Rationality, Rules, and Utility: New Essays on the Moral Philosophy of Richard B. Brandt

Rationality, Rules, and Utility New Essays on the Moral Philosophy of Richard B. Brandt EDITED BY Brad Hooker Westview Press Boulder, San Francisco, & Oxford All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Copyright © 1993 by Westview Press, Inc. Published in 1993 in the United States of America by Westview Press, Inc., 5500 Central Avenue, Boulder, Colorado 80301-2877, and in the United Kingdom by Westview Press, 36 Lonsdale Road, Summertown, Oxford OX2 7EW Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book is available from the Library of Congress. ISBN 0-8133-1568-9 Printed and bound in the United States of America

The paper used in this publication meets the requirements of the American National Standard for Permanence of Paper for Printed Library Materials Z39.48-1984. 10 9 8 7 6 5 4 3 2 1

Contents 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

Preface Brandt on Rationality and Value Stephen Nathanson Alternatives to Brandt's Theory of Reasons for Action B. C. Postow Brandt on Self-Control Philip Pettit and Michael Smith Brandt on Autonomy Michael Davis Brandt's Methods of Ethics R. M. Hare Moral Concepts and Justified Feelings Allan Gibbard The Evolution of Utility: A Philosophical Journey L. W. Sumner Expectation Effects, Individual Utilities, and Rational Desires John C. Harsanyi The Economics of Knowledge and Utilitarian Morality Russell Hardin The Problem of Group Egoism Gregory Kavka Welfare, Equality, and Distribution: Brandt from the Left William H. Shaw Brandt's Moral Philosophy in Perspective William K. Frankena Comments Richard B. Brandt

vii 1 17 33 51 67 81 97 115 127 149 165 189 207

-vReferences 249 About the Contributors

25 9 26 1

About the Book and Editor -vi-

Preface Moral philosophy is indebted to Richard Brandt. No other twentieth century philosopher has written more influentially either on practical rationality or on rule-utilitarianism. The papers in this collection explore Brandt's views on these subjects, as well as his views about the nature of utility. There has already been a collective festschrift for Brandt and his two longtime colleagues William Frankena and Charles Stevenson (Values and Morals, Alvin Goldman and Jaegwon Kim, eds., Dordrecht: Reidel, 1978). The present volume is not so much a second festschrift as a collection of new contributions to the continuing discussion of Brandtian ideas. The contributors here do an excellent job of setting out the aspects of Brandt's views that they want to commend or criticize. In addition, Brandt begins his "Comments" with a brief but extremely clear summary of his views. Hence, an introductory chapter outlining Brandt's views would be otiose. My own interest in Brandt's work owes a great deal to many enjoyable discussions I've had about it with R. M. Hare, Mark Overvold, and Alan Fuchs. I am grateful to them, and of course to all who agreed to write papers for the collection, including Conrad Johnson, whose early death unfortunately prevented him from completing his contribution. I am also grateful to Anthony Ellis, Peter Vallentyne, and John Heil for generously answering innumerable questions relating to this collection. Finally, Spencer Carr and Marykay Scott, of Westview Press, have been encouraging and helpful from start to finish. Brad Hooker -vii-

1 Brandt on Rationality and Value Stephen Nathanson The ideal of rationality stands at the heart of the philosophical tradition, providing the standards by which all other ideals and beliefs are tested. If belief in the physical world, the minds of others, the existence of God, moral values, or the validity of induction are found wanting by reason's standard, then honest, rational thinkers will be skeptical about these matters. It is not uncommon, however, for those who make these demands to leave their own conception of rationality either unclear or undefended. One of the great virtues of Richard Brandt's 1979 A Theory of the Good and the Right is that Brandt seriously addresses the question "what is rationality?" before moving on to discuss the extent to which morality is rationally justifiable. The theory of rationality he develops is one of the most thorough efforts by a philosopher to construct such a theory. Because Brandt pushes several lines of plausible thought to their limit, his discussion is extremely illuminating. It shows where those who want to follow his lead must go and helps to sharpen the ideas of those who decide against this route. The focus of Brandt's investigation is practical rationality. He wants to know what makes it rational to act in certain ways and irrational to act in others. While he does not discuss the rationality of factual beliefs, it is an important fact about his views that he assumes that empirically based scientific and common sense beliefs and theories are rationally credible. In this paper, I want to examine Brandt's theory of practical rationality. I will begin by sketching several types of traditional theories of rationality and will then describe Brandt's theory, comparing it with its predecessors. Finally, I will argue that Brandt's theory is not ultimately successful and will describe the lessons that I think follow from this assessment. -1-

I. Traditional Conceptions of Rationality One traditional view, which I call the classical conception of rationality, derives from Plato. In several powerful and influential passages in the Phaedo and the Republic, Plato draws a stark contrast between reason and desire, argues that it is the business of reason to determine our actions, and claims that a person who follows reason will seek knowledge and live justly. 1. Through his metaphors for the parts of the soul, Plato gives vivid expression to his belief that reason is what is most human in us, while the desires are portrayed as an insatiable, manyheaded beast that will lead us astray if we heed its promptings. 2. I want to highlight three features of this familiar view of rationality. First, it is a cognitivist conception, since it stresses the use of knowledge and reasoning to determine what we ought to do. Second, it includes a substantive theory of value, since it holds that anyone following reason will pursue certain ends or values (e.g., justice, knowledge, truth). Third, it views desires and emotions in a very negative way, seeing them as opposed to reason. Because our desires mislead us about what is genuinely valuable, they make it harder for us to act rationally. Hedonism provides a second model of rationality. According to hedonists like Bentham, Mill, Sidgwick, and C. I. Lewis, pleasure (or happiness) is the only thing that has intrinsic value and thus the only thing which it is rational to pursue for its own sake. While hedonists differ among themselves about whether there are different grades of pleasure, about whose pleasure one should seek, and about the best ways of obtaining pleasure, all agree that pleasure/happiness constitutes the good and that the value of all else derives from its capacity to produce pleasure or happiness. 3. For hedonists, people act rationally by performing those actions that result in the largest amount of pleasure and the smallest amount of pain. Hedonism differs from classical rationality in important ways. First, although hedonists do not disparage reason or knowledge, cognitive states and processes possess only instrumental value for them. For this reason, actions that produce pleasurable results without reflection are just as good as those based on reflection. Second, hedonists reject the negative appraisal of desires found in classical rationality. Indeed, hedonists appeal to the fact that people desire pleasure as evidence for their view. Mill (1861a: ch. IV) went so far as to say that "the sole evidence it is possible to produce that anything is desirable is that people do actually desire it." While Plato thought we need to resist our desires, hedonists see them as indicators of what is genuinely good. ____________________ 1. See, for example, Phaedo 64a-69e and Republic IV, 439-40. 2. Republic IX, 588c-592. 3. See Bentham 1789, Mill 1861a, Sidgwick 1907, and Lewis 1946.

-2In spite of these differences, classical and hedonistic views of rationality converge on a central issue. Both theories are fundamentally normative since both posit a substantive good that it is rational for all to pursue. If someone showed no interest in pleasure, hedonists would deem that person incapable of rational action, just as Plato would deem a person irrational if she had no interest in truth or virtue. Neither hedonism nor classical rationality is value neutral. A third type of view is the means/end or instrumentalist conception of rationality. According to it, actions are rational if they are likely to achieve whatever ends a person wants. While it is very easy to move from hedonism to instrumentalism, these two views are quite different in character. Unlike hedonists, who take our desire for pleasure as evidence of its value, means/end theorists argue that a thing's being desired confers value on it, that to be valuable is simply to be desired. Means/end theorists reject two of Plato's central assumptions: (a) that reason tells us what is good and (b) that desire is an obstacle to the good. From the means/end perspective, there can be neither value nor rational action without desire. When Hume's statement that reason is the "slave of the passions" is the most famous expression of the view that it is the passions that determine which goals we should seek. Reason merely tells us about possible objects of desire and instructs us how to obtain them (Hume 1738-40: bk. II, part III, sect. III) Means/end theories, then, provide a purely formal rather than a substantive conception of the good. They tell us it is rational to pursue what we desire but do not tell us what to desire. Finally, I should briefly mention three other views. First, value pluralists believe that there are many substantive goods and that actions are rational if they are likely to produce genuinely desirable results. Second, the "evil avoidance" theory suggests that while reason does not require us to pursue any specific goods, it does require us to avoid evils like death, injury and pain. 4. A third idea identifies rational actions with those that are in the agent's self-interest. While egoistic theories differ on what things are good, they all agree that reason requires only that one obtain things of value for oneself.

II. Brandt's Theory I have sketched some basic types of theory in order to place Brandt's theory in context and call attention to some issues to have in mind when we examine theories of rational action. These are the theory's view of the nature and value of cognitive processes as they bear on rational action, its position on whether ____________________ 4. This view is defended in Gert 1988. -3-

people's goals and desires are subject to rational appraisal, and finally its stand on the status and value of desires and emotions. Like classical rationalists, Brandt puts cognitive processes at the heart of his theory. He tells us that in its primary, most general sense, the term 'rational' ... refer[s] to actions, desires, or moral systems which survive maximal criticism and correction by facts. (Brandt 1979: 10) He even suggests that the expression "fully informed" might do as well, making it clear that rationality is primarily a cognitive concept. 5. Beginning with this overall conception, Brandt distinguishes three senses of "rationality". A person's action is rational to a first approximation, if and only if it is what he would have done if all ... the cognitive inputs influencing decisionlaction ... had been optimal as far as possible. (Brandt 1979: 11) Brandt's point here is that an action is rational (to a first approximation) if it is based on the best information available to the agent and the agent's reasoning is not illogical. Brandt calls this definition a first approximation because actions and decisions are not generated by cognitive inputs alone. They also require affective inputs, desires and aversions to particular things, states, and conditions. For actions to be fully rational, the desires or aversions that motivate them must also be rational. With this in mind, Brandt defines a second sense of rational. He calls a desire or aversion 'rational' if and only if it is what it would have been had the person undergone cognitive psychotherapy. (1979: 11) Since "cognitive psychotherapy" is nothing more than a process by which a person repeatedly calls to mind all relevant available information, this second sense of "rational" is essentially an extension of the first. 6. In the first sense, actions and decisions are to be maximally influenced by facts and logic, while in the second sense desires are to be subject to maximal cognitive influence. Finally, if both conditions are met, the action or decision is "fully rational" (1979: ii). Brandt's theory, then, like that of classical rationalists, is a cognitivist conception of rational action. Rational actions are those that we would perform if we reason correctly and are maximally influenced by available relevant information. ____________________ 5. In a later paper, Brandt writes: "In earlier drafts of the book, I used the term 'fully informed' [instead of 'rational'].... What I wanted to find is what desires and choices

6.

would survive maximal criticism by facts and logic." (Brandt 1989c: 125) For Brandt's fuller description of cognitive psychotherapy, see his 1979 111-13. -4-

Unlike the classical rationalists, however, Brandt rejects the idea that a theory of rationality should prescribe or endorse substantive values. His theory is supposed to be value neutral. As he writes, the proposed definition of 'rational' ... does not import any substantive value judgments into the concept 'rational'. (1979: 13) This does not mean that no value judgments will follow from the application of the theory. If an action is fully rational, Brandt thinks we can rightly call it "the best thing to do". But all that means is that the action would be chosen under ideal cognitive conditions. In effect, Brandt wants to derive values from facts. He wants no substantive value judgments among the premises of his theory. They are to be theorems rather than axioms or postulates. He criticizes philosophers who begin with "intuitions" about what is good or right (e.g., John Rawls), since these intuitive beliefs may be unwarranted, as well as those (e.g., R. M. Hare) who appeal to ordinary language, since language incorporates value commitments and thus begs the question of which values should be rationally preferred. Brandt's theory of value, then, is formal rather than substantive and naturalistic rather than normative. It is formal with respect to values because it does not specify in advance any particular things as having value. It leaves open the question of what "values" will emerge after beliefs and desires have been subjected to optimal cognitive inputs. 7. It is naturalistic in that Brandt wants to show that one can actually derive value judgments from neutral, empirical facts if one picks the right set of facts and the right method of derivation. The word "naturalism" has been applied to views that define rightness or goodness in terms of "natural", i.e., empirically discoverable properties. While some philosophers have assumed that this could be done by finding the correct meaning of words as we now use them, Brandt rejects this strategy. Instead, he opts for a method of "reforming definitions", recommendations about the best way for terms to be used. Rather than telling us how value terms are actually used, then, Brandt recommends a way of understanding them that enable us to reduce them to empirical concepts. In particular, he believes that attention to facts of human psychology will allow us to give verifiable meaning to the concept of "rational" and that we can then use this concept to answer questions about what things have value. In attempting to demonstrate "how psychological theory can be used to establish normative principles", Brandt sees that he is part of a long philosophical tradition. As he writes,

____________________ 7. In a later discussion, Brandt describes cognitive psychotherapy procedures as "the only ones which comprise a serious value-free way of criticizing intrinsic desires." (Brandt 1989c: 127) -5of course, it is not news that philosophers should think they can squeeze normative conclusions out of psychology: they have been trying to do this since Epicurus, and Hume is a prime example of the psychologizing of moral philosophy. (1979:2) Brandt's overall program is empiricist in orientation. He believes we know how to verify scientific and factual claims, and he wants to see how far we can get in the normative realm by using these same empirical methods. For him, there is no alternative because these are the only rational methods that exist. Brandt calls the alternative to his approach "intuitionism", using that term to refer to philosophical theories that assume that we already have well-justified normative opinions and that our theories must cohere with these opinions. He strongly rejects this, urging that We must avoid intuitionism even if this were to mean (as it does not) that we must end up as complete skeptics in the area of practice. (1979:3) Forced to choose between skepticism and the acceptance of some fundamental value premises, Brandt would choose skepticism. 8. Given this set of views, it is clear that Brandt's account is a version of an instrumentalist or means/end theory of rationality. Reason neither prescribes nor proscribes particular goals or values for people. Their desires do that for them. What reason does is to subject these desires to relevant information. If, at the end of this process, someone wants something and takes effective steps toward getting it, then that person acts rationally, no matter what it is that she wants. Likewise, if a desire is "extinguished" when subjected to relevant facts, then it was indeed an irrational desire. Any desire that would not extinguish after "cognitive psychotherapy" is a rational desire. Finally, unlike classical rationalists, Brandt casts no aspersions on the affective aspects of human life. Without desires, we would not be motivated to do anything. Desires, in fact, are the sources of value, since things acquire value by becoming objects of desire. While classical rationalists see reason as discovering what is good, Brandt sees desire as creating whatever goodness things possess. His theory is subjectivist in the sense that it makes value dependent on desire. Nonetheless, it is objectivist in two senses. First, given that someone desires something and that her desires have been suitably influenced by cognitive processes, it is an objective fact that the desired object is good (for that person). Second, since the ultimate test of a desire's rationality is that it would not extinguish if a

person were to subject it to cognitive psychotherapy, people can be mistaken about what is good for them. They may act on desires that they would ____________________ 8. For a similar use of the term "intuitionism", see Chisholm 1977: 23. Chisholm uses the term in a discussion of "the problem of the criterion" that is quite relevant to Brandt's methodological claims. -6not have if they had undergone cognitive psychotherapy. So, simply feeling a desire or thinking it to be rational does not make it so. The success of Brandt's theory would constitute a formidable philosophical achievement. It would provide a way to determine the rationality of people's actions without making controversial value assumptions and would show how both rationality and value are rooted in human psychology. Claims about rational action would be complex, since they deal with what a person would want under ideal conditions, but they would nonetheless be empirically supportable or criticizable. Finally, the theory shows how both cognitive and affective processes enter into the rational assessment of actions, and, unlike other instrumentalist views, provides a way to subject desires to rational criticism.

III. Doubts and Criticisms While Brandt's theory has many virtues, I believe that it is ultimately unsatisfactory and that any theory of its type must fail. My own view is that the rationality of action cannot be assessed without making substantive judgments of value. Because rationality is an irreducibly normative concept, any theory that attempts to define it in a value neutral way is mistaken. Since this is precisely what Brandt sets out to do, I do not think his project can succeed. 9. Before turning to specific criticisms of Brandt's theory, let me briefly clarify my own view. I hold that an action is rational for a person to perform if its foreseeable consequences are better than those of other actions the agent might do. An agent acts rationally if she performs the action with the best foreseeable consequences and does so because she is aware that it is the best thing to do. 10. Like Brandt's theory, this view includes a cognitive element, since assessing the rationality of actions requires us to take into account evidence about its consequences and the consequences of alternative actions. Likewise, when judging the agent's rationality in acting, we must consider whether the agent's awareness of foreseeable consequences played a role in causing the action. Actions are irrational if they are performed in spite of the agent's being in a position to foresee that other actions would have better consequences. In addition, if an agent does the best thing but does it through blind luck, then he has not acted rationally (even though the rational action has been performed).

Brandt need not disagree with what I have-stated so far. Our views differ, however, because my cognitive requirement is less stringent than his and my value requirement is more stringent. We can see the differences by considering ____________________ 10. For elaboration and defense of this view, see Nathanson 1985. 9. For a forceful version of this type of criticism, see Bernard Gert's discussion of an earlier work by Brandt in Gert 1977: 286-91. See, too, Gert 1988: especially 25-28. -7one of Brandt's examples of an irrational desire. Brandt calls a desire "mistaken" if it originated in false beliefs. He describes a student who begins work toward a Ph.D. because he falsely believes that his parents will be disappointed if he does not. The student now "wants an academic life for itself (but not because he has found it satisfying)" (Brandt 1979: 115). Is this a rational course of action? Brandt says that because this desire originates in a false belief about the student's parents, it might extinguish if, in cognitive psychotherapy, the student focused on the erroneous basis of his career choice. Whether the desire is rational or irrational depends on whether or not it would extinguish under these conditions. It is rational if it would not and irrational if it would. Brandt's cognitive requirement is more stringent than mine because I don't believe that a person has to undergo cognitive psychotherapy to determine whether her desires are rational or irrational. Nor do I believe that the criterion of rationality is whether the desires would extinguish or survive this process. Furthermore, facts about how a desire originate are often irrelevant to judgments about its rationality. What is relevant to judging the rationality of action is determining whether the foreseeable effects of the action are good or not. Predicting and evaluating these "future" facts is the only thing we need to attend to. In thinking about a career choice, we should ask whether the person enjoys the work, what other options there are, whether the costs in time, effort, and training are too great, whether there are other pressing reasons to choose a different alternative, etc. We do not need to ask about the conditions under which the desire originated or whether it would be altered by cognitive psychotherapy. The key question is: "Will this work out better for the person than other available options?" This is not to say that we should never ask about the origins of desires. Nor is it to deny cognitive psychotherapy is ever useful. It is significant, however, that we would normally consider or recommend cognitive psychotherapy only in cases where we have already decided that a desire is irrational. If, for example, someone is bored, under stress, takes little satisfaction from academic life, and yet persists in pursuing this career, we might inquire why she feels such a need to pursue a course that is making her seriously unhappy.

In this case, however, we already believe the desire to be irrational and recommend cognitive psychotherapy (or other measures) as a way of getting rid of the desire. We do not use facts about either the origin of the desire or its possible extinction by cognitive psychotherapy as a criterion of rationality. Likewise, if a person is happily pursuing an academic career, we would be unlikely to recommend either historical inquiry or cognitive psychotherapy as a way of "confirming" the rationality of her desires. We think we can judge the desire to pursue this career independently of whether it is based on false beliefs. Brandt's criterion, then, is excessively stringent from a cognitive point of view, since it requires knowledge beyond that which concerns the likely effects of -8pursuing this career. Nonetheless, the facts that he wants considered do not seem relevant, useful, or necessary in this case. Satisfying the cognitive psychotherapy test (either actually or hypothetically) is not a necessary condition of a desire's being rational. In addition, satisfying this test does not appear to be a sufficient condition of a desire's being rational. Again, imagine our subject being very unhappy in her pursuit of an academic career, undergoing cognitive psychotherapy, and finding that she still has a powerful desire to pursue this career. Nonetheless, doing so appears to produce stress, anxiety, and unhappiness. Her satisfactions are at best minimal and do not seem to justify either the suffering she undergoes or the sacrifice of potential happiness doing other things. In such a case, the desire seems to be irrational, even though it survives cognitive psychotherapy. 11. While this conclusion seems straightforward enough, it commits us to stronger value claims than Brandt is willing to make. To judge her desire to be irrational, we must be willing to say that pursuing an academic career is a bad for this person, even though she wants to do it, is fully aware of all relevant facts, and has subjected her desire to cognitive psychotherapy. Brandt, however, takes the agent to be the final authority on the value of her plan, as long as she has fulfilled the full information requirement he describes. She knows all the facts, including the high probability that she will continue to be frustrated and unhappy and the origin of her desire in a false belief about her parents' wishes. Finally, after repeated representations of these facts, she still wants an academic career. In this case, Brandt is committed to saying that it is rational for her to pursue an academic career. 12. This example makes clear the distinction between substantive and formal theories of value. If one is inclined to say that pursuing an academic career is good for this person because it is what she wants, then one is equating the good with objects of desire. On the other hand, if one thinks that in spite of this person's desires, she is making a mistake, one is committed to a substantive conception of value.

Note that a hedonist would have no problem saying that the person is making a mistake (provided that there is some other option which would yield greater happiness for her). In some places, Brandt appears to recognize hedonistic considerations. He comments, for example, that ____________________ 11. For my own earlier criticisms of the "extinguishability criterion", see Nathanson 1985: 107‐ 13. For other criticisms of Brandt's criterion, see Daniels 1983 and Lemos 1984. For a general discussion and criticisms that bear on Brandt's theory, see Brink 1989: ch. 3. 12. Allan Gibbard, though sympathetic with features of Brandt's program, rejects the full information criterion of rationality. See his helpful discussion of the "ideally coherent anorexic" who would rather starve than have "a figure plump enough to sustain life" in Gibbard 1990: 165-66, 171‐ 72, 175-77. -9the desire to get the Ph.D. will not extinguish if by this time the student has other reasons—for instance, finds academic work inherently satisfying. (1979: 116) Likewise, he thinks that the desire for an academic career may weaken if the person "reminds himself that a Ph.D. is for him very hard work, that he was not cut out to be an academic, and so on" (ibid.). In places, then, he does treat hedonistic considerations as relevant and does focus on foreseeable results of actions. Nonetheless, while these facts may influence the person's desires, they need not, and if they fail to have influence, Brandt must regard them as irrelevant. For a hedonist, however, these thoughts should be decisive. Pursuing one's desires is not rational if one knows that "satisfying" them will not in fact be satisfying, i.e., that it will yield no pleasure or happiness. Brandt's theory precludes him from saying this.

IV. Brandt on Hedonism While I do not believe hedonism is a fully adequate theory of value, it has advantages over Brandt's desire-based theory. It can account, as Brandt's theory cannot, for the irrationality of desires that do not extinguish after cognitive psychotherapy. It can do this because it is a substantive theory of value. Brandt himself seems to embrace hedonism at various places in his 1979 book, and it is worth considering both how close he comes to the theory and why he rejects it. Brandt believes that pleasurable and painful experiences play a large role in the genesis of desires. According to his theory, experiences and objects are either "natively" pleasant (or unpleasant) or have "learned pleasantness" (or learned unpleasantness). Natively pleasant objects are essential to our development because learned pleasantness is

acquired as a result of conditioning and association with what is natively pleasant. What we naturally find pleasant becomes emotionally connected to other (originally neutral) things, and these other things become pleasant derivatively through processes of association and conditioning. The same is true for unpleasant things (1979: 89-95). Finally, to experience something as pleasant is to desire its continuation. Therefore, we necessarily desire what we find pleasant (1979: 95-98). One might think that this account implies that people only desire what is pleasant. Brandt denies this. For him, even after association or conditioning produce a link between a pleasant A and some neutral B, B itself need not be something that brings pleasure. Association and conditioning can lead one to desire B, even though B is not thought to be pleasant. While there must be some causal link with pleasure to explain the genesis of a desire for something, the object of desire may be something that is neither natively nor derivatively pleasant. Here is -10Brandt's explanation of the difference between his genetic account of rational desirability and the hedonist's view that rational desirability is a function of expectable pleasure: Direct conditioning ... produces valence (i.e., attraction or aversion] by pairing the idea of an outcome with some pleasant or unpleasant state.... Because the past pairing of the idea of an outcome with some pleasant or unpleasant state was artificial [i.e., not the product of directly experiencing the outcome], ... the incentive value which the conditioning has attached to the idea of the outcome need not correspond with what a rational person would project as the hedonic promise of that kind of outcome....(1979: 136) The crucial point here is the gap between "incentive value" (how much we want something) and "hedonic promise" (the pleasure we can expect). If there is such a gap, then surely psychological hedonists are mistaken in their view that human action is always motivated by a desire for pleasure. But rational hedonists (or, in Brandt's terminology, "ethical hedonists") who affirm that pleasure is the only thing worth pursuing for its own sake need not accept Brandt's facts as a refutation of their view. From their perspective, if conditioning sometimes leads people to value things apart from their hedonic promise, that explains why people sometimes act irrationally. It does not show that the pursuit of nonpleasurable outcomes is rationally defensible. While Brandt explicitly rejects hedonism as a theory of value, there are times when he sounds very much like a hedonist, and he even recommends his theory on what seem to be hedonistic grounds. In explaining why we should care whether our actions and desires are rational (in his sense), Brandt writes:

Having rational and not irrational desires is a necessary condition of (or at least helps in) avoiding dissonance and unhappiness. Therefore you should be favorable towards rational desires and take an unfavorable attitude toward irrational ones (1979: 159) In another place (p. 161), he (correctly) notes that there are circumstances in which it would be irrational to stifle an irrational desire, if the desire is so strong that the individual would be made deeply unhappy ... by its suppression. This comment can only be understood, however, if we assume a connection between acting rationally and producing happiness. We cannot understand it in terms of Brandt's own extinguishability criterion. A person may rationally act on an irrational desire even though it brings bad consequences and even though he would be better off without the desire if attempting to get rid of the irrational desire would lead to even worse consequences, i.e., greater unhappiness. This is a -11judgment that a hedonist can make, but it is hard to see how it is consistent with Brandt's theory. Finally, Brandt seems to concede that the desirability of happiness/pleasure is fundamental when he says that his argument for taking a negative attitude toward irrational desires "does not work" for a person who is "uninterested in happiness and avoiding dissonance" (1979: 159). In this passage, Brandt seems to suggest that we should be interested in having rational desires (and avoiding irrational ones) because doing so is more likely to make us happy. These are not the only reasons why one might that Brandt is a hedonist. Another reason for possible confusion is that when Brandt talks about benevolence, he defines it as a desire for the happiness or welfare of other people and not simply as a desire that they get what they want. This makes for an odd asymmetry between his account of the good for oneself, which is simply the satisfaction of one's own rational desires (i.e., those that survive cognitive psychotherapy) and the good for others, which is their achievement of happiness and well-being. Discussing benevolence, he writes, what we seem to care about securing for other persons (e.g., our own children) is their happiness; and we seem to care about getting them what they want (or would want only if they knew more, etc.) only to the extent we think that so doing will bring them happiness or avoid distress and depression. (1979: 248) Because he defines the well-being of others as happiness, Brandt is led into the position of simultaneously attacking desire theories of the good of others, while defending just

such a view with respect to the good for oneself (see Brandt 1979: 249-53 and Brandt 1982).

V. Substantive Values and Normative Rationality Given the evident appeal of hedonism as a theory of value and Brandt's own apparent appeals to it, one might wonder why he rejects it. There are two reasons for this. First, hedonism is a substantive, normative theory of just the sort Brandt wants to avoid. It is based on the "intuition" that pleasure and happiness are good and is not derivable from neutral facts about how our actual desires are formed. (Recall the gap between "valence" and "hedonic promise".) Hence, hedonism is inconsistent with Brandt's naturalistic project. I believe that this is a misguided reason. Brandt's second reason is more compelling, however. It is that fully rational persons do not just want happiness.... [T]hey want achievement, the admiration of other persons, and so on. (1979: 246) -12Brandt is objecting here to the monistic feature of hedonism, its reduction of all values to one. For hedonists, everything that is rationally desired is desired as a means to pleasure or happiness. Yet, it is an important fact that people desire things apart from their tendency to produce pleasure. Because desire-based theories are pluralistic, they can account for this fact. What people desire (or would desire after cognitive psychotherapy) is good, even if it does not produce pleasure or happiness. The difficulty with this approach, however, is that it is too inclusive, too pluralistic, too indiscriminate. It grants value to things that ought not to be valued. In earlier work, I argued that rational actions are those with the best foreseeable consequences and that "best" is to be understood in terms of the production of a plurality of goods (Nathanson 1985: 139-46). Pleasure and happiness would certainly be on this list, but so might knowledge, friendship, moral virtue, creativity, and other values. An appropriate label for the kind of value theory I have in mind might be critical pluralism. Such a theory is substantive because it rejects the pursuit of a value free analysis that reduces the good to what is desired. It is pluralistic because it does not attempt to reduce all value to a single primary one. It is critical because it insists that statements of value can be subject to rational criticism. 13. When we judge a person's actions to be rational, we are appraising them and judging them to be appropriate to the attainment of that person's ends. In addition, however, we are saying that the person's ends are ones that we can understand as having value for that person. This is not the same as fully endorsing a person's ends or wanting them for ourselves. For example, if someone desires great power, we may not share or endorse

their placing of power at the center of their goals. Nonetheless, we can recognize power as something of value and can understand its appeal. The same is true of artistic achievement or fame. If we can understand the good involved in creativity or recognition, then we can affirm the value of these as goals, even if we ourselves do not pursue them in our own lives. Judgments about the irrationality of actions arise in a number of ways. First, we might judge someone's action to be irrational because it is foreseeable that the action will fail to produce ends that she desires. Second, we may judge an action to be irrational, even if it effectively promotes a desired end, if we regard the end as without value and not worthy of pursuit. If the end has no value, then effective steps to bring it about cannot be rational. A third kind of judgment of irrationality is more complex. Someone may act effectively to achieve a worthwhile end but may attach too much importance to that end and hence may fail to give proper due to other worthwhile ends. The ____________________ 13. On the criticism of values, see The Ideal of Rationality, ch. 12. For related views, see Nagel 1979 and Taylor 1982. -13pursuit of power or recognition, for example, may become an obsession, leading people to act in ways that undermine the value of their lives and the lives of others they care about. A person's devotion to art or a political cause (or even philosophy) may lead to neglect of other legitimate needs and interests. In some cases, we may regard such sacrifices as noble and important, as the necessary cost of greatness. In other cases, we may simply think them irrational. Such judgments are difficult to prove and are enmeshed with complex issues of fact and value. Nonetheless, they are an important type of judgment about irrationality, and a theory ought to recognize their intelligibility, even if it cannot provide a rule for determining their truth. Any adequate theory of rational action must leave room for the criticism of desires, ideals, and values. I have suggested elsewhere that we can criticize ideals and nonhedonic values for being inconsistent, harmful, immoral, bizarre, or based on false beliefs (Nathanson 1985: ch. 12). The point I would stress here is that we cannot judge the rationality or irrationality of actions without making some normative assumptions ourselves. If we can find no value in a person's aims, we must dismiss them as irrational (even crazy), even if they are pursued by the most effective means. In suggesting that such normative beliefs are a necessary part of judgments about rationality, I do not mean to overlook the possibility that we may narrowmindedly impose our own value judgments on others, making determinations about their good when we don't fully understand them or the things they take to valuable. What a person desires for herself is an important indicator of value, and respect for the autonomy of others requires us to take it seriously. Nonetheless, people's desires, even when they are based on full

information, are but the first and not the last word on what is good for them. Brandt recognizes this when he says that benevolence involves a concern for well-being and not simply for desire satisfaction. He loses sight of it, however, in his discussion of rationality. My point is that judgments of rationality presuppose beliefs about a person's good in the same way that attempts to act benevolently do. They involve substantive judgments of value, not value neutral judgments about the satisfaction of desire.

VI. Is Intuitionism Avoidable? While Brandt's account contains many important insights and arguments, his stringent methodological ideals prohibit him from accounting for the substantive normative components that any adequate theory of rational action must include. Brandt's rejection of normative "intuitions" could not be more forceful. As I cited above, he claims, "We must avoid intuitionism even if this were to mean (as it does not) that we must end up as complete skeptics in the area of practice." Brandt's desire to avoid intuitionism is not foolish. If his program succeeded, we -14could actually provide proofs of value claims. We could show how the methods of empirical inquiry can be used to establish what is rational for us to do. But the theory does not succeed, and this might suggest that we should become value skeptics. To make this choice, however, is basically to say that if a foundationalist epistemology cannot be constructed for values, then no value beliefs are rationally justified. This appears to be Brandt's message. 14. One might wonder, however, whether Brandt himself fully escapes the kind of intuitionism about rationality that he seeks to avoid. Admittedly, he avoids many of the standard normative assumptions of other theorists. Likewise, in rejecting appeals to ordinary language, he avoids many of the implicit commitments that are built into our use of normative terms. Nonetheless, there is a basic normative assumption that underlies his whole approach, his enthronement of scientific inquiry as the paradigm of rationality. Logical reasoning and empirical research are, for Brandt, the essence of rationality. If value beliefs can be derived from them, they can be rational. If value beliefs are not so derivable, then they must be consigned to the realm of the non-rational. But is not Brandt's method rooted in normative assumptions about the best ways to know about the world? Isn't Brandt an intuitionist about the rationality of factual beliefs? Does he not assume that scientific inquiry provides beliefs that all rational people will hold about the nature of the world? 15. As philosophers, however, we know that scientific rationality can be called into question. It is not going too far to suggest that the lesson of several hundred years of epistemological inquiry is that if we start with nothing, we end with nothing. The dream of foundationalism leads to the nightmare of skepticism, not only in ethics but in science as well.

Beginning with Descartes, foundationalists have defended a particular ideal about the establishment of beliefs. They had to be based on self-evident premises or derivable by deductive steps from such premises. Moreover, each person had to be able both to understand the proof for himself and to lay it out for others in such a way as to induce the same belief in any rational being. 16. Discussions of this project over the last decades, however, have led to a very different picture. 17. Our best supported scientific theories have the following properties: (i) they are based on fallible premises; (2) they are derived by methods of inquiry and reasoning that do not guarantee truth; (3) the arguments for ____________________ 14. For an impressive set of arguments against foundationalism in ethics as well as a worthwhile discussion of what makes a theory intuitionist, see Brink 1989: especially chs. 5 and 6. 15. I am using intuitionism here in something like Brandt's sense. The term has been used in a number of distinct ways by thinkers such as Sidgwick, Brandt, and Rawls. For a useful discussion, see Brink 1989: 107-13. 16. This ideal is most clearly expressed in Descartes' Rules for the Direction of the Mind. 17. Much of this discussion crystallized around Thomas Kuhn's influential discussion of scientific inquiry (Kuhn 1962). -15them are typically unknown to most people and can only be learned with difficulty; (4) if we present the arguments to other rational persons, they need not accept the conclusions we accept. If this is the model for our factual beliefs about the world, then it seems unnecessary to require more from our beliefs about values. Faced with this picture and the failure of the foundationalist model of individual rationality ("only believe what you yourself can prove"), one might well opt for skepticism as Brandt suggests. But skepticism only follows if we accept the idea that foundationalism provides the only model of rational thought, and there is no need to do that. 18. Instead, we can follow John Dewey, who saw that the quest for certainty is a dead end and championed an alternative model (Dewey 1929). Dewey began with the problems we face as human beings and argued that the test of a beliefs value is its role in solving our problems. The identification of problems and solutions, however, is an inherently normative activity. From this perspective, normative premises appear both in the motivation for factual inquiry and in the justification of factual beliefs. Hence, the requirement that we purge all normative beliefs from our inquiry into the realm of the normative is a misguided and unnecessary constraint. Ultimately, there is no way to construct a theory of rational action and rational desire without having a broader theory of rationality. If the rationality of scientific inquiry itself

requires normative premises, then there is no reason to accept the severe constraints Brandt's program imposes on inquiry into ethics and practical rationality. If epistemology cannot be naturalized, there is no reason to expect that ethics can be. 19. My conclusion, then, is that the subject matter of the theory of rationality is irreducibly normative. We cannot identify actions or desires as rational without making judgments about the worthiness of the ends for which they are performed. To be rational requires more than making the best use of all information. It requires being directed toward things of value. Brandt's theory fails because he refuses to acknowledge this. Nonetheless, the type of theory he seeks to establish has many attractive features. To have developed it so comprehensively is a great achievement, even if the primary lesson some of us derive is about the limitations of such a view rather than the conviction that it is true. 20. ____________________ 18. For a non-foundationalist attempt to defend science as the paradigm of rationality, see Brown 1988. I discuss Brown's theory in Nathanson 1991. 19. For a defense of this view, see Putnam 1981: especially ch. 6. 20. My thanks are due to Brad Hooker for helpful comments on a draft of this paper. -16-

2 Alternatives to Brandt's Theory of Reasons for Action B. C. Postow The label 'theory of reasons for action' can refer to a descriptive/explanatory theory, or it can refer to a normative theory. A descriptive/explanatory theory is concerned to give an account of reasons that can be cited to predict or explain action. These reasons are the (possibly confused or wrongheaded) ones for which people act. A normative theory of reasons, in contrast, is concerned to give an account of good reasons. These reasons are considerations that, although they may be overridden, play some role in making such and such an act one that the agent rationally ought to do. A consideration can be a reason, in this sense, regardless of whether the agent has any tendency to act on it or even to acknowledge it as something that it would be rational for her to consider. We shall be concerned exclusively with normative theories of reasons for action, not with predictive/explanatory theories. According to a common theory of reasons for action, which I shall call the Actual Desires View, reasons depend on the agent's actual desires. Richard Brandt rejects the Actual Desires View in favor of what I shall call the Ideal Desires View. On this view, reasons for action depend on the desires that the agent would have, were she to undergo procedurally defined ideal deliberation. The Actual Desire View and the Ideal Desires View can be combined in two different ways to yield two hybrid views: the Intersection View and the Union View. On the Intersection View, reasons for action depend on the desires in the intersection of the set of the agent's actual desires and the set of her ideal desires. On the Union View, reasons for action depend on the desires in the union of those two sets. After presenting these four views a bit more fully, I shall argue that every one of them succumbs to some fatal flaw or other (although the Ideal Desires View holds out the longest.) I shall then sketch a new view that shows promise of avoiding the flaws. -17The Actual Desires View defines rational behavior as behavior that maximizes (or at least produces a satisfactory amount of) the satisfaction of (some subset of) the agent's desires. 1. This view is perhaps the most natural way of generalizing the common-sense observation that people who enjoy climbing mountains have a reason to climb mountains, while other people may not have any reason to do so. This view of rationality is attractive to economists. One economics textbook expresses it this way: "We assume ... that individual behavior is efficient in the sense of aiming at, and actually achieving, maximum satisfaction. At least, this enables us to define what is meant by 'rational' behavior." (Reynolds 1976: 91)

On the Ideal Desires View, reasons for action correspond to the agent's ideal (not necessarily actual) desires, i.e. the desires that she would have were she to undergo ideal deliberation. Brandt's theory is perhaps the best developed version of the Ideal Desires View. Brandt says (1979: 113) "I shall call a person's desire, aversion, or pleasure 'rational' if it would survive or be produced by careful 'cognitive psychotherapy' for that person." 2. Cognitive psychotherapy is the "process of confronting desires with [all available] relevant information, by repeatedly representing it, in an ideally vivid way, and at an appropriate time." (Brandt 1979: 113) Brandt holds that cognitive psychotherapy would extinguish any desire that had been produced artificially by conjoining the thought of its object with a reward. Thus we can say that artificial desires would fail to ground reasons for action according to Brandt's theory (1979: 118). On the other hand, an agent's desire for a thing that she natively likes does provide that agent with a reason for action, for the desire would not be extinguished by cognitive psychotherapy. Thus, for example, an agent has a reason to avoid pain and to seek novel stimuli (1979:131). On the Intersection View, a desire grounds a reason for the agent to act only if the desire is both actual and also ideal in the sense that it would not be extinguished by ideal deliberation. Derek Parfit's Critical Present-aim Theory (CP) seems to be a version of the Intersection View (supplemented by two theses to which we shall turn presently). Parfit says (1984: 119) "Suppose that I know the facts and am thinking clearly. If my set of desires is not irrational [e.g. by being intransitive], what I have most reason to do is what would best fulfill those of my present desires that are not [intrinsically] irrational." The first sentence quoted seems to make this view an example of the Intersection View rather than of the Actual Desires View. One's present desires ground reasons for one to act only if one knows the facts and is thinking clearly. Those conditions would be otiose if they were meant merely to enable one to make the best calculation about how to ____________________ 1. The subset is generally taken to exclude actual past desires and sometimes future ones as well. If Parfit's thesis (a) below is adopted, the subset excludes any intrinsically irrational actual desires. One could specify similar subsets of the desires that would result from ideal deliberation. 2. Arguably, Brandt's theory is not a pure version of the Ideal Desires View since he holds that it can be rational to satisfy an irrational desire that one cannot get rid of (1979c 161). -18satisfy one's desires, for the second sentence says that what I have most reason to do is what would best fulfill my present desires—not what I (would) calculate would best fulfill them. So Parfit's requirement that I must know the facts and be thinking clearly in order for my present desires to ground reasons seems best interpreted as a requirement that my present desires must themselves survive scrutiny in light of knowledge and clear thinking if they are to ground reasons for me to act. In any case, interpreting his requirement this way yields a version of the Intersection View.

Parfit's Critical Present-aim theory (CP) can be thought of as the Intersection View supplemented by the following two theses: (a) if any desires are intrinsically irrational, then they cannot provide reasons for action even if they are in the intersection of the agent's ideal desires and her actually present desires; and (b) if there are any desires that are rationally required, then they provide reasons for action even if they are neither in the agent's ideal desires nor actually present in her (Parfit 1984: 119). 3. (Of course a holder of any of the other three views mentioned might also choose to adopt Parfit's theses (a) and/or (b).) The Union View states that a reason for action is grounded by any desire that is either actual or ideal. 4. It has been objected that the Union View is incoherent because it allows conflicting reasons in those cases where my current desires give me reason to do something while my ideal desires give me reason not to do it. But this result does not make the view incoherent. It would indeed be incoherent to say that there is most reason for a given agent to do something and also most reason for her not to do it. But the Union View is not committed to that result. It merely implies that there can be reasons both for and against doing something. Even the Actual Desires View implies that, however, for actual desires can also conflict with each other. The Union View might naturally suggest itself as an improvement on the Actual Desires View and the Intersection View. Neither of those views allows us to say that there is reason for an agent to do what would satisfy a desirable goal when it is only through ignorance or confusion that she lacks a desire for it. For ____________________ 3. Parfit holds that all theories of practical rationality can be perspicuously stated as versions of CP (Parfit 1984: 194). He seems mistaken in that, however. For example, Shelly Kagan points out that CP could not perspicuously express a theory according to which future desires directly generate reasons for action without the mediation of present desires (Kagan 1986: 752). Parfit's reply to Kagan on this point seems unsuccessful to me. He refers us to pages 121-22 of Reasons and Persons, where he states that some desires or aims can be rationally required according to CP (Parfit 1986: note 35). The rationally required desires described on those pages, however, are required of all agents, not just of particular agents in virtue of their particular future desires. Thus CP does not express the theory described by Kagan. 4. Why not consider the power set of the union, i.e. the set of all subsets of the union set? It is of course logically possible that reasons for action depend on any subset of this set. But this logical possibility does not seem to me to constitute an alternative view about reasons for action, for it includes no motivated way of specifying which subset reasons depend on. -19example, neither would allow us to say that there is a reason for a particular agent to get an education, if she would keenly desire one if her horizons were wider. The Union View remedies this defect, for it holds that ideal desires that are not actual nevertheless ground

reasons. In any case, the Union View suggests itself to fill out logical space after we have encountered the Actual Desires View, the Ideal Desires View, and the Intersection View. I shall now argue that there are serious defects in all four of the views that we have been considering. Here is an argument against both the Actual Desires View and the Union View. Typically, desires that would fail to survive deliberation (for example because they are based on ignorance or confusion) are thereby subject to rational criticism; they are to some extent mistaken or irrational. Such desires, however, ground reasons for action according to the Actual Desire View and thus according to the Union View as well. Further, there is no guarantee that there will be contrary reasons that override those reasons. Thus it is possible, according to the Actual Desires View and thus according to the Union View as well, that what there is most reason for an agent to do may happen to depend on desires that are, for her, mistaken or irrational. But that is unacceptable. Therefore, both these views are unacceptable. Perhaps it will be objected that the argument goes too fast, for it relies on the premise that desires that would fail to survive ideal deliberation are subject to rational criticism. That premise presupposes that there is an acceptable procedural definition of ideal deliberation. Since there may be no acceptable procedural definition of ideal deliberation, a presupposition of the premise is yet to be established. To this objection I would reply that we need not possess an acceptable procedural definition of ideal deliberation to acknowledge that some of our desires are irrational ones for us to have because they would fail to withstand our own scrutiny of them. 5. (They might fail, for example, because they would be seen to depend on beliefs that are inconsistent with most of our beliefs.) The Actual Desires View and the Union View make no provision for that acknowledgment, and so should be rejected. The Intersection View is not subject to the foregoing problem, for it does not let desires determine reasons unless those desires would survive ideal deliberation. The Intersection View, however, seems untenable on other grounds. It implausibly implies that if none of the agent's actual desires would survive her ideal deliberation (making the intersection empty) there is nothing that there is any reason for that agent to do. 6. If few or even none of the agent's actual desires ____________________ 5. Saying that it would be irrational for me to have a particular desire because I would lose the desire if I thought about it carefully, is of course different from saying that the desire is intrinsically irrational or that all rational people would lose a similar desire if they thought about it carefully. 6. This implication holds and is implausible whether or not there actually are any agents whose intersection would be quite empty. Some readers may be unmoved by that counterfactual observation, pointing out that in all actual cases there would probably be at least a few actual desires (e.g. the -20-

would survive her ideal rational deliberation, then she seems sorely in need of rational guidance about things that she ought to do despite her own indifference; the Intersection View precludes such guidance. If the Intersection View is supplemented by Parfit's theses (a) and (b), we must add the qualification that any actions that might be rationally required of or forbidden to all agents would of course be required of or forbidden to the particular agent in question. Even the supplemented Intersection View, however, does not provide sufficient guidance to particular agents. For a particular agent may lack a particular desire only because of ignorance or confusion, even though that desire is not rationally required of all agents. Suppose, for example, that a woman lacks the desire to join the clergy only because she confusedly thinks there is something about femaleness that is incompatible with being a member of the clergy. In that case, there would be a reason for her to become a member of the clergy, although the desire to become a member of the clergy is not rationally required. The supplemented Intersection View is undesirably silent about the rationality of acts the agent would want to do if she had undergone ideal deliberation. A related problem is that the Intersection View makes immune to rational criticism the agent's failure to enlarge her intersection set by correcting her badly mistaken actual desires, except insofar as that failure itself interferes with her limited stock of desires in the intersection. (If there is an independent rational requirement to aim to have the benefit of ideal deliberation, however, then the Intersection View supplemented by Parfit's thesis (b) is not open to this objection.) The Ideal Desires View does not suffer from the foregoing difficulties. Because it allows ideal desires that are not in the intersection with actual desires to determine reasons for action, it provides rational guidance to the agent whose actual desires are all mistaken. It states that there is most reason for such an agent to do what she would want after ideal deliberation. The Ideal Desires View also provides that the agent who has a few rational desires can be rationally criticized for not pursuing goals that she fails to care about only through confusion, etc. Many popular criticisms of the Ideal Desires View can be deflected by a guarded version of it. A well-known criticism by Allan Gibbard attacks the Ideal Desires View's notion that ideal deliberation requires full knowledge. Gibbard points out that what we rationally ought to do under unavoidable ignorance is not what we would do under full knowledge. Thus if one is lost in the woods it is pointless to say that one rationally ought to do what she would do if she had full ____________________ desire to eat) that would survive rational deliberation. Now the Intersection View implies that in such cases the only actions that there is reason for the agent to take, are actions that pursue the satisfaction of those few desires. Although this implication is less dramatic than the previous one, it is also quite implausible. On the contrary there seem to be reasons, of which the agent is ignorant, to pursue the satisfaction of (some of) her nonactual ideal desires.

-21knowledge, i.e. were not lost (Gibbard 1990: 18-19). A guarded version of the Ideal Desires View can sidestep this objection by stating explicitly that it is concerned only with what there is most reason for the agent to do all-thingsconsidered, not with what is the most rational thing for her to do given certain cognitive constraints. 7. (This latter concern, however, could perhaps be addressed in terms of the best hypothesis available to the agent about what there really is most reason for her to do.) A guarded version of the Ideal Desires View should adopt Parfit's supplementary theses (a) and (b). Recall that thesis (a) states that if any desires are intrinsically irrational, then they cannot provide reasons for action even if they are in the intersection of the agent's ideal desires and her actually present desires. Thesis (b) states that if there are any desires that are rationally required, then they provide reasons for action even if they are neither in the agent's ideal desires nor actually present in her. Thesis (a) meets the objection by Gert (1990) and others that patently irrational desires might survive ideal deliberation or even be generated by it. Thesis (b) meets the objection by Bond (1983) and others that there may be objective reasons for action that are independent of the agent's actual or hypothetical desires. Two additional weaknesses of the Ideal Desires View are noted by Michael R. DePaul and J. David Velleman. DePaul (1987: 478) points out that different temporal orderings are possible in one's consideration of the relevant facts. Ideal deliberation might yield different results depending on the order in which the facts are considered. Velleman points out that the relevant facts may be presented not only in different orders but also in different manners. One manner of presentation (corresponding, for example, to a particular literary style) might yield different ideal desires than a different manner of presentation (Velleman 1988: 365ff.). Thus a particular nonactual desire may be ideal for the agent—since it would result from (certain possible) instances of her ideal deliberation —but also not be ideal for the same agent—since it would not result from (certain possible instances of) her ideal deliberation. 8. Thus the Ideal Desires View would in principle yield either a contradictory or an indeterminate verdict concerning whether there is any reason for a particular agent to act in certain cases. Surely, however, that is a serious defect in a theory of reasons for action. A guarded version of the Ideal Desires View should therefore specify that each ideal desire of the agent's gives her a reason for action only if that desire would result from all deliberations that would count as ideal deliberations for her. Thus ____________________ 7. A similar limitation is observed by Parfit (1984: 153): "I distinguished between what I have most reason to do, and what, given my beliefs, it would be rational for me to do.... My main question is about what we have most reason to do." 8. A similar indeterminacy or self-contradiction would result from the fact that some of the agent's actual desires might survive (certain possible instances of) her ideal deliberation, yet be extinguished by (other possible instances of) her ideal deliberation.

-22all temporal orderings of the facts and all manners of presentation of the facts in ideal deliberation must yield a particular non-actual ideal desire, in order for that desire to ground a reason for the agent to act. No embarrassment would now be posed by the admission that a given nonactual desire might be ideal for a particular agent in that it would be produced by (some possible instances of) her ideal deliberation, while being nonideal in that it would not be produced by (some possible instances of) her ideal deliberation. For on the guarded version of the Ideal Desires View, these facts would not in principle result in an indeterminate or contradictory verdict concerning whether there is a reason for the affected agent to act. The nonactual desire provides a reason for her to act if and only if it would be produced by all deliberations that would count as ideal deliberations for her. Before I get to the objections that I think finally defeat even a guarded version of the Ideal Desires View, let me defend that view against two more objections. It may seem that the Ideal Desires View in any guise is vulnerable to the following objection. Desires (unless they are intrinsically irrational) are the sorts of things that can determine reasons for action. Their satisfaction and their frustration are quite real regardless of whether the desires themselves would survive ideal deliberation. Suppose, for example, that I know that my keen desire to avoid heights would not survive ideal deliberation, but that I have been unsuccessful in performing ideal deliberation to rid myself of the desire. Given that I do have the desire (as irrational as that desire may be), there seems to be a reason for me to decline an invitation to visit the observation deck of the Empire State Building. An obvious reply to that objection, however, is that perhaps it is not the irrational desire to avoid heights that determines the reason, for the reason may be a function of the entirely rational desire to avoid the unpleasant sensations experienced by acrophobes in high places. Here is another argument that might seem to show that nonideal desires can generate reasons, contrary to the Ideal Desires View. Suppose that I want a cup of frozen yogurt, but that if I knew all the relevant facts (e.g., that carrot cake is also available,) I wouldn't want the frozen yogurt at all but would instead want a piece of carrot cake. Still, given that I do want the frozen yogurt, there is surely a reason for me to get it. Thus reasons can be a function of actual desires that would not survive the knowledge that goes with ideal deliberation. If we accept this conclusion, we must reject the Ideal Desires view, for it refuses to allow reasons to be a function of such desires. But the argument about the desire for frozen yogurt fails to establish its conclusion. 9. That which in the argument is simply called a desire for frozen yogurt is actually a preference for frozen yogurt over all available alternatives. The agent's preference for it over all the alternatives that she knows to be available would ____________________ 9. I owe the argument in the current paragraph to Brad Hooker and Alan Fuchs.

-23presumably survive ideal rational deliberation; she just doesn't know that carrot cake is available. Thus, the explanation of the appearance that there is really a reason for her to choose the frozen yogurt is that there is a reason for her to choose the frozen yogurt rather than all the alternatives that she knows to be available. But there is no reason for her to choose it rather than carrot cake. Thus the argument provides no reason to think that reasons for action are a function of nonideal desires. Further, insofar as the agent's preference for frozen yogurt over the alternatives that she knows to be available is a preference that would not survive ideal deliberation (perhaps because the yogurt would make her sick) we would deny that there is a reason for her to choose it at all. This fact strengthens the impression that reasons for action are not a function of nonideal desires. I come now to a new argument that is fatal to the Ideal Desires View, even to the guarded version. I shall state the argument and then illustrate it. It seems that procedurally defined ideal deliberation can extinguish desires that are rationally permitted. 10. Even when the temporal ordering and the manner of presentation are varied, rationally permitted desires can be extinguished by procedurally defined ideal deliberation. If a rationally permitted desire is extinguished by all forms of ideal deliberation, then the guarded version of the Ideal Desires View gives it no role in determining reasons for action. But a rationally permitted desire ought to have some role in determining reasons for action. Thus the Ideal Desires View is flawed. For instance, procedurally defined ideal deliberation (such as vivid and repeated reflection on relevant facts) can cause (at least some) people to lose the desire to attend dinner parties. 11. This could happen regardless of the temporal ordering or manner of presentation of the facts. Nevertheless, that desire is presumably rationally permitted. Thus the desire ought to have some role in determining reasons for action, but the Ideal Desires gives it none. To that objection, there seem to be only the following possible ways to reply: (i) contend that, intuitively, the desires that would be extinguished by ideal deliberation are rationally impermissible after all; (2) reject the authority of intuition to determine which desires are rationally permitted; (3) contend that Brandt's or some other already-formulated conception of ideal deliberation does not eliminate any actual desires of the sort in the example; (4) modify or reformulate the procedural definition of ideal deliberation so that its results agree ____________________ 10. If a desire is rationally permitted then it is neither an intrinsically irrational desire nor a desire on which it would be irrational for the agent in question to act, other things equal. 11. I have adapted this example from Gibbard 1990: 20. (Gibbard uses the example to argue that Brandtian ideal deliberation would generate an irrational desire/neurosis.) Another example, supplied by Mark Vorobej in correspondence, is that it might be rational for a person to enter a burning building when there is an excellent chance of saving his loved ones, although if he allowed himself to dwell vividly on the terrifying blaze he would be unable to muster the courage. Entering the burning building or

attending the dinner party might even be more than rationally permitted to the individual in question; it might be what there is obviously most reason for her to do in the circumstances. -24with intuition. 12. None of these suggestions seems to work. According to my own intuitions, reply (1) is simply false. As for (2), no theorist that I am aware of has succeeded in dispensing with intuitions. Even the arch anti-intuitionist Brandt relies on intuitions, including the intuition that it is a deliberative error to overlook options that could have been thought of, the intuition that a desire whose genesis depends essentially on an unreasonable belief is itself irrational, and the intuition that it is irrational to count one desire twice in weighing the desirability of alternate actions. 13. As for (3), I know of no procedural definition of ideal deliberation that would never eliminate actual desires of the sort in the example. For reply (4) to work, we would need to modify the procedural definition of ideal deliberation so that relevance would not automatically be granted to all facts the vivid consideration of which would result in the alteration of our desires. But how would relevance be decided? I can think of no answer that is consistent with the spirit of the Ideal Desires View, i.e., that avoids prior substantive judgments about the rationality of desires. We have seen that the Ideal Desires View suffers from the difficulty that procedurally defined ideal deliberation might extinguish desires that are rationally permitted. Another problem is that it might generate desires that are irrational for the particular agent. For example, it might be irrational for me, on account of my particular situation, to commit adultery, but vivid and repeated dwelling on the possibilities might generate that irrational desire in me. 14. This might happen despite the fact that the dwelled-on possibilities include the risk of badly hurting the feelings of innocent people, as well as the possible delights of adultery. Now if the desire to commit adultery were intrinsically irrational, and not just irrational for the particular agent in question, then Parfit's thesis (a) would protect the guarded Ideal Desires View from the counterexample. It seems, however, that it might not be intrinsically irrational to desire to commit adultery, yet it might be irrational for me in particular. So far, we have considered four theories that base reasons for action on desires: the Actual Desires View, the Ideal Desires View, the Intersection View, and the Union View. We have found all four theories unacceptable. Now it would be hard to dispute the thesis that at least some reasons for action seem to depend on ____________________ 12. Replies (3) and (4) might appear to some readers to be incompatible with the guarded version of the Ideal Desires View, since that version speaks of all temporal orderings and all manners of presentation of the facts in ideal deliberation. But a formulation (or conception, or definition) of ideal deliberation need not specify any particular temporal order or manner of presentation of the facts. Rather it will specify ideal deliberation by answering questions of the following sort: Which facts must be taken into account?

How detailed must the deliberation be? How vivid must it be? Must it be done repeatedly? How long must it be continued? Must it be done while the agent is on an even keel emotionally? 13. The first two of these intuitions can be found in Brandt 1979 on pp. 71 and 89 respectively. The third can be found on p. 156 of Brandt 1983a. 14. This example was supplied by Mark Vorobej in correspondence. -25the agent's desires, broadly construed. For example, there seem to be significant reasons for avid mountain climbers to climb mountains, and those reasons seem to depend on the interests, projects, aspirations, likes, dislikes, etc., of the mountain climbers. And it also seems that there would be little or no reason for a person with very different interests, projects, likes, etc., to climb mountains. But how can we say that the agent's desires are relevant to reasons for action if we reject all four views that make reasons depend on the agent's desires? The Inclusive Data View (InDat), to be sketched presently, provides an alternative. 15. It makes reasons for action depend not only on the agent's actual and ideal desires but on a much larger set of data. Furthermore, where the other four views make reasons a direct function of desires, InDat makes reasons depend on the data in a more complicated fashion. InDat states that lower-order normative principles of practical rationality should be considered valid just in case they make maximally good sense (in a sense to be explicated presently) of data about desires, valuings, and other relevant matters. 16. Lower-order normative principles are specified to avoid pernicious self-reference for, as I shall explain shortly, the current view is itself a normative principle. An example of a lower-order normative principle (which may or may not be valid) is that an agent's desire for a thing that she natively likes provides that agent with a reason for action. Thus InDat can be thought of as a rule that licenses particular propositions like "there is a reason to pursue that which one natively likes". It provides that such a proposition is licensed if and only if it is part of a set of propositions that makes maximally good sense of the relevant data, given my characterization of the data and my account of what constitutes making sense of the data. Included as relevant are data about at least the following: what the agent desires (in a broad sense of 'desire'), what she values, what she thinks valuable, what she would desire or think valuable after a specified sort of ideal deliberation, what she would desire or think valuable from all standpoints (including both impartial and self-interested ones), and data about what all individuals in a specified domain do and would desire or think valuable in these circumstances. 17. The relevant data comprise not only those, but also any other facts and adequate theories that seem relevant, possibly including (but not ____________________ 15. This view may be thought of as a (possibly illegitimate) descendent of Thomas Nagel's theory in The View From Nowhere (1986). 16. After reading my explication of the notion of making sense of the data, some readers will think that I have instead explicated the notion of making sense in light of the data. I have no quarrel with this view. Those who hold it should think of InDat as stating

that lower-order normative principles of practical rationality should be considered valid just in case they make maximally good sense in light of relevant data, or just in case they make maximally good sense of and in light of relevant data. 17. The notion of standpoints is taken from Nagel 1986. A fully developed theory would of course have to specify a particular notion of ideal deliberation and a particular domain. The most inclusive version of InDat would include in the domain all individuals who desire or value; other versions would specify narrower domains. -26limited to) decision theory and any adequate theories of learning, motivation, intentional action, mental illness, personal continuity, and human nature. Here are several suggestions for ways in which a set of normative principles, including ones such as the principle that everyone has a reason to pursue that which she natively likes, might be said to make sense of (some of) the data. (a) It might disclose patterns in the data, e.g. by identifying salient similarities in desires that are taken to provide reasons. (b) It might "hook up" with a theory of learning, for example, by according significance to the same distinctions and variables or otherwise displaying points of contact. (c) It might distinguish veridical from nonveridical valuations in a way that takes account of the distinction between valuations that are based on factual confusion or ignorance and those that are not so based. (d) It might lend itself to an explanation of the appeal of normative judgments that it rejects. (d) It might avoid telling us that we rationally ought to do what is impossible. (e) It might reconcile conflicting pre‐ theoretical value judgments. (g) If the connection of the normative principles to the data can be established independently, so that we can speak of the principles making sense of the data, then the degree of sense that they make in themselves via internal coherence might be thought relevant to the degree of sense that they make of the data. In particular, sense increases with the number and importance of internal inferential connections and (relative) lack of division into subsystems that are relatively unconnected to each other by inferential connections. A system also increases in sense by having fewer rather than more underived principles and by answering more rather than fewer questions that it recognizes as legitimate. Inconsistencies detract from the sense that a set of principles might make, though trivial inconsistencies need not annihilate all sense. 18. I want to argue that InDat is a promising alternative to the Ideal Desires View, the Actual Desires View, etc. First, however, I must deal with the objection that InDat is not a genuine alternative to those views—that it is rather a meta-level principle that tells us when normative theories like the Ideal Desires View are valid. One possible cause of the suspicion that InDat is a meta-level principle is the fact that it sounds very much like a principle of Wide Reflective Equilibrium for Rationality (WRER), which would be generally acknowledged to be a meta‐ level principle. For simplicity, I stipulate that WRER says roughly that a theory of reasons for action is justified if the general propositions that it licenses make maximally good sense of relevant data. InDat licenses propositions about reasons for action just in case they make maximally good sense of relevant data. Now those principles may sound alike, but they are distinct precisely

because they are on different logical levels. It may be difficult to see this immediately, since we ____________________ 18. In this characterization of coherence, I draw elements from Bonjour 1985: especially p. 98, and from Ellis 1988: especially p. 68. An all-or-nothing version of the completeness requirement is posited as part of coherence by Sayre-McCord (1985: 171). -27may never have entertained the thought of anything like a normative-level homophone of the principle of Wide Reflective Equilibrium. Let us therefore dwell on the notion for a while. I have said that InDat is a rule for licensing propositions about reasons for action. It can also be thought of as a definition of the principles of reasons for action; it defines them as those principles that make maximally good sense of the data. This way of thinking of it will help us see that the logical level of InDat is comparable to that of Rawls's definition of the principles of justice as those principles of social cooperation that would be agreed to by self-interested persons under certain circumstances. 19. For a rival theorist—call him Pseudo‐ Rawls—could have offered, as a rival normative-level definition of the principles of justice, a quasi-homophone of the principle of Wide Reflective Equilibrium that parallels InDat. On Pseudo-Rawls's definition, the principles of justice are the principles of social cooperation that make maximally good sense of the data. This definition is a rival to the Rawlsian definition of the principles of justice regardless of whether it would yield the same principles; the definitions are rivals because they represent different rules for licensing a principle as a principle of justice. Pseudo-Rawls might try to defend his definition on the basis of meta‐ level WRE 20. or he might try to defend it on the basis of some other meta-level criterion, such as usefulness in bringing about the peaceful resolution of disputes of a certain sort. Whatever meta-level principle he adopted, however, the normative-level quasi-homophone of WRE would be a normative-level principle distinct from the meta-level WRE. Normative-level principles and the meta-level principles that concern them have distinct domains; therefore they are distinct principles. This is true even when the different-level principles sound very much alike, as do InDat and WRER, or the pseudo-Rawlsian quasihomophone of WRE and WRE. A familiar example of a pair of distinct principles that have the same structure and hence sound alike is provided by the rule of Conditional Proof in logic and the conditional proof strategy in metalogic. "This is, in a nutshell, the rule of Conditional Proof: to prove a conditional (A⊃B), simply assume A temporarily and see whether, given A, B follows. If it does, then you are permitted to infer (A⊃B)." 21 This rule of Conditional Proof is used in doing proofs within logic, while meta____________________ 19. "Together with the veil of ignorance, these conditions define the principles of justice

as those which rational persons concerned to advance their interests would consent to as equals when none are known to be advantaged or disadvantaged by social and natural contingencies." (Rawls 1971: 19) Rawls's definition is fruitfully regarded as a principle of his normative theory of justice, and Rawls seems to regard it so: "Once the whole framework is worked out, definitions have no distinct status and stand or fall with the theory itself" (1971: 51). 20. I leave aside the question whether such a definition combined with the meta-level WRE principle would make an attractive package. 21 . Klenk 1983: 172. -28logic is used to prove metatheorems about the rules of logic. Thus the rule of Conditional Proof is part of logic rather than part of metalogic. But a parallel conditional proof strategy is used in metalogic. In the metalogical conditional proof strategy, we infer the truth of a conditional metalogical proposition from the fact that we can derive its consequent from its antecedent. 22. Thus InDat need not be a meta-level principle just because there is a meta-level principle, WRER, that sounds very much like it. Perhaps the objector will grant that I have established that InDat may differ from WRER, yet still maintain that InDat is in fact the same as WRER. I would reply that they are distinct principles if they differ in their logical levels, and they do differ in their logical levels. InDat is a rule for licensing propositions about reasons for action, such as the proposition that everyone has a reason to pursue what she natively likes. WRER, insofar as it is a meta-level principle, is a criterion for evaluating rules for licensing such propositions about reasons for action. Furthermore, an advocate of InDat can coherently reject WRER as a meta-level principle. She can, for example, hold that the sole acceptable meta-level principle for the evaluation of a theory of reasons for action is the Darwall-style requirement that the theory must be such that an internally self-identified rational agent could regard it as an ultimate principle of justification (Darwall 1983). It would not be possible to accept InDat and reject WRER in this way if they were the same principle. Another possible reason that an objector might have for taking InDat to be a meta-level principle is that it seems to provide a test that can be used for evaluating the Ideal Desires View etc. This fact might seem to preclude its being on same logical level as the Ideal Desires View. Now one possible reply to this objection would be that InDat refers only to lower-order normative principles of rationality, not to very general principles such as the basic principles of the Ideal Desires View. But this reply may seem artificial. A less artificial reply is this. The Ideal Desires View construes reasons for action as a function of the agent's ideal desires. In contrast, InDat construes reasons for action as depending in a more complex way on a much wider class of data than the agent's ideal desires. To contradict each other in this way, it seems that the two views must be on the same logical level.

Just as InDat can be regarded as a test for the Ideal Desires View, the Ideal Desires View can be regarded as a test for InDat. For suppose that the normative principles that make the best sense of all the data turn out to imply that there is most reason for a particular agent to do something that ideal deliberation would lead her unequivocally to eschew. In that case a committed adherent of the Ideal ____________________ 22. It is common practice to use this strategy, for example, in the metalogical argument for the conclusion that if a proposition is provable on certain premises by the rules of logic, then the corresponding argument is semantically valid. See, for example, Mendelson 1987: 72. -29Desires View would conclude that there is something wrong with InDat—perhaps in its including, among the relevant data to be made sense of, things other than the agent's ideal desires. Since the Ideal Desires View and InDat can lead to conflicting conclusions about what there is reason to do, either view can be cited as grounds to reject the other. Of course InDat and the Ideal Desires View could lead to the same conclusions about what there is reason to do. This would happen if the lower-level normative principles of practical rationality that make maximally good sense of the relevant data happen to posit reasons for action that correspond to the agent's ideal desires. (For example, both views might lead us to hold that there is a reason to pursue what one natively likes.) Any two rival theories may agree in this way. They are, however, still rival theories, for they furnish rival rules for determining which statements about reasons for action are valid. Given that InDat is a genuine rival to the Ideal Desires View, the Actual Desires View, etc., what reason is there to hope that it will avoid the flaws that were seen to plague its rivals? In view of my account of making sense of the data, it seems reasonable to hope that the normative principles of practical rationality selected by InDat would satisfy the following conditions: (a) they would not recognize reasons corresponding to desires based on ignorance or confusion; (b) they would recognize reasons corresponding to desires that the agent lacks only because of ignorance and confusion; (c) they would recognize reasons corresponding to some desires (intuitively rational for the agent) that would be extinguished by maximally vivid and repeated contemplation of certain facts; and (d) they would not recognize reasons corresponding to some desires (intuitively irrational for the agent) that would be produced by maximally vivid and repeated contemplation. Conditions (a) and (b) would presumably be met because one way for a set of principles to make sense of the data is to distinguish veridical from nonveridical valuations in a way that takes account of the distinction between valuations that are based on factual confusion or ignorance and those that are not so based. As for (c) and (d), it seems reasonable to hope that they too would be met. For one way for principles to make sense of the data would be by identifying salient similarities in desires that we take to be rational or irrational for an agent. If the principles that made maximally good sense of the data did that, then it seems reasonable to hope that conditions (c) and (d) would be met. But if conditions (a) to (d) were all met, then InDat would avoid the flaws that we found

to plague the Actual Desires View, the Ideal Desires View, etc. For each of those theories foundered on its failure to meet one or more of conditions (a) to (d). InDat may appear to have an immediately fatal problem of its own. It says that a statement about what is valuable should be considered valid if it (is part of a set of principles that) makes maximally good sense of data about, among other things, what is thought valuable. When one thinks something valuable, however, -30doesn't one thereby think that the statement that it is valuable is valid? If so, doesn't InDat attempt to define value in terms of itself, and hence fail to define value? That would make the view a non-starter. Now it is true that InDat does not explicate the concept of value that is embedded in the data about what is thought valuable. 23. It is, however, appropriately neutral with respect to the concept of value that is embedded in those data. The normative principles of value that make maximally good sense of the data about what people think valuable, need not themselves attribute legitimacy to the concept(s) of value employed by those people when they think things valuable. For the notion of making sense of the data was explicated in such a way that this was not required. 24. I conclude that InDat represents a genuine alternative to the Actual Desires View, the Ideal Desires View, the Intersection View, and the Union View, that it shows promise of avoiding the flaws that we saw to be fatal to its rivals, and that it does not defeat itself at the outset by failing to define the embedded concept of value. Thus InDat merits further investigation. 25. ____________________ 23. Although I shall not defend it, I am intrigued by Paul Grice's view of the embedded concept of value. Grice suggests that an account of this concept involves a legitimate Humean Projection, in which that which characterizes thinking comes to characterize the object of thinking. Thus valuing x or thinking-of-x-as-valuable becomes thinking that x is valuable. See Grice 1991: 88. 24. Some ways of making sense of the data were said to include disclosing patterns in the data, distinguishing veridical from nonveridical valuations in a way that takes account of the distinction between valuations that are based on factual confusion or ignorance and those that are not so based, lending itself to an explanation of the appeal of normative judgments that it rejects, and reconciling conflicting pre-theoretical value judgments. All these could be done without explicating the notion of value or reason embedded in the datum judgments. 25. Earlier versions of portions of this paper were read at the Southern Society for Philosophy and Psychology and the American Philosophical Association. For helpful comments on earlier versions or relevant correspondence, I am grateful to Tom Carson, Bob Ehman, Alan Fuchs, Margaret Holmgren, Brad Hooker, Mike Lavin, Miriam Levering, John Nolt, David-Hillel Ruben, Fred Stoutland, Mark Timmons, and Mark Vorobej.

-31-

3 Brandt on Self-Control Philip Pettit and Michael Smith There are three questions about self-control that we discuss in this paper. The first is, what occasions self-control: what circumstances call for the exercise of the virtue? The second is, what constitutes self-control: what sort of psychological traits can realize it? And the third is, what recommends self-control: what is there to be said in favor of the virtue? Richard Brandt has put forward views, explicitly and implicitly, on all of these questions and the aim of the paper is to examine his views critically. We are in sympathy with many aspects of his approach but, as will appear, we defend a picture that differs significantly from that which we find in his works. The paper is in five sections. In the first section we present Brandt's views, as we understand them. And then in the three following sections we examine and criticize the answers that he suggests to each of our three questions. In the course of doing this we will have the opportunity to present, and to develop further, a picture we have already outlined elsewhere (Pettit and Smith 1993). The fifth section offers a brief resume of the argument.

I. Brandt's Account In giving an account of self-control we must do three things. First, we must say what the circumstances are that call for self-control. Second, we must say what self-control is in itself. what the difference is between two agents when one, in circumstances that call for self-control, exercises self-control, and the other, in those same circumstances, does not. And third, we must say why self-control is an attractive trait: why it is assumed, as people generally assume, that self-control is a virtue. In the terms introduced earlier, we must say what occasions, what constitutes and what recommends self-control. -33-

What Occasions Self-Control? Brandt tells us that self-control is required when an agent is relatively more strongly motivated to act otherwise than as she judges best. He notes that, on certain interpretations of what it means to judge a course of action best, this may seem to be impossible. For example, if such a judgment were the expression of an overall preference for acting in the way in question, then it would be impossible, by definition, for an agent to be relatively more strongly motivated to act otherwise. However, Brandt makes it clear that he rejects such an interpretation of best judgment. For, as he points out in an

important article on "The Structure of Virtue", "the fact of experience, that we sometimes fail to do something at the very moment we are judging that it would be 'best' is some evidence that 'is best' at most expresses a strong motivation but not an overall preference" (Brandt 1988: 70). This is all Brandt has to say about best judgment in "The Structure of Virtue". However, as is well known, he has more to say about the content of the judgment that an act is best in A Theory of the Good and the Right. In order to give Brandt's account of the circumstances that call for self-control determinate content, then, we must interpret it in light of his analysis of best judgment. In A Theory of the Good and the Right Brandt tells us that, though the expression "the best thing to do" "has no definite meaning at all, in ordinary use", we can say everything that it is useful to say using that expression by substituting the expression "the rational thing to do". The latter expression, he tells us "captures all that is clear in the former" (Brandt 1979: 15). Brandt would therefore presumably be happier himself with the following reformulation of his account of the circumstances that call for self-control. An agent is in a situation that calls for self-control when she judges something to be the rational thing to do, but finds herself relatively more strongly motivated to act otherwise. In order to flesh this out, we need to remind ourselves what Brandt means by the expression "the rational thing to do". Brandt holds that to say that something is the rational thing for us to do is in fact ambiguous (Brandt 1979: 11). On the one hand, it might mean simply that it is the thing we would do, if we had vivid awareness of all of the facts, but where our actual intrinsic desires are taken as given. Following Brandt, let's say that this is rational action "to a first approximation". On the other, it might mean that it is the thing we would do if we had vivid awareness of all of the facts, and if our intrinsic desires were themselves rational. Our intrinsic desires are themselves rational, according to Brandt, if they are the intrinsic desires we would have if we had vivid awareness of all of the facts. Following Brandt, let's speak here of "fully" rational action, as distinct from action that is rational to a first approximation. -34In order to see what this distinction amounts to, consider an example. Suppose an agent has just one actual intrinsic desire: a desire for pleasure. What is it rational for her to do? Brandt tells us this question is ambiguous. To a first approximation, it is rational for her to do what she would do if she were vividly aware of all of the facts, her desire for pleasure taken as given. Let's suppose that her options are just two. She can either eat a chocolate bar or eat a piece of fruit. Vivid awareness of all of the facts therefore means, in this case, awareness of how eating a chocolate bar and eating a piece of fruit each contribute to pleasure. We will therefore suppose that the rational thing for her to do to a first approximation, in this case, is to eat

a chocolate bar. For, we will suppose, eating a chocolate bar will give her more in the way of pleasure. A fully rational action is, however, different. For now we must begin by asking whether her actual intrinsic desires are themselves rational: that is, whether they are the intrinsic desires she would have if she were vividly aware of all of the facts. And let's suppose that they are not. For though she does not in fact desire health, she would desire it, and indeed desire it even more strongly than she desires pleasure, if she were to think vividly about what health is and involves. It then follows that the fully rational thing for her to do is to eat a piece of fruit. For that is what she would do if her intrinsic desires were themselves rational, and if she were then to think vividly about her options: that is, about how eating a chocolate bar and a piece of fruit each fare with regard to the production of both pleasure and health. Let's return to Brandt's account of the circumstances that call for self-control. An agent is apt to lose self-control, he tells us, when she believes that it would be best for her to act in a certain way, but she is relatively more strongly motivated to act otherwise. We noted that, by Brandt's own lights, this is better stated in the following terms: an agent is apt to lose self-control when she believes that it would be rational for her to act in a certain way, but she is relatively more strongly motivated to act otherwise. And we noted that, according to Brandt, the claim that it is rational to do something is ambiguous, and that it therefore follows that his account of the circumstances that call for self-control is itself ambiguous. There are, as we might now say, the circumstances that call for self‐ control "to a first approximation" on the one hand, and there are the circumstances that call for "full" self-control on the other. In order to see what these amount to, consider one of Brandt's own examples, a variation on the example we have been discussing thus far: exercising self‐ control in order to restrain indulgence in eating food. Let's suppose an agent is deciding whether to have one sandwich or two at lunch time. Among her various actual intrinsic desires we will suppose that she has desires for pleasure and health, and that these give rise to motivations to eat a second sandwich and to -35refrain respectively. She contemplates her various options and comes to believe that, taking her actual intrinsic desires as given, if she thought vividly about all of the options available to her and their various consequences, she would overall prefer to eat one sandwich rather than two. Unfortunately, however, she finds herself relatively more strongly motivated to eat a second sandwich none the less. Here we have a situation that calls for self-control to a first approximation. For the agent suffers from a form of instrumental irrationality. With her actual intrinsic desires taken as given she knows that, if she thought vividly about all of the options available to her, she would overall prefer to eat only one sandwich. But, compared with the alternative, thoughts about how refraining from eating a second sandwich will contribute to health

are "too abstract", or are thoughts about something that is too "remote in time", to remain vividly in her awareness (Brandt 1988: 70). Instead, less abstract thoughts about an immediate good, thoughts about how eating a second sandwich will lead to pleasure, occupy all of her attention, and so she ends up preferring overall to do that instead. So much for the circumstances that call for self-control to a first approximation. What about the circumstances that call for full self-control? Let's suppose again that an agent is deciding whether to have one sandwich or two at lunch time. But this time let's suppose that she only has an intrinsic desire for pleasure. She contemplates her various options and comes to believe that, if she thought vividly about all of the options available to her and their various consequences, she would overall prefer to eat two sandwiches rather than one. She therefore concludes that it is rational to a first approximation to eat a second sandwich. However, she then wonders whether her actual intrinsic desires are themselves rational, and comes to believe that, if she were to think vividly about health, she would have an intrinsic desire for health, and that her desire for health would in fact be stronger than her desire for pleasure. She once again contemplates her various options and comes to believe that, if she had rational intrinsic desires, and if she then thought vividly about all of the options available to her and their various consequences, she would overall prefer to eat one sandwich rather than two. This is then the fully rational thing for her to do. However, since she is not thinking vividly about what health is and involves, she does not in fact have a desire for health. The property of being healthy is, we might suppose, itself "too abstract" or its realization too "remote in time", to remain vividly in her awareness. Unfortunately she therefore finds herself relatively more strongly motivated to eat a second sandwich none the less. This is a situation that calls for full self‐ control, as distinct from self-control to a first approximation. In Brandt's view, then, when an agent is relatively more strongly motivated to act otherwise than as she judges it would be rational for her to act, she is in a situation that calls for self-control. The problem in such cases is either that, given her beliefs, her overall preferences do not reflect the relative strengths of her actual intrinsic desires (a situation that calls for self-control to a first approxima -36tion) or that her actual intrinsic desires are themselves irrational (a situation that calls for full self-control). Brandt's account of the circumstances that call for self-control is rather cumbersome, because of all of the distinctions we have had to make. Therefore, from here on, when we say that a situation calls for self-control we will mean that it calls for full self-control. When we want to say something about situations that call for self-control to a first approximation, we will make this explicit in the text. This will allow us to simplify the discussion that follows. No points of any significance will be lost.

What Constitutes Self-Control? We have explained Brandt's account of the circumstances that call for self‐ control. We must now say what Brandt tells us about self-control itself. Suppose two agents find themselves in circumstances that call for self-control. One exercises self-control and the other doesn't. What does the difference between the two agents consist in? Brandt tells us that the agent who possesses self-control possesses certain intrinsic motivations, motivations lacking in the agent who does not. Since we may suppose that both are alike with regard to the intrinsic desires that support their judgments about what it is rational to do, the intrinsic motivations that constitute the one agent's self-control must be extra intrinsic motivations. Self-control is therefore, according to Brandt, simply an extra intrinsic motivation adequate for acting in accordance with an agent's judgment concerning what it is rational to do in the circumstances that call for self-control. Brandt emphasizes the fact that the agent who possesses self-control need not possess a quite general intrinsic desire to act in the way that she believes rational when she has relatively stronger inclinations to act otherwise. For, as he reminds us, an agent may possess self-control in one area (say, in the consumption of food) but not in another (say, in the consumption of liquor). Self-control is rather, as he puts it, "a plural motivation" (Brandt 1988: 71). The agent who exercises self-control has certain intrinsic motivations adequate enough to adopt "a plan for specific situations, with rewards". For example, in the case of exercising self-control to stop overeating, she may have intrinsic motivations for going to see a therapist; for joining a group who congratulate each other for their success in losing weight; for promising to give herself rewards in return for restraint; and so on. Furthermore the agent will have intrinsic motivations adequate enough for the "development of motivation so that the problem calling for self-control does not exist" (Brandt 1988: 71). The agent who lacks self-control, Brandt tells us, simply lacks these extra intrinsic motivations. Brandt's view that it is the presence of a certain sort of motivation that constitutes selfcontrol fits, it should be mentioned, with his more general view that all -37virtues are motivational; they consist in dispositions to desire certain things intrinsically. "A virtue is a (certain kind of) relatively unchanging disposition to desire an action of a certain kind (e.g. helping one in distress, not stealing) for its own sake" (Brandt 1988b: 64). This motivational view of virtue contrasts with the view that at least some virtues are managerial rather than motivational in character; they involve cognitive skills rather than desiderative dispositions (Roberts 1984). We return to this point later.

What Recommends Self-Control? Finally, to the question of why self-control is supposed to be a good thing. There are two salient answers that might be offered. The first is that it is a good thing because the

consequences of possessing it are better than the consequences of not possessing it. The second is that, though that may indeed be true, there is a deeper reason for thinking well of self-control: viz., that it is rationally desirable. Brandt does not explicitly address the question of what recommends self‐ control and it is difficult to be sure about what he would want to say. On the one hand, his conception of what it means for something to be rational to do or rational to desire suggests that he would take the exercise of self-control to be something rationally required or supported. No matter what it is rational for an agent to do or desire—no matter what the content of his judgments as to what is best—it would seem to be rational for him to exercise selfcontrol, in the circumstances where self-control is called for: in these circumstances, as Brandt presents them, spontaneous motivations fail and self-control is the only way of doing what is judged rational. On the other hand, however, Brandt defends a general position on what recommends virtue which would not seem to square well with the view that self‐ control is rationally required or supported. This is the view that what makes certain motivational dispositions into virtues is the fact that they have good consequences. The characteristic mark of the virtue is "roughly that the trait must be one normally and importantly favorable either for the well-being of society (or some group thereof) or for the flourishing of the agent (or those dear to him, e.g., his family)" (Brandt 1988b: 76). This view would naturally call for qualification or amendment, if it is held that some of the motivational dispositions in question—self-control and perhaps some others—are not only beneficial but also rationally required or supported. We will not say anything more at this stage on Brandt's views about what recommends self-control. We return to the issue later, when we try to give a critical assessment of his views. -38-

II. What Occasions Self-Control? We have no quarrel with Brandt's general claim that self-control is required when an agent is more strongly motivated to act otherwise than she judges best: otherwise than how she judges it is rational for her to act. Indeed that claim fits very well with our own picture of practical unreason (Pettit and Smith 1993). But Brandt's claim has to be interpreted, as we saw, in the light of his analysis of what it means for something to be the rational thing to do or desire. And we believe that that analysis is crucially flawed. In this section we identify the central problem for that analysis and we present an amendment that enables us to get around it. The upshot is a revised picture of the sort of situation that occasions self‐ control.For Brandt, the term "rational" is used in three different contexts. i. An agent's action is rational to a first approximation when, taking her actual intrinsic desires as given, it is the case that, if she thought vividly about all of the options available to her and their various consequences, she would act in that way. 2. An agent's intrinsic desires are rational when they are the intrinsic desires she would

have if she were to think vividly about all of the facts. and 3. An agent's action is fully rational when it is the case that, if she had rational intrinsic desires, and if she thought vividly about all of the options available to her and their various consequences, she would act in that way. There are a number of questions that might be raised about this line of analysis but we are going to concentrate on one. A remarkable feature of (1), (2), and (3) is that none of them makes the rationality of what an agent does or desires in any way depend upon the circumstances in which she is to do or desire that thing. And this is problematic, as emerges when we think about what it is rational for someone to do or desire in circumstances of extreme irrationality. In order to see this, consider a rather fanciful case. Suppose an agent believes that he is Jesus Christ, and that, as a result, he has all sorts of irrational desires. And suppose further that the only way he could come to realize his predicament is by going to see a psychiatrist. What is it -39rational for such an agent to do? What is it rational for him to desire? (Imagine that we are giving him advice.) We take it that the answers are plain enough. The rational thing for such an agent to do and desire is to go and see a psychiatrist. Remarkably, however, Brandt's analysis of what it is rational to do and desire prevents him from saying any such thing. For Brandt, it is rational to a first approximation for an agent to visit a psychiatrist just in case, if she were to think vividly about all of the options available to her, her actual intrinsic desires taken as given, she would visit a psychiatrist. But the agent we have described has intrinsic desires that reflect his belief that he is Jesus Christ. These Jesus Christ-like intrinsic desires would lead him to do all sorts of things, if he were to think vividly about his options: healing the sick and helping the poor, for example. But we have no reason at all to suppose that they would lead him to go and see a psychiatrist. Going to see a psychiatrist simply doesn't fit into Christ's agenda. It follows that it is therefore not rational to a first approximation for the agent we have described to go and see a psychiatrist. But neither, on Brandt's analysis, is it rational in the full sense for the agent we have described to visit a psychiatrist. For it is fully rational for him to visit a psychiatrist just in case, if he had the intrinsic desires he would have after thinking vividly about the facts, and if he were then to think vividly about all of the options available to him, he would visit a psychiatrist. Now if the agent we have described were to think vividly about the facts, especially about the facts that constitute his identity, then we may well expect him to lose many of his actual intrinsic desires—all of those that are nourished by his false belief that he is Jesus Christ. And we may well expect that he would acquire certain other

intrinsic desires instead. But he would also, thereby, lose any reason he has to go and see a psychiatrist. For why would someone who knows that he is not Jesus Christ, who knows that he is he himself with the intrinsic desires it is rational for him to have, go and see a psychiatrist? His other desires would give him no reason to visit a psychiatrist. And neither would he form an intrinsic desire for such a visit. There is nothing intrinsically attractive for such a person about seeing a psychiatrist. We must therefore conclude that it is not fully rational for the agent we have described to go and see a psychiatrist. Despite these, to our eye, compelling points of criticism, we are sympathetic to the general form of Brandt's analysis of rational action and desire. Indeed the critique shows us how we can revise the analysis, and preserve that form. The lesson of the critique is that an analysis of rational action and desire must make explicit room for the fact that the situations in which an action is done or a desire is had is itself an independent determinant of what it is rational to do or desire. We suggest an analysis in which circumstances are given this role and in which one further amendment is made: the reference to vivid awareness is replaced by a requirement of full rationality and full information. This further amendment can be ignored by those who think, as Brandt does, that all that need be involved -40is vivid awareness. It signals the presence of an issue which is not relevant to our purposes here and which we do not wish to get in the way of our present debate with Brandt. 1. Our analysis is this: 4. It is rational for us to do, in circumstances C, whatever we, if we were fully rational and fully informed, would want ourselves to do in circumstances C; and 5. It is rational for us to desire, in circumstances C, whatever we, if we were fully rational and fully informed, would want ourselves to desire in circumstances C. The idea behind (4) and (5) should be plain enough. In order to find out what it is rational for us to do or want in certain circumstances C we must imagine ourselves in two quite different possible worlds. First, we must imagine ourselves in the evaluated world. This is the world in which circumstances C obtain. Perhaps these are circumstances in which we face certain options: to act this way or that. But perhaps also they are circumstances in which we have all sorts of false beliefs, like the belief that we are Jesus Christ, or circumstances in which, perhaps as a result of these beliefs, we have irrational intrinsic desires. And, second, we must imagine ourselves in the evaluating world. This is the world in which we are fully rational and fully informed: and informed, inter alia, about who we really are in the actual world where circumstances C obtain. The idea then is this. What it is rational for us to do or want in the evaluated world is whatever we, in the evaluating world, would want ourselves to do or want in the evaluated world. (4) and (5) direct us to imagine our fully rational and informed selves, in the evaluating world, forming a preference concerning what our less than fully rational selves should do or want in the circumstances of the evaluated world. The content of this preference then in turn tells us what it is rational for us to do or want in the evaluated world.

Our analysis of rational action and desire makes explicit room for the possibility that what it is rational to do or desire is sensitive to the circumstances in which an agent's less than rational self finds herself. Characterize the circumstances as those in which an agent's less than rational self has to choose between various options—to act in this way or that—but abstract away from the fact that her less than rational self has all sorts of psychological flaws, and her fully ra____________________ 1. The issue is whether the account to be given of rationality can be fully reductive. This relates to general questions as to whether partly circular, non-reductive biconditionals can be philosophically useful in the elucidation of the concepts they involve. See McGinn 1983: 6-8, 14; Peacocke 1983: ch. 2; Smith 1986; Peacocke 1986; Wright 1988; Johnston 1989; Pettit 1991; Smith 1992; Menzies and Pettit 1993; Pettit 1993: ch. 2. -41tional self may want her less than rational self to do and desire very, very different things from anything she can in fact do or desire. Perhaps she will want her to be doing and desiring exactly what her fully rational self would do and desire in her circumstances. (This is like Brandt's account of fully rational action.) However, characterize the circumstances as those in which, additionally, the agent's less than rational self has all sorts of irrational beliefs and desires, beliefs and desires which place severe limits on what she can in fact do, and her fully rational self will then plausibly have very different desires about what her less than rational self is to do or desire. For she will frame those desires with an eye to the limits provided by her less than rational self's defective psychology, and so she will take on board the fact that being fully rational is not a real option for her. In order to see this, consider the example that proved so difficult for Brandt. An agent falsely believes that he is Jesus Christ and has all sorts of irrational desires as a result. Is going to see a psychiatrist a rational thing for him to do and desire? Our own analysis suggests that this all depends on what the agent's fully rational self would want his less than rational self to do and desire in the circumstances of the evaluated world. If we abstract away from the false beliefs and irrational desires of the agent's less than rational self in characterizing his circumstances, then the agent's fully rational self will want his less than rational self to be doing and desiring very different things from anything he is at all likely to do or desire. He would want his less than rational self's actions and desires to reflect his belief that he is he himself and not Jesus Christ. He therefore would want him to have desires that are just like his own fully rational desires. And he would want him to do just what he would do on the basis of those desires. Let's say that these are the things that it is rational for the agent's less than rational self to do and desire, abstracting-away-from-less-than-ideal-aspects-of-his-psychology. (Again, this is like Brandt's account of fully rational action.)

Importantly, however, given his less than rational self's predicament, and especially given that being fully rational is not a real option for his irrational self right at the moment, the agent's fully rational self will also presumably have some far more realistic desires about what his less than rational self is to do and desire. He will want his less than rational self to do and desire whatever is required in order to regain his grip on reality. Since that seems to require a visit to a psychiatrist, so, it seems, his fully rational self will desire his less than rational self to do and desire just that. Our own analysis allows such circumstance-specific desires of a fully rational and informed agent to fix what it is rational for an agent's less than rational self to do and desire in those circumstances. Alongside what it is rational for an agent to do and desire abstracting-away-from-less-than-ideal-aspects-of-her-psychology, then, we might say that our own account suggests that there will be the things that it is rational for the agent's less than rational self to do and desire -42taking-on-board-less-than-ideal-aspects-of-her-psychology. This is what is fixed by the circumstance-specific desires of an agent's fully rational self: by the desires her fully rational self has for what her less than rational self does when she takes into account the irrationality of that self in characterizing the relevant circumstances of action. Thus, our own analysis of rational action and desire tells us that desiring to go and see a psychiatrist and acting upon that desire are indeed rational taking-on-board-the-less-than-idealaspects-of-the-agent's-psychology. And this is the right result. On our account, then, what it is rational for an agent to do, in any circumstances, is a matter of what, if she were fully rational and informed, she would want herself to do in those circumstances. But in characterizing an agent's circumstances we may pay more or less regard to the agent's own rational failings, and so, on our account, we can distinguish what it is rational for her to do abstracting-away-from-less-than-ideal-aspects-of-herpsychology from what it is rational for her to do taking-on-board-less-than-ideal-aspectsof-her-psychology. This amendment of Brandt's analysis of rational action and desire means that we have an amended answer to the question of what occasions self-control. The difference from Brandt's own answer is by no means trivial. We say that the need for self-control comes from the fact that what an agent judges it rational to do, abstracting-away-from-the-less-than-ideal-aspects-of‐ her-psychology, is not always the action which she is most strongly motivated to perform. But, unlike Brandt, we do not think that this is sufficient to require the exercise of self-control. We think that it is only necessary. For self-control is called for only if, in addition, the agent judges the exercise of self-control itself rational, taking-on-board-the-less-than-idealaspects-of-her-psychology: only if, having learned about those aspects, she continues to judge it rational for her to act in the manner prescribed. This is an important extra condition because, in many situations, it may not be met.

Consider an example. Suppose an agent's fully rational self desires excitement more than anything else, and so judges it rational for her less than rational self to go parachutejumping, abstracting-away-from-less-than-ideal-aspects-of-her‐ psychology. This is marginally more exciting than rock-climbing, and that is enough to recommend it as the thing to do. However, her less than rational self balks at the thought of parachutejumping: perhaps in the airdrome, perhaps on board the aircraft. She wants to sit tight and avoid the jump. Is this a situation that calls for self-control? Suppose the agent could get herself to jump by gritting her teeth. Should she just grit her teeth and get on with it? Not necessarily. It all depends on what her fully rational self desires her less than rational self to do, given that she is not motivated to go parachutejumping. Perhaps her fully rational self will desire her less than rational self to grit her teeth and get on with it. But perhaps she won't. If the cost of gritting her teeth is too great —if it will cause all sorts of pain and -43suffering, something the agent's fully rational self will want her less than rational self to avoid—then her fully rational self may prefer her less than rational self to get excitement in alternative ways. Perhaps she will prefer her to cut her losses and go rock-climbing instead. She will judge it rational for her to do this, and forget about self-control, takingon-board-less-than-ideal-aspects-of-her-psychology. 2. In such a case, we say the agent is not in circumstances that call for the exercise of selfcontrol despite the fact that, in one sense, she judges it rational to act in a way that she is not motivated to act—despite the fact that, abstracting-away‐ from-less-than-idealaspects-of-her-psychology, she judges it rational to take the parachute jump. We say that she is rather in circumstances that call for a change of mind. By our account, then, selfcontrol is called for only when the agent finds herself in circumstances where such a change of mind is not itself mandated; only in circumstances where the exercise of selfcontrol is itself rational taking‐ on-board-less-than-ideal-aspects-of-her-psychology.

III. What Constitutes Self-Control? Suppose two agents are relatively more strongly motivated to eat a second sandwich despite the fact that they each believe it is rational for them to eat only one, abstractingaway-from-less-than-ideal-aspects-of-their-psychologies. Suppose that they are also alike in their beliefs as to what can be done—for example, what therapies can be tried—to facilitate doing the rational thing. And suppose further that one of these agents has an extra intrinsic motivation of the kind Brandt describes, a motivation that the other lacks. The one has a desire to visit a therapist who will talk her out of eating a second sandwich whenever she finds herself in the situation just described. The other has no such desire. Then it seems to us that Brandt is right that what the agent who has the extra desire does, when she acts on her desire, should count as an exercise of self-control. And it also seems to us that Brandt is right that, in her case, the extra intrinsic desire she has therefore constitutes her possession of self-control in that case.

However, Brandt's position on what constitutes self-control is stronger than the claim that self-control may be constituted by an intrinsic desire. According to Brandt, a virtue like self-control simply is a desire to act in certain ways and only an agent with the extra intrinsic motivation that he describes possesses self‐ control. There is no other way in which an agent can possess, and so exercise, self-control. What does Brandt say, then, in support of this stronger claim? ____________________ 2. Gary Watson is well aware of this possibility in his discussion of desires that "coerce" rather than "compel" (Watson 1977: 326). -44He notes that, in a case much like that which we have been discussing, the agent who possesses self-control does things that the agent who lacks self-control doesn't do. And he argues that that shows that there is some independent source of motivation present: at least "enough to seek therapy and try to follow a constructive program" (Brandt 1988: 70). But, as is perhaps readily apparent, Brandt's premise simply does not support his conclusion. Even if the agent who possesses self-control does do things that the agent who lacks self-control doesn't do, it does not follow that what she does is explained by some extra intrinsic motivation. For, even by his own lights, there are alternative ways in which the agents could differ in their derived motivations. Consider an example. Suppose two agents are apt to lose self-control because, though they judge it rational to become more healthy, abstracting-away-from‐ less-than-idealaspects-of-their-psychologies, right now they are not thinking vividly about health, and so do not desire to be healthy to the extent that they should. Their desire for health is weaker than it would be if they were to think vividly about health. They are therefore, we will suppose, disposed to eat a second sandwich at lunch instead of refraining, despite the fact that they judge refraining the rational thing to do. But now, just suppose that one of these agents has a disposition the other lacks. Whenever she is in such circumstances, she is disposed to dwell on health, and, as a result of doing so, she regularly finds herself once more desiring health to the right degree. More to the point, in this particular case, she finds herself desiring health enough so that she refrains from eating a second sandwich. Does this person exercise self-control? It certainly seems as though she does. After all, how else should we describe her achievement? But it seems entirely wrong to suppose that we need credit such an agent with any special intrinsic motivations in order to account for her exercise of self-control. For, in this case, her exercise of self-control consists in her having certain vivid thoughts, and her having those thoughts is explained by a disposition she possesses to have those thoughts in the relevant circumstances. But a disposition to have certain vivid thoughts in certain circumstances need no more be a desire, it would seem, than having those thoughts need be an action. In this case, then, the agent's possession of self-control seems to be constituted by her disposition to have certain thoughts under certain conditions, not by an extra intrinsic motivation.

Consider another example. Suppose two agents are apt to lose self-control because, though they judge it rational to become more healthy, abstracting-away‐ from-less-thanideal-aspects-of-their-psychologies, and though they desire to be healthy to the extent that they should, they find it difficult to think vividly about how, say, refraining from eating a second sandwich at lunch time will help them to become more healthy. Instead they think vividly about how to satisfy another less strong intrinsic desire that they have. They think about how eating a -45second sandwich will give them pleasure. Because of instrumental irrationality of this kind, we will suppose that they are both disposed to eat a second sandwich rather than refrain. But now suppose further that one of these agents has certain other vivid thoughts that the other doesn't have. She thinks about an alternative way in which she can become more healthy. She thinks about the fact that, if she goes to see a therapist, the therapist will help her to eat less, and this, in turn, will cause her to become more healthy. And, let's suppose, these thoughts happen to be thoughts which remain vividly in her awareness, and so give rise to an even stronger motivation in her to refrain from eating the second sandwich and head off to the therapist instead. Does this person exercise selfcontrol? Again it seems to us that she does. But, again, it seems that we need credit her with no special intrinsic desires, over and above her desire for health, in order to explain her exercise of self-control. Her desire for health, together with the fact that she thinks vividly about certain indirect ways in which she can achieve health, is enough to explain why she heads off to see the therapist. And, correlatively, the fact that the other agent fails to think vividly about these indirect ways of achieving health may be enough to explain why she doesn't. In this case, then, the agent's possession of self-control seems to be constituted by her knowledge of the indirect ways in which she can achieve her goals, together with the fact the content of this knowledge can be represented vividly in her awareness, not by any additional intrinsic motivation. It is clear from these considerations that we should be more permissive in our account of self-control. "Self-control" directs us to that class of psychological mechanisms by which we can get ourselves to act in (roughly) the way we think it is rational to act, abstractingaway-from-less-than-ideal-aspects-of-our-psychologies, when we have a relatively stronger motivation to act otherwise. Brandt may be right that one such mechanism is the possession of certain intrinsic motivations of the kind he describes. But there are plainly other such mechanisms as well: for example, the possession of a disposition to have certain vivid thoughts under certain conditions and the disposition to think vividly about indirect ways of achieving our goals. And there may be yet further mechanisms too. Brandt is much too narrowly focused in his account of what constitutes self-control. We mentioned in our presentation of Brandt's view on the question before us that it connects with a general thesis that he defends about virtues: that all virtues are motivational dispositions and that it is a mistake, as a long tradition of thought holds, to

think that some virtues are managerial in character. The considerations rehearsed in this section are sufficient to give the lie to Brandt's general view. They show that some of the traits whereby we exercise self-control, some of the traits that count as virtues, are cognitive or related skills, not dispositions of a desiderative kind. -46-

IV. What Recommends Self-Control? Now to our third question and, as we shall see, a third area of contrast between Brandt's view and our own. We saw earlier that Brandt believes that certain traits are virtues— self-control, presumably, included—just so far as their possession is associated with good consequences. We agree that self-control generally has good consequences. But we believe that self-control is also attractive on other, arguably deeper, grounds: on the grounds that it is rationally desirable. In this section we ask whether Brandt is disposed to agree: whether he is of a mind, and whether he is in a position, to defend the same view. It appears that Brandt must be of a mind to think that self-control is rationally desirable. Suppose you take Brandt's view that to exercise self-control is to succeed, in spite of a strong opposing motivation, in doing what it is rational to desire and do. And suppose you believe, as he does, that this is the only route, in the circumstances, to such success. In that case you can hardly avoid holding that self-control is rationally desirable. But should Brandt be of a mind to regard self-control as rationally desirable, is he is in a position to take this view? It turns out that he is not. Consider an example discussed earlier. An agent is deciding whether to have a chocolate bar or a piece of fruit. She has only one intrinsic desire, a desire for pleasure, and this desire gives rise to a motivation to eat a chocolate bar. However, the reason she only has an intrinsic desire for pleasure is that she finds it hard to think vividly about what health is and involves. Her fully rational self, who does think about such things, has desires for health as well as pleasure, and, indeed, her desire for health is stronger. Let's suppose that the agent knows all this to be true. She therefore believes that it is rational for her to eat a piece of fruit, but she unfortunately finds herself relatively more strongly motivated to eat the chocolate bar. On Brandt's account, this agent is in a situation that calls for the exercise of self-control. Let's suppose, for simplicity, that there is only one way in which this agent could exercise self-control: by visiting a nearby therapist who will teach her how to think vividly about what health is and involves. The agent knows this too, and so begins to wonder whether she should exercise self-control. Would it be rational for her to have an intrinsic desire— or any trait—of the kind that Brandt thinks constitutes her self-control? In this case, would it be rational for her to desire to go and see the therapist? Would it be rational for her to act on that desire? Surprisingly, Brandt's own analysis of rational action and desire tells us that it would not be rational.

Brandt cannot say that it is rational to a first approximation for the agent to visit the therapist. For that is so just in case, if the agent were to think vividly -47about all of the options available to her, her actual intrinsic desires taken as given, she would visit the therapist. But of course, ex hypothesi, this condition is not met. The agent's actual intrinsic desires taken as given—her desire for pleasure—it is rational to a first approximation for her to eat the chocolate bar. And neither can Brandt say that it is fully rational for the agent to visit the therapist. For that is so just in case, if she had the intrinsic desires she would have after thinking vividly about the facts, and if she were then to think vividly about all of her available options, she would visit the therapist. But again, ex hypothesi, this condition is not met. The desire that it is rational for her to have, the desire for health, supports eating a piece of fruit, not visiting a therapist. Why would someone with that desire go and see a therapist? That desire would give her no instrumental reason to do so and neither would she form an intrinsic desire for such a visit: there is nothing intrinsically attractive, at least for the fully rational agent we are imagining, about visiting a therapist. The problem that blocks Brandt from saying that it is rational for an agent to desire and exercise a necessary measure of self-control, like visiting the therapist, is essentially the same problem that we raised in § II. That problem led us to present a revised analysis of rational desire and action, according to which it is rational for an agent in certain circumstances to desire or do whatever she, were she fully rational and informed, would want herself to desire or do in those circumstances. It turns out that this revised analysis, unlike Brandt's, does make it possible to say that self-control is rationally desirable. According to the analysis, an agent is in a situation that calls for self-control only if she is in certain circumstances in which she judges it rational to act in one way, abstractingaway-from-less-than-ideal-aspects-of-her-psychology, and yet she finds herself relatively more strongly motivated to act in another. Suppose, again, that the agent's options are to eat a piece of fruit or a bar of chocolate and that she has only a single desire, a desire for pleasure, which gives rise to a motivation to eat the bar of chocolate. The agent believes that her fully rational self would have an additional and stronger desire for health. And this leads her to believe that her fully rational self would want her less than rational self to eat the piece of fruit instead. The agent therefore finds herself more strongly motivated to eat the bar of chocolate despite the fact that she believes it rational to eat the fruit instead, at least abstracting-away-from-less-than-ideal-aspects-of-her-psychology. Suppose the only way in which the agent can exercise self-control, in the circumstances, is by visiting a therapist who will help her to think vividly about what health is and involves. Suppose, as Brandt generally supposes, that the only way in which the agent will go to the therapist is if she has an intrinsic desire to do so. And suppose further that the costs of visiting a therapist—in pain and suffering, say—are not so great as to

mandate a change of mind. Of the acts and desires that are real options for her, then, it is plausible on our account to sup -48pose that the agent's fully rational self would form a circumstance-specific desire that her less than rational self forms an intrinsic desire to visit the therapist and then acts upon her desire. For only so will she overcome her irrationality. Only so will her less than rational self do the things that are required to get herself back on the rational track. On our account, then, it is rational, taking-on-board‐ less-than-ideal-aspects-of-her-psychology, for the agent to desire to visit a therapist and to act upon that desire. It is rational for such an agent to possess and exercise self-control. A variant on this argument will go through, no matter what it is that the agent judges it rational to do, abstracting-away-from-less-than-ideal-aspects-of-her‐ psychology; no matter what sort of motivation outweighs her desire to do that; and no matter what measure offers itself as the means of self-control—provided, of course, that those means do not themselves mandate a change of mind instead. We can see therefore that on our approach, self-control presents itself as a trait that is rationally attractive, even rationally compelling, in a distinctive fashion. To fail to be self-controlled, when self-control is called for, is to fail in an important way to be rational.

IV. Conclusion The foregoing discussions should serve to outline the contrasts between the view of selfcontrol that Brandt defends and the picture to which we are drawn. Take the question as to what occasions self-control. For him and for us self‐ control is occasioned only when an agent who judges it best to act in one way is more strongly motivated to act in another. For him and for us what an agent judges best is what she judges that it is rational for her to do. But there the approaches come apart. For we offer a different and, we think, more satisfactory account of what it is rational for an agent to do, and this account leads us to distinguish, in a way Brandt doesn't, between two sorts of situations in which an agent is more strongly motivated to act other than as she judges best. Only some such situations present occasions where the exercise of self-control is appropriate. Others mandate a change of mind instead. On the question of what constitutes self-control, the contrast has a similar partial character. For Brandt and for us, self-control may sometimes be realized by the possession of an intrinsic desire for a measure—say, a therapeutic measure—whereby an agent is led, contrary to an otherwise compelling motivation, to do what she judges best. But we hold, on the basis of more or less commonsense examples, that that is not the only way in which self-control may be constituted. In some cases it may involve the presence of a suitable desire, but in others it can be realized by the presence of any of a range of cognitive habits. Self-control may be a managerial or a motivational matter.

-49Finally, to the question of what recommends self-control in those circumstances where self-control is really called for. Anyone who adopts Brandt's line on what occasions selfcontrol must want to think that self-control is rationally desirable. But here, again, Brandt's analysis of rational desire and action lets him down. It turns out, on that analysis, that it is not going to be rational for an agent to desire and pursue measures necessary for self-control. The problem is remedied, however, under our account, and so we can happily say that self-control is a rationally desirable trait.3 -50-

4 Brandt on Autonomy Michael Davis My thesis is that Brandt's A Theory of the Good and the Right offers a conception of autonomy different from, and at least as plausible as, any other now available. Many of Brandt's readers may find this surprising. Indeed, perhaps even Brandt would. What Brandt claims to offer is a conception of practical rationality, not autonomy. Yet if, as I shall argue, Brandt's conception of rationality can absorb the sort of case that makes autonomy seem a concept worth the attention it has recently received, autonomy theorists face a dilemma. On the one hand, they can deny that Brandt's practical rationality offers a conception of autonomy. That denial would leave utilitarians free to claim that they do not need one. Brandt's practical rationality will do. 1. On the other hand, autonomy theorists can admit that Brandt's conception of practical rationality is also a conception of autonomy. But that admission would invite hard questions about why autonomy has been so important to moral discussions over the last decade or so. Why talk about "autonomy" if the concept conveys nothing not conveyed less problematically by the more traditional concept of practical rationality? I shall conclude with a suggestion for answering that question.

I. Conceptions of Autonomy Autonomy is literally self-rule. The term may be applied to political entities ("an autonomous region of tile Russian Republic") or to individual persons, their acts, choices, or desires. My concern here is the latter, what is commonly called "personal autonomy". "Personal autonomy" has a wide range of applications, from the global ("Professionals need autonomy") to the local ("The patient's decision was not autonomous"). Whether applied globally or not, "personal ____________________ 1. Consider, for example, the effect on the argument in Haworth 1984. -51autonomy" can refer to either a state of affairs or a right. In what follows, I shall be concerned primarily with the general state of affairs a right to personal autonomy might protect, not with whether this factual autonomy is an ideal state actual people can only approach, a state actual people can be in to varying degrees (or in one respect but not another), a threshold (or all-or-nothing) state most people actually are (or could be) in, or some combination of these (cf. Feinberg 1989). I shall also have little to say about special cases of personal autonomy such as moral or artistic autonomy. My concern is personal autonomy as such.

Conceptions of personal autonomy seem to be of two sorts: (a) those for which "autonomous agent" is central; and (b) those for which "autonomous desire" is central. Agent-centered conceptions may say little (indeed, nothing) about "autonomous desires". For example, on Gerald Dworkin's recent account, a person is autonomous insofar as (and only insofar as) she has certain capacities ("the capacity ... to reflect on [her] first-order preferences, desires, wishes, and so forth, and the capacity to accept or to attempt to change these in light of higher‐ order preferences and values") (Dworkin 1988: 20). 2. Desires as such are neither autonomous nor non-autonomous. An act (or choice) is autonomous only insofar as it results from the exercise of the relevant capacities of the agent. Desire-centered conceptions, in contrast, consider a person to be autonomous (or to have personal autonomy) only if (or only insofar as) his acts (or other choices) are autonomous; and to consider his acts to be autonomous only if (or only insofar as) they derive from desires, motives, or the like that are themselves autonomous. Autonomy is primarily a characteristic of desires and only derivatively of acts or persons. Virtually all desire-centered conceptions of autonomy now available are either historical or structural. 3. In historical conceptions, a desire is autonomous if it results from a certain process (or is not the result of certain processes). 4. For example, my present desire to eat chocolate chip cookies would be autonomous if it ____________________ 2. For an even more global rendering of autonomy (one including elaborate environmental conditions), see Raz 1987. For a valuable sorting out of issues affecting global conceptions, see Benn 1988: esp. chs. 8 and 9. 3. The only possible exception I have encountered is in Thalberg 1978 [=1989]. This is ostensibly a criticism of various two-level structural conceptions (of "unfree actions") which he contrasts with his own admittedly sketchy one-level conception. But, in the course of explaining his one-level conception, he several times offers a "pragmatic" test, for example (1989: 128): "If he had been forewarned of this situation, and alerted to less coercive and otherwise no more disadvantageous alternative situations, he would have circumnavigated this one." Though originally stated for coercion, he later (1989: 131-32) applies similar hypothetical tests to manipulation, compulsion, and socialization. 4. For a convenient summary of the literature, see Christman's "Introduction" (1989: 320). Note that Christman's scheme of classification seems to rule out in advance both Dworkin's recent conception and Brandt's. -52actually originated in an appropriately reflective process, but not if it originated in hypnotic suggestion or early socialization into a "chocoholic" family.

For structural conceptions, a desire is autonomous if it stands in a certain relationship to other current desires of the person (whatever the desire's history). For example, according to Harry Frankfurt's well-known account (Frankfurt 1971), a person's (first-order) desire is autonomous if it is one with which that person actually identifies. The history of the desire is relevant, if at all, only insofar as it affects a person's willingness to identify with it. 5. According to another structural conception, a desire is autonomous if, and only if, it fits the person's actual life plan (Young 1980). For all their differences, desire-centered conceptions (and perhaps agent‐ centered conceptions as well) have certain common concerns. These may be taken both to provide an implicit definition of their subject and to set minimum conditions for their adequacy. We may consider those concerns under three categories: manipulation, compulsion, and socialization. We sometimes distinguish between desires (wishes, urges, and other motives) we have, thinking of some as "genuinely ours", "expressions of the real me", or "authentic", and others as somehow "foreign", "alien", "inauthentic", or "imposed". To act knowingly on foreign, alien, inauthentic, or imposed desires is somehow degrading, humiliating, or at least unpleasant. Some of these objectionable desires are clearly the work of others. One example common in philosophic discussions is the desire which one has solely as a result of hypnotic suggestion of which one is not aware. We would be outraged to discover that a hypnotist had, even temporarily, made us her unknowing marionette. Most of us view in the same way some desires we derive from advertising. We do not, for example, want to desire peanut butter simply because we have come to associate it with Peter Pan (much less because the ad misled or deceived us). Where we find that our desires have this origin, we are likely to say that we have been manipulated—and that such manipulation has reduced our autonomy (whether or not we also want to say it has reduced our freedom). Not all objectionable desires come from outside. Some of us experience desires which, though having no external source, still seem foreign. The alcoholic's craving for alcohol may seem like this. Both she and those who know her may think of the craving as a "disease", that is, as somehow imposed on her. The disease may have external origins (for example, early exposure to alcohol), but it would be just as much a disease if its source were the alcoholic's own genes. If the craving is strong enough to win out often against other considerations, including the alcoholic's life plan, we are likely to describe her as drinking compulsively (or drinking under compulsion). Such a description may excuse her from blame or ____________________ 5. Gerald Dworkin's early conception of autonomy also belongs here. See, e.g., Dworkin 1981. -53-

otherwise shield her from responsibility. But even when it does not, it signals that something is deeply wrong with the relation between the agent and the act. The agent is, as it were, under the thumb of an alien power. She lacks self-control. She is not autonomous in that respect. Sometimes, however, our perspective is different. The person having the desire may feel nothing untoward and yet we, observing the person, may still want to say that he is somehow not free. This can happen in cases both of manipulation and of compulsion, but it seems most common in cases of (what we are likely to call) "mere socialization". I may, for example, have tried without success to convince a servile housewife to support the equal rights amendment. In time, I may come to think that no amount of reasoning will win her over. Reason cannot reach her. She does not want equality. She has, as she says, been brought up to think that a woman should be her husband's helpmate. She cannot bring herself even to imagine the alternative to which my reasons appeal. She is not autonomous but (as we might say) "in bondage to her upbringing". Any adequate conception of autonomy should identify cases such as these as failures of autonomy and, as well, explain why we consider such failures to be bad (rather than merely indifferent or positively good). Autonomy is, as such, good (even if neither the only good nor the greatest). Any adequate conception should help us to understand why.

II. Brandt's Conception of Practical Rationality For Brandt, an act (or other choice) is "fully rational" if, and only if, it is rational to a "second approximation". An act is rational to a "first approximation" if the agent would have chosen it even if, taking his desires as they are, he had all the relevant information vividly before his mind. An act is rational to a second approximation if, in addition, the desires motivating it are themselves rational. 6. Brandt's conception of practical rationality is desire-centered. Brandt's conception is, however, neither historical nor structural. Rather, it is (we might say) "hypothetical" (or "counter-factual"). The desire must be capable of passing a certain test. A desire is rational if, and only if, it could survive "cognitive psychotherapy". The therapy consists of associating (with appropriate vividness, repetition, and balance) the desire in question with all relevant information available. The criterion of relevance is causal. Information is relevant if awareness of it would "make a difference to the person's tendency to perform a certain act, or to the attractiveness of some prospective outcome". The criterion of availability is individualized but objective. A piece of information is available if it would be among the beliefs the person held at the time he had formed his ____________________ 6. See, for example, Brandt 1979: 148. -54-

beliefs according to (a) the principles of logic, inductive as well as deductive, and (b) his "total observational evidence", including what he could have obtained at the time but did not. 7. For Brandt, then, my desire for a chocolate chip cookie is irrational if it would fade were I to remind myself that I have been putting on weight recently, that eating cookies is in part responsible, and that I will live longer and better if I do not put on any more weight. In contrast, the desire would be rational if the only way I could get it to fade would be to think of rotting garbage. For Brandt, what makes a desire rational is its ability to withstand exposure to relevant information, not its ability to withstand any conditioning whatever. 8. The rationality of a desire is an empirical question more or less independent of the desire's history or relation to other desires. No desire, whether native or learned, and whatever its connection with other desires, is necessarily immune to cognitive psychotherapy. So, for example, though my desire not to be burned at the stake rests more or less directly on a native dislike of pain, I still might come to desire being so burned if I repeatedly reminded myself of the cause being burned would serve. If my desire for the success of that cause could itself survive cognitive psychotherapy, my desire to avoid burning would be irrational (Brandt 1979: 132). Brandt takes it to be the tentative conclusion of psychology that four sorts of desire (or aversion) would generally be irrational (1979: 115-26). Two of these, desires based on "false beliefs" (such as hypnotists often use) or "generalization from untypical examples" (such as the testimonials used in advertising) need not detain us. Because such desires are directly produced by cognitive error, it is easy to see why cognitive psychotherapy should be effective against them. The third, "exaggerated desires produced by early deprivation", though not originally the consequence of cognitive error, still seems to depend on an incongruity with the world as it is. So, for example, someone whose desire for success at any cost was produced by her parents' early withholding of affection whenever she failed, might find her desire for success fade if she repeatedly reminded herself that now her family and friends would like her more were she to spend more time with them and less at the office. If so, her desire for success ultimately rests on social relations that no longer exist. Like the other two sorts of desire, this third has only (what we might call) "histórical force" (which is exactly what makes it vulnerable and so, irrational). The fourth sort of irrational desire, that produced by society when it would not have been produced by direct experience with the desired thing, often re____________________ 7. Brandt 1979: 12-13. This appeal to standard inductive and deductive logic raises difficulties not usually discussed. For some suggestion of what these might be, see Davis 1991. 8. Brandt 1979: 156. Note that his examples of successful de-conditioning (for example,

at pp. 114‐ 15 or 131 of Brandt 1979) do not indicate what "the facts" are. -55sembles the other three in having only historical force. But sometimes the situation is more complicated. The desire might be maintained in part by continuing reinforcement. For example, the servile housewife's servility may derive not only from what she learned as a child but also from the response of friends whenever she is at all assertive. If she would be much less inclined to servility after a year of being assertive in a less forbidding environment (one actually possible for her in this world), then (according to Brandt) her present servility is irrational. This analysis of rationality allows Brandt to provide a relatively convincing explanation of the motive any agent would have to be rational to a first approximation. An act rational in this sense is, by definition, one that avoids all mistakes deriving from inadequate reflection. It must be the one most likely to satisfy the agent's desires as fully as possible. Since (all else equal) we all want to satisfy our desires, we must (all else equal) want to act rationally to a first approximation (Brandt 1979: 153). Brandt is somewhat less successful at explaining why anyone would want to be fully rational in his sense (that is, rational to a second approximation). In place of an argument, he offers the observation (correct, I think) that, as a matter of fact, we do not like having to think of our desires as deriving from mere socialization or the accidents of past experience. We want instead to have desires that result from "sensitive reaction with the real world" (Brandt 1979: 157, 159). We want our acts (and the desires from which they derive) to be consistent with the best understanding of the world of which we are capable. Brandt explains what makes practical rationality a good thing by showing that it satisfies a certain second-order desire (the desire to act only on first-order desires that do not prevent us from acting on the best understanding of the world of which we are capable). This reference to a second-order desire may suggest that Brandt is in fact offering a structural conception of practical rationality. He is not. Brandt's test of full rationality, rationality to a second approximation, does not presuppose any actual second-order desire whatever. My first-order desire for, say, chocolate chip cookies, may survive (or fail to survive) cognitive psychotherapy whether or not I have the second-order desire to have only first‐ order desires of the appropriate sort (and, indeed, whether I have any second‐ order desires at all). Brandt refers to second-order desires only to explain why we should take the results of cognitive psychotherapy seriously, not to explain why the results are what they are. Brandt's therapy is, in part at least, a kind of reflection, presupposing a Dworkinesque capacity of the person who (at least hypothetically) undergoes the therapy. This may suggest that Brandt's conception of practical rationality differs in no important way from Dworkin's conception of autonomy. The suggestion should be dismissed—for two reasons. First, much cognitive psychotherapy may require nothing more than simple deconditioning (for example, association of

-56eating cookies with gaining weight, dying early, and so on to compensate for childhood associations with parental approval). Dworkinesque reflection is not always necessary. Second, even when such reflection would be necessary for the appropriate therapy, Brandt's conception of rationality does not require that the actual person reach her desire (or act) through that process or even be capable of doing so in her actual circumstances. The capacity for reflection need be present only "hypothetically". On Dworkin's conception, a person could not be autonomous if she had no capacity for reflection in her actual circumstances (however much she might be able to reflect if her circumstances were radically different). 9. On Brandt's conception, she could. Brandt's conception is, in this respect, less demanding than Dworkin's.

III. From Practical Rationality to Personal Autonomy That, then, in outline, is Brandt's conception of practical rationality. Is it also a conception of autonomy? I shall now offer two substantial reasons to think so. First, I shall show that Brandt's conception of rationality can distinguish manipulation, compulsion, and mere socialization from analogous influences to which we have no objection. Second, I shall show that Brandt's conception offers an explanation for why we find the autonomy-reducing influences, but not the others, objectionable. In short, I shall show that Brandt's practical rationality meets the minimum conditions we set for an adequate conception of autonomy. Let's begin with manipulation. What is wrong with getting someone to do something by hypnotic suggestion or the associative devices of advertising? The answer is that, typically, neither the hypnotist's suggestion nor the associations of advertising would survive cognitive psychotherapy. Their power to motivate depends on implanting false beliefs (or, more precisely, inclining us to make poor use of the information available). They are objectionable because they tend to undermine full rationality. On Brandt's conception of rationality, then, nothing is wrong with hypnotic suggestion or advertising that implants beliefs that would survive sensitive interaction with the real world. So, for example, an ad that leaves an accurate impression does not ____________________ 9. While Dworkin cautions against requiring too rich a mental life as a precondition of autonomy, he clearly thinks reflection of some sort must actually be part of a person's life for him to be autonomous. So, for example, he (1988: 17) tells us, "[A] farmer living in an isolated rural community, with a minimal education, may without being aware of it be conducting his life in ways which indicate that he has shaped and molded his life according to reflective procedures. This will be shown not by what he says about his thoughts, but in what he tries to change in his life, what he criticizes about others, the satisfaction he manifests (or fails to) in his work, family, and community." Brandt could, I think, recognize as autonomous people who were not reflective even in this weak sense.

-57threaten autonomy even if it achieves its effect through the same nonrational means as other advertising. That conclusion—and others to come—show that Brandt's conception of rationality differs in important ways from conceptions of autonomy now available. The differences may suggest certain objections to treating Brandt's conception as a conception of autonomy. I shall deal with those objections in the next section. For now, what is important is whether Brandt's conception has the resources to do the minimum we set for any conception of autonomy. Brandt's conception also has the resources to distinguish between compulsions and unobjectionable motives. Compulsions are strong motives, that is, motives hard to resist; but it is not merely their strength to which we object. We have other motives we find hard to resist, for example, for survival or justice, but we do not (ordinarily) label these "compulsions". We do not (ordinarily) think they reduce autonomy. How then does a compulsion differ from other strong motives? For Brandt, the answer must be that a compulsion is a motive that cognitive psychotherapy would weaken or destroy. Compulsions are objectionable because they are irrational. Their strength depends (in part at least) on something other than the way the world really is. Not the substance of the motive as such, or its strength as such, makes it a compulsion but the relation between these and the world. We can, then, easily distinguish between strong desires, like that for justice, which are not compulsions, and others, like the alcoholic's desire for drink, which may be (that is, would be if, despite any genetic base, it weakened in therapy). That brings us to socialization. On the one hand, we recognize that we must be socialized if we are to be human in anything more than the biological sense. Some socialization is necessary for autonomy. On the other hand, we do not want to be the product of "mere socialization". What could be the difference between the good socialization we do not consider adverse to autonomy and the bad socialization we do consider adverse? For Brandt, the answer is that good socialization would survive better use of the information available, while bad socialization would not. What is wrong with "mere socialization" is that it sets a distorting screen between us and the world as it actually is. The servile housewife is less autonomous than she would otherwise be precisely because (and only if) her insulation from the wider world freezes her desires, making her immune to my arguments. Her present desires are the product of unnecessarily limited experience. She would benefit from cognitive psychotherapy.

IV. Some Objections Considered I have not yet shown Brandt's conception of practical rationality to be an interesting conception of autonomy. All I have actually shown is that practical ratio -58-

nality must be a component of any plausible conception of autonomy. Since hardly anyone would deny that, that I can show it is hardly impressive. What I need to show, and what would be impressive, is that autonomy is no more than practical rationality (in Brandt's sense). That claim is not unprecedented. For Kant, moral autonomy necessarily consists of acting on practical reason (the categorical imperative). 10. Since, for Kant, moral autonomy is the general category of which other forms of autonomy are special cases, his position is surprisingly close to Brandt's. The difference is not in the role of practical rationality. For both, acting rationally is the opposite of being under external control. Rather, the difference between the two is that Brandt's conception of practical rationality (and so, of autonomy) is fundamentally empiricist while Kant's is fundamentally rationalist. For Brandt, contingent experience (the results of cognitive psychotherapy) determines what rationality requires. For Kant, an a priori test does the same job. Nonetheless, someone might object that Brandt's conception of practical rationality cannot provide a complete account of personal autonomy. How am I to answer this objection? As it stands, it is both plausible and unfair. It asserts that Brandt's conception is missing something. It does not propose a candidate. Its plausibility comes from unfairly leaving me to fill the blank. The ways of doing that are uncountable. Complete refutation is therefore impossible. The best I can do—without becoming tedious—is to dispose of only the most obvious ways of filling the objection's blank. While I cannot show that Brandt's conception misses nothing, perhaps I can briefly show that it misses nothing obvious. That, anyway, is what I shall now try to do. I shall begin with what may seem the most obvious omission. Brandt cannot (it may seem) distinguish between those desires that we merely have and those that are genuinely ours. Yet, this difference is at the heart of autonomy. So, for example, what makes good socialization good cannot be (as Brandt would have it) that wider contact with the world could not erase it. Socialization can only be good if it makes us autonomous, that is, makes us the kind of person whose desires, choices, and acts are (or, at least, can be) genuinely ours. But, no desire can be genuinely ours if we accept it uncritically. Brandt's conception of rationality in fact gives a specific interpretation to the concept of a-desire-being-genuinely-ours. For Brandt, a desire is genuinely ours if, and only if, we have it now and would still have it after cognitive psychotherapy. To claim (as the objection does) that this cannot be right, that autonomy does not allow us to count desires as genuinely ours unless we came by them "critically", is simply to claim that autonomy is a historical concept. That is unfair for two reasons. ____________________ 10. Compare, for example, Hill 1989 or Korsgaard 1989. -59-

First, the concept of autonomy is plainly vague enough to allow for plausible (if nonetheless mistaken) conceptions that are structural rather than historical. We have only to survey the literature of autonomy to see that. Brandt's conception, whatever its oddities, is not alone in ignoring history. So, its being ahistorical cannot be a reason to deny that it is a conception of autonomy (even if it were a good reason to deny that it is a good conception of autonomy). Second, the concept of autonomy is plainly vague enough to allow for plausible conceptions not requiring desires to undergo criticism before they can be genuinely ours. So, for example, conceptions of autonomy relying on mere identification with the desire have been popular even though they resemble Brandt's in providing an interpretation of "genuinely ours" permitting a desire to be genuinely ours even if it has not undergone criticism. Criticism cannot be built into the very idea of autonomy. A plausible conception of autonomy may indeed require critical evaluation of desires. But that must be proved by something other than appeal to the concept as such. That brings me to a second objection. Brandt's conception of rationality seems to have nothing to say about coercion. Consider a typical example of coercion: The robber points a gun at you and says, "your money or your life". In this way, he may take over part of your life for a time. You had planned to use the money for food. Now you will (as we say) be forced to hand over the money to him instead. He has changed what you will do (just as a manipulator might). But, on Brandt's conception of rationality (and anyone else's), the robber does not touch your rationality (unless, of course, you panic). Handing over the money may be (and usually is) the rational thing to do under the circumstances. If autonomy is just practical rationality, then your decision to hand over the money will be autonomous. Your act can be at once, and in the same respect, both autonomous and coerced. But (it will be said), surely this conclusion shows that Brandt's practical rationality has missed something important in autonomy. I think not. Look again at the example. The robber does change what you will do, just as a manipulator might, but he does not do it—as the manipulator would—by changing the way you see the world. The robber changes how you act by changing the world. He is "outside". If you have a new desire—the desire to hand over the money—it is new only in a derivative sense. You preferred handing over the money to dying even before you met the robber. He has merely made it reasonable for you to act on that abstract preference. You have lost control of the situation, but not of yourself. He has attacked your liberty, not your autonomy. Coercion enslaves us in a way different from the way manipulation, compulsion, and socialization do. Coercion as such merely interferes with liberty, our rule over things, not over ourselves. 11. ____________________ 11. Compare Thalberg's discussion of coercion (including his concluding contrast with manipulation), at 1989: 126-29. -60-

This is not to say that coercion cannot reduce autonomy. We can easily imagine cases, for example, a concentration camp, in which coercion is so frequent or sustained that it destroys autonomy. Coercion can, over time, beat us down until we lose all control of ourselves, until we become dumb beasts of burden, unthinking automatons. But that is a point about certain forms of coercion, not coercion as such. Indeed, coercion of that sort is more like socialization than like ordinary coercion. What then are we to make of the almost universal agreement that "coercion situations seem like the paradigm cases of the loss of autonomy" (Christman 1989: 8)? We need to make some distinctions. In particular, we need to distinguish between autonomy in the sense we have been using it so far, autonomy in fact, which simply consists of fully rational action, and a juridical sense, a sphere of action protected for the sake of autonomy in fact. The right to autonomy assures the agent enough elbow room to act rationally, whether he in fact acts rationally or not. Factual autonomy can be reduced or destroyed, but only juridical autonomy can be invaded or violated. Consider the robber again. The robber might in fact only attack your liberty, leaving your autonomy untouched. But he might still invade your right to choose autonomously. The right might forbid coercion like the robber's because if robbers (and others who attack liberty) were common enough, you would lose all control of yourself and become instead a hand-me-down slave, passed from one temporary master to another, their acts combining to deny you the time, incentive, and other conditions necessary to maintain practical rationality. By protecting your liberty, the right of autonomy might protect autonomy in fact from this cumulative destruction even if no particular robbery would itself reduce anyone's autonomy. So, the objection that Brandt's conception would not count ordinary coercion as a loss of (factual) autonomy is both correct and (arguably) irrelevant. The objection has, nonetheless, proved useful. It has forced me to distinguish between factual and juridical autonomy. This in turn allowed me to explain the close relation between autonomy and liberty without fusing them into a single confusing concept. This seems to me an important contribution to understanding autonomy. That brings me to a third obvious objection to treating Brandt's practical rationality as a conception of autonomy. Because Brandt's cognitive psychotherapy does not require that the being subject to it be capable of critical reflection, his conception of practical rationality applies to any being capable of sensitive interaction with the world, even dogs or other "dumb animals". 12. Yet, what could be ____________________ 12. For those who find this assumption too much to entertain, I would suggest Fox 1974: esp. pp. 154-64 (a chapter entitled "Socially and Emotionally Maladjusted Dogs: Canine Delinquents and Sociopaths"). Many of the suggestions Fox makes for curing particular mental disorders (in this -61-

more absurd than to suppose an unreflective animal could be autonomous? Hence, Brandt's practical rationality cannot possibly provide a complete conception of autonomy. My response to this objection is much the same as my response to the one before. I admit that Brandt's conception of rationality commits us to recognizing dogs as capable of autonomous action (assuming they are capable of action at all). Indeed, I think that a merit of his conception. After all, there is a perfectly straightforward sense in which some dogs are in fact (more or less) autonomous, that is, in full control of themselves. They do not fawn, ape us, do only what they are told, or behave stupidly. They seem to see the world through their own eyes, to exhibit an understanding of some things we lack, as if they had developed a sense of things through their own sensitive interaction with the world. The force of the objection comes, I think, from confusing this straightforward sense in which dogs (or any other unreflective animal) can be autonomous in fact with the quite different question of how much we are required to respect that autonomy. That we can interfere with a dog's autonomy is one claim; that we shouldn't interfere with it is quite another. One benefit of Brandt's conception is that it forces us to keep these two claims distinct, rather than, as most other conceptions seem to, implicitly resolve the second while explicitly dealing only with the first. This may still seem a small point when the subject is the autonomy of dogs. Perhaps it will seem much less so if the subject is changed to what control the retarded, the aged, or those not quite of age should have over how they are treated. 13. That brings me to the last objection I shall consider, the most important of the four, though the easiest to answer. Given Brandt's conception, many acts we would normally consider invasions of autonomy would in fact increase autonomy rather than reduce it. For example, the hypnotist whose suggestion causes my craving for chocolate chip cookies to fade would, on Brandt's conception, increase my autonomy—whether she acts with or without my consent. This may seem utterly perverse. The current interest in autonomy seems largely the result of concern about paternalistic intervention in our lives. We appeal to autonomy (in part) to help explain what would be wrong with a hypnotist taking on herself to change my mind by nonrational means. Any conception of autonomy that seems to approve such intervention, or even to recognize it as increasing ____________________ chapter and elsewhere) are forms of de-conditioning of just the sort Brandt includes in cognitive psychotherapy. 13. Note, for example, that Illinois' Powers of Attorney for Health Care Law, §§ 4-6, specifically permits revocation of the power of attorney "without regard to the principal's mental or physical condition." Autonomy (a right of self-government) is here granted even to the (legally) incompetent. Why? Is the purpose to avoid the inspection of decisions that can interfere even with rational choice? Or might there be another purpose? -62-

autonomy, would thus seem to mock the very motive from which it must seek validation. Again, we must take care to distinguish between factual and juridical autonomy. The objection to what the hypnotist did is not, it seems, the hypnotism as such. So long as the results are the same as would be achieved by cognitive psychotherapy, we can admit that the results might increase autonomy. I have, after all, been made more rational. So, the objection must be to the hypnotist's paternalism, her using hypnotic suggestion on me without my consent. She usurped control. If this is indeed the objection, it is an objection to the hypnotist's treating me as if I were not an autonomous agent capable of myself making a rational decision to use, or not to use, hypnotic suggestion to cure my craving for chocolate. The objection, in other words, is that the hypnotist has violated my right to control over a certain range of decisions about myself. The hypnotist has violated my juridical autonomy, even while increasing my autonomy in fact. Unless I have control over what hypnotists may do to me, I will be in danger of becoming their marionette. Brandt's conception of practical rationality thus helps clarify what we in fact find objectionable in what the hypnotist did (when we do object).

V. Final Comments I have, of course, not proved that Brandt is right about practical rationality or personal autonomy. That was not my purpose. I myself have criticized Brandt's conception of practical rationality elsewhere (Davis 1987). If those criticisms were sound (as I think they were), they will resound in any conception of autonomy relying on that conception of rationality. My purpose here has been the more limited one of pointing out that Brandt's conception of practical rationality provides an approach to autonomy sufficiently different from others, and sufficiently rich in itself, to deserve attention (including judicious amendment). Brandt's empiricism provides a content most conceptions of autonomy lack. That content in turn gives us a way, indeed, almost forces us, to distinguish between autonomy in fact (practical rationality) and the sphere of action protected so that we can be autonomous in fact (juridical autonomy). To be treated as an autonomous agent (that is, as an autonomous agent deserves) is to be treated as if one's action is fully rational (whether it is in fact or not). While Kant's conception of autonomy also forces us to recognize the substantial independence that the right of autonomy might have from the fact of autonomy, contemporary discussion seems to assume, almost without argument, that the fact of autonomy demands a right more or less coextensive with the fact. This implicit assumption seems to have generated a good deal of confusion of which I will give one significant example. -63Gerald Dworkin recently (1988: 15-16) asserted that "[personal autonomy is] intuitively a global rather than local concept". His argument was brief. That is surprising. Not only

have most contemporary writers on the subject treated autonomy as a local concept, that is, as applicable first to individual desires, choices, or acts and only derivatively to agents as such; Dworkin himself had earlier taken the same view. What then did Dworkin say on behalf of his intuition? Only this: It [autonomy] is a feature that evaluates a whole way of living one's life and can only be assessed over extended portions of a person's life, whereas identification [his old way of conceiving autonomy] is something that may be pinpointed over short periods of time. We can think of a person who today identifies with, say, his addiction, but tomorrow feels it as alien and who continues to shift back and forth at frequent intervals. Does he shift back and forth from autonomy to non-autonomy? (Dworkin 1988: 16) Dworkin seems to think the answer is obvious: No! In fact, so long as we take the question to be about autonomy in fact, the obvious answer would seem to be: Why not? No one is in fact fully autonomous anyway. Most of us are more autonomous some days and less autonomous others. So, no one can shift back and forth between autonomy and non-autonomy—unless autonomy in fact is, as seems plausible, a threshold (or all-ornothing) concept. But, if it is a concept of that sort, daily fluctuations in degree of autonomy could well carry an agent back and forth over the threshold (though not in just the way Dworkin describes). So, as an argument from intuitions concerning autonomy in fact, this argument is, if anything, counter-intuitive. The argument takes hold only if we interpret it as resting on intuitions about the right to autonomy. The right to autonomy is a "global concept", that is, one applying, if at all, to an agent over an extended period, not to her desires, choices, acts, or even brief timeslices. The person Dworkin describes forces us to recall this just because he seems so close to disintegrating that he hardly seems an agent at all. Such a person could not shift back and forth between (juridical) autonomy and (juridical) non-autonomy (even if some of his desires are in fact autonomous). While in the condition Dworkin describes, he is always in a state of (juridical) nonautonomy. His instability makes us want to deny that he has any right to be treated as a rational agent. That explains the intuition Dworkin appeals to. That, however, does not show that Dworkin can save his argument simply by making clear that his concern is juridical autonomy. Imagine a person who has one set of desires when asleep (or delirious) and another when awake, and who alternates between those two states from day to day. Such a person would fluctuate from day to day between (juridical) autonomy and (juridical) non -64autonomy (and such fluctuation is far from impossible). Apparently, the relation between juridical and factual autonomy is complex.

So, Dworkin's argument will not work on either a factual or juridical interpretation. Whatever power it has comes (in part at least) from confusing factual and juridical autonomy. That a writer as astute as Dworkin could go this wrong should give us all considerable incentive to pay more attention to the distinction between factual and juridical autonomy—and to pay more attention too to conceptions of autonomy that, like Brandt's, help us do that. Once we begin to pay more attention to that distinction, we shall, I think, come to see that juridical autonomy ("the right [of a rational agent] to be let alone") is more interesting than autonomy in fact (practical rationality). What then explains current interest in factual autonomy? The answer, I think, is the many amputations the concept of rationality has undergone over the last three centuries. Rationality has been reduced to reasoning in accordance with the principles of deductive and inductive logic, or consistently trying to maximize whatever one values, or some other relation among one's beliefs, desires, and choices. Rationality has come to seem subjective, that is, a state of an agent independent of the rest of the world. Brandt's conception of practical rationality, whatever its defects, is a conception of autonomy precisely because it is not subjective in this way. An agent cannot be rational unless his beliefs, desires, and actions are consistent with all available information, not only with what he happens to have but with what he should have. We might then guess that current interest in factual autonomy—and in "brains in a vat"— would decline once we found a satisfactory way to recapture the critical power that the concept of rationality had before Hume made reason the slave of the passions. Whatever criticism the details of Brandt's approach deserves, its overall direction invites us to think of reason not as a mere relation among ideas but as a relation between our ideas and the world. That, I believe, is an invitation we should promptly accept. 14. ____________________ 14. I should like to thank the National Endowment for the Humanities for allowing me to participate in its 1990 Summer Seminar on Autonomy. I should also like to thank the other participants, especially Gerald Dworkin, the seminar director, for the critical discussions that helped to shape this paper. -65-

5 Brandt's Methods of Ethics R. M. Hare Richard Brandt is perhaps the most solid of all the American moralists of his generation. By "solid" I mean that his arguments are consistently well constructed and lead to determinate conclusions—unlike those of some of our contemporaries who have enjoyed wide recognition. He always makes clear what he thinks, and why. It is therefore never a waste of time to examine his views, even if in the end one does not agree with all of them. For myself I have found that our views tend, as they develop, to converge; certainly I owe Brandt a very great debt. My title needs some explanation. I am not using the expression "Methods of Ethics" in exactly Sidgwick's way. It has, as I shall use it, two different meanings, corresponding to different senses of "Ethics". The first covers what Brandt calls, in the first chapter of A Theory of the Good and the Right (1979), "Strategies of Practical Criticism": answers to his question "How to Proceed?". This is a very general question about method, which could be asked about other areas of philosophy too. Probably it is the different answers to this question that will continue to divide Brandt from typical "ordinary language" philosophers like me, especially since Brandt comes down more firmly than Sidgwick in favor of his own chosen method. The second sense covers what Brandt (1979: ch. 12) calls "different derivations of the right": that is, different ways of construing, leading to different procedures for answering, the question "Which actions are morally right?". He there instances the Ideal Observer Theory, my own theory, and that of John Rawls. He adds (1979: ch. 14) to the list egoism, act-utilitarianism and utilitarian generalization. Rejecting all of these, he advocates what he calls "A pluralistic welfare‐ maximizing moral system", which is a development of his earlier rule‐ utilitarianism. I shall discuss each of these choices of method in turn, but only in relation to the choice between Brandt's favored methods and my own. I largely agree with his rejection of the other methods he considers, and with his reasons for it. I take -67first the alternatives considered in Brandt 1979: ch. i. Brandt rejects what he calls "appeal to ethical intuitions". Since I agree with him about this, and have myself argued in a similar vein (e.g. Hare 1973; 1981: 10 ff.), I need do no more than express my continuing agreement. It is really surprising how philosophers go on appealing to moral intuitions as the basis of their arguments in the face of objections like Brandt's. This leaves two possibilities which Brandt calls "appeal to linguistic intuitions" and "a reforming

definition". The first he attributes to myself among others. I will now try to defend my own version of this "strategy". It consists in first clarifying the questions we are asking when we ask moral questions. This will determine the meaning, and thus the logical implications, of moral statements. We may hope on the basis of this to find a method of moral reasoning which is supported by logic. This will proceed, essentially, by discarding contradictions in our thinking and in our wills. The hope (which seems also to have been Kant's) is that having eliminated these we shall be left with just one answer to a given moral question that we can accept. Brandt offers several objections to this method. The first is that if we try to find out what people mean when they ask moral questions, we shall get lost in the vaguenesses of "ordinary language". Perhaps this is true of many philosophers, who have indeed got lost. But we must not give up too soon. The careful study of our ordinary uses of the moral words can be extremely rewarding, and Brandt, by neglecting it, may have missed insights at least as important as those he gets from his study of the psychologists. It is true that the moral words have different uses in ordinary speech. Many other words, including the logical connectives, are in the same boat. For example "and" can sometimes represent simply the sign of conjunction, which is subject to commutation, so that "p and q" is equivalent to "q and p"; but sometimes, as in "He put on his parachute and jumped out" it can carry also the implication that the event reported by the first conjunct precedes in time that reported by the second. And "if" notoriously has many different uses. It would be surprising if "good", "right", and "ought" were any more univocal in ordinary language. The sensible course for logicians to take in the face of these ambiguities is to sort out the different senses, which is always a difficult task. Sometimes this sorting out is best done by constructing an artificial language in which the different meanings can be clearly distinguished, as has happened with "or", the sign for disjunction (exclusive or inclusive). But sometimes the necessary distinctions can be made in ordinary language without recourse to formal systems. A careful study of the moral words as we ordinarily use them has revealed features of their logic which could hardly have been discovered in any other way. For example, it revealed to Stevenson the distinction, within the meaning of moral statements, between the "descriptive" and the "emotive" or "expressive" (or better, "evaluative") elements. Brandt himself uses this distinction at Brandt 1979: 193 ff. and elsewhere. Further study has revealed the connection of descrip -68tive meaning with the so-called "universalizability" of moral statements, and the supervenience of moral properties. The prescriptivity of moral statements has also been the subject of useful study, though its results are more controversial; it is hard to see how we could determine what precisely could be meant by calling moral statements "prescriptive" without attending carefully to

the implications that they have as people ordinarily use them. This feature, and its relation to the commendation (or the opposite) that moral statements imply, is going to have a great bearing on the method of moral argument. If, as I think is the case, there are both prescriptive and non-prescriptive uses of the moral words, the distinction between them has important implications for the problem of weakness of will that has exercised philosophers ever since Socrates, and also for the dispute between internalism and externalism in ethics. So, even if, as Brandt thinks, the study of the meanings of the moral words will not take us the whole way, it will certainly get us started. He gives few examples of the troubles that, in his view, beset the "ordinary language" moral philosopher, and they do not suffice to make his point. One kind of examples can be dealt with easily by invoking the distinction between descriptive and evaluative meaning just mentioned. If the moral concepts of religious people are different from those of Bertrand Russell, or those of unruly east side boys from those of Moore or Sidgwick (Brandt 1979: 7), the explanation may be that, although all these people are using words like "good" and "ought" for commending, they have different standards of commendation, and therefore different descriptive meanings for their moral words. It is illuminating and socially helpful to say this, because, if it is true, then, although the parties have different standards, they mean the same evaluatively by "ought". There is therefore something to build on in mutual discussion if they are ready to engage in it; the evaluative meaning carries with it certain implicit logical rules which the participants can be got to recognize, and which can guide the discussion. I have argued elsewhere (Hare 1992: 98-112) that if the unruly boys are, at least evaluatively speaking, using "ought" in the same way as the rest of us, there is a possible educational approach to them which might lead them to change their standards; whereas, if they were using the word in an entirely different meaning, communication would have completely broken down. This is not, however, a full explanation of the different uses of the moral words in different cultures. Evaluative and normative language has certainly changed and developed in the course of history, even in its evaluative meaning and the logical rules which that determines. It probably carries requirements now which it did not in ancient Greek times. It is often said, for example, that the normative language of Aristotle was not governed by a rule of unrestricted universalizability, though in books 8 and 9 of the Nicomachean Ethics he gets a long way towards formulating the doctrine as governing our relations with our friends. It was left to the Stoics and Christians to extend it to cover all people, -69and to Kant to give it its definitive logical expression; though even he did not extend it to cover non-human animals, and so he had to be corrected by his disciple Leonard Nelson and by Peter Singer, following Bentham (see Singer 1975: 8; 1981: 120 ff.). These advances in moral culture got built into the moral language, which is now, as a consequence, different from what it was. This is like the way in which chemical language has changed; "H 2 O" used not to be a possible meaning of "water", because in those days

nobody had heard of oxygen and hydrogen; but now it is (see Hare 1984, and compare v. Wright 1941: ch. 3). Those who consult linguistic intuitions do not need to be put off by these developments; they have to trace them, and this too can be illuminating. Further, the best way to find out what ordinary people mean by their moral questions is seldom to ask them for an explicit definition; we can readily agree with Brandt that this can lead us astray. Nor are explicit definitions normally used in teaching children these words (or, indeed, most other words). Matters are even worse if we ask for a definition in purely descriptive terms, such as he sometimes seems to contemplate (see, e.g., 1979: 5, top); this will at best inform us of the moral standards of the person we ask, which will vary from one person to another. Brandt is tempted to ask for such descriptive definitions because of a lingering bias in favor of ethical descriptivism, though he has in fact by now abandoned this (see below). The right way to inquire into the meanings and logical properties of normative words as people use them is to observe carefully what they think their moral statements imply or are inconsistent with. This can become clear, even if they have no ready-made definition to offer. The Socratic method, which got moral philosophy started (cf. Aristotle Metaphysics * 987b i ff.), operated in this way. Plato made it less destructive, and actually produced suggested definitions of his own; and moral philosophers have been at it, with useful results, ever since. I certainly think that by following this method I have got to understand moral language a lot better than I did when I started; and this ought to be the experience of any dedicated moral philosopher. For these reasons the linguistic moral philosopher does not need to be troubled by the vagueness of ordinary moral language. But Brandt has a second complaint, that ordinary language embodies confusions (1979: 7). Anybody who says this invites the retort that the confusions may be his own; but we can admit that ordinary people are often confused. However, their language provides the means of getting out of these confusions —means which Brandt himself uses. His example is the familiar puzzle as to whether a person who does what he thinks is his duty must be therein doing his duty. Prichard discussed this (1932) and distinguished between two meanings of "duty" which he called "objective" and ____________________ * References for Aristotle will be to the Bekker pages. -70"subjective" duty; and the matter has been further clarified since. As Brandt says "there is nothing in English usage which positively excludes recognizing different meanings for the two different contexts", and "If one sees something mistaken in this argument, it is because one has studied moral philosophy and seen that there are possible different kinds of claim which must be distinguished" (pp. 7 f.). But what is this if not to distinguish between meanings—a linguistic activity.

The main task of moral philosophy is, indeed, to look carefully at our moral concepts and sort out these confusions. Often it is not the language that is confused, but the people who use it, through inattention to its niceties. But there is another more serious objection that might be brought against linguistic moral philosophy, though Brandt himself does not give it prominence. This is that, even if we did give a correct and clear account of the existing moral language, nobody would be bound to go on using it if he wanted to say something different. Do we want to make ourselves slaves to our existing conceptual scheme? If this means tying ourselves to the descriptive meaning of our moral words, we obviously do not. We want to be free to adopt new moral ideas, as was done when the gospel was preached that we ought to love our enemies, or is done now when people say that we ought to treat non-human animals better than has hitherto been thought obligatory. But to change the content of morality is not necessarily to change its form. One of the wonderful features of moral language, which those who brush it aside have often missed, is that we can ask the same moral questions but, because we have different standards, give different answers to them—answers we can then discuss, regulated in our discussion by the rules of use for the words which we share. To be governed by these common rules is to be governed by their logic. It is therefore worth digressing to ask how we are governed by logical rules in general—a vast topic which I shall be able to treat only summarily. Suppose that someone were to suggest handling the word "all" as casually as some philosophers have suggested we might handle "ought" as used in morals. He might want to use "all" in future to mean the same as "some" means now. This is closely analogous to the maneuver of those who want to use "ought" in a way that is not universalizable, so that he can say, without logical offense, "I ought, but nobody else ought in circumstances however similar". With "all", he might want to use the word in such a way that he could, without logical fault, say "All the books in the shelf are blue, but there is one which is not". This would be quite all right if "all" were being used in the sense of "some"; and someone could indeed do this. But it would not make us alter quantificational logic. This brings up the question of the sense in which the rules of logic are conventional—a subject on which there has been much confusion. There is a sense in which they are conventional and another in which they are not. In the first sense, unlike the second, the claim that they are conventional is quite harmless. -71All the words we use, including the logical words, owe their meanings to conventions for the use of those sounds, or marks on paper (whatever Plato may have maintained in the Cratylus). However, it is a necessary truth that if we go on using them in accordance with those conventions, and thus with the same meaning, we have to stick to the same rules, because the meaning determines the rules and vice versa. Conventionalism in the other, unacceptable, sense denies this. It claims that one could go on using the words with the same meaning but with a different logic. In our example, we could then claim that we could go on using "all" with its ordinary meaning and yet say, without logical fault, "All

the books are blue but there is one which is not". But actually we cannot do this without changing the meaning of "all" or of some other word in the sentence. The same applies to the suggestion that we might change the moral language. We could certainly do this if we wished, and if we could secure the understanding of those we were talking to. We should then be asking a different question when we asked what we morally ought to do. It would help avoid confusion if we used a different word for the new concept. We might, as Simon Blackburn has suggested, say "shmorally ought" instead of "morally ought" (Blackburn 1984: 220-21); and we might free the new word from the requirement of universalizability. Actually we do not need to do this, because we already have in language a prescriptive expression which is free from this requirement (the plain imperative); but someone might like to have a more sophisticated version of this. We should then be asking a different question when we wondered what we shmorally ought to do, or, in plainer language, what to do. There is nothing to stop us asking this question; we already often ask it in the plainer form. But this would not free us from the requirement of universalizability which governs the question "What, morally speaking, ought I to do?" in its old sense. The linguistic innovator could say that he was not interested in this old question. Amoralism is very fashionable, among moral philosophers as among avant garde intellectuals generally. It is not yet fashionable with "unruly" east side boys, who have their own quite firm moral rules different from ours; but it will probably become so. I have admitted (Hare 1981a: 182 ff.) that a moral philosopher cannot bring any conclusive logical arguments against a consistent amoralist; all he can do is show the disadvantages in becoming one. This is a problem which faces all of us, and faces Brandt in a different form, as we shall see. But I can claim this much, that if we go on asking the same questions, we are bound by the logical rules governing the words in which we frame them. The question remains of what these rules are, and I cannot claim without argument that they are as I say they are. For the arguments I must refer to my other writings. But I will just give my reply to a counter-argument of Brandt's, which is fimiliar. He says (1979: 9) "Suppose I demand of myself, on grounds of conscience, that I perform a certain act, but am not prepared to make a similar demand on others for similar cases." His implication is that I should then be using -72"ought" non-universalizably. Brandt does not actually say that I should be using "ought" at all. I might reply that "ought" is not a very suitable expression for what I am thinking. I would have done better to say "If I did not do this, I should be falling short of my own ideals and standards". If I used "ought" I might mean it hypothetically: "ought, if I am to live up to my ideals". Someone who meant it in this sense would not be required by logic to extend the prescription to other people who might have different and less demanding ideals. Whether this is the correct explanation of such cases would have to be a matter for careful inquiry into what people do mean when they say this kind of thing. The inquiry

might leave unscathed our ordinary sense of "ought" in which it is universalizable, as is evident from many more common examples (see Hare 1955 = 1972a: 21). That then is my defense of the appeal to linguistic intuitions. We must now turn to Brandt's own proposal of a "reforming definition". Whether this is of practical help will depend on how likely people are to discard their existing meanings for the moral words in favor of his proposed new definitions of "good" and "right". The basis of these is an assignment of "a clear and useful meaning to the term 'rational"'. The main traditional questions of moral philosophy can, he thinks, "be satisfactorily rephrased as questions about what it is rational to want for itself, rational for a person to do from his own point of view, and about which forms of action would be permitted by a moral system for a society which it would be rational for a member of that society to support" (1979: 1 f.). He later defines "morally wrong" as "would be prohibited by any moral code which all fully rational persons would tend to support in preference to all others or to none at all, for the society of the agent, if they expected to spend a lifetime in that society" (1979: 194). So everything goes back to his reforming definition of "rational desire". I do not wish to quarrel with this basic definition of "rational"; indeed, I have given my support to it elsewhere (Hare 1981a: 214). Brandt puts it thus: "I shall pre-empt the term 'rational' to refer to actions, desires or moral systems which survive criticism and correction by facts and logic" (1979: 10). This process of "criticism and correction" of desires he calls "cognitive psychotherapy" (1979: 11, 113). The term "rational" is of course notoriously ambiguous, and Brandt notes a number of different senses. That is why he says "pre-empt". This indicates that he does not want to stop, for example, Rawls or the economists he follows using the word in their way to mean, roughly, "prudent"; nor would he now, I suppose, wish to deny Allan Gibbard (Gibbard 1990: 6 ff., 45 ff.) his totally different use of the word, as a way of expressing a norm. The important question is this. Are the moral words, redefined according to Brandt's "reforming definitions" in terms of his definition of "rational", adequate substitutes for our ordinary uses of them? In other words, if we ask the questions that Brandt poses in terms of his new definitions, shall we be content -73to give up asking the old questions? Even if we grant all that Brandt says about what it is rational to want, or what moral system it is rational to support, in his sense of "rational", do we have to agree that we ought to do what such a system prescribes? Or will the old questions remain? This is a question which we can hardly answer without a careful analysis of the old questions, which Brandt in his 1979 book does not attempt. Someone who was interested merely in polemics might suggest reasons why we would not be content. This is best discussed in connection with Brandt's choice of a method of ethics in the second sense mentioned above (that in which Egoism, Act-utilitarianism, etc., are rival methods). But before we come to this I need to make some preparatory remarks about Brandt's approach. First I want to ask whether Brandt is still a

descriptivist, as he seems to have been when (in the '50s) he wrote Ethical Theory. There is some evidence in his 1979 that he is not. For example, on p. 168 he says "The moral code could hardly control interpersonal behavior if individuals could not communicate the relevant motivations and feelings; and a special terminology is necessary for directing and articulating the basic motivation, as when an agent must think through whether his moral motivation is on balance in support of a bill before the legislature." If the terminology Brandt is speaking of here is the moral language, he must think that it has this directive function. On p. 170 he speaks of "a person's normative statements as expressing his moral code". This is language which would be appropriate in a prescriptivist like me or an expressivist like Gibbard. I take this as an indication of the convergence I mentioned in my first paragraph. If Brandt really were an expressivist or a prescriptivist, he ought not to say the sort of thing that he says on 1979: 150 f. There he deprecates the person who "takes the statements (e.g. 'that is the best thing to do') simply to express alreadyformed intentions to act, or actual desires". "In that case, of course", he goes on, "there would be no gap between intention/desire and these judgments—but then there would be a cost elsewhere, since apparently there would be no such thing as appraisal of intentions and desires, at least no language in which such appraisals could be expressed." This seems inconsistent with his view, just mentioned, that normative statements can express a moral code and direct and articulate motivation. The motivations expressed by normative statements will, no doubt, be of a more complex sort than simple desires or intentions. But all the same it is perfectly possible that the moral language could have the function of expressing appraisals of simple desires or intentions—appraisals which were a kind of more complex desires or intentions. That indeed might be a rather misleading way of putting the views of people like Gibbard or me. My second preparatory question is this. Brandt says that his is not a two-level but a onelevel theory (1979: 252; he says "two-tier", but I prefer my own term). But -74in examining his theory we have to look carefully to see whether it does not in fact have two levels very like mine. It works by asking what moral code a rational person would support for a society in which he was going to spend a lifetime (Brandt is here using his own sense of the word "rational" as defined earlier). So he clearly has two levels. The first consists of the moral code itself which the rational person is deciding whether to support. The second is the thinking he does when he is making this decision. In comparing Brandt's theory with my own, as we shall be doing, it will be clearest to ask, of each of Brandt's levels in turn, how closely it resembles my corresponding level (critical or intuitive). In the present academic climate in America, in which plagiarism, very widely interpreted, is thought to be a cardinal sin, I must explain that in raising this question I am not in the least interested in questions of priority. I have learned more from Brandt (especially from his earlier discussions of rule-utilitarianism) than he has from me. His Theory of the

Good and the Right was written after my "Ethical Theory and Utilitarianism" but before my Moral Thinking in which the ideas in it were more fully developed; and we have frequently discussed these matters with each other and exchanged typescripts, to my great profit. Some of these discussions have appeared in print (Hare 1988, 1989b; Brandt 1988c, 1989a). What I am interested to do now is to pinpoint the remaining differences between our theories, basing myself mainly on his 1979 book, and neglecting both his earlier writings, including Ethical Theory, and his later, in which his views do not change much. Brandt's discussion of the concept of a moral code (1979: ch. 9) is one of the best things in the book, and far more thorough than anything I have attempted. I could readily take over, as an account of what I have called the intuitive level of moral thinking, nearly all that he says about moral codes. To have a moral code is to be disposed to think and feel and act in certain ways. I stress all these aspects of moral codes in my account of the intuitive level in my Moral Thinking (which, as I said, appeared after Brandt's 1979 book). He would therefore be wrong if he now attributed to me the view that "a morally wrong act is one of a kind which informed prudential persons would proscribe for absolutely everyone" (1979: 196). This is much too simple. In 1952 when I wrote The Language of Morals, and had not yet developed a two-level view, I did indeed say something rather like the view he attributes to me, that "Your doing A is morally wrong" means "Your doing A in the circumstances is an instance of a principle which prohibits the doing at any time of anything like A in relevantly similar circumstances, to which I hereby subscribe" (Brandt 1979: 196; cf. Hare 1952: 191). But I was there setting up an artificial model, and talking about what I later came to call critical thinking. The two-level account which I now advocate distinguishes this from intuitive thinking. About the latter I would adopt something much more like Brandt's ac -75count of a moral code. Intuitive or prima facie principles are, I say, general, in the sense that they can have exceptions. They have to be general, because to make them specific enough to cover all possible cases would make them too complicated for practical use. So I could readily agree with him that to have a moral code is not to have a set of hard-andfast exceptionless principles. One has a moral code, rather, if one is generally disposed to commend and condemn certain broadly described actions. One does not stop having the code if one either breaks the principles out of weakness or badness of will, or makes exceptions to them because one genuinely thinks it right to make them in unusual cases. Moral codes and intuitive principles are to be judged by their acceptance-, not their observance-utility. Let us take it then, at least for the sake of argument, that Brandt is right in what he says it is to have a moral code. The question next arises, how we are to decide what moral code to support. This must be a task for another kind of moral thinking, which I call "critical".

Brandt says (and I agree) that our thought about this has to be rational; and I could, with qualifications to be made later, accept his sense of this word. What moral code, then, would it be rational to support for a society in which one was to spend a lifetime? Here there may be a difference between our views, arising from the accounts we give of the impartiality that all moral codes have to have. Rawls, as is well known, secures this impartiality by his "veil of ignorance", which conceals from the parties in his "original position" the information they would need in order to be partial. The Ideal Observer Theory secures it by stipulating (in its most acceptable form) that the ideal observer is impartially benevolent. My own theory secures it by insisting that moral principles have to be properly universal, and, further, that we have to think of other people affected by our actions as if they were ourselves. How does Brandt secure the same impartiality? His proposal for a substitute moral terminology would be implausible and unacceptable unless he could secure it somehow, because, whatever view we take about the argument from ordinary language, it is quite clear that ordinary morality requires impartiality in the selection of moral principles. It is true that sometimes intuitive morality requires us to be partial; but this is not true of moral thinking at the critical level (see Hare 1979 = 1989a: 321 ff., 1981: 135 ff.). To understand Brandt's answer to this question, the best place to look is his account of benevolence (1979: 138 ff.), and of self-regarding and other-regarding outcomes of behavior (1979: 204 ff.). It is obvious that if we were all, like the ideal observer, impartially benevolent, a moral system would not need to explain what to do with those who are not. But in the actual world they are the problem. Brandt gives an interesting account, based on the work of psychologists, of how most people come to be at least somewhat benevolent. He also shows that it is -76indeed to their advantage to be so, because it leads people to avoid actions which might provoke retaliation, and for other reasons. But the fact remains that many, even most, people are not as benevolent as morality, as generally understood, would require, and that some people are quite the reverse, especially those who have the power or the cleverness to get away with it (Brandt instances the operators of Nazi concentration camps, 1979: 145). There is also a problem about the extent of our benevolence: does it, or ought it to, extend to non-human animals or to future generations, or to the weak, none of whom can retaliate? Most of us think that we have at least some duties to all these classes of beings. Would a rational person, or would he not, support a moral code that imposed these duties? Here it becomes necessary to look again at Brandt's use of "rational". If the rational desire or action is that which survives maximal exposure to facts and logic, we have to ask "What facts and what logic?". Clearly we have to be allowed to bring in any facts whatever that might be thought relevant; but what about the logic? This brings out an

important point about rationality noticed in Hare 1981a: 214 f. Rationality is in the first instance a property of thinking, and indirectly of the results of that thinking in action or motivation. It is a property of thinking directed to the answering of questions, and the requirements for rationality may vary according to the question that is being answered. And what question is being answered is determined at least in part by the meanings and logical properties of the words that are being used in asking it. So, if the question that is being asked is "What is it best (i.e. to my greatest advantage) for me to want or do?" the kind of thinking that will be rational in answering it will be different from what it would be if the question were "What is it morally right for me to want or do?" If the rational person deciding what moral code to support is asking the first of these questions, his answer, if he thinks he can get away with it, may be that of Callicles in Plato's Gorgias. He may want there to be a moral code in his society which determines the behavior of other people, but want himself to be free to disregard it. So he will say that it is to his greatest advantage to encourage others to behave morally, but himself to behave immorally when it suits him. This would be a rational way of answering the first question; but would it count as supporting the moral code? On the other hand, if he is asking the second question, he cannot answer it rationally without paying attention to the logical properties of the expression "morally right". And what these are is a question that Brandt dismisses together with "ordinary language philosophy". If these logical properties are as I think they are, then the rational person will have, in answering the second question, to universalize his prescriptions over nonhuman animals and future generations, and over those too weak or too simple-minded to resist him. He can escape this requirement only by stopping asking that question and becoming an amoralist. -77The consequence is that there is a large class of people who are exposed to moral arguments on my view but not on Brandt's. We may grant that his method suffices for argument with those who are benevolent, at least to the degree that they are. And we may also grant that the true consistent amoralist, who will not ask any moral questions unless with his tongue in his cheek, can escape the arguments of both of us. But there remains a very large class in between (perhaps encompassing most of humankind), who do want to go on asking moral questions, but are in dispute about the answers to them. If such people ask what it is right for them to do in Brandt's proposed sense, they may, in answering them, disregard the interests of dumb animals, future generations and the weak, if the benevolence they have acquired does not stretch that far. For they may well not want to support a moral system which makes such demands on them. On the other hand, if they ask the question in my sense, which I claim is the sense it has in ordinary discourse, we have a further weapon. We can ask them, because of universalizability, whether they think it right that someone should act in the way that they are proposing if they were the victims in an identical situation. If they want to go on asking the moral question, this will restrain them from giving certain answers.

I claim that there are big advantages for society in general in having a moral code, in my sense of "moral"—which is, no doubt, why we try to get our children to acquire benevolence and other moral attitudes. But it is unclear whether this incentive would be so strong on Brandt's view. Might not parents who sought the interests of their children bring them up not to be more benevolent than conduced to those interests? What actually happens is perhaps best explained on lines suggested by Peter Singer (1981; cf. Hare 1981b). Our genes have selected themselves differentially, and have implanted in us some degree of kin-altruism and reciprocal altruism; and the learning processes described by Brandt have reinforced these kinds of altruism. But they have also implanted in us some tendency to rationality, at least in Brandt's "facts-and-logic" sense. However, this rationality has a momentum of its own: once we start thinking rationally, we are constrained by the logic, and this tends to extend itself. We have, for good reasons of our own advantage, invented the moral language; but its logic, though it could fall short of requiring full universalizability over all possible victims of our actions, tends to get extended so as to transcend this limitation. So we find ourselves, for example, asking whether it would be all right for someone else, however weak he (or she) now is, to treat us, were the situations reversed, in the way that we should like to treat him now. It is a historical fact that morality has developed and progressed in this way. We have to ask whether it would have done so if the moral words had the definitions that Brandt proposes for them. The answer will depend on the sense given to "rational" in his definition, and on whether, if it is taken in the "logic-and-facts" sense, the logic -78is taken to be as I say it is in ordinary discourse. So to settle the dispute between us we are driven back to ordinary discourse. Before I end, there are two further points I must mention. The first is one on which I am very much inclined to come over to Brandt's view. He distinguishes between happiness and desire versions of utilitarianism. I have already (Hare 1981a: 101 ff.) claimed that both these versions can be expressed in terms of preferences, with different restrictions on what preferences are to count. If we count only synchronic or now-for-now and thenfor-then preferences, and exclude from consideration asynchronic now-for-then preferences, and also count only preferences for experiences of the preferrer, and ignore all other preferences, then we get something like the happiness version favored by Brandt, understanding happiness as the fulfillment of our synchronic experiential preferences. It is interesting that Brandt, who in his earlier account of rational desires includes asynchronic and non-experiential desires, when he comes to the choosing of a moral code to support, rules these out. He does so for very good reasons. They would complicate our moral thinking intolerably, and even lead to paradoxical conclusions. When I wrote Hare 1981a I still hankered after giving all preferences a place in moral thinking, and avoided the complications by a somewhat ad hoc maneuver, which I called the "requirement of

prudence" (the requirement that we should consider only those preferences that the preferrer would have if prudent). This move brought me for practical purposes into line with Brandt's view, but it left me unsatisfied. In Hare 1989b I came even closer to Brandt, but remained uneasy. Recently Mane Hajdin, in a very important paper (1990), has suggested a way in which I might exclude non-experiential and asynchronic preferences from consideration; indeed he claims that my method of reasoning requires their exclusion, because only synchronic experiential preferences will make it work. I am very much inclined to accept this suggestion, and to use other means of reaching conclusions which I had tried to reach by bringing in the other sorts of preferences. But since several good philosophical friends have argued against this course, I am still undecided. If I did take it, I should be in agreement with Brandt on this point. The last problem I shall raise is about a difficulty in Brandt's formulation of his "reforming definition" of "right". Consider the following case. Suppose I have a friend who is a homosexual, and in the society in which I live homosexuals are put in prison. Would it be right for me to speak frankly to my acquaintances about his homosexuality? If I do, the word will get around, and he will end up in jail. But I would certainly, as a rational person, support a moral code in which people could speak frankly about other people's sexual orientation, if I were -79going to spend a lifetime under it. So on Brandt's definition it would be right for me to speak frankly. But most of us would think it wrong to do so in the actual circumstances of such a society. On my own theory too it would be wrong, because I can certainly prescribe universally that in these circumstances one should not speak frankly. I think that Brandt is addressing this difficulty on 1979: 297 f. His suggestion is that there would in an optimal moral code be different rules. One of these might say that it is right to speak frankly about such matters; another that it is right to protect one's friends from imprisonment on the basis of a law which should have been repealed long ago. These rules come into conflict in this case. But this solution demands some way of settling the conflict. With a two-level system we can bring in the critical level to adjudicate it. But a one-level system, such as Brandt's purports to be, has no way of doing this. It could have rules of priority for settling conflicts between the first-order rules; but, as I argue in Hare 1981a: 32 f., this move fails to meet the need to give different relative priorities to two rules in different cases. And further, such a set of priority rules is an embryonic two-level system, because the thinking which assigns the priorities is of a higher level than the thinking which merely applies the rules. The only other way of avoiding this difficulty that I can think of is to complicate the rules of the code until they are specific enough to deal with all cases. Thus in the present case

the code might contain a rule which said "It is right to be frank about people's sexual orientation, unless doing so will put them in prison; but in that case it is right to keep quiet." If we took this line, our code would soon get extremely complex. This might be all right for a set of principles elaborated in critical thinking, whose principles can be as specific as is needed to cover particular cases. But then the code would not serve the purposes that Brandt is seeking, which demand a degree of simplicity; nor would it serve the purposes of intuitive thinking in my own theory. The best solution is to separate the two levels, and use critical thinking to deal with unusual particular cases where conflicts arise, but keep the principles of the intuitive level general enough to be helpful and manageable. This is what my own theory does. I shall be interested to know how else Brandt proposes to deal with such awkward cases. -50-

6 Moral Concepts and Justified Feelings Allan Gibbard I'll begin on a more personal note than I normally would in writing a critical paper. Richard Brandt has had more direct influence on my thinking in ethical theory than any other philosopher. This was partly an accident of history: At seventeen, I went off to Swarthmore College knowing it was a good place, but with no idea there was anything special about its philosophy department—and not much idea there was anything special about philosophy. Swarthmore's seminars for upper-level students were always an intense experience, but nowhere more so than in philosophy: three or four hours, one day each week, of intense discussion at the professor's home, with half one's working time reading and writing in the subject. Brandt's forthrightness, his philosophical insight, his determination to push to the heart of the matter, the way he revelled in critical inquiry— whether the criticisms were of us, of our unseen authors, or of him— these virtues all impress me now as they did then. More to the point of this paper, a number of the central themes of Brandt's moral philosophy still seem to me both right and crucial. Morality must have a point, Brandt insists, if its demands are to be justified. We mustn't just cling to it because it's ours and we find it lovable. Morality centrally involves moral feelings or attitudes, and assessments of these as justified or not. Morality too is crucially a social institution, to be assessed for its benefits and harms. To think about morality from the beginning, we should think about why to be concerned with morality, and guide ourselves by answers to this question. Brandt, though, never encouraged discipleship; he valued independence of mind in his students. We were off track only if we didn't take seriously the task of clarifying our thoughts, working them out, and criticizing them. And I myself— though on dozens of points of detail as well as in broad themes I came to think as Brandt did—never thought he had everything right. The book I wrote decades later, Wise Choices, Apt Feelings (1990), pursued agenda very much set by Brandt. Brandt's own book A Theory of the Good and the Right (1979) had convinced me to start moral theory with questions about ra -81tionality: to get to the root of the subject, think about why it might be rational to care about moral questions. Then interpret moral questions in whatever way makes them most worth caring about. Brandt's book had delineated his project in just this form.

I departed from Brandt's own way of carrying out this project, though, at both crucial turns: his account of what the term 'rational' should mean, and his proposal for how moral questions should be treated as special questions of rationality. In this essay I'll chiefly discuss moral concepts—though my differences with Brandt on defining rationality will enter in. Both with rationality and with more narrowly moral concepts, large elements of my own views stem from Brandt. Indeed for each, I'll be claiming, we can read Brandt almost as formulating the theory that came to seem right to me, but then pulling away. Brandt's views have two sides: the rationale and the theory meant to implement it. Brandt's rationale I largely accept, but a theory other than Brandt's, I'll be arguing, offers the best way to realize major elements of the rationale.

I. Moral Concepts I'll choose for particular study a fairly recent development of Brandt's proposals (Brandt 1985 1. ). This gives us something like the latest published word on the theory—though as with any inventive and active philosopher, Brandt may always have changed some of his views. His broad strategy is to start with moral thought, talk, and motivation as they exist— perhaps not much influenced by philosophy—and ask whether we should want such practices and why. Our moral practices we'll find valuable on the whole, Brandt concludes, and find valuable many of their particular features. The reasons are the familiar ones stressed by Hobbes, Hume, and Mill: Moral constraints protect us and foster mutual benefit. Still, the conceptual structure of current moral practice is in lamentable shape. We philosophers—if we think clearly and are determined to cut to the heart of matters—can improve these practices and enhance their benefits, with philosophical clarity as a bonus. The best way to do this, Brandt argues, is conceptually reformist: propose new concepts to replace the old, muddled ones. The new concepts should capture the benefits that current moral practice gives us, and enhance them by letting us design our morality with clear heads and a clear sense of the purposes for which we are doing it. Begin, then, with moral practice as we find it. In Brandt's characterization, we can identify three chief components: ____________________ 1. Althhough I refer to this article as 'Brandt 1985', my page references to it will be to the 1992 reprint in Brandt's collection, Morality, Utilitarianism, and Rights. -821. 2. 3.

Personal complexes of feeling-tied motivation: Propensities to such feelings as admiration, guilt, and disapproval, and consequent desires and aversions. Belief in justification: Believing "about all these attitudes/dispositions that they are justified in some appropriate way" (Brandt 1985: 62). Moral discourse and symbolic thought: Here enter moral terms and concepts.

(i) and (2) Brandt presents together as a person's motivational moral code. 2. Compared with these motivational codes, Brandt says, moral discourse (my third component) "is relatively an epiphenomenon" (Brandt 1985: 61). Still, we do need it, not only to communicate approval and disapproval, but also to engage in far more refined kinds of moral discussion and reflection (Brandt 1985: 64). For these purposes we need such intricate concepts as "prima facie obligation, obligation overall everything considered, excuse, reprehensibility everything considered, the praiseworthy or supererogatory, perhaps also moral rights and 'offenses'" (Brandt 1985: 65). This offers work to philosophers, at various levels of depth, including a level where we try to go "to the bottom of matters" (Brandt 1985: 66)—as with most work in metaethics. At the bottom of matters are concepts that cry for reform: the concept of justification in component (2), and the more specifically moral concepts in the moral thought and discourse of component (3). Component (i), our personal systems of moral feelings and feeling-tied motivations, will presumably also need reform, but this won't be the definitional reform that is the work of a reforming metaethics. 3. (2) and (3) are where the conceptual work's to be done. Brandt's own proposals for conceptual reform proceed in two major stages. The first he expects to be fairly uncontroversial: Express moral concepts in terms of moral feelings' and motivations' being "justified". The term 'reprehensible', for instance, means warranting disapproval: an action is reprehensible if and only if one would be justified in disapproving of a person on account of performing ____________________ 2. Condensing slightly less, we can say that a person's having a motivational moral code consists in the following: (a) She is intrinsically averse to acts of certain kinds, and not on grounds of self‐ interest. (b) The strengths of these aversions form a resultant, much as with a vector sum. (c) She tends to guilt if she goes against these aversions and without "excuse". (d) She tends to disapprove of others who go against them without "excuse", (e) She experiences certain further kinds of admiration and disfavor. (f) She disapproves of the absence of these aversions, (g) She thinks all of the above justified. (Brandt 1985: 61-62). 3. With components (2) and (3), of course, there is also much of substance to be done, but if we want to get to the root of matters, conceptual reform is the place to begin. -83such an action. This boils questions in the moral discourse of (3) down to questions of justification that figure in (2). The second stage is then to reform and make precise the concept of justification in play in these questions. How shall we do this? We could say—still uncontroversially, perhaps—that justified disapproval is disapproval mandated by a "justified" moral code. This leaves us with the term 'justified' as applied to moral codes, and Brandt despairs of finding any precise ordinary meaning for this (Brandt 1985: 66). It is at this point the reforms must begin.

What should be our guiding rationale? Look for a reform, Brandt suggests, by asking, "What is the importance of knowing whether an effective moral code is justified? What is the function of raising such questions?" (Brandt 1985: 67). People ask themselves, "What kind of moral code am I to teach myself, subscribe to, prefer, support for the educational system?" (Brandt 1985: 67). They ask "whether their own moral codes are subject to well founded criticism, whether there are considerations that can give them assurance rather than self-doubt" (Brandt 1985: 67). The reform, then, should yield a conception that "will answer this practical doubt" (Brandt 1985: 67). Brandt's proposal goes (very roughly) as follows: Let a "justified" moral code mean one it is "rational" to support. Then in turn, stipulate that when I say it is "rational" for you to support a moral code, I mean this: I hereby recommend supporting it, taking on as my objectives and beliefs your own objectives and beliefs—but corrected in complex ways (which Brandt specifies). A chief part of the correction is what Brandt in A Theory of the Good and the Right called "cognitive psychotherapy". In this free condensation, I've glossed over many of Brandt's refinements, and also a number of puzzles of interpretation that I have. My aim has been only to say enough to set the stage for the particular line of commentary I want to pursue. The chief steps in Brandt's proposals are these: First, moral concepts are characterized in terms of moral sentiments' being "justified" or not. (This part I accept.) Second, justified sentiments are sentiments mandated by moral codes it is "rational" to support. Third, to call something rational is to recommend it on a set of grounds that are built into an elaborate reform definition of the term.

II. From Threat of Circularity to Expressivism Begin with a threat of circularity in Brandt's proposals. Brandt's reforms have been adopted, imagine. What, with the reforms, do moral opinions consist in? What is it, say, to believe that theft is "reprehensible"? Isn't it to believe it rational to support a social practice that includes—among other things—the very belief in question, that theft is reprehensible? If so, the term "reprehensible" has been defined circularly: in terms of belief that a kind of action is reprehensible. -84More carefully, we could put a variant of this argument for circularity as follows: Brandt's reforms have been adopted, we're imagining. We now say, i. Theft is reprehensible if and only if disapproval of theft is justified. (Def.) ii. Disapproval of theft is justified if and only if it is rational to support one's society's having a moral code that enjoins disapproval of theft. (Def.) iii. A society's having a moral code that enjoins disapproval of theft consists, in part, in people's thinking that disapproval of theft is justified. (Def.) We can now use (iii) to substitute one equivalent for another in (ii), yielding iv. Disapproval of theft is justified if and only if it is rational to support (along with certain other things) people's thinking that disapproval of theft is justified.

This is to define the term 'justified' in terms of itself.Does Brandt fall into such a trap? I'm not sure; my guess is that the answer would depend on tricky questions of interpretation that have no clearly correct answers. What I'll argue is this: First, Brandt can avoid such a circularity (as perhaps he does). Indeed I'll suggest two ways he might do this. Second, though, either way of avoiding the circularity carries a price. One leads us away from Brandt's own questions as the first ones in systematic moral inquiry. The other sacrifices a part of the crucial insights that motivated Brandt's proposals in the first place.How, then, can we avoid circularity? One way would be to distinguish two senses of the term 'justified' as it figures in Brandt's proposals. The term appears in two different senses, we can say: the sense Brandt officially defines, and another sense he treats as logically prior. Look again at (iv). Agreed, the term 'justified' in its first occurrence in (iv) has the meaning Brandt is out to define. At the end of the statement, though, should we really speak of people's "thinking that disapproval of theft is justified"? That leads to the circularity. Shouldn't we instead speak of their state as a special, unitary attitude—one we might call "regarding-as-justified"? This state regarding-as-justified, we could say, doesn't consist in a belief that the thing in question is "justified" in the sense that (iv) in part defines. (iv), then, should really be put as iv. Disapproval of theft is justified if and only if it is rational to support (along with certain other things) people's regarding-as-justified the disapproval of theft. -85This way out I myself would find quite appealing. I am an expressivist about justification: I think that there is indeed a unitary state regarding-as-justified—a state that doesn't consist in believing some proposition " V is justified". My own theory is that this state consists in accepting norms. 4. Should Brandt, then, accept the suggestion that the term 'justified' figures in his theory in two senses? Perhaps he should: After all, we could say, Brandt himself thinks there is a state "regarding-as-justified", prior to his reforms—since he thinks that people who haven't accepted those reforms still accept moral codes, and so regard certain feelings of disapproval as justified. Now if Brandt takes this line, he still differs from an expressivist with regard to justification. He thinks that there is not only the unitary state "regarding-as‐ justified" that the expressivist recognizes, but a distinct state, believing that disapproval, say, is justified —believing that the state of affairs characterized by (iv*) obtains. An expressivist, though, could argue that this second concept of justification is superfluous. We've already got the state regarding-as-justified, and so what need have we for the term "justified" in the sense Brandt so carefully defines? Isn't our real question what to regard-as-justified? To be sure, we might well decide to regard-as-justified whatever we think is "justified" in the sense of Brandt's definition. (Call this justified B .) Then Brandt's term 'justified B ' will indeed be of great interest—but its interest will be that we've already settled what to regard-as-justified. Why not, then, adopt the expressivist's own strategy? Why not, that is, characterize a concept justified E by saying: to believe that something is justified E is to regard-as-justified that thing? The question on which moral inquiry hangs, after all, is what feelings and the like are justified E —justified in the expressivist's sense. Only once

we settle in our minds that all and only justified B things are justified E will we be able to use the two concepts interchangeably. Probably, though, this way of avoiding circularity isn't the one Brandt will want to take. He thinks, true enough, that there is a state regarding-as-justified, a state that we can be in before we've adopted his reforms. He thinks, though, that it is a confused and unsatisfactory state. We're therefore best off, he'll think, substituting believing things to be justified B for this confused state of regarding-as‐ justified. Can he, then, reform away the confusion and avoid circularity? I think he can, but at a price. Consider again the three chief components of our pre-reform situation: (i) an array of feeling-tied motivations, (2) regarding these as justified, and (3) moral discourse and symbolic thought, with moral terms and concepts. The regarding-as-justified of (2), Brandt thinks, is a confused state, and the moral thought and discourse of (3) consists in using terms and concepts that have no clear meaning. Can't the reform, then, be this? Take the notions involved in (1) ____________________ 4. See especially Gibbard 1990. -86as clear enough to use, and then look for good substitutes for the vague notions of (2) and (3). As before, proceed by refining our vague notion of justification. Explain justification in terms of the rationality of supporting the feeling-tied moral motivations in component (i). Leave out of our reform definitions all talk of whether it is rational to support one or another version of the concept-laden components (2) and (3) of moral practice. To regiment the moral terms of (3), use a pattern we can illustrate by again choosing the term 'reprehensible'. X is reprehensible for his act A if and only if any justified array of feeling-tied motivations would include disapproving of those who do acts like A for doing such a thing. An array M of feeling-tied moral motivations is justified if and only if it is rational to support M as the moral motivations for one's society. And in saying it is rational to support M, I'm recommending supporting it, on the basis of my audience's corrected desires and beliefs. Along the way, this gives a sense to the term 'justified', and so we can now replace our confused, pre-reform regardings-as-justified with clear beliefs that desires, aversions, patterns of feeling, and the like are "justified" in this sharp sense. These definitions are non-circular, and the trick in making them so has been to avoid all question of whether it is rational to support components (2) and (3) of moral practice—to support regarding certain moral motivations as justified, and using certain moral concepts. What we assess as rational or not is support of various alternative arrays of feeling-tied moral motivation: component (1) of moral practice, but not components (2) and (3).

Going this way, though, constitutes a hefty sacrifice of Brandt's guiding insights. To get to the bottom of our moral questions, Brandt proposed, we should think about the benefits of morality as a social practice. That seems right, let me agree. Now shouldn't this mean the benefits of the whole complex of moral practice (1)-(3)? To avoid the threat of circularity, though, we've now renounced being guided by the benefits of any part of this complex except the inarticulate motivations of (i). Many of the benefits of morality come from an interplay of concepts and motivations, but the proposal now is to find sharp, pragmatically motivated concepts that take no heed of these benefits. Is there a way, then, to follow Brandt's guiding insight more fully and still avoid circularity? There is, I think, but it would mean shifting the status of Brandt's proposals. It would mean, in effect, moving in the direction the expressivist tried to force out of the earlier way of avoiding circularity. To get Brandt's proposals best to capture his insights, I'll be claiming, we need to misconstrue them as proposals not for new linguistic practice, but for substantive ways of thinking. These ways of thinking will be couched in expressivistic vocabulary. To see this, let's shift focus. What, as Brandt would say, is the force of concluding that something is justifiedg? What is the force of finding that -87disapproval, say, is "justified" in the sense of the reform? It is to recommend support of an array of moral motivations that includes this disapproval—or perhaps to accept such a recommendation. What is recommended is that one set out to support such a motivational array. What "support" includes Brandt leaves rather vague, but it involves urging it on others and urging it on oneself. Now we could imagine my urging certain motivations because their being widespread serves certain of my purposes, but not finding that these advantages elicit the motivations in myself. Suppose I thought, for instance, that hating one's economic competitors spurs high effort and productivity, so that it would be best if I hated them—even though my competitors haven't done anything but work in ways that I myself hope to work once I rouse my hatred. Urging hatred on myself, on these grounds, might not make me hate anyone, for I might just not find anything to hate them for—useful though I think such hatred would be. Indeed there's a sense in which I wouldn't even regard such hatred as justified, if all that recommends it is its spurring effects, and not some iniquity on the part of those I'm urging myself to hate. Brandt is quite right, I'll agree, that before his reforms catch on, there isn't any one, clear factual opinion that goes with regarding disapproval, say, as justified. But he's also right, I would claim, that regarding one's disapprovals and the like as justified is part of the complex of psycho-social phenomena that constitute the workings of a moral system. Isn't that because regarding-as-justified is clear enough psychologically, even if it isn't a state of having a clear opinion about some clear matter of fact? I argued in Wise Choices, Apt Feelings for the existence of such a clear psychic state, and such a state seems tailor-

made to figure in Brandt's description of the social psychology of moral systems. It is the state, on my own view, of accepting norms that say to disapprove. Suppose, then, there is such a state—whether or not my own account of it is right. Then isn't the force of finding that disapproval is justified B —justified in Brandt's reformed sense—that it might well engage this state? Call this state regarding-as-justified E (choosing the subscript 'E', as before, because an expressivist will think it the basic state in terms of which the concept of justification should be explained). Won't Brandt's concept justified B have its intended force precisely if, when I conclude that a disapproval is justified B , I come on that ground to regard-as‐ justified E this disapproval? The base question of moral inquiry, then, is what to regard-as-justified E . Being justified B gets its force for me if I take what things to regard-as-justified E to depend on whether they are justified B . Once we have made such a determination, a concept like Brandt's justified B can indeed be useful. An expressivist, after all, doesn't have to deny the usefulness of concepts with much descriptive content packed in; his claim is that these are not the only the concepts we need. Fact-laden concepts are to be welcomed, so long as we don't expect them to crowd out the more basic justificatory concepts we need. -88The better way for Brandt to define a fact-laden 'justified', it should now be clear, is not the one I proposed in my second attempt to avoid circularity, but in my first. The second attempt ignored the pragmatic advantages of finding attitudes justified or not, and of having one or another system of moral terms and concepts. If by now we've agreed that an expressivistic concept of justification is needed—that we have to decide what sorts of things to regard-as-justified E before we know whether a concept like justified B has a force that will make it important—we can go back to the parallel-concepts schema that I dismissed earlier as playing too much into the hands of the expressivist. We've now given the expressivist what he wants. And Brandt too can have much of what he wants: We can, if we wish, let our moral conclusions rest on conclusions as to what kind of moral practice it is rational to support. We can ask which kinds of moral practices it is rational to support in their entirety, which version of an entire complex (1)-(3) it is rational to support—where we interpret components (2) and (3) using an expressivistic concept of justification. Our question in the end is which moral sentiments to regard-asjustified E , but— if we find it wise—we can let our answer hinge on the answer to the different question of which moral sentiments are justified B .

IV. Departures from the Official Theory Both here an in my book, I've tried to follow much of Brandt's guiding rationale, while departing from Brandt in the proposals I make to implement the rationale. Our rationales are much alike, whereas our theories differ. In particular, I depart from Brandt on three chief matters: (a) The motivational-affective system that figures directly in morality. I speak of guilt and impartial resentment; Brandt speaks of a broader range of feelings and motivations. (b) How rationality is tied to morality. For Brandt, moral questions boil

down to questions of which moral code it would be rational to support. I take moral questions rather to be questions of how it is rational to feel—which moral sentiments it is rational to have. (c) What the term 'rational' is to mean. In this concluding commentary, I'll explain these three departures and offer some justification for them.

(a) The Motivational-Affective System Like Brandt, I think that morality is tied in a special way to a certain range of feelings and feeling-tied motivations. We both think that moral concepts are somehow to be explained in terms of such array of feelings and motivations—the feelings and motivations that figure, as Brandt puts it, in a person's "motivational moral code". Brandt, though, appears more liberal than I in the feelings and motivations he will admit to the set. We both include guilt, and we would both, I take it, include impartial resentment or feelings of indignation. Brandt, -89though, speaks more broadly of "disapproval", and of kinds of "disfavor" or "contempt" that fall short of indignation (Brandt 1985: 61-2). Now to be sure, if I'm allowed a suitably narrow reading of these terms then I won't object. I'd be glad to allow feelings that are too weak to count as feelings of indignation, but which apart from strength are the same kinds of feelings. The terms Brandt uses, though, threaten to stretch farther, and to encompass feelings we won't regard as moral. I claim, then, that my own terms better specify the range of motivations we need to identify. I'm speaking so far of negative moral feelings. Brandt also includes positive feelings of "admiration". I too am glad to include positive moral feelings, but the term 'admiration' strikes me as too broad. I'd speak instead of feelings of moral approbation, and of moral self-approbation. To see the threat of including too much, think, say, of failure to dress with pizzazz. Fashion-conscious people will regard this as a fault, but not—with most such people—as a moral fault. Do their feelings count as moral on Brandt's account? I'm not sure. These people are averse to failing themselves to dress with pizzazz, and in a sense many will disapprove of such failure in others. They'll hold drab people in disfavor or contempt— and they'll admire others who dress with special pizzazz. Has Brandt, then, defined a "motivational moral code" so that typical fashion-conscious attitudes toward pizzazz and drabness are excluded—as presumably they should be? Perhaps Brandt rules these out when he specifies that moral aversions must be "for no further reason and in particular no reason of self-interest" (Brandt 1985: 62). Is dressing with pizzazz a matter of self‐ interest? I myself don't find the answer clear. Aren't the dandy's aversions to drabness much like an architect's aversion to designing cloying buildings, even if a cloying style is the one that will bring in commissions? The feelings in both cases are tied to ideals of admirable or despicable style in one's way of living—

and our concept of self-interest may not be clear enough to classify these attitudes as selfinterested or not. 5. My own proposal was that narrowly moral judgments are tied to specific narrowly moral sentiments: guilt, impartial resentment or anger, and moral approbation and selfapprobation. Doesn't this get things right for the case of dress? Some fashion-conscious people do seem to treat failure to dress with pizzazz as a moral fault. In saying this, we wouldn't mean they suffer from a simple conceptual confusion about what sort of thing could be a "moral" fault. Rather, we would mean something like this: such people respond to drabness with resent____________________ 5. Mark Overvold (in his 1980 paper) proposes, roughly, that self-interested motivations are desires for states of affairs that entail one's own existence. I think this includes too much, since it indudes a special concern with one's own moral purity—a concern that will be tied to guilt and moral self-approbation, and which we can read Kant as thinking the pure and basic moral motivation. -90ment and indignation, and they would feel guilty about it if they themselves were to fail to dress without pizzazz. (I once knew an Australian who evinced indignation when Americans wore sandals with socks. She seemed to treat it as a moral offense.) Listing these feelings as moral seems, in these cases, to get the classifications right. 6.

(b) Rationality and Morality Is the basic moral question what sort of moral code it is rational to support, or how it is rational to feel about the various kinds of things people do or might do? These two kinds of questions are related, to be sure: Having a "motivational moral code", in Brandt's conception, is chiefly a matter of feeling-tied motivations. Indeed to a point we can read Brandt and me as taking—by no coincidence—the very same position on how moral questions reduce to questions of which moral feelings are justified Brandt, though, explains a feeling's being "justified" or not in terms of which moral code it is "rational" to support. I treat "justified" and "rational" as amounting to the same thing. Why doesn't Brandt do so too, and so speak directly of which moral feelings are "rational"? Brandt and I, to be sure, mean different things by 'rational'; that's the difference I'll touch on under (c) below. And Brandt's official definition of'rational' has the term applying to acts but not to feelings. This can't be the reason, though, for Brandt's proceeding so indirectly, it would, after all, be easy to give a Brandtian-style definition of 'rational' as applied to feelings. A rational feeling, Brandt could say, is a feeling that would survive cognitive psychotherapy. Then he could mean by a "justified" feeling—justified disapproval, say—a feeling that is rational in this sense. This is far more direct than saying that a justified feeling is one such that it is rational to support a code the having of which includes having that feeling.

Directness aside, which way works better? Suppose we've given the term 'rational' a sense, either mine or the one I proposed for Brandt—a sense that allows feelings as well as actions to be "rational" or not. What should moral questions boil down to? (i) Its being rational to feel certain ways toward actions? Or (ii) its being rational to support a moral code the having of which consists, in part, in feeling certain ways toward actions? ____________________ 6. This, of course, raises many issues that don't divide me from Brandt, some of which I have discussed elsewhere. Specific emotions, many philosophers and psychologists maintain, can't be individuated except by the kinds of beliefs or judgments that go with them. Moral feelings include moral judgments, and so moral judgments can't be explained in terms of moral feelings. I argue against this view in my book (1990: 129147); see also Gibbard 1993. A widespread alternative to Brandt's and my ways of doing things is to delineate moral judgments by their subject-matter. I don't think this will work; for an extended argument, see Gibbard 1992m. -91I've argued that it's (i) and not (ii)—though I certainly want to allow that we might end up letting our answer to (i) rest on our answer to (ii). We might, that is to say, take how it is rational to feel to depend on which feelings are directed by a code it is rational to support. Moral questions, though, hinge on (ii) only indirectly, by hinging on (i). And this hinging is not a matter of (i)'s being defined in terms of (ii). So I've proposed. How, then, might we test Brandt's and my proposals? The test is to separate (i) from (ii) in thought-experiment, and see where moral judgment should follow. Here is a test case: Raskolnikov, imagine, supports a moral code that condemns murder. (Define "murder" here not as wrongful killing, but as picked out in fairly ordinary ways.) Indeed he supports a code that condemns murder, even when the agent finds he can produce more happiness by murdering someone than any other way. Having a moral code that permitted murder in order to increase total happiness, he thinks, would lead to mistakes or abuses by unrigorous thinkers swayed by temptations. He himself, though, is a thinker of high rigor, he's convinced, and since murdering the joyless widow for her gold would increase total happiness, he sees no reason himself to feel guilty if he does so. He sees no grounds for guilt-tied aversion to the act. Does he think such a murder wrong? In effect, he doesn't buy into the moral code he thinks it rational to support: he doesn't think it rational himself to have the guilt-feelings and the guilt‐ tied aversions that he thinks it rational to support in a moral code for his society. Does he think that the murder he's contemplating would be morally reprehensible? I would think not: He thinks guilt over such a murder would be irrational—and he might think that resentment too would be irrational on the part of anyone who really understood the case. More to Brandt's point, though, would our purposes be best served by a sense in which Raskolnikov does think such a murder to be morally reprehensible—though he sees no reason to be morally appalled in any way if he does it—or a sense in which to think what Raskolnikov thinks is not to think such a murder reprehensible if carried out

by a rigorous thinker, however rational it might be to promote a public moral code that condemns such murders? Now to be sure, for right-thinking people, it may well be in fact that the two senses would coincide. We haven't, after all, established that Raskolnikov is a right-thinking person. And to be sure, we do (and perhaps should want to) allow some conceptual room between thinking an act morally wrong and feeling morally appalled. I can think that my feelings are defective in that they aren't "tracking" the moral rights and wrongs of the matter. But my own view allows for this: I can think it rational to be morally appalled and irrational not to be, and yet find I don't actually feel the way—as I judge—it is rational to feel. Raskolnikov, as I've said, doesn't "buy into" the moral code he thinks it rational to support. That is to say, he doesn't think it rational to have the feelings that constitute having the code as one's own. If he supports the code, to be sure, he -92does advocate buying into it; that's part of what it is to support the code. ("Support" won't consist just in saying that the code is the one it is rational to support.) But Raskolnikov doesn't do what he advocates: He doesn't himself buy into the moral code he thinks it rational to support. Brandt presumably does, and so Raskolnikov and not Brandt provides us a test between the two definitions. Does it suit our purposes best to regard Raskolnikov as making moral judgments that he doesn't use as a guide to action or to feelings about action? Don't we do best to regard Raskolnikov as insincere when he tells others that murders—even of the kind he is secretly contemplating—are "morally reprehensible"? Indeed if we begin to use the term 'morally reprehensible' in Brandt's sense, then we won't, by calling such murders "morally reprehensible", be supporting a moral code that condemns such murders. For to support a moral code is to advocate buying into it, whereas to call such murders "morally reprehensible", in Brandt's sense, is to advocate supporting the moral code, not buying into it. Raskolnikov can sincerely say, in Brandt's sense, that murders are "morally reprehensible", but in saying this he isn't providing guidance on whether to murder or how to feel about murder. One who understood Raskolnikov's view of matters and accepted it would believe what Raskolnikov was saying, but not think that any feeling-tied aversion to such murders was rational. Should we, then, reform our moral terms so that calling such murders "morally reprehensible" doesn't constitute supporting a moral code that condemns such murders? Should we use the term "morally reprehensible" so that one can understand and accept the claim that such murders are morally reprehensible, and yet not draw from this any guidance for one's feeling-tied aversions? The reform seems excessively to detach moral language from moral practice, in a way that should not please Brandt.

(c) The Term 'Rational', as Applied to Acts and Feelings I'll speak mostly of acts, since it is for acts that Brandt officially defines the term 'rational'. What I say, though, would apply as well to a Brandt-style definition of 'rational feeling' that we could easily invent. In my book, I briefly took up the definition of the term 'rational' that Brandt had proposed as a reform in his 1979 book, The Good and the Right. I complained, in effect, that Brandt's definition of 'rational' hadn't sufficiently secured the recommending force of the term. In the 1985 article, Brandt switches ground in a way that precludes any such complaint. Whereas in 1979, 'rational' had received an entirely factual meaning, by 1985, it has a preformative element: The person who calls something rational is saying "I hereby recommend" on certain kinds of grounds. Thus, the recommending force of the term is now analytic on its face. -93To this analytically recommending force of the term, Brandt joins an intricate factual content. Roughly, if I say it is "rational for X to do D", I am recommending the act D to X on the basis of X's own corrected desires and justified beliefs. 7. Corrected desires "form a transitive system", are "mood independent, and capable of surviving repeated vivid reflection on relevant facts" (Brandt 1985: 69). Should we build this "vivid reflection" test into a definition of 'rational'? In my book, I imagined someone for whom vivid realization of what people's innards are like would produce an intrinsic aversion to eating with other people. If this person enjoys the fellowship of eating in company—because he manages to avoid dwelling on what their innards are like—the fact that full and vivid picturing of their innards would arouse such an aversion doesn't seem to speak against eating in company. It seems to speak against getting full information. Does this objection survive, now that Brandt has built recommending force analytically into the term 'rational'? In Brandt's new sense, I can't call it "rational" for my queasy person to eat alone—because I don't recommend his doing so. Still, neither can I call it rational for him to eat in company. For to do so, I would have to recommend dining in company as a means to forward his "corrected" desires—the desires that would survive repeated vivid reflection on relevant facts. And those include strong desires to avoid eating in company. Now to be sure, it's his desires somehow corrected that matter. It matters how his desires would respond to "repeated vivid reflection on relevant facts", if this means how they would respond if they responded rationally. Informed rational desires have an authority that ignorant desires lack, however rational the ignorant desires may be as a way of coping with ignorance. Desires are best founded when they constitute rational responses to full information.

This makes no sense, though, unless 'rational' means something different from what Brandt proposes meaning by the term. How might we follow Brandt's guiding rationale, and allow ourselves to say things like this? Brandt looks for a meaning that will have the right kind of recommending force. Why not simply give the term 'rational' this recommending force? How might we do this? What is it to recommend? To track down an answer, ask what a person accepts who accepts the recommendation. Can't we say, in my terminology, that the person ____________________ 7. This raises a puzzle for how moral terms are to be defined. In saying it is "rational" for agent A to perform an act, one recommends the act to the person X, on the basis of X's own corrected beliefs and desires. Brandt's definitions of moral terms, though, speak simply of its being "rational to support" a moral code, without saying rational for whom. It isn't dear, then, whether in calling something wrong, I'm recommending support of a moral code just to myself, on the basis of my own corrected beliefs and desires, or to everyone, to each on the basis of that person's corrected beliefs and desires. On the latter interpretation, another puzzle will arise: I am to base my recommendations not on my own justified beliefs, but on theirs. Suppose, though, I think their beliefs false, though justified given their misleading evidence. Why base my recommendations on beliefs that I myself think false? -94who accepts a recommendation comes to accept norms that say to do the thing recommended? That fits as expressivistic view of the term 'rational': that to think something rational is to accept norms that say to do it, and to call something rational is to express this acceptance of norms. 8. Which way of defining 'rational' is to be preferred is a large and difficult question, and my remarks here only point to an initial rationale for diverging from Brandt on the question. For the most part, I've been discussing not how to define 'rational', but how to characterize morality given a meaning for the term 'rational'. With Brandt, I accept that moral concepts invoke a motivational‐ affective system. My list of specifically moral attitudes, though, is narrower than Brandt's, and I've been arguing that Brandt's way may let too much in. With Brandt too, I accept that moral concepts are about which moral attitudes are justified. I've differed from Brandt, though, on what this involves. If an attitude's being justified is tied to the moral code it is rational to support, I've been arguing, we should not make the tie a matter of meaning. We should let 'justified' mean something much more direct, and then ask—as a substantive question— how an attitude's being justified in this sense depends whether it is rational to support a moral code that dictates the attitude. ____________________ 8. This formulation is a first approximation, which would need many qualifications. See Gibbard 1992r for an attempt at more refinement in my proposal.

-95-

7 The Evolution of Utility: A Philosophical Journey L. W. Sumner Utilitarianism is both a theory of the good and a theory of the right. As the former, it tells us which states of affairs have intrinsic ethical value. As the latter, it tells us to take the maximization of this value as our ultimate goal, and also connects the pursuit of this goal to our individual obligations and responsibility of utilitarianism is the one which combines the best theory of the good with the best theory of the right. However, there are many different possibilities in both cases, and there is little agreement within the utilitarian community on which to choose. Over his lengthy career, Richard Brandt has made many influential contributions to utilitarian debates concerning both the good and the right. He is probably best known for his defense of a particular form of rule utilitarianism. I think it is fair to say that Brandt was one of the first philosophers in this century to begin to appreciate the resources available to utilitarians for building a sophisticated and plausible theory of the right. However, because this aspect of his work has been so widely discussed, I will say nothing further about it here. Instead, I will focus on Brandt's companion work on the theory of the good. If we ask what value utilitarians aim to maximize in their theory of the right, the short (and unhelpful) answer is: utility. The crucial question then becomes: What is utility? This is an issue with which Brandt has been grappling, more or less continuously, for over thirty years now. However, the task of critically assessing his views on the issue is somewhat complicated by the fact that they have undergone considerable revision and development over that period of time. I do not say this in a reproachful spirit; au contraire, I find Brandt's willingness to reexamine his commitments in the light of new lines of thought both admirable and endearing. In this respect, as in so many others, he has provided the rest of us with a role model we would do well to emulate. I also have a further reason -97for welcoming the evolution of his philosophical thinking about the nature of utility: in all important respects, Brandt's later (and, I assume, current) ideas on this question seem to me much closer to the best utilitarian theory of the good than those of his younger days. Or so I will be arguing. To get some sense of the distance he has traveled, let us return to Brandt's starting point. His first extended treatment of the issue is to be found in Ethical Theory, which he

published in 1959. Here Brandt (1959: 304) draws a distinction between two classes of putative intrinsic values: personal goods are those which can belong to individual conscious beings, while impersonal goods are those which cannot. The former class includes such items as pleasure or happiness, knowledge, and virtue or excellence; among the latter what Brandt has chiefly in mind is a favored distribution of personal goods across individuals (such as equal shares), but he might also have included states of affairs (such as the stability or diversity of the ecosystem) which have nothing directly to do with individuals at all. Brandt goes on to suggest that personal goods may be thought of as so many ingredients or aspects of individual well-being. Utilitarians have traditionally embraced welfarism—the thesis that only welfare (or well-being) has intrinsic value. However, they have diverged considerably in their views on the nature of welfare, some identifying it with pleasure or happiness, while others have favored a more pluralistic account which includes non-hedonistic personal goods. Because of the historical association of utility with welfare, Brandt is reluctant to count a consequentialist theory as utilitarian in the strict sense if it includes any impersonal goods in its theory of value (1959: 355, 404-5, 429-30). In the end, however, his own (tentative) preference in that book was for just such an 'extended' theory which, in addition to incorporating a pluralistic account of welfare, treats equality of welfare as an intrinsic value. In Ethical Theory, therefore, Brandt rejected both welfarism as a theory of the good and hedonism as a theory of welfare. By the time he published A Theory of the Good and the Right twenty years later, he had come round to the opposite view on both counts. In the Preface of this work, which remains the most complete and connected presentation of his ethical theory, Brandt candidly advises the reader that "the positive views in the present book are substantially different from those of the earlier book at almost every point" (1979: vi). The most radical changes have occurred in Brandt's theory of the good, where the intervening years have inclined him increasingly away from his earlier pluralism toward the relative simplicity of classical utilitarianism. In what follows, I will review some of these changes in detail, and also suggest that we should regard them as improvements. However, I first want to venture an interpretive proposal. It seems to me that Brandt was motivated to rethink his theory of the good at least partly by virtue of coming to take welfare more seriously as the central value of an ethical theory. The concept of welfare plays only a marginal, and somewhat ambiguous, role in Ethical Theory. It has no entry of its own in the index and is -98introduced in the text largely as a convenient label for the category of personal goods: "We might say that a personalist about values is one who counts only states of welfare as having intrinsic worth." Even its value as a label is then immediately called into question; the above sentence continues "but 'welfare' is itself a rather vague term, although one widely used" (Brandt 1959: 304). As Brandt continues to assess alternative ethical theories, he finds the label useful (the chapter on utilitarianism is entitled "Moral Obligation and General Welfare"), but one is still left with the impression that the concept of welfare is doing no real work for him. At this stage in the evolution of his

thought, Brandt seems to have little interest in welfare as such. Instead of beginning by affirming the intrinsic value of well-being and then going on to determine what it consists in, which is the traditional utilitarian order of argument, Brandt first asks which states of affairs are worthwhile in themselves and then finds it useful to call (some of) them ingredients or constituents of welfare. Although I cannot be certain of this, I surmise that Brandt's conceptual focus began gradually to shift after the publication of Ethical Theory, with welfare gradually coming to play a more prominent justificatory role in his theory of the good. Certainly by 1966, when he published "The Concept of Welfare", he had become more interested in the question of how best to analyze this notion. Although the ostensible topic of that essay is the notion of social welfare as it occurs in welfare economics (the essay appeared in a volume entitled The Structure of Economic Science), Brandt seems to assume throughout his discussion that the concepts of harm and benefit are of pervasive normative concern. In any case, he sets out to clarify "the meaning of 'welfare' in our ... familiar discourse about the welfare of individuals" (Brandt 1966: 258). Still holding that this term is too vague for an analysis of its ordinary usage to be instructive, Brandt urges instead that "we assign some definite meaning to [it], doubtless within the rough and vague limits prescribed by present usage, and that we do so for definite, statable, and relevant reasons." (1966: 260; cf. p. 270) In considering the various meanings which might be assigned to the term, Brandt introduces the two alternatives which he later came to call the happiness theory and the desire theory. We will return below to the continuing debate which has engaged him ever since concerning the relative merits of these two candidates. What is important for the moment is that they are different analyses of what welfare is—different theories about its nature. The question of which provides the better analysis is not worth pondering for twenty-five years except against the assumption that welfare matters a great deal for ethics. From that point on, Brandt's order of argument seems to be: welfare is important, and therefore whatever it consists in (whether happiness or desire-satisfaction) is important as well. Since he shows no sign of assigning similar value to any other good, he has in effect become a welfarist. -99In Ethical Theory Brandt departed from welfarism (as he understood it) by assigning intrinsic value to an impersonal good, namely, equality of welfare. This is, however, one of the many views he came to repudiate in A Theory of the Good and the Right: "I no longer affirm that an equal distribution of welfare is itself a basic value." (1979: vi) Instead, he accepts the utilitarian goal of maximizing the sum total of welfare and defends a distributive principle for income (not welfare) on the ground that it will be welfare-maximizing (1979: ch. 16). This change of heart effectively relocates the distributive issue from the theory of the good to the theory of the right. This, it seems to me, is where it belonged all along, quite independently of whether one chooses to follow Brandt in subordinating the distribution of welfare to its aggregation. The construction of any form of consequentialist ethics is best conceived as unfolding in a sequence of stages.

1.

At the first stage, one selects a menu of intrinsic values: states of affairs worth pursuing or promoting for their own sake, no matter where or when they occur. Because consequentialists affirm the priority of the good, these goods must not presuppose any antecedent principles of the right. All of the personal goods which Brandt considered in Ethical Theory (pleasure or happiness, knowledge, virtue or excellence) are candidates at this point, as are some impersonal goods (such as the stability or diversity of the ecosystem). The menu of values on which a particular version of consequentialism is built constitutes its theory of the good. Nonconsequentialists can share such a theory of the good. The distinctively consequentialist move occurs at the second stage, at which some operation is specified for combining these particular goods into a synoptic or global goal. The most familiar operation, because it has been favored by utilitarians, is arithmetic addition or aggregation. But consequentialists can also choose operations which are distributionsensitive: an equal distribution of intrinsic goods across individuals, or a pattern which focuses exclusively on the minimum share, or one which brings together aggregative and distributive factors, or whatever. Once the goal has been defined, the third stage tells us what to do with or about it: whether, for instance, it is to be maximized. Maximization is the most familiar option at this stage, again because it has been promoted by utilitarians. But other choices, such as satisficing, are available. If we bring this grid to bear on Brandt's theoretical commitments, then his early assignment of intrinsic value to equality of welfare seems misplaced. An equal distribution of welfare cannot simply be added to welfare as a further intrinsic good. Instead, because it clearly presupposes some antecedent standards of fairness or justice, it belongs at the second stage, which begins the construction of a theory of the right. It is at this stage that Brandt has changed his views, abandoning his earlier insistence that equality be added as an independent constraint on aggregation in favor of the straight utilitarian line. Since these are ____________________ 1. See Sumner 1987, § 6.1. -100alternative theories of the right, rather than the good, I shall not consider their relative merits here. I simply want to point out that there is a perfectly good sense in which, malgré lui, Brandt has been a good welfarist all along. The only goods which have truly been basic in his ethical theory are those he has regarded as so many ingredients or constituents of individual welfare. If Brandt has been a consistent welfarist, he has not been consistent in his views about the nature of welfare. Here is where the real changes have occurred in his theory of the good, changes which have moved him ever closer to the classical forms of utilitarianism. One important step was the abandonment of his early pluralism about welfare. In Ethical Theory he was prepared to include knowledge or virtue along with pleasure or happiness

as ingredients of individual well-being. But his view of this matter had also changed by the time he published A Theory of the Good and the Right: Various philosophers have thought that some things, different from happiness and possibly not desired by anyone or everyone, are worthwhile in themselves and worthy of being produced for no further reason, for instance: knowledge and virtue. This view, however, seems to be obsolescent, and I propose to ignore it. (1979: 246) From this point on, Brandt consistently disregarded what have come to be called objective theories of welfare, focusing instead on what he took to be the legitimate contenders: the happiness and desire theories. 2. Since he regards objective theories as obsolescent, he has little to say about their deficiencies as accounts of the nature of welfare. I think that this is unfortunate. Objective theories are worth dwelling on a bit, not because they have any chance of success, for they do not, but because the reasons for their failure are instructive. At this stage it will help to lay out some conceptual apparatus which seems to be presupposed in many of Brandt's discussions, but which he never makes explicit (as far as I am aware). In the first place, we need to distinguish clearly between a theory about the nature of welfare and a list of its sources or ingredients. A theory of welfare tells us what welfare consists in—what it is for something to make a person's life go well. Such a theory does not provide a list of the things which (invariably or normally) make our lives go well, but it does specify the conditions which anything must satisfy in order to make it onto such a list. There is a useful analogy here with metaphysical theories about the nature of causation (or, indeed, anything else of metaphysícal interest). A theory of causation tells us what causation consists in—what it is for one thing to cause another. It does not tell us which things cause which other things; that is the task of the sciences. The nature of causation is one thing, particular causes and effects an____________________ 2. See Brandt 1983c: 87; 1989a: 33. -101other. Likewise, the nature of welfare is one thing, its particular sources or ingredients another. The accounts of welfare with which Brandt is principally concerned are genuine theories in this sense; they are not just lists of beneficial states of affairs. What they have in common is that they are all subjective. A theory is subjective if it makes an individual's welfare depend, in whole or in part, on her own attitudes or concerns. A theory which takes welfare to consist in the satisfaction of desires or preferences is obviously subjective in this sense, but so is a more hedonistic account which equates it with enjoyment or happiness. In either case, the account of what it is for someone's life to be going well (or badly) makes essential reference to that person's (positive or negative)

attitudes toward the conditions of her life. A theory is objective, on the other hand, if it takes an individual's well-being to be entirely independent of such attitudes on her part. With these conceptual resources in hand we can now identify two problems for any objective theory of welfare. In the first place, such 'theories' often consist merely of a list of states or conditions (knowledge, virtue, etc.) which are claimed to make a person's life go better, independently of the extent to which the person seeks them or finds them rewarding. Brandt is right to reject any such list on the ground that it provides no "general systematic answer" to the question of what welfare consists in (Brandt 1989a: 33). An objective list is not a failed theory of welfare; it is no theory at all. What we need from an objective theory is some account of how items get onto its list—an account which makes no reference to the subject's attitudes or concerns. An objective theory which tries to provide such an account faces a formidable obstacle which derives from the nature of welfare. To say that a life is going well is clearly to attribute some form of value to it; welfare is an evaluative notion. 3. But lives may exemplify many different modes or dimensions of value: moral, aesthetic, perfectionist, and so on. Welfare constitutes just one such dimension; to distinguish it from the others, we may call it prudential value. 4. What is unique to this mode of value is that it tracks how well a life is going for the person whose life it is. All other dimensions, by contrast, furnish assessments of a life from some standpoint external to its subject (the moral point of view, objective aesthetic standards, the subject's natural kind, or whatever). A successful theory about the nature of welfare must somehow preserve and explicate this relativity of prudential value to the subject's standpoint. It is easy to see how subjective theories might manage this, since they can appeal to some aspect or other of the subject's attitudes: her desires or preferences, what she finds enjoyable or ____________________ 3. Brandt was aware of the evaluative nature of welfare in his earliest discussions; see Brandt 1966: 260. 4. I borrow this useful phrase from Griffin 1986. -102rewarding, or whatever. But there seems to be no way for an objective theory to capture the subject-relativity of welfare. I conjecture that Brandt's dismissive treatment of objective theories of welfare stems from his realization that they face an impossible task. However, there are two quite different ways of regarding the items which typically appear on an objectivist's list of human goods: knowledge, virtue, achievement, and the like. One is to treat them as so many putative sources of well-being—states or conditions which make a person's life go (prudentially) better, regardless of the extent to which she cares about them or finds them rewarding. If this status is claimed for them then we can legitimately demand that the objectivist supply a theory of welfare capable of supporting the claim. It is this demand which seems impossible to meet. 5. However, the very same conditions can be advanced

as having intrinsic value even when they make no contribution whatever to the subject's well-being. After all, it can be said, a life full of knowledge or virtue or achievement may be a better life (better as a human life) without (necessarily) being better for its subject. In that case, these goods can claim a place alongside welfare on a list of intrinsic values. Construed in this way, what the objectivist has to offer is not a rival (non‐ subjective) theory of welfare but a rival (non-welfarist) theory of the good. This sometimes appears to be the way in which Brandt interprets the objectivist's position, 6. though I am not sure that he ever clearly distinguished between these two possible construals. They are worth distinguishing, it seems to me, if only because a non-welfarist theory of the good is a much more viable proposition than an objective theory of welfare. Like Brandt, I am still strongly inclined to reject it in favor of the welfarist alternative, but I cannot see that nonwelfarist theories of the good are susceptible to the same kind of decisive refutation as objective theories of welfare. 7. Having embraced welfarism as a theory of the good, and having rejected objective theories of welfare, Brandt has tended to focus his attention on the question of which subjective theory provides the better account of the nature of well‐ being. As he sees it, the only two plausible candidates are (what he has called) the desire theory and the happiness theory. Both, it must be remembered, are theories about the nature of welfare, not (extremely short) lists of its sources. Thus each theory should provide some account of what it is for a life to be going well for its subject. On the desire theory how well my life is going is a function of ____________________ 5. Of course, subjectivists about welfare can, and typically do, include most of the same items on their list of the principal ingredients of well-being. But they are able to explain why these items deserve to be on the list. 6. See the passage quoted above from Brandt 1979: 246 and 1983c: 87. 7. For Brandt's (relatively brief) treatments of this question, see Brandt 1983c: 89 and 1989a: 44‐ 45. I have defended welfarism against the rival claims of perfectionism in Sumner 1992. -103the extent to which it is satisfying my desires or preferences (or some subset thereof, such as those which are rational or informed). On the happiness theory it is a function of the extent to which I am finding my life satisfying or fulfilling or rewarding. Although there is a good deal of overlap between the two theories—because we normally experience the satisfaction of our desires as gratifying—the fit is not perfect. There is no logical guarantee that we will enjoy what we are seeking once we have attained it, and also no guarantee that we will not enjoy things which we have not sought. Thus there is a choice to be made between the two accounts. On this issue as well, Brandt has come to reject his own earlier views. In his 1966 paper "The Concept of Welfare" he defended a version of the desire theory: "I suggest, then,

that we are keeping reasonably close to the ordinary associations of 'welfare,' and are at the same time assigning a useful meaning to the term, if we decide to use 'X is on a higher level of welfare if p than if q' to mean 'X rationally prefers p to q'." 8. At the same time he was firm in rejecting any equation of welfare with happiness: "we should not want to say that a person's welfare is increased only if, and to the extent that, his happiness is increased." 9. Thirteen years later, in A Theory of the Good and the Right, Brandt argued that the desire theory is unintelligible and embraced its principal rival. Since the appearance of that book, he has remained a pretty staunch advocate of the happiness theory (although there have been some surprising developments, to which I will return later). In this respect as well, his later views have moved closer to those of the classical utilitarians. What led Brandt to change his mind? Let us pass over his earlier defense of the desire theory for now, and look at the reasons he later offered for rejecting it. The one which seems to have been most instrumental in the transition is the problem of desire change over time. This problem stems from the fact that every desire has two distinct time indices: the time at which the desire itself is held and the time at which the thing desired is to happen. At different times you may hold different, and incompatible, desires for the same period of your life. When you entered law school you planned a crusading career as a storefront lawyer; now that you have graduated you are more interested in working for a prestigious mainstream firm; later, you will wish that you had gone into policy analysis for the government. Choosing any one of these options would satisfy a desire you hold at some time for this period of your life. But which desire is it whose satisfaction would count as enhancing your well-being? When Brandt first came to deal with the problem of desire change over time, he regarded it as "the fundamental difficulty" (Brandt 1982: 179) for the desire ____________________ 8. Brandt 1966: 271; cf. 1972: 684-6. 9. Brandt 1966: 258; cf. 1967a: 414 "Happiness is not identical with welfare or wellbeing." -104theory, the lack of a principled solution to which rendered the theory unintelligible (Brandt 1979: 251). However, he later came not only to think the problem soluble, but to endorse a solution: "My suggestion is that, to be an intrinsic good, the desire for something must be just prior to the event of its satisfaction...." 10. In this he echoes James Griffin, who argues (while addressing a different issue) that "utility must, it seems, be tied ... to desires that are actual when satisfied." (Griffin 1986: 11) Brandt's suggestion is one which he himself earlier dismissed as arbitrary, 11. and this seems to be one occasion on which he got it right the first time around. If we once accept the basic premise of the desire theory—that how well some period of my life is going is a function of the extent to which it satisfies my desires—then there seems to be no principled reason for privileging desires which stand in a particular temporal relation to that period. Why should we think

that your priorities as a young lawyer dazzled by the prospect of wealth and power are a more reliable guide to your well-being than the ones which you later adopt in the light of much greater experience—or, indeed, the ones you embraced in your earlier, and more idealistic, days? The mere fact that a desire for a particular time of your life is operative at, or just before, that time seems to provide no assurance that it will be better informed, or more considered, than desires which are temporally more distant. Indeed, the rationale for favoring proximate desires seems to be drawn not from the desire theory at all, but from its rival. On the happiness theory, how well my life is going at a particular time, or during a particular period, is determined by how satisfying I find it at that time, or during that period. There is seemingly no role here either for earlier ambitions or for subsequent regrets. What desire theorists are concerned about, by contrast, is that desires be as informed and rational as possible. In that case, their best option when confronted with preferences for a period of your life which shift over the course of that life is to side with those which score highest according to these criteria. 12. Those might be the ones you hold during (or just prior to) that period, but then again they might not. It seems to me that desire change over time, considered as an isolated phenomenon, is less troublesome for the desire theory than Brandt took it to be. If the general project is sound of equating welfare with the satisfaction of (rational, informed) desires, then it should generate a solution to these temporal puzzles as a mere byproduct. However, I also think that Brandt was on the track of some important issues when he raised these problems about desire change—issues which he also recognized in other ways and which do constitute a serious challenge to that project. ____________________ 10. Brandt 1989a: 38. See, by contrast, Brandt 1991: 356-7. 11. Brandt 1979: 251; cf. 1982: 182. 12. This is Griffin's response to Brandt's original critique; see Griffin 1986: 16. -105To get to the heart of the matter, let us remind ourselves that a desire of yours for some state of the world is satisfied just in case that state occurs. 13. Nothing more is required on your part: you need not like the object of your desire once it has occurred or find it in any way rewarding; indeed, you need not be aware that it has occurred or even still exist when it occurs. In light of this, there is an obvious logical gap between the satisfaction of some desire on your part and any improvement—or indeed any perceptible change whatever—in your actual experience of the conditions of your life. It is one thing for your desires to be satisfied, quite another for you to be satisfied. It is this gap between desire‐ satisfaction, on the one hand, and personal satisfaction—or gratification or enjoyment or happiness—on the other, which is the root problem for the desire theory. It is connected to the problem of desire change, since both are rooted in the same deep feature of desires: the distinction between the time at which a desire is held and the time at which the desired state of affairs is to occur. It is a logical truth that these two temporal indices

cannot coincide: one can desire only what has not yet happened. Once the desired state has occurred then that desire is (necessarily) satisfied and therefore extinguished. (It can, of course, be succeeded by the wish that the state continue, or recur, but that is a desire with a different object.) I believe that it is this gap which Brandt was attempting to bridge in his specification that "the desire for something must be just prior to the event of its satisfaction." But it should be clear by now that narrowing the distance between the two temporal indices of a desire will do little by itself to solve the underlying structural problem. 14. The fact that you continue to want something right up until it happens—that your desire remains actual until satisfied—cannot guarantee that you will experience the desired state of affairs as fulfilling or rewarding (you may hate it when you finally encounter it). It cannot even guarantee that you will experience the state at all (you may never know that it has happened). If there is any solution to this problem then it must be sought in a different direction. In his various treatments of the desire theory, Brandt has shown that he is well aware of its basic structural weakness: "The problem with the desire-satisfaction theory arises from two facts: first, that occurrent desires at a time t are for something to occur (to have occurred) at some other time; and second, that desires change over time." (1979: 250) We have already seen how Brandt pursues the ____________________ 13. "A desire is 'fulfilled' in the sense in which a clause in a contract is fulfilled: namely, what was agreed (desired) comes about." (Griffin 1986: 14) 14. To be fair to Brandt, he never thought that it would. His temporal suggestion was part of a package of reforms designed to make the desire theory more palatable. I consider the other items in the package below. -106second of these features. But he is also alive to the difficulties caused by the first. For one thing, he is adamant on the issue of posthumous desires: "... a desire theory, with 'satisfaction' explained in this broad sense, may conflict with a virtually unanimous commonsense opinion that what happens after a person's death cannot be a benefit to him. I think the same is true for events of which he will never know." (1989a: 38; cf. p. 36) More broadly, he has argued that well‐ being cannot consist in (what he has called) pure desire-satisfaction, in which "it is not necessary for the person desiring [an event] even to know about [its] occurrence, much less to enjoy it." (1976: 452; cf. 1979: 148) Now my own view is that all pure desire-satisfaction theories are unsatisfactory as accounts of the agent's well-being, irrespective of whether the description of the wanted state of affairs entails that the agent is necessarily alive at the time the desired states obtain. I think it too simple that self-interest or benefit be defined in this general way, partly because it seems common-sense to hold that some event, however desired

beforehand, is not to one's benefit if it is boring or repugnant in the fact—a consequence which pure desire-satisfaction theories have to swallow.... (Brandt 1991: 356) 15. In other contexts, Brandt has reached the same result by considering the psychology of benevolence. After reviewing the various mechanisms by which benevolence is developed in individuals, Brandt concludes: Now there is no reason at all to think that a pure satisfaction of a desire— an event perhaps not even known about—as such elicits any unconditioned liking responses in the self. And hence, by our theory of sympathy/benevolence, there is no reason to think that the representation of the pure satisfaction of another's desire—perhaps totally unknown to him—will be the target of sympathetic/benevolent motivation. (Brandt 1976: 453; cf. 1979: 147-8) This takes us very nicely to the heart of the matter. If an unadulterated desire theory is inadequate then there are two ways to proceed: we can experiment with amending the theory so as to repair its faults or we can jettison it entirely and give the rival happiness theory a chance. Although Brandt's settled preference seems to be for the latter option, he has given the former a try. In a recent discussion, he has suggested the following modification of the desire theory: An event is intrinsically good or a benefit for a person if it is or was desired, or under certain conditions would be desired; if one is on the whole glad when and after it has come about (and not just because it was the realization of a preceding ____________________ 15. The other reason Brandt goes on to offer for rejecting such theories is the problem of desire change over time. -107desire); and if, in case the event does not enter the person's experience, the judgment that it is occurring is pleasing. 16. This analysis adds to the desire theory what Griffin (1986: 13) has called an experience requirement: in order for the satisfaction of a desire to count as an intrinsic benefit the subject must enjoy either the desired state of affairs itself or at least the knowledge that it has occurred. The result is a hybrid view which combines elements from both the desire and happiness theories. Since it handles those cases in which the satisfaction of a desire fails to have any positive intrinsic effect on the subject's life, it is a definite improvement over the desire theory. However, Brandt's proposal faces an obvious question: if some state of affairs satisfies the last two conditions, why does it need to satisfy the first as well? Suppose that you try some novel pastime for which you had no antecedent desire,

or which you confidently expected to hate, and find that you enjoy it. Does it not enhance your well-being despite satisfying no desire on your part? If we count pleasant surprises of this sort as ingredients of well-being, then desire-satisfaction is not even a necessary feature in a theory of welfare. Once the need for an experience requirement has been accepted, we have no reason to retain any remnants of the desire theory. Brandt's hybrid view then seems but a first step toward the ultimate destination of the happiness theory. Before turning our attention to this alternative, we may pause to ask why Brandt, having once dismissed the desire theory as unintelligible, came later to propose any version of it at all, even a hybrid one. 17. The answer, I think, is that ____________________ 16. Brandt 1989a: 34. Brandt adds, here and elsewhere, the stipulation that the desires in question must be rational—that is, they must survive (what he has called) cognitive psychotherapy. For the fullest exposition of Brandt's account of rationality, see Brandt 1979: Part I.I neglect the issues raised by this account, and by the resulting idealization of desires, because they merit an extended treatment of their own. What I am interested in are defects that persist in the desire theory, even when we assume desires that meet Brandt's high standards. Later in the same discussion, he adds two further conditions: (i) moral and altruistic desires must be excluded (since their satisfaction need not benefit their subject), and (2) the desire must be held just prior to the occurrence of the event which satisfies it. The latter condition has been dealt with above. I believe, however, that the former causes more problems for desire theories (pure or mixed) than Brandt realized. The only desires whose satisfaction can be connected with the subject's well-being with any plausibility at all are self-interested ones. But then the desire theory is in danger of saying that an individual's self-interest consists in the satisfaction of her self-interested desires, which is an unilluminating logical circle. Brandt was well aware of the difficulties generated for the desire theory by the breadth of our desires and the many grounds we can have for them: see Brandt 1979: 127 and 1982: 173. However, I am not sure he ever fully appreciated how hard it is for the desire theory to specify just the desires whose satisfaction counts as welfareenhancing, without arbitrariness or circularity. 17. We may also ask why Brandt says (Brandt 1989a: 34) that he proposed this theory ten years earlier in A Theory of the Good and the Right As I read this book, it repudiates the desire theory, in all its forms, as an account of the nature of welfare—how else are we to understand the argument of ch. 13, § 1? -108in the 1989 article in which this proposal occurs the status of the desire and happiness theories has been subtly redefined. In all of the earlier discussions they have been treated as rival theories about the nature of welfare. But in this case only the desire theory is considered as a candidate for that role. Brandt endorses the hybrid theory, outlined above, as a (stipulative) definition of the concept of welfare. He then goes on to raise the question "What events or states of affairs are intrinsic goods or benefits in this sense?" and considers the happiness theory as one possible answer to this question (1989a: 34, 40,

53-4). In terms of the distinction I drew earlier, the happiness theory is being here counted, not as a theory about the nature of well-being, but as a (very short) list of its sources or ingredients. Now none of this makes the desire theory, in any of its forms, one bit more plausible as an account of the nature of welfare. But it also cannot be the right way to construe the happiness theory. As a formal theory, it tells us that something counts as an intrinsic benefit for someone just in case she finds it agreeable or enjoyable or satisfying. An account of this sort has some initial plausibility, precisely because it avoids the problems we have outlined for the desire theory. But there is no plausibility to the idea that happiness (or pleasure or enjoyment) is the only intrinsic source or ingredient of wellbeing. If one accepts the happiness theory as an account of the nature of welfare, then whatever is experienced as rewarding or fulfilling for its own sake will count as an intrinsic benefit. The list of such states will be very lengthy indeed for creatures like us: pleasure may well find a place on it, but so will achievement, play, meaningful work, friendship, good sex, peace of mind, self-esteem, and so on and on. Essentially the same list will be in play if we opt instead for the desire theory: these are surely all things which we seek for their own sake. Under no plausible criteria for counting as an intrinsic benefit could happiness (or pleasure or enjoyment) be defended as the only such benefit. Brandt is well aware that there are two separate questions here: What are the conditions which something must satisfy in order to make our lives go better? and Which things satisfy these conditions? 18. Normally, he treats both desire‐ satisfaction and happiness as possible answers to the first question. However, in his 1989 discussion he relegates the happiness theory to the second question, leaving the desire theory as the only possible answer to the first. This explains his advocacy of (a hybrid version of) that theory. Or so I conjecture. If the desire theory won by default this time around then the happiness theory enjoyed a similar victory on a number of earlier occasions. Once Brandt settled into questioning the desire theory he also settled into accepting the happiness account, at first seemingly because (since no objective theory could be taken ____________________ 18. See, e.g., Brandt 1991: 359: "But the issues must be kept separate: whether gladness is a test of well-being, and whether the only good thing is pleasure." -109seriously) it was the only available alternative. (See, for instance, Brandt 1979, ch. 13; 1982: 183.) However, he has latterly come to embrace (what seems to be) a version of the happiness theory with somewhat greater enthusiasm. In a recent discussion, he has (with some diffidence) advanced the following proposal: ... that an act contributes to an agent's well-being if it enhances his gratifications. We might call it a gratification-enhancement theory.

Ordinary sensory pleasures of course count as forms of gratification.... The theory, however, is not hedonistic in the sense of saying that gratifications are the only good; various sorts of thing can be gratifying, just as many sorts of thing can be desired. Desire-satisfaction also is gratifying, in so far as the satisfaction of a desire is pleasant. (Brandt 1991: 357) Now this suggestion seems on the right track in two important respects. First, in contrast to the desire theory, it connects a subject's well-being to the quality of her experience of the conditions of her life. Second, in contrast to Brandt's 1989 defense of the desire theory, it preserves the distinction between the single criterion for something counting as intrinsically beneficial (that it be found gratifying) and the many things which might satisfy this condition. Of all the theories of welfare considered by Brandt during his philosophical journey, this seems to me the most plausible. Is it significant that Brandt has formulated this proposal in terms of gratification, rather than happiness? Not, it seems to me, once we recollect how Brandt construes happiness. In 1967 he offered the following account: There are two components of being happy. The first is dispositional: in order to be happy it is necessary that one like ... those parts of one's total life pattern and circumstances that one thinks are important. To say that one likes them is in part to say that one is "satisfied" with them—that one does not wish them to be substantially different, and that they measure up, at least roughly, to the life ideal one had hoped to attain; but it also implies, to some degree at least, a positively favorable attitude.... The second component of happiness is the occurrence (or nonoccurrence) of certain feelings or emotions.... [I]f a man is happy he will not be subject (except briefly—for example, in connection with disappointments about noncentral things) to gloom, anxiety, restlessness, depression, discouragement, and shame, for these feelings will not occur if he likes the total pattern of his life insofar as the parts he deems important are concerned. But are no positive feelings required? It seems we would not call a man happy if he did not frequently feel joy or enthusiasm or enjoy what he was doing or experiencing. (1967a: 413-14; cf. 1989a: 41.) I have quoted this passage at some length because the conception of happiness which it sketches seems to me so exactly right, and because it clearly supports the hypothesis that when Brandt later speaks of gratification he is referring to the -110same basic phenomenon. Doubtless more, much more, needs to be said about the nature of happiness before its identification with well-being can be fully and fairly assessed. But even in advance of that, the virtues of the happiness theory, and the ways in which it is capable of avoiding the pitfalls of the desire theory, should be evident.

Brandt himself has had more to say about the nature of happiness—or enjoyment, or satisfaction, or gratification. However, his favored account has the ironic effect of reinstalling the concept of desire at the center of his theory of welfare. The account unfolds in two steps, the first of which is a hedonistic analysis of happiness in terms of pleasure. Let me begin by considering what we should mean by "happiness". I propose to use it in the sense in which it is used in "the happiest year [or happiest hour] of my life", not in the sense of "true happiness". Happiness at a time does not imply ecstasy, a feeling of modest contentment will do, if not marred by pain, disappointment, grief, or anxiety. Roughly, a happy hour or year is one of predominantly pleasant moments (enjoyments) compared with unpleasant ones, with the understanding that the intensity of the pleasantness (or unpleasantness) is taken into account.... (Brandt 1989a: 41; cf. 1966: 266-7) The second step is a motivational analysis of pleasure in terms of desire: I think myself that the property which most clearly belongs to all pleasant states and activities is simply that at the time the person wants them to continue for their own sakes (which is consistent both with not thinking that one wants them to continue, and also with not wanting to continue them, everything considered, in view of consequences or of other things one ought to be doing). And, for one experience to be more pleasant than another is for the person to want to continue it, more intensely, for itself. (Brandt 1966: 268-9; cf. 1979: 35ff.; 1989a: 42) If we combine these two analyses then we get roughly the following result: happiness consists in a preponderance of experiences whose continuation one desires for their own sake. By this route the happiness theory has turned out in the end to be a special case of the desire theory. Something has obviously gone wrong. The defects which Brandt recognized in the desire theory cannot be remedied merely by specifying that the objects of desire must be experiences whose continuation is desired for their own sake. Whatever its object, a desire can only represent (or result from) an ex ante expectation that the continuation of some state or activity will be experienced as gratifying; the satisfaction of the desire cannot guarantee the ex post gratification. Furthermore, it is easy to think of intrinsically rewarding experiences (the birth of a baby, a perfect romantic moment) which would be utterly spoiled by pro -111longation or repetition; where many pleasures are concerned, more is not necessarily better.

If Brandt wishes to reduce happiness to pleasure, it is no help to further reduce pleasure to desire-satisfaction. It will be enough to characterize pleasure in terms of experiences which are liked or enjoyed, or found satisfying or fulfilling, for their own sake; this family of attitudinal notions is not further illuminated— indeed is distorted—by being analyzed in terms of the forward-looking concept of desire. But is it desirable to take even the first of Brandt's two steps: can happiness really be understood in terms of pleasure? As long as pleasure is construed attitudinally or motivationally, and therefore also broadly, I would not want to claim that a hedonistic account of happiness is an outright mistake. But it does seem to me potentially misleading, and therefore better avoided. Our everyday lives include certain sensations—the experience of orgasm, say, or of being stroked or massaged—which stand as our paradigms of physical pleasures. These pleasurable sensations are identified as such, and distinguished from painful ones, just by their characteristic feel. It is true that they are typically liked or enjoyed, but this is not necessarily so—it is logically possible to find pleasure distasteful, just as it is possible to like or enjoy pain. Brandt, of course, does not restrict pleasure to these core physical cases, and ordinary usage gives him much support for applying the notion across the full range of enjoyable experiences. But using the term 'pleasure' to cover all of these cases has invited confusion and misinterpretation ever since the days of the classical utilitarians. And the risk seems needless: it will do just as well to speak of enjoyment or satisfaction or gratification. Just as pleasure needs no further reduction to desire, these notions need no further explication in terms of pleasure. Brandt's hedonistic analysis of happiness, and thus of well-being, courts one particular danger which he has not entirely succeeded in avoiding. Because it is easy to think of pleasures as mental states, the analysis renders the happiness theory vulnerable to experience machine objections. The experience machine, philosophical readers will recall, is a device capable of synthesizing any states of mind we wish, including the illusion that they are not merely synthetic, while we float motionless in a tank. 19. Were we given the option of plugging into the machine for the rest of our lives (and, of course, forgetting that we had done so), would we take it? Why not, if all that matters to the quality of our lives is that our states of consciousness be as agreeable as possible? It seems to me that Brandt's own response to experience machine objections (Brandt 1989a: 49-51) rather misses their force. The question raised by these objections is whether the value of an experience, for its subject, necessarily remains ____________________ 19. The thought experiment was first suggested in Nozick 1974: 42-45; see also Nozick 1989: 104‐ 108 and Griffin 1986: 9. -112constant whether or not the experience is reality-based. This question many of us will surely answer in the negative, at least for some experiences. Sometimes, perhaps, all we care about is how the experience feels from the inside—but surely not always. In these latter cases what we value must be not just the experience itself, understood as

phenomenologically bracketed, but also its grounding in the world. With this Brandt may not disagree, but he runs the risk of misinterpretation by not plainly disavowing his own commitment to a mental state theory of welfare, and by his preference for hedonistic terminology. The problem is certainly not that Brandt lacks the resources for responding more frontally to experience machine objections. He has at least two routes open to him. One is to stipulate that enjoyable experiences are to be understood extensionally rather than intensionally. I am not having the experience of listening to Mozart or watching a baseball game unless I am actually at the concert hall or the ballpark and not floating in a tank. Less dogmatically, and perhaps more reasonably, he can invoke his requirement that in order to count as contributing to our well-being enjoyable experiences must be fully informed. As applied to desires, this requirement includes exposure to "all relevant available information", where this means "all the facts, reflection on which might tend to alter the desire" (Brandt 1989a: 34). 20. Now one might expect that relevant information about an enjoyable experience you are having might include the fact that it is being generated by electrodes implanted in your brain while you are floating motionless in a tank; exposure to this information might lead you to rethink the value you attach to the experience (or, in some cases, it might not). Why not then define happiness in terms of informed enjoyment or gratification, thereby completely avoiding all commitment to objectionably solipsistic implications? This seems the appropriate point to conclude my critical assessment of Brandt's continuing inquiry into the nature of utility or welfare. It should be clear by now just how far he has moved beyond his earliest ventures onto this terrain, and how much more reflective and resilient his current views are. I have, of course, still found grounds for doubt or hesitation about some of these views. They are, to my mind, still capable of further improvement. But I have also been impressed by the extent to which they circumscribe what seems to me a defensible theory of welfare—indeed, one of the strongest and most promising available anywhere—and by the extent to which the resources needed for improving them can be found within Brandt's own storehouse of ideas. It is also easy to miss the forest for the trees: if Brandt's views are not always right in every detail, they are generally headed in the right direction. Brandt's undoubted progress through his long philosophical journey seems all the more impressive when one reflects that it has mostly been achieved against the mainstream of contemporary theorizing. ____________________ 20. He recognizes the possibility of applying the same test to enjoyments in Brandt 1982: 175. -113He has been especially visionary in resisting the lure of the currently dominant desire theory; on this I am certain that the verdict of history will vindicate him. I am equally

certain that as long as he is able to continue his journey, we will continue to learn from its results. -114-

8 Expectation Effects, Individual Utilities, and Rational Desires John C. Harsanyi

I. Rule Utilitarianism, Act Utilitarianism, and Expectation Effects Act utilitarianism (AU) is the view that a morally right action is one maximizing expected social utility in the existing situation. (Here the adjective "expected" is short for the "mathematical expectation of".) In contrast, rule utilitarianism (RU) is the view that a morally right action must be defined in two steps. The first step is to define the optimal moral rule for such situations as the moral rule whose social acceptance 1 would maximize expected social utility in such situations. The second step is to define a morally right action as one in compliance with the optimal moral rule for situations like this. 2. Even though ethical views resembling RU can be found already in Harrod (1937) and to some extent even in the writings of the great utilitarian thinkers of earlier centuries, 3. Brandt has been the first scholar to recognize the great superi____________________ 1 Think of the social acceptance of a moral rule—or of an entire moral code (see note 2 below)—as its acceptance by most members of society as morally binding, even by most of those individuals who do not fully comply with it in their own actual behavior. 2. Strictly speaking, it is preferable to define morally right behavior as one in compliance with the optimal moral code, defined as the moral code (i.e., as the set of all moral rules) whose social acceptance would maximize expected social utility for the relevant society. The reason is that in general we cannot decide what moral rule would be optimal for a given class of situations without knowing what moral rules have been accepted by the society in question for other classes of situations. For example, we cannot decide what moral obligation to assign to a father toward his children without knowing what obligations this society assigns to other relatives toward these children: For obviously we should not assign conflicting responsibilities to different persons with respect to the same children, yet must ensure that some person or persons will look after these children. 3. For instance, Berkeley wrote (in his Passive Obedience, 1712): "In framing the general laws of nature, it is granted we must be entirely guided by the public good of mankind, but not in the -115-

ority of RU over AU, and has been the real founder of modern rule utilitarian theory (see especially Brandt 1959, 1963, and 1967b). Yet, Brandt has put insufficient emphasis, it seems to me, what may be the most significant advantage of RU over AU, viz. its socially very beneficial and very important expectation effects (see below). To be sure, the most obvious consequences of social adoption of a given moral code are what may be called its compliance effects. These include the social effects of people's direct compliance, and even of their possibly very partial compliance, with this moral code. They include also the social effects of their indirect compliance, such as their willingness to put social pressure on other people to obtain their compliance, and their tendency to have guilt feelings in case of their own noncompliance. Some of these compliance effects will be directly beneficial to society while others may have the nature of social costs, to be called positive and negative compliance effects, respectively. Thus, people's direct compliance with the social code will often directly benefit them themselves as well as other people. But the efforts needed to make their own behavior actually conform to the moral code, especially in difficult situations, will have the nature of social costs. So will the efforts needed to inculcate respect for the moral code in the next generation, the guilt feelings and the social stigma resulting from noncompliance, and the loss of respect for the moral code when people see widespread noncompliance. In the case of very demanding moral codes, the high social costs of compliance may make their adoption unattractive, even if their positive compliance effects as such would be highly beneficial to society. 4. Yet, these positive and negative compliance effects are not the only social effects of an accepted moral code. For people will not only themselves comply with such a moral code to some extent, but will also expect other people to do so. This expectation will give rise to what I propose to call expectation effects: it will provide people with some assurance that their interests will be respected by other people, which in turn will provide incentives for them to engage in some socially beneficial activities they otherwise would not undertake. Accordingly, we may divide the expectation effects of a moral code into assurance effects and incentive effects. Typically these assurance and incentive effects are associated with what may be called the specially protected moral rights and moral obligations established by the moral code, defined as rights and obligations that must be honored even if this means performing actions with lesser direct social utility (though it may be ____________________ ordinary moral actions of our lives.... The rule is framed with respect to the good of mankind; but our practice must be always shaped immediately by the rule." 4. For an excellent discussion of what I call positive and negative compliance effects, see Brandt 1979: 287-89 and 1992: 144-47. -116-

permissible to put these rights and obligations aside in some special cases for the sake of sufficiently important social interests). As an example for specially protected rights, let me discuss property rights, such as a person's property rights over his boat. According to our moral (and legal) rules, nobody can use this boat without the owner's permission. This is true even if such use would increase social utility because another person would derive much higher utility from using this boat during a given period than the owner would—though infringement of this rule may be permissible under some special conditions, for instance, to save a human life. Of course, property rights do have their social costs because, as we have seen, they will often prevent use of some asset by non-owners even when its use by non-owners would produce more direct social utility than its use by the owners would. But this social cost is often worth paying because such property rights give property holders an incentive to keep their property in good repair and, more importantly, give all of us an incentive for hard work, for saving and investment, and for entrepreneurial activities, so that we can acquire property. Property rights also provide some assurance for property holders that their interests in their property will be protected, giving them some financial security and thereby making them somewhat independent of other people's goodwill. By the same token, widespread property ownership in a society provides some protection from abuses of power by the government and by unduly powerful private interests. As an example for special obligations, let me discuss people's obligation to repay any borrowed money in due course. Our moral (and legal) rules require borrowers to repay the money they owe to the lender at the agreed time, together with any interest agreed upon—except if this would cause extreme hardship. This obligation obtains even if the borrower is much poorer than the lender is. Yet, when this is the case then by actually repaying the money the borrower will in fact decrease social utility because the lender's utility gain when the money is returned to him will be much smaller than the borrower's utility loss will be, since the former will have much less need for the money than the latter would have. Yet, in spite of this, the rule that borrowed money has to be repaid is a socially very useful rule, even in such cases. For if rich people believed that poor people were free of the obligation to repay loans received from rich people, then they would not lend money to poor people at all. Thus, both poor people and rich people benefit from this rule. Poor people benefit because this rule gives rich people an incentive to lend money to them under suitable conditions; whereas rich people benefit because this rule gives them some assurance that they normally will not lose their money by lending it to other people— even if the latter are much poorer than they are. Because of these socially very beneficial expectation effects, i.e., assurance and incentive effects, such specially protected moral rights and obligations often have -117-

very high social utility. 5. Thus, it is not surprising that the moral codes of all advanced societies make very wide use of such rights and obligations. No doubt, so would any ruleutilitarian moral code. In contrast, an act-utilitarian moral code could not recognize any specially protected moral rights and obligations without inconsistency. For, by its very definition, such a moral code would have to permit, and indeed would even have to require, infringement of any moral right and any moral obligation if this would increase, however slightly, social utility. By disregarding the expectation effects of alternative moral codes, Brandt has deprived himself of an important argument for the superiority of RU over AU. He has also deprived himself of what may be the simplest argument to refute the claim made in Lyons 1965 and Hodgson 1967 as to the "extensional equivalence" of RU and AU. For given the fact that a rule-utilitarian moral code can, whereas an act-utilitarian one cannot, recognize specially protected moral rights and obligations, the two moral codes obviously cannot be extensionally equivalent.

II. The Nature of Individual Utility Desire Theory vs. Happiness Theory Since social utility is usually defined in terms of individual utilities (as the sum or as the arithmetic mean of all individual utility levels), the concept of individual utility (or of individual well-being) is a central concept of utilitarian theory. Nineteenth-century utilitarians were hedonists and tended to define an individual's utility level as a measure of his or her happiness, that is, as a measure of the degree to which pleasure dominated over pain in his or her life at a given time. In contrast, most contemporary utilitarians define a person's utility level as a measure of the degree to which his or her desires or preferences are satisfied. Moreover, most of these authors would restrict this definition to this person's informed desires or preferences, i.e., to those hypothetical desires or preferences he or she would entertain if he or she 6. had all the relevant information and also made proper use of this information. For convenience, I shall follow Brandt in referring to these two theories of utility as happiness theory and as desire theory. 7. ____________________ 5. Ever since Bentham's well-known outburst against the somewhat inflated humanrights rhetoric of the French revolution in his Anarchical Fallacies ("nonsense upon stilts"—see Bentham 1843: 501), most utilitarian authors underestimated the social utility of recognizing a wide range of moral rights protected from infringement by the accepted moral code. 6. In what follows, for stylistic reasons, in similar phrases, I shall usually omit the female pronoun. 7. Though I am going to use the terms "happiness theory" and "desire theory," the term "preference theory" would be actually a better description of the latter. For a person's utility function

-118Brandt used to be a supporter of desire theory (Brandt 1966) but in recent years has switched his support to happiness theory (Brandt 1982 and 1989a). Yet I shall argue in favor of desire theory. The basic objection to happiness theory is that what we actually want are not only desirable subjective experiences, i.e., desirable states of our own mind such as pleasure and happiness, but rather include some desirable objective outcomes in the outside world such as having money, social status, success, worth-while accomplishments, knowledge and understanding of the world and of our own place in it, love and respect by others, and so on. Moreover, we do not want to live in a fool's paradise by merely imagining that we have these desirable things. Rather, what we want is real possession of these things in actual fact.

Altruistic Goals To be sure, a supporter of happiness theory may try to counter this objection by suggesting that, even if we do set ourselves some goals other than our own pleasure and happiness, we do so only because of the purely instrumental value these goals have in promoting our own pleasure and happiness. But this can hardly be true for our altruistic goals based on an unselfish concern for other people's happiness rather than on a selfcentered concern for our own. No doubt, when we try to promote other people's happiness we often find that by this means we do promote also our own. Even if we perform altruistic actions benefiting people without any close personal ties to us, even our own life may become somewhat happier, richer, and more rewarding. But what seems to enhance our own happiness most are very close and deep personal relations with our friends, lovers, and marriage partners, based on genuine mutual concern by both sides for each other's well-being. 8. ____________________ admits of definition in terms of his preferences as such but does not admit of definition in terms of his desires without making use of his preferences. This is so because in order to define his utility function we must know how he would rank his various desires according to their relative importance to him. Yet, it is only his preferences that provide this information. For instance, suppose we know that John desires both a wrist watch and a cassette recorder. This will not tell us which of these two desires are more important to him unless we know which of these objects he would prefer if he had to choose between them. 8. In fact, when we make major sacrifices for others, this may make us so much poorer or may worsen our own situation so much in some other way that our own happiness will seriously diminish. But it might happen that, even so, to make these sacrifices was the best thing for us to do even from our own selfish point of view, or that at least it was the lesser evil for us under the existing conditions. Even a soldier sacrificing his life for his comrades might have chosen the lesser evil, given the alternatives available to

him, because he might have felt that to survive knowing that he acted in a cowardly manner by letting down his friends would have been an even unhappier alternative for him than to lose his life. -119Of course, even if caring for other people does tend to promote our own happiness, it does not follow that this is actually our main reason for doing so. When we do things for other people, we may do them primarily for their sake rather than for our own sake. When parents help their children in various ways they may do so simply because they love them. They will feel themselves happier by doing so but their own happiness will be often a very minor consideration. Happiness theory in its pure form seems to suggest that when we are helpful to other people then, in ultimate analysis, our primary purpose is always to promote our own happiness rather than theirs. If this suggestion were right it would mean that none of our actions could be truly altruistic actions—surely an absurd conclusion.

Other Objectives with Intrinsic Value Apart from our own happiness and that of other people, we seem to recognize also other objectives as having intrinsic value, such as having a life rich in worth‐ while accomplishments, or having one rich in highly interesting experiences. This can be seen from the fact that we find it a meaningful question whether we would prefer a somewhat less happy life with many worth-while accomplishments or with many very interesting experiences to a happier life without such accomplishments and without such experiences. Indeed, we can understand that somebody might actually choose a life rich in accomplishments or one rich in interesting experiences rather than a somewhat happier but less productive or less interesting life—regardless of what our own choice would be in such a situation. Of course, a supporter of happiness theory might argue that if anybody actually chose the life of great accomplishments or that of great experiences then this very choice of his would indicate that he would expect to find more happiness in that life than in the life devoid of such ingredients. But such an argument would have little force because it would amount to an attempt to ensure the truth of happiness theory solely by an arbitrary terminological fiat.

Happiness Theory as a Special Form of Desire Theory Brandt has suggested that happiness theory can be regarded as that version of desire theory that measures our utility for an experience by the intensity of our desire for this experience and for its continuation at the time we have this experience (Brandt 1982: 182). 9. Subject to an important qualification (due to the

____________________ 9. Brandt's actual statement is: "The difference [as compared with other versions of desire theory] is that ... happiness theory assigns utility only to moments of experience [Brandt's own italics], -120problem of misguided desires—see below), I agree with Brandt that this is a good characterization of happiness theory. I agree with him also that the suggested approach to utility measurement will give quite good results in many cases. Yet, in many important cases this approach does break down and when it does this points to some serious limitations to the validity of happiness theory. For instance, let us try to use this approach to assess the utility of drug use for the user of a highly addictive drug. When he uses the drug he will tend to experience very intensive pleasure and happiness at the time, together with a strong desire to prolong and to repeat this experience. Hence, according to happiness theory, we must conclude that use of the drug will be of a very high positive utility for him. This conclusion, however, is surely very paradoxical. For we all know that, in actual fact, use of the drug will be of very high negative utility for him: It will lead to a selfdestructive, very stubborn, and possibly even irreversible, drug addiction, which will deprive him of any real control over his own life, ruining his health, his family life, and his ability to work. Desire theory, if based on the notion of informed desire, can easily resolve this paradox by pointing out that the drug addict's strong desire for the euphoria induced by the drug and for its continuation is far from being an informed desire. Rather, it is a desire that can persist only because this very euphoria clouds his judgment and makes him unable to visualize effectively the disastrous consequences of his continued drug use. On the other hand, this way of resolving the paradox is not open to happiness theory in its usual form. For, even though the drug addict's desire for a drug‐ induced euphoria and for its continuation is a very misguided desire in terms of his real objective interests (as defined by his hypothetical informed desires), we have no reason to deny that his great pleasure and happiness induced by the drug are genuine subjective experiences of great pleasure and happiness. Thus, as long as happiness theory retains its basic assumption of close connection between happiness and utility, it cannot avoid the paradoxical conclusion that drug use has very high positive utility for the drug user. 10. This completes my discussion of some objections to happiness theory. Next I discuss what I take to be Brandt's main objection to desire theory, viz. the difficulties arising from the variability of our desires in their intensity over time. ____________________ with the degree of utility depending on how intensely this experience is wanted (or its

10.

continuation is wanted) for itself at the time" (Brandt 1982: 182). The problem that drug-induced euphoria poses for happiness theory is closely related to that discussed by Smart (1961: 11-15) in connection with experiments involving electric stimulation of the "pleasure center" in rats' brains. -121-

The Variability in the Intensity of Our Desires As Brandt has pointed out, this variability poses serious problems for desire theory. For it means that if we want to define a person's utility level as a measure of the degree to which his desires are satisfied then we must choose the specific point(s) of time to be associated with the desires we want to use in this definition. Yet, as Brandt has rightly argued, choice of this (or of these) time(s) will pose very difficult problems, and will be hard to do in a principled manner (Brandt 1982: 176-84). Yet, these problems, it seems to me, can be avoided by making use of two important distinctions. One is the distinction between a person's local desires and his global desires, proposed by Griffin (1986: 13, 15). The former are his desires for specific experiences or outcomes, whereas the latter are his desires for one kind of life rather than for other kinds when he considers all these possible lives in their entirety. The other distinction is one commonly made (but the terminology I shall use to describe it is, once more, Griffin's) between a person's actual desires, i.e., his desires as they actually are, and his informed desires, i.e., the hypothetical desires he would entertain if he had all relevant information and made proper use of this information. Using these distinctions, we can redefine a person's utility level in terms of his global desires rather than all his desires, and indeed in terms of his informed global desires rather than his actual global desires. Moreover, once we have done this, we are free to define his utility level at any time t purely in terms of his informed global desires at that time t. For these desires, being global desires, will not only cover his interests at time t itself but rather will cover all his interests from that time on till the end of his life. Moreover, they will give no undue priority to his interests at one time over those at other times—say, by giving priority to his present interests over his future interests. For, we must remember that a person's informed desires, including his informed global desires, are not directly observable variables but rather are theoretical constructs and as such, by their very definition, are meant to be free of distortions caused by lack of imagination in assessing the future, or by emotional inability to postpone gratification, or by other similar factors. 11. ____________________ 11. This does not mean that a person's informed global desires must always give the same weight to his interests at various periods of time. It means only that the weights they assign to his interests at different periods must reflect the relative importance he wants to give to those time periods, and should not reflect merely various deficiencies in his

information-processing and decision-making ability. When we try to define a person's informed global desire in greater detail, we must of course keep in mind that neither this person himself nor any outside observer can predict this person's future behavior, his future attitudes, or future changes in his environment. This means that the best we can -122Thus, we arrive at the following definition: A person's utility level at any given time is a measure of the degree to which his informed global desires at that time are satisfied. It may appear that this definition makes a person's utility level independent of the extent to which his local desires are satisfied. But this is not the case. To be sure, our definition mentions only his global desires. Yet, the latter are basically desires for a high overall quality of life, 12. and the latter obviously does depend, not only on some global characteristics of his life, but also on many local facts, such as the extent to which his various local desires are satisfied. Indeed, satisfaction of any reasonable local desire of his will yield him some positive utility gain—which will be all the greater the higher the importance of this local desire within the totality of all his desires and within his life as a whole. 13.

III. Cognitive Psychotherapy and the Rationality of Our Desires Rational Desires Brandt's concept of a rational desire is presumably meant to capture the intuitive notion of a desire good to have in terms of our real interests—just as other authors' similar concepts are, including my own concept of an informed desire. But it seems to me that his definition of this concept fails to capture this intuitive notion in some important respects. He defines rational desires as those that would survive a hypothetical psychological procedure he calls cognitive psychotherapy, in which we would in our own mind confront our desires with all relevant information, by representing this information to ourselves repeatedly at appropriate times "in an ideally vivid way". Here the term 'relevant information' is understood to include all information "relevant in the sense that it would make a difference to [our] desires and aversions if [we] had thought of it" (Brandt 1979: 111; see also pp. 110-29). I have two objections to this definition. One is that we do not need the cognitivepsychotherapy criterion because we can decide whether a given desire ____________________ do is to consider a number of alternative possible scenarios, and then assign what we

feel are reasonable probabilities to all of them. In a person's quality of life I am including the extent to which his unselfish desires and preferences—at least those really important to him—are actually satisfied. 13. Note that in general there is no reason to expect that a person's (global) utility level will be simply the sum of these local utilities. For, in the terminology used by economists, these local utilities may behave like complements or like substitutes rather than like independent "goods". (The joint utility of two complements always exceeds the sum of their separate utilities whereas that of two substitutes always falls short of the latter.) 12.

-123of ours is rational or not simply by a rational judgment as to what our real interests are concerning this desire, that is, by asking whether the overall quality of our life would be better or not if we could get rid of this desire. (Reflection will show that this is the same question as whether this desire of ours is an informed desire or not, i.e., whether we would entertain it at all if we had all the relevant information and had also made proper use of this information.) For instance, suppose that Smith is a heavy smoker, who feels a strong urge to light up many times during a day. Yet, he is fully aware of the serious risks his smoking habit poses to this health in the long run. But he goes on smoking, perhaps because he simply turns his mind away from these risks whenever he feels a strong urge to smoke. On the other hand, suppose that in a 'cool hour', when he feels that his clear thinking is not clouded by any pro-smoking emotion (and, for that matter, by any anti-smoking emotion either), he tries to make an unbiased rational judgment as to whether, in terms of his real preferences, the pleasure he gets out of smoking justifies taking such serious risks to his health or not. If he answers the question in the affirmative then he can conclude that his desire to go on smoking is a rational desire, 14. whereas if he answers it in the negative then he must reach the opposite conclusion. Now suppose he does reach the conclusion that this desire of his is against his real preferences and, therefore, is an irrational desire. As we know very well, this rational judgment may not automatically extinguish this desire, and may not even weaken it to any significant extent. If he wants to weaken it seriously or to extinguish it then he may have to use Brandt's cognitive psychotherapy, i.e., may have to make a special effort to review the facts underlying his rational judgment, presenting them to his own mind repeatedly in a very vivid manner over a considerable period of time. Or he may have to try other procedures such as ones involving hypnosis or various medications, and so on. In other words, the strength of our desires is not under the immediate control of our rational judgments. Rather, it is controlled by our motivation system, whose responsiveness to cognitive psychotherapy, or to hypnosis, or to medication, will depend primarily on the physiological properties of our nervous system rather than on the rationality or irrationality of these desires themselves.

Let me now restrict my discussion to the responsiveness of our motivation system to cognitive psychotherapy. To be sure, if one of our desires is extinguished or is seriously weakened by this procedure, this will in fact indicate that it was an irrational desire. This is so because, by its very definition, cognitive psychotherapy amounts to repeatedly going through both the reasons we may have for sticking to this desire and the reasons we may have for trying to give it up. Thus, ____________________ 14. Admittedly, a person well informed about the health hazards posed by smoking is rather unlikely to arrive at this conclusion. -124if this procedure does extinguish or weaken it, this must mean that the latter reasons are stronger than the former reasons are, i.e., it must mean that this desire is irrational. Yet, if cognitive psychotherapy has no noticeable effect on the intensity of this desire, this need not imply that this is a rational desire: it may mean only that the responsiveness of our motivation system to the stimuli to which it was exposed during cognitive psychotherapy is rather low—presumably due to some physiological properties of our nervous system, which may have no connection with the rationality or irrationality of this desire itself. This leads me to my second objection to Brandt's definition of rational desires. I have already argued that the rational-judgment criterion makes Brandt's cognitivepsychotherapy criterion for rational desires redundant. Now, I shall argue that the latter criterion is also unreliable. This is so because, as I have just tried to show, if a desire of ours did survive a properly conducted cognitive psychotherapy, this might not show that it is a rational desire but might show only that the responsiveness of our motivation system to cognitive psychotherapy is rather low, perhaps for physiological reasons unrelated to the rationality or irrationality of this desire. 15.

Irrationally Acquired Desires I also have a third objection, though it is one to a relatively minor aspect of Brandt's theory. It is to his view that any desire of ours must be classified as an irrational desire if it was originally acquired in an irrational manner. I want to argue that this principle has only limited validity. For instance, suppose that somebody originally acquired his interest in classical music in order to please his grandmother, who was very fond of such music. No doubt, this was not a very rational way of choosing his musical interests. Nevertheless, it seems to me that his present desire to attend many classical concerts must be recognized as a rational desire if he now derives genuine aesthetic pleasure from attending such concerts.

Indeed, we cannot deny that some of our irrationally acquired desires are rational desires greatly enriching our lives. For instance, many very successful professional careers, many happy marriages, and many deep friendships had their origins in some irrationally acquired positive attitudes due to some silly accidents ____________________ 15. Let me use this opportunity to correct a misunderstanding. Brandt (1982: 173) attributes to me the view that a person's preferences based on sadism, envy, and malice, i.e., his malevolent preferences, should be excluded from his preferences used to define his individual utility function. This is not my view. All I have suggested is that people's malevolent preferences should not be used in defining our society's social utility function because we cannot rationally require the members of our society to help other members in satisfying their malevolent preferences. -125such as that somebody sat next to somebody else on some very unimportant occasion. True, some irrationally acquired desires, and in particular those acquired by uncritical acceptance of some attitudes, or even some prejudices, current in our social environment —which Brandt (1979: 117) calls artificial desires—will often cause considerable harm. For example, a boy with a natural aptitude for, and with a strong interest in, repairing motor cars may be prevented from becoming a motor mechanic because early in life he acquired an artificial aversion to this occupation by his parents' repeated disparaging remarks about this kind of work. More generally, the main damage that artificial desires and aversions seem to produce is needlessly narrowing the choices effectively available to people with such desires and aversions. This is perhaps a more serious problem in the case of artificial aversions than it is in the case of artificial positive desires. Thus, if somebody is kept out of the automechanic occupation by a prejudice against it then he may never find out what he has been missing. In contrast, if somebody is attracted to this occupation by a positive prejudice in its favor then he will presumably find out fairly soon if this is not the right occupation for him—but of course in the process he might waste a lot of valuable time. On the other hand, as my classical-music example shows, an artificial desire may also widen people's effective choices. For the person who got interested in classical music to please his grandmother may have never found out how enjoyable such music can be if he had not gotten acquainted with it early in life to please his grandmother. The main conclusion, it seems to me, we can draw from this discussion is that the only reliable criterion for the rationality of our desires is whether, in terms of our informed preferences, we would have a better life without these desires or not. All alternative criteria can be seriously misleading—whether they ask how well these desires would survive cognitive psychotherapy, or ask how rationally these desires were originally acquired by us.

Let me end this paper by expressing my admiration for Richard Brandt's many distinguished books and papers in ethics and in the theory of human motivation. Though I did not have the privilege of having him as my teacher, I feel I have greatly benefited over the years from his numerous publications as well as from occasional personal discussions and from correspondence with him. -126-

9 The Economics of Knowledge and Utilitarian Morality * Russell Hardin During the past fifty years, utilitarianism has been dichotomized into extreme and restricted, direct and indirect, and act- and rule- variants. These dichotomies are often taken to be the same. In an important sense, however, the last of the series is more narrowly defined than the other two, which do not reduce to the dichotomy of act- and rule-utilitarianisms. We can achieve utility indirectly without following rule-utilitarian rules. For example, we can create institutions, which then may carry out particular policies that are utilitarian. Of course, they carry out policies through the actions of individuals in the institutions, whose actions are determined by their functional role requirements rather than by direct act-utilitarian assessment. Our action in creating or supporting such an institution only indirectly contributes to utility. We can also devise a system of moral education, such as Richard Brandt proposes, in which, again, we strive to achieve best overall outcomes through intermediary devices, such as careful inculcation of rules simple and few enough for most to follow (Brandt 1979: 286‐ 300). One might count this form of utilitarianism as itself institutional. Hence, there are at least three forms of utilitarianism that fit into the larger dichotomous categories: act-utilitarianism under direct and extreme theories, and both rule‐ utilitarianism and institutional utilitarianism under indirect and restricted theories. A striking feature of these three forms of utilitarianism—act-, rule-, and institutional—is that they seem to be grounded in different epistemologies of the relevant actors. I will investigate the general claim that the epistemic conditions ____________________ * I thank Richard Brandt for discussing these issues while I was writing, Brad Hooker for extensive and insightful written commentary on an earlier draft, and the Andrew W. Mellon Foundation for general support of this and other projects. This paper parallels another (Hardin 1993) that addresses the law rather than moral theory. -127for various forms of utilitarianism differ and that their different conditions are a chief reason for the different utilitarianisms, perhaps even the chief reason. Epistemic issues pose problems even if there are no conflicts of moral principles or of interests. There are essentially three kinds of knowledge that come into discussion here: simple personal knowledge of the consequences of one's particular actions, personal knowledge of the utility of personally following general rules, and institutional or organizational

knowledge of the utility of general policies. The claims for rule-utilitarianism and institutional utilitarianism imply that the latter two kinds of knowledge do not reduce to the first kind of personal knowledge. My argument will be that personal knowledge of the utility of following general rules either does reduce to the simpler kind of personal knowledge or is incoherent, while institutional knowledge can genuinely diverge from simple personal knowledge. The focus on the economics of knowledge in part overlaps and in part complements what one might more nearly call the psychology of knowledge and commitment that drives much of Brandt's account.

I. The Economics of Knowledge The philosophical theory of knowledge is generally either about such foundationalist issues as how one can know something to be true or about coherentist accounts of the fit of some knowledge with other knowledge. For pragmatic inquiries, such as ethics, the more pressing question is why and how one would come to know various things. 1. There are two economic responses to this question. First, we come to know some things because it is directly in our interest to know them, in the sense that the costs of learning are likely outweighed by the anticipated benefits of knowing. For example, if you wish to lead the life of a doctor, you know to set about learning relevant things. The costs of learning may seem grievous, but they should eventually be outweighed by the benefits of a desirable career. Second, we come to know some things because the process of learning is itself valuable to us, so that we incidentally learn things that may or may not turn out to be beneficial. For example, you might enjoy reading a daily newspaper and occasionally you may learn something from it that later serves your interest, but you mostly learn things of no benefit to you beyond your pleasure in or consumption of the knowledge—and it is this pleasure that drives you to read. Theories of knowledge that are concerned with justified belief tend to focus on the object of a belief. An economic theory of knowledge focuses instead on the believer, on the act or process of coming to believe. To simplify somewhat, ____________________ 1. For broader consideration of this issue, see Hardin 1992a. -128standard theory of knowledge is about the criteria for establishing the truth of some bit of knowledge. The economic theory is about how and why you or I come to have the particular bits of knowledge we have. Attached to each of our bits of knowledge is our range of confidence in its truth. Some of those bits, perhaps most of them, might fail some philosopher's test of justified true belief. I cannot distinguish beliefs from my knowledge because I do not have a methodology for establishing the certain truth of any belief. I have only the methodologies I use to establish any of the beliefs I have. Hence, there is little point in distinguishing between beliefs and knowledge. In particular, there is

no place I can stand to determine that some one of my bits of knowledge is a justified true belief. An economic theory of knowledge says in essence that all one can claim of any piece of knowledge or belief is that it has had certain resources poured into testing or confirming its truth. There is, for an individual knower or believer, no independent way in which to determine that a particular bit of knowledge is true. The only knowledge is that he or she has invested some amount of effort in finding out or has come to know it in a certain way. From the perspective of the knower, therefore, little if anything can be true in some objective sense. (This follows, perhaps obviously, from basing the theory in the subject.) Moreover, there appears to be no other theory of knowledge that is plausible for the agent. Relying on the authority of another, for example, is merely a form of investment in knowledge. In any case, on the agent-centered view of the economic theory of knowledge, again, virtually everything is merely knowledge or merely belief— there is no distinction. Someone else could say my knowledge is lousy or very good, but that person's own knowledge would be as subjective as my own. That person would therefore also not be able to hold a standard of truth. When I face a moral (or any other practical) decision, I have available to me a lot of knowledge obtained in various ways about many things, few of them relevant to the choice I face. I may also be able to inquire after more knowledge that is specifically relevant to the choice at hand. If I am motivated by a particular, explicit morality, my actual choice will be a function of the bits of knowledge I have. Some of that knowledge might count as moral beliefs, some of it as beliefs about objective states of the world. But I might not know enough to keep these categories straight. As a user of my knowledge, my pragmatic problem is not so much how confident I am of the correctness of my knowledge, but merely of the confidence I have in that knowledge relative to other bits of alternative knowledge. I might hesitate to make strong claims for the rightness of my knowledge, but I might not have the luxury of hesitating very long in acting on it. When we later judge my action—either its rationality or its morality—we should judge according to the pragmatic considerations that drove me. The answer to the question, "Was it moral (or rational) of me to do x in situation S?", depends on whether my action -129was morally (or rationally) consistent with my knowledge at the time. The answer does not depend on whether my action was in some more general sense the best thing one could have done in that situation. Rather, the answer depends on my life history. Why does the economics of knowledge or belief matter for a utilitarian theory? A theory of practical knowledge is necessary for reaching normative judgments about normative claims. If I wish to say you were morally at fault for choosing one action rather than another, I will have to assume that you understood your actions well enough or that you were culpably ignorant for any failure to understand your actions. Some of your beliefs

might be wrong but not culpably so. For example, most Europeans in 1492 may have believed the world was flat. Richard Lewontin notes of scientists that an "observer who is not immersed in the practice of a particular scholarship and who wants to understand it is at the mercy of the practitioners" (Lewontin 1991: 140). That is to say, the methodology for my knowing many things in my own field is different from that for knowing virtually anything in other fields. The methodology of trusting the practitioners is a compromise. I cannot know as well as they can. But this is a, perhaps complex, variant of the "ought implies can" problem. I cannot know those things better; therefore it cannot be true that I ought to. Perhaps I could have come to know them better, but probably only by forgoing the stronger knowledge I have in my own field. But as of right now, I cannot know them. This view of the role of knowledge in a claim of moral responsibility calls into question the standard utilitarian move to look to the actual consequences of actions for summary judgment of their morality. This move requires that the judge have a superior methodology for reaching the correct judgment of what to do. Perhaps a particular judge does have. But this cannot be a claim for a transparent trump. The trump cannot be seen for what it is from its simply being laid on the table. Typically, it must be argued and defended against the actor's own judgment. Moreover, the actor being criticized for moral failure did not have the trump argument when it was needed. But it would be foolish to say those faced with decisions should not act until they have perfect understanding—they will never have perfect understanding. In a utilitarian account, actors can only be judged by what they know or reasonably believe and by how their actions fit with their knowledge. What we need, therefore, for making moral assessments of actions is a plausible theory of the actor's knowledge. The economic theory of knowledge is a compelling account of the actor's knowledge and of the meaning of its being called "reasonable". On the economic theory of morally relevant knowledge, the demands of morality might seem stricter for you than for me. You may know much more about relevant causal relations than I and may also be more accustomed to and, therefore, more adept at figuring out what action is right in some context. Indeed, it may typically be true that a moral theorist faces sterner demands than -130the average person faces. The moral theorist is specialized—just as the auto mechanic or the dentist is—and should be expected to perform better at her specialty than those who are not specialized. If specialization is valuable, then organizational pooling of specialists is likely also to be valuable. Institutions have several advantages over merely aggregated individual actions. They can coordinate large numbers of actors and they can counter perverse incentives. An especially important reason for both of these is that they can collect information that most individuals could not collect or that individuals acting spontaneously could not collect as efficiently. Without such information, it might be virtually impossible for a population to achieve some collective benefits.

Unfortunately, while institutions have epistemological advantages over individuals, they may also have epistemological disadvantages. I can come to know you in a way that no organization could. Indeed, it is plausibly a fallacy of composition even to suppose an organization can know you in many ways that individuals can. Among the ways that I can know you and a particular organization cannot know you may be many that are centrally important in determining your welfare or the welfare effects on you of various actions or rules. Hence, there may be real tradeoffs in opting for organizational resolutions of many problems. It is only somehow "on the whole" that an organization may be superior to aggregated individuals in handling certain problems. For other problems, individuals may be better left to their own spontaneous devices.

II. Act-, Rule-, and Institutional Utilitarianisms Act-utilitarianism is too familiar to require much discussion. Act-utilitarians hold that what an individual should do in any context is simply what would, from then and there, produce best results on the whole. There should be no prior constraints. In particular, there should be no standard moral rules that focus on kinds of action as opposed to the actual consequences of individual actions on a case by case basis. An early complaint against utilitarianism was that it required too much calculation and could therefore be self-defeating. One answer to this criticism is to argue that we do not really need to calculate much of the time. Rather, we merely follow good rules, perhaps even including those of ordinary morality. That might sound perversely circular. We rescue the claims of utilitarianism against other moralities, including ordinary morality, by importing into it ordinary morality. But this move would not be circular if we could show that the rules of ordinary morality are somehow socially selected because of their contribution to general utility. I will consider below this and other ways in which we might justify belief in the utility of following particular moral rules. -131The general concern with rule-utilitarianism is arguably motivated by concern with law and other institutional resolutions of problems, which were problems centrally addressed by David Hume (Hardin 1988: especially ch. 2) and John Stuart Mill (On Liberty and On Representative Government). In keeping with the usual drift of moral theory through much of this century, however, rule‐ utilitarianism has been developed primarily as a theory of personal morality. Oddly, then, its force may depend on the epistemological capacities of institutions while its actual application Can only call on the epistemological capacities of individuals, including what they may learn from watching institutions and large collectivities. How does institutional utilitarianism differ from rule-utilitarianism? Institutional utilitarianism is addressed to institutional resolutions of, usually, complex problems involving strategic interactions of many people. Even if all of these people share a commitment to utilitarianism, they may run afoul of strategic interactions that lead them into less than optimific results. It seems plausible that similar problems motivated the

invention of rule-utilitarianism. However, rule‐ utilitarianism is generally addressed to individual choosers to override their individual calculations in particular classes of cases. Unfortunately, it raises the question, asked by virtually every critic of rule-utilitarianism, how one is to know that following a particular rule is optimific when violating it seems, from direct calculus, likely to produce a better result. Implicitly, that question is about the comparative epistemological grounds of one's belief in the optimality of the rule versus that of the direct calculus. Rule-utilitarianism seems confused in the following sense. If rule‐ utilitarianism and actutilitarianism give different recommendations, a mere appeal to optimific results does not resolve the difference. Or, rather, act‐ utilitarians claim, it settles the issue in favor of actutilitarianism. Because institutional utilitarianism includes consideration of positive characteristics of institutions, its differences with act-utilitarianism can be resolved with a direct appeal to optimific results: Things would be worse without the institution and without its capacity to marshal many actions on behalf of the general welfare. Moreover, they might sometimes even be worse without the institution's incapacity to respond to specifically act-utilitarian considerations. If X-utilitarianism is distinctively different from act-utilitarianism, it must conflict with act-utilitarianism in its prescriptions in some cases even though these prescriptions must, in both theories, be derived from optimific considerations. X-utilitarianism is obviously not distinctively different from act‐ utilitarianism if it requires that, whenever actutilitarianism and X-utilitarianism give conflicting prescriptions for action, actutilitarianism trumps X‐ utilitarianism. Virtually every rule-utilitarian theorist supposes rules for action must sometimes violate act-utilitarian assessment. Is institutional utilitarianism distinctively different from act-utilitarianism? As I will argue below, act -132utilitarianism may recommend one course of action to me while institutional utilitarianism recommends that I be sanctioned if I do not follow a quite different course of action. But, in a compelling sense, this difference is a difference in the world, not in the theory.

III. Rule-Utilitarianism David Lyons has argued that rule-utilitarianism reduces to act-utilitarianism if calculation is carried out sufficiently minutely and completely (Lyons 1965). This claim is not about indirect theories in general but only about rule-utilitarianism in particular. To see why, consider the merits of institutional utilitarianism and act-utilitarianism. Does institutional utilitarianism reduce to act-utilitarianism? No, or at least it need not. The epistemology for institutional action may be inherently different from that for individual action. Unless the knowledge and constraints on knowledge that a relevant institution faces are identical to those for an individual, the two bodies of knowledge may include disparate, contrary "facts".

At Lyons's a priori level of argument, knowledge seems to be generally accessible. There are no knowledge costs that trade off against the benefits of using the additional knowledge. Lyons's argument essentially abstracts from the pragmatic, epistemological constraints that recommend creating institutions to resolve some problems. Hence, his argument genuinely is only about rule-utilitarianism, not more generally about all indirect theories of utilitarianism. Pragmatic constraints can sometimes recommend against creating institutions and for leaving responsibility for action to moral individuals. For example, it may make no sense to put fences around small ponds, to post lifeguards at them, or to seek legal sanctions against adults who fail to rescue anyone floundering in such a pond. But if pragmatic constraints recommend the use of an institution in some context, it is likely that the pragmatic constraints on individual and on institutional knowledge differ in that context. It is the difference between institutional and individual epistemologies that augurs against rule-utilitarianism. The typical individual cannot be supposed to know enough to know that violation of a rule in this case would be good but is supposed to know enough to know that following some rule is typically good. Perhaps it is plausible that sometimes one would know a general tendency without knowing anything about individual cases from which to calculate the general tendency. For example, one might justifiably believe politicians tend to be deceitful and might deduce from this that a certain politician is deceitful even without knowing much of relevance about this person or about any other particular politician. (One may know nothing more than that obviously a politician might have strong career interest in deceiving the public.) But strong rule-utilitarians -133must want more than this to be true. They must be able to say that following some rule, such as 'Do not lie', generally leads to the best results or even that it always leads to best results despite apparent evidence to the contrary in particular cases. Since we are here concerned with causal relationships, it seems preposterous to claim a unique relationship with certainty—to suppose following some particular rule is always beneficial. Lying to Immanuel Kant's threatening assassin by telling him that his intended victim is not hiding in one's home seems, for a utilitarian, a very good bargain. 2. Arguments that it is not a good bargain are likely to turn on causal considerations, such as whether one can really believe the assassin's threats. For Kant it was supposed to turn on a logical deduction from first moral principles without regard for mere consequences. 3. For a rule-utilitarian, it can only turn on causal tendencies to enhance welfare, not on rationalist or deontological inferences. The rule-utilitarian might assert that there are at least two rules in conflict here: not to lie and not to assist in a murder. Since the consequences of violating the latter rule seem considerably worse than those of violating the first, we may violate the first rule and lie. But this just is an act-utilitarian salvation.

IV. Institutional Utilitarianism John Rawls has presented one of the clearest arguments for indirect, institutional achievement of utilitarian results in a defense of the utilitarianism of criminal justice (Rawls 1955: especially 3-15). Recall a standard objection to utilitarianism in Rawls's early years (nearly forty years later it is still, so slowly does philosophy change, a standard objection). The objection is that utilitarianism could not support a principle of criminal justice. Rather, it must permit and even commend the deliberate sacrifice of an innocent if that would bring calm and save lives. Rawls noted that a serious utilitarian account of criminal justice must be an account of an institution, not of individualized action. Then he asked just who in the institution of justice would have the authority to intervene to sacrifice innocents. With remarkable gentleness, he reduced the question to its core silliness (Rawls 1955: 10-13). One may recognize in this argument a nascent version of his ____________________ 2. Writing in the Parisian press, Benjamin Constant accused a "German philosopher" of holding the absurd view that "to tell a falsehood to a murderer who asked us whether our friend, of whom he was in pursuit, had not taken refuge in our house, would be a crime." Responding to this provocation, Kant asserted that he did indeed hold that the duty to speak the truth is "an unconditional duty which holds in all circumstances" (Kant 1797 [1909: 361, 364]). 3. W. I. Matson argues that Kant's conclusion that one must not lie but must truthfully tell the assassin where to find his victim is wrong in Kant's own theory. He supposes Kant could have said otherwise only because he had lived too long (Matson 1954 [1967: 336]). -134later theory of justice. In this theory we create institutions of justice in the face of sociological and psychological constraints and then we follow their requirements. We do not commonly take direct personal actions for broad distributive justice. 4. Rawls's argument is essentially one of institutional utilitarianism, although it is often wrongly thought of as rule-utilitarian. It is institutional in its requirement of role holders and sanctions. A rule-utilitarian might claim that having everyone disposed to follow a particular rule (unless the particular consequences of the moment would be disastrous) would produce better results overall than having everyone disposed to set aside the rule whenever their assessments of relative benefits seemed to recommend violating the rule. It is such a conclusion that justifies institutionalizing many utilitarian rules into law and, hence, justifies recourse to institutional utilitarianism. But the institution can be designed to see to it that the rule gets followed very often, even when many people may not be motivated by utilitarian concern. The rule might also be institutionalized through organized educational efforts to lead people to have the relevant dispositions, although it would then be followed by individuals acting as rule‐ utilitarians, not as institutional roleholders or as subjects of institutional sanction. Still, when an individual acts, her act-

utilitarian assessment may trump either the sanctions of the institution or the dispositions of rule-utilitarianism. It is not easy to describe how this would be done for the ruleutilitarian disposition, which, as a motivation, is of a kind different from the normal utilitarian costs and benefits. Institutional utilitarian sanctions, however, are of a kind with the normal utilitarian consequences. Consider a narrowly focused institution for accomplishing a utilitarian purpose. Suppose we can show that democracy would work much better at representing the interests of the populace if we have 90-100 percent turnout instead of about 50 percent, as we typically have in the United States for presidential year elections (with even less in other election seasons). And, from the experience of other nations and perhaps from tests in various communities in our own nation, we can show that a small fine—perhaps $25—coupled with easier registration and voting would very likely increase turnouts to the desired level. Institutionalizing the system of fines and easier voting could then be utilitarian. It does not follow, however, that my voting on election day will be utilitarian. As always, it need not be true that my vote has any significant chance of making a difference. It can make a difference for a single-office election only in the closest of races in which the candidates receive equal votes or in which my candidate receives one less vote than another, so that my vote breaks or forces a tie. In a polity of several million voters, the odds of such closeness are vanishingly small (Hardin 1982: 59-61). Even if the value of my vote is thought to be in its register____________________ 4. I spell out a theory of institutional utilitarianism in Hardin 1988: chs. 2-4. -135ing a share of the public's interest, rather than its directly influencing the election, one more vote in a minority of millions will not likely be taken into account. Against my voting may be far more substantial interests at stake, such as the happiness of an invalid, my work on an important project with a tight deadline, or my unexpected need to be out of town on voting day. A high turnout might be good, but my turnout might nevertheless be bad. Unfortunately for me, it might not be utilitarian for the institutional officers who impose fines to have procedures for hearing such cases as mine. Procedures for exempting anyone could be far more costly than the fines that might occasionally be rescinded; and the availability of the procedures might induce more non-voting in the expectation of getting out of the fine. Hence, what it would be utilitarian for me to do on election day (not vote) would be punished by our utilitarian voting institution—and rightly so on the institution's utilitarian assessment (see further Hardin 1992b, 1993). The institution's utilitarian assessment would turn not only on the relative costs and benefits of my voting and not voting, but also on the costs of institutional action to ferret out and weigh those costs. Both the institution and I may be fully utilitarian in motivation, and yet the conclusions we reach about what is utilitarian seem in this case to be contrary. I

conclude, as a utilitarian, that I should not vote. The institution concludes, on utilitarian grounds, that I should be punished for not voting. Both are right for reason of different epistemologies. Act-utilitarianism and institutional utilitarianism are therefore distinctively different in some plausible applications. Actual institutional resolutions are typically plagued by problems of split motivation. Assume a curvilinear, nearly step-function provision that benefits an entire population by aggregating their individual contributions. It might be utilitarian to let some percentage of the relevant population free-ride on the efforts of others, because their additional contributions, while just as costly as anyone else's, are less valuable to the group. But if some free-ride, others may follow their example and tip the population into steeply declining provision. Hence, enforcement beyond efficient provision may be necessary to keep provision at a reasonable level. One could make a related argument about the general return on enforcement costs. Enforcements of, for example, the tax laws to the level of perfect compliance would cost radically more than enforcement to a somewhat lower level and we surely are right to opt for the lower level. But if that level is too permissive, noncompliance might multiply enough to unravel the system. There are two main arguments for the use of institutions to achieve desired outcomes, whether morally desired or otherwise. First, institutions can be used to overcome conflict or Sidgwick's dualism of practical reason (see Brink 1988a: 291-307)—the conflict between self and collective interest—not least because they can have coercive power or the power to impose a particular coordination. Second, they can be used to overcome epistemological and action problems of individuals, who may not know what relevant organizations know or cannot -136know it as efficiently and who are often not as effective in spontaneous cooperative action as they are in organized institutional action. In general, if all are utilitarian in their motivations and if there are no epistemological and action differences between individuals and institutions, there is no argument for institutions as pragmatically necessary devices for achieving utilitarian results. Under such conditions, institutional utilitarianism would not be a distinctive program. What determined act-utilitarians must be able to claim in defense of their position is that institutional utilitarianism is not a distinctive program, that institutional utilitarianism would not produce more optimific results than act-utilitarianism. Clearly this cannot be a conceptual claim. It is necessarily a claim about positive theories of causal connections between the motivating and coordinating powers of institutions and the actions of individuals under or in the face of those institutions. No sensible philosopher can claim to have such mastery of relevant positive theories as to conclude that institutional utilitarianism actually reduces to act-utilitarianism in a real society. The contrary claim that institutional utilitarianism and act-utilitarianism can diverge is prima facie compelling, as it may be in voting and tax compliance.

V. Sources of Knowledge Act-utilitarianism requires only knowledge that one can directly assess or guess, knowledge one already has or can gain. Rule-utilitarian knowledge must somehow be accepted from an external (authoritative?) source but, unlike other such knowledge, this must be thought to trump one's own knowledge. Otherwise, rule-utilitarianism is always trumped by act-utilitarianism when the two differ. In some accounts, rule-utilitarian knowledge is always supposed to trump act‐ utilitarian knowledge (Melden 1966: 184185). In institutional utilitarianism, the knowledge you need for acting in certain contexts is encapsulated in the directions handed out by some institution (for trivial example, do not make U-turns on certain streets). Your direct act-utilitarian knowledge may well be taken to trump the institutional-utilitarian knowledge. But that is not the whole story. The institution may still have good reason to act on institutional-utilitarian knowledge while you have good reason to act on act-utilitarian knowledge. The conflicts between institutional utilitarianism and act-utilitarianism and between ruleutilitarianism and act-utilitarianism are both conflicts of knowledge. But the first conflict can be resolved by attention to the different methodologies for institutions and for individuals to gain and apply knowledge. These methodologies can be readily and clearly articulated. In the case of the conflict between rule-utilitarianism and act-utilitarianism, the knowledge is all individual. It is true that there are different methodologies for gaining different bits of our knowledge. For example, I am much more reliable at gaining knowledge in -137the field of my specialization than in many other fields, and I go about it in very different ways, ranging from relying on a vague memory from a source whose credentials I may no longer be able to evaluate (maybe it was the Times, maybe it was gossip on the bus or subway) to relying only on a careful, systematic test of a particular claim. But there are no distinctively different methodologies for applying different bits of my knowledge to effect utilitarian results. For the individual there is a partial analog of the problem that institutional decision making may conflict with individual decision making. The individual may have to decide in advance what knowledge to have readily available when she later faces substantive decisions. The ex ante rationality of that prior decision then constrains what is plausibly rational action later. Because she comes ex ante to know x and not y, she may find that, on her best understanding, decision p is better than decision q when, with the opposite investment in knowledge, she would have found q better than p. Again, I may have conflicting bits of knowledge that come from different methodologies, such as direct assessment and a socially promulgated rule or norm. But this is merely a special case of the problem that my own epistemological constraints, the economics of my own knowledge, hamper me in my instant problem of decision. There may further be

a conflict between me and someone who judges me, even from the legal system. But typically there is only one actor involved in my choice of action. I decide and I act. This issue has been addressed only indirectly in most rule-utilitarian argument. Brandt supposes that an ideal utilitarian agent would have dispositions of varying strengths associated with various rules (Brandt 1988a: 347-350, 352-360; 1989b: 91, 96f). For example, the disposition against killing would be far stronger than that against lying. Here again, the present self chooses under constraints from the past self, who determined the dispositions. The methodology for the determination of these dispositions may be quite different from that for estimating the consequences of a present action. If the dispositions are determined institutionally, as through using the educational system and perhaps the civil and criminal justice system to inculcate rule-utilitarian dispositions, then our problem is merely an instance of the differences between individual and institutional epistemologies. If the determination of the dispositions is an individual-level task for the rule-utilitarian, we must wonder how that determination differs from other knowledge the person has. In part, a strong disposition would be a device for stiffening one's resolve or overcoming weakness of will. Indeed, if adopting dispositions is to serve the task of stiffening one's resolve to do right, the implication is that there is a right that one could have determined readily enough if only one's psychological problems were not in the way. Hence, if the device of adopting strong dispositions is part of a distinctively different utilitarian theory, then the dispositions must be more than a solution to such psychoogical problems. It must be an epistemological device for reaching better judg -138ments of what should be done, not merely for chivvying oneself to do what one knows is right. As an epistemological device, the adoption of dispositions lacks justification. In general, we may take for granted a fairly high level of understanding of the problems of ordinary individual knowledge and of organizational knowledge. The more difficult and much less discussed issue is how an individual can come to know the truth of general empirical rules. This latter concern is central to rule-utilitarianism in all its variants from G. E. Moore's social-evolutionary theory to Brandt's dispositional theory. Let us turn to that issue.

VI. Our Knowledge of Empirical Rules J. O. Urmson supposes Mill thought it possible to establish general or aggregatelevel principles independently of prior assessment of the rightness of actions in relevant particular instances. If this is supposed to be an account of a Millian individual moral theory, as Urmson seems to think it is, it is a remarkably implausible reading of Mill. Mill reputedly "makes it clear that right and wrong are derived from moral rules" (Urmson 1953 [1971]: 171). One might have thought the consequentialist Mill held the opposite view that following a rule is right or wrong only insofar as doing so produces utilitarian results or fails to produce them. And against Urmson's reading, what Mill makes clear is that, "Where there exists a genuine and strong desire to do that which is

most for the happiness of all, general rules are merely aids to prudence, in the choice of means; not peremptory obligations" (Mill 1832-3 [1984: 39]). In the passages of Mill from which Urmson works, what Mill actually argued is merely that one can choose the utilitarian action in manifold contexts without first having to work out de novo what the principle of that action should be or, more directly perhaps, what that action should be. Mill's example is an analog from prudential choice. The sailor does not set out to sea planning to figure out the positions of the stars de novo but works from lessons learned in the past, plausibly learned by others and merely passed on to the sailor, for example, in the form of charts of the skies, maps of the seas, and an astrolabe. 5. In the moment of instant decision whether a storm has carried my ship off course, I turn to the charts and find out quickly—I do not start by deriving Kepler's laws. Kepler's laws, the sky chart, the map of the seas, and the astrolabe are all "intermediate generalizations" whose use simplifies my task of determining where I am. ____________________ 5. It is because he is generally concerned with such authoritative knowledge that Mill had great interest in the problem of moral authority. On this latter interest, see Friedman 1968. -139These intermediate generalizations are only that. They are themselves aggregated out of masses of specific data, including Tycho Brahe's data on the stars and numerous explorers' maps of the seas. I rely on them not because I have previously derived them but because I have experienced their effectiveness in my previous navigations. In comparable ways, Mill supposed I might come to believe in the efficacy of following various moral rules in contexts in which I might be ill-placed to deduce the best action from the simple principle of optimizing welfare. In some degree, I think the analogy is not fully convincing. I will take up this point further below in a discussion of the problem of assessing group benefits from collective following of a rule. But I will use Mill's discussion as an account of how we can make moral recourse to rules. The most difficult issue for many rule-utilitarians is that this account is not what they have in mind. Rather, they seem to suppose that we can somehow have knowledge that comes through direct apprehension of the rule or generalization. The epistemology of rule-utilitarianism seems ill defined. If there is a way to know the utility of following a general rule independently of knowing how to judge particular cases, Lyons's move, discussed above in § III, to reduce rule‐ utilitarianism to actutilitarianism fails. If the only way to judge the utility of following a rule is by generalizing from cases, Lyons's criticism of rule-utilitarianism goes through. In the literature there seem to be four main devices or methodologies for knowing the utility of rule following without derivation from a survey of cases. These are the invocation of social norms, appeal to authority, claims of direct apprehension or intuition, and rationalist deduction. If these cannot stand scrutiny and no alternative can be found, rule-

utilitarianism is incoherent or reducible merely to act-utilitarianism or to actutilitarianism coupled with institutional utilitarianism.

VII. Knowledge via Social Norms Suppose our rules take the form of norms that have evolved socially. In the social evolution of norms, the role of testing is clear in many prudential matters. Lack of testing is the fundamental problem in the spontaneous discovery of good moral norms, even if these are primarily for collective benefit. The issue is the sheer difficulty of carrying out tests of large numbers of trial norms. What the market or even a partial market can do for selection of what serves individual interests cannot be done for collective interests, especially for collective interests at or near the whole-society level. Selecting norms for dyadic and very small‐ number interactions may be relatively easy because such norms may be very nearly congruent with enlightened self interest. This will often not be true for larger-number interactions (unless these are norms for mere coordination on actions or outcomes that all prefer if all are coordinated). -140If there is a motivational problem, as in Sidgwick's dualism of practical reason, we may never succeed in testing a collective-action norm because we may never have enough people following it to know whether it is generally beneficial. There are often great personal incentives to try to discover and to make use of scientific or useful truths, but typically there are no comparable direct personal incentives for discovering or making use of moral principles. The worst problem for an argument that social norms must be utilitarian, however, is that we can easily find examples of socially evolved norms that are not. Consider a woefully sad case. St. Kilda, an island lying 110 miles west of Scotland, was once inhabited by a people with little outside contact, whose population and society collapsed over the course of the nineteenth century. Apparently, a chief reason for the collapse was an odd norm for infant care. It is believed that a mixture of fulmar oil and dung was spread on the wound where the umbilical cord was cut loose. The infants commonly died of tetanus soon afterwards. The first known tetanus death was in 1758, the last in 1891. Around the middle of the nineteenth century, eight of every ten children died of tetanus. By the time this perverse pragmatic norm was understood and antiseptic practices were introduced, the population could not recover from the loss of its children (Maclean 1980: 121-124). The last remnant of about three dozen people left the island in 1930 (Maclean: 142). One might suppose such a norm cannot survive because the people, as in the case of the St. Kildans, finally cannot survive it. But a society with many norms is unlikely to depend for its survival on the utilitarian nature of every one of these. Hence, many could be nonutilitarian in even the most hopeful evolutionary vision. Social evolution is no guarantor of utilitarianism.

VIII. Knowledge from Authority If there were an authority for the rightness of our rules, rule-utilitarianism would be as unassailable as that authority. A quick look around philosophical circles suggests there is no unassailable intellectual authority. Indeed, the most assertive moral theorists are often also the most embarrassing. Only a true believer in a particular religion could be convinced there is an unassailable religious authority. There seem to be no other plausible candidates. But suppose there were an authority or authorities, as there are in the sciences and in practical pursuits such as medicine, engineering, law, the performing arts, perhaps sports, business, and many others. What does it mean for someone to be an authority in one of these endeavors? Knowledge from authority raises the general question of knowledge of rules. If I do not know what is true of some matter, how can I judge that someone else does know it? Sometimes this question admits of an easy answer. Perhaps I can -141simply look to the results of following the authority's judgments. This may work splendidly in, say, coaching ballerinas or pianists, repairing hearts, or building bridges. Each effort of the authority produces a result that can be checked out, even by comparison with the results achieved by other putative authorities. Unfortunately, following a moral authority's advice may have consequential effects, but these will not be subject to general appraisal in the same way. Kant would tell me not to lie to the assassin at my door; Mill would surely tell me to lie, as Constant would. Kant would insist that testing by results would be wrong—results do not matter. What we have to accept or question is the nature and force of the argument that connects the advice to some kind of moral principle. After a while, we might test our own, perhaps tedious and too late, judgments against those of a reputed moral sage and discover that she always gets it right much faster than we do, and we might slowly come to accept her as a moral authority therefore. But this just is to say we recognize that the act-utilitarian calculation invariably fits her instant insight. What we need here is not merely someone who reaches act-utilitarian judgments faster and more infallibly than we do. We need someone who can enunciate principles for action that transcend calculation, principles that might sometimes even run afoul of actutilitarian recommendations. And we would need our authority to be credible in trumping our own calculations in these latter cases, to be credible when speaking ex cathedra against our own best understanding. The very idea sounds today more like a play by Eugene Ionesco than a plausible story. The form that a rule-utilitarianism based on authority must take if it is to seem at all sensible is roughly that of Richard Brandt or of R. M. Hare (Brandt 1979: 286-300, 229233). These theories propose a system of morality rather like a system of law, except that

enforcement is less important than psychological internalization. The authorities are various teachers, including parents, who inculcate moral rules in children. This enables the children to grow up to be generally utilitarian in their behavior not because they calculate well but because they follow rules that achieve, on average, fairly good approximations to greatest overall welfare. Here, rules are a device to economize on decision making and moral judgment, and inculcation is a device to bridge Sidgwick's dualism of practical reason. In Brandt's system, the authorities derive their rules from consideration of cases. As in Hare's system, there are authorities merely as a matter of division of labor to allow specialization. Brandt's system is not inconsistent with act‐ utilitarianism except insofar as institutional utilitarianism is also. The inconsistency comes in through the facts that the system may be able to secure better results overall than direct application of the utility principle by all individuals would and that it may reach conclusions for typical cases that differ from what -142an individual would reach from the individual's own understanding of the facts. Finally, this system is not rule-utilitarian—its rules are merely devices for getting as closely as practicably possible to what act-utilitarian calculus recommends.

IX. Knowledge by Direct Apprehension A sufficient reason for the lack of unassailable moral authorities is merely that no one has a method for generalizing the truth about a large collection of individual instances in the absence of a method for establishing the truth about any of the instances. Perhaps there can be no such method in what is essentially an empirical matter. In logic or mathematics, one might give a possibility or impossibility proof over a general class of relations even without being able to prove relevant results about specific relations. But all of the arguments typically offered in defense of the rightness of various rule-utilitarian rules are at best claims for something like a Condorcet jury theorem of the truth: As proportionally more of us agree on a rule, the more confident we can be in its truth (Condorcet 1785 [1976]). Unfortunately, the empirical assumptions of Condorcet's theorem may typically not be met. An alternative vision might suppose we have direct intuitions about utilitarian rules even when we might have difficulty seeing their fit with actual causal relations. Intuitions of various kinds are clearly necessary for utilitarianism, as for virtually all moral theories. Indeed, intuitions of some kind are necessary for mathematics, for the sciences, for common sense. But intuitions in such fields are typically held subject to testing or to weighing against alternatives. Once rock-solid intuitions about the geometry of space have even been chiseled down by Einstein's theory of relativity and later advances in physics. To suppose that one could have a correct intuition about the utility of always telling the truth seems utterly preposterous. One might have systematic intuitions, such as about fairness or welfare or of the scaffolding of Kant's complex moral theory, from

which a relatively focused substantive conclusion could be deduced. But a direct substantive intuition about lying or any other ordinary action cannot be compelling. In any field but ethics, holding such intuitions with deep commitment despite lack of theoretical or empirical warrant would be seen as silly, as an intellectual disqualification. As Brandt says, "it is puzzling why an intuition—a normative conviction—should be supposed to be a test of anything." (1979: 21) If direct apprehension is what various rule-utilitarians have in mind, one may hope they unpack the notion to show how it is something more sound than moral intuitionism. If it is the latter, one may hope they will show how the more fundamental intuition of utilitarianism—that the goodness of an action is a function of the welfare effects it has— fits the odd substantive intuition that says do x when direct assessment of welfare considerations says do y. -143-

X. Knowledge by Rationalist Deduction Clearly, one of the problems of rule-utilitarianism is that its rules cannot come from an unquestionably authoritative source. Indeed, it is commonly framed as though its rules were derivable from reason by each individual, and this is the form in which ruleutilitarian argument seems most convincing. For example, Brad Hooker defines ruleconsequentialism as "the theory that an act is morally right if and only if it is allowed by the set of rules and corresponding dispositions the having of which by everyone would bring about the best consequences considered impartially" (Hooker 1991: 269; Brandt 1979: 300-305). Here reason might be sufficient to derive the rules in the following sense. Reason could take into account basic facts of human capacities that must be mobilized or countered to affect outcomes, but not accidental facts of history. Now suppose not everyone is acting as Hooker's ideal rules specify. Perhaps it would then be better for me also to violate the rules. Reason might commend coordinating on some action rather than another, when in fact almost everyone is coordinating on the other. It might then be harmful for me to follow reason instead of the practice of my fellow citizens. For a trivially clear example, suppose reason dictates that we all drive on the right when virtually everyone is driving on the left. (One might suppose this was the condition in Sweden shortly before Sweden changed its rule from driving on the left to driving on the right in 1967 (Hardin 1988: 51-53).) Anyone who followed reason here would act murderously and suicidally. The driving example is a case in which there are at least two possible choices for our general coordination, one of which is genuinely superior to every other, although any is radically superior to no coordination. Hooker's principle seems to require me to follow the rule of the superior coordination. But an inferior coordination may have been "chosen" already by historical happenstance, and my following the rule of the superior coordination might cause great harm.

Using reason alone, and ignoring the peculiar facts of the matter in my context, to decide how I should act has the perplexing problems of the generalization argument (Hardin 1988: 65-68; Hardin 1980). In that argument, we deduce what would be a good rule to follow if it were, as with Hooker's rules, followed by everyone. What we actually need in real life are rules for how to behave in the light of the behavior of others. We should therefore add to reason assessments of expectable consequences of our actions in context. But then rationalist, deductive theory collapses, as Kant may have sensed (Kant 1785). 6. It is no longer rule ____________________ 6. Pp. 92-4 in Paton's translation; pp. 425-27 in the Prussian Academy edition. Kant's resolution of this dilemma was to ignore the contingent and to rely wholly on reason, even if that led to murderous results, as in insisting on telling the truth no matter what the contingent considerations. -144driven; it is act or calculus driven. Or, rather, its rules are always subject to trumping by act-utilitarian calculation. Hence, its rules reduce to something like rules of thumb. The examples used for elucidating this problem are often chosen as cases in which one might think a deontological principle and a rule-utilitarian principle agree but actutilitarian calculation goes against both. In the face of partial or total noncompliance with the rule-utilitarian principle, an act-utilitarian might be supposed to act against it. For example, in the face of racial discrimination, one might think it better to participate in the system than to flout it and thereby cause great harm in a disturbance. Brandt argues that the story in this particular example is too amateurishly told if the act-utilitarian criticism of rule‐ utilitarianism seems to go through. One could flout the law and practice of discrimination in a constructive way, and that is, he says, what rule-utilitarianism must recommend (Brandt 1988a: 358-359). But this salvation suggests a slip into the argument of Lyons that, if the rules are to be adequately specified, they will finally reduce to actutilitarian injunctions. In one sense it is odd that such examples as the botched coordinations of the driving convention pose a problem for the rationalist deduction of utilitarian rules. They are, after all, one of the easiest of all classes of interactions to resolve, at least in principle. (The Swedes spent a substantial sum to change their driving convention. The British might consider a comparable change prohibitively expensive.) But in another sense, it should be clear why they might be hard for the rationalist rule-utilitarian. They depend on paying attention to the idiosyncrasies of the actual world in their resolution. One convention might easily be seen as inferior to another in its full empirical setting even though neither had any apparent natural superiority. The rationalist can escape the test by simply reframing the rule for coordinations as, in the abstract, to follow the extant convention. But this is not a happy escape, because it elevates miserable conventions, such as those of racial and sexual subordination, to moral status. Kant's rigid exclusion of empirical considerations from our attention while reasoning theoretically about ethics would be

utterly silly for a utilitarian. But bringing them together threatens the rationalist ruleutilitarian effort.

XI. Concluding Remarks Rule-utilitarianism and institutional utilitarianism are similar in their concern with aggregate, standardized problems that should be handled with standardized devices— such as rules or institutions. A good utilitarian reason for resort to rules or institutions is that these may reduce the costs of decision and error. What may often be even more important is that they can coordinate all on relevantly similar or related actions. In actual practice, however, the most important -145function of institutions may be in overcoming partial compliance. I may well think it better on the whole to have a particular aggregate resolution of some issue, but then I may also be tempted to free-ride on the efforts of others, perhaps only because others free-ride. The best world, then, would be one in which I and those others are sanctioned if we do not comply. Partial compliance is not merely a problem for rationalist (or any other) rule‐ utilitarianism, it is one of the fundamental reasons for institutional utilitarianism. For example, partial compliance with the duty to vote can be overcome with institutional devices even if it could not be overcome merely by individual commitment. For interactions not governed by institutions, one might still think rule-utilitarianism valuable. But once we leave aside the cases governed by institutions, including Brandt's education of the young to follow simple rules, there may be little scope left for ruleutilitarianism. Indeed, little of the traditional concern of moral theorists remains. There is still the very little bit of promise‐ keeping that is not sanctioned by contract and law nor otherwise motivated by self interest. 7. And there is beneficence in many and varied contexts that somehow escape institutional management (Hooker 1990; 1991; Hardin 1988: 38-41, 53-59; 1990; 1991). The variants of utilitarianism must be distinguishable in their implications if they are to be of any interest. There is an ample literature asserting the difference between act- and rule-utilitarianism. Typically, the difference turns on implicit claims for a distinctively different epistemology, claims that are not well articulated or grounded and that seem mistaken. Otherwise, rule-utilitarianism seems to be merely a psychologically interpreted, more refined revision of act-utilitarianism. There is a clear distinction between act- and institutional utilitarianism in their different locuses for reaching judgments. But that difference should be functionally acceptable to the act-utilitarian, who, after considering the sociology and epistemology of certain matters, must prefer to turn those matters over to utilitarian institutions. Doing so, of course, risks having the institutions trump one's judgment in some cases. This is a problem that is empirically inescapable. It is in the nature of the world in which we must act. It is kin to the criminal justice problem of choosing a level of enforcement in the confident knowledge that stiffer enforcement will

catch both more miscreants and more innocents and that weaker enforcement will catch both fewer miscreants and fewer innocents. The two kinds of error—catching too few miscreants and catching too many innocents—come in a package. The conflict between institutional and individual judgments is constitutional. Institutions and organizations are of value in large part because they have epistemologies that differ from the epistemologies of individuals. The supposed ____________________ 7. Almost all promise keeping may be adequately handled by self interest without need for an overriding moral motivation (Hardin 1988: 59-65). -146conflict between rule-utilitarian and act-utilitarian principles does not appear to be constitutional in a similar way. None of the forms rule-utilitarianism takes can reasonably keep rule-utilitarian judgments from being subjected to the correcting test of actutilitarian calculation by the very individual who applies the rule. Therefore, without further articulation of the odd epistemology of knowing rules for cases that one cannot directly evaluate, simple rule-utilitarianism is not a coherent program. Because of the conflict between institutional and individual judgments, two seemingly contrary conclusions follow. First, institutions should often be given the power to influence behavior to produce utilitarian outcomes. Second, individuals should continue to act from their own best understanding, including consideration of institutional constraints. This will often (although presumably not very often) mean that utilitarian individuals ought to do what utilitarian institutions ought to attempt to prevent them from doing, perhaps by fining or otherwise threatening to punish them (Hardin 1993). This is not simply a moral conclusion, because both the institutions and the individuals in this conflict may be motivated by the same moral concern, which is utilitarian welfare. It is an implication of differences in the epistemological constraints on institutional and individual knowledge. If one had ideal mastery of all relevant facts at issue in such a conflict, one might be able to conclude that either the individual or the institution was morally misguided. But often such a conclusion could not be grounded in an account of how the one or the other party should have known the facts better. And if it is not true that they should have known the facts better, then it cannot be true that they should have acted "better". For many cases we cannot say that certain individual actions are wrong; we can merely say that it can be right for the relevant institution to say or rule that those actions are wrong. Moreover, the utilitarian individual who gets blocked or sanctioned by a utilitarian institution may, in a cool moment, hold it right that the institution behaves that way. -147-

10 The Problem of Group Egoism Gregory Kavka A certain strategy for grounding morality is currently popular among major moral theorists, both contractarian and utilitarian. One first offers an account of the rational pursuit of an agent's own good, conceived of as happiness or desire‐ satisfaction. Then one defends certain (systems of) moral principles as those such rational, personal-goodpursuing agents would choose to live under together. This strategy is a particularly ambitious version of what I have elsewhere called "the reconciliation project"—the attempt to show morality to be consistent with rational prudence (Kavka 1984). It is ambitious in attempting to derive moral constraints from principles of individual rationality. The strategy has certain obvious attractions. It offers the hope of accommodating the "alienated" individual to the constraints imposed by morality, by showing how these constraints operate to promote her advantage. And it promises to lend greater definitiveness and accuracy to moral theorizing and argumentation by recasting some of the messy problems of ethics in the more precise language of rational choice theory. This strategy also faces some well-known problems. Its various practitioners do not agree among themselves either about the rational pursuit of an individual's good, or about the content of the moral constraints that rational personal‐ good-pursuing agents would choose to live under together. More importantly, for my purposes, there is the nagging problem of why such rational agents should comply with socially beneficial moral constraints if they think they could personally benefit from non-compliance. A variety of answers have been offered to this question, going back at least to Hobbes's response to "the Foole [who] hath sayd in his heart, there is no such thing as justice", and continuing up to David Gauthier's contemporary account of "constrained maximization" (Hobbes 1651[1968]: 203; Gauthier 1986: ch. 6). But all these answers appeal, in one form or another, to reciprocity: compliance with moral constraints benefits you because it facilitates the cooperation (and like compliance) of others, and you—like everyone else— need such cooperation to get along in the world. -149This suggests a further problem, which is frequently noted in passing, but seldom dealt with at length: the problem of group egoism, as I call it. Everyone does need the cooperation of others to get by in this world. But not all others. Only enough suitably placed others. Hence, if the only rational ground of compliance with moral constraints is reciprocity, the scope of moral protection would seem to extend only to cover potential reciprocators. More specifically, while all individuals may need cooperation from other individuals in their group, members of powerful groups may not need the cooperation of the members of weak groups. The weak may have nothing to offer, or what they have to offer may be gotten more cheaply by coercion than by cooperation.

The problem of group egoism has led some theorists interested in contractarian theories embodying the above-described strategy to despair of the promise of such theories. Thus, Jean Hampton writes (1991: 48-49) that such a "theory gives us no reason to respect those with whom we have no need of cooperating.... And I would argue that this shows that [it] fails in a very serious way to capture the nature of morality." And even Gauthier at one time (1977: 164) seemed to think that this problem (and related ones) rendered contractarianism "from a practical point of view, bankrupt". Recently, however, attempts have been made to address aspects of the group egoism problem within a contractarian framework, most notably by Gauthier himself. 1. The group egoism problem also affects certain utilitarian theories, particularly that presented by Richard Brandt in his influential book A Theory of the Good and the Right (1979). 2. In this paper, I show how the problem creates an unfilled gap in the center of Brandt's theory, then go on to discuss the resourceful—but ultimately failed—attempt by another utilitarian, Peter Singer, to bridge that gap. I then evaluate Gauthier's attempt to solve the group egoism problem within the context of his contractarian moral theory, and close with some reflections upon the problem's ultimate implications for moral theory.

I. Brandt's Gap On the surface, Brandt's strategy in A Theory of the Good and the Right seems straightforward. In part I of that book, he gives an account of a rational person's good. Roughly, it amounts to maximal satisfaction of those of the person's desires which are themselves rational, in the sense that each would survive a process of cognitive psychotherapy in which the person is made vividly aware of all rele____________________ 1. See Gauthier 1986: ch. 9; and Morris 1991: 76-95. Their positions are discussed in § III below. See also Kavka 1986: 439-46, and Kavka 1984. 2. Brandt is not the only prominent utilitarian to endorse the "rational contractor" strategy. See Hare 1972b: 167-73. -150vant information relating to the desire. In part II, he argues that people rationally pursuing their good would choose, or tend to support, over other moral codes for the society they expected to live in, a rule-utilitarian moral code. But his method of arguing for ruleutilitarianism reveals a substantial gap between what the account of rationality in part I would allow him to establish and the utilitarian conclusions he reaches in part II. Brandt contends that, on his account, benevolent desires can be rational. But he acknowledges (1979: 143-45) that a person is not necessarily being irrational if he lacks benevolent desires altogether. So to establish the rationality of choosing a rule-utilitarian moral code to live under, he considers the cases of the perfectly benevolent person and the selfish person separately, reasonably assuming that imperfectly benevolent people

will fall between these two extremes in the "generosity" of the code they would support (1979: 222). It is the rational selfish people, of course, who pose the hard problem. Brandt argues on Hobbesian grounds that even they will want—for their own protection—some minimal moral code of prohibitions on assault, murder, theft, and cheating. And, if they look to the long-run, they will tend to endorse egalitarian measures of the sort favored by utilitarians, in order to promote social stability. 3. Yet, selfish people who are specially advantaged and think the relevant long-run is beyond their lifetimes, may rationally demur (Brandt 1979: 218-21). More importantly, there may be whole classes of sentient beings that rational selfish people will wish to exclude from moral consideration. Brandt writes, In one major respect rational benevolent persons will support a moral system different from that which rational selfish persons will support. For the rational selfish person is interested in parts of a moral code which benefit others, only to the extent to which benefit in return is provided for himself. So he supports a system calling on him to submit to moral obligation to others only in so far as those others can help or hinder him. He wants only reciprocal ties. The benevolent man, in contrast, wants the benefits of the moral system extended to those who can do him neither harm nor good: to future generations, to animals, to the mentally defective, small children, and possibly fetuses. (Brandt 1979: 221) If Brandt had only added citizens of poor nations and powerless members of domestic underclasses to his list, this would constitute a flawless statement of the problem of group egoism. How does Brandt propose to deal with this-problem? Essentially, he leaves it unaddressed. With characteristic honesty, he admits that moral philosophy simply cannot demonstrate the rationality, for everyone, of being moral. But he claims (1979: 242) that this is all right because all that is needed for tolerable ____________________ 3. In this too, he is adapting and extending a line of argument presented by Hobbes. See Kavka 1988: 91-93. -151social life is unanimous agreement on some core principles of morality—essentially the Hobbesian minimal code prohibiting assault, theft, and so on. He says it is easy to show that all rational persons would agree on these principles, but he here simply ignores the problem of who is protected by these principles. Social life may not be tolerable for beings falling outside the protection of these principles, and Brandt has acknowledged— in the passage quoted above—that rational persons will not agree about who falls within the scope of (minimal) morality's protection.

Thus, while Brandt (1979: 222) may be correct that his arguments are strong enough to "warrant raising" the question of what sort of moral code would maximize expected utility, they leave a large gap in his apparent attempt to derive an account of the morally right from his theory of a rational agent's good. In particular, they do not help us answer the questions of whose utility is to be maximized, or why a rational person should favor maximizing utility for a group larger than needed to secure the cooperation and reciprocation she needs. By failing to address this group egoism problem, Brandt has left his rule‐ utilitarianism without proper foundations.

II. Singer's Bridge Though Brandt fails to deal with the group egoism problem, despite his recognition of it, the same cannot be said of another contemporary utilitarian theorist, Peter Singer. In his insightful book, The Expanding Circle: Ethics and Sociobiology (1981), Singer argues at length that reason operates on natural altruistic sentiments so as to expand the scope of moral concerns and principles to cover all sentient creatures. If Singer's argument is successful, it might bridge the gap left at the center of Brandt's theory, by showing how reason and nature work in concert to eliminate group egoism. Let us, therefore, look at the line of argument Singer presents. Sociobiology arose out of a puzzle in Darwinian theory: if natural selection favors organisms constructed and behaving so as to increase their own chances of surviving and reproducing, why are there so many examples of altruistic behavior in the animal kingdom, in which one member of a species takes risks or uses resources to help other members? Singer explains three mechanisms that socio‐ biologists have posited to solve this puzzle, and discusses some anthropological evidence to support the claim that each mechanism has had a role to play in the development of altruism in humans. Kin selection is the idea that genes that predispose the organisms containing them to act "altruistically" toward organisms that are relatives (and hence likely to carry the same gene) may thereby increase their representation in the gene pool over time. The second mechanism, reciprocal altruism, involves organisms' tendencies to benefit others in -152"expectation" of receiving benefit from these others in return. The hypothesis of group selection says that groups whose members behave altruistically toward one another are less likely to go extinct than are groups composed of non-altruists: hence surviving groups are made up of altruists. 4. As Singer recognizes, while these mechanisms together provide a plausible explanation of why human beings possess altruistic sentiments and actiondispositions, each suggests that altruistic concerns will be of strictly limited scope (Singer 1981: 15, 17, 20-21). 5. Kin selection explains altruism to kin, reciprocal altruism to potential reciprocators, and group selection to fellow group members. Unless there are further mechanisms for the creation or redirection of altruism, the specter of group egoism looms large on the horizon.

Singer's mechanism is reason, in its practical applications. Social life with others requires us to give reasoned justifications for our conduct: if we want our fellows to accept what we do, we must be prepared to offer general impartial justifications that reflect their interests as well as our own. But "reasoning is inherently expansionist. It seeks universal application." (Singer 1981: 99) Thus, while practical reason's initial function may have been to help people cooperate within limited groups, its imperialistic logic allows it to gradually stretch the scope of our altruistic concerns from kin and tribesmen to all humanity and even to animals. Ultimately, this logic points toward impartial concern for the interests of all sentient creatures, and indeed, there is a definite historical trend in that direction (though admittedly more so with respect to professed principles than actual conduct). There is an interesting lacuna in Singer's story of the naturalistic basis of generalized, reason-guided, altruism. This concerns the evolutionary origins of practical reason itself. Two distinct accounts are possible, and Singer hints at each in places, but never explicitly develops either one. 6. The first account treats practical reason and adherence to norms as direct objects of evolutionary forces. As developed by Allan Gibbard (1990: 23-30), this account suggests that the propensity to submit to social norms and public standards of reason facilitates coordination in a species such as ours, and hence provided a substantial survival advantage to our early ancestors. An alternative account would treat practical reason as a side-effect of the selective pressures operating to produce scientific or technical reason aimed at predicting and manipulating the non-human world. Consider the useful general proto-scientific maxim "treat things the same unless ____________________ 4. Group selection is controversial among evolutionary biologists because of a free-rider problem: egoist mutants or invaders in an altruistic group will do better than altruists and will turn it into another group of egoists before it is able to outlast other groups. For various informed views on group selection, see Brandon and Burian 1984. 5. For empirical confirmation that the scope of altruistic tendencies tends to vary inversely with their strength, see Fiske 1991: 60-61. 6. Singer 1981: 91-92 hints vaguely at the first account; p. 116 hints vaguely at the second. -153there is a relevant difference between them". It has obvious survival advantages as a rule of methodological practice, and—as a principle of interpersonal inquiry and discussion— it promotes the growth of knowledge. But if norms of technical reasoning took this form due to biological or cultural evolution, we would be predisposed to apply similar norms to interpersonal conduct. It is easy, for example, to see how ethical principles like the formal principle of equality ("Treat people the same unless there is a relevant difference between them"), or meta‐ ethical norms like the universalizability criterion would seem natural to us under these circumstances. Indeed, the distinguished Chinese physicist, Fang Lizhi makes the connection explicit:

In my field of modern cosmology, the first principle is called the "Cosmological Principle". It says that the universe has no center, that it has the same properties throughout. Every place in the universe has, in this sense, equal rights. How can the human race that has evolved in a universe of such fundamental equality, fail to strive for a society without violence and terror? And how can we not seek to try to build a world in which the rights due every human being from birth are respected? (Lizhi 1989: 44) Whichever of these two accounts seems more plausible, their availability strengthens Singer's argument by showing that practical reason, like our altruistic sentiments, may be viewed as a natural product of our biological history and evolution. If the impartial promotion of the interests of all sentient beings does in fact follow from the conjunction of natural practical reason and natural altruistic sentiments, Brandt's gap has been bridged, the rule egoism problem is solved, and the utilitarians are home free. But does it follow? Singer's argument that it does is surprisingly thin. He proposes "taking the element of disinterestedness inherent in the idea of justifying one's conduct to society as a whole, and extending this into the principle that to be ethical, a decision must give equal weight to the interests of all" (1981: 100). His argument for doing so is that the main alternative rational bases for ethics that he considers— egoism and objective value accounts—do not hold water. Utilitarianism alone remains as a rationally defensible basis for morality. As Singer puts it, "If moral rules are not to be recommended to the group on the grounds of their good consequences for the group, on what basis are they to be recommended?" (1981: 108) There are several major flaws in this central argument of Singer's. First, even if it is acknowledged that rational morality embodies some fundamental notion of disinterestedness or impartiality, it is a large unjustified leap to the substantive interpretation of disinterestedness as assigning equal weight to party's interests. Equal rights, equal liberties, equal opportunities, equal resources, reward proportioned to merit, and reward proportioned to contribution, are only a few of the -154substantive moral conceptions that might reasonably be defended on disinterested or impartial grounds. Utilitarianism thus expresses a particular substantive interpretation of disinterestedness, one that is not inherent in the notion of principles that can be rationally defended to other members of society. Second, this argument never really addresses the crucial issue of scope. If rational justification to a group can find no basis other than appeal to the interests of the group, as Singer says, this implies nothing about how large and inclusive the relevant group is. Singer himself favors extending the equal consideration of interests principle to cover all sentient beings, including animals (1981: 120-24). But non-human animals are not creatures we have to, or even can, "address" in order to justify our conduct. More

generally, since the public justificatory function of practical reason at most implies the necessity of impartial concern for the relevant public without determining who that relevant public includes, it provides no solution to the group egoism problem. Singer has therefore not established, from the nature of reason alone, that the circle of our natural altruistic concerns must expand to eventually encompass all persons, much less all sentient beings. But is he not right that there is some tendency for reason to push us in that direction, and, indeed has there not been an historical tendency for the scope of concern to expand? Perhaps, but this hardly suffices to solve the group egoism problem. For past expansions may be at least partly explained in terms of new possibilities for reciprocation opened up by new knowledge, technology, and social conditions. As the possibilities of reciprocity expand, people come to treat as objects of moral concern some whom they had previously ignored. But those remaining outside the bounds of reciprocity —e.g., animals—are still in jeopardy. More importantly, even if reason provides some psychological impetus to expand the circle of concern, there will be psychological processes pulling in the other direction that may impose limits on that expansion. As noted earlier, all of the sociobiological explanations of altruism imply its natural scope will be limited. Further, certain cognitive processes reinforce the tribalistic tendencies associated with altruism of limited scope: selective attention to information (good information about insiders and bad information about outsiders are more salient), wishful thinking (overvaluing yourself, your projects, and your associates), and plain old rationalizations in support of selfish interests. 7. These psychological tendencies might exert forces that impose a limit on the scope of ____________________ 7. On the power of interest to influence social judgment, Thomas Hobbes writes (1651: ch. 11 [1968: 166]): "For I doubt not, but if it had been a thing contrary to any man's right of dominion, That the three Angles of a Triangle should be equall to two Angles of a Square, that doctrine should have been, if not disputed, yet by the burning of all books of Geometry, suppressed, as farre as he whom it concerned was able." -155people's moral concerns that falls far short of equal concern for the interests of all sentient creatures. 8. Singer acknowledges that there are such counter tendencies that severely inhibit the effectiveness of reason's call to us to act on principles of equal concern for the interests of all. Yet he believes that our abhorrence of cognitive dissonance and the discomforts of living a life of systematic hypocrisy gives us good reasons to fight and overcome these tendencies (Singer 1981:142-45). But cognitive dissonance could also be avoided by adopting a substantive view of rational impartiality that is less demanding than Singer's exacting utilitarian standard. And systematic hypocrisy is not a necessary strategy for getting along in the case of group egoists: as nationalists, tribalists, or patriots, they can readily expect to win the approval, praise, and affection of those around them by openly

favoring the interests of their group and ignoring or denigrating the interests of outsiders. 9.

In the end then, Singer's argument will not allow us to bridge the gap in Brandt's justification of utilitarianism, because it cannot provide a non‐ question-begging solution to the group egoism problem. Of course, if we identify practical reason with the substantive principle of equal concern for the interests of all affected, we can obtain a rational "justification" for utilitarianism. But this simply bypasses Brandt's (and others') strategy of trying to justify a moral system by showing it would be chosen, or supported, by persons concerned to promote their own good. It demonstrates only that concern for the good of all implies concern for the good of all. And rather than determining the scope of the "all" in a manner that relates it to the good of the individuals asked to comply, the equal concern principle simply asserts that any "affectable" entity qualifies. In the end, no reason is given why rational agents should eschew group egoism in favor of morality, save for the observation that a substantive utilitarian interpretation of practical reason requires this.

III. Gauthier's Piecemeal Solution If utilitarians like Brandt and Singer have failed to solve the group egoism problem, contractarians who seek to derive morality from rational individuals' pur____________________ 8. In the end, Singer (1981: 156-64) seems to recognize this, but tries to reconcile limited partiality with what utilitarianism requires by pointing out the utilitarian advantages of rule-following. He gives no reason, however, for supposing that the limits on impartiality due to our natures will coincide, even roughly, with the forms of partiality allowed under a utilitarianism that recognizes the usefulness of general moral rules. 9. Thus, like Brandt (in his discussion of minimal core morality), Singer occasionally seems to fall into the error of supposing that the prudential necessity of an individual's restraint toward his fellow "insiders" implies the prudential necessity of his (or their) restraint with regard to "outsiders". This, of course, amounts to forgetting the problem of group egoism, not solving it. -156suits of their interests fare no better. Here I will consider the most prominent contractarian of this sort, David Gauthier. As noted above, Gauthier once despaired of the success of contractarianism because of the group egoism problem and related matters. But in his more recent systematic treatise on ethics, Morals by Agreement, he shows no signs of that former despair. This is not, however, because he offers a general solution to the group egoism problem. Rather, he follows a strategy that I have used in previous discussions of the problem—a piecemeal strategy. This consists in trying to bring morality and the implications of rational interest theory into closer accord by simultaneously (i) whittling down traditional interpretations of morality to make it less restrictive and demanding, and (2) arguing that rational pursuit of interests does require

extending the scope of our moral concern to include many of the main groups that common sense morality says must be included. 10. In essence, Gauthier argues that morality does depend on reciprocity, but reciprocity—properly understood— both extends further in some directions than we are initially tempted to think, and yields a morality that requires much less of us than some moralists think. In particular, reciprocity extends far enough to encompass the claims of future generations, but does not necessarily condemn imperialism or vast inequality in the international sphere. Here I will ignore issues of international justice to focus on Gauthier's treatment of intergenerational justice. But before turning to that topic, it is worth noting that one of Gauthier's foremost interpreters—Christopher Morris (1991)—has shown how certain elements of Gauthier's theory allow for the scope of moral standing to extend beyond the bounds of reciprocity. For on one reading, Gauthier allows his rational bargainers to take an interest in the interests of those with whom they do not interact reciprocally (Morris 1991: 91). Hence, if I need reciprocation from you, Gauthier's theory may require me to take account of the interests of a third party (for example, an animal, infant, fetus, terminallyill patient, or impoverished foreigner) whom neither of us reciprocates with, and whom I do not myself care about—provided you care about that party. We may, however, accept Morris's observation about this implication of Gauthier's theory without feeling it saves that theory from the group egoism problem. For, even on this interpretation, the moral standing of individuals (and groups) is contingent upon whether anyone (and if so, whom) in other groups happens to care about them. And it seems to assign moral standing in a rather arbitrary way: two entities may be just alike except with regard to possession of the fortuitous relational property of being cared for by me, yet— on this account—one of them has moral standing for you and the other does not. If Morris has not solved the group egoism problem for Gauthier in a satisfactory way, has the latter done so himself? To find out, let us look at how Gauthier ____________________ 10. See Gauthier 1986: ch. 9, especially pp. 269-70; Kavka 1984 and 1986: 439-46. -157handles the central case of saving for future generations. At first glance, relations between generations seem paradigmatically asymmetrical and non-reciprocal. We can affect future generations a great deal in virtue of what kind of a world we leave them to live in, but most of them—being born after our deaths—cannot affect us at all. Does a rational pursuit of interest model therefore imply the strongly counterintuitive conclusion that we have no obligation to save (invest and protect the environment) for future generations? Not according to Gauthier, who writes: Mutually beneficial co-operation directly involves persons of different but overlapping generations, but this creates indirect co-operative links extending throughout history.... Thus, although each individual might be

prepared to agree with his contemporaries that they should exhaust the world's resources without thought for those yet to be born, the need to continue any agreement as time passes, to extend it to those who are born as it ceases for those who die, ensures, among rational persons, the terms must remain constant, so the exhaustion of the world's resources does not present itself as an option. (Gauthier 1986: 299) And Gauthier offers his own view of what particular "constant terms" regarding savings would emerge from a rational bargain among generations. Each generation has a maximal claim based on what it would receive if all the benefits of social cooperation from previous generations were passed on to it rather than being consumed by those previous generations. Hence, each generation has a larger maximal claim than its predecessor. Each generations' minimal claim is its state of nature utility, assumed to be the same for all generations. And the just outcome—the one that would emerge from a rational bargain among generations (or their members)—requires savings that provide each generation with the same fraction of its maximal claim, measured from the common minimal claim baseline. 11. This means larger and larger absolute amounts are consumed by each succeeding generation, since an equal fraction of a larger total claim is larger. But each generation gets an equal percentage share of what it could have gotten, given its position in the order, if all previous generations had passed on their entire cooperative surpluses. Gauthier considers this criterion superior to a utilitarian solution to the problem of intergenerational justice on the grounds that the latter may require too great a sacrifice by present generations for future ones. As he puts it: ____________________ 11. Actually, Gauthier's stated criterion is a bit more complicated to allow for taking advantage of pareto-improvements (i.e., changes making every generation better off) over the equal fraction of maximal claims outcome. I ignore this minor complication. In the models I discuss, no pareto-improvements over the equal fractions outcome are possible, because total production (and hence consumption) is fixed. -158The utilitarian may find herself as the advocate of restricting current consumption to whatever is needed to ensure maximum productivity, and eschewing all personal luxury, the costs to present persons being outweighed in her calculations by the much greater benefits to future persons. (Gauthier 1986: 304-5) Unfortunately, neither of the two main elements in Gauthier's treatment of intergenerational justice can withstand close scrutiny. Consider first his attempt to show that reciprocity covers the relations among generations because of generational overlap. He speaks here of the terms of saving and investment having to remain constant because of "the need to continue any agreement as time passes" (1986: 299). But whose need is this supposed to be? Not that of the first generation subject to the agreement, whose members need continuation of the agreement at most only through their lifetimes. Does

their overlap with the next generation (assume only one for simplicity) create such a need for rational bargainers? Hardly. The asymmetry of their situation with respect to their successors means they are in an enormously superior bargaining position: if no agreement is reached, they can use what they want and deal with the individual members of the next generation on an ad hoc basis as they arrive. The next generation can threaten to withhold its members' cooperation unless they get decent terms, and may argue that they will need fair terms to ensure the cooperation of their successors because of overlap. But this threat of non-cooperation will be hollow since to carry out the threat would redound against the later generation's own interests. (Gauthier himself argues that such threats to act in a nonmaximizing way will not be made among rational bargainers (1986: 155-56).) And the first generation can point out that the next generation need not worry about the similarly hollow threats of later generations. Thus, in a rational bargain among self‐ interested generations about savings, the earlier generations would successfully exploit the bargaining advantage inherent in their temporal precedence to force agreement on a very low intergenerational savings rate, despite the fact of generational overlap. 12. Of course, this outcome could be altered in various ways by changing assumptions about the bargaining situation. If we assume, á la Rawls (1971: 292), benevolent concern for particular members of future generations (e.g., descendants), this would raise the savings rate upon which the generational representatives rationally agree. Or, if we legislate impartiality by dropping a Rawlsian veil of ignorance over the parties, so that they represent different generations and each does not know which she represents, we might expect a fair savings schedule ____________________ 12. The actual savings rate between generations of rational individuals would turn out higher than the rate required by their rational agreement for two reasons: people's benevolence toward particular members of future generations (particularly their own offspring and descendants) and their inability to consume all their assets during their lifetimes because of uncertainty about the length of their lives. -159to emerge (Rawls 1971: 286-92). But making either of these moves would depart from the strategy—shared by Gauthier—of deriving moral requirements from rational pursuit of individual interest, rather than by imposing conditions on the pursuit of interest designed to yield fair or moral outcomes. The group egoism problem will have been solved in this context essentially by assuming it away. I have argued that a rational bargain among self-interested generational representatives would yield a low savings rate due to the advantaged bargaining position of representatives of the early generations. 13. Gauthier criticizes utilitarianism for requiring too heavy sacrifices by early generations, i.e., too high a saving rate, at least initially. What does his own "proportional shares of maximal claims" proposal imply about the savings rate? Gauthier does not go into specifics on this question. But by constructing a pair of very simple quantitative models, we can see that essentially the same criticism

that he applies to utilitarianism—that it places an unfair burden on early generations— also applies to his own proposal. Our first model assumes constant production. The first generation produces benefits of cooperation of amount one unit. Any of this it does not consume itself, it passes on to the second generation which also produces one unit worth. Anything that the second generation does not consume from what it has produced and inherited, it passes on to the third generation—which also produces one unit—and so on. What is the just savings rate required by Gauthier's proposal for this simple model? Each generation's minimal claim is zero. Each generation would receive its maximal claim if all previous generations passed on their entire surplus to them. So the maximal claim for generation N is simply N, one unit from each generation up to and including itself. P is the proportion of each generation's maximal claim that it should receive according to Gauthier's criterion. We can readily determine P as follows. The total to be consumed by the N generations is the total produced, N. (The last generation is assumed to consume all it produces and all it has inherited.) If Gauthier's condition is to be satisfied, the Kth generation's consumption is P times its maximal claim, or KP. The sum of consumption of the various generations must therefore be iP + 2P + ... + NP = P(1 + 2 + ... + N). The sum of the integers up to and including N is [N(N+1)]÷2. (The mean number in the series is [{N+1}÷2] and there are N numbers in the series.) So, setting the sum of production and consumption equal, we get: N = [PN(N+1)]÷2. Solving for P, the Gauthier consumption rate, yields: P = 2÷(N+1). ____________________ 13. When I speak of "early" or "first" generations, I mean generations that are early or first among the generations involved in the hypothetical bargain over savings rates. If we assume the past is fixed, so the bargain is among present and future generations about savings rates for the future, then the "first" generation consists of presently existing people. -160Ironically, in view of the excessive demands he claims that utilitarianism places on early generations, Gauthier's criterion ends up allowing early generations to consume very little and requires them to save virtually all. If we imagine a horizon of only ten generations, then the consumption allowed is only 18% of maximal claims. If the agreement covers 50, or 100 generations, the consumption rate drops to 4%, or 2%, respectively. For the first generation, its maximal claim is only its own production of one unit. So the Gauthier criterion requires of the first generation out of one hundred that it save 98% of its paltry one unit available and consume only 2%! Later generations, to be sure, can only consume the same proportion of their maximal claims. But these maximal claims, defined counterfactually in terms of what would be available if the entire social surplus had been passed on to them by previous generations, are many times greater than the maximal claims of the early generations, even assuming no increase in productivity over time, as we have in this model. So the early generations—already punished by the natural lottery

in being born too soon and hence not being able to live much off the savings of their predecessors—are further punished by Gauthier's criterion, which greatly inflates the relative claims of those who come later. Are these conclusions a result of the oversimplifications of our model, particularly the unrealistic assumption of constant production? To find out, we may consider a slightly more complex model, in which the productive capacity of each generation grows arithmetically. That is, the first generation produces one unit (which can be used in any combination for consumption or saving), the second generation produces two units, and so on. What does Gauthier's criterion imply in the case of this model with its more realistic assumption of rising productivity over time (perhaps due to social, cultural, technical, and intellectual advances)? In this second model, total production is 1 + 2 + ... + N, which equals [N(N+1)]÷2. As before, each generations' maximal claim is equal to total production up through that generation: [N(N+1)]÷2 for the Nth generation. For convenience we may abbreviate the Kth generation's maximal claim as MC(K), remembering it is equal to [K(K+1)]]÷2. When Gauthier's criterion is satisfied, consumption is distributed so that each generation gets a constant proportion P of its maximal claim. By putting production equal to consumption, we get the equation: P [MC(1)] + P [MC(2)] +... + P [MC(N)] = [N(N+1)]÷2. When we solve this equation for P—the Gauthier consumption rate—we get a surprisingly simple result: 3÷(N+2). 14. This allows a slightly higher consumption rate than in ____________________ 14. An outline of the necessary calculations is as follows: P[MC(1)] + P[MC(2)] + ... + P[MC(N)] = P{[MC(1)] + [MC(2)] + ... + [MC(N)]}. Since for each integer K, MC(K)=[K(K+1)]÷2=(K+K2)÷2, the bracketed sum on the right equals: [(i + 2 + ... N) + (12 + 22 + ... N2)]÷2. We have seen that the left factor of the numerator equals -161our earlier constant production model: 25% (rather than 18%) for a horizon of ten generations, 6% (rather than 4%) for a fifty generation horizon, and 3% (rather than 2%) for a horizon of one hundred generations. But, while on this increasing production model, the earliest generations fare a bit better in absolute terms, they fare much worse in relative terms. Since maximal claims grow much faster from generation to generation with an increasing production model, later generations are getting proportionate shares of much bigger maximal claims. Hence, if the Gauthier criterion is satisfied, the ratio of consumption—in terms of units of goods or resources—between later and earlier generations increases the more productivity increases over time. Hence, the unsavory implications of Gauthier's proposal about intergenerational justice are not due to the oversimplifications embodied in our first

model. They are more a result of the odd idea that each generation's maximal claim point should correspond to all earlier generations foregoing the fruits of cooperation for that generation's benefit. This leads to inequities favoring later generations that are probably greater than any allowed by utilitarianism, and which could never be adopted in a rational bargain among self-interested generations in which the early generations have a bargaining advantage due to their temporal priority. In blunt terms, the characteristics of the bargain amongst generations that Gauthier envisions, and the actual quantitative implications of the distributive criterion he proposes, are utterly incompatible.

IV. Conclusions I have discussed the views of three authors who recognize—in one form or another—the group egoism problem. Of these, Brandt never attempts a solution, Singer "solves" it only by imposing a utilitarian interpretation of rationality, and Gauthier's strategy of piecemeal attack fails due to the inadequacies of his account of justice among generations. Are these difficulties merely accidental ones, that an improved version of the ambitious reconciliation strategy might avoid? Or does the group egoism problem reveal a fundamental weakness in the strategy itself? I am inclined to take the latter view: the strategy of deriving morality from ____________________ [N(N+1)]÷2, and a well-known formula (readily verifiable by a simple proof by mathematical induction) tells us that the right factor equals [N(N+1)(2N+1)]÷6. Setting consumption equal to production, and substituting in these values for the sums of integers—and integers squared—up to and induding N, we obtain this equation: (P+2){[N(N+1)÷2] + [N(N+1)(2N+1)÷6]} = N(N+1)÷2. Dividing both sides of this equation by N(N+1)÷2, yields: P ([2N+4]÷6) = 1. Solving for P gives the formula in the text, P = 3÷(N+2). If the simple formulas for P in the two models leads the reader to suspect that the author explored a variety of models and selected for presentation the ones with the simplest formulas for the consumption rate, he overestimates the author and underestimates the powers of chance. -162the choices of rational individuals pursuing their own interests is too ambitious to succeed. To achieve a reasonable degree of success, any version of the reconciliation project will have to be less ambitious than this. In particular, it will have to include four features. First, as Brandt and Gauthier readily concede, 15. it will have to accept weaker (or narrower) interpretations of the requirements of morality than many philosophers favor. Second, it will have to rely on various contingent probabilistic claims about the coincidence of group and individual interests. Third, it will have to make use of human

benevolence to broaden the scope— and uphold the force—of moral requirements. Fourth, the drive toward unanimity, the ambition of showing adherence to moral norms to be rational for everyone, must—despite its attractions—be abandoned. On this last point, Brandt seems more willing than Gauthier to bow to the inevitable. 16. There is an alternate way of attempting to reconcile rationality and morality, exemplified by Singer. One can drop the individualist and instrumentalist conception of rationality employed by Brandt and Gauthier, and import some notion of impartiality into the idea of rationality itself. This, as we have seen, has two drawbacks. The problem of scope, of whom we are to be impartial among, remains. And the attractive aim of the ambitious reconciliation strategy— deriving moral requirements from an instrumental conception of individual rationality—has been abandoned. But perhaps, in light of the group egoism problem, it has to be. ____________________ 15. Brandt 1979: 242; Gauthier 1986: 268. 16. Cf. Brandt 1979: 221 and Gauthier 1986: 1-2. -163-

11 Welfare, Equality, and Distribution: Brandt from the Left William H. Shaw In a previous essay (Shaw 1988), I recommended to Marxists that they approach morality in the way Richard B. Brandt sets forth in his 1979 book, A Theory of the Good and the Right. That essay showed the affinities between Marxism and Brandt's theory, and it argued that from a Marxist perspective a Brandt-like approach is superior to rival ethical doctrines because it fits better with the general program of historical materialism, with the Marxist critique of ideology, and with Marxism's historical, yet critical approach to morality. However, whether leftists find a moral theory like Brandt's attractive will depend not just on questions of metaethics, moral method, and Marxist theory, but also on the theory's implications for the more politically charged question of distributive justice. A commitment to socioeconomic justice, in general, and economic equality, in particular, lies close to the heart of all leftists, whether Marxist or not, and they will not be won to a utilitarian theory like Brandt's unless its practical political ramifications are suitably egalitarian. Brandt treats the topic of distributive justice in Chapter 16 of A Theory, where he puts forward a moral principle intended to govern economic distribution. In what follows, I rehearse the arguments for Brandt's principle and answer rejoinders to them. I also argue that the principle is one that egalitarians can accept. I shall not be concerned here with whether Brandt's principle is compatible with the little Marx had to say about justice or about distribution under socialism and communism, 1. but rather with whether his principle is compatible with the strong egalitarian sentiments of those on the contemporary political left. If I am ____________________ 1. For a review of Marx's reflections on justice and distribution and of the exegetical controversies these reflections have engendered, see Peffer 1990: ch. 8. Allen W. Wood argues, controversially, that Marx himself was hostile to the ideal of equality (Wood 1981). -165right, then Brandt's approach should be judged congenial to those of us on the left, not only—as argued in my earlier essay—at the level of ethical methodology, but also in terms of its (basically utilitarian) treatment of questions of economic distribution.

I. Utility and Equality The allegation that utilitarianism is a faulty moral theory because it overlooks, distorts, or violates our sense of justice is an old one. Mill responded to this contention in the fifth chapter of Utilitarianism, and much, much more has been written on the issue since then. When the critic of utilitarianism is on the political left and when the focus of discussion is not simply justice but, more specifically, distributive justice—that is, society's allocation of economic benefits and burdens, goods and income—then the critical contention is that utilitarianism is insufficiently egalitarian. This is the charge against which I wish to defend Brandt's moral system. The above charge can be raised at two levels. At the first level, the contention is that utilitarian policies and institutions are insufficiently egalitarian in the straightforward sense that they tolerate too much economic inequality. The egalitarian critic must not be supposed to be demanding perfect equality in the distribution of economic benefits and burdens because the critic may, as a matter of principle, acknowledge values other than equality, or may, as matter of practice, view perfect equality as an unattainable goal, the realization of which will inevitably be compromised by various factors. Egalitarians disagree among themselves about the dimension of equality that is of fundamental moral concern—is it equality of welfare, equality of resources, equality of opportunity, or equality of something else? I pass over these intramural disputes because, despite them, egalitarians unite in believing that economic equality is an important political ideal because income and wealth are at least good proxies for, or means to, those things or those respects in which it is inherently valuable for us all to be equal, even if income and wealth are not themselves things equality of which is intrinsically valuable. There is, however, a second level at which the charge of inegalitarianism is sometimes advanced against utilitarianism. This is the philosophically deeper allegation that, regardless of utilitarianism's actual distributive implications— even granting, ex hypothesi, that in the real world these implications are strongly egalitarian— utilitarianism slights the moral equality of persons at some more profound level. Although utilitarians adhere to Bentham's dictum that each person is to count as one and no one as more than one, critics of the theory believe that its commitment to welfare maximization fails somehow to respect individuals as equals. This sort of accusation is widespread in the literature, but it -166will be largely ignored here, for two reasons. First, the effort to show that utilitarianism (at the theoretical as opposed to practical level) is not deeply enough committed to equality, like the parallel contention that it is not deeply enough committed to desert, say, or to rights, requires a full-blown moral theory to support it. But it is well beyond the scope of this essay to survey and assess all the moral systems that rival Brandt's. Moreover, and this is the second reason, the deep worry becomes less interesting and less pressing—at least for political leftists—if utilitarianism is in fact sufficiently egalitarian in its practical distributive implications. If utilitarianism gives the right answer in

practice, the charge that it does so on the wrong philosophical grounds is not likely to disturb left-wing economic egalitarians who are without prior commitment to some nonutilitarian theory. By contrast, if utilitarianism were seriously inegalitarian in its economic implications, this fact would cause left egalitarians to reject theories like Brandt's. 2. It would also embarrass those friends of utilitarianism who feel strongly about equality, perhaps leading them to reassess their attachment to utilitarianism.

II. Brandt's Distributive Justice Principle Let us turn, then, to what Brandt takes to be the implications of his approach for the distribution of income and goods. Brandt proposes the following principle as one that a welfare-maximizing moral system would embrace and which, accordingly, should inform the relevant social institutions: The real income ... after any taxes should be equal, except (a) for supplements to meet special needs, (b) supplements recompensing services to the extent needed to provide desirable incentive and allocate resources efficiently, and (c) variations to achieve other socially desirable ends such as population control. (1979: 310) I shall call this Brandt's Distributive Justice Principle (or BDJP). BDJP concerns the allocation of money both as a matter of simplicity—money is something we can easily hand out in equal amounts—and because, a few exceptions aside, it is more efficient and better promotes welfare to allow individuals to make their own purchasing decisions rather than have goods allocated to them. Elaboration of BDJP is required, as Brandt acknowledges, both with regard to the size of the supplements and variations and with regard to the institutional structures best ____________________ 2. In fact, political progressives already do tend to reject utilitarian theories as inherently reactionary and hostile to their concerns. This is because right-wingers frequently defend inegalitarian programs (e.g., tax cuts for the wealthy, abolition of minimum wage laws, tolerance of unemployment, hostility to unions) in what sounds like utilitarian language (e.g., enhanced economic growth, trickle-down prosperity). -167suited to implement the principle. I do not undertake that task here. Rather, I aim to restate the case for BDJP, to rebut various criticisms of that argument, and to show that BDJP's commitment to economic equality is as robust as reasonable egalitarians would want. A few preliminaries are necessary. First, Brandt does not explicitly refer to his principle as a principle of justice. I do not know whether this omission was intentional, but I call the principle "Brandt's Distributive Justice Principle" anyway because it covers the domain traditionally covered by principles of distributive or economic justice and

because doing so fits with Brandt's own understanding of what it means to single something out as "just" rather than "right", say, or "obligatory". There is, in any case, no reason for utilitarians of Brandt's stripe to shy away from talk of justice. 3. Second, like the other principles of a welfare-maximizing moral code, BDJP has only prima facie force "of whatever degree of strength would be welfare maximizing". This means that in practice BDJP may have to be balanced against other moral norms. This fact does not give rise to an objection because, as I intimated, even staunch egalitarians are unlikely to value only equality. Of course, one would not be much of an egalitarian if economic equality were far down one's list of values, and Brandt does not say what degree of strength BDJP would have in comparison to other moral principles. However, since BDJP is the only principle of economic distribution he enunciates, presumably it is the most fundamental principle in that sphere and would be a salient part of the moral code welfare-maximizers would attempt to propagate in their society, teach to their children, and internalize in themselves. Although welfare maximization motivates BDJP, the principle tends strongly toward equality of welfare. Not only does equality of real income promote equality of welfare, but so also do (a) and (b), which authorize exceptions to income equality. These exceptions move us away from an equal distribution, but they move us toward equality of welfare by removing inequalities due to special needs and by compensating those "who are willing to work rather than enjoy leisure time". What justifies BDJP, though, is not an ideal of welfare equality but, rather, one of welfare maximization. Maximizing and equalizing, however, are not necessarily antithetical, and Brandt maintains that efforts to promote the former will strongly tend to produce the latter. Exceptions (a) and (b) modify equality of income in light of individual need or the type of work or service being provided. These are considerations that principles of distributive justice frequently and properly take into account. By contrast, exception (c) seems odd. We might reward people for having more ____________________ 3. Rereading the fifth chapter of Mill's Utilitarianism should convince one of this. For Brandt on "just" and "unjust", see his 1979: 307. -168than the average number of children in order to encourage them to be fruitful, or we might tax them more for doing so in order to curb population growth. However sensible either policy might or might not be in particular circumstances, neither policy seems to be a matter of (what one would normally take to be) justice or rights. I would therefore recommend that (c) be dropped from BDJP. This recommendation does not entail that (c)-type considerations should not influence the allocation of income, but rather that they not do so as a matter of economic justice. Because BDJP is a prima facie principle, it can appropriately be balanced against other considerations. In my view, exception (c) makes internal to BDJP considerations that would be more coherently seen as countervailing. I

say "more coherently" because BDJP is supposed to be, for Brandt, an independent moral principle, part of the ideal moral code for society. BDJP will more clearly constitute a moral norm or principle of justice that is distinct from the simple open-ended injunction to maximize welfare if the miscellaneous utilitarian policy concerns, to which exception (c) opens the door, are excluded from BDJP itself.

III. The Diminishing Marginal Utility of Money Subject to the exceptions mentioned above, BDJP requires equality of monetary income. Various consequentialist considerations can be marshaled on behalf of equality of income or, at least, against significant economic inequality: (a) economic inequality promotes snobbishness, indolence, and pursuit of (relatively less satisfying) luxury and status goods at one end of the social spectrum, discontent and envy at the other end; (b) it distorts the marketplace because, where supply is limited, those who have more money will simply outbid those who have less regardless of the value the latter place on the goods in question; and (c) economic inequality translates swiftly into political inequality. Brandt does not appeal to these contentions in A Theory of the Good and the Right, 4. resting his case for BDJP instead on the argument that "an equal distribution of income is the best strategy for maximizing happiness, in view of the declining marginal utility of money" (1979: 312). Accordingly, I shall focus on this argument, though I think the above mentioned considerations have merit as well. The diminishing-marginal-utility-of-money argument is a venerable one, from which many in the utilitarian tradition have drawn egalitarian sustenance. A. C. Pigou, for one, stated the case (1932 [1952]: 89) for equal distribution with admirable succinctness: ____________________ 4. He does, though, in his Ethical Theory (1959: 419-20). R. M. Hare sees envy as one of two utilitarian arguments for equal distribution (the other is diminishing marginal utility) in Hare 1991:127. -169Any transference of income from a relatively rich man to a relatively poor man of similar temperament, since it enables more intense wants to be satisfied at the expense of less intense wants, must increase the aggregate sum of satisfaction. The old "law of diminishing utility" thus leads securely to the proposition: Any cause which increases the absolute share of real income in the hands of the poor, provided that it does not lead to a contraction in the size of the national dividend ... will, in general, increase economic welfare. 5. In the three sections that follow this one, I defend the thesis of diminishing marginal utility both in general and with regard to money. In the final three sections of the essay I turn to the other premises linking diminishing marginal utility to equality of income. But, first, the thesis needs to be explained.

Economists have long pointed out that additional units of a commodity are worth less and less to a consumer. The more of something a consumer already has, the less will be the marginal utility of an additional unit of that good. Economists call this generalization the Law of Diminishing Marginal Utility. Paul Samuelson explains it this way: As you consume more and more, your total utility will grow at a slower and slower rate. This slower growth in total utility arises because your marginal utility (the extra utility added by the last unit consumed of a good) diminishes as more of the good is consumed. (Samuelson and Nordhaus 1985: 412) 6. Introspection and commonsense observation support this proposition, which reflects "the fact that your appreciation or taste for a good drops off as more of the good is consumed" (Samuelson and Nordhaus 1985: 412). 7. If I have only one pair of shoes, the benefit I receive from another pair is much greater than if I already possess twenty-five pairs. I may enjoy more overall happiness from twenty‐ six pairs of shoes than from two, but the incremental, or marginal, gain in my happiness from my twenty-sixth pair is much less than the marginal benefit of the second pair. Economists recognize that some people and some commodities may be exceptions, but they do not see these anomalies as undermining the basic soundness of the Law of Diminishing Marginal Utility. ____________________ 5. Economists credit Daniel Bernoulli in the eighteenth century with having been the first to discuss explicitly the declining marginal utility of money. Bentham viewed with favor the argument for equality based on it in his Theory of Legislation, "Pathological Propositions upon which the Advantage of Equality Is Founded". 6. For an equally authoritative textbook account, see Baumol and Blinder 1985: 358; for a more philosophical discussion, see Dyke 1981: 35-41. 7. Empirical psychology lends some support as well because laboratory studies show analogous effects regarding our perception of sound, light, and other sensations. For example, the larger the weight in the palm of a blindfolded person, the less the person will notice the addition of an extra (or marginal) unit of weight. See Samuelson and Nordhaus 1985: 412-14. -170The argument for BDJP assumes that the marginal utility of money declines just as the marginal utility of other commodities does. Introspection, everyday observation, and commonsense psychology support this assumption, too. Although increasing a person's income may (and usually does) increase his or her satisfaction, it does so at a diminishing rate. An additional $100 brings more satisfaction to a person who has only $5000 than it does to the same person when he or she has $50,000.

IV. Frankfurt on Diminishing Marginal Utility Not everyone has been persuaded. As part of his critique of economic equality as a moral ideal, Harry Frankfurt (1987) has attacked the proposition that an egalitarian distribution of economic assets will tend to maximize aggregate utility. He disparages the thesis of diminishing marginal utility, in general, and its application to money, in particular. Frankfurt begins by observing that people may derive more utility from certain goods after sustained consumption than they derive at first. This is true when appreciating or enjoying something requires repeated trials "which serve as a kind of 'warming up' process: for instance, when relatively little significant gratification is obtained from the item or experience in question until the individual has acquired a special taste for it, has become addicted to it, or has begun in some other way to relate or respond to it profitably" (Frankfurt 1987: 26). 8. Because capacity for gratification is less at earlier points in the consumption sequence than at later points, marginal utility in such cases increases as the consumption sequence proceeds. Frankfurt is pointing to something real. Some tastes are acquired, and their pleasures deepen with repeated trials. Because modern jazz is not immediately accessible to me, I may, for example, enjoy listening to a jazz record more on the twentieth time than on the second (although even this is more likely to be true if I play the record once a week than if I play it twenty times in a row). However, Frankfurt's "warming up" point does not suggest, let alone entail, that I enjoy my twentieth jazz record more than my second. The fact that the marginal utility of certain experiences (like listening to the record) can be said to increase, does not imply that the marginal utility of owning different jazz records does. Moreover, even where marginal utility of experiences or possessions does increase, this trend will not continue indefinitely. Frankfurt's own (passing) reference to addiction makes this point eloquently because nothing better illustrates what it means to be locked onto a downward marginal utility curve. Frankfurt ____________________ 8. In a similar vein, Charles Karelis (1986: 105-7) contends that at low levels of consumption where the consumer is basically dissatisfied, the marginal utility of consumption may increase: thus, the fourth forkful may give more relief to a hungry diner than the first. -171(1987: 27) basically acknowledges this, appearing content to rebut only the assumption "that every segment of the curve has a downward slope". But no believer in diminishing marginal utility makes such an assumption. The thesis of diminishing marginal utility does not imply that one's second bottle of beer is inevitably less satisfying than one's first, but only that at some point each further bottle of beer is less satisfying than earlier bottles. What the principle asserts, and what matters for the egalitarian argument it

supports, is not that one's marginal utility curve is uniformly smooth and downward in every segment, but rather that its overall course is southerly. On the basis of the above, Frankfurt apparently rejects the economists' Law of Diminishing Marginal Utility although he does acknowledge that the marginal utilities of "certain goods" tend to diminish and that "experiences of many kinds become increasingly routine and unrewarding as they are repeated". He adds, however, that this is "not a principle of reason. It is a psychological generalization" (1987: 26). This is rhetorically tendentious on two counts. First, given that the Law is really a claim about the overall shape of utility curves, it is not just "certain goods" whose marginal utility declines; it is the vast majority. Plausible exceptions are hard to find, and when they are found, they appear to be of restricted scope. Second, although economists call it a "law", the thesis of declining marginal utility is, of course, a psychological generalization. No one asserts it to be a "principle of reason", if by that one means a non-empirical principle. Moreover, if it is true that for most goods and most people, marginal utility declines as consumption increases, then this fact has important moral and policy implications even if its epistemological status is seen—at least in the eyes of philosophers—as lower than that of a "principle of reason".

V. Monetary Income and Utility Frankfurt also argues that even if the thesis of diminishing marginal utility holds for consumer goods in general, it is doubtful "whether this provides any reason at all for expecting a diminution in the marginal utility of money". The reason it is doubtful is simply that money is different. "It is quite possible that money would be exempt from the phenomenon of unrelenting marginal decline because of its limitlessly protean qualities." (Frankfurt 1987: 26) Or as Blum and Kalven, on whose work Frankfurt draws, put it: "Money is infinitely versatile. And even if all the things money can buy are subject to a law of diminishing utility, it does not follow that money is." (Blum and Kalven 1953: 58) Money is not just another commodity; it is the generic instrument of exchange. The diminishing marginal utility of goods entails that "a person tends to lose more and more interest in what he is consuming as his consumption increases", but this does not mean -172that he must also tend to lose interest in consumption itself or in the money that makes consumption possible. For there may always remain for him, no matter how tired he has become of what he has been doing, untried goods to be bought and fresh new pleasures to be enjoyed. (Frankfurt 1987: 26) Let us grant that money is different and that the diminishing marginal utility principle of consumer goods does not entail that the marginal utility of money declines. Still, is the latter not a very reasonable assumption? After all, common experience and common sense support the observation that the incremental value of $1000 to a millionaire is far less than it is to an laid-off blue-collar worker.

Blum and Kalven reject this argument as based on introspective intuition. They concede that most people would agree that if they had many times the amount of money they now have, they would place a lesser value on an additional dollar. But the argument stumbles on the problem of interpersonal comparisons. The argument can only proceed on the assumption that the money utility curve for yourself, which you derive from introspection, also holds true for other men in the society. This further hypothesis ... can only be "tested" by intuition. (Blum and Kalven 1953: 61) They go on to quote Lionel Robbins to the effect that introspection does not enable person A to measure what is going on in B's mind, nor vice-versa. However, Blum and Kalven's rebuttal is misguided, and their appeal to Robbins misplaced. My conviction that the marginal utility of my monetary income diminishes as my income increases might be a product of introspective reflection. Yet if others report the same conviction and if their behavior seems to confirm their reports, then my belief that in general the marginal utility of income diminishes as income increases does not rest on my introspection but rather on observation and what other people relate about themselves. By analogy, if others say that they are like me in finding their third helping of a rich dessert less satisfying that their first and the prospect of a fourth positively nauseous, and if their actions are consistent with what they say, then my understanding of the typical dessertutility curve is not based simply on an introspective analysis of my own case. Frankfurt (1987: 27) objects to the diminishing marginal utility of income thesis for a different reason, namely, that it overlooks utility thresholds due to saving: "Accumulating money entails, as warming up does, generating a capacity to derive, at some subsequent point in a sequence, gratifications that cannot be derived earlier." This proposition is murky. Some people oversave, and the mere fact that one is accumulating money can hardly entail that one will derive (or is now "generating a capacity to derive") more gratification from the future -173spending that saving makes possible than from spending one's money as it comes along. What is undoubtedly true is that "it may at times be especially worthwhile for a person to save money rather than to spend each dollar", and it is probably also true that the utility of a dollar that completes a savings program and thus makes possible some fresh and previously unobtainable type of satisfaction "may be greater than the utility of any dollar saved earlier in the program" (Frankfurt 1987: 27). This threshold effect is particularly evident in the case of collectors "who characteristically derive greater satisfaction from obtaining the item that finally completes a collection ... than from obtaining any of the other items in the collection" (Frankfurt 1987: 28n). 9. But this is manifestly a special case and, even if granted, is far from showing that the overall utility curve of money is other than downward.

Moreover, the whole notion of threshold effects has egalitarian implications because extra money is more likely to result in utility gains from threshold effects when dispensed to the poor than when given to the rich. 10. $2400 given to an unemployed worker is more likely to help him or her across some utility threshold than it is to help a millionaire, or if the same sum helps both across a utility threshold, the poorer person's utility gain is likely to be larger—for example, acquisition of a (used) car versus an upgraded car stereo system for the car one already has. 11. Extra money can help those in poverty benefit from utility thresholds that are without parallel for well-to-do citizens. The homeless, for example, are thwarted in their search for the jobs that would free them from poverty by the lack of an abode, a telephone, and the showers and clean clothing necessary for making themselves presentable to prospective employers. By contrast, bestowing an extra $50,000 on a multi-millionaire might make it convenient for her to complete her collection of classic sports cars, but it is hard to believe that the threshold this gift enables her to cross produces more utility than would sharing the money among some unemployed families. Further, if the multi‐ millionaire values completing her collection, she already has the resources to do so without any hardship to herself. She need not wait for an extra $50,000 to fall into her lap; she can sell one of her condominiums or hire less domestic help. If she has not already done so, it must be because she values these things more than she does completing her collection. ____________________ 10. The egalitarian implications are even stronger if Karelis (1986: 113) is right that, for people who have very little of it, the marginal utility of money typically increases. 11. A local store specializing in such things estimated the price of replacing my car stereo with a system they recommended as "really good" to be $2400. 9. This is a standard example, of which believers in diminishing marginal utility are aware. See Baumol and Blinder 1985: 359. -174-

VI. The Grocery Basket Argument This last point is important, and Brandt appeals to something like it when he argues that "expenditure of successive increments to one's income produces less happiness or welfare than that of the preceding increments, on the average and in the long-run" (1979: 312). The reason such expenditure produces less happiness follows from the fact that outcomes are preferentially ordered, some being more strongly wanted than others: So a person, when deciding how to spend his resources, picks a basket of groceries which is at least as appealing as any other he can purchase with the money he has. The things he does not buy are omitted because other things are wanted more. If we double a person's income, he will spend the extra money on items he wants less (some special cases aside), and which will give less enjoyment than will the original income. The more one's income, the fewer preferred items one buys and the more preferred items one already has. On the whole, then, when the necessities of life have been

purchased and the individual is spending on luxury items, he is buying items which will give less enjoyment (1979: 312) Brandt's argument derives from Abba Lerner, who first stressed that the thesis of the diminishing marginal utility of money follows from the simple assumption that consumers spend their income in a way that maximizes their satisfaction (Lerner 1944: 26). Compare the goods the rational consumer actually buys with the goods he could have bought in their place with the same income, but did not. Given that he prefers the chosen basket of goods to other things that cost no more, the consumer must regard this basket of goods as offering at least as much, if not more, satisfaction than would an alternative basket of goods. With additional income, the consumer now purchases things he initially passed over because he viewed them as providing less satisfaction. I now examine and answer four rejoinders to the Brandt-Lerner grocery basket argument. (i) First is that, according to "everyday observation", what is bought is usually only no less utile, and not more utile than what is rejected. In other words, the most fruitful use for new income, and the use to which it is actually put, is usually "more of the same", and not some new thing that was passed over when the present income level was achieved. What most people most want, specifically, is ever-greater quantities of a compound that consists of such items as square feet of living space, days of leisure, and years of education for their children, and when income rises, it is more of this compound, and not previously rejected things, that is bought.... But this means that the average person's marginal utility for income would ... be constant rather than diminishing. (Karelis 1986: 111) -175There are three problems with this argument. (a) Appeals to "everyday observation" and "what most people want" cannot conceal the baldness of the assertion that with increased income people buy more of the same goods, rather than new goods. To deny or play down the fact that with increased income people buy, not just more food, clothes, and entertainment, but different and better food, clothes, and entertainment, seems perverse. (b) Even if what people typically want with increased income is simply "more of the same", this fact still fits the grocery basket model. Suppose, for example, that with my initial income I choose an apartment with 800 square feet of living space rather than a more expensive apartment with 1200 feet, preferring to spend less on rent in order to spend more dining out in restaurants. When my income increases, I move to the larger apartment. I am (ex hypothesi) purchasing more of the same, but I am also choosing a commodity that I passed over before. If I had valued the larger space sufficiently, I could have rented the larger apartment initially and modified my eating habits. (c) The "more of the same" view does not entail that marginal utility remains constant instead of declining. Whether I am buying increments of living space or additional cans of beans, it is hard to believe that the marginal utility of these purchases remains constant.

(2) In response to the grocery basket argument, Frankfurt contends that the satisfaction a person receives from a certain good may vary according to whether she possesses certain other goods. Suppose Sally loves buttered popcorn but gets little satisfaction from unbuttered popcorn and even less from butter by itself. Popcorn and the amount of butter necessary to flavor it costing the same, Sally will buy popcorn in preference to butter if she cannot buy both. But the incremental increase in income that enables her to buy both butter and popcorn makes it possible for her to enjoy buttered popcorn, and her satisfaction from this is considerably greater than is the sum of her satisfaction from butter and popcorn taken separately. The particular increase in income that makes possible the purchase of both butter and popcorn has an enhanced marginal utility. 12. No doubt goods like butter and popcorn can complement each other in this happy and marginal-utility increasing way, but Frankfurt overlooks cases where possession or consumption of some new good can diminish rather than increase the satisfaction brought by the goods with which it is combined. Thus, as Sally's income increases yet further, she now buys flavored gourmet popcorn, which is sufficiently tasty that butter adds far less zest to the overall experience than it did before. The main trouble, though, with Frankfurt's rejoinder is that the Brandt‐ Lerner argument does not suppose that one's income-utility curve moves smoothly and consistently downward, that it does so immediately with the first dollar, or that it descends with each dollar-by-dollar increment. ____________________ 12. Frankfurt neglects to mention that both Lerner and Brandt have discussed cases like this; see Lerner 1944: 32-33, and Brandt 1959: 416. -176Brandt's contention about what holds "special cases aside" and "on the average and in the long-run" is unaffected by a few micro-level counterexamples. Buttered popcorn does not undermine the grocery basket argument because the sums are ridiculously small. Start someone with a weekly grocery budget of $100; even if being able to buy butter and popcorn together brings a utility bonus, this fact hardly undercuts the larger point that increases in a person's grocery budget will bring utility increases only at a diminishing rate. Perhaps Frankfurt's point would have weight against Lerner if Lerner were understood (unfairly) to be claiming to have deductively proved the diminishing marginal utility of income by showing it to be a logical entailment of consumer rationality, but Frankfurt's rejoinder certainly has no weight against Brandt's formulation of the argument. We are concerned with an empirical generalization about consumer satisfaction, one which is made plausible by the considerations Brandt and Lerner introduce, which is supported by everyday observation and commonsense reflection, and which could presumably be verified (or modified or refuted) by closer scientific study. A counterexample like that of buttered popcorn cannot refute an empirical generalization unless there is reason for thinking it represents a tendency at least as widespread or as probable as that identified by the purported generalization.

A final point needs to be made in response to Frankfurt: as long as the overall slope of people's income curves is downward, then the irregularities he points to are irrelevant. The reason they are irrelevant is that there is no reason to believe that the rich person's downward curve will have more upward blips or upward blips of more significance than will the curve of the poor person. Indeed, there is reason to believe that the reverse is true for it is the poor and not the rich, who have trouble putting butter and popcorn together. (3) Frankfurt advances another line of rebuttal to the grocery basket argument, challenging its apparent assumption that the rational consumer uses his first n dollars to purchase a certain good, G(n), and then uses his additional income i to buy something else, G(i). This supposition is unwarranted, Frankfurt says: instead of using his n + i income to purchase G(n) and G(i), the consumer might use it buy something, G(n+i), that costs more than either of these goods. In other words, the fact that a consumer defers purchasing a certain good until his income increases does not imply that he "rejected" purchasing it when his income was smaller. The good in question may simply have been beyond his financial reach at the time. If the consumer saved his initial n in hope of acquiring the i necessary to purchase G(n+i), then he clearly prefers it to either G(n) or G(i). On the other hand, if he spends n on G(n) because he despairs of ever having the i necessary to purchase G(n+i), then it would still be wrong, Frankfurt argues, to construe his behavior as representing a preference for G(n) over G(n+i). But Frankfurt's argument here boils down simply to the point that the utility of G(n+i) may be greater than the utility of G(n) and G(i) together, which -177is the utility threshold argument again. Brandt's presentation of the argument acknowledges the existence of special cases like this, and Frankfurt has given us no reason to think such cases offset the general tendency in question, still less that such cases are more common or have more utility significance at the high end of the income spectrum than at the low end. (4) A fourth and final rejoinder to the grocery basket argument comes from Blum and Kalven. They write: It is not plausible that the most important wants of a man with a $5,000 income remain his most important wants when he has an income of $25,000. As his income changes his way of life changes. He becomes in effect a man with a different hierarchy of wants and values. In the end, all that can be told from the data is that when a man has $5,000 he prefers most the uses he then makes of his income and that when he has $25,000 he again prefers most the uses he makes of his income. (Blum and Kalven 1953: 60) According to Blum and Kalven, then, we cannot say that with increasing income a person receives less satisfaction from the extra money spent because it is being spent on less preferred items. All we can say, rather, is that the person's wants have changed.

One's way of life can indeed change with one's income, and the rich may well have wants the poor do not—perhaps because poor people (like other people) tend not to want things they know are out of their reach. But this fact does not embarrass the Brandt-Lerner argument. First, Blum and Kalven seem implicitly to assume that because one's wants have changed, the amount of satisfaction received at the two different levels of income cannot be compared. Yet this agnosticism does not follow from their premise, and they give no independent argument for it. Moreover, agnosticism here would prevent one from asserting not only that the marginal utility of money declines but also that satisfaction is positively correlated with income, which is a fundamental postulate for neoclassical economists (and, one assumes, for Blum and Kalven, too). Second, in saying that a person's "most important wants" change, Blum and Kalven trade on an ambiguity. Assume that at an annual income of $5,000, Fred wants but cannot afford an automobile. At $25,000 he is able to buy one and solve his transportation difficulties. Is having an automobile still an important want? No and Yes. No, because Fred's want has been met. He does not "want" a car because he has one. Indeed, he takes for granted his vehicle and the welfare‐ enhancing convenience it brings. Yes, because ownership of an automobile remains a strong, dispositional want. Whether Fred thinks about it or not, possession of a car contributes significantly to his overall welfare, making it possible for him to hold the job and live the style of life he does. -178-

VII. Are Our Income-Utility Levels Alike? That the marginal utility of a dollar decreases as one acquires more of them seems so well and widely confirmed that the existence of threshold effects and other, occasional upward blips on one's income-utility curve are unlikely to shake our conviction that an additional $100,000 given, say, to Lee A. Iacocca, whose reported 1992 paycheck from Chrysler was $16.9 million, could not possibly have a discernible impact on his life, let alone bring him more satisfaction, happiness, or well-being than did the first $100,000 he ever received. And what is true of Mr. Iacocca seems true, mutatis mutandis, for other people and at other income levels. Yet even if the declining marginal utility of monetary income is a general truth, this by itself does not imply that income should be distributed equally. One reason for this (a second is discussed in the next section) is that if people's income-utility curves are at different levels and if we know what these are, then welfare-maximization requires an income distribution that equalizes marginal utilities, rather than an equal distribution of income. This is why Pigou, in the passage quoted previously, refers to the "transference of income from a relatively rich man to a relatively poor man of similar temperament". If we assume that people's temperaments are similar or, in other words, that they have similar income-utility levels, then, given the declining marginal utility of money, increasing the poor person's absolute share of real income increases net welfare.

This is an assumption Pigou—unlike Brandt—is happy to make. Pigou rejects the idea that rich and poor have different mental constitutions—either inherently or as a result of upbringing and training—such that the rich are able to derive more satisfaction from a given income than the poor are. And he deals effectively with the class-biased contention that any sudden rise in the income of the poor is likely to lead to "a good deal of foolish expenditure which involves little or no addition to economic welfare" (Pigou 1932 [1952]: 91). That the rich operate at a higher utility level than the poor is a proposition most people today would find both false and distasteful, but probably few philosophers or economists would assert that we know to be true Pigou's postulate that people have similar utility levels. 13. Pigou's assumption is a supposition about human nature, but it is congenial both to those in the utilitarian tradition and to those in the Marxist tradition. ____________________ 13. For his part, Frankfurt asserts (1987: 25) that we know it to be false, claiming that "it is evident that the utility functions for money of different individuals are not even approximately alike.... Everyone knows that there are, at any given level of expenditure, large differences in the quantities of utility that different spenders derive." -179Bentham, Mill, and Marx rejected the worst class prejudices of their day and emphasized the malleability of human nature and the ways in which social institutions shape our character and desires. Such assumptions about human nature may be unprovable. But that people react in the same way to varying amounts of money can be claimed to be a plausible working hypothesis. "In the ordinary affairs of life", Pigou urges, "we always assume that groups of prima facie similar men will be mentally affected by similar situations in much the same way.... We expect similar situations to produce similar mental effects." 14. Or as another economist writes with regard to taxation: The assumption that wants are equal ... though obviously not true, approximates more nearly to the truth than any other working assumption that could possibly be invented. Since the state must collect a revenue, it must have some definite assumption upon which it can proceed. The question is not, therefore, whether men's wants are equal, but whether there is any rule of inequality of wants upon which the apportionment of taxes could be made with a nearer approximation to the truth. 15. The argument here involves two steps: first, that Pigou's postulate is more plausible than any alternative assumption and, second, that since we must act on the basis of some assumption about individual's utility curves, we best act on the one that is most plausible. A related, but distinct line of argument is that, in addition to inherent plausibility, there are both moral and policy reasons for adopting Pigou's assumption. Being simpler than its rivals, that supposition provides an easier and more straightforward basis on which to

design institutions and to implement policies. By analogy, the law frequently ignores individual differences and acts on the assumption that people are the same, for example, when it sets one standard of negligence for all. 16. In addition, the assumption of human equality fits better with the principles and values that make up the moral code that utilitarians would want to govern their society than does the rival proposition that human beings differ in their ability to enjoy similar experiences. Given our democratic tradition and culture, publicly defending that rival premise as a morally legitimate basis for policy would be difficult, if not impossible. Its hierarchical implications would be hard to square with the equality of respect and concern for others that the ideal moral code would presumably enjoin and without which it would be less likely to win support and prove socially stable. ____________________ 14. As quoted in Blum and Kalven 1953: 61. 15. As quoted in Blum and Kalven 1953: 61n. 16. Blum and Kalven (1953: 62n) reject the analogy but on—I would argue—weak grounds. -180Although I find the assumption that people's utility curves are at basically similar levels plausible and believe that the above considerations justify taking it as a basis for public policy, Brandt does not avail himself of this assumption. Instead, he grants that "people's tastes and needs differ, and [that] the correlation between income and happiness-level also varies from one person to another, depending on personality" (1979: 312). Nevertheless, Brandt observes, we lack reliable information about individuals' incomeutility curves. This lack of reliable information, he explains, is the reason why utilitarians cannot maximize welfare simply by distributing money to those whose well-being will be most enhanced by it. 17. There are not publicly identifiable classes into which individuals with high, medium, or low levels of income-utility fall. And there is no straightforward way in which individuals could hope to show that they would benefit relatively more from extra income than other people would. Furthermore, or so I would suggest, even if people could hope to show this, permitting them to try would be socially divisive and would channel their energies in counterproductive directions. Thus, we are in no position to adopt the strategy of maximizing utility by distributing the national income such that the extra enjoyment produced by the last dollar each person receives is exactly the same. In addition, even if we grant that people's levels of incomeutility are different, the declining marginal utility of income implies that, given our lack of information about those differing levels, the happiness-maximizing strategy is to divide income equally. Following Abba Lerner again, Brandt illustrates the reason for this with a graph representing the marginal utility-income curves for Mr. A and Mr. B. Diminishing marginal utility of money means that both their curves slope downward. Suppose, though, that Mr. B's curve is lower because he always gets less satisfaction from his income. If we are to maximize enjoyment, we should give more to Mr. A rather than divide things equally. But if we do not know whose curve is higher and whose

lower, then we do not know which way to move away from equality in order to maximize happiness. Now one might think that it would make no difference in which direction we moved. The probability of shifting away from equality in a direction that increases overall utility is equal to the probability that we are moving in a direction that decreases overall utility. Given this, it would seem that we would be moving the right way half the time and the wrong way half the time. But while the probabilities are indeed the same, the loss from moving the wrong way is greater than the gain from moving the right way because both curves slope downward (as a result of diminishing marginal utility). As Lerner (1944: 32) writes, "every time a ____________________ 17. In the case of the ill or handicapped, on the other hand, we do have enough reliable information to do this. -181movement is made away from an equalitarian division the probable size of the loss is greater than the probable size of the gain." The same reasoning applies to any number of persons, implying that equal distribution will maximize probable total satisfaction. Several writers have reviewed this argument, whose basic validity has not been gainsaid. 18. Along with the other reflections of this section, the argument suggests the following conclusion. Because of the declining marginal utility of money, equal distribution of income is the best strategy for maximizing happiness, under the following disjunctive condition: that people receive roughly the same utility from any given level of income, that we are justified in postulating that they do, or that while their utility levels differ, there is no practical or efficient way for us to know whose level is higher and whose lower.

VIII. Incentives The declining marginal utility of money has powerful egalitarian implications, but it has these implications only if one grants an important further assumption. This assumption is frequently stated to be that transfers from rich to poor do not diminish the total amount of income to be distributed, that is, that redistribution "does not lead to a contraction in the size of the national dividend" (Pigou 1932 [1952]: 89) or that "total amount of income is unrelated to its distribution" (Friedman 1947: 410). Strictly speaking, though, the argument for income equality requires only the weaker assumption that any contraction of the national dividend occasions less loss of welfare than is gained by income transfers. Because money is not happiness, a relatively equal distribution of a total national income of x might well produce more total welfare than an unequal distribution of a significantly larger national income of x + n.

Nevertheless, moving from the status quo toward equality of income is a process, the dynamic effects of which may have untoward welfare consequences. Depending on the socioeconomic institutional framework of the society in question and the motivations of its citizens, some ways of pursuing income equality or the pursuing of it too rapidly or too relentlessly may be, or after a certain point come to be, counterproductive. This could happen for a variety of reasons. Among the more interesting and important is the possibility that people will lack sufficient economic incentive to invest in certain ways, to pursue certain careers, or to work at certain jobs. They might lack sufficient incentive, say, because the tax system deprives them of all income above a certain level and gives it to those who earned less than that amount. The result could be a reduction in ____________________ 18. See Friedman 1947: 409-11; Samuelson 1972: 647-50; Buchanan 1985: 56-58. -182national income sufficient to lower net social welfare despite the welfare gains from income transfers. Brandt's Principle of Distributive Justice addresses this problem by allowing equality of income to be modified by "supplements recompensing services to the extent needed to provide desirable incentive and allocate resources efficiently". For instance, BDJP might permit financial rewards to individuals beyond their (hypothetical) equal share of national income if this were necessary for people to enter unpleasant, stressful, or dangerous occupations, to work well in certain boring and monotonous jobs, to put in long hours, or to sacrifice current enjoyments in order to undergo training or acquire the skills necessary for pursuing certain socially beneficial careers. In such ways, a system permitting income differentials might entice people more successfully, than would a system of equal income, to behave in ways that expand society's wealth and, thus, its total welfare. Should egalitarians worry about this? Brandt makes the important point, noted earlier, that while such supplements undermine equality of income, they promote equality of welfare. By recompensing people for extra efforts, such supplements tend "to make the total prospects for different persons equal" (Brandt 1979: 320). In this respect, such supplements are analogous to those to the ill and the handicapped: the extra income allocated in this way serves to equalize people's welfare and, thus, should not disturb egalitarians (Lerner 1944: 32). But the phrase "supplements recompensing services" obscures a contrast between payments necessary simply to offset the welfare losses one would otherwise incur by undertaking a certain job or pursuing a certain career and payments that reward people above and beyond that level. For example, the motivational structure of people in a certain society might be such that extra payments are needed to entice those with special talents to apply themselves to tasks that only they can do or that they can do better than others. If the benefits to society from the talented applying themselves are great enough, utilitarianism could conceivably authorize their receiving a financial reward that goes beyond what is necessary to

recompense their welfare sacrifices. Incentive payments of this sort—call them "pure incentives"—smack of extortion, and paying them will distress egalitarians. On fuller reflection, however, there are reasons for doubting that utilitarians would authorize pure incentive payments—by contrast, with "welfare equalizing" supplements that eliminate disincentives—or, at least, that they would authorize them routinely. First, in our society the comparatively lush rewards of many talented professionals are due, at least in part, to restrictions on entry into those professional fields. Control over the number of medical school places, for example, has enabled the American Medical Association to boost physicians' earnings enormously over the decades, both absolutely and in comparison to other professions. In various areas, licensure and other occupational restrictions -183buoy the incomes of relatively privileged professionals by preventing paraprofessionals and others from performing equivalent services for less. Removal of unnecessary restrictions would go a long way toward freeing society from having to pay pure incentive payments. Second, many of the talented members of society, those with socially useful skills in limited supply, possess the talents they do, not because of innate ability alone, but also because of the comparative advantages they enjoyed in education, upbringing, and social environment. A society committed to substantive, as opposed to merely formal, equality of opportunity would find itself with a larger pool of talented or potentially talented citizens, thus reducing the possibility of any group being in a position to demand pure incentive payments. A third point supports the previous two: jobs that only the talented can perform tend to be more intrinsically engaging and rewarding than other jobs—in large measure because they involve the exercise of skill and expertise. Imagine that medical training and subsequent hospital apprenticeship were recompensed equally with other jobs and that young interns were no longer initiated into their profession by overwork. Suppose further that a practicing physician's work week were approximately the same as that of a hospital orderly. It is hard to believe that many with the temperament and talent for being physicians would prefer to be orderlies instead and that pure incentive payments would be necessary to entice them to be doctors. Many are the jobs whose salary and other advantages are significantly higher than necessary to induce people to do them—jobs, in other words, whose occupants enjoy what economists call "employment rent". Fourth, society has strong utilitarian grounds for resisting giving into demands for pure incentive payments—even if these seem necessary at the time— because it will not want to encourage such demands nor to affirm the social legitimacy of such motivation. Considerations of long-run social welfare thus undergird social efforts to promote the norm of equality and to reinforce jobrelated motivational considerations other than the desire for an extra and disproportionately high monetary reward.

BDJP permits supplements "to the extent needed" as incentives, but the four considerations just discussed argue for interpreting this phrase as not countenancing pure incentive payments as opposed to welfare-equalizing supplements. A further reason is this. If BDJP is interpreting as permitting pure incentive payments, this risks draining it of content and undermining its moral character because BDJP would, in effect, be licensing as morally legitimate any supplementary income that a person is in a position successfully to demand. On the one hand, such an interpretation risks circularity. What people demand and whether others meet those demands depend, in part, on the perceived justice of those demands. Under the interpretation of BDJP disfavored here, there would be no truly independent norm against which to measure their legitimacy. That is, the -184demand for a pure incentive payment would be legitimate if and only if those making it were able to enforce their demands. On the other hand, bear in mind that BDJP is supposed to be, not a pragmatic or temporary economic policy, but a moral principle. According to Brandt, having a moral principle involves, among other things, being intrinsically motivated to act according to it and feeling guilty when one fails to live up to it. Those who embrace the principle tend to disapprove of those who do not follow it and to admire those who do. It is doubtful that, under the disfavored interpretation, BDJP would be a principle that rational persons would seek to write into the moral code of their society, which would involve them propagating it as best they can, teaching it to their children, and internalizing it in themselves. The principle that people are entitled to a larger share of income than others if they can successfully extort it from society is an unlikely candidate for the ideal moral code. By contrast, the interpretation of BDJP preferred here is simpler and intrinsically more attractive; it is more likely to win adherents; it is easier to teach; it forms a more stable and enduring basis for internalization and conative attachment; it can weather better what Rawls calls "the strains of commitment"; and it fits more coherently with the other values and principles likely to be enshrined in the ideal moral code. Interpreting BDJP so as to exclude pure incentive payments does not imply that such payments should never be permitted. BDJP is a prima facie principle, and its institutional implementation may have to be modified in light of other moral considerations. Now suppose—my previous arguments notwithstanding—that we were in an exceptional situation where paying certain people pure incentive payments really would maximize total social welfare and by a significant margin. My recommendation is that, rather than interpret BDJP so that it authorizes those payments, we view them as in conflict with the principle but say that the situation is such that we permit, at least temporarily, other considerations to override BDJP, modifying its application in practice. (By analogy, sometimes the wise and right course is to yield to the extortionist, but this does not make the extortion legitimate.) In this way, BDJP retains its integrity and its distinctive content. Even if BDJP is compromised or overridden in certain circumstances, it lays down a norm of equality that would be a better candidate for the ideal moral code than the version of BDJP that permits pure incentive payments.

IX. Allocation of Material Resources The previous section focused on job income, but one might think that the issue of incentives and disincentives poses a larger obstacle to equal distribution under BDJP when one considers entrepreneurship, investment, and the efficient utilization of capital assets or material resources. Equality of income might be -185attacked on the grounds (i) that the prospect of fairly spectacular reward is necessary to entice independent inventor-entrepreneurs to spend the time and take the risks necessary to develop new products, services, and processes or (ii) that the prospect of a less spectacular, but nonetheless equality-busting reward is necessary to entice those with capital to invest it productively. Although these two issues deserve more complete discussion, two brief comments must suffice here. First, whether contention (i) really is a problem and, to the extent that it is, whether the problem might be addressed in ways other than by permitting income inequality need to be examined. One cannot simply assume that the creative impulse responds only to financial incentives or that these must be inegalitarian in character. For example, grants, subsidies, risk-pooling, insurance, and other forms of assistance might be designed to encourage independent inventorentrepreneurs to tackle problems others are ignoring. Second, contention (ii) points to the question, skirted by Brandt, of exactly what sort of economic system and what sort of property arrangements would be welfare maximizing. But even if, following Brandt, we restrict discussion to a market-oriented society in which people have private control of some productive assets, an important point can be made. People are in a position to demand pure incentive payments that distort income equality just to the extent that they control productive resources—either human or material—that others lack. Equalizing people's productive resources would minimize this problem. Milton Friedman makes a related point when discussing the fact that Lerner (like Brandt) is willing to compromise the principle of equality in order to provide incentives necessary to increase total income: This difficulty could have been largely avoided by considering instead the distribution of resources.... Measures to reduce inequality by altering the distribution of resources (such as social investment in the training of individuals, inheritance taxation, etc.) may interfere less with the optimum utilization of resources than measures that seek to redistribute income directly. (Friedman 1947: 410) Because a people's capitalism made up of citizens with equal productive assets would enjoy a fairly high degree of income equality prior to any redistributive taxation, equality of income could be maintained and promoted with less resource misallocation and market distortion. Friedman's point resonates, surprisingly, with the traditional Marxist dictum that socialists should be concerned more with control of the means of production than with distribution of the products of production.

In America today inequality of wealth and of the attendant control over society's means of production is prodigious. For instance, the richest 1 percent of American households owns 37 percent of the private wealth of the country and has a greater net worth than the bottom 90 percent of Americans added to -186gether. 19. Pushing for equal income, as BDJP requires, will undoubtedly necessitate an assault on inequalities of wealth as well. In addition, the declining marginal utility of money provides a direct and independent case for reducing inequality of wealth. Abolition of inheritance, to mention one obvious wealth equalizing strategy, would have enormous welfare-enhancing effects. 20. Throughout this essay, I have defended Brandt's Distributive Justice Principle, rebutting criticisms of it and of the declining-marginal-utility-of-income thesis that buttresses it. I have also tried to show that in practice BDJP is as egalitarian a principle of distributive justice as left-wingers would want and that the principle's egalitarian character is not undercut by the need for economic incentives. But an analysis of the full economic implications of a utilitarian approach like Brandt's must also involve a comparative assessment of different possible economic systems along welfare lines—in particular, consideration of rival forms of economic ownership, of rival ways of organizing production and distribution, and of alternative authority arrangements within units of production. Friends of Brandt's theory need also to examine in detail the comparative welfare consequences of pursuing equality of income in different ways and of pursuing in different ways and to different degrees substantive equality of opportunity, equality of wealth, and equality of productive resources in order to determine yet more specifically "whether and when and in what way and to what extent people ought to be treated equally" (Goldstick 1991: 70). Discussion of these issues involves challenging empirical questions, and it may be that a utilitarian assessment of them will yield only tentative and revisable policy recommendations, rather than some further principles that should be taught as part of the ideal moral code. But it is also possible that such discussion could lead to a revamping of BDJP. In other words, there might be Brandtian reasons for including within the ideal moral code an expanded principle (or set of principles) of economic justice that embrace other concerns beyond the distribution of income—for instance, control over capital assets or investment decision-making. 21. Whether there are such reasons and to what principle (or principles) of economic justice they would lead us are questions for future discussion and reflection, but any further development of Brandt's theory in this area is likely to strengthen, rather than lessen, its egalitarian import. ____________________ 19. See New York Times, April 21, 1992, p. Ai. 20. On the case for the abolition of inheritance, see Haslett 1986 and Ascher 1990. See also Michalos 1988. In his Principles of Political Economy John Stuart Mill favors limiting the amount any one person may receive by bequest or inheritance.

21.

Brandt suggests that a further principle about the allocation of educational opportunities might be needed (1979: 310). -187-

12 Brandt's Moral Philosophy in Perspective William K. Frankena Richard Brandt is one of the very few left who began working in moral philosophy in the thirties, lived through the intervening decades, and are still active in the nineties. In the language of a famous sentence by Kant, he has been an "I think" accompanying the ethical theorizing, meta-ethical and normative, of most of the twentieth century, as I too was in a somewhat different way. But Brandt has not contented himself with being a historian or interpreter, a commentator or critic, as I so largely have; from the beginning he resolutely set himself to develop a system of moral philosophy of his own that would compare well with those developed by others of his time, those of his predecessors, and live on into the next century or longer. He says in discussing psychological egoism (Brandt 1992: 97), "I confess to a desire to do philosophical work of some significance, and would exchange a good deal of personal enjoyment for the production by me of philosophical work as memorable as that of Aristotle or Plato", adding that "this desire is certainly not a desire for a state of euphoria of my self." Brandt has pursued this goal with singular devotion, with candor and flexibility, but with hard and careful work, clarity and vigor of thought, with an unusual knowledge of work in anthropology, psychology, and law, and also with a good deal of critical and constructive insight and originality—all to a very significant result, both in the form of a moral theory and of its application to current moral problems. He must be put very high on any list of those who have produced fullfledged systems of moral philosophy in English in the second half of this century. My project here is not to make a defense or critique of any part of Brandt's ethical theory, though I shall express a few doubts or questions about it, but it is also not just to expound it; it is rather to put his work in the perspective in which I now see it in a way that may be useful to others who are interested in thinking about it either critically or favorably, and in a manner that will make clearer than -189would otherwise be the case what he has contributed to the history of moral philosophy and of utilitarianism in particular. 1. One way to do this would be to write an account of this history and place Brandt in it; another would be to develop a systematic typology of moral theories, especially of utilitarian ones, and place Brandt's in that. Both courses are tempting to me, for I have always been historically and classificationally minded; and here I shall proceed by way of

a combination of them. The outcome can only be what it will be, but I hope it will throw light in various directions. Throwing light has been my own main endeavor all along, and this may be my last attempt to do so. In doing this, I shall have only British and American moral philosophers in mind, especially J. S. Mill, and rely mainly on the collection of Brandt's essays, Morality, Utilitarianism, and Rights (1992). I begin with some history of utilitarianism in moral philosophy. It has taken a variety of forms and has rested on various kinds of foundations. In form it has usually been hedonistic or at least quasi-hedonistic, conceiving of the good and utility in terms of pleasure, happiness, or desire or preference satisfaction, as in the theological utilitarians, Bentham, Mill, Sidgwick, and Hare; but it has also taken a rather different and opposing or "Ideal" form, conceiving of the good in terms of excellence or perfection, beauty or aesthetic experience, knowledge or truth, and goodness or virtue, as well as or instead of pleasure or happiness, as in Moore, Rashdall, Laird, and Ewing. Also, in its theory of morality or of the right, it has sometimes taken an act or direct utilitarian form, as in Moore, Smart, and perhaps Bentham, and sometimes a rule or indirect utilitarian one, as in Berkeley and others, but more commonly has been somewhat ambiguous on this point, as in Mill and perhaps Sidgwick, or else combined these two forms in some way, as Hare did. For grounding, utilitarianism has sometimes looked to intuition or self-evidence, as in the Ideal utilitarians and in Sidgwick; but it has generally been empiricist in theory of knowledge, taking either a theological direction by resting ethics on an empirical natural theology in one way or another, as in Berkeley, Gay, and Paley, or a non-theological one, as in Bentham, Mill, Sharp, Smart, Hare, and others. Thus, in what is called meta-ethics it has covered the waterfront—been cognitivist (definist and naturalistic, or intuitionist and non‐ naturalistic) or non-cognitivist (emotivist or prescriptivist or whatever), and either supernaturalist or not, mostly not. Now, Brandt regards himself as in general Millian in his ethical theory; this is clear from the number and character of his references to Mill over the years— just look at the index of the collection of Brandt's essays—and he has also called his theory Millian, though perhaps not in print. We may label a theory as generically Millian if, and maybe only if, (i) it is hedonistic or quasi-hedonistic in the above sense about the good or about the end to be promoted, that is, in its view ____________________ 1. I did so briefly for Brandt's earlier views in my Frankena 1964; 373-75. -190of what constitutes benefit, welfare, utility, or "flourishing", (2) it holds that the moral quality of actions, dispositions, and institutions depends ultimately and wholly, whether directly or indirectly, on the relative size of their contribution to the good so conceived of one's society, of humankind, or of the whole sentient creation, and (3) it regards these two parts of its doctrine as not based on theology, or on intuition or self-evidence, and as not being analytic or a priori in any sense, and yet as not being arbitrary, since a good case can be made for them on the basis of human experience and psychology.

In all three of these broad respects Brandt is a Millian. Re (i): he espouses a "happiness theory" of the good, as opposed to an "Ideal" and even to a desire or preferencesatisfaction theory, and he has a roughly hedonistic conception of happiness, equating the happiness theory with hedonism (Brandt 1992: 158) and happiness with enjoyment (1992: 164), though he does reject Mill's view of pleasure as a felt quality that itself has qualitative differences in favor of defining pleasure (happiness) in terms of having experiences one wants at the time to continue or repeat for their own sake. It follows that he must (and does) disown Mill's confusion between the view that desiring X is the same thing as finding it pleasant and the view that pleasure is the ultimate object of desire (two kinds of psychological hedonism), and cannot argue as Mill did that happiness is the good because it is the sole object of actual desire. What he believes instead is that one actually does what one most wants to do at the time, and that happiness is what one would want as an end for oneself, if one were rational in the sense to be defined below, even if it is not the only thing one wants as an end, since one can desire as ends things one will not experience oneself (as he says he does in the above quotation), desiring them actually and rationally, egoism and hedonism to the contrary notwithstanding. On such more specific points Brandt's views are more like Sidgwick's than like Mill's, but his position is still generically Millian. As for me, though I agree with much of what Brandt says so far, I have held to a more "Ideal" theory of the good, one more like those of Plato, Aristotle, Moore, Rashdall, Ross, Laird, and Ewing, though without the intuitionism or nonnaturalism involved in most of them. Brandt's relation to Millian point (2) we shall come to later, for it is only here that his utilitarianism proper comes into the picture; one can hold to almost any theory of the (non-morally) good that does not give priority to the right over the good (as Rawls has) without being a utilitarian—intuitionists, non-naturalists, and deontologists can and have been Millians and hedonists in their conceptions of the good and happiness, e.g. Butler, Price, Reid, and R. M. Blake. As for (3), I believe Brandt is a Millian about this too in a broad sense, though he differs from Mill on important details. Bentham and the Mills were the first in the history of ethics explicitly to reject all views that based ethics on intuition or self-evidence, on theology or any kind of metaphysics, or on deduction or any kind of deductive argument from non-ethical or non-evaluative premises, while also -191maintaining that ethics is neither a matter of convention nor entirely up for grabs. Thus while J. S. Mill follows Bentham in insisting that basic ethical principles are not matters of intuition and also are not subject to what is commonly understood by proof, viz. proof by reasoning, he goes on, in a very famous passage, to say of the utilitarian normative theory, We are not, however, to infer that its acceptance or rejection must depend on blind impulse, or arbitrary choice. There is a larger meaning of the word "proof", in which this question is as amenable to it as any other of the disputed questions of philosophy. The subject is within the cognizance of the rational faculty.... Considerations may be presented capable of

determining the intellect either to give or withhold its assent to the doctrine; and this is equivalent to proof. (Utilitarianism, near the end of ch. 1) This passage has often been lamented because of Mill's talk, here and later, about a kind of proof that is not a proof, and yet is a proof after all, and Broad (1930: 174) remarked that we all learned "at our mother's knee" some of what was wrong with what Mill went on to try to do. 2. But it nevertheless represents a crucial turn in the history of moral philosophy, for it involves a disclaimer both of the methods of establishing an ethics used by predecessors and other contemporaries and of the postures assumed later by ethical "skeptics" like James Balfour, Nietzsche, existentialists, logical positivists, other noncognitivists, and some relativists.Having written this, Mill in effect indicates that the rest of his little book is a presentation of the considerations he deems capable of determining the intellect to give or withhold its assent to his own utilitarian doctrine, i.e., to (i) and (2) as he understands them. Analysis of the book reveals that he tries to do the following things: a. Give an explanation making clear just what his doctrine is and is not. b. Answer objections to it. c. Advance objections to other doctrines. d. Show what sanctions are available or what motives the doctrine has. e. Prove (in a pretty strict sense) that psychological hedonism is true, or more accurately, that we are so constituted as to desire nothing which is not either a part of happiness or a means to it. 3. And to be fair to Mill we must read him as thinking, only after he has done his best with all of them, that he has given a "proof" in his "larger" sense, not just of ____________________ 2. Actually, the knees belonged to Sidgwick, Bradley, and Moore. 3. Mill also argues that his doctrine is implicitly subscribed to by his opponents. This is a favorite ploy of Bentham's, and Mill uses it in his remarks on Kant in ch. 1 . Brandt does not use it. -192(1) but also of (2), i.e. has done enough to determine our intellects to give their assent, though he talks as if his "proof' falls entirely under (e). He has been taken to task especially for what he does in trying to establish his psychological hedonism, itself somewhat confused (as I indicated), but also for thinking that by proving happiness to be the only thing we desire, he can "prove" that it is the good or the only thing desirable, and finally for thinking that if this is so, then it "follows" that the promotion of happiness and even of the general happiness, must be the criterion of morality in some way or other. Nonetheless, it seems to me that Bentham and Mill were right in thinking that the sorts of things just listed constitute the only kinds of considerations actually available for "determining" our intellects to accept any normative theory, and that presenting them persuasively may and must be regarded as necessary and sufficient, whether one calls

doing so a "proof" or not, given that we cannot and should not rely on "the way of intuition" or on that of deduction from theological or other non-ethical premises (or going from Is to Ought as if one can do so logically). I wish to claim that Brandt is also Millian in accepting this general picture of the way in which one can and should argue for utilitarianism. He certainly rejects "the way of intuition", both in the older form of a claim that some ethical or value judgments are selfevident and in the recent form of a reliance on one's "intuitions", either in reflective equilibrium or not; he also denies that one can deduce any such judgments from theological, metaphysical, or scientific conclusions, though, as I mentioned, he makes a good deal of use of social science. And he very clearly and explicitly tries to do carefully the kinds of things indicated in (a), (b), (c), and (d), though he disagrees with many of Mill's moves under these headings, e.g. Mill's idea that pleasures differ in quality and his notion that happiness has "parts" like virtue, both of which I am more sympathetic with. (e) is another matter: like most of Mill's readers Brandt rejects Mill's alleged proof that we are so constituted as actually to desire nothing ultimately but happiness, as well as his further claims that one can "prove" a theory of the good by proving that human nature is constituted in a certain way, and that, if psychological hedonism is true, then the intellect must also assent to a utilitarian theory of morality of some kind. Still, I think that Brandt is generically Millian even at this point in insisting that one's ethical theory must be consistent with the best available scientific ones, that one's moral psychology, especially one's theories of motivation and conscience, must rest on such scientific theories. 4. I would say then that he is or has been presenting a Bentham-Mill kind of case for a roughly hedonistic sort of utilitarianism without making the specifically Millian mistakes that so many have pointed out. At this point we must notice an approach to the problems of ethical theory, meta-ethical or normative, which has not yet been mentioned, but which has ____________________ 4. See, e.g., Brandt 1992: 5. -193found some favor among analytical philosophers in this century, and from which Brandt has gone to considerable pains to distance himself. This is the view that to find a foundation for ethics it is necessary and sufficient to ask what is the meaning or analysis of terms or concepts like "good", "right", and "just", as these are actually used in ordinary ethical discourse or thought, or by asking about their "use" or "logic" in that context. As it had been for Sidgwick, this view was the chief of Moore's "prolegomena to any future ethics", and it remained so in one form or another for many of his successors, some of whom rejected the intuitionism and utilitarianism that he thought he could establish in this way, e.g. for Broad, Ross, Ewing, Toulmin, Urmson, Hare, and Firth. Now, if one adopts this "ordinary language" approach, there are roughly two directions one can take, unless one opts for an "error theory" like J. L. Mackie's. One is to argue that one or more of the central terms or concepts of common-sense ethics are unanalyzable,

simple, or indefinable, and that the others can be explicated by reference to them; this view has usually been called intuitionism or non-naturalism, and would be a form of moral realism. The other is to contend that ordinary ethical terms or concepts can all be defined or analyzed with entire fidelity to ordinary usage by using only non-ethical ones, empirical or non-empirical (Moore called such views, respectively, naturalistic or metaphysical, but tarred them all with committing "the naturalistic fallacy"); or, alternatively, that ordinary ethical judgments have a non-cognitive function and/or a logic that enables them to be analyzed along other lines, without essential reference to simple or non-natural properties or to descriptive, factual, or scientific ones, as in Hare's universal prescriptivism. One can take any of these paths and be a utilitarian. In Principia Ethica (but not in Ethics) Moore took the intuitionist path about good but then argued that the right could be defined as action that is conducive to at least as great a balance of intrinsic good over intrinsic evil in the world as possible, thus seeing act-utilitarianism as analytically true, or true by virtue of the very meaning of right, though only given the simple non-naturalness of good he thought was established by use of the open question argument. Hare finally came to the conclusion that his universalist kind of noncognitivism was equivalent to a form of act-utilitarianism. It is harder to find an important example of an ordinary language analyst who is a naturalist in Moore's sense and also a utilitarian, but Moore apparently thought this was true of Bentham and Mill, and some of the theological utilitarians could have thought that good can be defined as pleasure and right as what God commands, arriving at their utilitarianism, not by definition alone, but by adding that God commands us to promote the greatest general happiness, i.e., not at first but at second remove. Perhaps I should have mentioned this ordinary language approach in connection with what I said earlier about the Bentham-Mill kind of justification of ethical first principles and of the principle of utility in particular. It might be thought that Bentham was taking this approach when he wrote that, if ought and -194right do not mean conformity to the principles of utility, they have no meaning at all (Bentham 1789: ch. i). I am not sure, however, that we can rest anything much on this dictum, or even that Bentham would have thought he was stating a fact about what ordinary people mean by those words. This just does not sound like him! Again, Mill writes in one of Brandt's favorite passages that we "do not call anything wrong unless we mean to imply that a person ought to be punished in some way or other for doing it", and this does seem to be a claim about ordinary language or thought. But it can hardly be taken seriously even as part of a naturalistic definition or analysis of wrong since "we mean to imply" does not necessarily entail "we mean" in Moore's sense, and anyway ought is used here by Mill in a way that still needs explanation. Besides, he had already written in his System of Logic (near the very end) that ethical judgments are not indicative but imperative, and belong not to science but to art—a view he reasserted later and never, so far as I know, gave up. This is why I have implied that Bentham and Mill hold we cannot deduce any ethical judgments from factual ones, not even from those of psychology; Mill could therefore not have believed that one can go logically from

"Happiness is desired as end" to "It is desirable or good as end", as has been alleged. Though he said he could "prove" the latter by proving the former, he did not think this was a proof in the strict sense, and said as much himself. He was in contemporary terms a non-cognitivist about ethical judgments, not a crypto-naturalist. He must have meant only that, if the intellect is clearly and fully convinced that happiness and it alone is desired as end, then it will be "determined" to assent to the doctrine that it and it alone is desirable or good as end. No doubt he thought that in saying what he does he was making points about commonsense discourse, but at least he was not thinking that wrong means "contrary to the general happiness" or "punishing this is conducive to the general happiness" or anything utilitarian. In short, I still believe my earlier account of the Bentham-Mill approach to utilitarianism to be substantially correct. However this may be, the main point here is that Brandt rejects all forms of the ordinary language approach to value and moral theory. 5. We need not now recount his reasons, but it is important that we see what his posture is. He means to be a naturalist or definist in Moore's sense of giving definitions of ethical terms by reference to natural or empirical ones, and so to be a cognitivist about ethical judgments, though there is a question about this to be looked at later. But he does not pretend to be describing or analyzing our ordinary ethical discourse; for him there are too many bugs both in it and in ordinary language analysis. Indeed, he rightly insists, more explicitly than almost anyone else ever has, that ethical questions using our usual ethical words or concepts must be restated in an important and clarifying way. Thus he rejects reliance on linguistic intuitions ____________________ 5. Cf. Brandt 1992: 4f., 19-37. He did not always take this position; see my account referred to in note 1. -195as well as ethical ones and is not proposing merely to "elucidate" everyday ethical thinking but rather to reform it by giving clear and revisionary definitions of its terms that use only empirical and logical concepts and will, he hopes, replace those defined without any significant loss in what we can or wish to say or do. It would be open to Brandt simply to propose straight-out hedonistic and utilitarian definitions of the good, the right, etc., not of course as descriptive findings about our ethical discourse, but as such reforming proposals. This would yield him his happiness theory and his utilitarianism at first removes, but it is not what he does, possibly because he feels such definitions would be too far from ordinary usage, but more likely because he thinks it wise to determine our intellects to assent to his doctrines by taking them along a more indirect route. This is to argue for defining good, right, blameworthy, and other ethical expressions in one way or another in terms of being rational in some one of various respects, e.g. in action, in desire, etc., after having proposed to define or redefine what it is to be rational in each of those respects (Brandt 1992: 4-6, 39, 66)—and so as to be a reforming naturalist about ethical and value judgments and about judgments of rationality itself. In Sidgwick's words, Brandt wishes to give our ethical notions "an ideal

element" by including a reference to how a rational being would respond, but one that is "entirely interpretable in terms of fact, actual or hypothetical, and does not introduce any judgment of value". 6. Thus Brandt does not begin developing his theory of the good or desirable by recommending that we define it as what is desired as end, or even as happiness, but by restating the question at issue to ask what a rational person would desire for herself, and perhaps also for others, as such or for its own sake, where by a "rational person" he means and urges us to mean one who is logical, conceptually clearheaded, and actively affected by the best available relevant information, experience, and imagination. Only then does he go on to argue that the good is enjoyment or happiness (and not anything else, not even desire-satisfaction) by trying to convince our intellects, in effect, not that this is what we do ultimately desire, but that it is the rational thing to desire for ourselves or for others. 7. Mill's so-called "proof" that happiness is the good does not come off, Brandt implies, because "x is desired as an end" does not entail "a rational person would desire it as an end". We must now do something more to identify Brandt's form of utilitarianism and to see how he argues for it. It will of course include or be conjoined with his happiness theory of the good, but that is not what makes it utilitarian, as we have seen. What makes it utilitarian is the fact that it includes a utilitarian theory about what is morally right, wrong, good, bad, virtuous or vicious, viz. one that holds the moral qualities of anything (actions, motives, character traits, persons, ____________________ 6. See Sidgwick 1907: 112; and Brandt 1979: 126f. 7. Cf. Brandt 1992: ch. 9. -196etc.) to be dependent, directly or indirectly, on its bearing on the general good. One is a utilitarian if and only if one subscribes to such a theory of morality (or, if one prefers, of what Williams calls "the ethical" 8. ); Mill implies as much when he distinguishes utilitarianism's "theory of life" (that happiness is the only thing desirable as an end) and its "theory of morality" (that the promotion of happiness is the criterion of morality), even though he also sometimes equates "the utilitarian doctrine" with the former. 9. Thus Brandt remarks that utilitarianism "is usually defined as a general normative theory about what actions are morally right or wrong" (1992: 197), and indeed usually so defines it himself. However, in the passage just quoted from, he does not, though possibly only for the purposes of the question he is there concerned with, saying he is going to follow Mill, and then quoting a good bit of Mill and defining utilitarianism as a "general theory about when actions and institutions are (instrumentally) desirable", without any reference to morality. By quoting what he does he seems to me to be following more of Mill than he should, coming close to defining utilitarianism in terms of its theory of life rather than in terms of its theory of morality, and even seeming to approve of Mill's claim that if

happiness alone is good as end, then its promotion must be the criterion of morality, which is not actually his view as I shall be expounding it. In any event, Brandt does not himself argue directly from his happiness theory of the desirable to a utilitarian theory of the criterion of morality, as Mill tried to do in his ch. 4. Brandt's is an indirect utilitarianism or "indirect optimific theory" more indirectly arrived at, for he sees, as Mill did not, not only that one cannot show happiness to be the good by showing it and it alone to be desired as an end, but also that, even if one could, one cannot thereby show that the promotion of happiness must be "the test by which to judge of all human conduct" and institutions, since one could still be a deontologist about this; and he certainly sees that one cannot conclude that the promotion of the greatest general happiness must be the criterion of morality in any way. In order to reach this conclusion one must first establish something about morality, and, on Brandt's view, also something about a rational-in-his-sense person's attitude to the general welfare. None of this Mill saw, but Brandt does, even though he almost seems not to in his use of Mill in the passage just referred to. I have implied that utilitarian or optimffic theories of morality can be direct or indirect, and stated that Brandt's is indirect. A direct theory holds that the moral quality of an action is a function of its own utility or contribution to the general good, as in actutilitarianism; an indirect theory holds the moral quality of an action to be a function of the utility or contribution of something else to which the action is related in some way, e.g. of the act-type or rule it comes ____________________ 8. Cf. Williams 1985: 6, 25, 50, 90. 9. See ch. 2, near the beginning; ch. 4, second paragraph. -197under, the motive from which it is done or the disposition or character trait it expresses, as in rule-utilitarianism proper, motive-utilitarianism, and traitutilitarianism, respectively. 10. Mill did not make any such distinction; he writes mostly like a rule-utilitarian, occasionally like an act-utilitarian, and sometimes like a trait-utilitarian. Brandt has usually referred to himself as a rule-utilitarian, and tends to think of Mill as one too, but, as we shall see, he is himself one of a rather special sort. The distinction between the two main kinds of utilitarianism first came into prominence in a debate about what kind of a utilitarian Mill was, started by Urmson and continued by Smart, Mandelbaum, Lyons, and others. But it was originally and clearly made by Berkeley long ago, as Brandt pointed out (after I told him so). In his Passive Obedience (1712) Berkeley is a theological utilitarian who contends, on the basis of natural theology, that it is "the general well-being of all men ... which God desires should be procured by the concurring actions of each individual" and hence that it is "the great end to which all moral obligations are subordinate". He then writes,

The well-being of mankind must necessarily be carried on in one of these two ways:—Either, first, without the injunction of any ... universal rules of morality; only by obliging every one, upon each particular occasion, to consult the public good, and always to do that which to him shall seem, in the present time and circumstances, most to conduce to it: or, secondly, by enjoining the observation of some determinate, established laws, which, if universally practiced, have from the nature of things, an essential fitness to procure the well-being of mankind; though, in their particular application, they are sometimes... the occasion of great sufferings and misfortunes, it may be, to very many good men. 11. And he goes on to argue against the first and for the second of these two ways of promoting the general well-being, and to claim that we can know by reason what at least some of these rules or laws are, thus becoming the first explicit rule‐ utilitarian, at least as far as I know, unless Aquinas was one, which I doubt. Berkeley did not base his arguments against act-utilitarianism on intuitions; he used more pragmatic arguments (and a theological one). Brandt likewise rejects act-utilitarianism, but, as he himself notes, he at first made an appeal to intuition in doing so, which he now disowns in favor of arguments more like Berkeley's non-theological ones (Brandt 1992: 1, 10, 113f., 142f., 229). Brandt is then a rule or indirect utilitarian of some sort about morality, but of what sort, and how does he arrive at it? In accordance with his general strategy, ____________________ 10. The expression "motive-utilitarianism" as recently used is ambiguous. It may mean either the view that motives are to be judged by their utility or the view that actions are to be judged by the utility of their motives. One can hold the former without holding the latter. Here I mean the latter. On trait-utilitarianism, see Frankena 1973: 64. Mill writes like a trait-utilitarian about nobility near the end of his System of Logic. 11. Reprinted in Calkins, ed., 1929: 434. -198he must argue that in some respect his rational person will be a utilitarian about morality —not necessarily about what she desires or does but about morality, for, as I said, besides his happiness theory of life he must bring in something about rationality and something about morality to reach his utilitarianism. What does he bring in about morality? Briefly, a certain concept of what morality is, or rather of what it is for a person or society to have a morality. Behind Brandt's views about this there is a good deal of anthropology and social psychology, but he also makes interesting use of what Mill says about conscience in chapter 3 and about morality in chapter 5. To quote just one of his many formulations of his conception of morality, one which happens to make no reference to Mill: I take it that the central feature of a morality is... that there is in an individual, or group, a system of intrinsic aversions to types of actions

(with a corresponding tendency to feel guilty or remorseful, if these are infringed, and to disapprove of others for the same reason—where "disapprove" can cover a range of emotions/attitudes, such as being horrified, appalled, shocked, disgusted, and anyway having a negative "resentful" attitude toward the person ...). And, I think, participants in a morality normally think that these attitudes are justifiable in some appropriate sense. 12. For Brandt there is nothing to "morality" but the business of having such attitudes and expressing them in moral judgments of various sorts. To have a conscience or moral code for him is to have a morality in this sense. It should be noticed that Brandt allows that having a morality may also involve having intrinsic desires or preferences for certain types of actions; it need not, but it may and, he thinks, should; but in any case the attitudes, which may be acquired in various ways, must be for or against certain kinds of actions as such, e.g. keeping or breaking promises. We must also understand that he means here to be proposing or recommending that we define a morality as such as a set of intrinsic attitudes, not to be analyzing our ordinary concept of a morality, if there is one. He does think he is following Mill here, but he is certainly elaborating and improving on what Mill says. I think myself that Mill may have had two theories of morality which he did not distinguish, one in chapter 4 and another in 5, but agree that Mill does sometimes seem to conceive of morality vaguely in Brandt's way. Notice too that my quotation does not mention rules; this is probably because Brandt is not conceiving of moral rules merely as general statements or laws that a person or society might proclaim, as Berkeley seems to, but more as general in____________________ 12. Brandt 1992: 77; for a fuller statement see p. 62. In connection with what I say here, see especially pp. 121ff. There appears to be some ambivalence in Brandt's mind about whether or not morality is an institution (Brandt 1992: 121, 81, 148, 202). It seems to me that, while an individual's code or conscience may not be an institution, that of a society would be one, in much the way its legal system is, and that an individual's morality may be part of an institution if it is just an internalization of her society's. -199ternal attitudes or dispositions that a person or group has and may or may not proclaim or formulate as "rules". There is nothing utilitarian about this account of morality or conscience; it is and is designed to be neutral as between utilitarianism and other moral theories, and it says nothing about the content of moral codes or about how this is or should be determined. Also it is intended to cover ideal moralities as well as actual ones, and it does not imply that what is morally right or wrong or good or bad is what is required, permitted, or forbidden by one's actual moral code or that of one's society; indeed, Brandt proposes that what is morally right or wrong be regarded as fixed or even defined by the code or conscience it would be ideal to have, whatever this is. But now we must ask how he gets

from this conception of morality to his utilitarianism. His course, very different from those of his various kinds of utilitarian predecessors, including Mill, is as follows. First, he again brings in his concept of being rational. Just as he redirected the question what the good, private or public, is to make it ask what it is rational to desire or pursue as end or what one would desire or pursue as end for oneself if one were rational in his sense (answering that it is happiness), so now he recasts the question of what is morally right to make it ask something else about how a rational person in his sense would respond. He proposes that the question be understood as asking what would be permitted or required (or forbidden) by the moral code or conscience that one would choose and foster for the society in which one expects to live if one were rational in the sense of being logical, conceptually clear-headed, and actively affected by the best available information, experience, and imagination that would bear on that question in particular. It will not do, he believes, simply to ask what is right; one should ask that clearer question and try to answer it as carefully and wisely as possible. That is, as was indicated, he is proposing that the morally right be understood as what is permitted or required by the ideal moral code (or the ideal set of intrinsic aversions to and preferences for act‐ types), and that this be viewed as the one that such a rational person would have, and seek to have prevail, in her society. So far we still have no utilitarianism; Brandt's utilitarianism or indirect optimific theory appears only in his reply to the further question: how would such a rational person asking his question answer it, or what general type of moral code ( in the sense defined above) would she favor for her society? One could give a non-utilitarian answer here, but Brandt's contention is that she would favor the code or conscience the prevalence of which in her society would maximize expectable utility or happiness in that society; the ideal code or conscience is the one she would choose and favor for her society and this is the one that would expectably maximize welfare in her society if it were to prevail therein. Or in other words, the right act-types are those that would be required or at least permitted by the morality (set of intrinsic attitudes) a rational person as defined would espouse for herself and her society, and these are the act-types whose being re -200quired or permitted by her society would expectably promote the greatest general good or happiness in it. And the right action for a person in a certain particular situation is not the one that would itself maximize utility, but the one that would come closest to conforming to the set of intrinsic attitudes (toward act-types) she would have if she were rational, i.e., according to Brandt, the set of intrinsic attitudes (or conscience) that she would see as maximizing utility in her society, if taught and prevalent. Only in finally reaching these conclusions does Brandt become a utilitarian.Roughly, then, his moral philosophy may be summarized in the following set of equations: i. The good as end = what a rational person would desire for herself and others for its own sake. 2. What she would thus desire = happiness. 3. Therefore, the good = happiness. 4. A morality = a set of intrinsic attitudes toward certain act-types as such.

5. 6. 7.

The morally right = what is permitted or required by the morality a rational person would want for her society. The morality she would want for her society = the one that would expectably maximize the good (happiness) for it, if taught and prevalent. Therefore, the morally right = what is permitted or required by the morality (conscience) that would expectably maximize happiness in one's society, if taught and prevalent.

Here (1), (4), and (5) are proposed reforming definitions, (2) and (6) are synthetic equations or factual claims, and (3) and (7) represent conclusions inferred from what precedes. But, notice, it is not and cannot be concluded that the morally right = what will maximize utility, actual or expectable, in one's society. Of course, the whole set of equations is preceded by Brandt's definition of a rational person. But what makes him a utilitarian and distinguishes his kind of utilitarianism is (7). Unlike Mill's, the utilitarianism Brandt thus reaches in his more indirect way, via a certain concept of a rational person and a certain concept of morality, is clearly itself an indirect one; the rightness of an act or even of an act-type is not a function of its utility but of the utility of the teaching and prevalence in a society of the set of intrinsic moral attitudes (the code or conscience) to which it is related. Is it a form of rule-utilitarianism? Yes, but only if "rules" may be seen as intrinsic attitudes toward act-types in Brandt's sense. Perhaps it might be better to call it codal or conscience-utilitarianism, since it makes the moral quality of acts and act-types a function of the utility of codes or consciences, or, in other -201words, makes utility the criterion not of particular acts (as act-utilitarianism does) or even of act-types (as some kinds of rule-utilitarianism do), but of codes or consciences (or moralities) as these are defined by Brandt; I have heard Brandt himself suggest this. How Millian is this form of utilitarianism? It involves a happiness theory of the good or desirable and of the end to be promoted in some way by morality, but does it involve a Millian theory of the right or of morality itself? This depends a little on how one reads Mill. He says, near the end of chapter 4, that if (or since) happiness is the sole end of human action (and institutions), then the promotion of it "must be the criterion of morality", and let us assume that he means the promotion of the general happiness here. The question is: in what way does he think such utility should be the criterion of morality? Near the beginning of chapter 2 he writes that utilitarianism "holds that actions are right in proportion as they tend to promote happiness, wrong as they tend to produce the reverse." This can be read as suggesting the act-utilitarian view that utility is directly the criterion of the moral quality of actions. But this is not what Brandt holds. A few pages later Mill writes that "the standard of morality may ... be defined [as] 'the rules and precepts for human conduct' by the observance of which [a happy existence] might be, to the greatest extent possible, secured to all mankind, and not to them only, but ... to the whole sentient creation." This seems to express a kind of rule-utilitarianism, as I have

also heard Brandt say, one which applies the principle of utility, not directly to actions, but rather to rules or precepts for action; but if it does, it expresses a kind of rule‐ utilitarianism that Brandt rejects, viz. one that holds the right act to be one that conforms to the rules whose "observance" maximizes utility. Brandt's view is rather that the right act is one that is required or permitted by the code or conscience it is most useful for the members of one's society to be taught, to have, and to express in words to others, not just for them to conform to in action. He finds some support for it in Mill's chapters 2 and 5, as I have indicated, but is surely going well beyond Mill in his exposition of it, besides using a different and novel way of arguing for it. Still, it remains true that his moral theory is Millian in a broad sense. Indeed, I think it plausible that Mill would agree with it if he could be asked. There is another point to be mentioned here. Whatever Mill thinks is to be judged by considerations of utility or, as he calls it, "expediency"—whether it is actions, rules, traits, consciences, or moralities themselves—he thinks the utility to be considered extends to "all mankind" and even to "the whole sentient creation", and Moore thinks it must extend to the entire universe, not just to that of sentient beings. Millian utilitarianism rejects Moore's line and so does Brandt. But Brandt speaks only of the maximization of the utility of the society one expects to live in, not of all humanity, let alone of the whole sentient creation. Of course, he could argue that his kind of rational person would favor a code the -202having of which would maximize happiness for human kind or all sentient beings, but he does not; he usually seems to have a more limited scope in mind. In one passage (1992: 190-2) he raises this issue in an interesting way but still seems not to subscribe to Mill's wordings. Perhaps, then, one can even ask whether Brandt is a utilitarian at all, but I haven't the temerity to do that, though I do define utilitarianism in a universalistic way in my Ethics. Except perhaps for this point, Brandt is a Millian utilitarian but more clearly and explicitly an indirect one than Bentham and Mill were, though he is an indirect one of an original and unique kind, and arrives at his utilitarian theory of morality without making Mill's mistakes, by a certain carefully crafted series of steps. I think he has also answered pretty well most if not all of the recent criticisms of utilitarianism and rule-utilitarianism. But I must add that he does not really hold there to be no place at all for act-utilitarian thinking—that one should always decide what one morally should do by looking to one's moral code or set of intrinsic attitudes, and never by looking directly to see what act would probably maximize utility in a particular situation. Instead, he allows that there may be situations in which even the most rational or most utility-maximizing code or conscience would yield no satisfactory answer, or even situations in which one may go against one's code or conscience, however enlightened it may be. Indeed, he is more ready to allow this than is Hare, who is an act-utilitarian about the right! His claim is only that it is utility-maximizing for us to have, and normally act and judge according to, codes or consciences that rule out (or rule in) certain kinds of actions. 13. Of course, there

is the further question just what act-types our moral codes or consciences should rule out or in, and Brandt has done a good deal of work on it, but his answers go beyond the scope of this paper (see his 1992: 315-87). I should explain, however, that Brandt does maintain that his rational person would not favor for her society either a monistic moral code that had only the act-utilitarian rule that one should do what maximizes utility for one's society, or the ethical egoist rule that one should do what maximizes one's own utility. He recognizes that one could have a morality or conscience in his sense consisting entirely of either of these principles, but contends that neither of these moralities (or any other monistic one) would maximize welfare in a society, and that any morality seeking to do so must contain a plurality of rules or intrinsic attitudes to act-types more or less like Ross's list of prima facie duties, or, for that matter, ____________________ 13. In this connection, see especially his 1992: 206-12. Brandt now tells me that he thinks a conscience-utilitarian will make exceptions of cases in which serious harm would come from following her specific rules, presumably without revising these. This seems to mean he is not a pure indirect utilitarian. Does it mean he is inconsistent? Perhaps not, but if she judges it right to act against her conscience, can her judgment be a moral one in his sense? Mustn't her moral judgments express her intrinsic attitudes? Or is Brandt just assuming she will have an intrinsic aversion to the act-type of doing serious harm even by acting on one of her specific rules? -203more or less like the morality of our own society (which he would like to revise in important respects). Having mentioned Ross, I should add that Brandt thinks a rational utilitarian would regard all her duties as prima facie in Ross's sense, since they may conflict in some situations. She would not regard them as absolute in the way in which Berkeley thought some of his rules or "laws" were, especially that of passive obedience to rulers. There are many other parts or aspects of Brandt's moral philosophy that one might also put into perspective, e.g. what he says about blameworthiness, excuses, rights, and virtues, or about why we should have a morality in his sense at all, or live by it if we do. And, of course, one might go on to present and examine his arguments for his various moves, steps, and conclusions. Doing these things is not part of my project, but I do want to make a few brief comments in addition to those already made. (i) I very much admire Brandt's achievement, but still have the doubts about utilitarianism expressed in chapter 3 of my Ethics and believe they apply even to Brandt's very sophisticated form of that view, since it too holds that the ultimate end, however indirectly it is to be achieved, is or should be simply to maximize utility or happiness, without any special regard for the manner of its distribution. 14. (2) I am also bothered by the way in which he seems to make the right relative to what maximizes welfare for one's own society, as contrasted with making it a function of what does so for humankind or sentient being as a whole. (3)

I think individuals and societies should have moralities in his sense, but I doubt that morality can be reduced to the business of having such moralities, mainly because it seems to me that such moralities can be evaluated on moral grounds or by appeal to moral principles, much as legal systems can. I do like Brandt's replies to Williams and Scheffler but think there is something to Williams's notion of "the ethical" as opposed to "the moral", though I wish Williams had been clearer about what he has in mind, and am myself uncertain about what should be said. My remaining comments are somewhat different in nature. (4) One relates to a part of Brandt's argument, viz. that in which, as I put it, he claims that a person who is rational in his sense will in effect be a utilitarian in her choice of a morality, again in his sense, for her society. Actually, Brandt sometimes brings in benevolence as well as rationality at this point, leaving it somewhat unclear whether or not his argument depends on the person's being benevolent, and, if so, in what ways (1992: 21, 68f., 83, 141). (5) I said earlier that there is a question about Brandt's being a cognitivist and, in Moore's sense, a naturalist, and this needs a little explaining. There may even be a difficulty here. For, as was indicated, he proposes that we take "X is wrong" to mean "X is forbidden by the moral code a rational person would favor for her society"; on this understanding it has a naturalistic meaning and is a purely cognitive judgment, since Brandt ____________________ 14. See Sidgwick 1907: 416f. -204also proposes a naturalistic definition for "rational". On the other hand, he certainly thinks, as some non-cognitivists do, that having a morality in his sense normally involves making such judgments as "X is wrong", and that in making such judgments as part of one's morality, one is expressing an intrinsic attitude one has, not or at least not just reporting some kind of fact like "I have an intrinsic aversion to such acts" or "Such and such a kind of person would have one." Thus it could be argued that Brandt is really thinking in terms of two different uses of "right", "wrong", etc.: one as part of his normative theory and one as part of his concept of what goes on in a morality. I am not sure what to say about this, but perhaps he could reply that we should in our morality use "wrong" as meaning "A rational person (as defined) would ..." instead of using it vaguely or without that sense—that this is just what he is proposing. But even so, must he not somehow keep an emotive or expressive function for "wrong" as well as this cognitive one? Perhaps it could have a purely cognitive meaning in the ideal code but have an emotive one in any actual morality, but then the ideal code would seem not to be a moral one by Brandt's definition of a morality. 15. Brandt's contribution to the history of utilitarianism, as I have been picturing it, is a very considerable one: a largely original form of that doctrine argued for in a clear and largely original way. Both the form of it and the manner of argument, I have tried to show, are generically Millian but specifically very distinctively Brandtian, and free from Mill's own kinds of mistakes of psychology and logic. Finally, Mill said he could not prove his

utilitarianism, any more than anyone can prove any ethical first principle, but thought he could "prove" it in a larger sense by presenting his various kinds of considerations determining the intellect to assent to it. I am sure Brandt would also not pretend to have proved his new doctrine, and he might not want to use the word "proof" at all, even in Mill's wider sense. 16. But I think we may say that, not in what I have recounted here alone, but in his entire work, he has shown that there is a "credible" form of utilitarianism, more so than Mill's and many others, and more so than recent critics of utilitarianism have realized, one which should be taken very seriously. As Mill observes at the end of chapter 4, however, whether this is so or not "must now be left to the consideration of the thoughtful reader". ____________________ 15. In this comment, I am partly indebted to a luncheon remark by Allan Gibbard. In connection with it, see Brandt 1992: 2-4. 16. Actually, Brandt is (somewhat tentatively) a relativist of sorts about this; while he agrees with Mill that ethics is "within the cognizance of the rational faculty", he doubts that all of our intellects will be persuaded to assent to the same ethical doctrine even if they are rational in his sense. (He may regard my disagreement with him as proof of this. I do not.) One may also ask if Brandt is a moral realist This label is ambiguous and vague; all one can say is that he shares some views with those who now call themselves moral realists (cognitivism and naturalism) but denies others favored by some or most of them. He may even be some kind of constructivist, though this too is unclear (albeit no fault of his). -205-

13 Comments Richard B. Brandt I am most grateful to the contributors to this volume for doing me the honor of exploring and criticizing my thoughts about moral philosophy. And, needless to say, I am grateful to Bradford Hooker, who dreamed up this collection of essays and has assiduously brought it to fruition. I have learned something from every one of the papers collected here. In some cases, the authors have agreed with me rather a lot, and I have written correspondingly brief replies. In some other cases the disagreement expressed is so fundamental that I have again been rather brief, since I think I have nothing more to say than what I have already said. I think almost all of the critics have confined their attention mainly to my 1979 book, A Theory of the Good and the Right, and to the essays collected in my 1992 Morality, Utilitarianism, and Rights. This target is not exactly stationary; my thoughts have changed over the years, and still do. In fact, I am currently preparing a volume which in some ways is rather different from anything that went before—different in ways some of the critics will applaud. I will, however, for the most part confine my replies to a defense of the views expressed in my already published work. It may be helpful if I list what I regard as the central themes of my thinking. First, about the nonmorally good. I begin with the conception of desires and aversions, and then raise the question how they can be criticized. I take it a criticism is sound if it consists in the desire being faced, vividly and repeatedly, by known relevant facts; if this procedure results in the extinction of the desire, the desire is impugned and must be viewed as "irrational", and we do not want to call its target "intrinsically good". 1. This view seems to identify the good with desire-satisfaction, but I criticized this view in my 1979 book (see ch. 13). Lately I have leaned toward a "happiness" theory of the nonmorally good (Brandt ____________________ 1. I concede that this result does not show an inconsistency between a desire which fails and the facts about the results of this process, but I explain (1979: ch. 8) why a desire which fails should be ignored. Actually, I broaden and clarify this conception to some extent in Brandt 1993, where I show, among other things, why desires which "discount the future" are irrational. -2071989a: 59-66). Even more recently, thinking that both the happiness theory and the desire-satisfaction theory have part of the truth, I have been inclined to sit on the fence.

I have explained a person's "moral code" or "conscience" as consisting of his intrinsic aversions to actions of a given type (like telling a lie), a disposition to feel remorse or guilt if he offends against these aversions (without excuse), and to disapprove of others who offend against them (without excuse), and to think all these attitudes are justified in an appropriate way (see my 1979: 165ff.). (A "social" moral code is defined in terms of the codes of the individuals in the society.) But in what way can your moral code be "justified" to you? My answer has been that the way to do this is to show that the code is part of the "social moral code" which you would support if you were fully rational (in the sense explained above), were free from factual error and conceptual confusion (1979: 185ff.), and expected to spend your life in that society. I then went on to suggest it might be educationally beneficial to define "is morally wrong" as "would be prohibited by any moral code which all fully rational persons would tend to support" (1979: 194). However, in deference to the fact that not everybody will support exactly the same moral code, I suggested a relativist reading—i.e., "would be prohibited by any moral code which I [or you], if I [you] were fully rational, would tend to support ... if I [you] expected to spend a lifetime in that society" (1979: 194). This leaves open that a moral statement expresses the actual moral code of the speaker, since it would normally be inferred, from a given statement, that the speaker had a personal moral code of a corresponding sort. So far this conception leaves open the kind of moral code to which a rational person would subscribe. However, as explained in my 1988a (reprinted as ch. 8 in my 1992 collection) and in my 1979: ch. 15, I think a rational person would accept a form of "indirect optimific" theory. This kind of theory favors the social moral code whose prevalence would maximize happiness, desire-satisfaction, or some combination thereof. 2. A moral code endorsed by an indirect optimific theory would incorporate some exceptions when either the normal rules conflict (etc.) or there are wide differences of moral opinion in the society. Such a moral code must deal with such cases as that of the Hopi, where the recognized obligations of children to their father are hardly compatible with what would probably be an ideal system otherwise. And the code must be considered as a whole. I have conceded that it may make a difference whether a person is perfectly selfish, or perfectly benevolent, or benevolent only to a rational degree. In the case of a purely selfish person, it may not be possible to incorporate require____________________ 2. The consequences of a social moral code's "prevalence" include the costs of teaching and reinforcing that code. -208ments to help animals, fetuses, future generations, or contemporary persons who are unable to benefit us. Most clear-headed people probably have some (native) sympathy and so would tend to favor codes protecting these groups. But, although that is where I

personally would go, I do not claim to show that everyone would, and my critics have suggested that I do not really establish any form of utilitarianism which pays attention to the well-being of absolutely all sentient creatures. I have published work supplementing the foregoing with discussions of the concept of a "moral right", of "blameworthiness" and "excuses", of traits of character and of virtue. And I have discussed practical issues such as the morality of suicide, of the accepted "rules of war", of the treatment of defective newborns, and of optimal tax and welfare policies (see my 1992: chs. 10-19).

I. Comments on Nathanson Stephen Nathanson has written a very thoughtful account and criticism of my views, especially those about how to appraise a person's desires. He rightly reports my negative view of theories which just take some values as certified by "reason", but he disagrees with this negative view, partly because he thinks it represents an undue respect for scientific method as the only way of getting truth, but also because he thinks that, even worse, it is "foundationalist" in its theory of knowledge. It is true I regard "scientific method" as the only way to get knowledge; but I view scientific method as essentially seeking a theory which explains observational data. 3. I think the suggestion, by many, that science involves value judgments is indefensible. But here is hardly the place to go into this. Unlike some others, I do not regard judgments of the value of states of affairs as being self-evident. How important are achievement, understanding atomic physics, knowing French, the ability to evaluate works of art, and even pleasure? The answers are by no means obvious to me. What I would like to have is a theory about value which shows how scientific method would answer these questions. As briefly explained above, we may have desires that could not survive confrontation with knowledge obtained (however indirectly) by scientific method. And this sort of knowledge can tell us the most efficient plan for satisfying the desires that would survive. This is the approach to non-moral values that I developed in my 1979 book. ____________________ 3. Though I (a) admit that access to "observational data" itself is a matter of theory involving justified acceptance (e.g.) of the reliability of memory, and (b) do not hold that any observational data are matters of certainty but only of independent credibility. -209Incidentally, there I was not attempting to give an analysis of "rational" in its ordinary use. 4. That term could be dropped from my exposition. What I was trying to do was show how certified empirical psychology could be utilized in criticism of a person's goals.

Nathanson focuses his criticism on my suggestion that, if a person wants to pursue a Ph.D. degree only because she thinks her academic parents want her to, and if she would lose this desire if she came to see that they do not want that, then her desire is properly undermined, and the desire would vanish by well‐ known principles of extinction. Now I agree entirely that appraisal of a career choice should also include consideration whether one enjoys one's work, has other options, etc. But if a person pursued a career teaching economics solely because her parents preferred this, then if the force of this reason is overturned, her career choice does seem to have been a mistake. Nathanson alludes to the possibility that a person might remain fixed in his desire, despite being shown that his basic reason is mistaken and that all other considerations are negative. I can hardly deny this is possible; here, more investigation of the person's motivation is called for. Of course I do not deny that hedonistic considerations are relevant; I suggest full knowledge would not undermine them. However, Freud refused analgesics, in his last illness, since he preferred his suffering to losing the clarity of mind which was important for a major achievement. So desires for things other than pleasure are important to some people. But admitting this does not require us to deny that desires for most things occur because of contiguity conditioning with states pleasant in the past. I agree that I have been somewhat ambivalent about the status of pleasure, and Nathanson brings this out nicely. Nathanson himself, however, rejects hedonism in favor of a pluralism about values, but does not accept these just on the ground that people want them and they would withstand "cognitive psychotherapy". He thinks we can appraise them rationally, and when we do we are making substantive judgments of value. He says we may dismiss the values of another as being "inconsistent, harmful, immoral, bizarre, or based on false beliefs". But to do this is in part to employ value judgments of our own. Nathanson's approach seems intuitionist. As such, it ignores the need to provide a factual foundation for an appraisal of our values. 5. Nathanson thinks my desire for something more rests on a mistaken view of the logical foundations of science. Here I disagree, but this is not the place for me to spell out my views about the foundations of the scientific enterprise. ____________________ 4. Although I did try this earlier see Brandt 1983a: 143-64. 5. I do not mean to deny that there is knowledge of synthetic propositions by selfevidence. I agree with Langford 1944. But with respect to the implications of this for ethics, see Brandt 1944. -210-

II. Comments on Postow Professor Postow's comments on my work continue criticisms set forth in her very thoughtful 1989 paper. 6. Relevant to this controversy between us is an explanation I offered of "It would be irrational of you not to take an umbrella to the game, in view of the fact there is a 70% forecast of rain, and you dislike getting wet more than you dislike

lugging an umbrella", which I proposed as follows: "I hereby discommend that you not take an umbrella, while taking as my objective maximizing satisfaction of your transitive mood-independent ultimate desires, as they would be if they had been subjected to repeated vivid reflection, and having as my beliefs about options for actions and consequences those which are justified on your evidence—since this recommendation exemplifies a strategy for decision making which will in the long run satisfy your corrected desires as effectively as any other strategy." I still like this explanation, although I concede that many people would not recognize all of it as involved in their judgments about rationality. I take it the most controversial part of this is the part about "satisfaction of transitive mood‐ independent ultimate desires as they would be if subjected to repeated vivid reflection". I have not held that this is part of the ordinary conception of rationality, but I have thought that most people would—partly for some reasons I state, among others—want to ignore actual desires that could not survive being faced with facts. They would also want to ignore desires that are intransitive or mood‐ dependent (and would rely on desires they would have if they reflected more, with more knowledge). I called this process of reflection "cognitive psychotherapy". Postow rightly says that I think such reflection would not be critical of desires an agent has to avoid pain or seek novelty. She also defends my proposal that we should criticize desires by cognitive psychotherapy, against various objections that have been raised. 7. Postow, however, regards it as a fatal objection to my view that a person, employing cognitive psychotherapy, might be led to reject going to dinner parties by reflection on people's innards. 8. I don't see why she is moved by this suggestion; such reflection seems to me to be irrelevant 9. and to ignore the attractive features of going to dinner parties. Of course, there sometimes are good reasons for not attending a dinner party. One may have recently had a fight with someone who will be there. But innards are irrelevant, unless prone to malfunction at dinner parties in socially disruptive ways. ____________________ 6. Brandt 1989c is a reply to that paper. 7. Railton 1986: § 2 also defends me against such objections. 8. As she notes, here she is following a suggestion of Gibbard's. 9. See Brandt 1979: 111-15. -211I might add that my own conception of "cognitive psychotherapy" has undergone some modification, as I explain below at the end of my discussion of Gibbard. In her earlier paper she posed several other difficulties for my view, and argued that I actually appeal to "intuitions" myself, although I officially abjure this. There is much to be said in my defense, but not here. Nevertheless, I concede that her criticisms merit serious consideration, and some of them perhaps call for tightening the view I put forward.

Postow's approach to practical rationality is rather like the method of searching for wide reflective equilibrium in morals. She proposes that a principle about action, e.g., "An agent's desire for a thing that she natively likes provides that agent with a reason for action", may be considered valid if it "makes maximally good sense of and in light of relevant data", such as not only data about what the agent desires, thinks valuable, and would desire after ideal deliberation, but also facts and theories about learning, motivation, mental illness, personal continuity, etc. She proposes numerous examples of relationships to such data which would render a given principle one that "makes maximally good sense of and in light of relevant data". The major differences she cites between the application of her proposal and that of what I called "cognitive psychotherapy" is that hers would recognize reasons "intuitively rational" for the agent when cognitive psychotherapy would reject them, and that hers would reject reasons that cognitive psychotherapy would approve but which are "intuitively irrational" for the agent. She ends by saying correctly that her theory merits further investigation. Still, her theory is incomplete, with the result that it is hard to know how it could be applied—a problem it shares with Wide Reflective Equilibrium.

III. Comments on Pettit and Smith Pettit and Smith offer various criticisms of and amendments to proposals I have made about rational choice and the nature of self-control. Some of these I find helpful and hereby adopt into my conceptual framework. I think we agree that self-control is important when one thinks it is irrational to do/want certain things, and already has some motivation not to do them: to eat in a way contrary to health, to watch TV instead of work, put others down, lose one's temper, etc. Although some motivation is there, it is often not strong enough to control behavior: one thinks it irrational to overeat and put on weight, but finds oneself guzzling chocolates, perhaps partly because, at the moment of choice, the ideal of health (or of pleasure over the long term) is too abstract or remote to be effective. So what can one do? As I see it (and as therapists see it) there is some motivation toward not doing these things: enough so that, if one knows about this pos -212sibly efficacious option, one may visit a therapist. On the therapist's advice, one may adopt a program of eating sparingly just for today (one day at a time), of joining a weight-watcher's group and risking being shamed if one has to confess overeating, of congratulating and rewarding oneself if one does well. As a result of gradual success, one might learn that one can cope with the problem. Ultimately, the initial disparity between the restrictions one wants and thinks it rational to adopt, and strength of motivation to adopt them, is normally (we hope) removed. Do we want to call this new state, brought about as suggested, one of "self-control"? I think so, especially if we remind ourselves that it succeeds against the pull of contrary present desires.

Pettit and Smith want to make an amendment here with which I agree. Their amendment is similar to one suggested by Gibbard some time ago; Gibbard pointed out that if I am lost in the woods, it is not very helpful to be told that it is rational to do what I would do if I were fully informed about the facts (such as where I am and the best way to get out). What is really rational is for me to adopt a strategy which a fully informed person would advise for one in my actual situation of ignorance, e.g., maybe just walk always in a downhill direction. So Pettit and Smith urge that my conception of "rationality" does not pay attention to the actual agent's particular circumstances, e.g., his belief that he is Jesus Christ, or that he has wayward desires. Obviously, a fully informed person would not believe he is Jesus Christ, and presumably would want his actual self disabused of any such belief and any desires dependent on it. If he recognizes that his actual self does have this belief, and if the only way to be disabused would be to visit a psychiatrist, it would be rational for him to recommend that to his actual self. And, for the case of a person who rationally wants to quit guzzling chocolates, it is rational for him to go to a therapist and take advice about how to overcome this temptation. Pettit and Smith are right that a fully informed and rational person will aim to make recommendations to his actual self about what to want/do, in full awareness of the nature and desires of his actual (but improvable) self, with all its limitations, including right now the inability to stop eating chocolates. Doing this may also make it rational to change his mind about what it would be best to do if he were not taking his actual self into account: perhaps the pleasure of eating chocolates is so great that it is best, given his actual psychological makeup, to lead a life which includes this. The same would be true for parachute jumping, for someone who finds himself painfully terrified when the time to jump approaches. I have not, however, supposed that a person who seeks help from a therapist to overcome his eating problem has an extra intrinsic desire. He does have an existing desire to overcome, say, his eating problem, but though, as things are, that desire is not strong enough to resist chocolates, it is strong enough to motivate him to go see a therapist as a means. (There are other things he might do which would be helpful, e.g., stop thinking vividly of the taste of chocolates.) How is it -213possible that someone may now be motivated to visit a therapist in order to control eating habits, but not able to stop eating? The answer is that this is just the way things are! What then do I mean by a state of "self-control"? I mean that the discrepancy has been removed between what one would, if fully rational (and of course, informed), prefer that one's actual self do and what one is most motivated to do as one is. But this self-control can be needed with respect to quite different things such as diet, temper, physical fear; fear of rejection, and so on indefinitely—and a person may be self-controlled in one area but not in others. I must confess, however, that what I just claimed about the state of self‐ control hardly squares with my own original definition of a "virtue" as a "relatively unchanging disposition to desire an action of a certain sort ... for its own sake". As I define the

improved state of affairs, that disposition is not unchanging (it didn't exist before, and may fade out). Moreover, the psychological qualities involved in leading up to this state —what we might call "self-control" in a different sense—include not only having motivation, but also knowing about options and understanding and respecting the advice of the therapist. So I should really not explain self-control as a straight motivational disposition, i.e., as the greater strength, in some area, of desire to abstain from doing what one thinks is irrational. Perhaps what I should have purported to be explaining is just the notion of "having gained self-control" in a certain area. I could then have added a separate account of the personal qualities involved in getting this kind of self‐ control. Pettit and Smith may wish to define "self-control" as a general disposition and ability to take whatever steps are useful in order to extinguish the discrepancy between what one rationally would want or choose one's actual limited self to want and do, and what one, as one is, wants or chooses to do. This has the virtue of being a unitary conception of selfcontrol. Perhaps it is the most useful definition of "self-control". On the other hand, it ignores the actual final state when the discrepancy has disappeared. Furthermore, we should question whether there is any general trait of self-control of this sort. Perhaps we think there is if we think "self-controlled" people are those who almost always in the end succeed in wanting and doing what they think it is rational for them to want and do, just as they are, without taking any steps to change themselves. But it is not clear that the qualities needed to achieve self-control (my sense) in one area bring self‐ control in other areas. Is there a general "disposition and ability" involved in getting self-control in all areas? Of course, I fully agree that "gaining self-control" involves cognitive components: e.g., knowing there are therapists, maybe knowing some psychological theory about the effects of reinforcement and of finding that one can cope, etc. Indeed, it is not necessary to see a therapist; if one knows enough psychological -214theory, one will think of useful ways to get the same result. These facts support Pettit and Smith's more cognitive and managerial (but diversified?) picture of self-control as an account of what is involved in getting to the final state where there is no disparity between actual desires and rational ones. In general, I am inclined to agree with the amendments that Pettit and Smith offer or imply—especially their insistence on its being rational to take the nature and limitations of one's actual self into account in deciding what it is rational to recommend that one's actual self want and do. But I would exclude their supposition that taking steps to avoid being governed by irrational desires/actions must somehow involve an "extra intrinsic desire". I also accept that my original affirmation that self-control is a purely motivational substantive disposition like truthfulness or kindness was too simple. And I agree that taking the steps necessary to bring actual motivation into line with rational motivation is itself rational, or even "rationally required". But I think that I did not earlier doubt this.

IV. Comments on Davis Michael Davis's paper, like all Davis's earlier work, is ingenious. He proposes that my "cognitive psychotherapy" conception of rational desire/action provides a reasonably good analysis of the concept of "autonomy", where "autonomous desire" is taken to exclude manipulation (e.g., by a hypnotist or misleading advertising), compulsions (strong irrational motivations), and "bad" socialization (which sets a distorting screen between us and the way the world really is). Cognitive psychotherapy, according to my conception, involves vivid, repeated exposure of a given desire to all relevant information, the thought being that this association will, according to motivation theory, effect the extinction of unacceptable desires. Clearly, this kind of therapy eliminates the influence of manipulations (say, displacing my desire for chocolate chip cookies produced by advertising in favor of a desire to abstain from unhealthy foods). This therapy would also at least weaken compulsive motives. And, as for "bad" socialization, cognitive psychotherapy would displace (e.g.) a desire for success at any cost if this desire results from early family disapproval of failure. It would also displace a housewife's servility in favor of a more assertive attitude that would be reinforced in a more favorable environment. "Good" socialization is thus socialization which survives being fully faced with available information. A reason we prefer changes arising from this "cognitive psychotherapy" is that we want to have desires that result from "sensitive reaction with the real world". This conception seems immune to various objections that might be raised. One such objection is that it does not require a desire to have survived previous criticism. But must an autonomous desire have undergone previous criticism? No, it is enough if it would withstand such criticism now. -215Another possible objection is that my theory seems not to condemn the effects of coercion. Suppose a robber demands your money. Such coercion is not manipulating in the sense of changing your view of the world; it is changing the world, attacking your liberty not your autonomy. And it should be condemned as a violation of liberty, not as a violation of autonomy. What do we say to a hypnotist who saves me from my irrational desire to eat chocolate? Is he invading my autonomy? No: the effect of this is to increase autonomy, since the result is more in line with what cognitive therapy would accomplish. But the situation is different if the hypnotist has acted without my consent; then he has usurped my right to juridical autonomy. Here, Davis points out, we must distinguish between factual autonomy and the "juridical" sense—the right not to have one's factual autonomy invaded. Davis points out that it might be objected that this psychotherapy conception of autonomy allows that animals may be autonomous; but that is not objectionable (why should we not allow that a dog or cat has autonomous desires?) since it leaves open how much we are bound to respect their autonomy—their cases differing from those of the retarded, aged, or children.

Thus far I have largely summarized what Davis has to say, with approval. His distinctions seem to me fair enough. Whether my conception of "cognitive psychotherapy" provides a reasonable way to analyze "autonomous" desire or action is another question, one which I have not intended to address. What I have wanted to do is provide a conceptual scheme which would enable us to identify rational—information based—desires and actions. The conception I have developed does seem to me promising as an analysis of "autonomous desire" and "autonomous action". But I leave it to Davis to defend the conception! In earlier work, Davis expressed serious doubts about the normative force of cognitive psychotherapy (Davis 1987). Indeed, he thinks my conception allows that cognitive psychotherapy occur in a "social vacuum", unaffected by the fact that we have to live with people firmly committed to certain moral prohibitions, as if the attitudes of people around us are not important parts of our real world. For example, if I submitted my aversion to incest within the nuclear family to the proposed form of cognitive psychotherapy in abstraction from attention to the attitudes of those around me, I might rid myself of the aversion, whereas I certainly could not have done this if I conducted my "therapy" in full knowledge of other people's attitudes toward me and my aversion. If this objection is well‐ taken, it would have the effect that various counterexamples to utilitarianism which I have rejected on the ground they could not withstand cognitive psychotherapy could no longer be dismissed and I could no longer regard the choice of an indirect utilitarian moral code as rational! I am unrepentant on this score. Moral objection to incest within the nuclear family is based on the predictably disastrous effects of permitting it. So permit -216ting it can be criticized on indirect-utilitarian grounds. We need not worry, therefore, about whether cognitive psychotherapy would disabuse people of their opposition to such incest if we ignored this fact about it. It is not clear to me that there are any similar apparent counterexamples to utilitarianism which cannot be appraised in a similar way, and thus which stand in the way of an indirect utilitarian moral code being viewed as rational. (There are other problems with justifying an indirect utilitarian code; see my reply below to Kavka.) Still, Davis's reasoning in that earlier paper is typically inventive and deserves reflection.

V. Comments on Hare Let me begin by expressing my appreciation for Professor Hare's generous and careful review of my work—and for the way he does moral philosophy in general. His work is always informed, clear, careful of details, and insightful. I have learned a great deal from him. Hare and I share objections to certain ways of doing moral philosophy. We both reject "old-line" naturalism, according to which terms like "good" and "morally right" are synonymous with some expression which, if I may use this phrase, can be explained in an

"empiricist descriptivist language". Likewise, we both reject non-naturalist descriptivism, according to which these terms name properties, not explainable in an empiricist language, about which we can have intuitive or self-evident knowledge. Moreover, Hare and I agree that having a "moral code" consists in having aversions to performing acts of a specific type (e.g., telling a lie), of being disposed to feel discomfort (guilt, remorse, etc. or at least compunction) in response to awareness that one has infringed these aversions (without some "excuse"), and of being disposed to disapprove of others (feel anger, coolness, etc.) who do so without "excuse", and to think that such stances can be "justified" in some appropriate way. A person's set of attitudes, of this sort, toward forms of conduct is at least roughly what Hare means by a person's "intuitive morality". I agree with Hare that when a person employs moral language he is (usually) expressing such a moral code, at least in the sense that others who understand the language, on hearing him say (apparently sincerely) some act is morally wrong would take it that he had a moral code unfavorable to that action—and that the speaker would not say this unless he had, or thought he had, a moral code condemning this sort of action. There is more agreement between Hare and me, and in fact I am now inclined to accept some, but not all, of his criticisms of the reasoning on which I relied to show the importance of "reforming definitions" of moral terms as distinct from relying on "linguistic intuitions". With so much agreed, where do we disagree? -217The major disagreement between us concerns the theory of the justification of moral judgments and attitudes. My view has been that the content of such a code is justified for a person if it can be shown that he, when fully informed of the facts, and with "rational" nonmoral motivations, would be willing to support such a code (or a new one) for his society if he expected himself and his children to spend a lifetime in that society. His relevant nonmoral motivations will include not only desires not to be injured in various ways (assaulted, raped, defamed) but also the degree of sympathetic altruism which is "rational" for him. An objection to this proposal, an objection emphasized in Kavka's contribution to this volume and apparently seconded by Hare, is that such a theory will not reach a morality impartial between selfish agents and those unable to do as much for them in return (e.g., the poor, animals, future generations, fetuses, maybe small children). We should discuss how far my proposal will reach, but a morality to some extent so circumscribed may be all we can rationally justify. In order to appraise my proposal about justification, I will compare it with Hare's view, which may turn out to be very similar. Hare's view is that we can base justification of a moral code, mostly, on "critical moral thinking", which applies to cases where there is conflict of interests. The force of "critical thinking", he says, derives from the meaning of moral judgments, which Hare takes to express overriding universalizable prescriptions, or imperatives, which the speaker is

resolved to obey himself. Overridingness I shall ignore, since it makes no difference to the argument. Moral judgments are universalizable in the sense that they apply to everyone and contain no proper names—so "It is wrong to do A in C" is to be construed as "Let no one in circumstances C do A!" There is another element in this type of thinking: a psychological principle we shall come to, which has been called, by Gibbard (1988), the "conditional reflection principle". With how much of this need I disagree? I need not question the part about moral judgments involving prescriptions, since, as we have seen, a moral code involves aversions, which entail prescriptions. As for universalizability, my conception of the choice of a moral code is the choice of one for "his society" (and I see no reason why this should not be construed as world-wide, where the problems are essentially world-wide). 10. Further, since the code is to concern act-types it will not involve proper names. So far there seems to be no disagreement. There is, however, the role of the "conditional reflection principle". We shall come to this shortly. Hare explains that moral judgments are universalizable in the sense "that they entail identical judgments about all cases identical in their universal proper____________________ 10. See Brandt 1983b. -218ties" (Hare 1981: 108). 11. He holds this on the basis of a claim that, were a person to say that an act A was right but another, with the same universal properties, was wrong, we should react by the same kind of incomprehension as we would to hearing someone say that two figures are exactly the same but one is triangular and the other not. This reaction, he says, is a reliable clue to the meaning of the moral statement. I shall not contest this claim. So far, this is rather, but not exactly, like my suggestion that moral attitudes are directed only to act-types, definable in abstract terms. But it is also different, since what makes two actions "the same" is defined more narrowly. Indeed, it seems to require that the "act" be construed only in terms of its impact on the recipient's experiences. Thus Hare (1981: 108) says universalizability implies that "if I say that I ought to do a certain thing to a certain person, I am committed to the view that the very same thing ought to be done to me, were I in exactly his situation, including having the same personal characteristics and in particular the same motivational states." My view is not far from this. I hold that "moral" attitudes are attitudes toward act-types because these—like attitudes to telling lies, theft, injury to others—are the ones we call "moral". But there are also pragmatic reasons for restricting a "moral code" to attitudes about act-types. As Charles Stevenson pointed out in his last publication (1983), a

speaker could not expect to be taken seriously if he insisted on making "self-excepting" moral judgments. Moreover, it might be hard or impossible to learn moral attitudes if the act-type is permitted to vary purely because of the individuals involved, and the ability of moral judgments to be action-guiding would be impaired. So "same act-type" cannot vary just with individuals, but it seems it must be explained somehow, to make clear what qualifies and what does not. Presumably, as an act-type, it will be explained at least as limited to what one does, causes, or brings about. Hare thinks acceptance of universalizable prescriptions in his sense is important. He thinks one reason for its importance is that nearly everyone would agree that in the choice of moral principles we must be "impartial" among all sentient creatures (including animals—though Kant might query this); and he thinks the stated conceptions, properly understood, will secure such impartiality, and that nothing short of them will. Now if we restrict an "act-type" to causing a kind of experience, as Hare seems to do, the effect appears to be that whether another individual is of a different race or nationality, or a child or an animal is not included in the concept of a ____________________ 11. Cf. my criticisms of Hare in Seanor and Fotion 1988. Hare has changed his position considerably since that volume; see at the end of his critique of me that volume. For this and other reasons (some of them the force of Hare's replies), I do not endorse all the arguments I put forward in that volume. -219type of action toward this individual. So, if we universalize in this restricted way, we shall treat members of weaker groups equally—we cannot view treating them in a certain way as right if we would view the same treatment for ourselves as wrong, the situations being identical. This seems to secure the kind of impartiality in the treatment of these groups that we moral philosophers want to secure. Apparently, what brings impartiality is this restriction, not universalizability as such, which Hare thinks depends on the meaning of moral terms. We should notice that it would be consistent with my theory of justification to make the same restriction, although, as we shall see in a moment, there is some question whether we would want to do this in all cases of moral judgments. However, the most distinctive part of Hare's view is his reliance on the conditional reflection principle. This principle affirms that if I understand fully and vividly—as Hare thinks is necessary for a rational appraisal—the harmful impact of my action on someone else, I will also want (prescribe) that this same act not be done to me were I forthwith in that person's situation. As Hare puts it (1991: 109), when I consider doing what would injure another person, then

if I fully represent to myself his situation [as a result of my action, which Hare supposes would be unpleasant or aversive], including his motivations, I shall myself acquire a corresponding motivation, which would be expressed in the prescription that the same thing not be done to me, were I to be forthwith in just that situation. But this prescription is inconsistent with my original 'ought' judgment, if that was ... prescriptive. For ... the judgment that I ought to do it to him commits me [by universalizability] to the view that it ought to be done to me, were I in his situation. So universalizability (in the sense restricting act-types to causing experiences) plus the further principle about what motivations I shall acquire in a given situation, has the implication that I must abandon my original 'ought' judgment (if my desire that this not be done to me is stronger than my desire to do it to the other person). If we take all this into account, how does "critical moral reasoning" go? (i) If I am considering whether it is morally permissible for me to do a certain act to you (viz., whether I can universally prescribe my doing this to you), I first identify how this will affect your experiences. (2) I then determine how strongly you will wish this not to be occurring to you. (3) I then represent to myself this impact on you, and the strength of your motivation in response to it, and as a result of the "conditional reflection" principle, I shall find I don't want to be in your position, your experiences and motivations taken into account, any more than you do. (Indeed, my reaction to being in your position with your desires and experiences will exactly mirror yours.) (4) So I shall find that I cannot want that this sort of experience happen to me. (5) But, in virtue of the principle of universalizability, I cannot prescribe doing—or think it right to do—what I was -220thinking of doing to you without also prescribing that the same thing be done to me, since the two events are identical. (6) So, if my desire to do what I was contemplating doing to another person is weaker than my desire that the same sort of experience not be caused in me, I have to give up prescribing—and doing—what I was thinking of doing to the other. (If, on the other hand, the response of the other person is non-negative or at least weaker than my desire to do what I had considered doing, then I am morally free to do it.) Evidently, the psychological principle of conditional reflection is crucial to this whole process. Let us now appraise the status of this principle. First, we might ask whether a person's motivational response to my action must be viewed as part of what I "do to him". This seems doubtful in light of the fact that the patient of my action may have all sorts of irrational preferences. Suppose the other person strongly wants not to suffer the slightest bit of pain. How should I take account of this? Must I follow his preference, even if I do not mind a bit of pain and think his attitude is irrational? Why should I take on such preferences? It is true that, if I do not, I am not universalizing to the case of a person with his disposition to have irrational motivations, and so far not to him as he is. (Hare says I must take into account the fact that a cyclist does not want his bicycle moved, for any

reason at all or no reason; so the cyclist may be taking a quite irrational attitude.) Now Hare's defense of the universalizability principle is the supposed lack of comprehension of a judgment that it might be right for someone to do this to me, but wrong for me to do it to him. But should "universalizability" be defined so that the possibly irrational motivational reaction of a person to an act is part of what "I do"? No, "my act" must surely be defined more narrowly, so as at least to exclude abnormal responses, and include only psychologically necessary ones. 12. In any case we must ask whether this principle is true. Is it true that when I know how strongly the patient will prefer (however irrationally) that the act (including its impact on him) not be performed, I shall have an equally strong preference now that this not be done were I in his position, and therefore must judge (by universalizability) that I ought not to do it to him? Hare thinks that if I do not share the patient's preference, this shows I did not represent his situation to myself as fully and vividly as is necessary for rational appraisal—in short, he thinks it shows I did not fully and vividly understand the other person's situation. This seems to me an unwarranted definitional fiat about "fully and vividly understand". 13. Moreover, to make a different-point, does it even follow from the ____________________ 12. Hare seems to reject this kind of criticism; see his 1992: 166. 13. Hare thinks that he leans toward the view that if one represents to oneself fully one's "future preferences" they will be "replicated in the present" (Hare 1981: 94, fn. 4). To be sure, it can be simply stipulated that one does not represent (and appreciate) "fully" unless this effect occurs. Furthermore, I agree that dear failure to represent vividly may be reflected in one's motivation. But the view that -221fact I know I would dislike a person's situation if I were in it, that I shall now want not to be in it to the same degree as he does, or even as much as I think I would if I were in the situation—especially if I am clearly aware of the fact that I am not going to be in that position? (To adopt Hare's answer is a bit like saying that, knowing I shall not like it if I am fat a year from now, I shall be aversive to that situation equally strongly now— strongly enough to restrict my diet. 14. ) What all this shows is that Hare's view requires, in addition to universalizability, some views about psychology that we may reject. One may ask, moreover, how essential this principle really is for an appraisal of a person's moral code. On my conception, a person will support a moral code for society which will prohibit things being done to him which he will strongly dislike, on grounds of self-interest. And, since a "social moral code" will include his own moral code as an instance, it will therefore prohibit his doing the same sort of thing to others. (This may not involve accepting the irrational dislikes of others.) But suppose there is some kind of experience which the agent does not dislike, but someone else does, not necessarily irrationally. For instance, we are told that a kind of surgery can have the effect that a person no longer minds pain. If someone has had such surgery and so does not object to

pain, would this mean he need not include a prohibition of causing pain in his moral code, and so he could feel morally free to cause pain to others? What the principle of conditional reflection seems to me to do, then, is secure that, if a patient of an action dislikes something, the agent will necessarily also dislike being in the position of the patient, and so will want to prohibit causing this experience in someone else. So a person, rationally choosing which moral code to support, would want to include a prohibition of this. Insofar as people are alike in what they dislike, the principle does not add anything. How far are they alike? Aside from the possible case of surgery preventing minding pain, there are possibly some cases: some people may not like to be tickled; by conditioning some may not like certain kinds of food; some may not like to have to be alone. We might say, however, that every agent will want to avoid unpleasant states of himself and, if so, he will want a moral code prohibiting others from causing these in him, and his causing them in others. What the principle of conditional reflection does, then, is provide a warranty that an agent will want morality to prohibit people's causing states of affairs disagreeable to somebody. How important this is seems open to debate, but it certainly plays a role in Hare's argument. Hare explains how his method of reasoning can be extended to multilateral cases. I can compare (he thinks) my desire to perform a certain action, with the ____________________ representing to oneself fully one's future preferences necessarily replicates them in the present seems to me contrary to experience. 14. Cf. Hare's reply at Seanor and Fotion 1988: 216. I remain unconvinced. -222total harm of that action on all the others affected. He observes (1981: 111), "in principle unanimity can be reached by our method of reasoning, once each fully represents to himself the situation of the other." This result leads to impartiality in the choice of a moral system.. It also tends to lead to act-utilitarianism. 15. The type of reasoning Hare seems to use for paradigm cases seems to concern only hurting or helping others, e.g., my causing unpleasant experiences for somebody but being unwilling to have the same thing done to me (as universalizability requires). Certainly these are the types of case to which the foregoing argument applies—what he calls the "Golden Rule" argument. But does this kind of "critical thinking" apply easily to all kinds of moral evaluation? Hare allows that divergence of ideals is possible, and he concedes that, when there is, not much of an argument of any sort is available to settle a dispute (Hare 1963: 146-56). But, further, suppose I feel free to kill an enemy soldier because of his status as an enemy soldier, although he does not want to be killed; is this

then to be condemned? Hare might reply, as he did earlier to the case of a judge handing down a nasty sentence, that there are many persons whose desires are involved. In the sentencing case, there are those who want to be protected from crime. In the war case, there are comrades, citizens to be liberated or protected, etc. But this answer makes moral assessment very complicated. More important still, consider moral issues about sexual freedom (e.g., the prohibition of incest), property rights, promises ("A promise is a promise" even if breach involves no injury), suicide, eating the bodies of dead human beings, and dueling. For at least some of these cases, and possibly many more, my proposal about how to criticize one's moral code seems more applicable than Hare's device, which seems to require fitting them (so far as any argument is possible) into the paradigm of causing distress to someone. I concede, however, that a sensible attitude toward at least most of these can best be defended by reference to the long-term benefits/costs for various classes of individuals. How does this kind of "critical" moral thought relate to a person's "intuitive" morality? Hare thinks that the results of "critical" thinking are (somehow) decisive if intuitive morality has conflicting demands, or if somehow intuitive morality doesn't seem to fit the situation, or when we are deciding what kind of intuitive morality to teach our children. Hare thinks that we should teach an intuitive morality, but which one? We are to decide this by taking into account how seriously detrimental given types of action would be, how frequently the occasion for them would arise, how hard it will be to teach the moral prohibitions forbidding these types of action, how likely people are actually to obey these prohibitions, how inconvenient living by these prohibitions would be for agents, ____________________ 15. Only tends, since the agent may have strong aversions to other things, besides the distress of the people her acts affect—she may be averse to breaking her promises, or to inequalities in well‐ being. -223and what the "expectation" effects (emphasized by Harsanyi) would be (see Hare: 1988: 224-28). This raises the question of what I morally ought to do if the requirements of "critical" thinking and my intuitive morality conflict. I myself find it hard to believe that I have done the moral thing if I have not followed the instructions of my "intuitive" morality, in so far as, after careful reflection, I think it is justified. This consideration raises the next point I wish to discuss. Am I right in describing (and defending) my view as sponsoring a "one-level" view of morality—not two, as in Hare's "critical" and "intuitive" levels? Hare thinks we need to bring in "critical" morality especially when principles of intuitive morality give conflicting directives, or when one of its requirements seems not to fit a certain kind of case. I agree with Hare that thinking is called for in these situations.

Suppose two principles conflict. This is like two laws giving conflicting directives. In the law, we leave matters to an appellate court, which, I suggest, comes out with an "interpretation" which the court thinks makes a useful long-term precedent and also preserves analogies in the law. In morals, when principles conflict, what I think we have to do is to ask ourselves whether one principle should be given priority, or perhaps whether a more sophisticated general principle should be adopted as a replacement for both, on the ground that the teaching and prevalence either of the new ranking or of a more sophisticated replacement principle would be better in the long run. It might be better even if it is too complicated to build into the conscious emotional attitudes of everyone. 16. The new principle is to be accepted for the same types of reason that justified the originals, that it appeals to informed self-interest and rational sympathetic altruism. (One might substitute for this Hare's conception of "critical" thinking.) So, henceforth, the person who has thought will have the new principle in her moral code, not quite as conscious and automatic as the old principles, but still there as part of her "effective code". It may well be that Hare agrees with this. This is very different from saying that one should resolve all conflicts of moral principles by going to an act-utilitarian resolution, as Hare seems to on occasion. And it is not saying that a sound "intuitive" morality must direct us just to do the act-utilitarian best thing, a point Hare himself concedes, since he thinks an intuitive morality is to be appraised by its "acceptance-utility", not by whether its prescriptions about action coincide with those of act-utilitarianism. We might agree with him, moreover, that a person's choice of which moral code to support for her society, in view of her nonmoral motivations and an expectation of ____________________ 16. I have mentioned that a moral principle may be more or less salient. Some may not be as salient as ones about lies, theft, direct injuries. They may concern behaviors that occur less frequently, may not be taught as early in life, may just be more complicated, and may not be strongly connected with guilt feelings or indignation about others. For instance, a principle about the distribution of income. -224spending a life in that society, will arguably be framed along utilitarian lines— she will choose the code the teaching and prevalence of which will, everything taken into consideration, maximize public well-being. However, it looks as if Hare is willing to set intuitive principles to one side, as compared with critical thinking, in problem cases where principles conflict or seem not to fit. He says that the notorious sheriff should hang an innocent man if he knows it will maximize expectable utility, but "in practice he never will know this". And doctors should murder a lonely, drunken, but apparently healthy beggar if his heart and kidneys are needed to save two other patients, if the doctors are really sure of the facts. This level of assurance, however, "will not be forthcoming in many situations, if at all" (Hare 1981: 133-5, 164).

What practical difference does my proposal make? Suppose we could cause some harm to someone. Act-utilitarians would think they would be right to cause it if the advantages to themselves of doing so are even just slightly larger than the harm to the other person. Non-act-utilitarians might well think it would be better, in the long run, for people to be motivated to restrain themselves in this kind of situation. Hare would seem to go for the act-utilitarian resolution of this case, whereas if we considered the long-run desirable motivation, we would go the other way. What would I say about the case Hare describes as a difficulty for me—the case of the homosexual friend in an intolerant society? Hare says: "I would certainly, as a rational person, support a moral code in which people could speak frankly about other people's sexual orientation, if I were going to spend a lifetime under it." But, he queries, what might I, in consistency, say about what I ought to do if, in that society, homosexuals are put in prison? Hare wonders whether, in that case, I would still say it is morally permissible to speak openly of my friend's homosexuality. My rough answer, formulated frequently in the past, 17. is that the truly optimal code would be disjunctive: it would provide that one be free to speak except when this would cause disastrous harm, and that one work for a situation in which it is possible harmlessly to speak freely. This is like saying: normally keep your promises, but if so doing would be very seriously hurtful, then don't at least for certain types of situation, since there is a stronger obligation not to cause so much hurt, and, if you like, make a corresponding restriction in your rule about promise-keeping. Incidentally, I don't think incorporation of this disjunctive rule would make morality unacceptably complicated; I think ordinary morality already has such disjunctive rules! I agree with Hare that in various situations the learned aversions (dispositions to feel guilty, etc.) are not enough; they must be amplified by reflection about what kind of principle would be optimal, in view of the costs or benefits of its in____________________ 17. Brandt 1979: 297-99; 1983c: 98-99; 1988a: 358-59; 1989b: 96-97. These last two references are reprinted in my 1992: 87-88, 155-56. -225corporation in one's society. But, so construed, my proposal may well have an advantage over Hare's two-level code, at least if the "critical" level is act‐ utilitarian and has the priority Hare sometimes gives it. Despite the weight Hare gives to "intuitive" moral principles, he seems to be really an act-utilitarian about what is right, whereas I am an unabashed indirect utilitarian, of a kind incorporating "critical" thinking into a one-level moral code. Maybe there is really no disagreement here. 18. I won't comment here on Hare's remarks about hedonism, since I indicate my current position on this in my replies to Sumner and to Harsanyi.

VI. Comments on Gibbard Allan Gibbard begins his critique with some reminiscences about student life at Swarthmore College. Students like Gibbard (fall, 1962), Newton Garver, Gil Harman, Richard Henson, David Lewis, Steve Nathanson, John Troyer, and Peter Unger made the place exciting for the instructors. Gibbard wrote a piece for my seminar which he revised while in the Peace Corps in Ghana before going to graduate school, and published in the Australasian Journal, under the title "Rule Utilitarianism: Merely an Illusory Alternative?". In his contribution to this volume, his criticisms of me go deeper. Gibbard's first critical comment concerns a possible circularity in my explanation of a moral code being "justified". I have supposed that a person's morality consists roughly of aversions to acting in certain ways definable in abstract terms (lies, injuries of others), dispositions to feel guilty if one infringes without excuse and to disapprove of others who do, plus a belief that the foregoing are justified in some appropriate way. I did not specify what that way is. I left the concept as open as when John Locke (1706: bk. 1, ch. 3) said, "I think there cannot any one moral rule be proposed whereof a man may not justly demand a reason...." Thus, a person might offer as a "justification" that the rule is a matter of divine will, or that it just formulates what "we" do (as when the Hopi say, "It is the Hopi way"). However, I then went on (but not as an analysis of this ordinary belief that a person's moral code is justified) to propose a useful conception of "justification", which I suggest (a) must include properties helpful in resolving practical doubts about one's moral code, (b) must permit that knowing something to be justified not require a priori synthetic knowledge or a commitment to a coherence theory, (c) must not simply authorize a person's actual moral beliefs, ____________________ 18. Except that my view leaves open the truth of a certain kind of relativism: two people, when all factual disagreements have been removed, might differ in their moral conclusions. Hare's view may leave open the truth of a certain kind of relativism too, if we include his view about "ideals" and construe him as admitting some non-utilitarian considerations into moral reflection. -226and (d) must have the result that the attitude is recommended to at least most informed persons. I then went on to propose a definition I believed had these properties: a moral code is justified to a person if she would support it for her society, on the basis of the non-moral preferences she would have if she were fully and vividly informed, if her preferences were mood-independent, transitive, not a consequence of stimulusgeneralization from abnormal past experiences, and so on. I assume that desire for protection of self (and children) from injury and frustration, and a rational degree of empathic altruism will be among the relevant desires. I see no circularity here: I think people think a moral system is justified in some appropriate way (inchoate, unspecified), and I later go on to propose a sense to be given to "justified" (not one that is already the sense of their use of "justified") that will do a clarifying job.

Gibbard proposes that we regard the belief that one's code is justified (in an unanalyzed way?) as a "unitary state" that consists in "accepting norms". I am not clear what this comes to. Apparently it means a disposition to embrace an imperative, both in actions and avowals, as a result of normative discussion (with others) with its demands for consistency and willingness to be persuaded by others—a process leading to coordinating consensus. Gibbard apparently thinks that my proposal about "justification", in so far as it differs from this, does no useful work and indeed is misleading. He is right that I don't agree. Gibbard raises objections to several special features of my account. First, he wants to limit the motivational-emotional part of a moral code to feelings of guilt (I'd prefer "remorse") and indignation (impartial resentment). This makes sense, I think, if we restrict ourselves to cases of injury to others. But I wanted to define these motivations more broadly so as to include, for instance, Rashdall's thought that being drunk is "intrinsically disgusting" and Kant's thought that masturbation is degrading. Thus I leave the motivational-emotional part of a moral code broad enough to include whatever attitudes amount to condemning the person. (This is not, of course, to say I endorse these examples.) I don't think I've conceived of the motivational-emotional part of a moral code too broadly. Something is an item of a person's "morality" only if (a) she has some aversion to acting in certain ways (explainable in abstract terms), (b) she feels remorseful or guilty if she acts in these ways (without excuse or in the presence of a stronger aversion—prima facie obligation—to do something conflicting), (c) she condemns others who act in these ways (without excuse etc.), and (d) she thinks the foregoing attitudes can be justified in some appropriate way. All these features must be present. Yet I don't think "failure to dress with pizzazz" qualifies as something towards which on can have a moral attitude. (Does anyone think of a rule requiring pizzazz that it is any more "justified" than most of the rules of etiquette? Would its supporters really condemn someone who infringed?) I therefore disagree with Gibbard, and think those who think it qualifies are guilty of a conceptual fault. -227Another objection Gibbard raises concerns a distinction between rational moral feelings and rational support for a moral code. My thought is that we can't appraise moral feelings as rational solely by "cognitive psychotherapy". For we can't separate the appraisal of a moral feeling from the appraisal of a whole moral system of which it is a part. And we want to appraise a whole moral system on some such basis as whether it is better for society than another, or whether it is one we'd feel happier living with. For instance, we may think some obligations are not strong enough to justify acting in conflict with strong personal interests—e.g., giving up a weekend of skiing in order to be home to keep a promise to pay the paper girl (see Regan 1985: 56-67). We need to have varying strengths of moral aversions, and can obtain an ordering only by thinking, e.g., referring to the long term benefits (etc.) of different systems. 19. But might not it be possible, as Gibbard suggests, that one be motivated to "support" a whole moral system but feel no guilt about

committing a murder which the system prohibits—indeed without having any features of the moral code he "supports"? I agree that if a person feels no guilt (etc.) about committing a murder, his morality is not opposed to murder, but if he wants to know whether his participation in a murder-condemning code is justified or rational he had better ask which kind of code he can rationally support (not just should accept). Such a code must extend to himself, among others. Gibbard objects that I take "corrected desires" to be ones "mood-independent and capable of surviving repeated vivid reflection on relevant facts". Actually, when I now explain this phrase, I substitute a longer list of conditions for "corrected desires": (a) not discounting the future, (b) not being affected by temporary motivational states (such as satiation, present or past), (c) not being the result of stimulus generalization from abnormal types of case, (d) not a consequence of failure to associate the target with unpleasant facts (such as that getting drunk leads to headaches), (e) not a result of failure to make discriminations such as between present ability to afford something and past poverty, (f) not a result of false or unjustified beliefs such as the belief that God wants the Sabbath day to be kept holy, and (g) not a result just of praise or expressions of alarm by parents or others whose appraisal of objectives has no sound basis. This list doubtless needs expansion and explanation. I agree with Gibbard that "is rational" has commending force, but I don't think my commending something makes it rational, nor does discommending make something irrational! So I don't agree with Gibbard that "to think something rational is to accept norms ____________________ 19. Incidentally, I disagree with what seems to be an implication of Gibbard's first point on my characterization of "moral practice"—to the effect that desires/aversions are "consequences" of propensities to certain feelings. I would not wish to underwrite the implied psychology of the origin of moral desires/aversions. I agree with Martin Hoffman that if we have an empathic aversion to a certain kind of action, it follows, as an effect, that we'll feel remorseful if we do that, and condemn others who do. See Hoffman 1989. -228that say to do it, and to call something rational is to express this acceptance of norms". Although I have indicated some points about which I disagree with Gibbard, I want to go on record as saying that his 1990 volume is fabulously rich in original insights and informed by very wide learning not only in philosophy but in the social sciences generally, particularly in psychology and sociobiology. He is too generous in depicting his theory as some kind of development of mine; it is far more, although I would like to think his judgment on this were true!

VII. Comments on Sumner Wayne Sumner's commentary is a careful account of my intellectual biography, together with not only deep challenges to theses he thinks go astray but also thoughtful proposals about how one might meet some of the difficulties. His analyses are illuminating, and many of his critical points well-taken. Still, I don't agree with all of his arguments and conclusions. Let me make clear where I don't agree now, and why. I take it that "welfare" (or well-being) is to be explained as an intrinsically good state of individual sentient creatures, so that nothing like distribution of income or punishment for illegal actions qualifies as welfare. A welfarist about the intrinsically good is one who holds that all intrinsically good states are those of welfare—of individual beings. Consequentialists hold that the morally right action is some function of the intrinsically good, but not necessarily the one which maximizes welfare. I formerly held that there are intrinsic goods in addition to welfare, viz., an equal distribution of goods among individuals. I dropped this view in mid-life, and also tended to drop the view that states of individuals such as knowledge and virtue, except in so far as they are components of happiness or are objects of a person's desire, are intrinsic goods—although in fact I have been and still am a bit ambivalent about this. I do not recall exactly why I made this change; possibly it was because the conflict between hedonism and desire-theories of value came to occupy center stage in philosophical discussions. Or possibly I associated these views with reliance on intuitions and self-evidence in thinking about values, which I had come to reject. Sumner offers suggestions about how this change might now be defended. He suggests that a theory of welfare, as a theory about when a person's life is going well, must be subjective. That is, it must "track how well a life is going for the person whose life it is". So, a theory about the nature of welfare must preserve "and explicate this relativity ... to the subject's standpoint"—how life goes well for him. A subjective theory of welfare does this by making welfare a function of the person's desires or enjoyments. Objective theories do not. One could query, however, whether a failure to be a "subjective" theory in this sense is a defect. Suppose one thinks that knowledge, just as such, makes life -229better (richer) for a person. In what sense does this fail to express the "subject's standpoint"? Does "life going well" for a person imply "she likes it" in some way? Does "life going well" imply that it "goes well for the person"? Sumner thinks a proponent of "objective" goods would do better not to propose them as components of welfare (of life going well for a person) but rather of the good in a non‐ welfarist sense, but he does not elaborate here. (In a recent paper elsewhere (Sumner 1992), he discusses perfectionism, which he regards as the chief opposition to "subjective" theories of the intrinsically good. He tentatively rejects this theory, but is a bit open minded.)

As for the contest between hedonistic and desire theories of welfare, I have repeatedly pointed to the difficulties, for desire theories, in allowing for the fact that a person's desires change over time. Sumner seems to agree that this is indeed a difficulty, but affirms that the more serious objection is that one cannot hold that life has gone well in a certain experience, however much one has wanted it, even rationally, if one strongly dislikes it, or is indifferent to it, in the event. I think both these objections constitute serious difficulties for the desire theory of welfare; I have emphasized the objection about changing desires because it is not based on any controversial ideas about what makes life go well. Rather, it simply points out that one cannot even set up an intelligible program of rating events, on the desire theory. Sumner rather likes the account I gave of happiness in an encyclopedia article of 1967: that a happy person is one who likes the important features of his life pattern, does not wish them different, and whose life includes positive emotions like joy, and few negative ones. (This account is very like a recent one by a psychologist (Diener 1984).) But Sumner goes on to object to my explanation of this in terms of "predominantly pleasant moments" and then following on by analyzing "pleasant experience" as one that "makes one want, at the time, to continue it"—not in view of consequences. (I explain this conception in a moment.) But it is not entirely clear to me why Sumner objects to explaining a state of happiness as one of predominantly pleasant feelings. The disagreement about the analysis of "pleasant experience" is about a point in the philosophy of psychology. What is there about an experience which is its pleasantness? Not, I think it is now agreed, the presence of some special introspectible common quality in the experience itself. (It is thus different from pain, which is a special kind (or kinds) of sensation.) Long ago, E. L. Thorndike defined a "satisfying" state of affairs as one an animal does nothing to avoid, but often does things to attain or preserve. Notice that a child, enjoying an ice-cream cone, will not throw it down, but will hold on if someone tries to tug it away. This disposition is causally dependent on the taste of the ice-cream, which energizes the disposition to retain (Pfaffman et al 1979). So I suggest that "The experience E is pleasant for P at t" is the same as "An experience of the kind E is going on in P at t, and is the differential cause at t of an increment in Ps disposition to -230retain E beyond t". Essentially this view was defended in Alston 1967, and is at least very close to the views of Sidgwick (1907: 122) and Kurt Lewin (1938: 135). This concept is very different from desire ex ante, and even from being pleased that something happened ex post, both of which are at times different from the event desired. Some philosophers, including Sumner, object to this view on the ground that there are many highly enjoyable experiences that we would not want to continue, that would be spoiled by this. We don't want the orchestra to go on playing after a fitting finale; but if we are enjoying it, we surely don't want it to stop before the end—and will hope to get an encore! Are there moments of bliss which we would not wish to go on for another moment?

Sumner also objects to the view because the analysis of "pleasant" is inconsistent with the fact that desire "can only represent ... an ex ante expectation that the continuation of some state or activity will be experienced as gratifying". However, I think this takes an excessively narrow view of "wanting to retain" when something nice is going on. I do not, moreover, agree that it is of the nature of a desire that it be for a future event. We can be glad that a certain event occurred—is that not desire? And we often wish we had done things differently in the past. So I am not convinced by Sumner's criticism of this view of pleasure; nor do I think espousing it in any way inconsistent with rejecting the desire-theory of welfare. I have not suggested that an event is a pleasant experience just because it was desired at some other time, however rationally. I have, however, argued that an enjoyment is not really a part of a person's good if it would not have occurred if the person had been fully informed (Brandt 1979: ch. 7). Sumner has a final objection to this view: that it implies that well-being is a mental state. He thinks this renders the view susceptible to experience-machine counterexamples. Would we really prefer a love-experience, however nice, which derived from being attached to electrodes in a machine? (This sounds as if Sumner is himself appealing to a desire-theory!) I have expressed puzzlement elsewhere about just what this counterexample is supposed to prove (1989a: 49f.). I do not have difficulty with supposing that welfare is or requires satisfying mental states. Here I am tempted to go along with Railton (1989: 170) that this "science-fiction example ... cannot ... afford a crucial test against experiential conceptions of the good because it draws upon intuitions about what we want for its own sake which were developed in settings where the drastic split the machine effects between experience and reality does not typically exist." I agree with Sumner's point that a reaction to an experience is not fully rational if it would not have occurred given full information about its sources. I am insufficiently clear about what the "experience-machine" conception really is to know whether the pleasures in it would be possible if the person were informed about what his real status is. -231One consequence of this discussion is that I find myself unsure what exactly Sumner's own view really is—that a life goes well if it is "happy"—i.e., "whatever is experienced as rewarding or fulfilling for its own sake will count as an intrinsic benefit" and in the list of such states will be included "achievement, play, meaningful work, friendship, good sex, peace of mind, self-esteem, and so on and on". This is roughly the same list as if we opted for the desire theory. Notice that the list of states is not confined to happiness or enjoyment. Does this mean that Sumner's view is that a person's life goes well in respect of any feature of it reflection on which is able to elicit a favorable attitude in him at the time it is occurring? Is it limited even to a person's experiences? Where do I end up? I agree with Sumner in being a welfarist about the intrinsically good —that it consists of states of individual sentient creatures. I further agree that it consists in moments they find satisfying or enjoyable (but I am unclear just how much of a

restriction this is for Sumner.) I agree that an event is not intrinsically good just because it is (was) desired, even rationally, by the subject. 20. I am not convinced that the sole reason an event is made good by desire or by being satisfying is that this captures relativity to the subject's standpoint. I agree with Sumner that something's being satisfying (in some sense needing to be specified further) is the crucial condition of its being intrinsically good. I think a desire theory also founders on the fact of changes in desires. Did I back‐ slide to a desire theory in holding that a satisfying life, as compared with a life in which one's desires were fulfilled but not satisfying, is one we'd overall prefer? I certainly was not saying that we would prefer a life in which those events occurred which we had wanted, irrespective of whether we liked them in the event. (Still, I would happily delete the paragraph in which this statement occurred!) Sumner seems to object most strongly to my proposal that happiness is to be analyzed primarily in terms of predominantly pleasant states of mind, and, more forcefully still, to the view that a pleasant experience is one which makes us disposed to continue or repeat it. He suggests that in the end this is capitulation to a desire theory. I do not agree that it is. He also objects to the result that my view makes welfare essentially a mental state. All Sumner's criticisms are trenchant, and based on a very careful reading of my text. I am grateful to him for them, as doubtless will others be who, unlike me, tend to agree with the force of many or even all of them. ____________________ 20. The quote from me cited in his footnote 10 was a mistake. What I should have said is that, if one were to go for a desire-theory at all, this would be the most plausible location for the relevant desire. Sumner is right that a desire at this time is not necessarily better informed than ones occurring at other times. -232-

VIII. Comments on Harsanyi Ever since his articles in the Journal of Political Economy in 1953 and 1955, his paper in Mind on ethics conceived as hypothetical imperatives, and his subsequent criticisms of Rawls, John Harsanyi has been marked as a leading writer on rational behavior and rational social norms. 21. One can only regard his views with the highest respect. So I turn eagerly to considering what should be said in reply to his paper about topics which have long been of interest to both of us. First, the matter of cognitive psychotherapy and the rationality of desires. Harsanyi thinks we need not worry about how to criticize intrinsic desires, since we can always evaluate a given desire by how well it fits in with the "overall quality" of our lives. But it seems to me we can identify the overall quality of a life only by a criticism of our desires: can we know whether achievement is a constituent of a high quality of life before we have appraised the rationality of the desire to achieve? Apparently, Harsanyi is just not worried about a question concerning personal desires which does worry me. My answer was that rational assessment of one's desire involves facing them with vivid and repeated

representation of relevant facts; the desires that remain after this procedure can be viewed as not "irrational". Harsanyi does not comment on my examples of how such therapy would realistically work. 22. It is, incidentally, a mistake to think that I suppose that a desire is irrational if it was first acquired in an irrational manner. Consider a student who aspires to a Ph.D. because he thinks this will please his parents, or a person who developed a liking for classical music because his grandmother insisted on his listening to Mozart. As I said (1979: 116), if a person finds academic work inherently satisfying, his ambition will not be diminished by awareness that it began because of a supposed desire of his parents; and if he comes to like Mozart's music, his valuation of it will not be altered by the reflection that he developed an interest in Mozart because of his grandmother's influence. Harsanyi agrees with this. And I agree that an aversion to working as a motorcycle mechanic may be harmful if it prevents someone from having what would be, for her, a fulfilling life. Someone with an aspiration to be such a mechanic can rationally appraise it by trying out this line of work, by talking with those in the occupation, and perhaps by noting that her parents actually know nothing about such a life. But what I want to ____________________ 21. The papers I mentioned are reprinted in Harsanyi 1980. See also Harsanyi 1977: ch. 4. I have also profited from seeing many other papers of his, some of them issued by the Center for Management Research, Berkeley. 22. See my 1979: 115-26. For my more current view, see the additions in my reply to Gibbard. -233insist on is that it simply begs the question to say that we have a (rationally supported?) conception of a life of "high quality"; we need to know what are rational desires (or rational enjoyments) in order to appraise such a conception. I turn now to the issue of whether I am right to reject the desire theory of welfare in favor of a happiness theory. Harsanyi prefers to speak of a "preference" theory rather than a "desire" theory of utility, but since (I think) "preference" can be explained in terms of relative desire-strength, I see this as no difference. Again, I am somewhat ambivalent about the choice between the desire theory and hedonism, though I lean toward the latter (Brandt 1989a: 53-4). But I think both theories may be right in what they affirm, partly wrong in what they deny. Harsanyi's main reasons for preferring the desire theory are that we don't want just subjective experiences—recall Nozick's "experience-machine" example—and that we would prefer an interesting life of accomplishment to a happier life. I do not wish to argue that his views on this are mistaken, but I suggest that when we try to spell out exactly what some of the nonhedonic components of a good life are, matters become murkier. Do we want knowledge as such—perhaps as true beliefs which are not

experiences at all? Or do we want all kinds of knowledge, e.g., about the love life of other people or their dying moments? At most, perhaps, we want some kinds of knowledge. But which? Would we really want some state of affairs for itself if we thought it would make no one happier? Harsanyi has another reason for preferring the desire theory of individual utility to the happiness theory. He reports (favorably) my proposal that "experience E is pleasant" be construed as "E is occurring and is the differential cause of the person wanting to continue/repeat it at the time". But he goes on to argue that the happiness theory, implausibly, must underwrite drug-addiction because it is so pleasant. However, to say this is to ignore the fact that drug-addiction leads to very unhappy long-term results. Harsanyi also thinks that the happiness theory implies that it is irrational to harbor altruistic desires for the welfare of other people. I do not see this implication; a happiness theory of the intrinsically good can perfectly well admit altruistic desires, but say that altruistic desires should be directed at the happiness of other people, not the agent's own happiness. The happiness theory of the intrinsically good is not a form of egoistic psychological hedonism. Harsanyi says that the theory "in its pure form" holds that "when we are helpful to other people ... our primary purpose is always to promote our own happiness rather than theirs." This is not at all part of the happiness theory. What it holds is that, if you want to help someone else, you should do what will add to her happiness. My chief objection to the desire theory, at least in one form, is that it holds that an event which brings about what one did/does desire is intrinsically good even if one heartily dislikes the event when it occurs, or even if one never knows about it, or even if it occurred after one's death. Here I go along with Sumner's view in his paper. (A version of the desire theory that is not open to the objection -234about postumous benefits is outlined in Overvold 1980 and 1982. I discuss Overvold's version in Brandt 1991.) My second objection to the desire theory is that, since a person's desires change over time, it is not clear which desires, according to the theory, determine the goodness of an outcome. Harsanyi thinks we can overcome this problem if we distinguish between a person's local desires (e.g., desire for a high from cocaine) and his global desires (dislike of a life of addiction), and informed versus uninformed desires. Does this work? Suppose someone initially wants a big wedding, likes it while it is occurring, but regrets it later as puerile nonsense. On the global desire-satisfaction theory, how do we identify her informed global desire? Should she (we) estimate the strength of her desire both prior to the wedding, while it was occurring, and then at all times afterwards? Such a calculation might not be too difficult if the desire is largely stable—if it doesn't change as one becomes more "mature". But will her "informed global desires" at the time of the wedding be that definite? How is someone to estimate how important a big wedding will

seem years later? Will "informed global" desires not also change over time? Moreover, Harsanyi concedes that satisfaction of global desires is dependent partly on the extent to which local desires are satisfied, since the former are "basically desires for a high overall quality of life" (but presumably not happiness). Yet with all these complications he holds that the utility-level of a person at any time t is to be defined in terms of his informed global desires at t. Harsanyi says these desires "will cover all his interests from that time on till the end of his life". I find this conception puzzling if it does not mean he is to speculate about what his global desires will be at the many points of later life—not just at the time t. So I do not see how Harsanyi proposes we can determine rationally what is a "high overall quality of life"—and isn't that the main question? I think we need more details on how all this will be worked out, before we conclude that the informed global desire theory really is defensible. The importance of this objection is in the fact that it does not depend on our preferences, whether for a life of happiness or one of achievement (etc.). The objection is that the desire theory does not make clear an intelligible program for decision-making. If this objection goes through, the desire theory must be rejected. Finally, I turn to what Harsanyi says about rule-utilitarianism and its advantages. Harsanyi defines "rule-utilitarianism" as the thesis that a morally right act is one that conforms with the moral code the social acceptance (adoption) of which would maximize social utility. My conception is slightly different. I begin by defining a "person's moral code" in terms of motivations to do certain things and dispositions to feel guilt at not doing them and to condemn others who don't do those things. I add that deviations may be justified in order to avoid serious social losses, and may be excused if the action does not show discrepancy -235between the person's moral code and the code widely regarded as optimal. As a consequence, I think that the basic question for a given person is what kind of moral code she should adopt and recommend to others, and this choice should reflect the fact that in actuality many people in her society do not accept one and the same moral code. This fact may necessitate having disjunctive clauses in the code she should adopt and recommend. Doesn't this mean that actual individuals will favor somewhat different and conflicting moral codes? Yes, they will; but (following Berkeley) I have suggested that at least the main features of the code will be a matter of agreement among rational people. So I regard these proposals of mine as a somewhat insignificant gloss on Harsanyi's definition of a "rule-utilitarian" theory. I agree with Harsanyi entirely that the rule-utilitarian concept of an optimal moral code is superior to the act-utilitarian one, primarily for pragmatic reasons. First, having good rules of conscience will, by and large, lead people to make more beneficial decisions. Second, collective acceptance of such a code enables people to form more definite expectations about what other moral people will do, with a consequent enhancement of

incentives. I have not emphasized this last as much as Harsanyi has, or as I should have done, but I have discussed it briefly. 23. I am pleased by Harsanyi's full statement here.

IX. Comments on Hardin Professor Hardin's paper is essentially a review of, and adjudication between, actutilitarianism and indirect utilitarianism (which he divides into two forms: ruleutilitarianism and institutional utilitarianism.) 24. His analysis leaves rule‐ utilitarianism the most implausible of these three. In contrast, while I admit a role for institutional utilitarianism, I rank rule-utilitarianism the most plausible. What is the main difference between direct and rule- (or conscience) utilitarianism (although Hardin views the educational aspect of my rule-utilitarianism as at least closely related to institutional utilitarianism) as theories about the rightness of other-regarding actions? The main difference is their view of the aims or motivations, in such contexts, of the moral person. According to the direct utilitarian there is only one basic motivation: to maximize the total well-being of everyone affected. According to the indirect utilitarian, however, we should also be strongly motivated to speak the truth, keep promises, not injure other sentient beings, etc., but these motivations themselves are subject to appraisal as long‐ range strategies for achieving the same goal as that of the direct utilitarian— maximizing total well-being. Choice between the two theories, therefore, from ____________________ 23. See my 1988: 346ff. This is reprinted in Brandt 1992: 142ff. 24. See also Hardin's learned and tightly argued 1988 book. -236the general utilitarian point of view, will be made on the basis of showing which is the better long-range strategy for maximizing the general well-being. Hardin's third form of theory, institutional utilitarianism, may also be appraised as a longrange program of maximizing public well-being—by structuring institutions in optimific ways. In his book he treats tort law, contract law, and criminal law (law by governments generally, including statutes, right to trial by jury, right not to be compelled to testify against one's self, etc.) as institutions. What, then, is the program of "institutional utilitarianism"? The idea is that what individuals do should be fixed by institutional constraints, but the structure of the institution itself is to be selected on the basis of the probability of the whole promoting the long-range public welfare, as contrasted with alternative possible structures. So taken, the fact remains that all decisions are made by individuals. It is individuals who decide how an institution like the law is to be set up (Congress), and how it is to be interpreted and applied (judges and appellate courts). In these decisions they need not act as utilitarians of any sort (they may have a prudential eye open), and even if they try to, they may feel bound to decide so as to preserve analogies in the law, and to try to make the law an intelligible systematic whole. But they could make decisions as utilitarians of some sort. Indeed, in his book Hardin speculates

that if Sherlock Holmes had been a police officer he might well have risked problems by withholding evidence, if he thought that would do the most good! Hardin's main charge against rule-utilitarians is that they cannot offer a convincing explanation why their long-range strategy can do better than that of act‐ utilitarians. Indeed, he seems to suggest that there is no way of showing that a given moral rule of conscience (e.g., to speak the truth) can be appraised other than, apparently, by noting whether act-utilitarians who spoke the truth in specific situations produced more utility overall than they would have done by telling a useful falsehood. However, if there is a problem how rule-utilitarians can appraise favorably telling the truth as a good-maximizing strategy, there is an equal problem how act-utilitarians are to decide which kind of action will do the most good. For they have to appraise actions ex ante, predicting which one among several will expectably maximize well-being. How will they do this? I suggest they must do this by reliance on what they know about human beings in society, say, how disruptive it is apt to be for a human being to, say, rely on erroneous information in making plans, as well as how beneficial it may be to informants to have their auditors making plans on the basis of false information. (Of course, they can take into account any specific knowledge of how disastrous given misinformation will be.) Essentially the same for the various moral motivations which rule or conscience utilitarians will claim should play a part in the motivation of a moral person. Motivations, of course, have been learned in some way: by teaching of -237parents, by experience of being the target of actions by others who do not share these motivations, and so on. Their utilitarian justification, of course, must come mainly from understanding the impact (on sentient beings) of actions called for by these motivations. So both direct and indirect utilitarians must rely, in one way or the other, on knowledge of how certain types of action will affect sentient creatures. (Act‐ utilitarians, of course, take into account the specific situation.) Is there an advantage in being motivated in the way that rule- or conscience utilitarians have in mind? Yes, because dealing with cases on an individual basis will be not only time-consuming but also a temptation to rationalization. In addition, there are the expectation effects Harsanyi emphasizes. (How securely could one trust a whole-hearted act-utilitarian to observe property-rights, or to return borrowed money at the time promised?) Sophisticated rule-utilitarians will not, however, unlike Kant on lies, ignore the differences between cases: they will see that there are some promises that need not be kept, e.g. ones extracted by misrepresentations by the promisee. Sophisticated ruleutilitarians will also have learned a rough order of weight among obligations

(motivations). What they will do is have a settled system of motivations, very complex of course, a bit like common-sense morality (as I explained above). This view is often criticized for not taking cognizance of situations in which it would be harmful to act on the justified general motivations (aversions to act‐ types). Suppose the agent's society is divided about what is moral. Or suppose the agent must deal with people who are selfishly motivated to ignore moral demands. Rule-utilitarianism need not be so inflexible in such cases. The theory holds that, when a puzzling case arises, the agent should reflect on which type of rule is a better guide to maximizing benefit. As I've indicated, disjunctive rules are possible. We could have a rule telling us what to do where there is rough moral agreement in the society, but directing us to avoid harm where this is not the case. In case virtually all individuals are in fact, say, disinclined to speak the truth (and hence few people would believe purported information), the motivation to speak the truth would be pointless, and the motive dropped. So, as a long-term strategy for maximizing benefit, I think the rule-utilitarian theory comes out on top of the act-utilitarian strategy. Hardin seems to think that some form of institutional utilitarianism is the best bet for maximizing welfare—and certainly the rules of institutions do play a large role in our lives. He says this theory is addressed, usually, to "complex problems involving strategic interactions of many people". He does not provide many examples, but does mention a possible proposal for compulsory voting, rules about which side of the road to drive on, rules about whether it is allowable to make a U-turn at a certain intersection, about serving in the armed forces, and about compulsory payment of taxes. Of course I acknowledge that we need to -238know the socially expected and collectively best thing to do in such cases. 25. But Hardin thinks we should attend to institutional utilitarianism because institutions can assemble much better information about probable outcomes than can an individual. I concede to Hardin that institutions probably can assemble much better information in many contexts than we can individually. But, as he points out, not when we are dealing with specific individuals about whom we have special knowledge. Thus, institutional rules are hardly likely to be helpful for many decisions which concern us most frequently —how much to give to charity, how to treat students or colleagues or friends or lovers or children, how fast to drive (within limits), how much to drink, which promises to keep, and so on almost without end. I also concede that government may have facts about how fully to punish tax evasions, based on knowledge of the costs of prosecution and the gains from full payment, which no individual is likely to have. But is this in fact very often the case even on matters calling for moral decision? When Congress passed a uniform code for criminal sentencing in 1987 it is hard to believe it was well‐ informed; it made absurd penalties

mandatory, as has been made clear by various writers. Many states make it criminal for a physician to assist in a suicide of an ill patient anxious to die. What special information do the legislatures rely on? Moreover, politicians, if they want to be reelected, often have to bow to popular prejudices, phobias, and misconceptions. True, institutional power can eliminate much of the problem thought to plague ruleutilitarianism, the problem of only partial compliance, by imposing sanctions which prevent it. But this has its limits: government cannot mandate how much people are to give to charity, whether they will fight in a war they think unjustified, or do any of the things people feel morally bound to do whatever the state may say. The most it can do is impose sanctions, the justice of which is much debated in the literature on the conflict between law and conscience. 26. Hardin questions how rule-utilitarians can justify the various moral motivations they advocate (not to injure, lie, steal, break promises, etc.), and asks if this is a matter of "authority". I think the individual does not "choose" these dispositions: they are the accumulated result of parental punishment, example, and admonitions, of the influence of teachers and role-models, and of public debates. But also important can be information about how the "rules" protect and promote one's own and the general well-being. The constellation of motivations that ____________________ 25. Although notice that both act- or rule-utilitarians will find situations in which their theories lead them to act contrary to social expectations and even to the law. 26. None of this is to deny the important point that society needs a system of laws, mostly enforced, fashioned after utilitarian reflection. -239mature rule-utilitarians would find in themselves would be very rich. When a puzzling case arises, they would ask whether a given rule, or perhaps a new rule, should have priority. As I briefly explained earlier, to come up with an answer rule-utilitarians would have to rely on a lot of general knowledge about people, their aspirations, etc. So, while their earliest, simple set of rules may come from an "authority" such as parents or teachers, new rules should come from reflection, in the light of all they know about the human scene, about which code (the prevalence and teaching of which) would do most good. That code would not be act-utilitarian. What I have said in criticism of Hardin does not address the question why we should be utilitarians at all, of any sort. For that, see my 19883, and in this volume both my introductory remarks about justification and my replies to Hare, Gibbard, and Kavka.

X. Comments on Kavka Gregory Kavka, in a typically lucid and thoughtful piece, argues that I do not show that a fully rational person would support a social morality which gives equality of

consideration to groups in no position to provide reciprocal benefits, the most conspicuous such groups being future generations, animals, and the mentally defective. The same claim might be made about whether a fully rational, relatively well off adult would support a social morality that provides benefits to small children, to people of poor countries, or to powerless members of the underclass. Whereas I had argued that a kind of rule-utilitarian moral system would be supported by rational people, I did not show that it would be supported by selfish people, however rational. There are, of course, two different questions. 27. The first is: what kind of moral system would a rational selfish person support? The second is: would a rational person live by the moral code he supports where he thinks it conflicts with self‐ interest? The second question will take different forms depending on the kind of morality represented in the motivations of a given agent; presumably, if rational, a person might pay little attention to the demands of his "morality" if it were open to sophisticated objections. The serious question is why a rational agent would pay attention to an optimal morality, if he subscribed to it, where its demands conflicted with his self-interest. To my mind, this second question is whether a rational agent would pay attention to the demands of a ruleutilitarian morality, if he subscribed to it, if he thought it conflicted with self-interest. Of course, the answer to this question will depend on exactly what an "optimal" ____________________ 27. Which Kavka has discussed before. See in Kavka 1984. -240morality requires of one. For instance, will one's optimal morality require a wealthy man to give to the poor to the point incomes/resources become equal, or only when the costs of giving are small and the benefits received are large? Kavka's criticisms of me are concerned with the first of these two questions. I would put this as asking whether it is possible to recommend successfully to a rational person the support of a social morality whose demands conflict with his self-interest. Evidently, whether it is possible will depend on the actual non‐ moral motivations of this person. Perhaps he is, and wants to continue to be, a benevolent person. If that is not itself an irrational desire, then perhaps a rule‐ utilitarian system of morality can be recommended successfully to at least many rational individuals. One might go on to ask how we can recommend to a rational person a morality that is universalizable in the sense of aversions (etc.) to types of action (lies, theft) definable in abstract terms, as contrasted with actions explainable only by use of proper names or indexicals, say ones performed by specific persons or certain social groups. Or we might ask how we can recommend to a rational person a morality that is impartial in the sense that its demands not be subject to reasonable rejection by anybody. It might be said that we do not have a morality at all if these conditions are not met. But we could reply to this by saying that, whether or not we call the motivations he has "a morality", a rational selfish person would not wish to restrict his motivations and behavior to this extent. If he

does not, if he is interested only in asking what sorts of motivations we successfully recommended for him to have, I think we have stopped talking about morality (a fact that may not move a sophisticated selfish rational person). In a successful recommendation of a moral code to a rational person, what motivations other than some degree of benevolence can we expect to reply on in her? We can appeal to her aversions to physical damage or psychological damage and threat thereof. We can also appeal to her aversion to losing her property. We can also appeal to her aversion to any of these things happening to her family or friends. How about desires for a modest standard of food, lodging, health care? Freedom to express her thoughts, practice her religion, join with others in projects of reform? A selfish person might want all these things, but only as far as is consistent with her wish not to do her full share in supporting, or paying for them. And she may be even less willing to do her full share in supporting these goods if she will not herself be able to share fully in them, or if the benefits are extended to future generations, animals, or impoverished countries on the other side of the world. However, in his 1984 paper, Kavka himself provides some reasons which explain why a fully rational person might be ready to support my form of rule‐ utilitarianism as a social morality. Kavka could point out that this conception of an optimal morality is more attractive than some may have thought: it will permit acts of self-defense, it holds out the prospect of life in a society of mutual -241trust, and it is very unlikely to require the ultimate sacrifice—death in a moral cause. Kavka points out that if morality requires us to treat the impoverished well, this will make for social stability and discourage revolution. Much the same for justice on the international level: the wealthier nations will benefit in various ways from giving aid to the poorer nations. Again, conservation for the benefit of future generations tends to be beneficial to the present generations, not to mention the fact that most people do care about the welfare of their own children, or even grandchildren, who will clearly benefit from conservation. Kavka also points out that many people have "self-transcending" interests, like concern for the arts, religion, democratic government, the prospering of the human species, and that promoting these things gives meaning to their lives. Kavka's own points go a long way toward filling the gap he detected in my argument that rule-utilitarianism could be successfully recommended to most people in our actual world. I myself would have emphasized them in the past, if I had thought of them! But these points do not fully succeed in rescuing my argument. Kavka considers the reasoning by which Peter Singer has attempted to fill this gap, by appeal to sociobiology and history. Kavka does not contest that an evolutionary argument suggests that some degree of altruism can be expected to develop widely. He thinks, however, that this altruism will necessarily be of limited scope. Here I differ with Kavka on the facts. If kin selection explains development of some degree of altruism in small,

mostly related groups, this does not imply that the trait is necessarily restricted to kin— and I think empathy/sympathy obviously is not. But this does not tell us what degree of altruism is rational in the sense I have outlined. Moreover, although Singer has a great deal more to say, I agree with Kavka that his arguments do not suffice to show that a rational person must assent to a morality that treats equally the interests of all persons, in whatever groups, and animals. Kavka then turns to David Gauthier's view that morality can be construed in terms of a rational bargain between self-interested individuals or groups. Gauthier attempts to do so partly by interpreting morality to make it less restrictive and demanding, and partly by urging that rational pursuit of interest does require extending moral concerns at least to future generations, although hardly to the international sphere or to animals. Kavka concentrates his criticism on Gauthier's discussion of justice between generations, and argues that it arrives at very counterintuitive results, results which should give pause to Gauthier. This result does seem to support Kavka's claim that it has not been shown that a rational selfish person would support all of a rule-utilitarian morality (though I am not clear exactly what a rule-utilitarian view of just savings for future generations would be like). -242Kavka concludes that the general project of showing that morality can be construed as an extension of rational self-interested motivation has to be abandoned. Some of the same points can be made in answer to the second question of whether conforming with the rules of an optimal morality to which one is committed is rational even when this conflicts with self-interest. For suppose we define "our self-interest" in terms of the desires we would have if we underwent cognitive psychotherapy, minus the desires we have because we have internalized some system of morality. The question is whether, if we were rational, we would prefer, moral commitments aside, to reduce our commitment to our present (optimal?) morality, in such ways as to further our selfinterest. Now when we count up all the various benefits of living morally (including having other people's the trust and esteem), my guess is that most of us, if rational, would prefer not to reduce our commitment. Or at least we would not, if we didn't abstract from our moral commitments, but instead considered everything, including what our moral commitments achieve. On the other hand, this "we" certainly does not mean "everybody". For there are lots of people who have virtually no advantages in life, who have not thought about morality and its point much less about a rule-utilitarian account of it, who do not care that others will think badly about them if they behave immorally. For them, I would not wish to argue that living up to an optimal morality, or even some very diminished form of it, pays, as compared with following self-interest, in the restricted form available to them. Living morally would pay for them only if they were different sorts of person. Something like this might apply to a highly selfish person of any sort.

So I accept Kavka's complaint that I did not do as much as possible to support the ideal of a rule-utilitarian morality, and I am happy to avail myself of the lines of reasoning of his 1984 paper. But even after his points are taken fully into account, the reasoning for a rule-utilitarian moral system comes up short, in that it could not rationally convince everyone, as they are. We cannot successfully recommend morality to all rational persons. What we can say is that most of us, as we are, could not rationally, on the basis (at least mostly) of nonmoral motivations, consent to reducing our commitment to our morality, which is some more or less close approximation to an optimal morality.

XI. Comments on Shaw Shaw's "A Moral Theory for Marxists" (1988) discussed the view about distributive justice I put forward in my Theory of the Good and the Right. Both his earlier -243paper and the one in this volume are thoughtful and inventive, and largely supportive of my views about distribution. Roughly, Shaw's thesis is that my form of utilitarian morality mandates the income equality he favors, although it falls short of his wider ideal for social organization. I argued that a welfare-maximizing moral system would hold (or imply), as a moral principle with prima facie force, that income after taxes should be equal, except for supplements to meet special needs, supplements to recompense services to the extent needed to provide desirable incentives and to allocate resources efficiently, and variations to achieve other desirable ends such as population control. (Items in a moral code differ in salience, depending on various factors. This item has low salience.) Shaw thinks that the last clause should be dropped, if the moral principle stated is one of distributive justice. I agree. I had argued, following Abba Lerner, that the goal of after-tax equality of income is a desirable one, primarily because of the declining marginal utility of income, and our ignorance of the specific utility-curves of people with different incomes. Harry Frankfurt and others (e.g., Blum and Kalven) have criticized the view that the utility of income declines as income increases, in the long run. Shaw subjects the reasoning behind their objection to a very careful critique, which in my opinion is conclusive. However, even granted the declining marginal utility of income, we could advocate a social policy of not distributing income equally, but rather so as to make equal the marginal utility of everyone's income—so permitting incomes to be very unequal where necessary to achieve this effect. However, as Lerner argued and Shaw agrees, we do not have the requisite information about the marginal utilities of incomes of people at various income levels. As a result, if our distribution diverges from equal income in the wrong direction, we shall do more harm than we would gain when we diverge from equality in the right direction. So the best policy, other things being equal, is to aim for incomes equal after taxes. Shaw agrees. In the case of the ill and handicapped, we know enough of the utility curves deriving from their disabilities to recommend palliative extra grants of income, for

the sake not only of increasing equality of well-being but also of maximizing social wellbeing. These income grants are well spent. I had argued that we may also need an exception where necessary to provide incentive. People may not invest in socially important ways, or go into certain occupations which are stressful, boring, or require long training. I had pointed out that special compensation for people who go into stressful (etc.) jobs might be justified not only as promoting equality of long-term welfare, by compensating for these losses, but also as maximizing social welfare by inducing people to work where they are needed. Shaw points out that compensation to recompense losses is one thing, but compensation just to get needed services performed is another. This, Shaw suggests, looks like "extortion" and is needed largely as a result of unjust advantages in education (etc.) and of professional limitation on -244entrance into certain preferable fields. In any case, he thinks this compensation ignores the fact that the better-paying positions are intrinsically more rewarding. (It also ignores the empirical fact that an increase in tax rates on those with more lucrative jobs does not lead people in important jobs to work less hard or long (Pechman 1987: 76-7).) Shaw thinks that extra income to compensate people for losses in taking unpleasant or dangerous jobs, or for training, is acceptable and should be incorporated in a moral principle about justice, but not extra income to get socially needed things done, if the reason that individuals are not willing to do these needed things unless highly paid is just that they can get away with such demands. He has a point here, but he does not show how such a distinction could well be made in an income-tax system. But inequality of income is often also defended on the ground that the prospect of it is necessary to encourage entrepreneurial activities, or investment in the right places. This ground seems to me much more plausible than the idea that income differentials are needed to motivate managers in large industries. In any event, if our system does need to encourage entrepreneurial activities and economic investment, this can be handled by variations in the tax code. Shaw thinks I make a mistake in confining attention to market-oriented capitalist economic organizations, with relevantly different income-tax rates and welfare grants perhaps in the form of a negative income-tax for the poor. What I should do, he thinks, is also make an assessment, along welfare lines, of "rival forms of economic ownership, of rival ways of organizing production and distribution, and of alternative authority arrangements within units of production". And we should consider not only equality of income but also equality of wealth, which might require abolishing inheritance. So an amended moral principle of distributive justice might well involve concerns about "control over capital assets or investment decision-making". The primary reason I did not discuss such issues was their complexity. I am sympathetic toward socialist forms of economic organization, but we must ask how they can be

managed so as to avoid the failures of recent systems. One can only welcome details about how such systems could be worked out. And even if we came to the conclusion that a largely socialist system would be superior to the largely capitalist one, we must worry about the transition costs. Thus it is far easier to judge reasonably that a change in tax rates or welfare grants will improve things than it is to judge reasonably that changing our whole economic system will do this. So I am unrepentant about focusing in a way that Shaw finds too narrow. 28. ____________________ 28. True, changes in tax rates and welfare grants will not make the world perfect, but then the human material we have to work with is not perfect either. -245-

XII. Comments on Frankena William Frankena's paper aims to put my work in ethics in historical perspective, something I have never thought much about, but for the outcome of which aim I am grateful. Frankena is a virtually infallible source of information about the history of ethics. 29. So he is the ideal person to undertake this task. I am pleased Frankena views me as a Millian to the extent he does. Moreover, I am pleased he thinks I have sometimes got things right. I certainly should have, given that I have benefited from extended commentary from him over many years. Nevertheless, I have some questions about interpretation. Frankena points out I have said a defensible morality is one the teaching and prevalence of which would maximize utility in the person's society. This contrasts my view with that of Moore and Mill, who held that the relevant utility is that of the world or all sentient creation. I may have said that, but if so my statements missed my fundamental intention. I did hold that the morality defensible for one society is often not defensible for another, because the physical or institutional backgrounds are different, so that sometimes what is even possible in one group is not in another. I discussed this question in the context of the question whether a resident of India has a "moral right" to a certain standard of living, and pointed out that problems like this are essentially worldwide and must be considered from a global perspective. I intended to say that, where the morality of one group makes a difference to the well-being of another group, the well-being of the wider group is highly relevant to what is the optimal morality. (Whether I succeeded in showing that a rational person would adopt such a universalistic morality is another matter.) Frankena also queries how consistently I can hold, as I do, that the normally best (and simple) set of rules of conscience may be set aside to avoid disaster or gain a great good. I have regarded this permission as very similar to the criminal law about "justified" illegal actions. We could, perhaps, regard these exceptions as recognition of another special and weighty simple rule: to act in a way to avoid disasters, whatever other rules may say; and perhaps we should, since this rule is itself simple (although requiring judgment) and ought to be included in the conscience we should teach our children. But I

have also suggested that, in cases of this sort, a person can act like an appellate court, think out what more specific rule allowing for such a case would be beneficial and properly viewed as a precedent. Such a rule may not be simple enough to teach generally, just as a court's decision might be too complex to embody in the written law. And such a rule ____________________ 29. Though I demur somewhat at his sharp distinction (in what he says about Mill) between the "meaning" of a moral statement, and what "we mean to imply" when we use it. And I am not quite persuaded that Mill was as much a noncognitivist as Frankena thinks. See my 1952: 461, 466f., 469f. -246perhaps would not be attached to guilt feelings and so not quite properly regarded as part of a person's moral code. But it could still be viewed as a part of a person's effective moral motivations. I do not regard this view as a relapse into act-utilitarianism. As Frankena points out, the simple moral conscience will consist of principles rather like Ross's prima facie obligations—i.e., plural, some with low salience, possibly conflicting, and not absolute as Berkeley thought some of them are. But unlike Ross, I do not view promoting a certain kind of distributive system as a basic obligation. For me but not for Ross, the moral importance of a certain distribution derives from its effects on aggregate well-being. Frankena himself does not regard an equal distribution as an intrinsic nonmoral good. But he thinks that some experiences are intrinsically good apart from their contributing to, or being a part of, a person's happiness. He thinks they can be good because they manifest a certain kind of perfection. Here I am somewhat ambivalent: Frankena holds, and I agree, that an experience is not intrinsically good if it is not pleasant to some degree, but I do not feel certain that the degree of some experience's intrinsic goodness is fixed by any relation to pleasant experiences alone. On the other hand, while I admit to admiring experiences out of line with their pleasantness (or contribution to pleasure), being admired is one thing, being intrinsically good another. So on the fence I sit, thinking how difficult it is to ascertain precisely what features other than pleasantness add to intrinsic value (Hare 1989b: 44ff.). One critical point of Frankena's I do not understand: he says that moralities can and should be appraised on moral grounds by appeal to moral principles. I agree, of course, that actual moralities can be—by comparison with ideal moralities (the moralities that would maximize the well-being of sentient creatures). I think, however, he is referring to something else. But what? Like Kavka, Frankena asks whether the kind of utilitarian moral code I advocate could be the rational choice of just anybody, or only of the benevolent. I am prepared to admit that purely selfish people (who might not be living an optimal life even in self-interested terms) might reject rules that would protect people (and animals) who cannot reciprocate. Thus some people may be so built that acceptance of an impartially justified conscience-

utilitarian morality is not rationally justified for them. To my mind, that is unfortunate; but it does not imply that such a morality is unjustified for the rest of us. Finally, Frankena asks whether in the end I hold an expressive or a cognitive theory of moral judgments. Here I claim a right to change my mind. I now think that ethical statements, as actually made, are expressive in the sense that intelligent auditors of the language would infer from their use that the speaker has a corresponding moral code (roughly), and that a trustworthy speaker would not make such a statement if she did not believe (in some sense) that she had a corresponding moral code. Moral statements are not straight assertions that the -247speaker has a certain moral attitude. But suppose a speaker said, "This action is required by a moral code which rational people would support because they know its teaching and prevalence would maximize social benefit". This statement would also be taken as some evidence that the speaker has a corresponding moral code. Moreover if, as I have suggested, a moral statement implies that the speaker thinks the corresponding moral code is justified, and we try to spell this out, it may become less clear that an expressivist and my kind of cognitivist reconstruction are all that different. -248-

References Alston, William P. 1967. "Pleasure", in Paul Edwards, ed., The Encyclopedia of Philosophy, vol. 6: 341-47. New York: Macmillan. Aristotle. Metaphysics. Many editions. —. Nicomachean Ethics. Many editions. Ascher, Mark L. 1990. "Curtailing Inherited Wealth". Michigan Law Review 89: 69-151. Baumol, William J. and Alan S. Blinder. 1985. Economics: Principles and Policy, 3rd edition. San Diego: Harcourt Brace Jovanovich. Benn, Stanley I. 1988. A Theory of Freedom. Cambridge: Cambridge University Press. Bentham, Jeremy. 1789. Principles of Morals and Legislation. Many editions. Bentham, Jeremy, 1843. Anarchical Fallacies. In The Works of Jeremy Bentham. John Browning, ed.,. vol. II. Pp. 488-534. Edinburgh: William Tait. Berkeley, George, 1712. Passive Obedience, or the Christian Doctrine of Not Resisting the Supreme Power, Proved and Vindicated upon the Principles of the Law of Nature. Many Editions. Blackburn, Simon. 1984. Spreading the Word. Oxford: Clarendon Press. Blum, Walter J. and Harry Kalven, Jr. 1953. The Uneasy Case for Progressive Taxation. Chicago: University of Chicago Press. Bond, E. J. 1983. Reason and Value. Cambridge: Cambridge University Press. BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge, MA: Harvard University Press. Brandon, Robert N. and Richard M. Burian. 1984. eds., Genes, Organisms, Populations: Controversies Over the Units of Selection. Cambridge, Mass.: MIT Press. Brandt, Richard B. 1944. "The Significance of Differences of Ethical Opinion for Ethical Rationalism". Philosophy and Phenomenological Research 4: 469-94. —. 1952. "The Status of Empirical Assertion Theories in Ethics". Mind 61: 458-79.

—. 1959. Ethical Theory: The Problems of Normative and Critical Ethics. Englewood Cliffs, NJ: Prentice-Hall. —. 1963. "Toward a Credible Form of Utilitarianism", in H. Castañeda and G. Nakhnikian, eds., Morality and the Language of Conduct. Pp. 107-43. Detroit: Wayne State University Press. -249—. 1966. "The Concept of Welfare", in Sherman Roy Krupp, ed., The Structure of Economic Science. Pp. 257-76. Englewood Cliffs, NJ: Prentice‐ Hall. —. 1967a. "Happiness", in Paul Edwards, ed., The Encyclopedia of Philosophy. vol. 3. Pp. 432-35. New York: The Macmillan Company & The Free Press. —. 1967b. "Some Merits of One Form of Rule-Utilitarianism". University of Colorado Studies in Philosophy. Pp. 39-65. Reprinted in Brandt 1992. —. 1972. "Rationality, Egoism, and Morality". The Journal of Philosophy 69: 681-97. —. 1976. "The Psychology of Benevolence and Its Implications for Philosophy". The Journal of Philosophy 73: 429-53. —. 1979. A Theory of the Good and the Right. Oxford: Clarendon Press. —. 1982. "Two Concepts of Utility", in H. B. Miller and W. H. Williams, eds., The Limits of Utilitarianism. Pp. 169-85. University of Minnesota Press: Minneapolis, MN. Reprinted in Brandt 1992. —. 1983a. "The Concept of Rational Action". Social Theory and Practice 9: 143-64. —. 1983b. "The Concept of a Moral Right and its Functions". Journal of Philosophy 80: 29-45. Reprinted in Brandt 1992. —. 1983c. "Problems of Contemporary Utilitarianism: Real and Alleged", in Norman Bowie, ed., Ethical Theory in the Last Quarter of the Twentieth Century. Pp. 81-105. Indianapolis: Hackett. —. 1985. "The Explanation of Moral Language", in David Copp and David Zimmerman, eds., Morality, Reason and Truth. Pp. 104-19. Totowa: Rowman and Allenheld. Reprinted in Brandt 1992. —. 1988a. "Fairness to Indirect Optimific Theories in Ethics". Ethics 98: 341‐ 60. Reprinted in Brandt 1992.

—. 1988b. "The Structure of Virtue", in Peter A. French, Theodore E. Uehling, Jr. and Howard K. Wettstein, eds, Midwest Studies in Philosophy Volume XIII. Pp. 64-82. Notre Dame: University of Notre Dame Press. —. 1988c. "Act-Utilitarianism and Metaethics", in Seanor and Fotion. Pp. 29-41. —. 1989a. "Fairness to Happiness". Social Theory and Practice 15: 33-58. —. 1989b. "Morality and its Critics". American Philosophical Quarterly 26: 89-100. Reprinted in Brandt 1992. —. 1989c. "Practical Rationality: A Response". Philosophy and Phenomenological Research 49: 125-30. —. 1991. "Overvold on Self-Interest and Self-Sacrifice". Journal of Philosophical Research 16: 353-63. Reprinted in John Heil, ed., Rationality, Morality, and Self-interest. Totowa, NJ: Rowman and Littlefield. Forthcoming. -250—. 1992. Morality, Utilitarianism, and Rights. Cambridge, England: Cambridge University Press. —. 1993. To appear in the proceedings of a conference on preferences, held at the University of Saarlandes, July, 1992. Brink, David O. 1988. "Sidgwick's Dualism of Practical Reason". Australasian Journal of Philosophy 66: 291-307. —. 1989. Moral Realism and the Foundations of Ethics. Cambridge: Cambridge University Press. Brown, Harold. 1988. Rationality. London: Routledge and Kegan Paul. Buchanan, Allen. 1985. Ethics, Efficiency, and the Market. Totowa, NJ: Rowman & Allanheld. Calkins, M. W. 1929. Ed., Berkeley. New York: Scribner's Sons. Chisholm, Roderick. 1977. Theory of Knowledge. 2nd ed. Englewood Cliffs, N J: Prentice-Hall. Christman, John. (ed.). 1989. The Inner Citadel. New York: Oxford University Press.

Condorcet, Marquis de. 1785. "Essay on the Application of Mathematics to the Theory of Decision-Making". Reprinted in Keith Michael Baker, ed., Condorcet: Selected Writings. Pp. 33-70. Indianapolis: 1976. Daniels, Norman. 1983. "Can Cognitive Psychotherapy Reconcile Reason and Desire?" Ethics 93: 772-85. Darwall, Stephen L. 1983. Impartial Reason. Ithaca: Cornell University Press. Davis, Michael. 1987. "Realistic Utilitarianism and the Social Conditions of Cognitive Psychotherapy". Social Theory and Practice L3: 237-59. —. 1991. "Gewirth and the Pain of Contradiction". Philosophical Forum 22: 211-27. DePaul, Michael R. 1987. "Two Conceptions of Coherence Methods in Ethics", Mind 96: 463-81. Descartes. 1628 or 1629. Rules for the Direction of the Mind. Not printed until 1701. Dewey, John. 1929. The Quest for Certainty. Many editions. Diener, Ed. 1984. "Subjective Well-being". Psychological Bulletin 95: 542-75. Dworkin, Gerald. 1981. "Concept of Autonomy", in Rudolph Haller, ed., Science and Ethics. Pp. 203-13. Amsterdam: Rodopi Press. Reprinted in Christman. Pp. 54-62. —. 1988. The Theory and Practice of Autonomy. New York: Cambridge University Press. Dyke, C. 1981. Philosophy of Economics. Englewood Cliffs, NJ: Prentice-Hall. Ellis, Ralph D. 1988. "Factual Adequacy and Comparative Coherentism in Ethical Theory". The Southern Journal of Philosophy 26: 57-81. Feinberg, Joel. 1989. "Autonomy", in Christman. Pp. 27-53. Originally published as ch. 18 of Feinberg's Harm to Self. New York: Oxford University Press, 1986. -251Fiske, Alan Page. 1991. Structures of Social Life. New York: Free Press. Fox, Michael W. 1974. Understanding Your Dog. New York: Coward, McCann, and Geoghegan. Frankfurt, Harry G. 1971. "Freedom of the Will and the Concept of a Person". Journal of Philosophy 68: 5-20. Reprinted in Christman. Pp. 63-76.

—. 1987. "Equality as a Moral Ideal". Ethics 98: 21-43. Frankena, William. 1963. Ethics. Englewood Cliffs, NJ: Prentice-Hall. 2nd edition, 1973. —. 1964. "History of ethical theory in America from 1930 to 1960 in Philosophy", in R. Schlatter, ed., The Princeton Studies. Pp. 373-75. Englewood Cliffs, NJ: Prentice-Hall. Friedman, Milton. 1947. "Lerner on the Economics of Control". The Journal of Political Economy 55: 405-16. Friedman, Richard B. 1968. "An Introduction to Mill's Theory of Authority", in J. B. Schneewind, ed., Mill: A Collection of Critical Essays. Pp. 379-425. Garden City, NY: Doubleday. Gauthier, David. 1977. "The Social Contract as Ideology". Philosophy and Public Affairs 6: 130-64. —. 1986. Morals By Agreement. Oxford: Clarendon Press. Gert, Bernard. 1977. "Irrational Desires", in J. R. Pennock and J. Chapman, eds., Human Nature in Politics. Pp. 286-91. New York: New York University Press. —. 1988. Morality: A New Justification of the Moral Rules. New York: Oxford University Press. —. 1990. "Rationality, Human Nature, and Lists". Ethics 100: 279-300. Gibbard, Allan. 1988. "Hare's Analysis of 'Ought' and its Implications", in Seanor and Fotion. Pp. 57-72. —. 1990. Wise Choices, Apt Feelings: A Theory of Normative Judgment. Cambridge, MA: Harvard University Press. —. 1992m. "Moral Concepts: Substance and Sentiment". Philosophical Perspectives 6: 199-221, James E. Tomberlin, ed. (Atascadero, CA: Ridgeview Publishing Co.). —. 1992r. "Reply to Blackburn, Carson, Hill, and Railton". Philosophy and Phenomenological Research 52: 969-80. —. 1993. "Reply to Sinnott-Armstrong". Philosophical Studies 66 (forthcoming). Goldstick, D. 1991. "Distributive Justice and Utility". The Journal of Value Inquiry 25: 65-71. Grice, Paul. 1991. The Conception of Value. Oxford: Clarendon Press.

Griffin, James, 1986. Well-Being: Its Meaning, Measurement and Moral Importance. Oxford: Clarendon Press. -252Hajdin, Mane. 1990. "External and Now-for-Then Preferences in Hare's Theory". Dialogue 29: 305-10. Hampton, Jean. 1991. "Two Faces of Contractarian Thought", in Vallentyne. Pp. 31-55. Hardin, Russell. 1980. "The Emergence of Norms". Ethics 90: 575-87. —. 1982. Collective Action. Baltimore: Johns Hopkins University Press for Resources for the Future. —. 1988. Morality within the Limits of Reason. Chicago: University of Chicago Press. —. 1990. "Incentives and Beneficence". Social Justice Research 4: 87-104. —. 1991. "Altruism and Mutual Advantage". Paper presented to the conference, Altruism: Exploring the Intellectual Concept, School of Social Service Administration, University of Chicago, 7-8 November 1991. —. 1992a. "Commonsense at the Foundations", in Bart Schultz, ed., Essays on Henry Sidgwick. Pp. 143-60. Cambridge, Cambridge University Press. —. 1992b. "Institutional Morality". Paper presented at the conference, The Design of Institutions, Australian National University, July 16-17. —. 1993. "My University's Yacht: Morality and the Rule of Law", in Randy E. Barnett and Ian Shapiro, eds., NOMOS 26: The Rule of Law. New York: New York University Press. Hare, R. M. 1952. The Language of Morals. Oxford: Clarendon Press. — 1955. "Universalisability". Aristotelian Society 55: 295-312. Reprinted in Hare 1972a and in Hare 1989a. —. 1963. Freedom and Reason. Oxford: Clarendon Press. —. 1972a. Essays on the Moral Concepts. London: Macmillan. —. 1972b. "Rules of War and Moral Reasoning". Philosophy and Public Affairs 1: 16773.

—. 1973. Critical Study of Rawls 1971. Philosophical Quarterly 23: 144-55, 241‐ 51. Reprinted in Hare 1989a and in Norman Daniels, ed., Reading Rawls. Oxford: Blackwell Publishers. —. 1976. "Ethical Theory and Utilitarianism", in H. D. Lewis, ed., Contemporary British Philosophy 4. Pp. 113-32. London: Allen and Unwin. Reprinted in Hare 1989a and in A. Sen and B. Williams, eds, Utilitarianism and Beyond. Cambridge: Cambridge University Press. —. 1979: "Utilitarianism and the Vicarious Affects", in E. Sosa, ed., The Philosophy of Nicholas Rescher. Dordrecht: Reidel. Reprinted in Hare 1989a: 231-44. —. 1981a. Moral Thinking: Its Levels, Method and Point. Oxford: Clarendon Press. —. 1981b. Review of Singer 1981. New Republic, February 7. —. 1988. "Comments", in D. Seanor and N. Fotion, eds., Hare and Critics. Pp. 199-293. Oxford: Clarendon Press. -253—. 1989a. Essays in Ethical Theory. Oxford: Clarendon Press. —. 1989b. "Brandt on Fairness to Happiness". Social Theory and Practice 15: 33-58. —. 1991. "Justice and Equality", in John Arthur and William H. Shaw, eds., Justice and Economic Distribution. 2nd edition. Pp. 118-32. Englewood Cliffs, NJ: Prentice-Hall. —. 1992. Essays on Religion and Education. Oxford: Clarendon Press. Harrod, R. F., 1937. "Utilitarianism Revised". Mind 45: 137-56. Harsanyi, John. 1977. Rational Behavior and Bargaining Equilibrium in Games and Social Situations. Cambridge: Cambridge University Press. —. 1980. Essays on Ethics, Social Behavior, and Scientific Explanation. Dordrecht: Reidel. Haworth, Lawrence. 1984. "Autonomy and Utility". Ethics 95: 5-19. Reprinted in Christman. Pp. 155-69. Haslett, D. W. 1986. "Is Inheritance Justified?" Philosophy and Public Affairs 15: 12255. Hill, Thomas, E. 1989. "Kant's Theory of Practical Reason". Monist 72: 363-83.

Hobbes, Thomas. 1651. Leviathan. C. B. Macpherson, ed. Harmondsworth: Penguin, 1968. Hodgson, D. H., 1967. Consequences of Utilitarianism. Oxford: Clarendon Press. Hoffman, Martin. 1989. "Emotions and Justice in Society". Social Science Research 3: 283-311. Hooker, Brad. 1990. "Rule-Consequentialism". Mind 99: 67-77. —. 1991. "Rule-Consequentialism and Demandingness: Reply to Carson". Mind 100: 269-76. —. 1994. "Rule-Consequentialism, Incoherence, and Fairness". Proceedings of the Aristotelian Society. Forthcoming. Hume, David. 1738-40. A Treatise of Human Nature. Many editions. Johnston, Mark. 1989. "Dispositional Theories of Value". Proceedings of the Aristotelian Society supp. vol. 63: 139-74. Kagan, Shelly. 1986. "The Present-Aim Theory of Rationality". Ethics 96: 746-59. Kant, Immanuel. 1785 [1964]. Groundwork of the Metaphysic of Morals. Translated by H. J. Paton, 3rd edition. New York: Harper, 1964. —. 1797 [1909]. "On a Supposed Right to Tell Lies from Benevolent Motives". Many editions. E.g., pp. 361-65 in Thomas Kingsmill Abbott, ed. and trans., Kant's Critique of Practical Reason and Other Works on the Theory of Ethics. London: Longmans, 1909, 6th edition. Kavka, Gregory. 1984. "The Reconciliation Project", in David Copp and David Zimmerman, eds., Morality, Reason and Truth. Pp. 297-319. Totowa, NJ: Rowman and Allenheld. Reprinted in John Heil, ed., Rationality, Morality, and Self-interest. Rowman and Littlefield, forthcoming. -254—. 1986. Hobbesian Moral and Political Theory. Princeton: Princeton University Press. —. 1988. "Some Neglected Liberal Aspects of Hobbes's Philosophy". Hobbes Studies 1: 91-93. Klenk, Virginia. 1983. Understanding Symbolic Logic. Englewood Cliffs, NJ: PrenticeHall.

Korsgaard, Christine M. 1989. "Kant's Analysis of Obligation: The Argument of Foundations I". Monist 72: 311-40. Karelis, Charles. 1986. "Distributive Justice and the Public Good". Economics and Philosophy 2: 101-25. Kuhn. Thomas. 1962. The Structure of Scientific Revolutions. Chicago: University of Chicago Press. Langford, C. H. 1949. "A Proof that Synthetic A Priori Propositions Exist". Journal of Philosophy 46: 20-4. Lemos, Noah. 1985. "Brandt on Rationality, Value, and Morality". Philosophical Studies 45: 79-93. Lerner, Abba P. 1944. The Economics of Control. New York: Macmillan. Lewin, Kurt. 1938. The Conceptual Representation and the Measurement of Psychological Forces. Durham, NC: Duke University Press. Lewis, C. I. 1946. An Analysis of Knowledge and Evaluation. LaSalle, IL: Open Court. Lewontin, R. C. 1991. "Facts and the Factious in Natural Sciences". Critical Inquiry 18: 140-53. Lizhi, Fang. 1989. "Keeping the Faith". Translated by James H. Williams and Orville Schell, New York Review of Books, December 21. Locke, John. 1690 and 1706. An Essay Concerning Human Understanding. Lyons, David. 1965. The Forms and Limits of Utilitarianism. Oxford: Clarendon Press. Maclean, Charles. 1980. Island on the Edge of the World: The Story of St. Kilda. New York: Taplinger. Matson, W. I. 1954 [1967]. "Kant As Casuist". Reprinted in Robert Paul Wolff, ed., Kant: A Collection of Critical Essays. Pp. 331-36. Garden City, NY: Anchor, 1967. McGinn, Colin. 1983. The Subjective View. Oxford: Oxford University Press. Melden, A. I. 1966. "Utility and Moral Reasoning", in Richard T. De George, ed., Ethics and Society: Original Essays on Contemporary Moral Problems. Pp. 173‐ 96. Garden City, NY: Doubleday. Mendelson, Elliott. 1987. Introduction to Mathematical Logic. 3 rd edition. Monterey, CA: Wadsworth and Brooks/Cole Advanced Books and Software.

Michalos, Alex C. 1988. "A Case for a Progressive Annual Net Wealth Tax". Public Affairs Quarterly 2: 105-40. -255Mill, J. S. 1832-3? "On Marriage", in John Stuart Mill, Collected Works, vol. 21, Essays on Equality, Law, and Education. John M. Robson, ed. Toronto: University of Toronto Press, 1984. —. 1843. System of Logic. —. 1848. Principles of Political Economy. Many editions. —. 1861a. Utilitarianism. Many editions. —. 1861b. On Representative Government. Many editions. Morris, Christopher. 1991. "Moral Standing and Rational Choice Contractarianism", in Vallentyne. Pp. 76-95. Nagel, Thomas. 1979. "The Fragmentation of Value", in Nagel's Mortal Questions. Pp. 128-41. Cambridge: Cambridge University Press. —. 1986. The View From Nowhere. New York: Oxford University Press. Nathanson, Stephen. 1985. The Ideal of Rationality. Atlantic Highlands, NJ: Humanities Press International. Revised edition forthcoming (LaSalle, IL: Open Court, 1994). —. 1991. Review of Brown 1988. Philosophy and Phenomenological Research 51: 44851. Nozick, Robert. 1974. Anarchy, State, and Utopia. New York: Basic Books. —. 1989. The Examined Life: Philosophical Meditations. New York: Simon & Schuster. Overvold, Mark. 1980. "Self-Interest and the Concept of Self-Sacrifice". Canadian Journal of Philosophy 10: 105-18. —. 1982. "Self-interest and Getting What You Want", in H. B. Miller and W. H. Williams (eds.), The Limits of Utilitarianism. Pp. 186-94. Minneapolis: University of Minnesota Press. Parfit, Derek. 1984. Reasons and Persons. Oxford: Clarendon Press. —. 1986. "Comments". Ethics 96: 832-72.

Pechman, Joseph. 1987. Federal Tax Policy. Washington: The Brookings Institution. Peacocke, Christopher. 1983. Sense and Content. Oxford: Oxford University Press. —. 1986: "Reply to Michael Smith". Synthese 68: 577-80. Peffer, R. G. 1990. Marxism, Morality, and Social Justice. Princeton: Princeton University Press. Pettit, Philip. 1991. "Realism and Response-Dependence". Mind 100: 587-626. —. 1993. The Common Mind. Oxford: Oxford University Press. —. and Michael Smith. 1993. "Practical Unreason". Mind 102: 53-79. —. and Peter Menzies. 1993. "Found: The Missing Explanation". Analysis 53: 100-109. Pfaffman, C., M. Frank, and R. Norgren. 1979. "Neuromechanism and Behavioral Aspects of Taste". Annual Review of Psychology 30: 283-325. Pigou, A. C. 1932 [1952]. The Economics of Welfare. 4th edition. London: Macmillan. -256Plato: Cratylus. Many editions. —. Gorgias. Many editions. —. Phaedo. Many editions. —. Republic. Many editions. Postow, B. C. 1989. "Criteria for Theories of Practical Rationality: Reflections on Brandt's Theory". Philosophy and Phenomenological Research 49: 69-88. Prichard, H. A. 1932. "Duty and Ignorance of Fact". Proceedings of British Academy 18. Reprinted in Prichard's collection, Moral Obligation (Oxford: Oxford University Press, 1949). Putnam, Hilary. 1981. Reason, Truth, and History. Cambridge: Cambridge University Press. Railton, Peter. 1986. "Facts and Values". Philosophical Topics 14: 5-31. —. 1989. "Naturalism and Prescriptivity". Social Philosophy and Policy 7: 151‐ 74.

Rawls, John. 1955. "Two Concepts of Rules". Philosophical Review 64: 3-32. —. 1971. A Theory of Justice. Cambridge, MA: Harvard University Press. Raz, Joseph. 1987. "Autonomy, Toleration, and the Harm Principle", in Ruth Gavison, ed., Issues in Contemporary Legal Philosophy: The Influence of H. L. A. Hart. Pp. 31333. New York: Oxford University Press. Regan, Donald. 1985. "On Preferences an Promises: Response to Harsanyi". Ethics 96: 56-67. Reynolds, Lloyd G. 1976. Microeconomics: Analysis and Policy. Homewood, IL: Richard D. Irwin, Inc. Roberts, R. C. 1984. "Will Power and the Virtues". Philosophical Review 93: 227‐ 84. Samuelson, Paul A. 1972. "A. P. Lerner at Sixty", in Robert C. Merton, ed., The Collected Scientific Papers of Paul A. Samuelson, vol. 3. Pp. 643-52. Cambridge, Mass.: MIT Press. —. and William D. Nordhaus. 1985. Economics. 12th edition. New York: McGraw-Hill. Sayre-McCord, Geoffrey. 1985. "Coherence and Models for Moral Theorizing". Pacific Philosophical Quarterly 66: 170-90. Schneewind, J. B. 1977. Sidgwick's Ethics and Victorian Moral Philosophy. Oxford: Clarendon Press. Seanor, D. and N. Fotion. (eds.). 1988. Hare and Critics. Oxford: Clarendon Press. Shaw, William H. 1988. "A Moral Theory for Marxists". Polity 20: 479-97. Sidgwick, Henry. 1907. The Methods of Ethics. 7th edition. London: Macmillan and Co. Singer, Peter. 1975. Animal Liberation. New York: New York Review and Random House. -257—. 1981. The Expanding Circle: Ethics and Sociobiology. New York: Farrar, Straus, & Giroux. Smart, J. J. C., 1961. An Outline of a System of Utilitarian Ethics. Melbourne University Press: Melbourne, Australia. Smith, Michael. 1986. "Peacocke on Red and Red'". Synthese 68: 559-76.

—. 1992. "Valuing: Desiring or Believing?" in David Charles and Kathleen Lennon, eds., Reduction, Explanation and Realism. Pp. 323-59. Oxford: Clarendon Press. Stevenson, Charles. 1983. "Value-judgments: Their Implicit Generality", in Norman Bowie, ed., Ethical Theory in the Last Quarter of the Twentieth Century. Pp. 13-37. Indianapolis: Hackett. Sumner, L. W. 1987. The Moral Foundations of Rights. Oxford: Clarendon Press. —. 1992. "Two Theories of the Good". Social Philosophy and Policy 9: 1-14. Taylor, Charles. 1982. "The Diversity of Goods", in A. Sen and B. Williams, eds., Utilitarianism and Beyond. Pp. 129-44. Cambridge: Cambridge University Press. Thalberg. Irving. 1978 [1989]. "Hierarchical Analyses of Unfree Action". Canadian Journal of Philosophy 8: 211-26. Reprinted in Christman. Pp. 123‐ 36. Urmson, J. O. 1953 [1971]. "The Interpretation of the Moral Philosophy of J. S. Mill." Reprinted in Samuel Gorovitz, ed., Mill: Utilitarianism, Text and Critical Essays. Pp. 168-74. Indianapolis: Bobbs-Merrill, 1971. Vallentyne, Peter. (ed.) 1991. Contractarianism and Rational Choice. Cambridge: Cambridge University Press. Velleman, J. David. 1988. "Brandt's Definition of 'Good'". The Philosophical Review 97: 353-71. Watson, Gary. 1977. "Skepticism about Weakness of Will". Philosophical Review 86: 316-39. Williams, Bernard. 1985. Ethics and the Limits of Philosophy. Cambridge, MA: Harvard University Press. Wood, Allen W. 1981. "Marx and Equality", in John Mepham and David-Hillel Ruben, eds., Issues in Marxist Philosophy, vol. 4: Social and Political Philosophy. Pp. 195-221. Atlantic Highlands, NJ: Humanities Press. Wright, Crispin. 1988. "Moral Values, Projection and Secondary Qualities". Proceedings of the Aristotelian Society, supp. vol. 62: 1-26. Young, Robert. 1980 [1989]. "Autonomy and the Inner Self". American Philosophical Quarterly 17:35-43. Reprinted in Christman. Pp. 77-90. -258-