims that one's overall obligations are not and that one may appeal to the former when trying to determine the latter. This is surely plausible, but it may appear to indicate a priority of prima facie obligation over overall obligation that is inconsistent with (I). Not so. As Snare has pointed out, the logical or conceptual independence of overall obligation from prima facie obligation is not inconsistent with the former being epistemologically dependent on the latter.48 Similarly, one might claim that sometimes it is appropriate to appeal to one's prima facie 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 priority of prima facie obligation over overall obligation. For it may well be that certain matters are, and that we find them to be, of fundamental moral 44

45 46 47 48 49

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 and Belzer (1991), pp. 365-6. 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. Ross (1930), pp. 41 and 46-7. Ross (1930), pp. 10-12 and 16. Snare (1974), p. 244. Cf. again Ross (1930), p. 46. Cf. also Pietroski (1993), p. 513.

173

significance —well-being, perhaps, or respect, or integrity, or fidelity, or justice, or freedom, or something else — such that our overall obligations turn on, 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, we will agree that there are prima facie obligations, both in principle and in practice, having to do with the promotion of these values, so that sometimes, when we deliberate about what we ought overall to do in some situation, we may bear in mind that these prima facie obligations apply in this situation. But this is quite consistent with the analyses of overall and prima facie obligation that have been proposed here. 5.4.4 Residual obligation Steve missed his dinner date with Dave. Although he was perfectly justified in doing so - for Carl was in dire need of his ministrations - still there was a cost involved, a moral cost. He broke his promise, and Dave was disappointed. For this reason, Steve's helping Carl appears to have left a certain "moral residue," as it is sometimes put. Ross distinguishes two types of moral residue, one involving emotion, the other action. He says: When we think ourselves justified in breaking, and indeed morally obliged to break, a promise in order to relieve some one's distress, we do not for a moment cease to recognize a prima facie duty to keep our promise, and this leads us to feel, not indeed shame or repentance, but certainly compunction, for behaving as we do; we recognize, further, that it is our duty to make up somehow to the promisee for the breaking of the promise.50 Few would quarrel with this, and yet it is a fact that my account of prima facie obligation does not imply that it is true. Is this a problem with the account? I cannot see that it is. There might be a problem if the account implied that 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 a conceptual one; it is normative.51 To say that it is appropriate to feel such compunction (as Ross obviously believes; it is not just that we are led to feel it, 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 mis50 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 conceptual one. Indeed, although the view may be natural, it is clearly not uncontroversial. Some would claim that, as long as one does what is morally justifiable, 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 this is false, nonetheless we would do well to note two additional points. First, even if compunction (whatever that is, exactly) is called for in cases like Steve's, certain other "negative" moral emotions (such as guilt or shame) certainly are not; second, even if compunction is called for, still, as long as one has clearly done what one overall ought, an overall sense of moral ease would seem appropriate.54 Similarly, some would claim that, as long as one acts inculpably, then even if, like Steve, one breaks a promise to someone, one does not owe that person anything. Indeed, /would claim this (where it is understood that to say that S owes S' something is to say that 5 has an obligation to S' to do something. Here I am talking of what in Section 1.3 I called "obligationto" as opposed to "mere obligation." I shall say more about obligation-to in Section 5.5). I acknowledge that this claim may well initially appear quite implausible; 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 closely related claim that, for example, Steve has a prima facie obligation (even if not a prima facie obligation-to-Dave) to make amends to Dave. This seems 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 clearly doesn't have such a prima facie obligation; any scenario in which Steve is incapable of making amends to Dave would have this implication. (Steve might 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 gives rise to a residual obligation to make amends, even when one is not incapacitated in the relevant way. Suppose that Steve had a prima facie obligation to visit Dave on Monday, grounded not in any commitment to Dave but simply in beneficence. Had he then helped Carl en route and thereby rendered himself unable to visit Dave on Monday, certainly no explanation or 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" (as opposed to the "Remainders Thesis"), in Gowans (1994), pp. 91-2. 54 Cf. Zimmerman (1993c) on the possibility of the "resolution" without "dissolution" of ambivalence. 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 perhaps the condition that gave rise to the desirability of Steve's visiting Dave on Monday 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, that gives rise to the desirability of some further action? If so, then it is likely that some residual prima facie obligation will have emerged. But not necessarily; the agent might not have the relevant capacity. Indeed, it may not be only the original agent who incurs such an obligation. If Steve was walking with Willis when he came upon Carl, then perhaps, while Steve was busy tending to Carl, Willis had the obligation to call Dave and let him know what was happening.57 Here an obligation emerges from Steve's breaking his promise to Dave, but it is Willis's, not Steve's. In general, then, we may say that, if residual obligations emerge from the failure to satisfy prima facie obligations, this is perfectly consistent with the account of prima facie obligation presented here. Whether or not they do, however, is a complicated business, and it is certainly no defect in the account that it fails to imply that they do. Finally, it should be noted that residual obligations can arise not just from the failure to satisfy prima facie obligations. They can arise also where an overall obligation is not satisfied (whether or not this involves the failure to satisfy a prima facie obligation). As discussed in Section 3.2, often, when an overall obligation is not satisfied, a shift in obligations occurs, and a new overall obligation emerges. Thus Mandy became obligated to babysit Bill's child, and Victor became obligated to visit Bert. Also, as discussed in Section 4.4, often, when an n-level overall obligation is not satisfied, an tt+1-level overall obligation emerges. Thus you became obligated to vote for Dick, and Alan became obligated to attend the meeting on the second floor. These are all naturally classified as residual obligations; just like Steve's residual obligation to apologize to Dave, they have emerged from the 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. Wesley Newcomb Hohfeld made a lasting contribution to our understanding of 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 two are 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 liberty] 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 obligations. This issue has already been briefly addressed in Section 1.3. There the following four propositions (as formulated by Ross) were distinguished: (1.4) (1.5) (1.6) (1.7)

a right of A against B implies a duty of B to A; a duty of B to A implies a right of A against B; a right of A against B implies a duty of A to B; 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 follows: (1.4') if S' has a claim at T in W against S that S do A at T', then S has an obligation 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 a claim at T in W against 5 that S do A at T'. Combined, these statements yield the thesis that claims and obligations are in a certain way correlative to one another. The following points should be noted. 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 liberty at T in W as regards S' not to do A at T'. Second, the correlativity of claims and obligations is compatible with the normative priority of claims over obligations. It seems plausible to say that obligations derive from claims, rather than vice versa. For example, it 58 Hohfeld (1919), p. 35ff. Hohfeld was explicitly concerned with giving a taxonomy of legalrights,but his account is plausibly extended to moralrightsalso. Also, Hohfeld tended to prefer the term "privilege" to the term "liberty," although he acknowledged the appropriateness of the latter (at Hohfeld, 1919, pp. 42-3). Most commentators appear to prefer "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 it involves, 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 possibility that there be posthumous rights and also with the possibility that there be rights held by persons not yet born or even conceived. In such cases it would have to be true that S' has a claim at T even though S' does not exist at T. I doubt whether this is indeed possible, but I don't wish to rule it out here.59 Note that allowing for this possibility does not require allowing for the possibility that there be post- or prevital obligations. Clearly it is impossible for someone who has existed but will not exist to have obligations, for such a person no longer has any choice among accessible worlds. I would say, too, that it is impossible for a not-yet-existing person to have obligations (even remote ones), for I think that there being worlds accessible at some 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, introduced in Section 1.3, between mere obligation and obligation-to. S's having an obligation-to-S' is not the same as S's having an obligation towards or concerning S\ Whether or not S has an obligation-to-5' to do A is a question of whether or not S owes it to S' to do A; whether or not S has an obligation concerning S' is a question of whether or not the obligatory action, A, itself somehow involves S'. Often, of course, the two will coincide. If Pru has promised Zoe to take her to the zoo, then (let us assume) Pru has an obligation-to-Zoe to take Zoe to the zoo. Here, Pru's obligation-toZoe also concerns Zoe. But sometimes the one to whom an obligation is owed is not the one whom the obligation concerns. If Pru has promised Ollie to take Zoe to the zoo, then Pru's obligation still concerns Zoe, but it is an obligation she owes to Ollie, not Zoe.60 Fifth, (1.4') and (1.5') serve to emphasize a related distinction, that between mere claims and claims-against. For all that I have said so far, it may be that someone can have a claim concerning something that is not a claim-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 obliga59 60 61

Cf. Feinberg (1980), pp. 173-6 and 180-2. Cf. Hart (1979), p. 18. It is sometimes asserted that we can, and do, have claims against entities that are not themselves individual persons. If there is to be correlativity in this sort of case, then it must be that these entities can and do have obligations. This is an issue that I shall take up in Chapter 9.

178

tions generally; they assert that there is a correlativity between claimsagainst and obligations-to in particular. Objections that have been raised against the correlativity thesis would appear to succeed, if they succeed at all, only in impugning the general thesis; they leave the more particular thesis untouched. For example, David Lyons notes: It is often said...that one should be charitable and generous — that one has a duty to behave in that way. This appears to make good sense, even though no other person could have a right to one's charity or generosity (for if the other had a right, then what would be required of one would not be classifiable as charity or generosity). Of course, it is controversial whether or not one has an obligation to display charity or generosity. As noted in Section 1.3, it is arguable that all obligation is obligation-to, and thus it might be argued that, if no one has a claim to charity, then no one has an obligation to display charity. This, however, serves only to confirm the correlativity of claims-against and obligations-to. If there is an obligation to display charity but no claim to charity, then the obligation simply isn't an obligation owed to anyone (or anything)63 in particular, and so the more particular correlativity thesis is not 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 urging upon the world community the moral principle that all basic human needs ought to be recognized as claims.. .worthy of sympathy and serious consideration right now, even though, in many cases, they cannot plausibly be treated. . .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, is simply that, even if we accept the propriety of talking of basic needs as constituting claims in those cases where no one is in a position, and hence no 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 correlativity thesis is not threatened. There is one sort of case, however, which I think poses a more serious challenge 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. For instance, Kant is famous for asserting that there is a duty not to commit suicide and a duty to develop one's talents.66 Of course, this might simply be interpreted as an assertion that we have duties concerning ourselves, in which case they would not pose any challenge to (1.5'). And, even when so weakly interpreted, some would dismiss the assertion on the grounds that morality has to do only with how we treat other people. This seems mistaken, though. Considerations concerning self-respect would seem to indicate that 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 saying that some of our duties concern ourselves. It would seem, rather, that we owe it to ourselves to do certain things for ourselves. Yet it also seems odd to say that we have claims against ourselves. This is the challenge to (1.5').68 Other than dismissing self-regarding duties altogether or simply relegating them to the category of duties that merely concern the agent, there are 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 against ourselves; it is just that we are not likely to press or stand on these claims if they appear to be in danger of being flouted. I am not sure what to make of this response. A second response is to try to plumb a middle ground between mere obligation concerning oneself and obligation-to-oneself. According to this response, self-regarding obligations, though not strictly owed to oneself, nonetheless have an essential connection to oneself in that they are grounded in considerations of self-respect; thus they do not merely concern oneself by accident, as it were. I find the second response promising, although I concede that, unelaborated, it remains pretty obscure. It is not my intention to try to clarify it here, 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 that is (1.5') conjoined with the stipulation that 5 is distinct from S'. ((1.4') will still 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 to be so construed. This means that rights are themselves prima facie and so, like the obligations to which they are correlative, subject to being overridden.69 It means also that all those distinctions made earlier with respect to prima facie obligations — concerning whether they are direct or indirect, in principle or in practice, unconditional or conditional, and whether they are permanent, universal, or necessary - apply also to rights. Consider the (alleged, and suitably qualified) right of a promisee against a promiser to the latter's fidelity. Corresponding, as it supposedly does, to a direct prima facie obligation, it will be a direct right. As such, it can and will ground all sorts of incidental rights; for example, if the promiser must drive to Boston to ensure that he manifests fidelity, then the promisee has the right that the promiser drive to Boston. Also, as was noted earlier, the prima facie obligation of fidelity is a conditional one, the condition being that a commitment is made; if and when the commitment is made, then the obligation becomes unconditional (at least in practice, if not in principle). So too, then, a (potential) promisee's right is conditional on the promise being made; if and when it is made, the right becomes unconditional. Finally, if a (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 necessarily constitute overall wrongdoing. If the right is merely prima facie and one fails to satisfy it because one satisfies instead another right that overrides it, then one does no overall wrong (unless there is a yet more stringent right that one has failed to satisfy). Useful terminology to adopt here is Judith Thomson's:70 we may say that, if a right (that is, a claim) is not satisfied, then it is infringed; if such infringement is overall wrong, then the right is violated. We may also say that, if a right has been violated, then the violator has not only done (overall) wrong, he has done a wrong; equivalently, he has wronged the right-holder. Some philosophers, however, conceive of rights as correlative not to merely prima facie obligations but to overall obligations. This is to conceive of 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 corresponds to 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 absolute right of the lowest degree - call it a grade-1 right -just in case someone owes him a certain overall obligation; he has a grade-2 right just in case someone always owes him such an obligation; he has a grade-3 right just in case everyone always owes him such an obligation; there is a grade-4 right just in case everyone of a certain sort is such that everyone always owes him such an obligation; there is a grade-5 right just in case everyone is such that everyone owes him such an obligation; and there is a grade-6 right just in case it is necessarily the case that everyone is such that everyone always owes him such an obligation. More precisely: (5.30) S' has a grade-1 absolute claim at T in W against S that 5 do A at T 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 that S 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 done iff 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 done iff 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 iff for all worlds W, there is a grade-5 absolute claim in W that A be done. 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 can come 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 simple characteristic — innocence, for example — or a relatively complex characteristic — innocence-cum-parenthood, say.) In addition, each of the foregoing statements is supposed to be adequate to an account of so-called negative rights (where, to be more explicit, reference to doing A would be replaced by reference to not doing A). The foregoing account of unconditional absolute rights can be straightforwardly 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 at T', on the condition that p is true iff S has an overall obligation at T 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 do A 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 (truthfully or not; presumably not) that there is a grade-6 conditional absolute right to fidelity. Now, certainly no proponent of rights will quarrel with the claim that some moral rights are grade-1 absolute. For even those who regard rights as correlative to prima facie obligations will accept that sometimes the obligation in question is not merely prima facie but overall. But problems arise 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 grade 1 or higher). Let us examine these issues. To claim that a moral right is grade-2 absolute (or higher) is to claim that its unoverridden status holds on more than one occasion. The greater the degree, the greater the generality of this status. What is problematic is that, given the vast variety of considerations that may come into conflict with that consideration which the right in question represents, the greater the generality of "unoverriddenness" that one seeks to establish in the right, the less likely it seems that one will succeed in establishing it. Still, some have claimed success in such an endeavor. Alan Gewirth, for example, has argued 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 overridden in any circumstances."71 Here he appears to be contending that the right in question is at least grade-4 absolute. (He claims that there are other such rights, too, grounded in the general principle that there is an absolute prohibition against degrading persons.)721 shall not dispute what Gewirth says 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, for example, is likely to wish to grant such an absolute status to the right to be compensated for injury, or to the right that others not damage one's property, and so on. In general, then, it would appear that the greater the degree (that is, the higher the grade) of absoluteness, the fewer absolute rights there are. But this claim must now be qualified; for it can be argued that there are just as many absolute rights as there are prima facie rights. One can argue this by employing a rather obvious technical move (one that Gewirth himself repudiates, though for reasons specific to his own particular agenda of identifying certain rights as being especially significant),73 and that is to build into the description of the content of arighta list of conditions under which it would be overridden if its content were not qualified in this way. For example, it is often said that the right to life (understood as the right not to be killed) is, though particularly weighty, not absolute (in a sense of "absolute" that would seem to indicate at least grade 5, since the right is referred to simply as "the" right to life, with no restriction to particular persons or times), inasmuch as it is sometimes overall justifiable to kill people. One response is to "downgrade" the right by restricting it to all innocent people (thus understanding it to be a grade-4 absolute right, with C interpreted as innocence). But this response is frequently rejected on the basis that it is sometimes overall justifiable to kill innocent people. To this one might respond that it is never overall justifiable to kill innocent people unless a certain (probably disjunctive) set of conditions Q through Cn obtains74 and then declare that there is a grade-4 absolute conditional right that innocent persons have not to be killed, where the condition in question is that this set does not obtain. (Or, indeed, one might "upgrade" the right back to its original status and claim that there is a grade-5 absolute conditional right that all persons have not to be killed, where the condition in 71 72 73 74

Gewirth (1981), p. 2; the right is asserted on p. 8 and defended in ensuing pages. Gewirth (1981), p. 16. Gewirth (1981), p. 5. Cf. Donagan (1977), p. 72.

184

question is that the set of Ct through Cn does not obtain and the persons in question are innocent.) Indeed, this tactic of redescribing the content of rights could be used to argue that there are in fact no prima facie rights at all; rather, all rights are absolute, all (or certainly most) having this sort of extensive, complicated, conditional content. But why think of rights in this way? The most prominent proponent of the view that all rights are absolute is 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 but what he calls "valid claims." A claim is valid, he says, when "the balance of reasons" favors "recognition" of the claim.76 Elsewhere, he says that a valid claim is "a decisive case, invulnerable or conclusive." 77 It seems clear that Feinberg here takes rights to be claims that are not overridden; for this reason the obligations that they impose are not overridden either. Feinberg reaches this conclusion through an examination of how legal rights function. He says: "Having a claim".. .is an expression very much like the legal phrase "having a prima facie case." A plaintiff establishes a prima facie case for the defendant's liability when he establishes grounds that will be sufficient for liability unless outweighed 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) the same as having a right to X, but is rather having a case of at least minimal plausibility that one has a right to X, a case that does establish a right, not to X, but to a fair hearing and consideration. Claims, so conceived, differ in degree: some are stronger than others. Rights, on the other hand, do not differ in degree; no one right is more of a right than another. Now, I think that Feinberg may well be correct in what he says here, as long as we restrict his account to legal rights. Suppose that Paul and Paula have parted; their divorce was bitter, and now each is laying claim to a certain indivisible piece of property. A judge rules that the item belongs to Paula. 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 commit him to the view that rights are not merely prima facie, would appear to be inconsistent with what he says elsewhere (Feinberg, 1980, p. 225ff.) when he denies that all rights are absolute and endorses Thomson's distinction between the violation of a right and the mere 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 so everyone 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 prima facie but absolute, we may not automatically infer from this (as he appears to do)80 that moral rights are similarly not prima facie but absolute. Notice the implication. It would mean that our everyday discussions about moral rights are to a very great extent ill formed. As noted earlier, we could not say, without qualification, that someone who has been injured has a moral right to compensation; we could not say that a property-owner has a moral right that others not damage his property; and so on. We could not say these things because it can happen, as we all know, that it is sometimes overall morally right for an injurer not to give compensation, for someone to damage another's property, and so on, due to overriding considerations. For the same reason, many current debates would have to be reformulated. If a fetus has a right not to be killed, and this is a (presumably at least grade-4) absolute right, then no abortion is ever overall permissible; appeals to the rights of the mother would be misguided, because either mistaken or irrelevant. If a psychiatric patient, who has just expressed to his therapist a clear intention to commit murder, has a right to confidentiality, and this is an absolute right, then no breach of such confidence is ever overall permissible; appeals to the rights of his intended victim would be misguided, because either mistaken or irrelevant. But we don't normally find such appeals misguided, and that is because we don't normally take moral rights to be absolute. It is this implication of treating all moral rights as absolute — the implication that our everyday discussions of rights would need to be radically revised — that seems to me the best reason to reject such treatment; for there appears to be no compensating advantage. Others have given other reasons. For example, it has been argued that the denial that certain rights are prima facie (such as the right of property-owners that others not make unauthorized use of their property) cannot account for the obligation that others have to compensate for certain actions (such as the overall justified, but nonetheless 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 Feinberg says that 5's having a (legal) claim to X presents a "prima facie case" that S has a right to X, he is not saying that this imposes a prima facie obligation on others not to interfere with S's use of X. He is, rather, using "prima facie" in a purely evidentiary sense; he is saying, that is, that S's having a (legal) claim to X provides some reason (which is in principle 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 certainly seek to explain the (alleged)82 obligation of compensation without resorting to the notion of the justifiable infringement of a right.83 The difficulty in 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 rights are absolute. However the significance of this difficulty should not be exaggerated, for there is a complementary difficulty that besets the view that moral rights are prima facie, and that is the difficulty of specifying exactly what the conditions are under which whatever moral rights there may be are overridden. (Exactly when, for example, does fidelity trump gratitude? Exactly when does gratitude trump fidelity?) Still, it may be that these difficulties are not completely parallel, for this reason. Suppose that rights are prima facie and that the difficulties in specifying exactly, for all rights, the conditions 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 still make explicit reference to such rights; we can name them, discuss them, make certain determinations of their relative weights under certain circumstances, and so on. But on the view that all moral rights are absolute, if we cannot fully specify their content, then we cannot make explicit reference to them at all. It is hard to see how, under such circumstances, we could hope to justify our moral judgments by an appeal to rights. Yet we commonly attempt to do just this. In sum, we simply do not take all moral rights to be absolute, and I know of 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 obligation? Here I have no precise account to offer. If I had given a precise account of prima facie obligation-to, then, given the correlativity of prima facie obligations-to and rights, I could have derived a precise account of rights (or claim-rights and liberty-rights, at least). But I have not given such an account of obligation-to; the account that I have offered in this chapter is an account of mere prima facie obligation, and I confess myself unable to say 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 obligations pertain in that case. Nonetheless, it is of course possible that an accounting of the pertinent rights is of great help in determining what ought to be done. Rights, we commonly think, are (frequently) especially weighty considerations. I have already acknowledged that overall obligation may (on occasion) be epistemologically or explanatorily dependent on prima facie obligation; that is, that determining what prima facie obligations pertain is (on occasion) helpful in, and perhaps even required for, determining what overall obligations there are. If certain rights are correlative to certain particularly weighty (that is, difficult to override) prima facie obligations, then it clearly can help, when trying to determine what we ought overall to do, to determine whether these rights are at issue. Still, for all that, the hope that a determination of the pertinent rights will by itself provide for a quick, uncomplicated determination of what we ought overall to do seems to me in many cases 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 deontically best if he accepted the invitation and submitted his review on time. The trouble 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 because he has taken on a great many other tasks, but the fact is that this task, though accomplishable, would remain unaccomplished. It would be deontically better for him to decline the invitation than to accept it but fail to submit the review on time; for then, at least, someone else could be invited to do it, someone less qualified but someone who would get the job done. What ought (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 overall obligation implies: Rick ought to accept the invitation. But a number of philosophers believe that this is the wrong answer; they believe that Rick ought to decline the invitation. This is not because they are unsympathetic with the general idea that one ought to do one's deontic best; on the contrary, they are impressed by the fact that Rick would do better if he declined the invitation than if he accepted it, and it is for this reason that they believe he ought to decline it. This issue can be couched in terms of two broad doctrines, which have come to be called "actualism" and "possibilism."2 What the actualist says is roughly this: (6.1) S ought (overall) to do A iff what would happen if S did A is deontically superior to what would happen if S performed any alternative action; 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 deontically superior to what could happen if 5 performed any alternative action, where "could" expresses what in Chapter 2 I have called "personal possibility."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 be just one way that things would or could turn out if 5 did A. Still, they should suffice for present purposes. Note the following, however. My use of "would" in the statement of actualism presupposes that, if 5 does A at T2 and an event E occurs at T3 as a consequence of this, then it was true at T? that E would occur at T3 as a consequence of 5's doing A at T2. This is so even if the probability (whether objective or otherwise) at Tt of E's occurring at T3 as a consequence of 5's doing A at T2 was less than 1. In saying this, I don't mean to deny that 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 there is a sense of "would" according to which it is appropriate to say this, and that it is this sense that I'm employing here. This fact about how "would" is to be construed here also affects how "could" is to be construed, for the two are intimately related. (See Subsections 2.2.2 and 2.2.3.) Suppose that S can at T, do A at T2, that there is a probability of .1 at Tj that Ej will occur at T3 as a consequence of 5's doing A at T2, and that there is a probability of .9 at 7\ that E2 will occur at T3 as a consequence of 5's doing A at T2. Suppose that, despite these probabilities, E1 would (in my sense) occur and E2 would not occur at T3 if 5 did A at T2. Then Et could (in my sense) happen and E2 could not happen (unless 5 could make E2 happen by some other means). For more on the relevance of probability 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, as far as I know, only recently. At first actualism was favored;4 then possibilism 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 Rick declines the invitation, then he takes Track 2. Clearly this is not the best he can do, since Track 1 is better, and so anyone who is working within the tradition that one ought to do one's deontic best (as the actualist avowedly is) must surely accept that Rick does wrong in declining the invitation. And yet the actualist says that it is not wrong for Rick to decline the invitation, indeed that Rick ought to decline it. Isn't this plainly unacceptable? The actualist has a response here. He will point out that, according to him, Rick does do wrong in not accepting the invitation. He doesn't do wrong with respect to not-accepting as such, but he does do wrong with respect to not-(accepting-and-submitting). Why? Because, even though it is true that accepting would have been deontically inferior to its alternative, nonetheless accepting-and-submitting would have been deontically superior to any of its alternatives. But this response comes at a considerable cost, for it highlights the fact that 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 actualism, Rick is obligated to accept-and-submit but is not obligated to accept, even though he can of course not-accept. Some actualists have explicitly and boldly embraced this implication. For example, Frank Jackson and Robert 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 that were he to stop smoking he would compensate by eating more."7 Notice that (2.40'd) is implied by the following (where <} signifies logical 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 cannot both 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 case that 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, the actualist 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 notaccept; 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 line of argument (the issue will be discussed further in the next chapter), the point to be made here is that the actualist does not reject (2.47) because he accepts the possibility of moral dilemmas. Suppose that what would prompt Rick not to submit the review on time, were he to accept the invitation to do so, is sheer laziness; he just "couldn't" (more accurately: wouldnX) be bothered. Even in this case the actualist would say that Rick ought to notaccept. 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 situation to portray it as a moral dilemma, for the actualist must admit that Rick could do all that he ought (namely, accept-and-submit, the doing of which would 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, most objectionable about the actualist's view: it sanctions wholly avoidable wrongdoing. Let me not be misunderstood here. I do not mean that actualism sanctions doing what the possibilist calls wrong (although this is true on occasion). I mean that the actualist sanctions doing what the actualist calls wrong. True, there is no flat-out inconsistency to the effect that in one and 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) (in fact, O(A)), even though S cannot avoid doing wrong if he does A but can avoid doing wrong if he doesn't do A" The case of Rick is just such an occasion, for the actualist says that Rick ought to not-accept, even though Rick cannot avoid doing wrong if he does not-accept — in that, if he does not-accept, he cannot avoid not-(accepting-and-submitting), which is wrong by the actualist's own lights — but can avoid doing wrong if he doesn't not-accept (but rather accepts-and-submits). Why should Rick's obligations be thought to be thus tailored to his own easily avoidable moral failings? Of course, Rick's failure to submit the review on time might (in some other version of the case) be due not to laziness but rather to his having taken on a great many other tasks about which he is understandably concerned, and here one might be inclined to accept what the actualist says about Rick's obligations. This would be a mistake, though, for reasons that I shall discuss in Section 6.4. I have just claimed that no acceptable account of obligation sanctions wholly avoidable wrongdoing, where this is understood to consist in the conjunction of these propositions: S may, or ought, to do A; S cannot do A without doing 23; it is wrong for S to do B; and S can act in a way that includes 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 this criticism (first presented in Zimmerman, 1990b) "begs the question, since the issue between 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. But here one might be tempted to invoke an argument of Derek Parfit's to the effect that it can sometimes be right to cause oneself to do wrong. Parfit gives a case of the following sort.11 Patty is a politician, and Enid is her enemy. Enid tells Patty, truthfully, that she will kill all Patty's children if and only if Patty refuses to be filmed performing some obscene act, the film to 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 to the press, thereby ruining Patty's career). On the assumption that Patty cannot save her children except by submitting to being filmed performing the obscene act in question, Parfit claims that Patty ought to do this, even though she knows that this will cause her to act wrongly later (by committing the crimes), since she knows that she will choose saving her career over forgoing the crimes. Hence not only would it be right for Patty to cause herself 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 case do we have a sanctioning of wrongdoing of the sort that I am claiming to be unacceptable. First, we can understand "cause" in such a way that a cause'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 "submit to being filmed" and C for "committing the later crimes"). Here the submitting renders the later commission of the crimes inevitable. On this understanding of the situation, the later commission is not wrong; hence, contrary to Parfit's description of the situation, there is no wrongdoing caused. The second way to construe the situation - which is presumably Parfit's way - is to understand "cause" in such a way that a cause's effects are avoidable, even when the cause has already occurred (but the effects haven't). In this case, the situation may be represented as in Chart 6.3. Here, indeed, the submitting is obligatory but the commission wrong. But this does not constitute a sanctioning of wrongdoing in the foregoing sense. For although it is being said that Patty ought to submit and that this would cause 1: S & C ... 2: ~S & ~C ... 3: ~S & C ... C h a r t 6.2 11 Parfit (1984), p. 38.

194

1: S & 2: S & 3: ~S & 4: ~S &

~C C ~C C

... ... ... ...

Chart 6.3 her to commit the crimes, which would be wrong, the fact is that it is open to her to do what it is said she ought (namely, submit) while refraining from doing the wrong in question. 6.2.3 A nondiachronic case Recall from Section 4.3 the case of Charlie, captured in Chart 4.4, where the actions in question, whichever track Charlie chooses, would be simultaneous. Suppose that, whether or not Charlie accelerates, he's going to change lanes. Then actualism implies that it is not the case that he ought to not-accelerate, even though it is the case that he ought to not-change-andnot-accelerate. (Possibilism, of course, implies that he ought to not-accelerate.) Thus, as Jackson and Pargetter acknowledge,12 the diachronic nature of the reviewing case is not essential to the present discussion. Indeed, the case of Charlie brings out equally well the fact that actualism implies the falsity 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 significantly different from that with which we have been working so far. It is this:13 (6.1') 5 ought to do A iff what would happen if S did A is deontically superior to what would happen if S did not do A. This version is no more successful than the first. Like (6.1), (6.1') also implies that the theses just discussed are false, for it yields the same prescriptions as (6.1) for both Rick and Charlie. And like (6.1), (6.1') also sanctions wholly avoidable wrongdoing. Indeed, (6.1') seems to me even more problematic than (6.1), in that the former yields some counterintuitive prescriptions 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 respect to which we may suppose the following to be true: if you were to vote for Tom, that would be better than if you were to perform any alternative action (that is, better than if you were to vote for Dick, or Harry, or no one). It follows that it is not the case that, if you were to vote for Dick, or Harry, or no one, that would be better than if you were to perform any alternative action (for it wouldn't be better than voting for Tom). But now note this. It is consistent with the foregoing to suppose not only that, if you were to vote for Tom, that would be better than if you did not votefor Tom, but also that, if you were to vote for Dick, that would be better than if you did 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 you ought to vote for Tom, the latter (but not the former) also implies that you ought 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. I argue: 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 to actualism may well respond: if p then q; p; hence q. This is not an uncommon situation in philosophy, and in this case, as in others, an indisputable resolution of it is probably not forthcoming. But there is room for some maneuvering. For instance, the actualist might try to bolster his view by pointing out that, although it implies the falsity of the theses just discussed, it is nonetheless consistent with certain closely related theses. (For example, 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). (See Subsection 2.3.1 for a discussion of such implications.) It also implies that wrongdoing is 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 rejection of the original theses, it is surely a tack one is tempted to take only if some good reason can be given for rejecting possibilism, which implies the truth of the original theses without qualification. Here, though, the actualist thinks that he can provide such reason; at least, Jackson and Pargetter think that they can. I turn now to a consideration of certain arguments that they 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 to accept the invitation, for although it is true that accepting is part of the deontically best course of action open to him, it is also true that it is part of the deontically worst course of action open to him. Surely, then, there must be more to say before the question of what Rick ought to do is settled. In themselves the fact that accepting is part of the best and the fact that it is part of the worst show nothing. The first points to the answer that Rick ought to accept, but the second points equally to the answer that he ought not to accept. Possibilism errs in allowing the first fact to settle the matter.17 This is a puzzling objection. It rests on the assumption that acceptance's being part both of what's deontically best and of what's deontically worst points equally to its being obligatory and to its being wrong. No reason has been given for this claim, and the possibilist of course rejects it. For while the possibilist accepts the claim that, where a conjunction of acts (A&B) is obligatory, then so is each conjunct (see (2.40'd) for a more precise rendition 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, to think that an act, B, cannot feature in each conjunction. Consider an analogy. Suppose that a state of affairs, S, is part both of the actual 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 [it says] that one ought to do the best one can..., [it disregards] the fact that the best one can do may also be the very worst one can do... When that happens, the possibilist must advocate 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 persuasive 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 5 obtains is settled. In themselves the fact that S is part of W\ and the fact that it 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 advise Rick to decline the invitation. Is she advising him to act immorally? Is she hoping for this? Clearly not. Her advice is the right advice, and this implies that what she advises him to do is right.19 This objection cannot be accepted. It would prove too much, for it could just as easily be applied to the voting case. Knowing your propensity for overlooking the best-qualified candidate, I might well advise you to vote for Dick (rather than Harry). This advice would be just as appropriate here 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 implicit reliance 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 whether or not to accept the invitation, but that nothing she can do would persuade him to submit the review on time. Thus, although Rick can of course so act that Rick submits the review on time, Addie cannot so act that Rick submits the review on time, and so the best that she can do does not match up with 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 ought to do the worst one can do" (to use Goble's phrase - see the last note). After all, if Rick would submit the review on time if he were to accept the invitation, then the actualist would agree that he ought to accept, even though accepting is also (part of) the worst he can 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 be taken up in some detail in Chapter 9.) Under the circumstances, then, Addie has a choice between advising Rick to accept the invitation (and thereby seeing to it that he does accept the invitation but does not submit the review on time) and advising Rick to decline the invitation (and thereby seeing to it that he does not accept the invitation and so, again, does not submit the review on time). We may assume that the latter course of action is deontically better. Here possibilism implies that Addie ought to advise Rick 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 in which 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 some action 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 that S' perform a wrong action, where this is understood to imply that the wrongness of the action forms part of the content of the advice and thus part of the objective of the advisor. After all, it is not Addie's objective (or hope), when advising Rick to decline the invitation, that Rick should act immorally; that is something which she correctly takes as given, that is, as beyond her remedy (though not 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 reject it too. For there is no necessity that what would happen if S' did A be deontically better than what would happen if S' acted otherwise, even though what would happen if S advised S' to do A would be deontically better than what would happen if 5 acted otherwise. Suppose that Rick always acts contrary 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 invitation, even though Rick ought to decline it. Similarly, both the possibilist and 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 both the possibilist and the actualist agree that Rick ought to accept-and-submit, but both must acknowledge that it might happen that Addie ought not to advise Rick so to act, in that such advice could easily backfire and thus be deontically undesirable. There is therefore no support for actualism to be found in considerations concerning advice. 6.3.3 Detachment The third objection that Jackson and Pargetter raise is this. They claim that factual detachment for conditional obligation is valid; that is, that the unconditional obligation to perform an action may be inferred from the conditional obligation to perform it when the condition in question is satisfied. Thus, given the fact that Rick ought to decline the invitation if he won't submit the review on time and the fact that he won't submit the review on time, it follows that he ought to decline it (period).21 In response, we need only note that, as remarked in Chapter 4, factual detachment for conditional obligation is not valid. Consider the voting case once again, as represented in Chart 4.1. You ought to vote for Dick if you don't vote for Tom. Suppose that you won't vote for Tom. It doesn't follow that you ought to vote for Dick. On the contrary, you still ought to vote for Tom. Now, elsewhere, Jackson and Pargetter recognize this very point.22 But then they try to save their original thesis by drawing a distinction between ways in which obligation may be subject to conditions. First there is what they call "hypothetical obligation." A statement of hypothetical 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 hypothetical obligation is obligation that involves a subjunctive conditional and can 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, factual detachment is valid for statements of hypothetical obligation. But then there is what Jackson and Pargetter call "restrictive obligation."23 A statement 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 have called "conditional obligation proper." As Jackson and Pargetter note, and as I have noted repeatedly, factual detachment is not valid for statements of such obligation. Making this perfectly correct distinction helps Jackson and Pargetter, though, only if it can be shown that Rick's obligation to decline the invitation, if he won't submit the review on time, is a hypothetical rather than a restrictive one. This is of course what they contend, but their rationale is problematic, for three reasons. Consider, first, the voting case. There is an inconsistency in Jackson's and Pargetter's position with respect to this case. They declare the obligation that you ought to vote for Dick, if you don't vote for Tom, to be a restrictive one. This seems absolutely right. But then they say that, since (factual) detachment for restrictive obligation is not valid, it is not the case that you ought to vote for Dick. Yet we have seen that, on their version of actualism (captured in (6.1')), you ought to vote for Dick (since your doing so would be better than your not doing so). Second, Jackson's and Pargetter's position overlooks the apparent isomorphism between the voting case, where it is granted by all concerned that restrictive obligation is at issue, and the reviewing case. The mark of restrictive obligation, as Jackson and Pargetter rightly note, is that it concerns obligation relative to a restricted set of options. You ought to vote for Tom, given all your options; but, with the option of voting for Tom discounted, you ought to vote for Dick. Similarly, Rick ought to accept the invitation, given all his options; but, with the option of submitting the review 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 to vote for Dick. Similarly, Rick ought to submit the review on time; if he won't do this, then he ought at least to (have the courtesy to) decline the invitation. Jackson and Pargetter in fact admit that the propriety of the locution "at least" signals restrictive obligation. 5 Since it is proper to use it in the reviewing case, again we find a tension in their position. Third, their position is undermined by their own characterization of hypothetical obligation. They say that what follows "if," in cases of hypothetical obligation, involves not a restriction of options but a reordering of them. 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: ~A & ~S 2: A & ~S 3: A & S Chart 6.4 As noted in Section 4.1, this is tantamount to saying that, if Tom were a crook, then your options would not be ranked as in Chart 4.1 but in some other way, such as in Chart 4.2. But notice how different this is from the reviewing case. In saying that Rick ought to decline the invitation if he won't submit the review on time, we are surely not saying that, since it is true that he won't submit the review on time, his options are to be reordered in some such way as in Chart 6.4. I think, then, that we simply must accept that Rick's obligation to decline the invitation, if he won't submit the review on time, is a restrictive one — that is, a conditional obligation proper. Since factual detachment is acknowledged by all concerned to be invalid for such obligation, we cannot argue that it is such detachment that warrants the claim that the actualist wants to make, namely, that Rick ought to decline the invitation. Recall, however, that in saying that factual detachment is not valid for conditional obligation proper, I am saying that we cannot derive a primary unconditional 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 vote for 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 unconditional obligation to vote for Dick. Similarly, if Rick won't submit the review on time, then he ought to decline the invitation; if he doesn't even do that, he compounds his wrongdoing by violating a secondary unconditional obligation to decline the invitation. Jackson and Pargetter claim that there 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 his obligations. For, they say, he ought to decline (even though he ought also to accept-and-submit). But my possibilist account also allows me to say that Rick meets some but not all of his obligations in declining the invitation. He meets the secondary obligation to decline; he fails to meet the primary obligation 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 is this. 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 invitation to review it, for then the invitation would not be extended to anyone else. That is, under the circumstances, his accepting the invitation would 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 be individuated coarsely and thus that Rick's accepting the invitation would be 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 the invitation is not identical with his delaying the review but is related to it in some less intimate way (expressed by the term "by" in "Rick delays the review by accepting the invitation"), or (6.9) is false, since, even if Rick's accepting the invitation is his delaying the review, this identity holds only contingently, and he ought not to allow it to hold. That is, the possibilist, working under the assumption that acts are to be individuated coarsely, would accept that (6.10) if S ought to do A, and he ought to allow his doing A also to be his doing B, then S ought to do B, but he would have no reason to accept (6.9). (Recall here the discussion in Subsection 2.2.5 concerning Jones, who helped the old lady, as he ought to have done, but who in doing so startled her, as he ought not to have done. Someone who believes that Jones's helping the old lady and his startling her are identical must reject (6.9) and resort to some such device as saying that one and the same act can be obligatory under one description but not under another.) 6.4 WHAT IT IS BEST TO DO While it seems to me that the foregoing should suffice to put actualism to rest, 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 is only 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 taken on a great many other tasks or for some other reason would find it very difficult to do so, even though, strictly, he could do so? Isn't it plausible to say what the actualist says in this case and declare that Rick is not obligated to accept the invitation? Consider the medicine case first presented in Section 4.2. There are two medicines, A and B, that Audrey can administer. Let us suppose that, if she administers A on Monday and A on Tuesday, then the patient will be fully cured; if she administers B on Monday and B on Tuesday, the patient will be pretty well off, but not as well off as on the first course of action. Any other course of action would be disastrous for the patient. It seems that possibilism requires Audrey to administer A on Monday. But what if we now say that, if Audrey were to administer A on Monday, she would not administer it on Tuesday, and this not because oflaziness but for some other wholly understandable reason? For instance, the second dose of medicine A may be very hard for Audrey to lay her hands on, although strictly she could do so (it's in a safe, and she doesn't know the combination).29 Or perhaps it wouldn't be difficult for Audrey to do this, but it's rather that she doesn't know that two doses of A will cure the patient; on the contrary, she has very good reason to believe that a second dose ofA would be fatal, although she knows that with two doses of B the patient will be pretty well off. Under these circumstances, don't we want to agree with the actualist that Audrey ought not to administer A on Monday but should give the patient B instead? Or consider, finally, this case.30 Gary is playing a game. If he wins, great good fortune will abound; if he loses, it will be disastrous; if the game ends in a draw, there will be no significant consequences. He has in his hand a token, which he can place in one of two boxes. The game consists simply in the following. If he places the token in thefirstbox, then he either wins or loses; if he places it in the second box, it's a draw. What determines whether he wins or loses is whether he places the token in the first box with his 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, which with disaster. Ought Gary to place the token in the first box? Possibilism appears to imply that he ought. But don't we want to say, with the actualist, that whether or not Gary ought to do this depends on which hand he would use and what would happen as a result? 29 30

See Goble (1993), pp. 140-1. I accept that Audrey can, strictly speaking, lay her hands on the medicine. See the discussion in Subsection 2.2.3. 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 decline the invitation (when he would fail to submit the review on time due to his having so many other tasks to attend to), that Audrey ought to administer medicine B on Monday (when she would have great difficulty in laying her hands on a second dose ofA on Tuesday, or when she would fail to administer A on Tuesday in the belief that it would be disastrous to do so), or that Gary ought not to place the token in the first box (if he would do so with his right hand, say, and then disaster would ensue). After all, there is in principle a distinction between what it would be wrong for someone to do and what he would be blameworthy for doing, and while I readily grant that the agents in these examples would not be blameworthy for failing to accept the invitation, to administer A on Monday, and to place the token in the first box (with the left hand), respectively, it does not follow that they would do no wrong in failing to do these things.31 Still, it is also not clear to me that it is incorrect to say that the agents ought to decline, administer B, and put the token in the second box, respectively. But does this pose any problem for possibilism? I cannot see that it does. It was noted in Section 1.4 that the present theory is not committed to any particular account of what deontic value consists in. It is perfectly consistent with the present theory to say that, in the relevant sense of "best," it would be best for Rick to decline, best for Audrey to administer B on Monday, and best for Gary to place the token in 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 agent ought to do in these cases, while of course rejecting the reason for this diagnosis that the actualist himself gives. But here the actualist may well protest.32 His point is not just that Rick ought to decline the invitation, that Audrey ought to administer medicine B 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 to administer-A-on-Monday-and-administer-A-on-Tuesday, and Gary ought to put-the-token-in-the-first-box-with-his-left-hand. Thus we do have an objection to possibilism after all, in that in each case we have a 31 32

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. 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 also grant that he ought to accept-and-submit? (Of course, actualism may well yield this judgment, but actualism is unacceptable. It implies that Rick ought to decline even when he would fail to submit the review on time out of sheer laziness.) The reason why we are tempted to say that Rick ought to decline, where his failure to submit the review on time would be due to his having so many other tasks to attend to, is, I suggest, because of the great difficulty 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 ought to accept-and-submit. (This would be a matter of supererogation, an issue to be discussed further in Chapter 8.) Or again, the great risk involved in Audrey's case may lead us to say that it is not the case that she ought to administer A on Monday; but this should lead us also to say that it is not the case that she ought to administer-^4-on-Monday-and-administer-^4-onTuesday. So too with Gary's putting-the-token-in-the-first-box-withhis-left-hand. The risk, we may say, is too great — either (as in Audrey's case) 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 the complex action — and for that reason the complex action in question is not obligatory.33 In sum, actualism has unacceptable implications, whereas possibilism does not. It is clear which is preferable. 33 Goble's reason for saying that Rick ought to accept-and-submit, that Audrey ought to administer-^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 Goble indicates, but this is not determinative of what is obligatory, or not. Regret is appropriate in many cases where no obligation is violated (I can quite appropriately regret someone's having died in an earthquake); and (overall) regret may be inappropriate even where 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 objectionable when the subject is moral dilemmas. A dispassionate inquiry into moral dilemmas may appear especially inappropriate — their very nature demands a passionate response — and those metaethicists who deny that such dilemmas can even arise may seem to have reserved for themselves a unique place among out-of-touch academics. To which I reply (as politely as possible): the criticism is misdirected. No one is denying that life can present people with choices that are excruciating, both morally and personally. No one is denying that anyone who is confronted with such a choice, whether as participant or simply as spectator, should be deeply moved by it. The question is: just what is at stake in such choices? What is the nature of the situation in which such choices arise? To shun such inquiry is to court confusion, and that is what is objectionable. Moral dilemmas, as typically construed, are conflicts of moral obligation, but not all conflicts of obligation are dilemmas. A conflict of obligation involves there being two or more obligations which cannot all be satisfied, although each can be. In earlier chapters I have distinguished two main species of moral obligation- overall obligation and prima facie obligation — and from this emerge three main types of conflict: overall versus overall, overall versus prima facie, and prima facie versus prima facie. Let us consider the last type of conflict first. Clearly, conflicts between prima facie obligations are possible. I began Chapter 5 with a story about one such conflict: Steve had a prima facie obligation to render first aid to Carl and also had a prima facie obligation to 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 beneficence required rendering the aid, but it was fidelity that required keeping the date. The obligations stemmed from different considerations, and such considerations may conflict. Even when conflicts between prima facie obligations arise within one and the same "category" of obligation, the considerations that conflict are distinct from one another. While gratitude required both that Gertrude send Lucy lilies and that she send Chrissie chrysanthemums, it was, more particularly, gratitude-to-Lucy that required the first and gratitude-to-Chrissie that required the second. There is nothing particularly mysterious about such conflicts of obligation, nor about conflicts between prima facie and overall obligations. (After all, typically, and perhaps always,2 the most stringent prima facie obligation will be an overall obligation.) Clearly not all such conflicts merit the label "dilemma," however. This label should surely be reserved for occasions when the choice between conflicting obligations is a hard one; and not all conflicts of prima facie obligations (whether or not some of these are also overall obligations) need involve hard choices. Still, some clearly can. There are two principal ways in which a choice between conflicting prima facie obligations might be hard. It might be hard to tell which prima facie obligation is the most stringent, since more than one might appear equally compelling. Or it might be hard to make the choice for that prima facie obligation which is readily recognized as most stringent, since the moral cost involved in failing to satisfy some opposing prima facie obligation (or obligations) is so great.3 In both types of case, the choice may be an especially wrenching one, and I have no quarrel with calling such situations dilemmas. Nonetheless, it is not such situations that are the focus of inquiry in this chapter. Rather, I shall be looking into cases of conflict between overall obligations. Such cases, too, appear to warrant the label "dilemma," since they imply that overall wrongdoing cannot be avoided. I shall henceforth reserve the label "dilemma" for cases where overall wrongdoing is inevitable, and it is only into dilemmas in this sense that I shall be inquiring. 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 what it 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 agent with a hard choice, but it doesn't involve inevitable wrongdoing; for the agent may luckily make the right choice, despite his ignorance. Similarly, dilemmas (in the present sense) don't arise simply in virtue of there being a tie for what it is deontically best for the agent to do. Clearly, some cases of such ties can be wrenching (Ruth Barcan Marcus describes what seems to be such a case, where there are identical twins whose lives are in jeopardy and only one can be saved),6 but they don't imply inevitable wrongdoing. On the contrary, in cases of ties for what is deontically best, as the present 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 bears repeating: such cases do not constitute dilemmas, in the present sense. This bears repeating because at least one major recent work on dilemmas defines dilemmas in just this way. Walter Sinnott-Armstrong says that a dilemma arises when there is a conflict of two or more requirements (for all intents and purposes, a conflict of prima facie obligations), none of which is overridden.8 If there is an unoverridden prima facie obligation to do some act A, then A is, as (5.18) implies, featured in some world accessible to the agent to which there is none that is deontically superior. But this does not imply that A is overall obligatory, merely that it is overall permissible; hence it does not imply that refraining from doing A is overall wrong. SinnottArmstrong grants this.10 Hence what he calls a moral dilemma is not a moral dilemma in the present sense. I noted in Subsection 2.2.6 that the present account of overall obligation rules in the possibility of certain dilemmas that I there called prohibition dilemmas, as long as it is possible that each world accessible to the agent be such that some other accessible world is deontically superior to it. In such a situation the agent cannot avoid doing something wrong. But I there distinguished prohibition dilemmas from obligation dilemmas, which consist in conflicts of overall obligation. Like prohibition dilemmas, obligation 5 6 7

See note 55 to Ch. 2. Marcus (1987), p. 192. Sinnott-Armstrong distances himself somewhat from what Ross says about prima facie obligations, but nothing he says about requirements is incompatible with 'what I have said 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 differ in that, whereas obligation dilemmas arise from a surfeit of obligations, prohibition dilemmas arise from a dearth of obligations. Prohibition dilemmas involve inevitable overall wrongdoing because they involve the possibility that it would be overall wrong for 5 to do A even though there is no act B that S ought overall to do instead; obligation dilemmas involve inevitable overall wrongdoing because they involve the possibility that it would be overall wrong for S to do A even though 5 ought overall to do A. As I indicated in Subsection 2.2.6, the present account of overall obligation rules out the possibility of obligation dilemmas, barring certain complications to be addressed in Section 7.4. It is this issue that is the focus of this chapter. Henceforth, when I write of "dilemmas" without qualification, I shall mean obligation dilemmas exclusively. What renders (obligation) dilemmas especially problematic is not just that they involve inevitable overall wrongdoing (after all, prohibition dilemmas do too), but that they involve a certain moral inconsistency. The inconsistency consists in the implication that one and the same act can be both 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 Y for 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, and so 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 dilemmas must attend. I shall argue that such tension is in fact not acceptable and hence 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 between T3 and T4 is left open. ((7.7) of course concerns dilemmas that involve only two actions that cannot both be performed, although basic dilemmas can be envisaged that involve an indefinitely larger number of actions that cannot all be performed.) It is basic dilemmas that appear to have been at the center of much recent debate on moral dilemmas.

7.2 THE CASE AGAINST BASIC M O R A L DILEMMAS One argument against the possibility of basic moral dilemmas is simply this: my account of overall obligation implies that they are impossible. That this is so is easily seen. If clauses (a) and (b) of (7.7) are both true, then the deontically best worlds accessible to S at Tt are, on my account, worlds in which Sboth 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 reject the possibility of basic moral dilemmas. One reason is this: our own moral experience indicates that we do not take ourselves to be, or to have been, in dilemmas in situations where many people tend to diagnose a dilemma. I have in mind such cases as those that proponents of dilemmas have (or at least appear to have - more on this in the next section) themselves diagnosed as dilemmatic. A case presented by Jean-Paul Sartre is probably the most celebrated of all.11 It is the case of a student 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 he ought overall to choose it, and yet he cannot choose both. Now certainly such 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 such a case? Speaking for both of us, I am sure that we would seek advice as to what 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 dilemma after all; for if the advice we seek is moral advice (that is, advice as to what we 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 point out to us; and this indicates that we believe that at least one of the options that we're faced with is not (or, at least, may well not be) overall obligatory after all. In similar fashion, those of us who have been faced with such choices frequently look back at the situation in question and wonder whether we made the right choice; such wondering again indicates that we believe that we were not (or, at least, may well not have been) in a genuine dilemma after all. Of course, one reply to this is simply that our beliefs in such cases are mistaken; another reply is that, while our beliefs in such cases are not mistaken, there are other cases where we do not, or would not, hold such beliefs. Perhaps so, but the present considerations surely indicate that basic dilemmas aren't as common or as widespread as many of their proponents appear to maintain. There are other, more forceful reasons for rejecting the possibility of basic moral dilemmas. One argument concerns these two theses, first presented 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 explicitly stated): 212

(7.9)

if 5 ought overall at T, to do both A at T3 and B at T4, then 5 can at 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 conjunction 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 the principle 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 more to say on the matter. But the principle of agglomeration deserves further consideration. Some proponents of dilemmas have explicitly rejected the principle of agglomeration. In every case that I know of, the point seems to be this: it may be reasonable for S to do A and reasonable for S to do B, but it doesn't follow that it is reasonable for 5 to do both; and this applies to moral and nonmoral reasons alike.12 For example, it may be reasonable (morally or nonmorally) for Alex to accept Elvis's invitation to dinner and reasonable for her to accept Irving's, but not reasonable for her to accept both. But we must ask what "reasonable" means here. There are two principal senses: (a) what reason requires, and (b) what reason permits;13 and there are two other principal senses: (1) what is overall reasonable, and (2) what is prima facie reasonable. 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 facie obligation; this explains how conflicts of prima facie obligation are possible. Steve has a beneficence-based reason to tend to Carl and a fidelitybased reason to dine with Dave, but there is no X such that Steve has an X-based reason both to tend to Carl and to dine with Dave (for he cannot do both these things).14 Similar considerations pertain to (b2). Consider (bl) next, where we now take this to concern what is overall morally 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) is true. It may well be overall permissible for Alex to accept Elvis's invitation and 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 best worlds accessible to Alex are worlds in which she accepts one or other invitation but not both.15 But the failure of agglomeration for categories (bl), (a2), and (b2) gives us no reason to think that it fails for category (al); that is, we have been given no reason by proponents of dilemmas to think that (2.40'b)/(7.8) is false. 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 is true. For, as Paul Pietroski notes,16 nondilemmatic cases seem clearly (barring the actualist's concerns, which I addressed in Section 6.2) to conform to the principle of agglomeration, and it is difficult to explain this other than by appealing to the principle itself.17 Of course, a proponent of dilemmas could 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) (2.43 /// ) (2.36') (2.37'a)

ifO(y4),thenC(/l); \£~C(A&~B) & C(~B), then, if O(A), then O(B); ifO(A),thenP(/l); if O{A)y then ~P(~A).

As noted in the last section, the first two of these jointly imply that, in cases of 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 of these 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 do A at T3.

The contrapositive of the second yields this proposition (when A is substituted 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 ought overall 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 preserve the possibility of dilemmas, then, one must reject at least one of the four theses that function as premises in this argument. Again, I shall not seek to defend (2.46) here; I have said all that I want to say in Chapter 3. As to (2.43'"): the strong intuitive plausibility of this proposition was noted in both Chapter 2 and Chapter 6; I have no more to add here. And (2.36') and (2.37'a) are surely extremely plausible. A proponent of basic dilemmas might nonetheless seek to undermine this second argument by noting the following: I have already acknowledged that it is possible for there to be the analogue of a basic dilemma when it is prima facie, and not overall, obligation that is at issue and also when it is permissibility (whether prima facie or overall), and not obligation, that is at issue; hence I must admit that the analogues of this second argument break down somewhere in these cases, and it may be that this would give us reason to think that the second argument itself likewise breaks down. Let us now examine this matter. Consider prima facie obligation first. I of course grant that the analogue of (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 at Tt not to do A at T3. Moreover, the account of prima facie obligation that I have given in Chapter 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 not accept: (7.17) if S may prima facie at Tt do A at T3, then it is not the case that S ought prima facie at Tt not to do A at T3. This is because prima facie obligations (and prima facie permissions) are, as noted above, parti-resultant rather than toti-resultant. Thus, for example, although Steve has a beneficence-based prima facie obligation (and prima facie permission) to tend to Carl, he also has afidelity-basedprima facie obligation not to tend to Carl but to have dinner with Dave; again, this explains how conflicts of prima facie obligations are possible. But I cannot see how this provides any support whatsoever for the proponent of basic dilemmas. 215

As to permissibility (whether prima facie or overall - let's focus on overall): 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, the analogue 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 may overall at Tt not do A at T3, and this is for the obvious reason that ties for what is deontically best are perfectly possible. Once again, though, I cannot see how this provides any support 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 permissible not to do, or, contrapositively, that what it is overall permissible to do it 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 obligation that I have provided. Perhaps one way to defend it stems from considerations of supervenience. Many philosophers will agree that, if an act is obligatory (whether prima facie or overall), as of a certain time, then this moral fact supervenes on a certain set of nonmoral facts that obtains at that time. This set of nonmoral facts grounds the obligatoriness of the act; it renders the act obligatory. Similarly, if refraining from an act is permissible (whether prima facie or overall), as of a certain time, then this moral fact also supervenes on a certain set of nonmoral facts that obtains at that time. This set of nonmoral facts grounds the permissibility of refraining from the act; it does not, therefore, render the act obligatory. Now, as has been repeatedly observed, it is of course quite possible that an act should be prima facie obligatory and yet refraining from it be prima fade permissible. This is because 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 obligatory, the other not. But an act cannot be overall obligatory while refraining from it is overall permissible, for such moral facts arefofr'-resultant;they are grounded in one and the same set of nonmoral facts (namely, the set of all the nonmoral facts18 that obtain at the time in question), and it cannot be 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, either because they are skeptical about the supervenience of moral on nonmoral facts or because they believe that it begs the question to say that, if a set of nonmoral facts grounds the permissibility of refraining from an act, then that same set cannot ground the obligatoriness of the act. If so, I do not know what else to say in defense of (2.37'a)/(7.14). Further defense of this claim seems to me unnecessary, though, for it is the kind of proposition that should serve as the touchstone of an ethical theory. If my account of overall obligation had contradicted it, then it is the account, not the claim, that would 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 well trodden. Almost all the considerations that I have raised against the possibility of basic dilemmas have been raised, many times, by others.19 But they deserve to be stated again, because, despite what I take to be the great strength of these considerations, it seems that nowadays proponents of moral dilemmas far outnumber their opponents. This is a puzzling phenomenon. Of course, it may be — may well be — that the appearance of disagreement 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 of dilemmas may, I think, in every case be taken to be opponents of what I have called "basic moral dilemmas." But what is it that proponents of dilemmas are proposing? Well, consider the article by E. J. Lemmon that instigated the recent debate on dilemmas. There Lemmon distinguishes various types of situations that he calls moral dilemmas, but none of them seems to be a basic moral dilemma. In particular, what he calls "the simplest variety of moral dilemma in the full sense" seems best understood as simply a conflict of prima facie obligations, for the example he gives is this: 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 she has been unfaithful. I ought to return the gun, since I promised to do so... And yet 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, evidently resolved by not returning the gun.20 It could be that Lemmon regards this situation as what I have called a basic moral dilemma, but his talk of resolution indicates otherwise. Conflicts of prima facie obligation are resolved when the most stringent prima facie obligation is acted on; no such resolution is available for conflicts of overall obligation. Of course, conflicts of overall obligation might be subject to some sort of nonmoral resolution (perhaps satisfying one of the obligations would serve the agent better than satisfying the other), but it is apparent that Lemmon is concerned with a moral resolution to the conflict that he illustrates. It is clear that other prominent proponents of dilemmas are also at times concerned only with conflicts of prima facie obligation, so that there is in reality no opposition on this score between them and opponents of dilemmas. For example, Bernard Williams allows for the (moral) solubility of at least some of the conflicts of obligation that he is concerned with, even if these are not "soluble without remainder,"21 that is, even if their resolution leaves a certain sort of moral residue. Bas van Fraassen talks of "what ought to be for one reason and what ought to be for another reason,"22 and this is strongly reminiscent of the/wfr'-resultant nature of prima facie obligation. Marcus declares that "it is a better fit with the moral facts that all dilemmas are real, even where the reasons for doing x outweigh, and in whatever degree, the reasons for doing y."23 And Christopher Gowans endorses an approach to dilemmas according to which "even though S ought to do A and S ought to do B, there is nonetheless some sense in which, in the final analysis, (say) S ought to do B [and not ^4]."24 All such talk is most readily understood in terms of conflicts of prima facie obligation and, as such, seems wholly unobjectionable (except that Marcus's claim is surely too liberal; as we saw earlier, not all conflicts of prima facie obligation warrant the label "dilemma"). Nonetheless, it is also clear that proponents of dilemmas are at times concerned with unresolvable conflicts of obligation, and this may seem to indicate a concern with basic moral dilemmas. But let us not be hasty. There can be situations that are (morally) unresolvable that don't constitute basic dilemmas. For example, if more than one course of action is overall 20 21 22 23 24

Lemmon (1987), p. 105. Williams (1987), p. 129. Van Fraassen (1987), p. 141. Marcus (1987), p. 193. Gowans (1987a), p. 25.

218

permissible, then there is no moral resolution to this conflict between alternatives.25 Even if one chooses to call such a situation "dilemmatic," it is certainly not a basic dilemma; for wrongdoing is not inevitable. Indeed, as noted earlier, it is just such a situation that Sinnott-Armstrong calls "dilemmatic," and so, although I am an opponent of dilemmas, I do not oppose his view. And it may be that others have adopted his view. For example, Williams at one point talks of "extreme cases of moral conflict, tragic cases," where the following may be said: The agonies that a man will experience after acting in full consciousness of such a situation are not to be traced to a persistent doubt that he may not have chosen the better thing; but, for instance, to a clear conviction that he has not done the better thing because there was no better thing to be done.26 So described, a tragic case appears to be "no more" than a particularly excruciating conflict between equally stringent prima facie obligations than which there is none more stringent. Of course, I accept the possibility that such cases may arise. Similarly, van Fraassen talks of conflicts "which cannot 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 concerned with proposing the possibility of basic moral dilemmas may be adduced simply from the fact that they frequently don't even bother to address the formidable arguments presented in the last section against this possibility. Such an attitude betrays a total lack of concern with the possibility of basic dilemmas. And even when attention is apparently paid to these arguments, we need to be careful what to infer from this. For example, Lemmon rejects the principle that "ought" implies "can," and yet, as I have said, it is clear that he is concerned with "prima facie ought." Of course, I believe that he is mistaken to reject the principle even when so understood, but the point here is that his apparent concern with the arguments presented in the last section is, again, merely apparent. Others also may appear to have addressed some of these arguments (Williams and Marcus both reject the principle of agglomeration;29 van Fraassen rejects the principle that "ought" implies "may"),30 but, again, it may be that their concern is not with these argu25 At least, none of the alternatives is preferable in terms of deontic value. There are other morally 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 after all be any conflict between their views and that of opponents of dilemmas. Still, proponents of dilemmas do sometimes make remarks which suggest that they accept the possibility of basic moral dilemmas, the possibility that an agent will inevitably do overall wrong. Williams talks of tragic cases as cases where the agent accepts that "he has done something that he ought not to have done, or not done something that he ought to have done;"31 Marcus talks of being "damned if you do or ... damned if you don't";32 and although such talk can be construed as advocating merely the possibility that an agent will do prima facie wrong, it seems wise to grant that it may be intended as an endorsement of the possibility that an agent will inevitably do overall wrong and to examine it as such. What, then, are the reasons that have (perhaps) been advanced by proponents of dilemmas for accepting the possibility of basic dilemmas? A number of reasons appear to have been given. I shall now look into some of 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 just obvious that basic dilemmas can occur. Cases such as Sartre's, that of Agamemnon at Aulis (cited by Williams),33 that of Sophie's choice34 (cited by 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 clear that what's at issue is a basic dilemma? How could this be, even from the proponents' point of view? For they of course grant that there can be conflicts 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 constitutes a basic dilemma rather than merely a very hard but resolvable case. Another reason for accepting the possibility of basic dilemmas that may have been proposed is that cases of symmetry imply that such dilemmas are possible. One such case is that provided by Marcus of the twins whose lives are in jeopardy. It is clear, however, that such cases do not imply that basic dilemmas are possible; they simply involve equally overall permissible, not equally overall obligatory, alternatives. Dilemmas have often been diagnosed in cases not of symmetry but of incommensurability.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, as we 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 that each is equally overall obligatory. It may be that such cases are or can be especially wrenching, in that something of incomparable value is inevitably forgone, 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 the twins, which would appear to involve two comparable alternatives, seems to 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 situations are, as Williams says, not "soluble without remainder," that is, that no matter what one does, there will be a moral cost involved. But a moral cost can be incurred without an overall obligation's being violated; when Steve broke his dinner date with Dave, he incurred such a cost. Michael Stocker, however, suggests that in certain cases, cases where one is wrongly coerced by another to help implement some evil plan, there is a special reason to talk of "dirty hands" and that, in these cases, wrongdoing is inevitable.36 An example would be Williams's case of Jim and the Indians, cited in Section 2.1.37 But once again, while such cases may be especially wrenching, there is no need to regard them as involving inevitable overall wrongdoing. Obviously we may agree that it is usually wrong, overall, to implement evil plans, 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 consequentialist would presumably say. Perhaps he ought overall not to shoot the Indian, as the Kantian would presumably say. Perhaps there is both a strong reason to shoot and a strong reason not to shoot, as the Rossian would presumably say. It may not be clear to Jim just what he ought overall to do, in which case he is in a quandary. It may be that no option is deontically better than the other, in which case he is in a situation involving either symmetry or incommensurability. But none of this implies both that Jim ought overall to shoot and that he ought overall not to shoot. 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, it would 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 avoid conflicts of obligation of the sort that Marcus cites. After all, perhaps the only way to have avoided being in a position to save one but not both twins would have been to avoid the scene of the tragedy altogether, in which case both lives would have been lost. And even when one ought to avoid such conflicts of obligation, this can be explained without invoking basic dilemmas. For example, one ought (usually) to avoid making incompatible promises. Why? Not because they put one in a basic dilemma, but simply because they (usually) involve incurring unnecessary moral cost. It might be argued, however, that such cost can give rise to further overall obligations and that this is, on these occasions, a sign that one was in a basic dilemma. Perhaps one would have an overall obligation to explain to the twins' next of kin why not both were saved; perhaps one would have an overall obligation to apologize to, or even to compensate, the disappointed promisee. But it is clear, as the discussion of residual obligation in Subsection 5.4.4 attests, that such overall obligations can arise out of the failure to meet prima facie obligations that are not themselves overall obligations. There is no need to diagnose a basic dilemma in such cases. There is one type of moral residue in particular that is frequently cited as a sign of moral dilemmas, and that concerns emotions. As noted in Subsection 5.4.4, Ross believes that a certain feeling of compunction is called for when a prima facie obligation is not satisfied. But in some cases more than mere compunction is called for. Williams notes that it is sometimes appropriate to regret whatever it is that one has done, and he takes this as 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 it was 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 doing A 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 A and that it was overall obligatory for S to do A. We must surely grant the second premise. Sartre's student had cause for regret no matter which choice he made; Agamemnon likewise; Sophie too. But the first premise is false. Regret is appropriate whenever a moral cost is incurred, and this can happen without an overall moral obligation's being violated. This last point is widely accepted, but then proponents of dilemmas suggest this modification to the argument: replace talk of regret with talk of remorse or, equivalently, guilt feelings.40 Instead of (7.23) and (7.24) we would

then have, respectively: (7.25) it is appropriate for 5 to feel remorse for (guilty about) doing A [or not doing A] only if it was overall obligatory for S not to do A [or to do A]; (7.26) it is possible both that it would be appropriate for 5 to feel remorse for (guilty about) doing A and that it would be appropriate for S to feel 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 not obtain between regret and overall wrongdoing. Certainly being guilty is intimately related to overall wrongdoing; moreover, it seems plausible to say (although I shall express reservations about this shortly) that one feels guilty about, or remorse for, doing something only if one believes that one is guilty with respect to doing it. Given this, it seems plausible to say, along with van Fraassen, that "it is appropriate to feel guilt if and only if one is guilty";41 for then the belief that is part of the emotion will correspond with the 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 not doing A]. The question is, when is it that one is guilty with respect to doing [or not doing] 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 to satisfy 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 argument presently under consideration. But why, then, accept the second premise, (7.26) (once we are clear that it is remorse in particular, rather than some other negative emotion such as regret, that is at issue)? I know of no reason having been advanced for accepting it, other than to insist that basic dilemmas are possible, and to cite this as a reason would of course be to beg the question. (Actually, I'm not being quite honest here; I shall qualify this remark in the next section.) There is another answer to the question of when it is that one is guilty with 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 believed that he was failing to satisfy an overall obligation not to do A [or to do^]. 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 action and 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 warrant the claim that actual overall wrongdoing is necessary for actual guilt and thus necessary for appropriate remorse; on the contrary, presumed overall wrongdoing will suffice.43 Thus this, the strongest argument for basic dilemmas of which I am aware, fails. It is worth noting, finally, that the account of the appropriateness of remorse, (7.27), which I have provisionally conceded to the proponent of dilemmas, is in fact open to question. Suppose that Tracy, a truck driver, has just run over Reggie, a small boy who darted out into the street in front of Tracy's truck. Tracy did no wrong, let us assume; nor did she or does she believe that she did wrong. She is therefore not guilty, according to 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 supports only the "only if," and not the "if," of (7.29), since more than such a belief is required for culpability; freedom is also required. 43 Qualification: suffice along with other conditions, not among which is actual wrongdoing. See the last note.

224

we would all be a little put out if she showed no remorse for what she had done. True, there is, in a sense, nothing for her to feel guilty about, and yet wouldn't such a reaction bespeak a coolness of character that would likely prompt an absence of remorse in cases where it was clearly called for? I think that what we should say here is that Tracy's reaction is intrinsically appropriate but (probably) extrinsically inappropriate, inasmuch as it (probably) manifests a tendency to react in an intrinsically inappropriate way on other occasions.44 It seems, then, that (J.21) is at best acceptable only as an account of the conditions of the intrinsic appropriateness of remorse. But, even when it is so understood, there is reason to question it. This is because the claim concerning the connection between feeling guilty and being guilty, on which I originally based (7.27), is itself dubious. Robert Roberts has contended, with considerable plausibility, that feeling guilty doesn't presuppose believing oneself to be guilty so much as construing oneself as guilty.45 One example might be that of a lapsed Catholic, who believes the lapse to be justifiable and yet feels guilty about not attending Mass. Roberts likens such cases to optical illusions which are understood to be illusory and yet persist. Given such cases, it is arguable that it is sometimes appropriate, and not merely extrinsically so, for agents to feel guilty, even though they have not done overall wrong and did not believe themselves to have done overall wrong. If this is correct (and here I shall not seek to settle the question), then (7.27) is to be rejected.

7.4 N O N B A S I C M O R A L 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 dilemmas are so numerous and so vocal, it deserves to be covered anew. Let me now turn to some relatively fresh considerations, several of which are designed to undermine the proponents' case even further. I have so far been concerned with what I have called "basic moral dilemmas," 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 45 46

Cf. Zimmerman (1988a), pp. 134-5. Roberts (1988), pp. 183, 186-9, 195-6, and 201. 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 may in 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 moral dilemmas are not basic. Consider, for example, this sort of situation (where Tj 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 of binds have in fact already been given. For instance, recall from Subsection 3.2.2 the case of Mandy, who on Friday tells Phil that she will be able to help him move furniture on Sunday, thereby making a sort of informal commitment to do so, but who on Saturday promises Bill that she will babysit his child on Sunday, thereby making a formal commitment to do so. Her situation may be pictured as in Figure 3.2. When Mandy makes her promise to Bill on Saturday, she puts herself in a bind, for she can no longer avoid doing overall wrong on Sunday. If she keeps her promise as she ought, then she'll do the remote wrong of not helping Phil; if she helps Phil as she (once) ought, then she'll do the immediate wrong of not babysitting Bill's child. Here it is most important to note that the remote wrong, if it were committed, would be a genuine wrong, as I argued in Subsection 3.2.2. It seems to me perfectly appropriate to call binds a type of moral dilemma, for they involve wrongdoing that is, in a way, inevitable. It is true that the wrongdoing was not always inevitable, but it has become so (in Mandy's case, it became so on Saturday). It is also true that there is no single time at which the conflicting "oughts" apply (for then we would have a basic dilemma); one applies at one time (from Friday to Saturday, in 226

Mandy's case), the other at another (from Saturday to Sunday). Nonetheless overall wrong will have been done no matter what the agent chooses at the time of action to do, for by then it is too late to extricate himself from the dilemma. Consider now once again the argument concerning remorse, but with time indices explicitly provided as follows: (7.25')

(7.26')

it is appropriate for 5 to feel at T4 remorse for (guilty about) doing A [or not doing A] at T3 only if there is a time Tsuch that it was overall obligatory at Tfor 5 not to do A [or to do A] at T3; 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 be appropriate for S to feel at T4 remorse for (guilty about) not doing A at T3;

hence, (7.22") it is possible both that there be a time Tsuch that it was overall obligatory at T for S not to do A at T3 and that there be a time T such 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 construed as consisting in actual wrongdoing (see (7.28)). I think that we should also accept the second premise. Now I said earlier that I knew of no reason having been advanced to accept (7.26) other than to insist that basic dilemmas are possible, but I also said that I wasn't being quite honest in making this claim. I can now explain, for / now wish to advance a different reason for accepting (7.26'). Look at (7.22"). It doesn't say that basic dilemmas are possible. That basic dilemmas are possible is expressed by the following proposition: (7.22'") it is possible that there be a time Tsuch that it was overall obligatory at Tfor S not to do A at T3 and it was overall obligatory at Tfor 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 such identity is posited, and this is why it is acceptable. As long as binds are possible — and they are — we must accept (7.22"); for the following, which constitutes a bind, implies (7.22"): it was overall obligatory at Tt for S not to do A at T3 and it was overall obligatory at T2 for 5 to do A at T3. But this doesn'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 for her on Monday to feel remorse for not having helped Phil; and if she helped Phil on Sunday, then it would be appropriate for her on Monday to feel remorse for doing so, since she would thereby have failed to keep her promise to Bill. In this way I think we can concede to proponents of dilemmas their strongest argument and yet deny that basic dilemmas can occur. One objection to what I have just said is this.47 If it is to be appropriate for Mandy to feel remorse on Monday for what she did (whatever it was) on Sunday, then her belief concerning her guilt that renders her remorse appropriate must itself be accurate. Thus she must either believe on Monday that she was obligated from Friday to Saturday (but not beyond) to help Phil on Sunday, or believe that she was obligated from Saturday to Sunday to babysit Bill's child on Sunday, or both. But such time-specific beliefs concerning remote obligations are surely extremely rare, and hence it 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 is to 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 and that she ought to have babysat Bill's child, then her remorse for having failed 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 explicit time indices with respect to her obligations; at least, this is so as long as she doesn't attach incorrect time indices to her obligations. Thus it need not be at all rare that someone, who feels guilty about an action performed when he was in a moral bind, does so in an appropriate way. Binds are not the only type of nonbasic dilemma. Consider this situation (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 of Zimmerman (1987) on the occasion of its presentation at the 1984 meeting of the Eastern Division of the American Philosophical Association.

228

We may call this sort of situation a subsidiary dilemma. Such situations are possible; indeed, examples have already been given. Recall Alan, for instance, who oughtt to attend the 10:00 A.M. meeting on the first floor but fails to do so and, as a result, ought2 to attend another 10:00 A.M. meeting on the second floor. He can do each of these things, but he cannot do them both. 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 not such 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 can be distinguished. There are also variations on these types ("subsidiary binds," "interpersonal binds," "interpersonal subsidiary dilemmas," and so on). Moreover, the account of overall obligation that I have proposed, while repudiating the possibility of basic dilemmas, accommodates the possibility of at least some nonbasic dilemmas. There is an aspect to this accommodation at which I have only hinted but which I wish now to acknowledge explicitly. With respect to binds and subsidiary dilemmas, my account implies that such dilemmas can arise only by way of some wrongdoing. It thus gives detailed expression to the traditional Thomist distinction between a perplexity simpliciter and a perplexity secundum quid.4S The distinction may be broadly characterized in this way: a perplexity simpliciter arises if, through no wrongdoing on anyone's part, it happens that whatever can be done will be wrong; a perplexity secundum quid arises, on the other hand, if it happens that whatever can be done will be wrong, but this predicament has been induced by wrongdoing on someone's part. Alan Donagan, one of the few philosophers to have entertained this distinction in recent years, claims, as does Aquinas, that morality would be inconsistent if it admitted perplexities simpliciter but not if it admitted perplexities secundum quid; he claims further that morality is not inconsistent, and so does not admit perplexities simpliciter, although it does admit perplexities secundum quid.49 In this he echoes Georg von Wright, who puts the point pic48 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 only arise through antecedent sin... It is only as a consequence of a fall that a man can come to be in a predicament."50 Von Wright talks of the agent's situation before the "fall" as "prelapsarian." Let us say that (7.30) and (7.31) represent two different types of predicament into which agents may enter by way of a moral lapse.

It is no accident that the illustrations that have been given of nonbasic dilemmas involve such lapses. (No illustration has been given of an interpersonal dilemma, and you have no way yet to tell whether my account accommodates such dilemmas or whether, if it does, it does so only by virtue of some lapse. Again, I shall discuss this later, in Section 9.4.) The illustration of a bind has it that Mandy wrongly promises Bill to babysit his child. Such a lapse is required on my view, for only by way of such a lapse is it possible that what formerly was deontically best (helping Phil move furniture) 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) by making her promise to Bill, babysitting Bill's child would not have become the best thing for her to do; it would have remained second-best.51 But given the deviation onto a different path, helping Phil drops from first- to second-best. Hence she becomes obligated to babysit Bill's child; but hence, also, she is no longer obligated to help Phil. Her lapse does not generate 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 very much whether she helped him, or by Bill's child desperately needing to be babysat by Mandy in particular), so that its being obligatory from Saturday on for Mandy to babysit the 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 - the rendering second-best of her helping Phil - or she could not. If she could not, then her helping Phil simply never was the best that she could achieve, contrary to the initial hypothesis. If she could, then her failure to prevent the event either constituted a lapse on her part or it didn't. If it did not, then the best that she could achieve involved allowing her helping Phil to be rendered second-best; but in this case her helping Phil simply never 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 obligation. 52 Opponents of dilemmas have sometimes muddied the waters by claiming that perplexities secundum quid are possible but without acknowledging that such perplexities cannot take 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 by way of a lapse, because basic dilemmas are impossible, period, for the reasons given in Section 7.2. Cf. Feldman (1986), pp. 204-5.

230

It is clear, too, that subsidiary dilemmas can arise only by way of wrongdoing. Alan's obligation to attend the meeting on the second floor arises only out of his failure to attend the meeting on the first floor. Were he to do what he ought! to do, it never would be the case that he ought2 to do differently. I confess that here talk of "dilemma" is quite possibly misleading, for it seems to be at best an exaggeration to say that it is inevitable that Alan do overall wrong. While it is true that he cannot avoid doing wrong if he does what he ought2 to do, the fact is that he can avoid its being the case that he ought2 to go to the meeting on the second floor; he can avoid this by going to the meeting on the first floor, and in so doing he would do no overall wrong at all. In this respect, subsidiary "dilemmas" are both like and unlike binds. They are like binds in that the latter also do not involve wrongdoing that can in no way be avoided by the agent; if Mandy had not acted wrongly on Saturday by making her promise to Bill, she would have avoided her bind altogether. But subsidiary "dilemmas" are unlike binds in that the latter do involve there being a certain time (Saturday, in Mandy's case) at which wrongdoing has become inevitable. So perhaps conflicts of subsidiary obligations really should not be called a species of dilemma at all. I include discussion of them here for the simple reason that their form resembles that of other conflicts that are properly called dilemmatic (compare (7.31) with (7.7), (7.30), and (7.32)) and because they can arise only by way of a lapse.

231

8 Supererogation

Steve, introduced in Chapter 5, stopped to render first aid to Carl, who had had a car accident, thereby failing to keep his dinner date with Dave. This case was used, following W. D. Ross, as an illustration of one prima facie obligation overriding another and thereby constituting an overall obligation. But was Steve really overall obligated to do what he did? I am sure that many would say that he was not. On the contrary, they would say that he 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 in going beyond some prima facie duty or obligation. This is one way to interpret, for instance, the following passage by Joel Feinberg: A janitor has a duty to spend eight hours cleaning his employer's floors. He works ten hours for eight hours' pay. Duty required eight hours; he gave duty plus 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 obligation (of fidelity, perhaps) to work eight hours; in working ten hours, he goes beyond this duty and thus performs a supererogatory action. But even if that is what this passage suggests, it is not, I am sure, what most people have meant by the term "supererogatory," and it is not what I shall mean by 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, failing to do what one is overall obligated to do, and in neither case is one acting supererogatorily.

1

Feinberg (1968), p. 397.

232

8.1.2 Going beyond perfect obligation Some writers conceive of supererogation in terms of going beyond perfect overall duty or obligation. This leaves open the possibility that one nevertheless has an imperfect duty to do the act in question. In Subsection 2.1.4 several different senses of "(im)perfect duty" were distinguished, and some of these have been proposed as appropriate for the definition of supererogation. Kant claims that we have an imperfect duty of charity. One suggested interpretation of this was that, while we have an overall obligation to act charitably, we enjoy a certain leeway with respect to how to satisfy it, in that we may choose between displaying charity to A and displaying it to B and displaying it to C and so on. The duty is thus, in a certain sense, disjunctive (see (X) in Subsection 2.1.3). In this way, it might be said, when 1 act charitably towards a particular person, I am acting supererogatorily, since I have no overall obligation to do that particular act; I could have satisfied my overall obligation to display charity by acting charitably towards someone else entirely. This won't do. Suppose that I have an overall obligation either to act charitably towards A or to act charitably towards B or to act charitably towards C (there are no other candidate recipients of my charity), and that therefore none of these individual acts is such that I have an overall obligation 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 simply fulfill my duty of charity. Of course, if I act charitably towards all three, I might then do something supererogatory; but if so, this does not consist simply in the performance of a duty that is imperfect, in the present sense. Another suggested interpretation of "imperfect duty" was that the duty in question is a duty not so much to act in a certain way as to adopt a certain maxim for action, and it has been proposed that a supererogatory action is one that conforms with a maxim that satisfies such a duty.3 But while this may be the best that can be done by way of squaring Kant's theory with supererogation, it falls short of what is required for supererogation as this is normally understood; for it does not accord the supererogatory act the status of lying beyond duty altogether.4 2 3

Cf. Heyd (1982), pp. 121-2. 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 the action in question. 4 Cf. Baron (1987), p. 244.

233

Another suggested interpretation of "imperfect duty" was that of an obligation that is not an obligation-to, that is, of an obligation that is not correlative to some right. (See Subsection 5.5.1 for a discussion of this correlativity.) I think many writers have understood supererogation in terms of acts that are not owed to anyone in the way that an act to the performance of which someone has a claim-right is owed to that person. Feinberg's example of the janitor can certainly be construed in this way, as can his example (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 one duty 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) thus supersedes duty.6 But again this won't do, unless all overall obligations are obligations-to. For otherwise one may have an overall obligation to do something that goes beyond an obligation-to, and this can hardly rank as supererogatory. For example, even if a lifeguard who is "off duty" goes beyond any specific duty-to when saving a swimmer, whereas a lifeguard who is "on duty" does not, if the former nonetheless has an overall obligation to save the swimmer — if, that is, he would do overall wrong in not saving the swimmer — then he can hardly be said to be acting supererogatorily 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 no overall 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 the agent has no overall obligation not to perform it. Thus it must be neither obligatory nor wrong; it must be (not just personally optional but) morally optional. And it must also be especially valuable, in some way. After all, not all optional acts are supererogatory; some merely serve to fulfill a disjunctive obligation rather than to "go beyond" obligation. 5 6 7 8 9

Feinberg (1968), pp. 392-3. Actually, Feinberg would not call giving the stranger a match supererogatory, since it does not involve any great sacrifice or risk. More on this later. Kamm (1985), pp. 119-20. See New (1974), p. 179ff.; Heyd (1982), pp. 119-20. The two most extensive treatments of supererogation to date concur. See Heyd (1982), pp. 120, 125; Mellema (1991), p. 11 and passim. 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 recently remarked on. We may call it "suberogation."10 Whereas a supererogatory act goes beyond duty, a suberogatory act, we may say, falls short of decency. Perhaps Steve acted in a supererogatory manner when he stopped and helped Carl; let us now assume so. Let us assume, that is, that he would have done 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 misfortune? This would have been a nasty thing to do, but perhaps it wouldn't have been wrong. If it wouldn't have been wrong, it would have been suberogatory: optional but especially disvaluable, in some way. (Here and henceforth in this chapter, when I use the term "optional," I mean that which is morally optional.) How to render more precise the sense of "valuable" that pertains to supererogation (and that of "disvaluable" that pertains to suberogation) is a very tricky question that I shall address later. (Be forewarned, though: I will not attempt to resolve this question.) Nonetheless, this much should be said now: a supererogatory act is more valuable than any merely optional — that is, optional but not supererogatory — alternative; for a supererogatory act "goes beyond" what duty requires and, where there are moral options, 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 be misunderstood. In saying that supererogatory acts go beyond duty by being more valuable than any merely optional alternative, I am not saying that all supererogatory acts are more valuable than all nonsupererogatory (whether obligatory or merely optional or wrong) acts. Suppose that in Situation 1 John, a janitor, has two alternatives, that of working ten hours (act A) and that of working eight (act B); and let us suppose that A would be supererogatory 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 of working eleven hours (act D); and let us suppose that C would be obligatory and D therefore wrong. Whatever it is that makes A more valuable than B, it seems plausible to say that both C and D are therefore more valuable still, even though A is supererogatory while neither C nor D is supererogatory (indeed, the latter is wrong). The point is, however, that A is more valuable than any nonsupererogatory alternative, that is, more valuable 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 says the following: [I]f we contrast the man who performs an heroic act of duty with the man who performs a trifling act of supererogation,.. .we can say that the obligatory act of the 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 that does not acknowledge any sense in which the supererogatory "goes beyond" 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 and more valuable, in some way, than any nonsupererogatory alternative. Insofar as it involves optionality, supererogation is of course agent- and time-relative, in that an act's being optional for one agent at one time is no guarantee that it is optional for that same agent at another time, let alone for 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 for 5 at T2 to do A at T3; and it can also happen that it is not supererogatory for S at T1 to do A at T3 but supererogatory for S at T2 to do A at T3. As to the first possibility: perhaps at T2 S makes a morally optional promise to do A at T3, thereby rendering the act obligatory; or perhaps at T2 S makes an optional promise not to do A at T3, thereby rendering the act wrong. As to the second possibility: perhaps at T2 S commits an act that converts his doing A at T3 from obligatory or wrong to optional. (Such an act would itself 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 construed as act-tokens. As noted in Subsection 2.2.5, the claim that Jones ought to help that old lady on Monday is probably not to be understood as implying that there is a particular act-token that is obligatory; rather, it is best understood either as implying that a certain act-type (or some other abstract entity)14 is obligatory or as implying that a certain disjunction of act-tokens is obligatory. Similarly, the claim that it is optional, indeed supererogatory, for Jones to help that old lady on Monday is probably not to be under11 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 a certain act-type is optional (supererogatory), or as implying that a certain conjunction of act-tokens is such that each token is optional (supererogatory). Similar remarks pertain to the suberogatory. Insofar as the suberogatory is optional, it can happen that an act is suberogatory for an agent at one time but not at another. Furthermore, if "suberogatory" is to be applied to act-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 of supererogation and suberogation. Even if it is accurate as far as it goes, however, this of course does not imply that there are, or even can be, any supererogatory or suberogatory acts; still less does it imply that those acts that we normally call supererogatory are in fact so. The philosophical literature is full of reasons for thinking that supererogation is impossible. Many of the most influential normative theories seem to rule it out. It appears, for instance, to be inconsistent with certain strains of Christianity,16 with Kantianism,17 with utilitarianism and, more generally, consequentialism,18 and with Ross's theory;19 in every case the problem seems (roughly) to be either that, according to the theory, no action that is nonobligatory can be valuable in the requisite way, or that, according to the theory, no action that is valuable in the requisite way can be nonobligatory. In addition, there may be good reason to deny the possibility of suberogation even if that of supererogation is accepted. This is because there is a certain asymmetry between these concepts that might not at first be apparent.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 case that the greater the extent to which an act falls short of decency, the more suberogatory it is; on the contrary, its status as optional is thereby threat15 16 17 18

Cf. Mellema (1991), p. 29ff. Cf. Heyd (1982), p. 26ff.; Mellema (1991), Ch. 3. Pace Hill (1971). 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 be that this point is reached immediately, as it were, so that the status ofsuberogation is never attained. It seems to me not accidental that the most plausible of the relatively few examples of allegedly suberogatory acts that exist in the literature (including, if it is plausible, the example given earlier in this chapter of Steve's taunting Carl) are most readily understood as examples where the agent has managed, in some morally repugnant way, not to violate an obligation- to. For instance, Chisholm and Ernest Sosa talk of minor acts of discourtesy (such as taking too long in a restaurant when others are waiting) and of not-so-minor acts committed while competing with others (such as advertising in a way calculated to drive a competitor out of business), and in such cases they explicitly say that the act is permissible because the agent is acting "within his rights."22 Similarly, Julia Driver gives several illustrations of what she takes to be suberogatory actions (a single person takes a seat on a train, thereby preventing two other people from sitting together, although another seat is available; a man refuses to donate a kidney to his brother, although the latter will die without it), and she too explicitly talks of the agent as acting "within his rights."23 But if the realm of obligation is not exhausted by what there is an obligation-to to do, then these examples lose some of their force. Perhaps we can agree that, typically, one acts within one's (liberty-)rights, that is, that one violates no one else's (claim-)rights, when one is discourteous to another or "plays hardball" in competitive contexts or refuses to donate a bodily part, but this does not 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 is not optional, and hence not suberogatory. There would seem to be powerful reasons, therefore, for thinking that supererogation is impossible, and further, independent reasons for thinking that suberogation is impossible. However, determination of whether or not these moral phenomena are in fact impossible would require a detailed investigation of the theories to which I have just alluded (something that I shall not attempt) as well as a detailed account of the pertinent senses of "valuable" and "disvaluable" that apply to supererogation and suberogation (something that I have already said I shall not attempt). In the absence of such investigation, it would seem premature to rule out the possibility of supererogation and suberogation. Yet it may seem that this is just what my 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 precludes supererogation.24 If this belief is correct, then, unless cogent arguments can be given, independently of considerations about maximization, to support such preclusion (as arguments were given in Chapter 7 against the possibility of basic moral dilemmas), such preclusion may be taken as a sign of the inadequacy of the proposed account of obligation. But is the belief correct? 8.2.1 Supererogation and discretion Supererogatory acts, it has been said, are at once both optional and especially valuable. But a maximizing theory links value with obligation, and thereby hangs the problem. As Michael Slote puts it, when discussing consequentialism in particular: "[C]onsequentialism, precisely because it insists that one do the best one can, leaves no room for moral supererogation: barring 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 that maximization precludes supererogation altogether? It might be replied that, while a maximizing theory does admittedly imply that some acts can be optional, it doesn't allow for the possibility that acts be fully optional, whereas supererogatory acts are not just optional but fully optional. An act is fully optional only if it is not just not obligatory but also such that its performance does not constitute a way to fulfill some obligation; it is, as it were, wholly beyond the reaches of obligation. Now it is true that a maximizing theory is hardly likely to classify an optional act as fully optional. Consider a case that has the oversimplified form depicted in Figure 8.1. Here the agent has four and only four alternative courses of action, the first two equally valuable, the others less so. In this case, acts A and B are optional, whereas C is wrong. But neither A nor B is fully optional, 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 to do B, performance of either will fulfill an obligation. 24

Included in this number are some who accept and also some who reject a maximizing theory. See, for example, the following, all of whom appear essentially to hold the opinion 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. The minor point is this: a maximizing theory can in principle allow for the possibility 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 supererogatory acts must be fully optional. Consider the case in Figure 8.1. Why is it that it is obligatory to do either A or B? The answer is, of course, that any course of action involving neither A nor B is wrong. In doing either A or B, the agent is thus fulfilling the obligation not to do something else. It cannot be accepted 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 be obligatory), so that, if the supererogatory act is performed, that very performance, though of course not itself obligatory, will nonetheless constitute the fulfillment of an obligation, to wit, the obligation not to perform the alternative act. Why is it, then, that maximizing theories have been thought not to accommodate supererogation? Well, consider the classic maximizing theory of act utilitarianism, which is traditionally seen to be an amalgamation of hedonism (the view that all and only states of pleasure are intrinsically good and all and only states of pain are intrinsically bad) and consequentialism (the view that one ought to produce the most favorable balance of intrinsic goodness over intrinsic badness that one can). It is often said that act utilitarianism rules out supererogatory action,26 but this is questionable. What it certainly does do is imply that many of the acts that we would normally call supererogatory are in fact not so, since it implies that these acts are either obligatory or wrong. But this is not the same as implying that there are no supererogatory acts, let alone the same as implying that there can be no such acts. Indeed, the consequentialist aspect of act utilitarianism (which is what makes the latter a maximizing theory) rules in the possibility of optional acts (as long as ties in sum intrinsic value are in principle possible, and there seems to be no reason to deny this). This fact might suggest to some that it is act utilitarianism's hedonistic, rather than consequentialist, aspect that renders it inimical to supererogation, and that another version of consequentialism would accommodate supererogation more readily. But actually it is not that easy. If we turn to the nonhedonistic consequentialism of G. E. Moore, we find that Moore's theory, too, has a hard time accounting for supererogation. While this theory (which allies consequentialism with a theory ofvalue according to which not just states ofpleasure 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, maximizing theory - it seems not to square with our common intuitions concerning supererogation any better than act utilitarianism does. For example, just as act utilitarianism implies that many instances of self-sacrifice that we would 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 27

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. Moore (1965), p. 102.

241

only by way of extreme self-sacrifice, then Moore's theory implies that one must, one is obligated to, undertake such sacrifice.28 This observation applies equally well to the neoconsequentialist view of Fred Feldman, couched in terms of possible worlds rather than in terms of consequences,29 and it is this view that inspired the account of obligation that I have proposed.30 Must we therefore conclude that maximization does indeed preclude supererogation? Not yet. It could be that the problem lies with (neo)consequentialism in particular rather than maximizing theories in general. Although (neo)consequentialism allows for acts' being optional, its focus on the promotion of intrinsic value neglects the fact that it is in the area of personal interest 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 maximizing theory were to admit into the final determination of right and wrong certain agent-relative values? What if, in the final computation of what is best and what not, an agent's own projects and commitments were given a weight disproportionate to that which their intrinsic value alone supplies? This possibility has been explicitly admitted in the account of obligation that I have proposed; recall that in this account S's being obligated at T to perform some act is analyzed in terms of the value for S at T of accessible worlds in which he does and does not perform the act. Does this feature of the account provide the means of reconciling maximization with supererogation? 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 to think that supererogation essentially concerns agent-relative values. Some people might claim, for instance, that, although it is obligatory to refrain from killing babies whenever possible, it is supererogatory to refrain from killing fetuses. Whether or not this is correct, it is certainly intelligible, and no tie to agent-relative values is, or need be, involved. The much more important point is this: even though a maximizing theory that incorporates agent-relative values in the way indicated allows, as do all maximizing theories, for certain acts to be optional, it does not appear to allow for certain acts to be discretionary in the way that supererogation requires. Consider a case where A and B are each optional, according to the maximizing theory in question, and where their values (and for this reason the values of the top-ranked worlds that include them) are, respectively, predominantly agent-neutral and predominantly agent-relative. For example, A might be 28 29 30

Cf. Heyd (1982), pp. 77-8. Cf.Mellema(1991),p. 87. See Ch. 2, n. 7.

242

a charitable act of some sort (in which the agent has no vested personal interest), while B is the purchase of some high-priced compact-disk player (in which the agent has considerable vested personal interest). Now consider 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 accessible B-world), because the agent's vested personal interest is marginally increased. The maximizing theory would imply that in this modified case neither 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 maximizing theory would once again imply that neither A nor B is optional; on the contrary, A would be obligatory and B wrong. In such cases, engaging in the sort of self-sacrifice that A would entail is clearly not left to the agent's discretion.31 It is probably at this point that many will diagnose a fatalflawin all maximizing theories. It is not, as it may at first have seemed, that such theories cannot admit that certain acts are optional; clearly they can. It is rather that such theories cannot admit that certain acts are discretionary, in the sense that it is left to the agent's discretion whether or not to perform the act when its performance and nonperformance are of different values. This, it might be claimed, is something that is essential to supererogation, but something that maximizing theories cannot accommodate. Somewhat more precisely, and applied specifically to the account of obligation that I have proposed, the argument is this: (8.1) necessarily, for any act A, if the performance of A is supererogatory and its nonperformance merely optional, then all accessible ~Aworlds are less valuable than some accessible /1-world; (8.2) if my account of obligation is correct, then, necessarily, for any act A, 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, then it is not supererogatory; hence, (8.4) if my account of obligation is correct, then, necessarily, there is no act A such that the performance of A is supererogatory and its nonperformance 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 considerations 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 ofA is suberogatory and its nonperformance merely optional, then all accessible ^4-worlds are less valuable than some accessible ~;4-world; (8.6) if my account of obligation is correct, then, necessarily, for any act A, 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 it is not suberogatory; hence, (8.8) if my account of obligation is correct, then, necessarily, there is no act A such that the performance of A is suberogatory and its nonperformance merely optional.

8.2.2 The division of values There is but one possible way for me to resist the conclusions of these arguments while hanging on to maximization. Like any maximizer, I must endeavor both to have my cake and to eat it - that is, endeavor both to say that, where A is supererogatory, its performance is superior to its nonperformance and, where A is suberogatory, its performance is inferior to its nonperformance, and yet also to say that, since A is optional, its performance 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 to the moral evaluation of an act.34 One set of values must be said to be pertinent 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 whether or not to perform it, a maximizing theory must declare its performance to be nondeontically superior but deontically equivalent to its nonperformance; and where A is suberogatory, a maximizing theory must declare its performance to be nondeontically inferior but deontically equivalent to its nonperformance. In this way, the first two premises of each argument will be seen to be true, but only because of equivocation on the term "less valuable," 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 are superior to their nonsupererogatory alternatives. Simon seems to miss this in Simon (forthcoming).

244

arguments are to be declared unsound and their conclusions thereby resisted. This solution should come as no surprise. In previous chapters I stressed that my concern, when giving my account of obligation, was with the deontic superiority of one action, or world, over another. This was intended to leave open the possibility that one act, or world, be superior, even morally, to another in a nondeontic way.35 Let me now elaborate. Suppose that consequentialists are correct and that 'what determines whether or not an act is morally obligatory has to do with and only with how 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 that acts are open to a sort of moral evaluation that has nothing to do with their deontic status. For example, some consequentialist might say that an act is morally meritorious if and only if its agent intends by means of it to achieve what is intrinsically best. Since intended results need not match actual results, it would follow that an act can be morally obligatory without being morally meritorious, and vice versa. A morally meritorious act would be superior in a morally relevant way to a nonmeritorious one, but this superiority would be nondeontic. Note that, on this consequentialist view, there is a common core to an act's deontic and its nondeontic moral status. Both have to do with the promotion of intrinsic value. What distinguishes them is that the one has to do with the actual promotion of such value, the other with its intended promotion. A nonconsequentialist can similarly distinguish between an act's deontic and its nondeontic moral status. For example, someone might say that whether or not an act is morally obligatory has to do with and only with whether it actually respects people's rights. In this case, deontic superiority consists in superiority concerning such respect. Nonetheless, it is surely possible that a nonconsequentialist of this sort should accept that an act is morally meritorious if and only if its agent intends by means of it to respect people's rights. Since, as before, intended results need not match actual results, it would follow that an act can be morally obligatory without being morally meritorious, and vice versa. Also, as before, there would in this case be 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 that an 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 exploit this possibility in order to accommodate supererogation. Unless there are further restrictions on supererogation that I have so far overlooked (an issue to which I shall return in the next subsection), I therefore conclude that maximization does not preclude supererogation. Even if this conclusion is accepted, though, it may be thought that it comes at too high a price; for some may find such a division of values to be a deus ex machina with no rationale other than to save maximization. But this 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 he doesn't have to,36 what's apparently meant is that it is valuable in a way that doesn't render it obligatory. (Thus "ought" is here used in a nonbinding sense. See Section 1.2.) Why should this common appeal to a double standard37 be deemed unavailable to a maximizing theory? The problem with most maximizing theories is that they don't make such an appeal, not that they can't, and this is why they cannot (as they stand) account for supererogation. Second, and more importantly, there is no such thing as a nonmaximizing 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 in respect of value" (as Moore puts it)38 - that is, it does not in some way rank an obligatory act higher than its nonobligatory alternatives - in which case I 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 obligation 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 nonperformance, 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 suberogatory;

hence, (8.12) if either the performance of A or its nonperformance is either supererogatory or suberogatory, then they are deontically equivalent; 36 37 38 39

Cf. Feinberg (1968), pp. 392-3. Cf. also Chisholm (1968), p. 417. Cf. Attfield (1979). Cf. also Raz (1975). See note 37 to Ch. 1. This is not to say that every theory of obligation exactly fits the form of (I), as I acknowledged in Subsection 2.1.2.

246

hence, (8.13) any superiority (due to being supererogatory) or inferiority (due to being suberogatory) of the performance of A to its nonperformance, or vice versa, is nondeontic. And so, if the strategy of dividing values comes at too high a price, then supererogation and suberogation must simply be declared impossible. But is the price too high? I don't see why. Surely it is plausible to say that there is more to morality than the deontic. An act, or its agent, may be praiseworthy or blameworthy, virtuous or vicious, and so on,40 and only if such matters are directly reducible to the act's being obligatory, permissible, or wrong - which is surely not the case - would the act's deontic status exhaust its moral status. If this is so, then supererogation and suberogation can involve nondeontic matters such as these. The question is which. Well, this is the question that I said I would not try to resolve: precisely what senses of "valuable" and "disvaluable" are at issue in supererogation and suberogation. Many different answers have been proposed. Almost all accounts of supererogation claim that, if an act is supererogatory, then either it or its agent is in some way praiseworthy, but just what goes into praiseworthiness is controversial. Some claim that, if an act or agent is to be praiseworthy, the act must be unusually valuable, and hence that there are no trifling acts of supererogation;41 others allow for trifling acts of supererogation that are, because trifling, only minimally praiseworthy.42 Some philosophers claim that, for an act to be supererogatory, its omission must be such that either it or its agent is not blameworthy.43 Some philosophers claim that a supererogatory act, even though it is not obligatory, is such that it ought (in a nonbinding sense) to be done.44 Some philosophers 40

41 42 43

44

I would prefer to say that agents, rather than acts, are praiseworthy or blameworthy, virtuous or vicious, and so on, but this might seem to put some strain on my contention that there is a variety of ways to evaluate acts, morally. Recall, however, from Section 1.5 that, even when it comes to obligation, I am in fact not at all sure that it is best to categorize the evaluation in question as drt-evaluation. Perhaps it is better categorized as a matter of a^enf-evaluation, in which case the strategy of dividing values is under no strain at all. E.g., Feinberg (1968), p. 400. 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. 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 the act can at best be "quasi-supererogatory." See Mellema (1991), Ch. 5. 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 "recommended," in a quasi-obligatory way, by morality.45 Some philosophers claim that an act is supererogatory only if it has good consequences, or is intended to have good consequences.46 Some claim that an act is supererogatory only if it is performed for someone else's sake. And so on. A similar variety of conditions may be proposed for suberogation. The matter is clearly both complex and controversial. It is perhaps easier to envisage a division of morally relevant values when the nondeontic values are not said to supply a reason for acting, harder when they are said to supply such a reason. Suppose we agree that an act is supererogatory only if it or its agent is praiseworthy. On some accounts of praiseworthiness, an act or agent is rendered praiseworthy by the motive underlying the act; on some accounts of motivation, the motive underlying 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 an act. Thus some accounts can declare the performance of a supererogatory act to be wholly on a par with its nonperformance with respect to reasons for action; the nondeontic superiority of the performance will be explained without appeal to reasons for action. On the other hand, some accounts of supererogation will declare the superiority of the performance of a supererogatory act to constitute a special reason for performing it. If this is essential to supererogation, then, if supererogation is possible, there must be not only two sets of morally relevant values but also two sets of moral reasons for acting. But, while it may be difficult to see just how such an account should go, it is clearly not without precedent. On the contrary, the common invocation of what is ideal rather than obligatory seems, often at least, to constitute an appeal to a nondeontic set of reasons for acting.

8.2.3 Continuity It may seem that something has gone wrong in the foregoing defense of the compatibility of maximization with supererogation and suberogation, in that I have overlooked a condition of supererogation and suberogation which 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 Gregory Mellema characterizes as follows: "There is a common and continuous scale of value shared by supererogation and duty, and the difference between supererogation and duty is reflected in the degree to which the value is realized in each. The value is realized to a greater degree in supererogation than in obligation."50 How can the strategy of dividing values be reconciled with this talk of a common and continuous scale of value? Whether or not such reconciliation is possible depends on just what is involved in this condition of continuity that (allegedly) applies to supererogation. It may seem that it is simply being claimed that the value exhibited by supererogatory action, like that exhibited by obligatory action, is a morally relevant value. Clearly I agree with this. But this cannot be all that is meant; for, although this gives sense to the talk of there being a common scale ofvalue, it does not give sense to the talk of there being a continuous scale of value. It may seem that it is simply being claimed that all supererogation involves the sort of "oversubscription" (to use Feinberg's term) that is displayed by John when he works a supererogatory ten hours instead of the merely 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 supererogation that do not involve such oversubscription, as when, for instance, one gives a stranger a match.51 Second, the sort of continuity that cases of oversubscription 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 action being that of working). Perhaps what is being claimed is that, if A is supererogatory and some alternative B merely optional, then both A and B must exhibit a common type of morally relevant value, A exhibiting more of this value than B. This is compatible with what I have said (although I doubt that it is true); for even if B is merely optional, it may nonetheless exhibit a certain nondeontic 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, as some claim, then a trifling favor will not be supererogatory; nonetheless, as a favor it may be good in a certain way that is irrelevant to its deontic status 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, ifA is supererogatory and B merely optional, then both A and B owe their respective statuses wholly to a certain single type of morally relevant value that they both exhibit, their statuses being different solely due to the fact that the former exhibits more of 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 continuity (understood according to the latest interpretation) as follows: certain accounts of supererogation clearly include it, and thus, even if the condition hasn't been established as necessary for supererogation, still it is perfectly intelligible, and hence something must have gone wrong with the argument for (8.12) and (8.13). Perhaps the sort of account of obligation and supererogation that might be thought most obviously to include the condition of continuity (understood 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 be less than optimifically benevolent or beneficent, to give others less than one is in a position to give them, so long as one is sufficiently beneficent, that is, gives others 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 satisficing to avoid being wrong. Then, with respect to supererogation, Slote says this: [S]atisficing forms of.. .consequentialism.. .are consequentialistic because they treat overall consequences as the only thing relevant to act-evaluation and because, more particularly, they regard any act with overall consequences less good than those of some alternative as morally less good than that alternative. But they allow for supererogation and for the moral permissibility of less than optimal behavior.53 Note: "the only thing relevant." Even if such a satisficing theory is to be rejected for some reason, doesn't its form demonstrate that a theory can satisfy the condition of continuity (understood according to the latest interpretation), and hence that my contention, that supererogation mwsdnvolve a division of values, is false? Well, recall that my contention that supererogation requires a division of 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 such thing as a nonmaximizing theory of obligation. Perhaps the former claim will be granted, but what about the latter? What, in particular, about satisficing theories of obligation? There is of course a tension between the claim that obligation involves maximization and the claim that it involves satisficing; but it is only a surface tension. To resolve it, one need only recognize that I am saying that obligation involves maximization of some sort, and this is clearly compatible with its involving satisficing ofanother sort. In particular, if two acts, one of which optimizes a certain value and the other of which merely satisfices that value, are both equally permissible, then clearly they are to be ranked as deontic equals; their deontic status is not determined solely by the degree to which they realize the value in question. For instance, a theory might say that an act is permissible if and only if it displays beneficence to at least degree 5, supererogatory if and only if it displays beneficence to a degree greater than 5. This is intelligible, but what it implies is that it is not simply beneficence that is pertinent to the determination of right, wrong, and obligation. Rather, it is beneficence-to-at-least-degree-5 that is at issue; any acts that display this characteristic are deontically equivalent, whether or not one displays a greater degree of beneficence than another. And it is not beneficence-to-at-least-degree-5 that determines whether and to what extent an act goes beyond duty; on the contrary, it is beneficence-to-adegree-greater-than-5. Thus the condition of continuity (understood according to the latest interpretation) is not satisfied by this theory. Despite there being a common core (namely, beneficence) to that characteristic which determines an act's deontic status and that which determines its nondeontic status, the characteristics are distinct. (One would, of course, like a rationale for making the cutoff point between mere optionality and supererogation be beneficence to degree 5. One such rationale might be that other people's needs are satisfied when and only when they are treated with beneficence to at least degree 5, and that one is obligated to satisfy other people's needs but not obligated to do more than this. Another rationale might give a different story. Once the rationale, whatever it is, is given, it will be easier, I think, to see that on any theory that accommodates supererogation, even a satisficing theory, there is 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 just illustrated, then an act's deontic value would turn on whether it meets other people's needs, whereas its nondeontic value would turn on something else entirely.) 251

Perhaps this will have served to show how modest the strategy of dividing values really is. Drawing a distinction between deontic and nondeontic values merely reflects the fact that supererogatory acts and their merely optional alternatives are both alike (with respect to optionality) and unalike (with respect to being beyond duty), and, as such, it seems to me wholly beyond reproach. It should be noted, finally, that I have not been arguing that supererogation and suberogation are possible. For all that I have said, they may yet be impossible. Certainly not every theory of obligation can, as it stands, accommodate them. Those theories that deny any division of moral values cannot, and others that do allow for such division may nonetheless undermine themselves. For example, any theory that says that something's being good from the ideal point of view is itself a deontic reason for trying to achieve it would appear to have undercut any advantage that this division of values initially afforded it with respect to the accommodation of supererogation. Ross's theory seems to have just this difficulty. In distancing himself from consequentialism by divorcing the right from the good, Ross may at first seem to have proposed a theory that can account for supererogation in a way that Moore's theory (unaccompanied by any division of values) cannot. For the special prima facie duties of which Ross speaks that vie for the status of duty proper do not owe their deontic status to 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 this would appear to allow Ross to declare that the desirability of an act, or of what the act brings about, provides at most a nondeoritic reason for its performance. But he does not say this. Rather, he declares the desirability of an act, or of what it brings about, itself to be a deontic reason for performing 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's being supererogatory does (or would) constitute a moral reason to perform it, and perhaps no such reason can be nondeontic after all.56 Or perhaps, once it is clearly recognized that what is especially valuable about supererogation is nondeontic, any inclination to say that what is (apparently) 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 act supererogatory because they wish to ensure that the agent is not blamed for its nonperformance. But if an agent can avoid blame and yet do wrong, then there may be reason to doubt that no "supererogatory" act is morally wrong.)57 Whether or not any of this is so, I shall not seek here to determine. My aim has been merely to argue that maximization does not, as such, 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 of groups. Groups can achieve things that individuals alone cannot. This is an obvious fact that has so far been neglected in this book, but it is a fact that poses 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 claim is captured in (I). For the time being, let us work with the following simpler paraphrase of (I): (9.1) for any person 5 and times T and T", S ought at T to do at T" the best that he can at T do at T'. The problem that besets this account may be illustrated as follows.1 Suppose that there are two voters, Vincent and Virgil, and two candidates for election, Smith and Jones, and suppose that the voters' options may be ranked according to the matrix given in Chart 9.1. That is, it would be (deontically) best if both Vincent and Virgil voted for Smith; next best if one voted for Smith and the other abstained; next best if both voted for Jones; next best if one voted for Jones and the other abstained; and worst if either both abstained 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 of what the illustration presupposes. In particular, we must recognize that the numbers represent the values of certain outcomes achievable by the group of Vincent and Virgil, that is, achievable by them, jointly. It is thus being presupposed that there are certain worlds accessible to this group, and also 1

The problem has received extensive treatment in Regan (1980). It is anticipated in Gibbard (1965), Sobel (1968), and Barnes (1971). My illustration borrows from Regan (1980), p. 18, and Feldman (1986), p. 155.

254

Vincent

Vote for Smith Vote for Smith 10 0 Vote for Jones 7 Abstain

Virgil Vote for Jones 0 5 3

Abstain 7 3

0

Chart 9.1 that these worlds may be ranked according to their relative values. Let us look at each presupposition in turn. There can be no doubt that groups can sometimes achieve things that subgroups or individuals cannot. Perhaps a group of four people can lift a heavy 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 subgroup 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, the group itself can do? It may seem that we should reject this too. Perhaps onequarter of the assembled company can squeeze into the back of a Volkswagen, but the whole group cannot; perhaps Paul can marry Paula, whereas the group of Peter and Paul cannot; and so on. But these facts should not blind us to the truth that whatever is accessible to a proper subgroup or individual member of a group is also accessible to the group. If one-quarter of the assembled company can squeeze into the car, then certainly the group as a whole can so act that one-quarter of its number squeezes into it; if Paul can marry Paula, then certainly the group of Peter and Paul can so act that Paul marries Paula. In general, then, we should reject the first and accept the second of the following two claims (where "{S^,..., SH}" signifies the group of persons Sj through SM, and where an individual member S( of a group is, for the sake of convenience, treated as the subgroup {5,}): (9.2) if Wf is accessible to {Su ..., Sn} from W at T, then W is accessible to 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 W at 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 presently concerned. 2 3

This illustration is borrowed from Humberstone (1983), p. 29. Cf. Humberstone (1983), p. 29; Feldman (1986), pp. 158-9.

255

The second presupposition of the illustration is that worlds that are accessible to a group have values (and may be ranked accordingly); that is, they have deontic values relative to the group. Some philosophers will reject this presupposition. If a group (as opposed to individuals in the group) cannot comply with the Categorical Imperative, then presumably Kant would reject the presupposition, since according to him deontic value relative to an agent is rooted entirely in that agent's ability to comply with the Categorical Imperative. Similarly, if a group cannot display gratitude, then Ross would presumably find the presupposition suspect, at least in this regard. But other philosophers would not hesitate to accept the presupposition. Consequentialists, for example, will accept that worlds that are accessible to groups have, like all worlds, an intrinsic value. And Ross, insofar as he accepts that intrinsic value is relevant to the determination of obligation and hence counts as a source of deontic value, will also accept the presupposition, at least in this regard. There is a twist to this second presupposition. If Chart 9.1 is to be used in the determination of Vincent's and Virgil's individual obligations, it must be that the numbers that serve to indicate the relative desirability of outcomes do so in an agent-neutral way. That is, the unanimous election of Smith 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-relativization that is built into (I) must be seen to be insignificant. By conjoining (9.3) with the second presupposition, we may assert the following: (9.4) if a world of deontic value Vis accessible to some proper subgroup of { St, ..., Sn} from W at T, then a world of deontic value Fis accessible to {Su ..., Sn} from W at T. That is, whatever deontic good (or bad) a subgroup can achieve, the group can achieve also. But, as with (9.2), the converse does not hold. That is, we should reject this: (9.5) if a world of deontic value Vis accessible to {S2, ..., Sn} from W at T, then a world of deontic value Vis accessible to some proper subgroup, if any, of {Su ...,£„} from H^at T. The problem with (9.1) that I am about to expose has, therefore, only limited 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 that theories that avoid this problem are superior. First, if they avoid it by failing 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 theory of obligation. Second, if they avoid it by avoiding the presuppositions in question, they face a problem of their own. The first presupposition is surely 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 information is needed. Can Vincent act in such a way that Virgil will vote for Smith? If so, (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 deontic value) is achieved.5 This of course poses no problem for (9.1). Nor does this scenario: Vincent can act in such a way that Virgil will vote for Smith (so that, we may say, Virgil is not intransigent relative to Vincent concerning Smith's being unanimously elected), but Virgil cannot act in such a way that Vincent will vote for Smith (because Vincent is intransigent relative to Virgil concerning Smith's being unanimously elected; Vincent insists on voting for Jones). Here (9.1) directs Vincent to act in such a way that Virgil will vote for Smith and also directs Vincent himself to vote for Smith; but it directs Virgil to vote for Jones (since this is the best he can do, given Vincent's intransigence). It may at first seem odd that (9.1) should direct an agent to perform an action which, if performed, guarantees that less than the best possible outcome will be achieved (in the sense that ten units of deontic value will be forgone in favor of five), but on reflection this is surely right. After all, it's not Virgil's fault that voting for Jones is the best he can 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 two wrongs to make a right. Suppose that Virgil is just as intransigently in favor of Jones as Vincent is. Then not only does (9.1) direct Virgil to vote for Jones (since this is the best that he can do under the circumstances), it directs Vincent 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, even though this yields only five units of deontic value rather than ten; even though, that is, the best that both could do is not done. Somehow, Vincent's intransigently voting for Jones is transformed from being wrong to being right 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 and are thus "directly collectively self-defeating," as Parfit puts it. See Parfit (1984), pp. 54 and 95ff. 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 that one ought to do the best one can, inasmuch as it demonstrates that there is a sense in which universal satisfaction of (9.1) is compatible with the best that 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 light a limitation to (9.1) that is easily remedied. (9.1) was drawn up with the individual alone in mind; it ignored the possibility of cooperation. The problem brings this out starkly: through a mutual failure to cooperate, Vincent and Virgil, though doing the best they can individually,6 fail to do the 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 will be 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" what he would do at T" if his group did at T' the best that it could at T do 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 acts according to it, it has what may seem to be another attractive feature. Notwithstanding the discussion of supererogation in Chapter 8, it might still be thought that theories of the sort captured by (9.1) are too demanding, that they require of each of us that we exhaust ourselves in the pursuit of the good (for example, by giving and giving to charity), since many others 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 the good) — although, if everyone were to pursue the good, none of us would have 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 individually, I mean simply that each does the best he can in fact do under the circumstances. I don't mean that each does the best he can do "all on his own," i.e., exclusive of any cooperation 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 would do 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 which would be required of us under the counterfactual condition where everyone 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 moral failures 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 doing his part depends on another's doing his, and where the latter is intransigent relative to the former concerning the group's generation of the best possible outcome. For example, my part in generating the best possible outcome may be to enter a room whose door only you can unlock. If you adamantly refuse to do your part, then I cannot do mine. Third, whether or not the discussion of supererogation in Chapter 8 is acknowledged to have dealt successfully with the apparent overdemandingness 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 failures of another is surely far too sweeping. Just consider a case where all that is needed to prevent a child from dying is the donation of one dollar. There are one hundred people present who can each give one cent; you can give one dollar without any hardship whatsoever; none of the others gives money 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 that it directs Virgil to vote for Jones, given Vincent's intransigence; for had he voted for Smith, zero rather than five units of deontic value would have been 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 for Jones, given Vincent's intransigence, but which will not permit a similar intransigence on Virgil's part. Donald Regan has produced a very elabo8 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, we may give this rough version of Regan's proposal: (9.7) for (a) (b) (c)

any person 5 and times T and T", S ought at T to: hold himself ready at T (and up to T') to cooperate with others at T'; identify prior to T ' other would-be cooperators at T'; and do his part at T ' in the best action that the group of actual cooperators 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 cooperate does wrong; anyone prepared to cooperate but frustrated by other noncooperators, who nonetheless does his best under these less than ideal circumstances, does what he ought. Doesn't this capture just what we want to capture? Unfortunately not. There are two principal objections to be made, based on the fact that the proposal prescribes a certain procedure. 12 First, Regan's theory violates the thesis that "ought" implies "can": S may be unable to do what (9.7) requires. (In particular, S may be unable to identify other would-be cooperators.) Second, Regan's theory ignores the possible deontic cost involved in undertaking the specified procedure. 13 It could be that identifying other would-be cooperators is itself a very costly activity, and to overlook this is surely inconsistent with the view that one ought to do the best one can. 9.2.3 Group obligation A proposal made prior to Regan's has been resurrected by several others in response to the difficulties that Regan's faces.14 It is, in effect, that (9.1) be generalized thus: (9.8) for any persons St...Sn and times T and T", {Su ..., Sn} ought at T to 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. 1713. Regan responds to such criticism (in Regan, 1980, p. 169ff.), but the response seems unsatisfactory. 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 be temporally scattered.

260

(Where n equals 1, (9.8) is equivalent to (9.1), insofar as we are treating each individual S{ as {S,}.) This proposal takes advantage of the presuppositions that give rise to the problem in the first place. If we acknowledge that groups can sometimes achieve outcomes that individuals alone cannot, and if we acknowledge that these outcomes are themselves of deontic value, then why not accept that groups can be the agents of right and wrong and can 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 correspondence between individual and group obligation. For example, given the 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 proper subgroup, as in the case of Vincent and Virgil. And we must also reject the following 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 cooperation - that is, everyone does his part in the best action performable by the group - then neither the group nor any subgroup does wrong, we should not accept that, if all its members act as they ought, then neither the group nor 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', then it 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 following: (9.12) if 5^ ought at T to do At at T\ and.. .and Sn ought at T to do An at T'n9 and if St does At at T\ and.. .and Sn does An at T'n9 then it is not the case that {Sl9 ..., Sn} or any subgroup of {Sh ..., Sn} goes wrong 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 wished to capitalize in the attempt to solve the problem before us. Their suggestion 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) of course yields the same story about obligation and wrongdoing in the individual case as (9.1) does — each of Vincent and Virgil ought to vote for Jones and hence does no wrong in doing so — unlike (9.1), (9.8) still diagnoses a wrongdoing in this case: Vincent and Virgil jointly do wrong in not electing Smith. Thus we need not accept the disturbing and unsatisfactory contention that nothing goes wrong in the election ofJones. This is an intriguing and elegant proposal, but I think it must be rejected. First, it is not clear that groups of more than one member can have obligations. I have acknowledged that, if the problem with which we're concerned is genuine, then wre must grant the presupposition that groups can achieve outcomes that have a certain deontic value. But this does not mean that we must accept that groups can have obligations. When I presented (I) in Chapter 2,1 noted17 that it might be that we should say that an agent, to be the subject of obligations, must be a moral agent; if not, we may have to accept that nonhuman animals have obligations (for they, too, may be able to achieve outcomes that have a certain deontic value). What is involved in an agent's being a moral agent is unclear, but arguably it must have the capacity to conceive of its action in moral terms,18 and it seems safe to say that no group with more than one member (as opposed to individuals in the group) has this capacity. Furthermore, some have contended that, for something to be obligatory, it must be something about which decisions can be made, and that joint outcomes achievable by groups with more than one member (as opposed to the individual component outcomes achievable by individual members of the group) are not entities of this sort.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 has deontic value and, often, has a degree of deontic value that no outcome achievable by some proper subgroup has, they cannot be obligated to achieve 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 extension of (9.1) and to accept, therefore, that Vincent and Virgil jointly do wrong when they unanimously elect Jones. What I wish to deny is that this is all that goes wrong in the election ofJones. The problem is simply this: on the present proposal, the group wrongdoing is not attributable to any of the 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 that Vincent is intransigently in favor ofJones; not wanting to have his friend do wrong, Virgil decides to be intransigently in favor ofJones too, believing that Vincent will thereby be absolved. (He resolves not to listen to entreaties even from Vincent himself; he believes that Vincent, being his friend but subscribing to an account of obligation other than (9.8), may attempt to dissuade him from acting in this way.) Now, if Virgil acts in this way, (9.8) - as Virgil recognizes - implies only that Vincent and Virgil jointly do wrong and that neither does wrong singly. But this is too much to swallow; the scenario still smacks of two (individual) wrongs making a right. The proponent of (9.8) might reply in this way. What's wrong with intransigence, when it's met with matching intransigence? Given Vincent's intransigence, Virgil's intransigence makes no difference; Vincent's intransigence alone rules out attainment of the group's best. So too, mutatis mutandis, for Vincent's intransigence making no difference. Thus, under the circumstances, there is nothing wrong with the attitude of either.21 But there is something wrong with this attitude. True, there is a sort of overdetermination here, such that each agent's being a party to achieving less than the best outcome in a way renders the other's being a party to it redundant. 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's moral obligation is to do the best one can while avoiding intransigence; that is (to coin a term), one ought transigently to do the best one can. Doing the best one can must be accompanied by the adoption of a certain attitude. This is in essence Regan's proposal shorn of its objectionable aspect: one must do the best one can (see clause (c) of (9.7)) while holding oneself ready to cooperate (clause (a)), but there is no obligation to undertake something 20 Here "faulting" someone means ascribing wrongdoing to that person. Thus "attributing" a group wrongdoing to an individual involves finding that that individual has done wrong with respect to the group outcome in question. It is natural to use the phrase "S is responsible for X" instead of the phrase "Xis attributable to S," but one should be wary of 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 impossible 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 be such that one would do one's part if all others did theirs. When conjoined with 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 the hypothetical case where not all others cooperate but where a sufficient number do so and their doing so would generate an outcome superior to that 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 to cooperate in the generation of any superior outcome, not just the best possible, so long as this outcome would in fact be the best available (given the behavior of any noncooperators). Suppose, for example, that the best that Bob, Carol, and Ted can jointly do is to raise and push aside a tree that has fallen on Alice. Suppose also that, if Bob refuses to cooperate, the best that Carol and Ted can jointly do is to raise the tree so that Alice can crawl out from under it. Suppose finally that, if both Bob and Carol refuse to cooperate, the best that Ted can do alone is go for help. What we want is for Ted to be prepared to help raise and push aside the tree if both Bob and Carol are prepared to cooperate, and (9.13) gives us that. But what we also want, and what (9.13) does not give us, is for Ted to be prepared to help raise 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 neither Bob nor Carol is prepared to cooperate with Ted. Then (9.13') is designed to direct Ted to (a) go for help but also (b) be prepared to cooperate with both 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 prepared to cooperate (something that (9.13) doesn't say). Nonetheless (9.13') is problematic. The requirement of counterfactual cooperation stipulated in clause (b) is too strong. Suppose that Fred were force-fed some mind-altering drug. Would he do the best he could under these circumstances? Quite conceivably not (especially if the drug altered his moral perception). Would this show that he was in fact unprepared to cooperate? It would not. A second problem with (9.13') is that it appears not to conform with the thesis that "ought" implies "can." While it is of course always in one's power to do the best one can, is it always in one's power to be such that one would do the best one could if circumstances were different? Fred appears to provide a counterexample. Let us assume that, having ingested the drug, he would not (although of course he could) do the best that he could do. There seems to be no reason to assume that whether or not this is 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 understood as 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 just one member.) W e 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 the best 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 it that G can so act that p is true is do his part in the group action in question. N o w it may be (as noted before in the discussion ofyou, me, and the locked door) 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 and G can, then, perform his or its respective part if the other does; otherwise it would not be the case that they could jointly perform the action in question. Thus each of S and G can be open to the other concerning p; each of them can so act that the other can so act that p is true. If either S or G cannot so act that p is true, it is because the other has avoidably prevented this. I think that we should accept that transigence is openness, as this is defined in (9.14). (9.15) might therefore appear to give expression to the solution to which I have alluded. Indeed, (9.15) is notably superior to any of 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 and not 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's preparedness to cooperate be general. The fact is, however, that (9.15) faces serious problems; the claim that one 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 5 remain transigent while performing an action that precludes such transigence. Suppose, for example, that it is time for Vincent to vote, that Virgil has not yet voted, but that Virgil remains intransigently in favor ofJones. Then clause (a) of (9.15) requires that Vincent now vote for Jones. In doing this, Vincent cannot of course remain transigent concerning Smith's unanimous election; yet clause (b) requires that he still be transigent in this respect, since it is still possible for {Vincent, Virgil} to so act that Smith is unanimously elected, and this can be achieved only by Vincent's voting for Smith. It may seem that the obvious solution to this first problem is simply to stipulate in clause (b) that T is earlier than T", to stipulate, that is, that transigence concerning the best outcome achievable by the group is required only up until the time at which 5 must act in order to do what is in fact the best that he can do. But a second problem is that, even with this amendment, (9.15) would still on occasion issue in a basic moral dilemma. This is because it can sometimes happen that the only way in which 5 can be open to G concerning some group-achievable outcome is by doing his part in the achievement, or what would be the achievement, of that outcome. Suppose, for example, that it is not yet time to vote, that both Vincent and Virgil 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 suppose now 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 it is the mere fact that Vincent is going to vote for Jones that makes it the case that he is closed to Virgil concerning Smith's being unanimously elected; so too, mutatis mutandis, for Virgil's being closed to Vincent. Their being mutually intransigent in this regard is due, therefore, not to any orneriness on anyone's part, but simply to their mutual disconnectedness. Their inability to interact renders each inflexible to, unmanipulable by, the other; neither can affect the way the other votes, and hence, however the one votes, it is perforce afait accompli for the other. Now, (9.15) condemns even such closedness, but it ought not to. For, under the circumstances, even if T is earlier than T", the only way in which Vincent can at T be open to Virgil concerning Smith's being unanimously elected at T" is by actually voting for Smith at T". Thus clause (b) of (9.15) implies that Vincent ought at T to vote for Smith at T', even though, given the fact that Virgil will vote for Jones, clause (a) implies that Vincent ought at T not to vote for Smith at T" but to vote for Jones instead. Perhaps this result should not surprise us. If Vincent and Virgil are disconnected in this way, then, even though there is a sense in which they can jointly, or as a group, achieve Smith's unanimous election, such action hardly seems paradigmatic of what we normally think of as joint, or group, action. It is certainly not concerted action, action that is undertaken by a group of interacting individuals. For this reason, even if we accept (in the spirit of (9.8)) that Vincent and Virgil have, as a group, done wrong, there would seem to be no reason to think of such wrongdoing as attributable even in part to any member(s) of the group. If anything, it is "attributable to nature."23 A third problem with (9.15) is that it requires that S be open to any group's best group-achievable outcome. This is surely too sweeping; once again, it can issue in a basic dilemma. For it can happen that what S must do to remain open to one group's best achievable outcome conflicts with what he must do to remain open to another group's best achievable outcome. For example, it could happen that Bob's and Carol's and Ted's jointly doing their best involves Ted's taking ten paces to the right, while Carol's and 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 costliness of doing what it requires. Even openness can be costly, as when (perhaps - it depends on just what deontic value consists in) S's openness to G, having become public knowledge, prompts someone to perform a harmful act that would otherwise not have been performed. 23 This is an ironic use of "attributable." Nature can do no wrong.

267

9.3 T H E 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 stipulation. The requirement that S be open to G must be conditional on such openness not preventing S from doing what is in fact the best that he can do under the circumstances. One ought transigently to do one's best, but the transigence must be unintrusive. I propose, therefore, this revision of (9.1) as a way of resolving the problem 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 T 7 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 T 7 , \£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 ' and do ,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 this transigence be unintrusive. Just as (9.1) only gives rough expression to (I), so (9.1') only gives rough expression to what I take to be the proper revision of (I). But before I set out this revision more precisely, let me address certain issues. In particular, let me explain how it is that the requirement that transigence be unintrusive is designed to solve the four difficulties that (9.15) faces, and then let me 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 do what is best only if he fails to meet his requirement to be transigent. Hence the sort of basic dilemma outlined in the first problem with (9.15) cannot arise on (9.17). Nor can the second sort of basic dilemma outlined in the second problem. If the best that Vincent can in fact do is vote for Jones, and the only way in which he can be transigent concerning Smith's unanimous election is to vote for Smith, then he is not required to be transigent in this respect. 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 the best 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 do so) leave it up to G whether or not S does his part, and it is in virtue of this fact 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 now think of them as mutually connected. Vincent's being connected to Virgil concerning Smith's being unanimously elected consists in this: Vincent can act in such a way that it is left up to Virgil whether both of them vote for Smith. (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. Perhaps Vincent can simply inform Virgil — truthfully — that he will vote as Virgil wishes.) That Vincent can act in this way is, of course, no guarantee that he will. While disconnectedness ensures lack of influence, connectedness doesn't ensure influence. If Vincent opts for intransigence concerning Smith's being unanimously elected, then, even though he is connected to Virgil in this respect, Virgil cannot influence Vincent in this regard, and so Virgil's access to the best that is jointly achievable is effectively obstructed by 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 is not achievable by some group of which he is a member, and also, as the falsity of (9.5) implies, that a group can sometimes achieve a good that no member can) is to do his part in the best group action that those who can in fact be moved to cooperate can perform. If S's holding himself ready to cooperate in another group action in fact conflicts with his holding himself ready to cooperate in this group action, then it conflicts with his doing what is in fact best, and (9.17) therefore implies that he ought not to hold himself ready to cooperate in the other group action. Finally, if S's holding himself ready to cooperate is unduly costly, then it 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 discussion 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 that counts (with respect to acting as one ought), the former implies that how one 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 that is not "exclusively act-oriented," where (it seems) a theory is exclusively act-oriented just in case one's satisfying it depends solely on whether (and not 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 is so in the case where at least one of Vincent and Virgil can both do his best and be transigent concerning Smith's unanimous election. In this case, whereas (9.1) implies that both Vincent and Virgil do as they ought if they both intransigently vote for Jones, (9.1') implies that both do as they ought only if they both transigently vote for Smith. For, suppose that Vincent can both do his best and be transigent concerning Smith's being unanimously elected. 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, and so 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 neither Vincent nor Virgil can both do his best and be transigent concerning Smith's being unanimously elected, then (9.1') does not differ from (9.1) in its implication that each of them ought to vote for Jones. This is so even though they would and could jointly do their best by unanimously electing Smith. Consider, for instance, the case where Vincent and Virgil are mutually disconnected with respect to Smith's being unanimously elected. Here the group does not do its best (and, if groups of more than one member can be obligated, the group does wrong), but no individual does wrong. But here, I submit, this result is not anomalous, precisely because of the disconnectedness. It is when transigence can be unintrusive that a group's failure 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 that is jointly achievable, then (9.1') certainly implies that S does no wrong if he 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, so that an agent can on occasion do his part simply by letting others get on with 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 attitude of (in) transigence is always within the agent's control — it is an optional attitude — and clearly its adoption will sometimes involve action (as when Vincent tells Virgil that he will vote as Virgil wishes); and if, as I suspect (and as noted in Subsection 2.2.1), nothing is in one's control except by way of a decision, and if a decision can be said to constitute a minimal type of action, then in fact adopting (in) transigence will always involve action.

270

always actively contribute to the achievement of the best group-achievable outcome. There are three sorts of cases where his remaining idle in this regard is permissible. First, it may be that he cannot actively contribute to the outcome. Second, it may be that, although he can actively contribute to some outcome, he ought nonetheless to remain idle, because only in this way can the group achieve its best. (If Rachel is a superior swimmer to Roberta and Roberta's joining in the rescue of a drowning person would only mess things up, then Roberta ought to stand by while Rachel carries out the rescue.) Third, it may be that S is permitted a choice as to whether to remain idle, because there are more than enough people actively contributing to the group-best and his active contribution, though not counterproductive, would not help either. In this last sort of case, if S does not actively contribute to the outcome, he gets a kind of free ride; the best is achieved without any effort on his part. Such free riding seems to me unobjectionable, though, if accompanied by the transigence that (9.1/) requires. Fourthly, the illustrations used so far (of Vincent and Virgil; of Bob and Carol and Ted and Alice; of Rachel and Roberta) might be most naturally understood to concern group actions whose individual components occur simultaneously, but such simultaneity is not essential to the illustrations. Indeed, I have already mentioned the possibility that Vincent's vote precede Virgil's, and it may be that Bob and Carol and Ted must perform actions that occur at (slightly) different times in order to save Alice. Now, it might seem that the temporal order of the individual component actions in a group action can make a difference as to just what the members' individual obligations are - that whether or not to participate is optional for the earlier links in the chain in a way in which it is not for the last link; for by the time the last link has a chance to contribute, all the others have contributed, their participation is settled, and so it is now clearly up to him alone whether the outcome is achieved. But in fact to single out the last link in this way would be a mistake. For each link there is the question of whether he ought to participate, and the answer is contingent on others' actions, whether or not these actions have already taken place. Thus, where ten 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 message along too. On the other hand, if at least one of Archie, Barry,..., and Ian has not cooperated, then it is not the case that Jack ought to cooperate (for he 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 case that Archie ought to cooperate (for his effort will be wasted; he could presumably do something better — unless, of course, he could prevail upon the noncooperators to cooperate). In each instance, then, the timing of others' actions is irrelevant to the determination of an agent's individual obligation, except insofar as such timing renders the others insusceptible to the agent's influence. 5 In summary, then, (9.1') condemns, as (9.1) does not, one's obstructing another's access to the best possible outcome, when avoidance of such obstruction is compatible with one's doing one's best. In this way, the problem facing (9.1) is solved. However, (9.17) only gives rough expression to the revision of (I) that I think is called for. At this point I offer the following 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 deontic value for S at T of W'. Let us say that any act and time that meet the conditions laid out in (I) for A 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 W at 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 value of 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 and forgo the details. Furthermore, if a group of more than one individual can be obligated to do something, as (9.8) contends, then (I") can be extended to 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 posed by Bernard Williams and first presented in Section 1.2. It concerns these three statements: (1.1) (1.2) (1.3)

someone ought to help that old lady: Jones is the only person who can help her; 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 wish now to point to a possible reading of (1.1) which Williams doesn't consider and which may account for the inclination, if there is one, to accept the inference in question. It is this: as Williams says, (1.1) does not express an individual obligation, but it does express a group obligation. That is, one might 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 individual in the group to help the old lady. But if we now say, with (1.2), that only Jones is so situated that he can carry out the group's obligation, then we may indeed infer, with (1.3), that he, as an individual, is obligated to help the old lady. For the best that he can achieve can be no better than the best that the group can achieve, and if the group can achieve its best only by his acting in this way, then he can achieve his best only by acting in this way. This of course does not help Williams's cause, which is to establish some important link between binding and nonbinding senses of "ought," but it does at least provide a way to grant him his claim that the inference is to be accepted even though the first premise does not express an individual 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 requirement concerning transigence, as long as it is acknowledged that, in any situation, the best that one can do is to perform some act transigently. In other words, transigence is just one of the features of (and thus just one of the features that contributes to the deontic value of) what one does and is simply to be included in the calculus of values along with all the other value-contributory properties of actions (whatever they may be). I have no real quarrel with this. As long as transigence is acknowledged as being deontically significant, my main point in this chapter has been accepted. 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 not be misled. We can revert to (I) only if we stipulate that transigence, as spelled out in clause (b) of (I"), is a necessary feature of the best that an agent can do. There is, therefore, no gain in simplicity. In fact, this way of proceeding strikes me as pretty messy. We would need to be able to talk of what would otherwise be best if transigence were bracketed, in order to be able 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. The result would be an account of obligation extensionally equivalent to the one 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 this extension, the question of group moral dilemmas could be treated in a man274

ner analogous to that employed in Chapter 7 when individual moral dilemmas were discussed. That is, basic group moral dilemmas would be impossible - being ruled out by analogues to (2.46) (the thesis that "ought (singly)" implies "can (singly)"), (2.40'b) (the principle of agglomeration for "ought (singly)"), and so on27 - but certain nonbasic group moral dilemmas (group binds, group subsidiary dilemmas) would be possible. However, the fact that individuals never act in isolation gives rise to the possibility of another sort of dilemma that I have called an "interpersonal moral dilemma." As stated in Section 7.4, the form of a (two-person) interpersonal moral dilemma is this: (7.32) (a) (b) (c) (d) (e)

St ought overall at T1 to do A at T3; Sj can at Ti do A at T3; S2 ought overall at T1 to do B at T4; S2 can at T1 do B at T4; and 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; Donald ought to buy his daughter a Cabbage Patch doll for her birthday tomorrow; 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). How might this case arise? If Donna and Donald are mutually disconnected, then this case could arise only by way of some lapse; hence the dilemma would constitute a perplexity secundum quid, as explained in Section 7.4. This is so for the following reason. Suppose that Donna cannot see to it whether Donald buys the doll, and suppose that Donald does as he ought and buys the doll. Then every world accessible to Donna is one in which Donald buys the doll; hence no world accessible to her is one in which she buys it; hence it is not the case after all that she ought to buy it. Thus it can happen that Donna ought to buy the doll only if Donald fails to act as he ought; her obligation presupposes his lapse. The same applies, mutatis mutandis, where Donald cannot see to it whether Donna buys the doll. The point here is not simply the obvious one that a conflict of obligations implies that some wrong is done; it's that the sort of conflict present27 28

Cf. McConnell (1988a), p. 27. This illustration is borrowed, in modified form, from McConnell (1988a), p. 28.

275

ly at issue, in the circumstances presently at issue, presupposes that some wrong 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 if the other fails to satisfy his obligation. If Donna and Donald are connected, however, the story is different. But if we make the presuppositions operative throughout this chapter - in particular, the presupposition that deontic value is agent-neutral - the result is still 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 a group can always achieve its best and, if it does so, then, as (9.11) says, no member does wrong. If a situation arises, then, in which some member cannot avoid some member's doing wrong, this can only be because the group does wrong; the members' having conflicting obligations presupposes the group's lapse. But even if we do not admit the possibility of group obligation, still such conflicting obligations presuppose a lapse on someone's part. There are three possibilities. Either Donna's buying the doll is deontically superior to Donald's doing so (more precisely, every world accessible to {Donna, Donald} in which Donald buys the doll is deontically inferior to some world accessible to {Donna, Donald} in which Donna buys the doll), or Donald's buying the doll is deontically superior, or they are of equal deontic value. Suppose that the first is the case. Given the agent-neutrality of deontic value, it must be that, if Donna is transigent concerning her buying the doll, then Donald can and ought to ensure that she buys the doll and that he does not buy it himself; hence, if in fact he has the obligation to buy it, this presupposes both that Donna does not do her best and that she is intransigent concerning her (and thus the group's) best, and so it presupposes a double lapse on her part. So too, mutatis mutandis, for the case where Donald's buying the doll is deontically superior to Donna's doing so. As to the final possibility: if the group-best is for one or the other of Donna and Donald to buy the doll, then it could not be individually better for Donna or Donald to buy it unless either Donna or Donald acts so as to preclude achievement of what is group-best; this, too, requires a lapse on someone's part. Finally, if Donna and Donald are connected but deontic value is not agent-neutral, it appears that interpersonal moral dilemmas can arise without any lapse on someone's part. Their dilemma would be rooted simply in the agent-relativity of deontic value. If it is thought that no moral dilemma of any sort can arise unless it constitutes a perplexity secundum quid, this would constitute a reason for rejecting the possibility that deontic value be significantly agent-relative. 276

Postscript

I have endeavored in this book to develop and defend a certain analysis of the concept of moral obligation, an analysis that can be put to good use in the resolution of a number of philosophical issues. In the Preface, I made mention of several questions and claimed that they would receive answers in the ensuing pages. And so they have, although some of the answers have been somewhat tentative. My hope is that these answers will have served collectively to set the concept of moral obligation in high relief Such clarification is valuable not just for its own sake. Although the account of obligation that I have presented is, appropriately, neutral regarding competing substantive theories of obligation, it should provide a useful framework within which these theories can be developed and assessed. Thus the exercise in metaethics undertaken in the foregoing pages has significant implications for normative ethics. Of course, it has not been my purpose to expound on these implications. Undoubtedly, too, my account has implications in the area of applied ethics, but tracing these is a very complex matter, and I offer no suggestions on how best to carry out this task. To repeat what was said in the Preface, then, anyone seeking solutions to such problems as those besetting the unfortunate Indianapolis would be ill advised to look 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 here regarding the questions that have been addressed, I shall be content.

277

Appendix: List ofpropositions

(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) 40 278

(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

(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) 90 279

(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

(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) (4.53) (4.54) (4.55) (4.56) (4.57) (4.58) (4.59) (4.60) (4.61) (4.62) (4.63) (4.64)

130 130 130 135 135 137 138 138 138 140 140 140 140

(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) 165 280

(5.23) (5.24) (5.25) (5.26) (5.27) (5.28) (5.29) (5.30) (5.31) (5.32) (5.33) (5.34) (5.35) (5.36) (5.37)

165 166 170 171 172 177 177 182 182 182 182 182 182 183 183

(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

(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

281

(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

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-Hill Aqvist, 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 University Press. 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: The Simplest 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, and Deontic Foci (A Rejoinder to Fred Feldman)." Philosophical Studies, 57: 157-74. Chisholm, Roderick M. (1963). "Contrary-to-Duty Imperatives and Deontic Logic." Analysis, 24: 33—6. Chisholm, Roderick M. (1968). "Supererogation and Offence." In Thomson and Dworkin (1968): 412-29. Chisholm, Roderick M. (1974). "Practical Reason and the Logic of Requirement." In Practical Reason, edited by Stephan Korner. New Haven: Yale University Press, 1974, 1-17. Chisholm, Roderick M. (1976). Person and Object. La Salle, 111.: Open Court. Chisholm, Roderick M. (1989). Theory of Knowledge, 3rd edition. Englewood Cliffs: Prentice-Hall. Chisholm, Roderick M. and Sosa, Ernest (1966). "Intrinsic Preferability and 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." American Philosophical Quarterly, 26: 133-41.

Davidson, Donald (1980). Essays on Actions and Events. Oxford: Clarendon Press. Davis, Lawrence H. (1979). Theory of Action. Englewood Cliffs: PrenticeHall. Donagan, Alan (1977). The Theory of Morality. Chicago: University of Chicago Press. Donagan, Alan (1987). "Consistency in Rationalist Moral Systems." In Gowans (1987b): 271-90. 283

Dreier, James (1993). "Structures of Normative Theories." Monist, 76: 22-40. Driver, Julia (1992). "The Suberogatory." AustralasianJournal ofPhilosophy, 70: 286-95. Feinberg, Joel (1968). "Supererogation and Rules." In Thomson and Dworkin (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 L o g i c . " Philosophical Perspectives, 4: 309—41.

Feldman, Fred (1993). "On the Consistency of Act- and MotiveUtilitarianism." Philosophical Studies, 70: 201-12. Follesdal, Dagfinn and Hilpinen, Risto (1970). "Deontic Logic: An Introduction." In Hilpinen (1970): 1-35. Foot, Philippa (1980). "The Problem of Abortion and the Doctrine of Double Effect." In Killing and Letting Die, edited by Bonnie Steinbock. 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, and Supererogation." Ethics, 85: 219—26. Frankena, William K. (1963). "Obligation and Ability." In Philosophical Analysis, edited by Max Black. Englewood Cliffs: Prentice-Hall, 1963, 148-65. Frankfurt, Harry G. (1969). "Alternate Possibilities and Moral Responsibility." Jowratf/o/P/n/osop/*)/, 66: 829-39. Frankfurt, Harry G. (1988). The Importance of What We Care About. Cambridge: Cambridge University Press. Fried, Charles (1978). Right and Wrong. Cambridge, Mass.: Harvard University Press. 284

Gert, Bernard (1988). Morality: A New Justification of the Moral Rules. New

York: Oxford University Press. Gewirth, Alan (1981). "Are There Any Absolute Rights?" Philosophical Quarterly, 31: 1-16.

Gibbard, Allan (1965). "Rule-Utilitarianism: Merely an Illusory Alternative?" 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 ofHuman Action. Princeton: Princeton University 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 and Morals, 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 Supererogation." 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." In Hilpinen (1970), 59-104. Hohfeld, Wesley Newcomb (1919). Fundamental Legal Conceptions. New Haven: Yale University Press. Hudson, Stephen D. (1986). Human Character and Morality. London: Routledge and Kegan Paul. Humberstone, I. L. (1983). "The Background of Circumstances." Pacific Philosophical Quarterly, 64: 19-34. Humberstone, I. L. (1991). "Two Kinds ofAgent-Relativity." Philosophical Quarterly, 41: 144-66. Hume, David (1888). Treatise of Human Nature, edited by L. A. SelbyBigge. 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, edited by Philip Pettit, Richard Sylvan, and Jean Norman. Oxford: Basil Blackwell, 1987, 91-110. Jackson, Frank (1991). "Decision-theoretic Consequentialism and the Nearest and Dearest Objection." Ethics, 101: 461-82. Jackson, Frank and Pargetter, Robert (1986). "Oughts, Options, and Actualism." Philosophical Review, 95: 233—55. Jackson, Frank and Pargetter, Robert (1987). "Two Puzzles about Conditional 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: Clarendon Press. 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 ofDuty. Albany: State University of New York Press. Mellema, Gregory (1993). "Quasi-Obligation and the Failure to Be Virtuous." Journal of Social Philosophy, 24: 176-85.

Mendola, Joseph (1987). "The Indeterminacy of Options." American Philosophical 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 Division Meeting of the American Philosophical Association. Raz, Joseph (1975). "Permissions and Supererogation," American Philosophical 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 to Sen." 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." Philosophical Review, 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 Ethical Theory, 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 Existentialist Philosophy, 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 Moral Theory." Nous, 20: 179-97. Scheffler, Samuel (1984). The Rejection of Consequentialism. Oxford:

Clarendon Press. Schwartz, Thomas (1985). "Dr. Krankheit and the Concept of Compliance." American Philosophical Quarterly, 22: 81—7.

Sen, Amartya (1983). "Evaluator Relativity and Consequentialist Evaluation." Philosophy and Public Affairs, 12: 113-32.

Seung, T. K. and Bonevac, Daniel (1992). "Plural Values and Indeterminate Rankings." Ethics, 102: 799-813. Shaw, P. D. (1965). "'Ought' and 'Can.'" Analysis, 23: 118-20. Simon, Caroline J. (forthcoming). "Supererogation: Puzzle or PseudoProblem?" 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 Compound Acts." Journal ofPhilosophy, 83: 341-5. Snare, Frank (1974). "The Definition of Prima Facie Duties." Philosophical Quarterly, 24: 235-44. Sobel, Jordan Howard (1968). "Rule-Utilitarianism." Australasian Journal of Philosophy, 46: 146-65. Sobel, Jordan Howard (1976). "Utilitarianism and Past and Future Mistakes." Nous, 10: 195-219. Sorensen, Roy A. (1991). "Moral Dilemmas, Thought Experiments, and Conflict Vagueness." Philosophical Studies, 63: 291-308. Stocker, Michael (1971). "'Ought' and 'Can.'" Australasian Journal of Philosophy, 49: 303-16. Stocker, Michael (1990). Plural and Conflicting Values. Oxford: Clarendon Press. 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. New York: Harper and Row. Unger, Peter (1977). "Impotence and Causal Determinism." Philosophical Studies, 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 Moral Theories." Metaphilosophy, 19: 253—63. Vallentyne, Peter (1989). "Two Types of Moral Dilemmas." Erkenntnis, 30: 301-18. Vallentyne, Peter (1992). "Moral Dilemmas and Comparative Conceptions 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, MichaelJ. (1986). "Subsidiary Obligation." Philosophical Studies, 50: 65-75. Zimmerman, Michael J. (1987). "Remote Obligation." American Philosophical 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 Philosophical Quarterly, 27: 345-55. 291

Zimmerman, MichaelJ. (1990b). "Where Did I Go Wrong?" Philosophical Studies, 58: 83-106. Zimmerman, Michael J. (1992). "Cooperation and Doing the Best One Can." Philosophical Studies, 65: 283-304. Zimmerman, Michael J. (1993a). "Obligation, Responsibility, and Alternate Possibilities." Analysis, 53: 51-3. Zimmerman, Michael J. (1993b). "Supererogation and Doing the Best One Can." American Philosophical Quarterly, 30: 373-80. Zimmerman, Michael J. (1993c). "A Plea for Ambivalence." Metaphilosophy, 24: 382-9. Zimmerman, MichaelJ. (1994). "Rights, Compensation, and Culpability." Law and Philosophy, 13: 1—32.

Zimmerman, MichaelJ. (1995a). "Prima Facie Obligation and Doing the Best One Can." Philosophical Studies, 78: 87-123. Zimmerman, Michael J. (1995b). "Responsibility Regarding the Unthinkable." Midwest Studies in Philosophy, 20: forthcoming.

292

Index of names

Anscombe, G. E. M , 55, 55n., 56, 278 Aquinas, St. Thomas, 229, 229n., 278 Aqvist, Lennart, 73n., 74n., 278 Attfield, Robin, 246n., 278 Atwelljohn, 171n., 278 Austin, J. L., 93, 93n., 94, 94n., 278 Badhwar, Neera, xiii Baier, Kurt, 239n., 241n., 278 Barley, Blake, 260n., 278 Barnes, Gerald, 254n., 278 Baron, Marcia, 233n., 253n., 278 Belzer, Marvin, 173n., 282 Bennett, Jonathan, 243n., 278 Berlin, Isaiah, 58n., 278 Bonevac, Daniel, 58n., 285 Brand, Myles, 38n., 278 Broad, 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., 279 Clarke, Randolph, xiii Clinton, Bill, 45 Coley, James, xiii Conee, Earl, xiii, 208n., 217n., 225n., 228n., 260n., 279 Copp, David, xiii

Davidson, Donald, 43n., 55, 55n., 56n., 279 Davis, Lawrence H., 4In., 279 Donagan, Alan, 17n., 40n., 141, 142, 142n., 143, 166n., 175, 184n., 217n., 229, 229n., 230n., 279 Dreier James, 18n., 280 Driver, Julia, 235n., 238, 238n., 280 Dworkin, 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., 280 Foot, Philippa, 38n., 280 Forrester, James, 127n., 280 Forrester, Mary, 247n., 280 Frankena, William K., 80n., 90n., 91n., 280 Frankfurt, Harry G., 52, 52n., 66, 67n., 85, 85n., 86, 87, 88, 89, 280, 287 Fried, Charles, 17n., 40n., 280

293

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, 282 King, John, xiii

Gert, Bernard, 213n., 281 Gewirth, Alan, 183, 184, 184n., 281 Gibbard, Allan, 254n., 281 Goble, Lou, 127n., 191n., 193n., 196, 197n., 198n., 202n., 204n., 205n., 206n., 281 Goldman, Alvin I., 55n., 97n., 281 Goldman, Holly S., 31, 31n., 97n., 117n., 136n., 189n., 191n., 281 Gorr, Michael, 38n., 281 Gowans, Christopher, 175n., 218, 218n.,281 Greenspan, Patricia S., 114n., 121n., 122n., 136n., 191n., 281 Haji, Ishtiyaque, xiii Hart, H. L. A., 7n., 178n., 281 Henderson, G. P., 95n., 281 Heyd, 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., xiii Hilpinen, Risto, 73n., 280, 281 Hintikka, Jaakko, 31, 31n., 144n., 145n., 282 Hoffman, Joshua, xiii Hohfeld, Wesley Newcomb, 176, 177, 177n., 282 Hudson, Stephen D., 213n., 282 Humberstone, I. L., 3n., 191n., 255n., 259n., 282 Hume, 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

Landers, Ann, ix . Lebus, Bruce, 220n., 282 Lemmon, E. J., 217, 218, 218n., 219, 219n., 282 Lemos, Ramon M., 17n., 50n., 282 Leplin, Jarrett, xiii Lewis, Carl, 46 Lewis, David, 18n., 60n., 172n., 282 Loewer, Barry, 173n., 282 Lyons, 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., 283 McCloskey, H. J., 50n., 173n., 283 McConnell, Terrance C , xiv, 16n., 44n., 59n., 93, 94, 94n., 108n., 142n., 175n., 176n., 217n., 225n., 230n., 241n., 262n., 275n., 283 McGinn, Colin, 7In., 283 McKinsey, Michael, 31, 3In., 97n., 131, 131n., 132, 132n., 136, 136n., 137, 137n., 138, 138n., 260n., 283 Mele, Alfred, xiii Mellema, Gregory, xiii, 3n., 55n., 234n., 237n., 242n., 247n., 249, 249n., 252n., 284 Mendola, Joseph, 60n., 284 Mill, John Stuart, 39n., 284 Montague, Phillip, 247n., 284 Montefiore, Alan, 93, 93n., 94, 284 Moore, 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., 284 New, Christopher, 234n., 237n., 239n., 241n., 284 Nozick, 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, xiv Postow,B. C.,260n.,284 Powers, Lawrence, 95n., 284 Prichard, 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., 284 Raz, Joseph, 243n., 246n., 284 Reath, Andrews, xiv Regan, Donald H., 254n., 257n., 258n., 259, 259n., 260, 260n., 263, 269, 270n., 278, 279, 284, 285 Richman, Robert J., 95n., 285 Roberts, Robert C , 225, 225n., 285 Robinson, Richard, 95n., 285 Rosenkrantz, Gary, xiv Ross, Alf, 72, 72n., 118n., 122, 285 Ross, 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., 285 Russell, Bertrand, 11, lln., 12n., 285 Sartre, Jean-Paul, 211, 211n., 212, 220, 223, 285 Sayre-McCord, Geoffrey, xiv, 75n., 285 Scheffler, Samuel, 29n., 239n., 285 Schwartz, Thomas, 264n., 285 Sen, Amartya, 18n., 285 Seung, T. K., 58n., 285 Shaw, P. D., 81n., 285 Sider, Ted, xiv Simon, Caroline J., 244n., 285 Sinnott-Armstrong, Walter, 58, 58n., 82, 82n., 83, 83n., 84, 84n., 90n.,91,91n.,93,93n.,94, 95n.,98n., 110, 111, llln., 209, 209n., 213n., 214n., 219, 220n., 285 Slote, Michael, 18n., 61n., 239, 239n., 250, 250n.,251,285 Smart, J. J. C , 239n., 241n., 285, 286 Smith, Holly M., 262n., 286 Snare, Frank, 170, 171, 171n., 172, 173, 173n., 286 Sobel, Jordan Howard, 191 n., 254n., 286 Sorensen, Roy A., 60n., 286 Sosa, Ernest, xiv, 236n., 238, 238n., 279 Stocker, Michael, 19, 19n., 95n., 98n., 221,221n.,239n.,243n.,286 Strawson, Galen, 14n., 286 Stump, Eleonore, 235n., 286 Styron, William, 220n., 286 Swank, Casey, 213n., 286 Tannsjo, Torbjorn, 192n., 286 Thomason, Richmond H., 108n., 191n.,286 Thomson, Judith Jarvis, xi, 13, 13n., 17n.,23n.,27n., 181, 186n., 262n., 284, 286

295

Tirrell, Lynne, xiv Twain, Mark, 90

von Wright, Georg Henrik, 229, 230, 230n., 287 Vorobej, Mark, 130, 130n., 287

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., 287 van Rijen, Jeroen, 89n., 287 Velazco y Trianosky, Gregory, xiv, 248n., 287

Walton, Douglas N., 38n., 287 Washington, George, 89, 90 White, Alan R., 82, 82n., 90n., 91, 91n.,92n.,93,93n.,95n., 115, 287 Widerker, David, 85, 85n., 86, 94, 287 Williams, 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, 40 accessibility, 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, 276 see also "can"; options act-categories, 11-18 act-evaluation, 20, 244-6, 247n., 250-1 act-individuation, 203 see also actions, types and tokens act-oriented theories, 270 actions compound, 22, 23n., 34-5, 37, 63, 66-8, 237 courses of, 24, 80, 240 group, Ch. 9 passim types and tokens, 53-7, 164n., 171-2, 236-7 actualism, Ch. 6 passim objections to, 191-6 advice, 1,3, 13, 198-200,212 agent-evaluation, 20, 247n. agents, moral, 27n., 261-2 agglomeration, 68, 71, 212-14, 219, 275 alternatives, 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, personal analysis, 75-8, 170, 172-3 attitudes, 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 of binds, 226-31, 275 blame, 13, 16-17, 44n., 85-8, 93, 108-10, 131, 132n., 205, 224n., 247, 253, 263n. and "can avoid," 81-2, 86-90, 95-6 see 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, 204 see also accessibility; "ought," and "can" Categorical Imperative, 18, 28-31, 116,256 character, 20, 71, 90, 225 charity, 38-9, 116, 179, 233, 243, 258-9 choice, see accessibility, and choice commensurability intramoral, 6, 58-9, 61n., 132, 134-5, 220-1 moral and nonmoral, 1—2, 6 comparability, see commensurability compunction, 174-5, 222 conflicts of obligation, see dilemmas; obligation, conflicts of conjunction, see actions, compound connectedness, see disconnectedness

297

perfect and imperfect, 38-40, 57, 233-4 positive and negative, 38 proper, 5, 141

conscientiousness, 16-17, 95 consequentialism, 18-19, 27, 30-1, 221, 237, 239, 241-2, 245, 250, 252, 253n., 256 continuity, see value, continuity of control, see, "can," and control cooperation, 42-5, Ch. 9 passim counterfactual, 263—5 procedural, 259-60 unrequited, 258-9 culpability, see blame

sans phrase, 5, 141

self-regarding, 180 see also obligation excuses, see justification, and excuse experience, 211—12

deontic status, 14-15, 221, 233, 237-8, 245-7, 249-52 deontic value, see value, deontic deontology, 17n., 18-19, 27, 30-1 detachment 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, 66 dilemmas, 117, 192-3, Ch. 7 passim avoidance of, 221—2 basic, 211-27, 229-30, 239, 266-8, 275 group,275 interpersonal, 229-30, 274-6 nonbasic, 225-31, 275 obligation-, 62-3, 209-10 prohibition-, 62, 209-10 subsidiary, 228-9, 231, 275 disconnectedness, 267—70, 275—6 discretion, 239-44 disjunction, 35-7, 62-6, 68-9, 72-3, 144n., 233, 236, 239-40 divine command theory, 18—19 division of values, see value, division of duty absolute, 15n.

fault, 257, 262, 263n., 269 fidelity, 141-3, 146-7, 150-1, 155-64, 166-7, 170-1, 174, 181, 183, 187, 208, 213, 215, 232 free-riding, 271 freedom, 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 nonbinding intransigence, 198-9, 257-9, 261, 266-7, 269-70, 276 justice, 18, 30, 163-4, 174 justification and excuse, 86, 93-5 lapses, 104, 225, 230-1, 236, 275-6 see also wrongness and wrongdoing laziness, 189, 192-3, 203-4, 206 liberty, see freedom; rights, liberties luck, 44, 49, 110, 206n., 209 lying, 89-90

ceteris paribus, 5

conditional, 5, 143, 146, 163, 169 correlativity to rights, 6—9, 39 see also rights, correlativity to obligations

making amends, see reparation maximization, see value, maximization of maxims, 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 obligation absolute, 5, 57, 141 actual, 5, 141 all else being equal, 145-60 all things considered, 5-6 cancellation 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, 141-4 see also dilemmas defeasibility of, see augmentation and 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, 113 formal, 12 group, 178n., 179n., 260-3, 270, 273-6 hypothetical, 200-1 immediate, 52, 96-113, 226 in practice, 169, 170n., 174, 181 in principle, 168-9, 170n., 174, 181 incidental, 150-3, 155, 157, 161-3 and indefiniteness, 57—63 and intention, 17 and knowledge, 17, 50, 108-10, 164-9, 173-4 levels of, 131-40, 152n., 176 material, 12 mere, 10-11, 13, 20-1, 39, 175, 178, 180, 187 moral and nonmoral, 1-2, 10 necessary, 165-70, 181-2 objective and subjective, 10—20 overall, 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., 181 prima facie, 5-10, 18, 20, 94, 111, 123, 136, Ch. 5 passim, 207-9, 213, 215-20, 222, 232, 252 primacy of, 99-100, 102 remote, 52, 96-113, 178, 226, 228 residual, 142, 174-6, 222 restoration of, 112-13 restrictive, 200-2 resultant, 141n. subsidiary, 20, 128-40, 202, 228-9, 231 supersession of, 97, 111—13 and timing, 105-6, 271-2 see also obligation, dynamics of -to, 7-10, 39, 175, 178-80, 187, 234, 238 universal, 165-6, 170n., 181 see also "ought"; paradoxes of obligation openness, 265—7 see also transigence options indeterminacy of, 60 infinity of, 60-2 moral, 27, 33, 59, 234-44, 246, 249-52, 271 personal, 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, 201-2 restriction of, 97, 119-20, 201 "ought" binding and nonbinding, 2—5, 9—10, 64, 91-2, 94, 174n., 246-7, 274 and "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 nonbinding iteration of, 63, 70-1 and knowledge, 49-50 and material implication, 63, 69-70, 118n.,123n., 144 and "may," see permissibility passive-transformable statements of, 5,91 and tense, 37, 82-4 see also obligation oversubscription, 249 paradoxes of obligation, 72-5, 191, 122-8, 131 Chisholm's, 127-8, 134 Gentle Murder, 127 Good Samaritan, 73-5, 122, 151n. Knower, 74-5, 122, 126 Reparation, 126-7 Ross's, 72-3, 118n., 122 Second Best Plan, 124-5 SuzyMae, 95n., 122n. perfection, 70-1 permissibility, 32-7, 55, 59, 63-5, 68-70, 81-2, 152-4, 164, 186, 209,213-21,238-9,247, 250-1,271 perplexities secundum quid, 229, 230n., 275-6 simplidter, 229 personal possibility, see accessibility possibilism, Ch. 6 passim objections to, 196-203 possible worlds, 31, 45, 147, 242 see also accessibility, of possible worlds praise, 87-8, 247-8 preparedness to act, 42, 260, 264-6, 269 probability, 190n. see also value, probable promises, 18, 38, 59-60, 110-11, 141-5, 166n., 174-6, 178, 222 see 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., 222-4 remorse, see guilt reparation, 126-7, 154, 156-9, 163-4, 184, 186-7 respect, 17-18, 174, 180,245 responsibility, 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, 238 conditional, 183-5 correlativity to obligations, 6—9, 176-80, 183, 187-8, 234 see also duty, correlativity to rights incidental, 181 infringement of, 181, 187 legal, 177n., 185-6 liberties, 177, 187, 238 to life, 184-6 overriding of, 181, 183-7 violation of, 181 risk, 206, 234n. satisficing, 250-1 self-imposed impossibility, 72, 95-101,110,113,117,122, 175 self-imposed necessity, 95-7, 105, 113 Simple Assumption, 132, 133n., 134-8 suberogation, 235, 237-8, 244, 246-8, 252-3 supererogation, 3, 206, Ch. 8 passim, 258-9 supervenience, 216-17 300

transigence, 263—6, 276 unintrusive, 268-74 utilitarianism, 18, 50, 115, 143, 237, 241 value actual, 11,15-17, 109 agent-relative, 18, 27-30, 159-60, 242-3, 256,276 continuity of, 248-53 deontic, 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, 244-53 instrumental, 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, 248-51 see also value, division of perceived, 11, 15-17, 31, 79-81 probable, 11, 15-17, 79-81, 109, 205 substantive views of, 15-19, 23, 27-8,30-1,50,139,277 time-relative, 27-30, 159 vice, 30n., 247 virtue, 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—5 compounding of, 128, 132-5, 202 levels of, 132, 134, 138-40 minimization of, 128, 134, 199 sanctioning of, 193—5, 196n. seriousness of, 17, 29, 99, 138-40 timing of, 100-5 wrongs, 33n., 181 two making a right, 257-8, 262

301