Social Learning Theory, Self-Regulation, and Morality Author(S): Thomas E

Social Learning Theory, Self-Regulation, and Morality Author(s): Thomas E. Wren Source: Ethics, Vol. 92, No. 3, Special Issue: Symposium on Moral Development (Apr., 1982), pp. 409-424 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/2380729 Accessed: 26/02/2009 04:25

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=ucpress.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Ethics.

http://www.jstor.org SOCIAL LEARNING THEORY Social Learning Theory, Self-Regulation, and Morality*

Thomas E. Wren

Moral psychology divides nicely into three relatively autonomous domains: cognitive-developmental theories of moral judgment, psychoanalytic theories of motivational processes, and social learning theories of moral behaviors and inhibitions. For instance, a well-known review of the literature on moral development opens by correlating these three domains with the romantic view of man as innately perfectible, the reformationist view of man as naturally inclined to evil, and the empiricist conception of the mind as a tabula rasa. Its author, Martin Hoffman, is a prominent social learning theorist, but the same principles of division recur in similar reviews by Lawrence Kohlberg and Jane Loevinger, to name just two major figures in the cognitive-developmental and psychoanalytic tradi- tions, respectively.' Both of these domains have been surveyed in detail by philosophers over the last several years. However, considerably less has been said philosophically about the third one, at least as far as morality is concerned. This is regrettable for many reasons, not the least of which is the dominant position social learning theory holds in present-day Ameri- can psychology. The neglect is not altogether benign: social learning theory has grown up out of behaviorism, and many philosophers regard behaviorist accounts more as capitulations to the opacity of the data than as explanations. It is to remedy this neglect as well as for the intrinsic interest of the issue that I shall discuss some recent variations in social learning theory on a theme that moral philosophy has long regarded as a formal feature of

* Research for this project was made possible by grants from the Spencer Foundation and the National Endowment for the Humanities, which are here gratefully acknowledged. Earlier versions of this paper were presented at the universities of Oxford, London, and Hawaii, and I am grateful for the comments I received on those occasions. 1. M. Hoffman, "Moral Development," in Carmichael's Manual of Child Psychology, ed. P. H. Mussen, 3d ed. (New York: John Wiley & Sons, 1970), vol. 2; L. Kohlberg, "Moral Development," in International Encyclopedia of the Social Sciences, ed. E. R. Seligman (New York: Macmillan Co., 1968), and "Education for Justice: A Modern Statement of the Platonic View," in Moral Education: Five Lectures, ed. N. Sizer and R. Sizer (Cambridge, Mass.: Harvard University Press, 1970): J. Loevinger, Ego Development (San Francisco: Jossey-Bass, Inc., 1976). The most recent, and most comprehensive, review of the literature is A. Blasi, "Bridging Moral Cognition and Moral Action," Psychological Bulletin 88 (1980): 1-45. Ethics 92 (April 1982): 409-424 ? 1982 by The University of Chicago. All rights reserved. 0014-1704/82/9203-0001$01.00

409 410 Ethics April 1982 morality as such: the theme of personal responsibility or self-direction. Because of constraints of space I shall pass over the many interesting things social learning theory has said about the contents of morality, that is, specifically moral (or immoral) behaviors, including the stereotypical ones of helpfulness, truth telling, cheating, and sexual deviance. I shall pass over these constraints in order to discuss the cognitive and moral form which professional and nonprofessional moralists alike commonly ascribe to them in the course of assigning praise, blame, responsibility, and so on. My general thesis is that, for all the research done on morally toned behaviors, there remains within social learning theory a terribly important metaethical deficiency. This deficiency is the altogether inade- quate account taken of the role which self-direction or self-directed prescribing plays in authentically moral conduct and, for that matter, in the whole process of socialization. To give point to my charge I shall take up in turn the very theorists who have come the furthest from the early, avowedly noncognitive views of learning theory regarding the socialization process: Albert Bandura, Walter Mischel, F. H. Kanfer, and Justin Aronfreed. They have developed the concept of socialization from the noncognitive level of merely overt conformity to the much more cognitively sophisticated and potentially moral level of self-regulation, a semitechnical term which some social learning theorists use to designate a level of personal autonomy or nonre- liance on external incentives. Here if anywhere in the vast literature of social psychology we ought to be able to discern a self-prescribing character corresponding to that which is traditionally identified with morality.2 That we cannot is, I think, a cause for profound concern on the part of philosophers and others interested in morality and human nature; but it is not, I hope to show at the end, a cause for despairing of moral psychology altogether. I. SOCIAL LEARNING THEORY AND THE LAW OF EFFECT Social learning theory is not a single theory but rather a family of theories that attempt to explain behavior primarily in terms of learning and learning in terms of the behaviorist credo, the Law of Effect. Consequently, its discussions of morality are distinguished from their cognitive-developmental and psychoanalytic counterparts by the motivational role it assigns to rewards. That is, the moral agent's motivation for being and acting in a moral fashion is understood to be his more or less conscious pursuit of benefits such as social approval, whose reinforcing value con- sists in their culturally contingent attachment to deep-seated drives,

2. Philosophers will recognize that I am taking an "internalist" position on the question of whether there is included, as part of the very meaning of a moral principle, some corresponding motivation on the part of anyone who accepts it (see W. D. Falk, " 'Ought' and Motivation," Proceedings of the Aristotelian Society 48 [1947-48]: 111-38; W. Frankena, "Obligation and Motivation in Recent Moral Philosophy," in Essays in Moral Philosophy, ed. A. I. Melden [Seattle: University of Washington Press, 1958]). Wren Social Learning Theory 411 needs, or aversions. As in psychoanalytic theories, moral development is understood as the acquisition of someone else's norms, either the norms of one's parents (Freud) or those of the society at large (social learning theory). But no unconscious motivational processes are posited, nor are any self-propelling stage-developmental sequences of cognitive structures-in fact, even the most adult phases of morality have minimal cognitive complexity. Like all other forms of human development, moral development or "socialization" is envisioned as a more or less smooth and ordered change in the person's feelings or affects and his behaviors, not in his mode of thinking and evaluating. To the extent that social learning theory comprehends this change, it does so in terms of the Law of Effect, which in the present context declares that a moral agent comes to adopt the morality that he does because its adoption has resulted in more reinforcing events.3 Here "morality" means a socially endorsed pattern of behaviors, with additional connotations of altruism ranging from simple noninterference to sacrificing one's life for others. Just what "reinforcement" means depends on which of two motivational theories is adopted: the expectancy theory or the tension- or drive-reduction theory. To characterize them in the ordinary language terms favored by philosophers, the former theory conceives desire in terms of anticipated pleasures, whereas the latter conceives pleasure-at least one of its most important modalities-as the resolution of desire. Both theories are compatible with the Law of Effect, although the psychologists we shall discuss below generally cast their research in expectancy categories. In short, social learning theory regards moral actions as having the same earthy dependence on reinforcement schedules as all the other kinds of human behavior have. Their respective motivations all work in the same way, competing on the same level. As Alston has pointed out,4 the standard picture offered by American social psychologists is that of a motivational field in which various tendencies, moral, nonmoral, and immoral, engage in a single interaction for control of behavior. As we shall see, even those learning theorists who have come the furthest in detente with cognitivism-viz., those mentioned above as recognizing the cognitive dimension of self-regulation-have this one-level, nonhierarchical picture of human motivation. What makes it nonhierarchical, and so a much more diluted sort of self-regulation than they seem to realize, is the lack of any suggestion of a higher order of motivation whose object is the ordinary first-order motivations or sets of motivations that are keyed to the environment.

3. Contemporary presentations of the Law of Effect sometimes stress that the reward need not be cognized, i.e., perceived as a reward (see the influential textbook by R. Brown and R. J. Herrnstein, Psychology [Boston: Little, Brown & Co., 1975]). The original formu- lations of the Law of Effect were less explicit. 4. W. Alston, "Self-Intervention and the Structure of Motivation," in The Self: Psy- chological and Philosophical Issues, ed. T. Mischel (Oxford: Blackwell's, 1977), p. 91. 412 Ethics April 1982 This one-level picture is entirely in keeping with the Law of Effect which shapes it. It is not, however, in keeping with the conception of morality that operates in the minds of most moral agents and in the theories of most moral philosophers. Social psychologists have picked up part of the ordinary conception of morality in their tendency to regard prosocial actions as coextensive with morality. But this is only one part of the concept of morality, if indeed it is a necessary part at all. I take it as self-evident that, if an action like returning a library book or throwing oneself on a live grenade is to count as moral, it must display some kind of self-determination in virtue of which it is more than mindless conformity or unwitting reflex.5 Indeed, it seems plausible that the common intuition that self-regulation is a necessary condition for moral action somehow lies behind the fact that the pejorative stereotype of a moralistic personality is virtually synonymous with that of excessive self-regulation or repression. II. SOME SOCIAL LEARNING THEORIES OF SELF-REGULATION One might expect the connection between the concepts of self-regulation and morality or moral development to be apparent. After all, self-regulation is the outcome of an agent's progression from passive heteronomy to a relative independence from externally provided reinforcements for the initiation or maintenance of certain actions. However, as matters now stand the connection in psychology between these two concepts is more apparent than real, since the picture drawn there of self-regulation is too simple to include the quintessentially moral self-prescribing which, following Hare and others, I shall hereafter simply call "prescriptivity." To recycle two of Ryle's terms, my thesis is that when social learning theorists treat the morally central phenomenon of self-regulation, they unwittingly replace the traditional parapolitical model of morality with a paramechanical one which is essentially nonhierarchical in the sense just given, and hence not truly prescriptive in the sense of a demand an agent makes of himself. Whenever some conscience-like motivational construct like an internalized value is portrayed in social learning theory as dominating the hurly-burly of the self-regulating agent's diverse tendencies and desires, it is thought to dominate not by any privileged status or authority which it has (as in a political system) but only by its greater power (as in a mechan- ical system). The underlying idea of this view of moral motivation is not the idea of political dominion but rather of control by sheer strength, employed in a good cause-social harmony-but utterly coercive none- theless. 5. At this point a tough-minded behavioral scientist might object that my line of argu- ment forces data into outmoded speculative molds. So much the worse for morality, he might say, if psychological inquiry discloses no prescriptive component. About the only riposte space permits is that such a move would amount to much more than revising some philosophical doctrine of what morality means-it would be to alter our primitive self- conceptions of what we mean, of "what it is like" to be a moral person. T. Nagel makes a related point in his "What Is It Like to Be a Bat?" in Mortal Questions (New York: Cam- bridge University Press, 1979). Wren Social Learning Theory 413

The most simple and familiar paramechanical representation of in- trapsychic conflict is that which portrays an organism as caught in an inherently nonrational force field of desires, simultaneously wanting to move toward and away from some global state of affairs-for instance, a rat for whom getting food also involves getting a shock. Philosophers will recognize this as Thomas Hobbes's model, as well as that of the ancient atomists. A modern version, which has prepared the way for today's social learning approaches to self-control, is the so-called approach-avoidance model of conflict developed by Kurt Lewin and a little later by Neal Miller.6 In this model, as well as in the related ones of approach-approach (the subject is pulled by two competing goods), avoidance-avoidance (two evils), and double approach-approach (two good-evil pairs), conflict is envisioned as a competition of desires. In each case the respective pro and con desires vary with regard to a variety of factors such as the distance from the goal, the strength of the underlying drive, and (here some small bit of rationality enters the story) the probabilities with which the goal or threat is perceived and expected. But for all their complexities, every conflict is understood in the final analysis as being resolved in favor of the strongest desire. In fact, even this formulation is misleading: victory is not something which is awarded by the agent to the strongest desire the way a prize is awarded by a judge to the fastest runner, but rather simply is the fact of its having overpowered the other contenders. Here there is no question of a desire's entitlement or being "declared" the winner- notions which suppose a set of rules under which a contender's efforts are evaluated-but only the question of whether or not it has prevailed. The approach-avoidance model was originally drawn up in language appropriate to drive- or tension-reduction theories rather than in terms of expectancies. However, like the Law of Effect on which it is based, it is compatible with either of these two conceptions of desire. (Thus the point made in the preceding paragraph about conflict resolution can be restated as the triumph of the strongest or most vividly anticipated affect. This is the version that applies to the social learning theorists under discussion.) Most of the experimental work on conflict situations has studied conflicts of a rather elemental nature, usually with nonmoral problems and non- human subjects. But in the following sections we shall see the same logic at work in contemporary psychological analyses of higher sorts of conflicts, in which conscience or its equivalent is conceived in the same coercive rather than parapolitical way as the approach-avoidance vectors of hunger and fear are supposed to operate on rats in a laboratory.7

6. K. Lewin, The Conceptual Representation and the Measurement of Psychological Forces (Durham, N.C.: Duke University Press, 1938); N. Miller, "Studies of Fear as an Acquirable Drive: I. Fear as Motivation and Fear-Reduction as Reinforcement in Learning of New Responses," Journal of Experimental Psychology 38 (1948): 89-106. 7. Miller, and later Brown, equipped rats with little harnesses attached to apparatuses which measured in units of physical force how strongly the rats were pulling toward the goal box. He then correlated this measurement with a measurement of distance from the goal (or in subsequent experiments by Brown, the "temporal distance") and so determined 414 Ethics April 1982

Bandura: Modeling The first of the self-regulation constructs which we shall consider is that of imitative modeling, as developed by Bandura and his associates.8 It is, quite simply, a theory about the power of example. They argue that any behavior, including specific acts of self-control, can be acquired through imitation. What is learned in the modeling process is, in their view, essentially an image, which serves as a guide or standard of appropriate- ness for subsequent behavior. It is also motivational, though only in conjunction with other reinforcements whereby it is realized in action.9 In Bandura's early work, observation of a model seemed to be a sufficient condition for the occurrence of imitative behavior, which reflects his behaviorist background ("observation of a model" is, supposedly, itself a piece of overt behavior). However, he soon made it clear that internal acts or states of attention and retention involving verbal or nonverbal images are also necessary. For this reason, as well as because of the element of expectation involved in the reinforcement processes, Bandura's theory has a considerably more cognitive tone to it than does simple behaviorism or, for that matter, the early psychoanalytic theory of repression by cathexes lying outside the ego. But it would be wrong to make too much of this point, for the only status which the image acquired by observing a model enjoys in the learner's consciousness is the minimal functional status of an affect- laden trigger (more or less decisive, depending on other circumstances) of behavior. It determines behavior in the same general way as do the early expectancy constructs applied to animal learning, namely, by way of portending those stimuli-noxious or pleasant, depending on whether the self-control is envisioned as repressing or inciting-that somehow have come to be associated with the acts in question. What is added by Bandura to this familiar stimulus-response (S-R) picture of the Law of Effect is, first of all, the idea that the association of affective stimuli and acts can be wrought vicariously as well as by instru- mental and classical conditioning. Bandura makes the important point that reinforcement by vicarious association works in virtue of the cogni- the strengths of the approach and avoidance tendencies, i.e., fear and hunger. Because both sorts of tendency were involved at the same time-the rats feared getting shocked as they approached the desired food-and because each tendency had a different gradient or rate of increase, it was necessary to infer their separate strengths indirectly by a fairly elaborate procedure, but the basic idea was simply that of attaching empirical, cardinal values to desires (see J. S. Brown, "Principles of Intrapersonal Conflict," Conflict Resolution 1 [1957]: 135-54). 8. W. Bandura, "Influence of Models' Reinforcement Contingencies on the Acquisi- tion of Imitative Responses," Journal of Personality and Social Psychology 1 (1965): 589-95, Principles of Behavior Modification (New York: Holt, Rinehart & Winston, 1969), Social Learning Theory (Englewood Cliffs, N.J.: Prentice-Hall, 1977, and (with R. Walters) Social Learning and Personality Development (New York: Holt, Rinehart & Winston, 1963). 9. As learning theorists they are committed to the axiom that the acquisition of any given repertoire of behavior does not guarantee its performance (see E. C. Tolman, "There Is More Than One Kind of Learning," Psychological Review 56 [1949]: 144-55). Wren Social Learning Theory 415 tive character of the mediation between observing the model and acting accordingly. But this is a very minimal sense of "cognitive": no inferen- tial or computational processes are posited beyond the general, undiffer- entiated one of imaginative representation.' Bandura admits that the observer's attention to the model is selective in terms of the intentions of both parties-the model and the observer-but he gives no account of what philosophers might call the conceptual side of learning: the general- ization from a set of particular behaviors to an abstract form or structure, be it an athletic skill or a moral disposition such as truth telling, capable of inhabiting any number of potential behaviors. In Bandura's view, the affect attached to a cognitive process is not a function of the meaning of the image, as cognitive-developmentalists like Kohlberg would hold, but only of the logically contingent associations which that image has en- joyed in the learning history of the agent. In particular, moral ideals and principles enjoy whatever motivational power they might have inde- pedently of whatever cognitively privileged status they might have as impartial or "good" reasons. A second addition by Bandura and his associates to the standard S-R picture is the idea of self-reinforcement, an idea which was quickly picked up and developed in a number of studies of adults and children adminis- tering their own rewards and punishments. This idea is a by-product of the more general one of modeling another's behaviors, since the notion of self-reinforcement at stake here is really a specific and somewhat more intricate form of modeling. The child imitates the judgmental behavior of an adult whom he has observed reinforcing-verbally as well as ma- terially-the child's own behaviors. In lieu of anyone else to reinforce, the child then proceeds to heap praise or blame upon himself. The connection between this sort of self-reinforcement and imitative modeling is perhaps best seen in D. L. Rosenhan's discussion of the paradox altruism presents for learning theory (since altruism is presumably engaged in without thought of reward). Rosenhan's proposal, which he points out is in line with Bandura's view, is that being charitable ("contributing") can contain within itself a special kind of self-reinforcement as follows: "In the context of imitating a model who has just been charitable, this view would hold that children tell themselves that they are 'good' because they have contributed, which allegedly is nearly as good as having someone tell them that they are good, or even giving them money for having been good.""II Unfortunately, how children graduate from the childish pleasure

10. To be fair, we should recognize the increasingly cognitive dimension of Bandura's recent work, a tendency which typifies the behavioral sciences in general. Nevertheless, he continues to minimize any semantical import that evaluative symbols used in self-regulation might have, writing, for instance, "Values can be invested in activities themselves as well as in extrinsic incentives. As we have seen, the value does not inhere in the behavior itself but rather in the positive and negative self-reactions it generates" (Social Learning Theory, pp. 139-40). 11. D. L. Rosenhan, "Learning Theory and Prosocial Behavior," Journal of Social Issues 28 (1972): 151-63. 416 Ethics April 1982 of telling themselves that they are good to a more adult form of altruism, or whether such a graduation is even possible, remains unexplained. To summarize, Bandura provides an account of two moments of self-regulation, the general one of imitative modeling and the derivative self-reinforcing one of imitating another's way of distributing praise and blame. In each case behavior is modified or inhibited, thanks to something the agent himself has done to change the motivational field within which he functions. So conceived, this process seems to deserve the title "self-regulation," but it is clearly an instantiation of the paramechanical model and not the parapolitical one. The image of the model which the agent cognizes is effective thanks to the vicarious but otherwise classical conditioning of affect (typically but not necessarily negative) to certain behavioral responses and outcomes. No evaluation is made, from a struc- turally higher perspective, of the quality of the motivations and behaviors that are inhibited, or at least no such evaluation is part of the inhibiting or self-regulating process per se. Mischel and Kanfer: Self-Manipulation The studies of self-regulation pursued by W. Mischel and F. H. Kanfer are, in slightly different ways, continuous with Bandura's conceptions of inhibition by imitative modeling and self-reinforcement.'2 They have fo- cused more on the active, self-interventional aspects of internalization than on the impact of a model's example, though Mischel in particular has also made numerous contributions to the literature on modeling.'3 But the basic paradigm remains that of expectation of, and consequent motivation by, the proposed behavior's affect-charged consequences, all in accordance with the Law of Effect. The acquisition of self-regulating skills-or as a philosopher would call them, dispositions-proceeds in the same way as that of any social-behavior pattern, namely, by observa- tional learning, conditioning, direct and vicarious reinforcement, gener- alization, and discrimination. As Mischel points out, these processes can be the antecedents of extreme dependency, aggression, and other maladap- tive or antisocial patterns as well as of prosocial patterns of self-control

12. W. Mischel, Personality and Assessment (New York: John Wiley & Sons, 1968), "Processes in Delay of Gratification," in Advances in Experimental Social Psychology, ed. L. Berkowitz (New York: Academic Press, 1974), vol. 7, and (with H. Mischel) "A Cognitive Social-Learning Approach to Morality and Self-Regulation," in Moral Development and Behavior, ed. T. Lickona (New York: Holt, Rinehart & Winston, 1976); F. H. Kanfer, "Self-Regulation: Research, Issues, and Speculations," in Behavior Modification in Clinical Psychology, ed. C. Neuringer and J. Michael (New York: Appleton-Century-Crofts, Inc., 1970), "The Maintenance of Behavior by Self-generated Stimuli and Reinforcement," in Psychology of Private Events, ed. A. Jacobs and L. Sachs (New York: Academic Press, 1971), and (with P. Karoly) "Self-Control: A Behavioristic Excursion into the Lion's Den," Behav- ior Therapy 3 (1972): 398-416. 13. See W. Bandura and W. Mischel, "Modification of Self-imposed Delay of Reward through Exposure to Live and Symbolic Models," Journal of Personality and Social Psy- chology 2 (1965): 698-705. Wren Social Learning Theory 417 and delay of gratification.'4 What is deemed special about self-regulation is simply that in it some of the reinforcing consequences of behavior are dispensed by the individual rather than his external physical or social environment. This point is slightly obscured in Mischel's delay-of-gratification paradigm, where by hypothesis the self-control exercised in waiting for a delayed, larger reward fetches an externally administered payoff. But the agent is active nevertheless, since he achieves this sort of self-control by fixing his attention on the more remote consequences. Furthermore, Mis- chel recognizes that people typically use a variety of other covert and overt self-manipulative strategies (e.g., muttering to oneself) to maintain their resistance to temptation: in this respect his model overlaps with Kanfer's "closed-loop learning model."' 5 In the latter model of self-regulation, the individual acts as his own behavior modifier, his goal being to influence the probability of a given response at the end of some behavioral chain. As in any therapy game, the only basic rule is that it must succeed: at the end of the learning process the subject must be able to function in the absence of immediate external supports. The individual can be weaned away from these supports by whatever manipulative strategies the wit of the modifier himself can devise. But in these strategies of self-regulation one substrat- egy that is always employed is that of setting up "ad hoc performance standards" in terms of which the behavior is reinforced. Kanfer calls this central element a performance promise or contract, though it need not be an overt act in its own right.'6 Once this intrinsically arbitrary standard has been set, its discrepancies with the person's actual performances can be discerned and reduced by diverse reinforcements. A typical example of the closed-loop model in action is that of a smoker who gives himself little prizes for smoking only a given number of cigarettes. In the rather inflated jargon of Kanfer and Karoly, the person's self-monitoring and self- evaluation produces a judgment, which then serves as an SD (drive stimulus) "either for positive self-reinforcement (SR+), if the outcome of the comparison is favorable, or for self-presented aversive stimulation (SR-), if the comparison is unfavorable.""l7 How the reinforcements are "self-presented" depends on the nature of the reward or punishment. Kanfer stresses the presentation of external tokens, such as an extra helping of dessert, which one bestows upon oneself. But Mischel recognizes that people have other, more internal resources with which to reward or punish themselves, namely, self-praise and blame; and Bandura has recently described the same phenomenon in 14. Mischel, Personality and Assessment, p. 188. 15. Kanfer and Karoly. 16. "The use of the term 'contract' should not be construed to imply more than a convenient analogy. Its purpose is to facilitate a rudimentary conceptual organization and hypothesis generation by borrowing from an area that has already integrated the relation- ships between standards, performance criteria and payoff conditions for many social prac- tices" (ibid., p. 408). 17. Ibid., p. 406. 418 Ethics April 1982 terms of "self-esteem costs."18 Nevertheless, for all their moral overtones these internal sanctions are thought to function as morally neutral affective experiences. In this respect they are indistinguishable from the external tokens in Kanfer's model. It is true that most of the contents in our internal sanctions of self-praise and blame are aligned very straightfor- wardly with those of moral values, which is only to be expected in soci- eties whose public mores are also the private morality of most of its members. But the relation can be reversed with no logical contradiction, as in Huck Finn's famous moral struggle about whether he should betray the runaway slave, Jim. Huck was surely correct in his prediction that not turning Jim in would induce a heavy overhead for himself in terms of guiltlike affects and self-esteem costs, but Huck's willingness to suffer that sort of distress displays considerable moral decency. In other words, there is a philosophical gap between morality and the kinds of self-regulation that Mischel and Kanfer describe. They are simply not talking about the same sort of psychological phenomena that Twain and moral philosophers have portrayed. The ability to wait for a delayed reward-or whatever other disposition might have been aimed at as the terminus of a self-regulated behavioral chain-is a tendency which, like all tendencies discussed in behavioristic sorts of theories, is conceived in conjunction with the Law of Effect as the mathematical probability of certain overt acts or act-types taking place. Furthermore, at least for the theorists I am discussing here, this probabilistic conception is based in turn on another, less purely behaviorist conceptualization of the acquired tendency as the resultant of motivational forces that, when summed, outweigh the competing inclinations (e.g., to go for the immediate reward). Here again the dominant model is paramechanical. The self is regulated, to be sure, but not by any sort of inner self. No supervening principles are involved, no prescriptive moral ideals are imaged, whereby the inclination, say, to wait for the birds in the bush gets an edge over the inclination to take the bird which is at hand. It is (thanks to the expected schedule of payoffs) just stronger, and regularly so. In short, by understanding conscience, or the conscience-like motivations discussed by Kanfer and Mischel, solely in terms of reinforcements (i.e., as the expectation of pleasurable and painful affects) social learning theory puts the apparently executive function of self-regulation on the same level as the other desires and tendencies it is supposed to direct. Consequently, conscience has no special, parapolitical relationship to the other motivations making up the agent's personality. But the whole idea of moral motivation-as opposed to such fortuitously prosocial affec- tions as an otherwise ruthless gangster's fondness for children (Kant's "pathological love")-is that an agent acts or abstains from acting in accordance with some norm that is independently cherished and antecedent- "pre-scribed"-to the specific, first-order motivations embedded

18. Mischel, Personality and Assessment, p. 166; Bandura, Social Learning Theory, p. 143. Wren Social Learning Theory 419 within the behavior. This kind of self-regulation may be considerably more rare than moralists like to think, but that is not the point. The point is that all we are given by Mischel and Kanfer are more one-leveled approach-avoidance models of conflict, in which the negative tendencies associated with the act in question simply outweigh the positive tendencies or vice versa without the benefit of that moral prescriptivity which philosophers and other nonpsychologists take to be at the core of moral behavior.

Aronfreed: Conscience As far as moral psychology is concerned, the most sophisticated model of self-regulation to appear so far among social learning theorists is that introduced by Aronfreed, who explicitly takes up the task of illuminating the contribution that moral judgment makes to conscience.'9 Although he is not the only social learning theorist to take note of the element of self-reproach, Aronfreed's conceptions of conscience and the related, morally toned affects of guilt and shame are in themselves unusual if not unique among authors in that tradition. Also, and still more importantly, he distinguishes within the agent's motivational system between two radically different forms of self-regulation. This distinction, novel among social learning theorists, is between controlling one's own behavior by "evaluative cognition" (which includes Aronfreed's idea of conscience) and controlling it by self-manipulative stratagems whose reflective cognition is not evaluative but only what philosophers of language since Aus- tifl have called a performative, in this case a self-directed threat or bribe proposed in the so-called performance contract. This is not to say that the second kind of self-regulation, which is, of course, that described by Mischel and Kanfer, has no cognitive complexity. Like any performative, it can be regarded as having an outside and an inside, comprising an essentially overt (albeit future-dated) performance X and an essentially covert performative speech act Y-where X would cause or be some rewarding or punishing event or state of affairs, and Y is the internal event of promising oneself that, given the conditions specified in one's self-sanctioning, one will indeed perform X. But as we have just seen, this sort of complexity, although sufficient to set self-regulation off from still less cognitive motivational structures, falls considerably short of moral prescriptivity since the self-regulation it is tied to is fundamentally paramechanical. The question now before us, then, is whether the first kind of self-regulation, Aronfreed's so-called evaluative cognition, can advance the discussion in any significant way. That is, to what extent does it involve a new, truly hierarchical or parapolitical model of conscience and moral inhibition? The answer to this question varies depending on how one reads Aronfreed, since, unfortunately, he is not altogether clear.

19. The phrase is Aronfreed's, in his "Moral Development from the Standpoint of a General Psychological Theory," in Likona, ed., p. 55. 420 Ethics April 1982

According to one reading, which is Alston's,20 Aronfreed's account of moral inhibition is that of a tendency which competes on all fours with other first-order tendencies or desires and so is not a higher-order tendency which influences the agent in opposition to what is his currently strongest tendency or desire. In spite of his sensitivity to the moral domain's special evaluative dimensions, on this reading Aronfreed would be just as closely tied to the nonhierarchical model of moral motivation or, more generally, self-regulation, as were the social learning theorists just considered. Consider, for instance, Aronfreed's analysis of how values engage behavior. After noting that evaluation can be thought of as a cognitive schema consisting of classificatory operations, ranking procedures, etc., he insists that an evaluative structure is not merely a cognitive schema for the economical coding of information. It is the quality and magnitude of the affectivity that becomes associated with particular classifica- tions which permit the structure to enter into the operations of value and to exercise some control over behavior.2' Alston interprets this passage as meaning that "the only way in which the adoption of a principle forbidding adultery can enter into the control of behavior is that negative affect has been associated with such acts." That is, in the course of socialization affect is attached to the actions or outcomes which are the objectives of one's principles or values rather than to the principles or values themselves; and this first view would lead us once again into the paramechanical model in the manner described in the preceding sections. For Alston, the fact that evaluative cognition inserts into the picture of self-control the operations and deliverances of self- judgment is not enough to explain how a moral agent (or for that matter any kind of self-intervening agent) can oppose the dominant trend of his lower-order desires and aversions. Internalized control by conscience might be more cognitively complex than the more rudimentary sort of self-control, but, Alston suggests, "so long as it is affective anticipations that are crucial" even Aronfreed's otherwise discriminating analysis fails. Or in the categories introduced above: because of his use of the expectancy model, Aronfreed misses the prescriptive element of truly moral self-regulation. By way of introducing my own interpretation, I would suggest that Alston's reading of Aronfreed does not take into account the fundamental ambivalence in Aronfreed's lengthy book Conduct and Conscience con- cerning the role of affective anticipations, an ambivalence which is only partly resolved in subsequent essays.22 Like Alston, I take the issue23 to be 20. Alston, pp. 95-96. 21. Justin Aronfreed, Conduct and Conscience: The Socialization of Internalized Con- trol over Behavior (New York: Academic Press, 1968), p. 278. 22. Besides the above-cited "Moral Development" essay, see his "Some Problems for a Theory of the Acquisition of Conscience," in Moral Education: Interdisciplinary Approaches, ed. C. Beck (Toronto: University of Toronto Press, 1971). 23. Note what is not at issue in Alston's paraphrase of Aronfreed as claiming that one must ("the only way," said Alston) associate negative affect with acts of adultery before a Wren Social Learning Theory 421 whether what becomes effectivelyy loaded" in Aronfreed's model are (1) the acts and their outcomes, or (2) the evaluative representations themselves. The former position is tantamount to the paramechanical, nonhierarchical model criticized above. Since it is the one explicitly stated throughout Aronfreed's book, it seems fair enough to say that his general treatment of self-regulation is as oblivious to the prescriptive character of morality as are those of the other learning theorists we have discussed. But the seeds of the second position are also present there,24 in a chapter on fear, guilt, and shame which makes no sense if read Alston's way. In that chapter we see Aronfreed trying manfully to tie the affective and motivating load of moral feelings directly to moral principles, all in a way that would satisfy the parapolitical model's requirement that self-control pro- ceed by appeal to authority, that is, in the name of principles or ideals to which one has a prior attachment. Accordingly, I shall now argue that the differentiation which Aronfreed makes there between the aversive states of fear on the one hand and guilt and shame on the other can be parlayed into a rudimentary but genuinely parapolitical and hierarchical model of moral motivation, in which self-regulation is truly if inexplicably prescriptive.

III. ARONFREED REVISITED: A CONCLUSION Fear, guilt, and shame are introduced by Aronfreed as three forms of 'anxiety," the generic construct embedded in his definitions of these three sorts of feelings. He is noncommittal on the nature of anxiety, probably since it is standard practice when applying the Law of Effect to define functionally such crucial behavior-determining constructs as reinforcement, incentive, and, in the present case, anxiety.25 That Aronfreed's construct of anxiety is no more elusive than similar categories in learning theory hardly makes it problem free, but let us accept it for now and turn our attention from the affective genus to its cognitive differentiae of fear, guilt, and shame. The first of these, fear, is determined by the anticipation of the act in question. These consequences are typically produced by external agencies such as a reproving parent, but the whole point of our review of theories of self-regulation was that they can also be produced, really or in anticipa- principle forbidding adultery can be an efficacious moral value. Since both authors are fully aware that affective associations can be acquired vicariously, we are not being presented with the ludicrous caricature of moral education as commencing with overt transgressions: as though, say, sexual values were best taught by starting with a few robustly distressful acts of adultery designed to set the affective stage required for the virtue of chastity. 24. And more obviously, in Aronfreed, "Moral Development." 25. Just as a reinforcer is anything which increases the probability of a response, so anxiety seems to be any psychological state which is avoided by the behaving organism: paradigmatically for Aronfreed, it is the generalized aversive state induced in a child by punishment. Thus he writes that, although anxiety sometimes interferes with the performance of an act, "the more important function of anxiety is to provide the motivation, and indirectly the reinforcement, for nonpunished behavioral alternatives to the punished act" (Conduct and Conscience, p. 55). 422 Ethics April 1982 tion, by the agent for himself, as in a performance contract. But now let us consider Aronfreed's treatment of the other two "cognitive housings" of anxiety: guilt and shame. These arise in turn from assessments of the benevolence or the social visibility of the act in question. Although he allows for overlaps and anomalies,26 Aronfreed clearly believes that in cases of guilt and shame the general distress of anxiety is administered internally, through a set of cognitions which are self-referring and hierarchical in a way that the expectations characterizing fear are not. The feelings of guilt and shame are motivating, but not because the agent seeks to escape their discomfort-that response would be the moral self- indulgence of F. C. Sharp's man "who always rode in the street car with his eyes closed, because he could not bear to see ladies standing when he had a seat."27 Rather, they motivate primarily because they are expres- sions of symptoms of some other, still-unexplained motivational process, one which is itself intensified by the real or anticipated experiences of guilt- or shame-feelings but is not to be identified with them. This view is obviously at odds with Alston's view of Aronfreed. The persuasiveness of Alston's critique is tied to its challenge of the very idea of "affective anticipations" being crucial to the concept of self-regulation, especially moral self-regulation. Aronfreed's general use of that idea-as well as Alston's critique-suggests that Aronfreed's paradigm of self- regulation is the "once burned, twice shy" sort of agent whose cognitive evaluations amount to a prudential cost-benefit analysis of his environment and so, as above, to a paramechanical model of self-regulation. Indeed, up to the chapter on fear, guilt, and shame, Aronfreed's book very strongly supports that suggestion and so is open to Alston's criticism. But in that chapter moral feelings emerge as in some sense capable of having lives of their own, as "functionally autonomous," even though often alloyed with fear or with each other. Why this happens is left unexplained, perhaps inevitably. But if my interpretation of that chapter is correct, then, contrary to Alston's charge, Aronfreed there envisions an agent for whom at least two action-guiding principles are not only hierarchical in the sense of directing other motivations but also truly loaded with affect: those of not harming others and of keeping in good social standing. These action guides can be formulated verbally as the principles of nonmaleficence and affiliation, though they need not be verbalized in order to function quite well as the sources for guilt and shame. Both are formal in that they are cherished by the agent prior to any specification of their contents, exactly in the manner desired by Alston. Furthermore, they

26. For instance, many guiltlike reactions to transgressions could, in certain social contexts, really be fear reactions, and vice versa. Thus reparation, apology, confession, even unspoken self-criticism could be mechanisms for warding off external threats rather than ways to express or resolve one's guilt. Conversely, a certain punishment could seem dreadful not so much because of the physical pain it causes as because it dramatically illustrates the degree of harm one has done to others (or for shame, the extent of the social rejection). 27. F. C. Sharp, Ethics (New York: Century, 1928), p. 76. Wren Social Learning Theory 423 are able to engage behavior evenwhen their objects have not antecedently aroused positive or negative affect, although it may be true (contra Alston) that the agent may not be able to acquire them or other deep-lying values without some such antecedent affective experiences. It seems, then, that Aronfreed has some inklings of a parapolitical conception after all, since the values which provide the criteria for assess- ing first-order motivations are presented by him as having their own, virtually self-contained affectivity. Of course a person may fail to heed those values, but when they are heeded they function as an executive agency, not just as another vector in the force field. Or to change the metaphor: it is as though a filter were put in the field, allowing some forces better access to the agent's behavior and deflecting others, all without directly altering any of their strengths. But this, I fear, is as far as the social learning theory approach takes one in the philosophical and psychological enterprise of accounting for the prescriptive element within moral experience. If moral principles, values, or ideals are to enjoy an executive status they must derive their motivational significance from some other source than the field of desires and inclinations that they regulate. Social learning theorists, including Aronfreed, have allowed the Law of Effect to establish the outer limits of inquiry, with the unfortunate result that they lack a metatheoretical context for explaining how values come to have the special sort of affectivity that Aronfreed seems, in his best moments, to want to impute to them. One need not be a Freudian to undertake this larger task, though, of course, Freud's was one of the most ambitious metatheoretical efforts. The philosopher Rom Harre has recently proposed such a context in terms of social collectives, whose driving and integrating principles are self-expressive themes such as honor and mutual respect.28 My own view is somewhat more psychodynamic than Harre's and yet more socially interactive than Freud's: it is that we must take certain conative structures as givens, not in the sense of their being innate biological structures (though they may be that too) but rather as constitutive components of the agent's present personality, self-concept, and "reason for going on. "29 The constitutive desires, as I would call them, are the proper subject matter of depth psychology, and probably include not only the relatively straightforward affiliative tendencies implied by Aronfreed but also oedipal and even more problematic sorts of strivings. I want to say, then, that the affective fea- tures and motivating potential which Aronfreed so correctly ascribes to "evaluative cognitions"-moral ideals, values, and principles-derive from these deep-seated strivings. In short, moral cognitions are prescriptive self-regulators to the extent that they articulate an agent's profoundly intimate, constitutive desires.

28. Rom Harr&, Social Being (Oxford: Blackwell's, 1980). 29. B. Williams, "Persons, Character, and Morality," in The Identities of Persons, ed. A. Rorty (Berkeley: University of California Press, 1977). 424 Ethics April 1982 To work this or any comparable view into a full-fledged theory of moral (or simply human) motivation would take us far beyond the behaviorist limits imposed by the Law of Effect and the correspondingly spare definitions Aronfreed gives of guilt and shame as forms of anxiety. The Law of Effect has been hailed as a star of the same magnitude as the Principle of Induction, and this estimate may be correct. But once one begins to wonder why certain kinds of reinforcers (e.g., guilt feelings) reinforce, a radically different kind of inquiry is called for, one which would work in the crepusculum between speculative philosophy and empirical psychologies such as social learning theory.