Crime, Violence, and Self-Esteem:
Review and Proposals

Thomas J. Scheff, Suzanne M. Retzinger, and Michael T. Ryan

The role played by self-esteem in precipitating, influencing, or preventing violent behavior is problematic. In this chapter, we will survey studies of the relation of self-esteem to crime and violence, propose a new approach, and make recommendations for public policy. The literature review will include both a survey of studies that have used quantitative methods and a selective review of studies of emotion and self-esteem in violent behavior.

The first half of our review is a broad survey. But we have narrowed the focus of the second half to the role of self-esteem in crimes of violence. We made this decision for two reasons. First, level of self-esteem may be crucial in crimes of violence, particularly violence toward intimates. Violence of this type often appears to involve intense emotion; it is sometimes referred to as a "crime of passion." As we suggest below, the kinds of passions involved in family violence may be related to selfesteem.

Second, in crimes against property and against the public order, the connection with self-esteem is not clear. This is not to say there is no connection; rather, studies simply do not provide a clear picture. Our discussion of quantitative studies criticizes weaknesses in the design and methods of existing research. Perhaps future, more adequate, studies may yet uncover such relationships. But most crimes against property (theft, for example) may be "economic" crimes, not crimes of passion, and therefore not related to self-esteem. Most prisoners in the American

corrections system have been convicted for crimes against property. It is possible that economic, political, and cultural forces, rather than psychological ones, swell our prison populations. These are important issues, but they will not be our focus here.

Quantitative Studies

Explanations of deviant behavior have a long tradition in psychology and sociology. As the influence of interactionist social psychology grew in the 1950s, theorists began including the idea of a self-concept in such explanations (Wells 1978). These theorists assumed that behavior is influenced by thoughts and feelings about self. The self-concept was seen as a product of socialization and interaction with significant others, an approach that could form a bridge between psychology and sociology. More recently, low self-esteem has been proposed as a cause of criminal behavior. Empirical validation, however, has been slow in coming. Examining the relevant studies will help us to assess the current state of this research.

Crime and violence are concepts that encompass diverse forms of specific behavior. The relevant research represents a wide assortment of approaches and concerns, and it also interprets the idea of self-esteem in a number of ways. The most common conception is global self-esteem, attitudes and feelings held toward the whole self, that is, "the degree to which an individual feels that he is personally and socially adequate" (Rosenbaum and deCharms 1962, 291). Other studies, however, concern the behavioral effects of levels of self-esteem that are temporarily induced (by insult or praise) under specific experimental conditions. A few have studied the consequences of interaction between global and induced self-esteem.

The population samples used in these studies also vary widely. Many are limited to a particular age group (e.g., children or adolescents), gender, social class, or other subgroup such as delinquents, making comparisons of the findings difficult.

Various types of behavior relevant to crime and violence have been investigated. Experimental studies have sought to link levels of selfesteem (both global and induced) with aggression, as well as with other forms of antisocial behavior such as cheating or dishonesty. In such studies, researchers first administer a questionnaire designed to measure self-esteem and then apply treatments that are assumed to raise or lower self-esteem. Subjects' responses to one or more experimental conditions

are then measured. Because self-esteem is measured or induced before observing how the subjects respond, these studies are better able to show the causative effects of self-esteem than are correlational studies, which simply show an association between the measured variables.

The rich diversity of experimental research in this area is impressive, and certainly much is owed to the investigators for their ingenuity and persistence. It is therefore especially disheartening that the experimental studies have tended to be inconclusive, often demonstrating effects that are weak, nonexistent, or sometimes contradictory. Although much has been learned, the parts still fail to add up to a recognizable whole. Furthermore, because these studies are conducted in laboratory settings, the extrapolation of results to real situations is uncertain. Such studies may lack what is called "ecological" validity.

Other studies have investigated the relationship of self-esteem to hostility and aggression, as well as to criminal behavior. But most research conducted outside the laboratory has attempted only to demonstrate correlations between levels of self-esteem and behavior, showing the possibility of a relationship without establishing causation. Even with such a limited goal, the correlations reported have been weak at best. Clearly, longitudinal studies are required to ascertain whether level of self-esteem plays a causal role in violent or criminal behavior. Very few such studies have been conducted, however, and these have focused principally on only one field, juvenile delinquency. Nevertheless, because of the care with which this research was done, we review it at some length below.

Many early studies of aggression focused on the "frustration-aggression" hypothesis (Dollard et al. 1939), which views aggression as a response to any stimulus that blocks "goal-directed" activity (i.e., the attempt to gain a valued reward such as money, food, or social approval). This line of research, popular into the 1960s (see, for example, Berkowitz 1962; Buss 1961), suggested that individuals differ in their predispositions to anger and aggression and therefore in their susceptibility to frustration. Such differences came to be seen as the product of the individual's self-concept or self-esteem.

Rosenbaum and deCharms (1962) conceptualized self-esteem as a "mediating response" that determines expectations of reward or punishment, success or failure, and so on. Persons with low self-esteem, having more negative expectations, should therefore be more sensitive to disapproval or rejection, and hence more easily "frustrated" and prone to aggression. At the same time, however, expectations of punish-

ment and failure should be expected to inhibit the overt expression of aggressive impulses by such individuals. A series of experiments showed support for both hypotheses: subjects with low self-esteem appeared to be more angered by verbal attacks than were subjects with high self-esteem, yet the former group also seemed more inhibited in expressing aggression.

Though not conclusive, these results suggest why there have been so many weak or contradictory findings. Depending on which proposed effect of self-esteem is more strongly elicited or on how aggression is measured, self-esteem may be positively or negatively correlated with aggression; or the countervailing influences may simply cancel each other out. Kingsbury (1978), for example, found that subjects with low self-esteem who were insulted by a confederate "punished" the confederate in a bogus learning task with levels of electrical shock that were substantially higher than those administered by similar subjects who had not been insulted. Among subjects with high self-esteem, there was no significant difference between those who were insulted and those who were not, supporting the notion that persons with low self-esteem are more easily angered by insults. But subjects with low self-esteem also administered higher shock levels than subjects with high self-esteem did, both when insulted and when not insulted (though the latter difference was not statistically significant), contrary to the prediction that those with low self-esteem should be more inhibited in expressing such aggression.

In another experiment, however, Worchel (1958) allowed subjects to use a questionnaire to express aggressive feelings toward an instructor who had insulted them. Subjects with high self-esteem exhibited levels of aggression that were substantially greater than those expressed by subjects with low self-esteem, supporting the idea that individuals with low self-esteem are more inhibited. But the contingent conditions for inhibition are neither specified by the theory nor explicitly controlled for in such experiments, making it difficult to generalize their findings. Cohen (1959) found evidence that people with low self-esteem are more vulnerable to the exercise of power over them than are people with high self-esteem, suggesting that at least some of the effects observed among subjects with low self-esteem may be inadvertent results of the experimenter's or other influence, perhaps in anticipation of approval or punishment. These considerations imply that the notion of a single, linear relationship between self-esteem and behavior is overly simplistic.

Rosenbaum and deCharms (1962) also noted that it was not clear

whether the differences they observed between the two sets of subjects could be generalized to task-oriented frustration. Epstein and Taylor found that when subjects competed in a "reaction time" test in which the winner administered a shock to the loser, the subjects used higher shock settings only when the opponent was also clearly aggressive, leading the researchers to conclude that "perception of the opponent's aggressive intent is a far more potent instigator to aggression than frustration in the form of defeat" (1967, 286). Geen (1968) similarly found personal insult to instigate aggression more effectively than did frustration per se. Feshbach, in an influential theoretical statement, concluded that "violations to self-esteem through insult, humiliation, or coercion are . . . probably the most important source of anger and aggressive drive in humans" (1971, 285). Feshbach's forceful statement is significant for the theory that will be proposed later in this chapter. Concurrent with investigations of the frustration-aggression hypothesis were studies of environmental factors that lead to aggressive behavior. McCord, McCord, and Howard (1961) reanalyzed data from the Cambridge-Somerville Youth Study of the 1930s, in which the home lives of 650 nine-year-old boys had been observed over a five-year period. Half the sample, who were judged to be "maladjusted" and potentially delinquent, formed the control group. Subjects in the other, "normal" half were observed to fall into three categories: those who were consistently "aggressive," those who were merely "assertive" (occasionally aggressive), and those who were consistently "nonaggressive." Researchers found that the aggressive boys had experienced significantly higher levels of direct parental attacks (physical abuse, verbal threats of punishment or abandonment, derogation, and rejection) than the other boys had faced. As the authors observed:"These attitudes and actions are all, presumably, direct attacks on the child's sense of security; they all tend to undermine the boy's conception of himself as a person of worth and significance; and they all carry the implication that the world is a dangerous, hostile environment" (1961, 84). Additionally, the aggressive boys were more likely to have experienced parental conflict in the home, as well as deficient (weak and inconsistent) parental controls, which, at least in current clinical opinion, might also be seen as contributing to low self-esteem. The authors noted that the aggressive boys displayed a greater tendency to reject their parents and the parents' values, more often choosing a "reference group" outside the home, such as a gang of delinquents or other group of neighborhood boys. (The relationship of this point to self-esteem will

be seen more clearly in our discussion of Kaplan's theory of deviance, below.) Parents who typically responded with aggression to crises in and outside the home, however, did not seem to be more common among the aggressive group of boys, suggesting that the modeling of such behavior was less important than the emotional climate of the familial relationships.

These observations support the earlier findings of Glueck and Glueck (1950), who worked with a somewhat different sample of lower-class boys. They found that many delinquent boys came from homes similarly characterized by severe punitiveness, parental rejection, inconsistency, and parental discord. Delinquents were described as more insecure and anxious and less cooperative and trusting than their nondelinquent counterparts. Bandura and Walters (1959) also reported comparable findings in their extensive study of middle-class delinquents.

Toch (1969) interviewed sixty-nine prison inmates and parolees who had a history of violence. He concluded that there were ten characteristic approaches to interpersonal situations that promoted violence. The most common, "self-image compensating," involved aggression in defense of self-image (as retribution against an individual who may have cast aspersions on one's self-image) or to promote self-image (to demonstrate worth). In both cases, the need to resort to aggression was believed to stem from feelings of extremely low self-esteem.

These studies are correlational only, even though they provide far richer and much more detailed evidence than that available from laboratory studies. They convincingly show an association between the presumed antecedents of low self-esteem and aggressive or delinquent behavior, but they fail to demonstrate that the relationship is a causal one.

The most comprehensive theoretical elaboration of the relationship between self-esteem and criminal behavior—and the most convincing empirical investigation thus far—is based on Kaplan's (1975, 1980) "esteem-enhancement" model of deviance. Kaplan proposes that delinquent behavior serves to enhance self-esteem for individuals who have experienced failure and lowered self-esteem.

Fundamental to this theory is the assumption of a "self-esteem motive"—defined by Kaplan (1975) as "the personal need to maximize the experience of positive self-attitudes and to minimize the experience of negative self-attitudes"—which stems from the child's early dependence on adults. As Kaplan describes it, the need to behave in a way that evokes positive adult behavior sensitizes the child to adult reactions to his or her self. Over time, therefore, the role of others will be adopted as a guide to

appropriate behavior, until a symbolic association is formed between the imagined responses of others and one's own responses to self.

When an individual's past experiences in a particular membership group have led to negative self-attitudes, according to the theory, the individual will associate that membership group with self-derogating experiences, ultimately losing the motivation to conform to normative expectations—and gaining the motivation to deviate from such expectations. Deviance becomes attractive to the extent that it represents opportunities for self-enhancing experiences. Whether deviant behavior is adopted and which type is chosen depend on circumstances—what kinds of deviant activity are visible and available, the perceived attractiveness of these opportunities, and so on. Whether the involvement is continued, says Kaplan, depends in turn on the extent to which the deviant activity is in fact felt to be self-enhancing or self-derogating.

To test this theory, Kaplan conducted a survey of seventh-grade students each year for three years. This age group (eleven to thirteen years old) was a compromise, likely to predict adult behavior, yet minimizing prior involvement in deviant activities. More than three thousand students were surveyed each year. Self-derogation and self-esteem were measured using seven items from Rosenberg's ten-question scale (those items Kaplan judged to measure "affective" components of self-esteem). Deviant behavior was measured by self-reports of twenty-eight "presumably deviant acts." Questions pertaining to other aspects of the theory (such as attitudes toward normative expectations) were also asked.

In the first analysis of the data (Kaplan 1980), several steps in the causal chain specified by the theory were investigated. Most relevant to our purposes here is the analysis of the effects of self-esteem on subsequent delinquency. Kaplan found that among subjects presumed not to have already adopted deviant behaviors, those with initially lower levels of self-esteem (and those with greater antecedent increases in self-rejecting attitudes) were more likely than others to subsequently adopt deviant patterns.

The effects of social class (defined in terms of the mother's education) and gender were also examined, for it was argued that deviance would be more normative for lower-class students than for middle-class students and less normative (i.e., more "deviant") for females than for males. The relationship of self-rejection to subsequent deviant behavior was shown to be stronger and more consistent for middle-class students, weaker and in many cases insignificant for lower-class subjects. In both classes, the association was stronger for females than for males.

Variables that might have intervened between self-esteem and deviance, such as the subjective association of negative self-attitudes with membership group experiences and the development of contra-normative attitudes, were also investigated.

Finally, the proposed increase in self-esteem following the adoption of deviant behavior was examined. Here, a clear relationship was not immediately found. Additional conditions were considered: (1) lack of defenses against or vulnerability to self-derogating experiences in normative settings (reflecting an inability to satisfy the self-esteem motive within normative membership groups); (2) the perceived self-enhancing potential of the normative environment (presumed to indicate the degree to which "deviant" acts were actually defined as such by the subjects); and (3) gender (taken as a rough estimate of the subjects' vulnerability to the self-derogating consequences of deviant behavior patterns, because females were expected to be more dependent than males on approval from—and hence less likely to reap the rewards of deviating from—the normative setting). In each case, the expected results were obtained, to varying degrees, although the authors acknowledge that their measures represented only coarse approximations of the conditions required by their theory.

Because of the piecemeal approach taken by this analysis, it fails to provide a convincing confirmation of the overall theory. Kaplan selectively points out only the significant effects that appear to support the esteem-enhancement model, based at least partially on "predictions" apparently formulated after the data were already in hand. Furthermore, because the twenty-eight different self-reported measures of deviance are analyzed separately, it is difficult to assess the strength of the relationships proposed. Multivariate analyses of other longitudinal data by Bynner, O'Malley, and Bachman (1981), Wells and Rankin (1983), and McCarthy and Hoge (1984) used path analysis techniques to simultaneously estimate the effects of the entire esteem-enhancement model. All three of these studies concluded that the direct effects of self-esteem on deviance were negligible or nonexistent, although two (excepting Wells and Rankin 1983) found support for the self-enhancing effects of delinquent behavior. These three studies assume considerable importance for our review, because they fail to support Kaplan's main findings.

Subsequent analysis of the Kaplan data by Kaplan, Martin, and Johnson (1986) posited three deficiencies in all of the earlier analyses.

First, the measurement of self-rejection or self-esteem had been inadequate, since the theory proposes that it is the combination of global feelings of inadequacy and the individual's association of such feelings with membership group experiences that disposes the person to deviance. Second, measures of deviance had most often been specified as a continuous variable covering specific kinds of behavior, whereas the theory asserts that it is "not the amount of deviance but rather the engagement in any of a range of functionally equivalent deviant behaviors that is predicted" (Kaplan, Martin, and Johnson 1986, 393). Third, and more fundamental, the earlier analyses had erred in focusing only on the direct relationship of self-esteem to deviance, whereas "self-rejection, theoretically, has direct effects on deviant dispositions, not on deviant behavior" (391). Thus, the disposition to deviance needed to be defined as an intervening variable in the model. The variables were respecified, using multiple indices. Strong support was found for the theoretical model. Estimated coefficients for the direct effect of self-rejection at time-1 on disposition to deviance at time-2 (= .76) and disposition to deviance at time-2 on deviance at time-3 (= .52) were both large and significant beyond the .001 level.

An elaboration of the model was provided by Kaplan, Johnson, and Bailey's (1986) addition of the latent construct "early involvement in deviant activities" (as reported at time-1) to the earlier analysis. Early deviance was expected to show a direct negative effect on self-esteem, by violating internalized norms and by evoking negative responses from the individual's social milieu. This effect was then expected to indirectly influence subsequent deviance both positively and negatively. That is, if self-rejection, or low self-esteem, increases the disposition to deviance, the likelihood of deviance should also increase. But low self-esteem was also expected to show a negative effect on deviant behavior in the new model, as a result of feelings of inefficacy and an increased need to conform to normative standards. (Negative self-feelings presume that the individual still desires positive regard from others and has internalized the norms that proscribe deviant behavior.)

Thus, as in Rosenbaum and deCharms's (1962) proposal regarding the countervailing effects of low self-esteem on aggression, self-rejection was hypothesized to motivate the person to further conform to membership group constraints, even as he or she becomes disposed to subsequently deviate from them. In addition, early deviance reported at time-1 was expected to have a direct positive effect on the disposition to

deviance at time-2, as well as a direct positive effect on later deviance at time-3 through maintaining or increasing opportunities to learn deviant behavious from others.

Inclusion of the new variable also permitted a test of whether the previously noted indirect influence of self-rejection on subsequent deviance was an independent causal effect, as predicted by the model, or was dependent on earlier deviance. The estimated coefficients indicated support for the model and were all found to be significant beyond the .001 level of probability:

effects of early deviance at time-1 on self-rejection at time-1 (.568), disposition to deviance at time-2 (.224), and deviance at time-3 (.340);

direct effects of self-rejection at time-1 on disposition to deviance at time-2 (.426) and deviance at time-3(-.180);

and direct effects of disposition to deviance at time-2 on deviance at time-3 (.510).

In the most recent refinement of the model by Kaplan, Johnson, and Bailey (1987), yet another variable was added: "deviant peer association" at time-2, measured by self-reports of deviant behavior engaged in by friends and peers at school. It was predicted that such associations would positively affect subsequent deviant behavior, and this prediction was supported. The direct effect of deviant peer association at time-2 on deviance at time-3 was estimated to be .387.

The added variable changed the effect of disposition to deviance at time-2 on deviance at time-3 from .510 to .197, while the direct effect of disposition to deviance at time-2 on deviant peer association at time-2 was estimated to be a substantial .635. Interestingly, no opposite effect was found—apparently, deviant peer associations do not affect disposition to deviance, as some theorists have asserted (cf. Matsueda 1982). The picture that emerges is one in which the disposition to deviance most often led students to adopt associations with deviant peers; these associations, much more than the disposition to deviance itself, provided the strongest impetus to subsequent deviance.

Kaplan and his associates certainly deserve praise and encouragement for their efforts. Having provided a comprehensive longitudinal study, complete with extensive theoretical underpinnings, they continue to test and refine the model with painstaking diligence. The final verdict on this research may still be years away, however, and it must be pointed out that several unresolved issues remain. Most notably, the posited

self-enhancing effects of deviant behavior, on which the theory is substantially based, have yet to be convincingly demonstrated. Nor has the possible bias from sample attrition been adequately addressed. Only 33.7 percent of the original sample completed all three tests; analyses of the initial questionnaires indicated that the most significant correlate of sample attrition was self-report of early deviance, a variable subsequently included in the structural equation model. Further, because it is impossible to know how the missing students may have differed from the final sample over time, the results cannot be generalized.

Similarly, students' self-reports of deviance may be unreliable. For more than half of the twenty-four items also reported on by school officials, for example, students reported fewer than 50 percent of the activities the officials "knew" had occurred. For three of the items, more than 60 percent of those who reported an act the first year subsequently denied it the next.

What does the esteem-enhancement model tell us about the relationship of self-esteem to crime and violence? In the end, the answer may be as equivocal as the one implied by Rosenbaum and deCharms (1962) more than a quarter century ago: "It depends." As more variables are added to the model, the practical utility of its insights may diminish, leaving us awash in a sea of conditional probabilities.

The implication of the numerous studies of self-esteem may be that the concept as it has been used so far is complex and contradictory, more indicative of propensities to particular kinds of influence and patterns of avoidance than of specific behaviors. Because such issues have not been resolved, quantitative studies have thus far been unable to significantly advance our practical understanding of the relationship between self-esteem and crime and violence.

The problem of defining the concept of self-esteem surfaces in all quantitative studies of the subject, not only those that seek to relate it to crime, violence, and aggression. Morris Rosenberg (1988), a pioneer in the quantitative study of self-esteem, estimates that approximately ten thousand such studies of the concept have been conducted. Each study has used a self-esteem scale, usually consisting of a small number of questions, most often ten to thirty. In the absence of a standard, agreed-upon conceptual definition of self-esteem, these scales have proliferated: there are now at least two hundred in use (Rosenberg 1988). These scales, for the most part, are not comparable. This lack of comparability alone would create chaos, a Tower of Babel, in any field of study.

Given the vast number of quantitative studies, there are surprisingly

few overall assessments of their findings. We located only five general reviews of the entire field and one review of a very large subfield, adolescent self-esteem. Four of these six reviews are quite critical. Diggory does not mince words: he concludes his overview with a commentary on the "utter bankruptcy of it all" (1966, 62). Crandall's assessment is also harsh: "Despite the popularity of self-esteem, no standard or operational definition exists" (1973, 45). He also criticizes the lack of evidence for validity and the theoretical and conceptual confusion. (His review is extremely useful, for he lists at least several items from thirty of the most widely used scales.) In their review of studies of adolescent self-esteem, Savin-Williams and Demo join the critical chorus: "Despite 1,500 articles on adolescent self-esteem . . . we know relatively little of its correlates, determinants, or predictors" (1983, 121). The most recent general review is also the most critical:

What has emerged . . . in the self-esteem literature is a confusion of results that defies interpretation. Hypotheses have been tested about the relationships between self-esteem and hundreds of other psychological variables. Many of these hypotheses have been formally supported, but most observed trends have been weak and insubstantial. There are few replications or systematic extensions, and it is difficult to know which findings are worth pursuing. Moreover, because different investigators begin with different assumptions, their findings stand in obscure relation to one another. (Jackson 1984, 2–3)

The two positive reviews, by Wells and Marwell (1976) and Wylie (1979), are more comprehensive than the negative ones. Both positive reviews report in detail the procedures used and the results found in the self-esteem studies. Both accentuate the positive: the authors sympathize with the systematic nature of the quantitative studies, and they praise the precision and reliability of the procedures. The optimistic and laudatory tone stands in stark contrast to the tone of the four critical reviews.

In one crucial respect, however, these two reviews are in complete agreement with the critics: the absence of significant findings . Wells and Marwell (1976, 72–73) go to great pains to document the contradictions among various studies. They report disagreement among twenty-nine studies on the direction of the relationship between self-esteem and seven variables, such as authoritarianism, admitting unflattering descriptions about oneself, persuasibility, and so on. Because of these contradictions, we are unable to interpret any of these relationships; that is, the findings have been nil. In the broadest and most careful of the re-

views, Wylie (1979, 690, 692) agrees, pointing to the "striking incidence" of findings that are weak, null, or contradictory.

Readers of the last two studies may conclude that these reviews represent the state of the field as healthy. But it is possible to construe them, in concert with the four critical reviews, in a different way: even reviewers who are completely sympathetic to the intentions of the quantitative studies acknowledge that these studies have produced no results . In our opinion, the implication of all six of the general reviews is not that the field is healthy but that it is in a state of crisis, and has been for some time. This interpretation has been underlined since the publication of the two positive reviews (in 1976 and 1979). Several thousand quantitative studies later, there are still no results that are both strong and replicated. How many inconclusive studies will be conducted before basic premises are examined?

It should be emphasized that the series of studies by Kaplan was not referenced by the six general reviews. His findings seem to contradict the conclusion we draw from the reviews, that the relationships reported between self-esteem and deviance have been weak or null. Some of his correlations were substantial, and his work may thus seem to be an exception. As already indicated, however, three subsequent studies by other investigators failed to replicate his findings. For this reason, our critique of the quantitative studies is not overturned by Kaplan's work.

We do not claim that the quantitative studies have been useless. On the contrary, we believe they were necessary. Their very lack of success suggests the need for new directions in theory and method that might be more suited to the problem at hand. The existing paradigm in the field is a general one, treating self-esteem as an attitude that can be studied like any other attitude. Perhaps what is needed is a new paradigm more closely connected with the particular problem of self-esteem. If, as suggested below, self-esteem turns out to be a complex cognitive-emotional structure, then a new paradigm might clear the path toward proving causation.

A Theory of Self-Esteem

As we have indicated, the vast body of quantitative studies does not establish level of self-esteem as a cause of crime and violence. Perhaps much more exploratory research and theoretical work will be necessary before valid quantitative studies can be designed. As a beginning step in

this direction, our focus now shifts to a theory of self-esteem and several studies that seem to support it. Like the quantitative research we have just criticized, these studies do not prove causation. None of them attempts such a task; they are all exploratory in design. We believe, however, that these studies, unlike the quantitative studies, may point toward a new direction in research and public policy.

Our purpose is to build a prima facie case for the importance of self-esteem in the causation of violent crimes. Public policy does not wait for final proof in other realms, and even a case that is merely plausible should provide grounds for debate and discussion of possible new law and policy in the area of corrections. We see no need to be defensive about advocating the importance of self-esteem, because the case that can be made for it is at least as good as the case that has been made for existing laws and policies.

Self-Esteem and Emotion

The theory and research to be discussed here suggest a new way to conceptualize self-esteem in terms of emotions, the basic emotions of pride and shame. The idea of connecting self-esteem to these emotions is not completely new, but it has only recently been made explicit. Our theory emphasizes what already seems to be understood in everyday usage, that self-esteem concerns how we usually feel about ourselves. High self-esteem means that we usually feel justified pride in ourselves, low self-esteem that we often and easily feel ashamed of ourselves or try to avoid feelings of shame. All of the correlates of self-esteem (perceptions, beliefs, and concepts regarding the self; attributions regarding others' perception of self; and behavior toward self and others) may derive from these self-feelings—whether we are usually proud or usually ashamed of ourselves.

In this view, level of self-esteem is a summary concept, representing how well one does overall in managing shame. For persons with high self-esteem, shame is painful but not overwhelming. Such persons have sufficient experiences of pride in their lives that they can usually manage the shame they experience.

Persons with low self-esteem appear to lack sufficient experiences of pride to be able to manage shame; for them, shame is a calamity, to be avoided at all costs. When it cannot be avoided, its effects are often disruptive or even catastrophic. What makes shame so problematic in our society is that, in adults, it is nearly always hidden. We have learned to

be ashamed of being ashamed (Scheff 1988). Both laypeople and researchers often ignore pride and shame as causal agents, because these emotions usually have low visibility and because we have all been trained to ignore them.

In modern societies, it is taken for granted that shame is a rare emotion among adults. This belief is reflected in the anthropological division between shame and guilt cultures: supposedly, traditional societies rely on shame for social and self-control, modern societies on guilt. A similar premise is found in psychoanalytic theory, which places near-total emphasis on guilt as the adult emotion of self-control, with shame deemed "regressive," that is, childish. (For an early critique of both premises, see Piers and Singer 1953.)

There is also an opposing tradition in the scientific literature, however, which maintains that shame is a primary emotion, generated by the constant monitoring of self from the point of view of others. Such monitoring, it is argued, is not rare but incessant, even in solitude. This thread can be found in Darwin (1872), MacDougall (1908), Cooley (1902), Lynd (1958), Goffman (1967), and Lewis (1971). Their arguments can be summarized by three propositions:

1. In adults, social monitoring of self is continuous. We are, as Cooley put it, "living in the minds of others without knowing it."

We do not think much of [self-feeling] so long as it is moderately and regularly gratified. Many people of balanced mind and congenial activity scarcely know that they care what others think of them, and will deny, perhaps with indignation, that such care is an important factor in what they are and do. But this is illusion. If failure or disgrace arrives, if one suddenly finds that the faces of men show coldness or contempt instead of the kindliness and deference that he is used to, he will perceive from the shock, the fear, the sense of being outcast and helpless, that he was living in the minds of others without knowing it, just as we daily walk the solid ground without thinking how it bears us up. (Cooley 1902, 208; emphasis added)

2. Social monitoring always has an evaluative edge and gives rise, therefore, to either pride or shame.

These first two propositions, taken together, create a puzzle. If social monitoring is so incessant, and if it produces either pride or shame, why do we see so little of either emotion?

3. Adults are almost always in a state of pride or shame, but these emotions have such low visibility that we seldom notice them.

This review suggests that although pride and shame may determine

behavior, they usually go unnoticed. Their lack of visibility involves both the stimulus and response sides. On the stimulus side, adults are careful to conceal their manifestations of pride and shame (Goffman 1967, 101–104). On the response side, these emotions are socially unacceptable, to the point that one is not supposed to discuss, express, or even feel them (Scheff 1984). We live in a shame-denying society, in that the denial and disguise of pride and shame are not only matters of individual propensity but also institutionalized patterns of collective behavior. Our very vocabulary denies shame, projecting it instead onto the outer world. We say, "It was an awkward moment for both of us," rather than leveling about emotions by saying, "We were both embarrassed."

Perhaps this institutionalized denial is the reason that self-esteem has never been adequately defined; it touches, after all, on the collective secret of pride and shame. For purposes of discussion, we will here define level of self-esteem as each person's ratio of pride to shame states . This definition is rudimentary, but it may help to stimulate further discussion and research. For example, even though the definition is crude, it could serve as a basis for classifying most of the thousands of items in the existing self-esteem scales. Showing that scale items are facets of either pride or shame, or perhaps are sources or consequences of these emotions, would be a first step toward making the scales comparable.

Because most states of pride and shame are carefully disguised, this definition concerns, for the most part, unacknowledged, low-visibility pride and shame. It parallels Satir's (1972) definition of what she calls "low pot": one is "low pot" when one experiences undesirable feelings but tries to behave as if the feelings weren't there. Her definition of self-esteem does not actually name what we now think is the specific emotion involved in "feeling low," shame. Rather, it follows the contemporary custom of not explicitly naming shame.

Recent developments in theory and method now provide a way of detecting low-visibility shame and demonstrating its role in behavior. The first step was taken by Gottschalk and Gleser, who developed reliable scales for inferring emotion from verbal statements. Although much of their focus is on detecting low-visibility anger and anxiety, they also furnish a scale for low-visibility shame (or "shame-anxiety," as they call it [Gottschalk, Winget, and Gleser 1969, 49–52]). Although compatible with Gottschalk and Gleser's approach, the approach taken by Lewis (1971), a psychoanalyst, is much broader. Her laborious, moment-by-moment analysis of the audiotapes of several hundred psycho-

therapy sessions provides the rationale for inferring low-visibility shame from observable external cues, both verbal and nonverbal (loudness, pitch, rate of speech, intonation, and so on). In her study, she demonstrates that, although neither patient nor therapist seemed aware of it, episodes of low-visibility shame occurred in every session.

Lewis divides unacknowledged shame into two basic types: overt, undifferentiated shame; and bypassed shame. Overt, undifferentiated shame involves painful feelings that are not identified as shame by the person experiencing them. Rather, they are labeled with a wide variety of terms that serve to disguise the experience of shame: having low self-esteem, feeling foolish, stupid, ridiculous, inadequate, defective, incompetent, awkward, exposed, vulnerable, insecure, helpless. Our culture provides a great many such codewords. Lewis classifies all the terms listed above as shame markers, because they occurred only in a certain context and only in association with specific types of nonverbal markers . The context always involved a perception of self as negatively evaluated, by either self or other—the basic context for shame. In this context, Lewis always found a change in the patient's manner, characterized by such nonverbal markers as speech disruption (stammers, repetition of words, speech static like "well" or "uhhh," long pauses), lowered or averted gaze, blushing, and, especially noticeable, a sharp drop in the loudness of speech, even to the point of inaudibility.

Both the verbal and nonverbal markers of overt shame can be characterized as forms of "hiding" behavior. The verbal terms hide shame under a disguising label, and the nonverbal forms suggest physical hiding: covering the face with the hands, averting or lowering one's gaze to escape the other's eyes, and using speech disruption or overly quiet speech to hide the content of one's speech and thoughts. At times, but not always, the ideation of the person undergoing overt shame may also involve hiding: "I wanted to disappear"; "I wished that the earth had opened and swallowed me."

Because Lewis's work was based on audiotapes, her markers for shame are limited to verbal and paralinguistic cues. In contrast, our study of the "moment of truth" in "Candid Camera" television shows was based on videotapes, and we found that very flagrant gestural markers of overt shame sometimes accompany verbal and paralinguistic cues or may occur alone (Scheff 1985; Scheff and Retzinger n.d.). When subjects learned they had been caught on camera in what they thought was a private moment, some of them showed extreme hiding behavior: they not only hid their face with both hands, but some also simultane-

ously turned away from the camera, and others even attempted to hide completely. (One man crawled beneath a desk.) We also observed gestures that we interpreted as hiding behavior, although they appeared in what Tomkins (1963) characterized as a "miniaturized" form. For example, at the moment of truth, many subjects began to bring one or both hands up to the face as if to cover it. Instead of covering it, however, they ended up only touching it.

Because the concept of hiding brings together many different types of ideational, verbal, and nonverbal markers, it is a significant aspect of overt shame. It is of considerable interest, therefore, that one of the markers, face touching, has been independently validated by a study with a very different methodology. Edelman (1989) surveyed five different European countries, focusing on the experience of embarrassment. Although the subjects also associated various other gestures with that emotion, face touching was mentioned in all five countries.

To summarize: overt, undifferentiated shame is marked by (1) a context in which self is perceived as negatively evaluated, by either self or other; (2) "hiding" behavior; and/or (3) the use of undifferentiated terms such as those listed above. In these instances, the negative evaluation of self appears to cause so much pain that it interferes with the fluent production of thought or speech, even though the pain is mislabeled.

Like the overt pattern, bypassed shame occurs in a context in which self is perceived as negatively evaluated. But unlike the markers of undifferentiated shame, which are often flagrant and overt, those of bypassed shame may be subtle and covert. Although thought and speech are not obviously disrupted, they take on a speeded-up but repetitive quality that Lewis refers to as "obsessive." Typically, her patients repeated a story or series of stories, talking rapidly and fluently, but not quite to the point. They appeared unable to make decisions, because of seemingly balanced pros and cons ("insoluble dilemmas"). They complained of endless internal replaying of a scene in which they felt criticized or in error. Often they reported that when they first realized the error, they winced or groaned and then immediately began to obsessively focus on the incident. In such situations, the mind seems to be so caught up with the unresolved scene that one feels unable to become directly involved in events in the present, even though there is no obvious disruption. One is somewhat distracted.

The two patterns of shame appear to involve opposite styles of response. In overt shame, the victim feels so much emotional pain that it obviously retards or disrupts thought and speech. In bypassed shame,

the victim avoids the pain before it can be completely experienced, through hyperactive and rapid thought, speech, or actions.

Adler's (1956) theory of human development anticipated Lewis's discovery of the two basic types of unacknowledged shame. Although he did not use the term, Adler described a "feeling of inferiority," that is, shame, that played a central role in his theory. He argued that children's primary need is for love (for what Bowlby [1969] called a "secure attachment"). If love is not available at crucial points, the child can proceed along one of two paths. One path is to develop an "inferiority complex"—to become prone to overt, undifferentiated shame. The other is to compensate by seeking power—to avoid feeling shame by bypassing it, through what we have termed hyperactive and rapid thought, speech, or actions.

Although the two formulations are compatible, Lewis's work marks a significant advance over Adler's. His theory, true to the psychoanalytic genre, uses concepts that are static and highly abstract. For this reason, his propositions, though provocative, are virtually untestable. No one has envisioned the observable markers of feelings of inferiority, the inferiority complex, or the compensatory drive for power. Nor has the process implied by this theory been spelled out in sufficient detail to allow observations to evaluate its accuracy. Through what concrete steps does deprivation of love cause either an inferiority complex or the drive for power? Adler's theory is couched in concepts that are only "black boxes"; the wiring within and between these boxes is not specified. His "theory" is not testable as it stands.

In contrast, Lewis's formulations provide the foundation for a testable theory, because they describe or at least imply observable markers for the major concepts involved and for the events in the connecting causal chain. She also specifies the behavioral manifestations of Adler's structures, allowing their presence or absence to be detected in actual episodes of social interaction. To our knowledge, hers is the first general theory of human behavior with these desirable characteristics.

Although oven shame and bypassed shame appear to be very different in terms of behavior, the difference is one of outer style; the different appearances mask an underlying similarity. Both the slowed-down pattern of overt shame and the speeded-up pattern of bypassed shame are disruptive; both involve rigid and distorted reactions to reality. Both kinds of shame are equally invisible—one is misnamed, the other ignored. These two basic patterns explain how shame might be ubiquitous, yet usually escape notice.

A series of studies by Scheff (1986, 1987, 1989) has shown how Lewis's theory and method can be applied to audiotaped discourse to demonstrate the cycle of insult, humiliation, revenge, counterrevenge, and chronic resentment that characterizes interminable quarrels and silent impasses. Retzinger (1988) reports similar results with videotapes of marital quarrels, using standard methods for rating visual, as well as verbal and paralinguistic, cues. The analysis of audio and videotapes may provide a new path toward establishing a causal relationship involving self-esteem, emotion, and behavior.

The concept of low-visibility shame may give us a key to what has been a very important unsolved problem: the origins of destructive anger. The cause of angry violence, of anger that goes out of control, has always been something of a mystery. We know that anger alone does not usually result in violence, because most anger is of brief duration and relatively low intensity. What, then, is the additional ingredient that results in violence?

The additional ingredient, according to several theorists, is what Lewis (1971) calls unacknowledged shame . She argues that the combination of unacknowledged shame and anger causes a feeling trap, an alternation between shame and anger, taking the form of a chain reaction that can lead to explosive violence. She calls the result humiliated fury —feeling ashamed that one is angry, angry that one is ashamed, and so on, in a sequence that will not subside.

The psychoanalyst Heinz Kohut (1977) uses a somewhat different term, "narcissistic rage," for what seems to be the same state; he identifies it as a compound of shame and rage. A similar concept, which he called "impotent rage," can be found in Nietzsche. Interpreting the feeling of impotence as a variant or cognate for unacknowledged shame, Gottschalk, Winget, and Gleser (1969), Lewis (1971), and Wurmser (1981) all suggest that violence results from a mixture of shame and anger. Similar suggestions have already been mentioned above, particularly in the emphasis that Geen (1968) and Feshbach (1971) place on the role of insult and humiliation in causing aggression.

Lewis's discovery of unacknowledged shame and the affinity between shame and anger provides the basis for an explicit theory of the role of emotion in aggression and violence (Scheff 1987, 1988). According to this theory, pride and shame states almost always depend on the level of deference accorded a person: pride arises from deferential treatment by others ("respect"), and shame from lack of deference ("disrespect"). Gestures that imply respect or disrespect, together with the emotional

response they generate, make up the deference/emotion system, which exerts a powerful influence on human behavior. Because both the level of deference being accorded and the person's emotional reactions are usually not acknowledged, this system has seldom been noticed, much less carefully studied. We will develop this analysis further in our discussion of the Attica prison riot, below.

The theory outlined here is supported by a series of exploratory studies, each study conducted independently, and independent of Lewis's work and the theory developed from it. Katz (1988) analyzed descriptions of several hundred criminal acts: vandalism, theft, robbery, and murder. In many of the cases, Katz found that the perpetrator felt humiliated and had committed the crime as an act of revenge. In some cases, the sense of humiliation was based on actual insults: "[A] typical technique [leading to murder] is for the victim to attack the spouse's deviations from the culturally approved sex role. . . . For example, a wife may accuse her husband of being a poor breadwinner or an incompetent lover . . . or the husband may accuse his wife of being 'bitchy,' 'frigid,' or promiscuous" (1988, 16).

In other cases, it was difficult to assess the degree to which the humiliations were real or imagined. Whatever the realities, Katz's findings support the model of the shame/rage feeling trap. In his analysis of the murder of intimates, he says: "The would-be killer must undergo a particular emotional process. He must transform what he initially senses as an eternally humiliating situation into a blinding rage" (1988, 19). Rather than acknowledging his or her shame, the killer masks it with anger, which is the first step into the abyss of shame/rage, ending in murder. Katz reports similar (though less dramatic) findings with respect to the other kinds of crimes he investigated. Because shame may be the key ingredient in a low level of self-esteem, Katz's study seems to implicate low self-esteem in the causation of criminal acts.

One issue not addressed by Katz's study concerns the conditions under which humiliation is transformed into rage. Because not all humiliation leads to blind rage, there must be some element not indicated in Katz's cases. Studies of family violence by Lansky strongly suggest what this extra element might be. In order to lead to blind rage, the shame component in the emotions that are aroused must be unacknowledged .

Lansky has published several studies of family violence. The first (1984) describes six cases, the second (1987) describes four. Another recent study (Lansky n.d.) analyzes a session with a single couple. In all eleven cases, Lansky reports similar emotional dynamics: violence re-

sulted from the disrespectful and insulting manner that husbands and wives adopted toward each other. Although some insults were overt (cursing, open contempt and disgust), most were more covert (innuendo or double messages). Underhanded disrespect seemed to give rise to unacknowledged shame, which led in turn to anger and violence, as predicted by Lewis. It was difficult for participants to respond to innuendo and double messages; these forms of communication seemed to confuse them. Instead of admitting their upset and puzzlement, they answered in kind. The cycle involves disrespect, humiliation, revenge, counter-revenge, and so on, ending in violence.

Both spouses often seemed unaware of the intense shame their behavior generated, as Lansky described in one of the cases:

A thirty-two-year-old man and his forty-six-year-old wife were seen in emergency conjoint consultation after he struck her. Both spouses were horrified, and the husband agreed that hospitalization might be the best way to start the lengthy treatment that he wanted. As he attempted to explain his view of his difficult marriage, his wife disorganized him with repeated humiliating comments about his inability to hold a job. These comments came at a time when he was talking about matters other than the job. When he did talk about work, she interrupted to say how immature he was compared to her previous husbands, then how strong and manly he was. The combination of building up and undercutting his sense of manliness was brought into focus. As the therapist commented on the process, the husband became more and more calm. . . . After the fourth session, he left his marriage and the hospital for another state and phoned the therapist for an appropriate referral for individual therapy. On follow-up some months later, he had followed through with treatment. (Lansky 1984, 34–35; emphasis added)

The wife's humiliation of the husband in this case was not disguised through innuendo; rather, her disparagement was overt. Her shaming tactics were disguised by her technique of first building him up by stating how "strong and manly" he was, then cutting him down. Perhaps she managed to confuse herself with this tactic as much as she confused him.

Lack of awareness of shame can be seen in Lansky's (n.d.) report of a conjoint session with a violent man and his wife. In this case, the wife dressed in a sexually provocative way, and her bearing and manner toward the interviewer were overtly seductive. Yet neither spouse acknowledged her activity, even when the interviewer asked them whether the wife was ever seductive toward other men. Although both answered affirmatively, their answers concerned only past events; there was an astonishing lack of comment on what was occurring at that very moment

in the interview. It would seem that blind rage requires not only shaming and shame but also blindness toward these two elements.

Collective Violence

The studies by Katz and Lansky are only exploratory, but they may be used in reinterpreting earlier studies of collective violence, such as those concerned with the 1971 prison riot and killings at Attica State Prison in New York. It is not just the fact that the guards in the Attica prison were humiliated that may explain the riot; rather, their humiliation was unacknowledged and, for this reason, transformed into blind rage.

This general point is especially clear in Lansky's work. He argues that, before clinical intervention, unacknowledged shaming and shame lead to humiliation, rage, and violence. Then he shows that after an intervention in which the therapist comments on and helps the individuals acknowledge the shame process, the level of violence sharply drops. In the riots to be discussed here, this second step, the acknowledgment of shame, was of course absent. The details of the shame and anger process at the interpersonal level may help us understand causation at the level of organizations such as prisons and mental hospitals.

The studies by Katz and Lansky may also have implications for social policy. Lansky's work in particular—both the clinical studies already mentioned and his organizational studies, to be reviewed later—would seem to have policy implications. We will return to this issue at the end of this chapter.

Our analysis of the Attica riot implicates unacknowledged shame, and therefore level of self-esteem, as a causal agent in large-scale, collective violence. The extensive report by the McKay Commission (1972) and a study by Stotland suggest that the level of self-esteem of both prisoners and guards is a key issue in corrections policy. As Stotland describes it: "For both troopers and guards, sense of competence, violence, and self-esteem . . . are linked" (1976, 88). He further notes that "a person's self-esteem can be threatened by failure [and] insults" (86). Because we believe that these studies have both direct and indirect implications, we will devote considerable attention to them.

The analysis of self-esteem and aggression parallels more recent work on shame and humiliation leading to aggression (Katz 1988; Lansky 1987; Retzinger 1987, 1989). These studies suggest a powerful affinity between shame and rage. When a person is shamed, anger is quick to follow if shame is not acknowledged (Horowitz 1981; Katz 1988; Lan-

sky 1987; Lewis 1981; Retzinger 1987; Scheff 1987). These studies suggest two conditions in which unacknowledged shame leads to rage, one concerning personality, the other concerning situational elements . Persons who are prone to shame may respond quickly and violently to their perceptions of slights against the self. But even persons who are not especially prone to shame may also respond violently to insult and humiliation in certain situations—for example, if they perceive themselves as isolated from others, with no one to turn to in their anguish. The status of the social bonds the shamed person or group shares with others (Bowlby 1969) may be crucial. Strong bonds can facilitate the management of shame; weak or severed bonds can impair it.

Stotland's analysis of the role of self-esteem in riots can be expanded by reference to the theory and research reviewed here. Stotland describes frequent episodes of shame and humiliation, but he does not give these episodes the crucial causal role that we do. As suggested in our quotations from his articles, he notes shame and humiliation only in passing.

In our theory, rage is used as a defense against a threat to self, that is, feeling shame, a feeling of vulnerability of the whole self. Anger can be a protective measure to guard against shame, which is experienced as an attack on self. As humiliation increases, rage and hostility increase proportionally to defend against loss of self-esteem; there is a decrease in the ratio of pride to shame states.

Another case that can be reinterpreted in this way is the 1970 shooting of demonstrators by National Guardsmen at Kent State University in Ohio. In their interpretation of the violence at Kent State, Stotland and Martinez conclude:

The events . . . leading up to the killings were a series of inept, ineffectual, almost humiliating moves by the Guardsmen against the "enemy." . . . The answer to these threats to their self-esteem, to their sense of competence, was violence. . . . Another aspect . . . which added to the threat to the self-esteem of the Guardsmen [was that] during their presence on . . . campus . . . the students insulted Guardsmen . . . [and the Guardsmen] were not in a position to answer back. Their relative silence was another humiliation for them. (1976, 12; emphasis added)

In these cases, hostility can be viewed as an attempt to ward off feelings of humiliation (shame) generated by inept, ineffectual moves, a sense of incompetence, insults, and a lack of power to defend against insults.

We will also use the events of the Attica prison riot to demonstrate the connection between self-esteem, shame, and violence. "With the ex-

ception of Indian massacres in the late nineteenth century, the state police assault which ended the four-day prison uprising was the bloodiest one-day encounter between Americans since the Civil War" (McKay Commission 1972, xi). Our analysis is a further attempt to explain the brutal events at Attica, where prisoners rebelled, seized hostages, gained control of a large part of the prison, and held it for several days. Finally, state troopers and guards attacked, killing both hostages and inmates. Even after the inmates' power had been destroyed, violence by the guards and troopers continued. The McKay Commission reported numerous instances of "brutality and humiliation of the inmates"; prisoners were killed, wounded, and humiliated (forced to crawl naked through mud, for example).

Our reanalysis concerns the underlying emotional dynamics behind self-esteem, as it exists in the context within and between prisoners, guards, and prison administration. The continuing violence against the inmates after they were no longer a threat points to anger and hostility as a defense against shame.

The conditions at Attica before the crisis made it easy to understand the need for reform:

For inmates, "correction" meant daily degradation and humiliation: being locked in a cell fourteen to sixteen hours a day; working for wages that averaged thirty cents a day in jobs with little or no vocational value; having to abide by hundreds of petty rules for which they could see no justification. . . . . All their activities were regulated, standardized, and monitored for them by prison authorities . . . their incoming and outgoing mail was read, their radio programs were screened, their reading material was restricted, their movements outside their cells were regulated, they were told when to turn lights out and when to wake up. (McKay Commission 1972, 3)

The degrading conditions the inmates faced mark "overregulation" by the institution (Lansky 1983), a breeding ground for shame and humiliation.

The crisis at Attica began with the appointment of a new commissioner, Russell G. Oswald, an "enlightened and progressive correctional administrator," who saw the abject conditions and the necessity for improvements. Oswald's philosophy was that "an atmosphere of community life-style, even though in a confined situation, [held] greater promise for successful rehabilitation" (McKay Commission 1972, 130–131). He granted the inmates new rights, such as mail and visiting privileges, and revised censorship procedures to allow privacy with attorneys and public officials, greater access to media, and increased public knowledge

of prison conditions. He also ordered that Muslim prisoners be given porkless meals and ruled that inmates were entitled to defend themselves when charged with infractions. These new policies granted civil rights to the inmates, removed some of the humiliating conditions, and gave the prisoners a greater sense of self-respect.

Although these were enlightened and necessary changes, the new policies were not well received by the guards. The commissioner's policies led to a decline in the power and status of the guards relative to the inmates.

Africa's all-white correctional staff from rural western New York State was comfortable with inmates who "knew their place," but unprepared and untrained to deal with the new inmate. . . . Unused to seeing their authority challenged, officers felt threatened by the new inmate. Viewing the recent relaxation of rules and discipline, the interventions of the courts, and the new program for the inmates, they felt that their authority was being undermined by Albany and that their superiors were not backing them up . The officers became increasingly resentful and insecure . The result was . . . daily confrontations between the new inmate and the old-style officer. (McKay Commission 1972, 107; emphasis added)

The courts were also responsive to the inmates' complaints; for example, punitive damages were assessed against the warden on May 14, 1970, for censorship of mail to lawyers and public officials.

The reaction of the guards was intense. Guards felt that the courts were "interfering in matters in which they had no competence. Judges, they felt, knew nothing about either prisons or prisoners. . . . The decisions were wrong, guards felt, and they —not the judges —would have to live with the consequences " (McKay Commission 1972, 125; emphasis added). The new rules created a sense of helplessness for the guards, who had no voice in decisions that directly affected them. In this way, they were as powerless as the inmates had been before the new policies came into effect. The administration was being responsive to the inmates' needs, but not to the needs of the guards, because the conditions of the inmates were more visible than the problems of the guards.

Helplessness and feelings of incompetence are intrinsically connected with shame states, which were not acknowledged by the guards or administration, but which created a response of increased hostility toward the inmates. The McKay Commission reports: "Corrections officers frequently found themselves demanding adherence to rules which inmates would not accept. As the number of confrontations increased . . . so did

their intensity. An officer's orders to stop talking . . . were first questioned, later ignored, and finally ridiculed" (1972, 120).

While feelings of competence and control increased among the inmates, feelings of incompetence and shame increased among the guards. Greater attempts to control the inmates, who responded with less tolerance for disrespect, led to increased hostility and lack of respect on the part of the guards. The guards began concentrating on inmates who were considered "troublemakers," searching for behavior that would justify discipline, harassing inmates who advocated prison reform, searching cells, and locking up men for investigation.

Under new contracts, more experienced guards began taking posts that involved the least contact with inmates. The less experienced guards were closer in age to the inmates, which caused resentment. The inmates faced not only inexperienced officers but also new ones. The new guards did not have "supervisory training or experience; they were totally unprepared for the jobs. . . The inmates could never learn what was expected of them from one day to the next," and the officers had difficulty ascertaining what caused the inmates' uncooperative behavior, leaving both sides with few incentives to "establish any rapport or respect " (McKay Commission 1972, 126, 127; emphasis added). There was no opportunity or desire to develop mutual understanding. The lack of social bonds, as indicated earlier, breeds shame: separation and shame are intricately connected. Thus the tensions and conflict between inmates and guards steadily increased.

Hostile behavior and lack of respect by both sides grew more serious. When a guard attempted to regulate a prisoner, the inmate punched him lightly in the chest. This defiance was supported by other inmates, and the guard, powerless to do anything, backed away. The inmate was placed in special custody the next day, a punishment that other inmates perceived as excessive. Violence by the guards increased, in response to the humiliating situation of the previous day. With no way to verify the facts, a rumor spread among the inmates that the prisoner in special custody had been brutally beaten.

In response to the rumor, inmates attacked guards and began to take over the prison. The guards responded with the sense that they were helpless to do anything commensurate with the seriousness of the outbreak; nor did they immediately inform the administration. The takeover quickly escalated: keys were taken from guards, inmates found and improvised weapons. The takeover involved five cell blocks, shops, the

auditorium, the school, the commissary, and forty-two officers and civilian hostages. Even the warden admitted to being helpless. These developments were spontaneous; the McKay Commission found no evidence that the uprising was planned.

The staff expected the prison to be retaken immediately. Instead, negotiations developed, causing further resentment among the staff, for the tradition in correctional institutions is to refuse to negotiate with inmates holding hostages. Guards arrived with their own weapons, but they were restrained by their superiors, causing more bitterness. The guards were angered not only by the increase in inmate power and the mistreatment of the hostages but also by the recognition given to the prisoners, "as if they were equals." The guards felt that they were not being given recognition by the administration or the media. They watched as the inmates denounced them on television. No one stood up for the guards; they were ignored.

Although the governor ordered that corrections officers should be excluded from the assault force because of their emotional involvement, the order never reached the state police or correctional supervisors. The result was a bloody massacre.

In the aftermath of the assault, hundreds of inmates were stripped and brutalized by corrections officers, troopers, and sheriffs' deputies. The suffering of wounded inmates was prolonged by failure to make adequate arrangements for medical attention.

Stotland's analysis, that the "guards' and troopers' self-esteem must have suffered by the rise in status and power of the inmates" (1976, 93), can be interpreted further. The guards were shamed by their powerlessness relative to the inmates and the administration. Feeling that they had no one to turn to and being humiliated by both sides, the guards engaged in angry attempts at control that finally ended in brutal violence.

An analysis of the deference/emotion system (Scheff 1988) may be used to explain the sequence of events. The administration, guards, and inmates were involved in a complex emotional tangle, with the guards caught between the other two factions. The guards were shamed the most—both by the administration, who failed to consult them before making changes, and by the inmates, who challenged their authority. Shown a lack of respect by both sides, the guards were in a position to enter the shame/rage spiral, which led to further disrespectful acts toward the prisoners; this in turn trapped them in their own spiral, and so on, in a repetitive loop of insult, humiliation, revenge, back and forth between guards and prisoners.

Although the commissioner's new policies had been more humane,

giving inmates more rights, the guards had never been consulted, and they had not been trained to deal with the new level of inmate rights. Thus they reacted with old disciplinary actions. Under the old rules, the inmates' behavior could be viewed as disrespectful. The tension between guards and inmates increased in part because of the perceived lack of respect shown to the guards by the inmates and the resulting humiliation felt by the guards. The prisoners in turn perceived the guards' behavior as lack of respect; the prisoners knew they had been given new rights and felt that they did not deserve to be treated in the old style. To reciprocate this lack of respect, the inmates humiliated the guards further by questioning, ignoring, and ridiculing them.

The situation can be interpreted as a triple shame/rage spiral (Scheff 1987). The guards were shamed by the behavior of the administration and the inmates, were powerless to confront the administration, and became hostile toward the inmates, who in turn were shamed by the guards' lack of respect and reacted with an angry lack of respect toward the guards. The theory predicts no limit to the intensity of the emotions generated in this chain reaction, a prediction amply borne out in this case.

To summarize our interpretation: First, the initial level of self-esteem of the prisoners was unacceptably low; prison policy and the informal system used by the guards combined to constantly humiliate the inmates. Second, reforms were introduced in a way that humiliated the guards, because they were not allowed to participate in the changes. The riot itself appears to have been a result of changes in the level of self-esteem of the guards. Perhaps because these changes were not acknowledged, the humiliation they caused led to blind rage and violence.

Short-Range and Long-Range Policy Implications

Our analysis may have implications for social policy. The level of self-esteem of those involved with social agencies, both staff and clients, may be crucial not only in explosive crises but also in day-to-day operation. If we construe self-esteem to involve the management of shame, then a host of issues that are usually unnoticed become important.

This review suggests that both formal and informal policies should consider the material well-being of staff and clients and the level of respect that is being accorded to them. Because issues of pride and shame typically go unacknowledged, reorienting policy along these new lines is not a simple matter.

At the policy level, it may be possible to make changes that would

upgrade the level of self-esteem of both staff and clients. Lansky (1983, 1984) has proposed that mental hospital units need to avoid what he calls overregulation of patients. He suggests that overregulation humiliates clients and interferes with their rehabilitation. His experience as an administrator shows that policy changes can be made without alienating staff, by continuing education and by including staff in the planning of change.

Lansky's unit has established what seems to be something of a record for successful treatment. For example, in more than fourteen years of dealing with openly suicidal patients, none of the approximately 350 inpatients or recent discharges originally admitted in acute suicidal crisis has actually committed suicide or even made a serious attempt. Although we cannot provide records of comparable units, it appears to us that the effectiveness of Lansky's unit may be unparalleled in this respect.

In addition to training the staff to be sensitive to shame and shaming, Lansky's unit calls in families to participate in the treatment program. Family participation in treatment of suicidal patients appears to be extremely important. By commenting on and interrupting a family shaming process and by bringing the family into treatment, the unit can successfully take what seem to be great risks: "We often allow such patients to go on weekend passes after management sessions with family, or go to jobs, even while they are suicidal" (Lansky 1983, 108). The policy of avoiding overregulation, as carried out in Lansky's unit, seems to be effective.

At the level of state policy, Lansky's unit could be used as a model to train staff at other mental hospitals. Lansky has apparently been successful at helping staff understand the patients' intense need for respectful treatment, avoiding overregulation of patients' activities, and training staff to notice and comment on the shame process. Perhaps lectures and videotapes concerning Lansky's techniques could be used to construct a training program.

It may also be possible to build on Lansky's methods so that they would be applicable to any kind of social agency. Self-esteem workshops could be made available for prisons, mental hospitals, and welfare agencies, sensitizing staff to the relationship between pride and shame, respectful treatment, and self-esteem.

For the sake of brevity, we will give only one example of the kind of training we have in mind. In order to maintain and build client selfesteem, staff must be acutely aware of the fact that signals of respect and

disrespect are carried less by what is said than by how it is said. That is, respect is signaled much more by our manner than by the content of what we say. In routine contacts, we often forget that our facial expression, bearing, and intonation can be disrespectful, even though our words are proper. Clients, who are already in a disadvantaged position, are likely to be extremely sensitive to an insulting or demeaning manner, no matter how subtle.

Even with only brief training, staff members seem to learn that becoming aware of their own manner is advantageous both to them and to the clients. For example, staff who are unaware of the difference between content and manner may tend to be somewhat embarrassed to talk openly about the gritty details of a client's situation. But avoiding the details works to the disadvantage of the client. First, it interferes with resolving the client's problems. Second, the lack of acknowledgment or resolution of staff-client embarrassment helps to perpetuate the client's already low level of self-esteem, as explained by the theory reviewed above.

Becoming aware of the importance of their own manner often has a liberating effect on staff members' ability to deal with clients. Staff may quickly realize that if their manner is respectful, they are free to talk about anything with little risk of embarrassment. This change may be extremely valuable, because it promotes cooperation in solving the client's real problems and also may help in increasing the client's level of self-esteem.

The self-esteem workshops and policy changes suggested here are both short-range interventions that could be introduced into social policy very quickly. Both interventions would involve upgrading the professional standing of staff through policy changes and continuing education. Both changes could be instituted with very little cost, requiring only relatively small changes in the relevant laws and organizational policy, as well as brief training programs.

The idea that the self-esteem of both staff and clients is a key ingredient in the effectiveness of social programs also suggests the possibility of long-range changes in policy. For example, the upgrading of the status of social agency staff through policy changes and continuing education is limited by certain features of the existing framework of social programs. One such feature involves the centralization and isolation of mental hospitals, prisons, and similar social agencies, relative to the communities from which their staff and clients are drawn. The isolation of these units seems to affect not only the self-esteem of the inmates but

also that of the staff. As long as staff and inmates are separated from their communities of origin, those communities can emotionally disown them, treating them as aliens rather than as members, reducing the social status of both staff and inmates. This can certainly serve as a brake on any proposed program to increase their status and self-esteem.

With respect to mental hospitals, a proposal to change the framework has already been made. The Task Force on the Seriously Mentally Ill has submitted a proposal to the state of California that calls for decentralizing the mental health system along the lines we have suggested above: small treatment units that can be located in or near the community from which their staff and clients are drawn. If our analysis of self-esteem, crime, and violence proves to be valid, then the same reasoning should also apply to the corrections system. The day of the huge centralized prison is probably over. Small units located in or near the communities they serve may be the wave of the future.

No doubt the long-range changes in the mental health and corrections system advocated here would be expensive initially. Given the relationships we have discussed between self-esteem and program outcomes, however, it is possible that in the long run the initial investment would be more than repaid by increases in the efficiency and humaneness of the new programs. Perhaps both the short-range and long-range changes might be tried out first on a small scale, with careful studies to evaluate their effectiveness. If they proved effective in the small trial units, this record could encourage their introduction on a large scale.

Bibliography

Adler, A. 1956. The Individual Psychology of Alfred Adler . Edited by H. Ansbacher and R. Ansbacher. New York: Basic Books.

Bandura, A., and R. Walters. 1959. Adolescent Aggression . New York: Ronald Press.

Berkowitz, L. 1962. Aggression: A Social Psychological Analysis . New York: McGraw-Hill.

Bowlby, J. 1969. Attachment and Loss . Vol. 1, Attachment . New York: Basic Books.

Buss, A. H. 1961. The Psychology of Aggression . New York: Wiley.

Bynner, J. M., P. M. O'Malley, and J. G. Bachman. 1981. "Self-Esteem and Delinquency Revisited." Journal of Youth and Adolescence 10(6): 407–444.

Cohen, A. R. 1959. "Some Implications of Self-Esteem for Social Influence." In Personality and Persuasibility, by I. L. Janis et al., 102–120. New Haven, Conn.: Yale University Press.

Cooley, C. H. 1902. Human Nature and the Social Order . New York: Scribner.

Crandall, R. 1973. "The Measurement of Self-Esteem and Related Constructs." In Measures of Social Psychological Attitudes, edited by J. P. Robinson and P. R. Shaver, 45–168. Ann Arbor, Mich.: Institute for Social Research.

Darwin, C. 1872. The Expression of Emotion in Men and Animals . New York: Philosophical Library.

Diggory, J. C. 1966. Self-Evaluation: Concepts and Studies . New York: Wiley.

Dollard, J., L. Doob, N. Miller, O. Mowrer, and R. Sears. 1939. Frustration and Aggression . New Haven, Conn.: Yale University Press.

Edelman, R. 1989. "Self-Reported Expression of Embarrassment in Five European Cultures." Journal of Cross-Cultural Psychology, forthcoming.

Epstein, S., and S. P. Taylor. 1967. "Instigation to Aggression as a Function of Degree of Defeat and Perceived Aggressive Intent of the Opponent." Journal of Personality 35: 265–289.

Feshbach, S. 1971. "The Dynamics and Morality of Violence and Aggression." American Psychologist 26: 281–292.

Geen, R. G. 1968. "Effects of Frustration, Attack, and Prior Training in Aggressiveness upon Aggressive Behavior." Journal of Personality and Social Psychology 9: 316–321.

Glueck, S., and E. Glueck. 1950. Unraveling Juvenile Delinquency . New York: Commonwealth Fund.

Goffman, E. 1967. Interaction Ritual . New York: Anchor.

Gottschalk, L., C. Winget, and G. Gleser. 1969. Manual of Instructions for Using the Gottschalk-Gleser Content Analysis Scales . Berkeley and Los Angeles: University of California Press.

Horowitz, M. 1981. "Self-Righteous Rage and the Attribution of Blame." Archives of General Psychiatry 38: 1233–1238.

Jackson, M. 1984. Self-Esteem and Meaning: A Life-Historical Investigation . Albany, N.Y.: SUNY Press.

Kaplan, H. B. 1975. Self-Attitudes and Deviant Behavior . Pacific Palisades, Calif.: Goodyear.

.———1980. Deviant Behavior in Defense of Self . New York: Academic Press.

Kaplan, H. B., R. J. Johnson, and C. A. Bailey. 1986. "Self-Rejection and the Explanation of Deviance: Refinement and Elaboration of a Latent Structure." Social Psychology Quarterly 49(2): 110–128.

———. 1987. "Deviant Peers and Deviant Behavior: Further Elaboration of a Model." Social Psychology Quarterly 50(3): 277–284.

Kaplan, H. B., S. S. Martin, and R. J. Johnson. 1986. "Self-Rejection and the Explanation of Deviance: Specification of the Structure Among Latent Constructs." American Journal of Sociology 92: 384–411.

Katz, J. 1988. The Seductions of Crime . New York: Basic Books.

Kingsbury, S. J. 1978. "Self-Esteem of the Victim and the Intent of Third-Party Aggression in the Reduction of Hostile Aggression." Motivation and Emotion 2(2): 177–189.

Kohut, H. 1977. "Thoughts on Narcissism and Narcissistic Rage." In The Restoration of the Self, by H. Kohut, 124–160. New York: International Universities Press.

Lansky, M. 1977. "Establishing a Family-Oriented Inpatient Unit." Journal of Operational Psychiatry 8: 66–74.

———. 1983. "The Role of the Family in the Evaluation of Suicidality." International Journal of Family Psychiatry 3: 107–118.

———. 1984. "Violence, Shame, and the Family." International Journal of Family Psychiatry 5: 21–40.

———. 1987. "Shame and Domestic Violence." In The Many Faces of Shame, edited by D. Nathanson. New York: Guilford.

———. n.d. "Murder of a Spouse: A Family System Viewpoint." Typescript.

Lewis, H. 1971. Shame and Guilt in Neurosis . New York: International Universities Press.

———. 1981. Freud and Modern Psychology: The Emotional Basis of Mental Illness . New York: Plenum.

Lynd, H. 1958. Shame and the Search for Identity . Glencoe, Ill.: Free Press.

McCarthy, J. D., and D. R. Hoge. 1984. "The Dynamics of Self-Esteem and Delinquency." American Journal of Sociology 90: 396–410.

McCord, W., J. McCord, and A. Howard. 1961. "Familial Correlates of Aggression in Nondelinquent Male Children." Journal of Abnormal and Social Psychology 62(1): 79–93.

MacDougall, W. 1908. An Introduction to Social Psychology . London: Methuen.

McKay Commission (New York State Special Commission on Attica). 1972. Attica: A Report . New York: Praeger.

Matsueda, R. C. 1982. "Testing Control Theory and Differential Association." American Sociological Review 47: 489–504.

Piers, G., and M. Singer. 1953. Shame and Guilt: A Psychoanalytic and Cultural Study . New York: Norton.

Retzinger, S. 1987. "Resentment and Laughter: Video Studies of the Shame-Rage Spiral." In The Role of Shame in Symptom Formation, edited by H. Lewis, 151–182. Hillsdale, N.J.: Erlbaum.

———. 1988. "Marital Quarrels: The Role of Emotion." Ph. D. dissertation, University of California, Santa Barbara.

———. 1989. "Marital Conflict: The Role of Emotion." International Journal of Family Psychiatry, forthcoming.

Rosenbaum, M. E., and R. deCharms. 1962. "Self-Esteem and Overt Expressions of Aggression." In Decisions, Values, and Groups, vol. 2, edited by N. F. Washburne, 291–303. New York: Pergamon.

Rosenberg, M. 1988. Personal communication to T. Scheff.

Satir, V. 1972. Peoplemaking . Palo Alto, Calif.: Science and Behavior Books.

Savin-Williams, R. C., and D. H. Demo. 1983. "Conceiving or Misconceiving the Self: Issues in Adolescent Self-Esteem." Journal of Early Adolescence 3: 121–140.

Scheff, T. 1984. "The Taboo on Coarse Emotions." Review of Personality and Social Psychology 5: 146–169.

———. 1985. "The Primacy of Affect." American Psychologist 40: 849–850.

———. 1986. "Micro-Linguistics and Social Structure: A Theory of Social Action." Sociological Theory 4: 71–83.

———. 1987. "Interminable Quarrels: Case Study of a Shame-Rage Spiral." In The Role of Shame in Symptom Formation, edited by H. Lewis, 109–149. Hillsdale, N.J.: Erlbaum.

———. 1988. "Shame and Conformity: The Deference/Emotion System." American Sociological Review 53: 395–406.

———. 1989. "Cognitive and Emotional Components in Anorexia: Reanalysis of a Classic Case." Psychiatry, forthcoming.

Scheff, T., and S. Retzinger. n.d. "Hiding Behavior: Toward Resolution of the Controversy over Shame." Typescript.

Stotland, E. 1976. "Self-Esteem and Violence by Guards and Troopers at Atrica." Criminal Justice and Behavior 3: 85–96.

Stotland, E., and J. Martinez. 1976. "Self-Esteem and Mass Violence at Kent State." International Journal of Group Tensions 6: 4–54.

Toch, H. 1969. Violent Men . Chicago: Aldine.

Tomkins, S. 1963. Affect/Imagery/Consciousness . Vol. 2, The Negative Affects . New York: Springer.

Wells, L. E. 1978. "Theories of Deviance and the Self-Concept." Social Psychology Quarterly 41(3): 189–204.

Wells, L. E., and G. Marwell. 1976. Self-Esteem: Its Conceptualization and Measurement . Beverly Hills, Calif.: Sage.

Wells, L. E., and J. H. Rankin. 1983. "Self-Concept as a Mediating Factor in Delinquency." Social Psychology Quarterly 46(1): 11–22.

Worchel, P. 1958. "Personality Factors in the Readiness to Express Aggression." Journal of Clinical Psychology 14: 355–359.

Wurmser, L. 1981. The Mask of Shame . Baltimore: Johns Hopkins University Press.

Wylie, R. C. 1979. The Self-Concept . Vol. 2, Theory and Research on Selected Topics . Rev. ed. Lincoln: University of Nebraska Press.