Deindividuation and Anger-Mediated Interracial Aggression: Unmasking Regressive Racism Ronald W. Rogers and Steven Prentice-Dunn University of Alabama A factorial experiment investigated the effects of deindividuation, anger, and race-of-victim on aggression displayed by groups of whites. Deindividuating situational cues produced an internal state of deindividuation that mediated aggressive behavior. Deindividuation theories were extended by the finding that the internal state of deindividuation was composed not only of the factors SelfAwareness and Altered Experience, but also Group Cohesiveness, Responsibility, and Time Distortion. As predicted, nonangered whites were less aggressive toward black than white victims, but angered whites were more aggressive toward blacks than whites. Interracial behavior was consistent with new, egalitarian norms if anger was not aroused, but regressed to the old, historical pattern of racial discrimination if anger was aroused. This pattern of interracial behavior was interpreted in terms of a new form of racism: regressive racism. Mob violence that has occurred since the time of the Roman republic has been attributed typically to short-term economic motives and political issues (cf. Rude, 1964). Economic and political motives, however, were inadequate to explain the torture, mutilation, and burning that frequently occurred in outbursts of interracial violence. Lynch mobs convinced social scientists that "the fundamental need was for a better understanding of the causes underlying the resort to mob violence" (Southern Commission on the Study of Lynching, 1931, p. 5). The major purpose of the present experiment was to examine interracial aggression within a group context, especially a context conducive to deindividuation. Deindividuation is a process in which antecedent social conditions lessen self-awareness and reduce concern with evaluation by others, thereby weakening restraints against the expression of undesirable behaviors (e.g., Diener, 1977; Zimbardo, 1970). Prentice- Dunn and Rogers (1980) provided the first confirmation of deindividuation theory's major assumption that deindividuating situational cues produce an internal state of deindividuation that mediates the display of aggressive behavior. The deindividuating cues lowered self-awareness and altered cognitive and affective experiences. This deindividuated state weakened restraints against behaving aggressively that are normally maintained by internal and external norms of social propriety. In the present study, therefore, we hypothesized that deindividuating situational cues would produce more aggression than individuating cues, and that an internal state of deindividuation would mediate the effects of deindividuating cues on antisocial behavior. Many problematic forms of interracial conflict occur in group contexts. The major contribution to our understanding of interracial aggression has come from the Donnersteins' research program (cf. Donnerstein & Donnerstein, 1976), which has focused on The authors gratefully acknowledge Kevin O'Brien, situations involving one aggressor and one Henry Mixon, George Smith, and Rod Walls for their victim; no published studies have examined assistance in collecting the data. Requests for reprints should be sent to Ronald W. interracial agression displayed by a group Rogers, Department of Psychology, University of Ala- of whites toward a black individual. The bama, University, Alabama 35486. present experiment examined interracial Journal of Personality and Social Psychology, 1981, Vol. 41, No. 1, 63-73 Copyright 1981 by the American Psychological Association, Inc. 0022-3514/81/410I-0063S00.75 63 64 RONALD W. ROGERS AND STEVEN PRENTICE-DUNN aggression in a group setting in which angry aggressors were deindividuated. This social situation approximates many naturalistic situations. Studies of interracial aggression have consistently shown that the strength of aggression directed toward a different-race victim varies as a function of, for example, potential censure (Donnerstein & Donnerstein, 1973), threatened retaliation (Donnerstein, Donnerstein, Simon, & Ditrichs, 1972), and the victim's expression of suffering (Baron, 1979; Griffin & Rogers, 1977). Donnerstein and Donnerstein (1976) have reported that, in a variety of conditions, white subjects manifest less direct aggression toward black than white victims. Griffin and Rogers (1977) interpreted their white subjects' more lenient treatment of blacks than whites in terms of "reverse discrimination" (cf. Dutton, 1976): To avoid appearing prejudiced, whites treated blacks more favorably (i.e., less aggressively) than they treated whites. Reverse discrimination is the overt manifestation of white people's viewing themselves as egalitarian and feeling threatened by the prospect of appearing prejudiced. Blacks would not be expected to display reverse discrimination, and studies of blacks' aggression have confirmed they do not (Wilson & Rogers, 1975). Both blacks' and whites' behavior, however, can be traced to the same underlying source: Both races seem to be "reacting against the older, traditional patterns for their races" (Griffin & Rogers, 1977, p. 157). For whites, the historical pattern of appropriate behavior toward blacks was racial discrimination and inferior treatment. Although whites may have negative attitudes on several specific issues such as blacks' economic gains (Ross, Vanneman, & Pettigrew, 1976) and race riots (Davis & Fine, 1975), survey data indicate that the new norm is an egalitarian view of the races (Brigham & Wrightsman, in press; Campbell, 1971; Taylor, Sheatsley, & Greeley, 1978). This new norm is especially prevalent among college students. Surveys at the university where the present study was conducted confirmed that the current norm among white students is an unprejudiced, egalitarian view of the races (Rosenberg, Note 1). Theoretically, reverse discrimination is a product of this relatively new egalitarian view of blacks (Dutton, 1976). For blacks, the historical pattern of appropriate interracial behavior was to inhibit aggression toward whites and to displace it to fellow blacks. The new norms favor more militancy, antiwhite attitudes, and overt hostility toward whites (Caplan, 1970; Wilson & Rogers, 1975). The new norms for blacks and whites represent dramatic departures from the deep-rooted values of the past. Both races have been found to act on these new norms if they are not emotionally aroused by a verbal insult. Thus, blacks are more aggressive towards white than black targets (Wilson & Rogers, 1975), and whites are more aggressive toward white than black targets (Griffin & Rogers, 1977). But what happens to behavior based on these new norms if the aggressors are insulted? Baron (1979) reported a three-way interaction effect among race-of-victim, insult, and pain cues. An examination of the conditions comparble to those to be studied in the present experiment (i.e., Baron's nopain-cues condition) indicated that when white subjects were not insulted, black victims received less aggression than white victims (i.e., reverse discrimination); if insulted, the level of aggression expressed toward blacks increased, but did not significantly differ from the level directed toward the white victims. Since we wish to understand interracial aggression in general and not merely whites' behavior toward blacks, let us also examine blacks' aggression toward whites. To interpret the interracial aggressive behavior of blacks, Wilson and Rogers (1975) suggested that emotional arousal produced a regression to a chronologically earlier mode of responding. The data from that experiment were interpreted as evidence that blacks' behavior could be understood as a product of "the conflict between new militant norms and the residue of oppression" (p. 857). Anger-mediated aggression should not be as firmly under the cognitive control of new norms of in-group solidarity and pride. The black students had been exposed to the traditional DEINDIVIDUATION AND INTERRACIAL AGGRESSION 65 values of in-group rejection and out-group however, the effects of anger in a group conpreference for many years before the ap- text have not been investigated. Virtually all pearance of the Black Power movement. studies of deindividuation and aggression They undoubtedly retained some residual have involved unprovoked aggression. Yet, symptoms. Thus, when they became emo- anger adds an important theoretical and aptionally aroused, the new values, which had plied dimension to our understanding of mob not been fully internalized, gave way to the violence. It was hypothesized that insult, or older, more traditional pattern. Similarly, anger arousal, would facilitate the expresthe young white adults in the present study sion of aggression among members of small had been exposed during their socialization groups. to the older tradition of belief in black inOne class of deindividuation theories posferiority. tulates that deindividuated behavior is not The foregoing considerations converge to influenced by usual discriminative stimuli. suggest an interaction between race-of-vic- It may be derived from these theories that tim and insult variables. If whites are not prior insult would have less impact on deangered, we predicted that they would dis- individuated than individuated group memplay reverse discrimination, directing weaker bers. On the other hand, it may be derived attacks against blacks than against whites. from Diener's (1980) theory that "because If angered, we hypothesized that whites self-regulation is minimized or eliminated would regress to the traditional pattern, dis- the deindividuated person is more susceptiplaying more aggression toward blacks than ble to the influences of immediate stimuli, toward whites. emotions [e.g., anger], and motivations" (p. One class of deindividuation theories sug- 211, italics added). gests that victim characteristics (e.g., difThere are several limitations to deriving ferent race) become less salient under con- predictions of interaction effects from these ditions of deindividuation. For Festinger, two classes of deindividuation theories. First, Pepitone, and Newcomb (1952), the defin- neither theory explicitly states how the variing characteristic of deindividuation is that ables of insult or different-race victim would individuals are not paid attention to as in- interact with deindividuation. Thus, other dividuals. As elaborated by Zimbardo interpretations are possible. Second, the (1970), deindividuated behavior is not under form of the interaction effect may vary as the controlling influence of usual discrimi- a function of the strength of the deindividnative stimuli; it is "unresponsive to features uation state. The present study is certainly of the situation, the target, the victim" (p. not an experimentum crucis, but perhaps it 259). Based upon this theoretical position, can shed light on the interaction of deindiany differential treatment of different-race viduation and anger-mediated interracial victims should vanish when group members aggression. become deindividuated. On the other hand, Diener's (1980) theory of less extreme forms Method of deindividuation postulates that crowd members are more responsive to external Design and Subjects stimuli as a result of the focus of attention A 2 X 2 X 2 factorial design was employed with three shifting away from the self. It is plausible between-subjects manipulations: (a) deindividuating versus individuating cues, (b) white versus black to infer that any differential treatment of cues victim, and (c) no insult versus insult. Ninety-six male different-race victims should be enhanced by introductory psychology students participated in the deindividuation. Therefore, the present study experiment to earn extra credit. Twelve subjects were was designed to test these two alternative randomly assigned to each cell. predictions. Verbal attack, or insult, is a potent and Apparatus well-established antecedent of aggression in The shock apparatuses were modified Buss aggression dyadic situations involving one aggressor machines connected to a polygraph. Each of the four and one victim (see review by Baron, 1977); aggression machines had 10 pushbutton switches that 66 RONALD W. ROGERS AND STEVEN PRENTICE-DUNN could be depressed to deliver "shocks" of progressively increasing intensity. Of course, shocks were not actually delivered. A Grason-Stadler noise generator (Model 901 A) was used to produce white noise in the deindividuating cues condition. Procedure The procedure was highly similar to one we had used previously (Prentice-Dunn & Rogers, 1980). Subjects arrived in groups of five; four were naive participants and one was our assistant. The study was explained as a combination of two experiments. The subjects had signed up for an experiment entitled "Behavior Modification" and were to be tested together. Our assistant, ostensibly another introductory psychology student, had volunteered for a study labeled "Biofeedback." After the experimenter determined who had volunteered for each topic, the biofeedback subject was sent to another room to receive detailed instructions for the biofeedback study. After hearing explanations of the concepts of behavior modification and biofeedback, the four white behavior modifiers were told that the response of interest in both studies was heart rate. It was indicated that the biofeedback subject would be attempting to maintain his heart rate at a predesignated, high level. Whenever his heart rate fell below the predetermined level, the behavior modifiers would administer an electric shock. The purpose of having groups of four behavior modifiers was explained as an attempt to establish a laboratory analog of a ward at the local state hospital where behavior modifiers actually worked in small groups. We explained to all subjects that they received their extra credit points for simply showing up and that they could discontinue at any time. Each subject was asked if he had any quesitons about or any objections to the use of electric shock. All questions were answered and no one declined to participate. In addition, written informed consent was obtained. Two mild sample shocks were administered to the behavior modifiers (i.e., the subjects) via finger electrode. The shocks were from Switches 4 (.3 mA) and 6 (.45 mA) on the aggression machine and each lasted for 1 sec. These samples were administered to convince the subjects that the apparatus really worked and to give them some idea of the shocks they would be delivering. The behavior modification subjects were then taken to an adjoining room, seated at aggression machines with partitions that blocked observation of others' responses (thus, responses were experimentally independent), and given instructions for operating the shock apparatus. Each time the biofeedback subject's heart rate fell below the predetermined level, a signal light would be illuminated on their panels. It was explained that the higher the level chosen and the longer the switch was depressed, the stronger the shock administered would be. The "shock" received by the biofeedback subject was alleged to be the average of the intensities and durations selected by the four behavior modifiers. The final instruction given to the subjects was that any of the 10 shock switches would be sufficient for the purposes of the experiment. It was explained that the equipment had been designed with different shock intensities because we had not known how strong the shocks would have to be to increase heart rate. We explained we had discovered that the different shocks all had equal effects on the biofeedback subject's heart rate, so the naive subjects could choose any intensity they wished on each trial. These instructions were designed to eliminate any potential altruistic motivation, and they made clear that use of the lowest possible intensity on every trial would fulfill the requirements of the experiment. Use of any intensity greater than "1" would only result in additional pain to the biofeedback subject. Each group was presented with 20 signal lights over the course of the experiment. The interval between the appearance of any two signals was initially chosen randomly, ranging from 20-50 sec. The intervals were then held constant across subsequent trials. The experimenter then left to bring the biofeedback subject from a waiting room to the experimental room. The doors were left open, so that subjects heard the final instructions given by the experimenter to the biofeedback subject about his role. Thus, the naive subjects would easily hear, but not see, their future victim. Experimental Manipulations The first manipulation attempted to differentiate maximally between deindividuating situational cues and individuating ones. In the deindividuating cues condition, the experimenter did not address subjects by name. They were informed that the shocks they used were of no interest to the experimenter and that he would not know which intensities and durations they selected (anonymity to the experimenter). Subjects were further informed that they would not meet or see the biofeedback subject (anonymity to the victim). The experimenter indicated that he assumed full responsibility for the biofeedback subject's well-being (no responsibility for harm-doing). Finally, white noise was played at 65 dB (SPL) in the dimly lit room under the guise of eliminating any extraneous noise from the hall or other experimental rooms (arousal). Prentice-Dunn and Rogers (1980) have shown that such manipulations decrease the subjects' feelings of identifiability and self-awareness. In the individuating cues condition, the subjects wore name tags and were addressed on a first-name basis. As in Zimbardo's (1970) study, the "unique reactions" of each subject were emphasized, and the experimenter expressed his interest in the shock intensities and durations used by the subjects. Subjects were informed that they would meet the biofeedback subject on completing the study. It was emphasized that the biofeedback subject's well-being was the responsibility of each individual behavior modifier. The room was well-lit and no white noise was broadcast. A second independent variable, race of victim, was manipulated through the use of four experimental assistants, two whites and two blacks. Assistants were assigned to the treatment cells randomly, with the exception that they appeared an equal number of times in each treatment combination. Analyzing this "assistants" factor as an additional variable in the factorial DEINDIVIDUATION AND INTERRACIAL AGGRESSION design yielded no main or interaction effects. Thus, the data from the assistants of each race were pooled in the analyses reported below. The third independent variable was introduced when the behavior modifiers overheard a conversation in an adjoining room between the experimenter and the biofeedback subject. This conversation took place immediately after the naive subjects received their instructions. This insult manipulation was operationalized as a series of questions and answers between the experimenter and the biofeedback subject (i.e., our assistant). Care was taken that the insulting remarks applied to all of the subjects and were devoid of any racial content or connotation. In the insult condition, the biofeedback subject, when asked if he objected to the behavior modifiers shocking him, responded that the equipment looked complicated and he wondered if people who appeared as dumb as the behavior modifiers did could follow instructions properly. When the experimenter reiterated the biofeedback subject's option to withdraw from the experiment, the biofeedback subject answered that he hoped the behavior modifiers were not as stupid as they appeared. Finally, when asked by the experimenter if he knew the behavior modifiers, the biofeedback subject said that he didn't know them personally, but that he knew their type; he could tell that they thought they were "hot stuff." In the no insult condition, the biofeedback subject simply stated that he had no objections to these particular behavior modifiers shocking him. Postexperiment Session Following the last shock trial, subjects completed a questionnaire containing manipulations-check items (10point Likert rating scales) and 17 items tapping an internal state of deindividuation. A second questionnaire assessed suspicions about the experiment. Five subjects suspected that shocks were not actually delivered, two subjects correctly guessed the race-of-victim hypothesis, and four subjects believed that the insult manipulation was staged. These subjects were deleted from the data analyses. The data in this unequal n design were analyzed with the complete least squares model recommended by Overall, Spiegel, and Cohen (1975) because this model meets the criterion of estimating the same parameters as those estimated in an orthogonal design. After each experimental session, each subject was thanked and given a full debriefing that was based on Mills' (1976) recommendations. Finally, a questionnaire was given to each student in a stamped envelope addressed to the Psychology Department's Committee on Ethics. These anonymous responses were returned by 58% of the subjects. 98% of the respondents understood why the deception had been necessary and did not resent it. One subject (2% of the sample) indicated that he did not think the deception was necessary, and another subject stated that his participation had not been voluntary. Neither of these two subjects explained their responses in the space provided on the questionnaire. Fortunately, both of these students, along with every other respondent (100% of the sample), indicated that (a) the experiment 67 should be allowed to continue and (b) that they would be willing to participate in another similar experiment. Results Aggression Deindividuating cues. A multivariate analysis of variance was performed on the sums of the shock intensity scores and the shock duration scores. This analysis yielded a main effect associated with the situational cues manipulation, Wilks's lambda (A) = .843, F(2, 152) = 6.79, p < .001. Univariate analyses of variance disclosed main effects for the situational cues variable on both shock intensity, F(l, 77) = 12.01, p< .001, and on shock duration, F(l, 77) = 5.37, p < .03. Compared to subjects in the individuating cues condition, subjects exposed to the deindividuating cues used higher shock intensities (Ms = 5.3 and 6.4, respectively) for longer durations (Ms = 1.6 & 2.8 sec, respectively). Neither of the two predicted 2-way interaction effects with deindividuating cues were significant (Fs < 1). Insult and race. To determine if the insult manipulation had been successful, an analysis of variance was performed on the sum of the two items assessing anger toward the victim. The only significant effect was that the insulted groups (M = 5.1) expressed much more anger than the noninsulted groups (M=1.9), F(l, 77) = 29.71, p< .0001. A multivariate analysis of variance indicated that the insulted groups expressed more aggression toward the victim than the noninsulted groups, A = .904, F(2, 152) = 3.97, p < .02. More importantly, the multivariate analysis of variance revealed a Race X Insult interaction effect, A = .912, F(2, 152) = 3.59, p < .03. The aggression centroids are shown in Figure 1. Univariate analyses of variance yielded identical interaction effects for both the intensity data, F(l, 77) = 4.63, p < .04, and the duration data, F(l, 77) = 4.57, p < .04. A Duncan's multiple-range test (p < .05) yielded an identical pattern of results for both measures. As may be seen in Figure 1, when white subjects were not insulted, they expressed less aggression toward black than white victims; however, if the white subjects 68 RONALD W. ROGERS AND STEVEN PRENTICE-DUNN only one of the factors are shown in Table 1. This factor structure, although not predicted, is readily interpretable. Thefirstfactor, which is composed of four items from the original Self-Awareness factor, is highly Internal State of Deindividuation similar to Fenigstein, Scheier, and Buss' (1975) & Buss' (1980) concept of Public Prior research (Diener, 1979; PrenticeDunn & Rogers, 1980) had established that Self-Consciousness (coefficient alpha = .72). a subjective state of deindividuation is com- The items loading on this factor are highly posed of at least two factors, Self-Awareness similar to Fenigstein et al.'s items (e.g., "I'm and Altered Experience. The 17 items in the concerned about what other people think of Prentice-Dunn and Rogers (1980) study that me.") The original Altered Experience facloaded greater than .4 on only one of these tor decomposed into the second, third, and factors were included in the postexperiment fourth factors. The second factor might still questionnaire. Although the present study be labeled Altered Experience (alpha = .62). differed from the former one in that half of The third factor represents another facet of the subjects had been emotionally aroused altered experiencing, but refers specifically by an insult and in that half aggressed to Time Distortion (alpha = .35). The fourth against a different-race victim, an initial factor is composed of items traditionally principal-axes factor analysis with varimax used to measure cohesiveness, thus, it is larotation yielded two factors highly similar beled Group Cohesiveness (alpha = .61). to those previously reported. However, five The final factor might be labeled Responeigenvalues were greater than unity, hence sibility (alpha = .69); it reflected the degree five factors were rotated to a varimax solu- to which all group members, including the tion. This solution accounted for an addi- subject himself, had responsibility for harmtional 26% of the variance. The factor load- doing. The remaining two items loaded on ings that are greater than or equal to .4 on more than one factor. The question assessing a feeling of togetherness among group members loaded on both the Group Cohesion and Altered Experience factors. Finally, the item measuring inhibition loaded on both the BLACK Public Self-Awareness and Altered Experience factors. o 145r To replicate our previous study, we asked o whether the original two factors could discriminate between the individuating and deindividuating cues conditions. The sums 130of each subject's raw scores on the SelfAwareness items and the Altered Experience items were submitted to a multivariate oanalysis of variance. A significant difference was obtained between the two situational cues conditions, A = .882, F(2, 164) = 5.34, 5 c p < .006. Subjects in the deindividuating cues condition had higher scores on the Al: 100 tered Experience factor (M=65.8) than subjects in the individuating cues condition (M=62.8), but lower scores on the SelfI Awareness factor (Ms = 33.1 & 38.5, reINSULT NO INSULT spectively). A multivariate analysis of variFigure I . Aggression as a function of insult and race ance comparing the two situational cues conof victim. ditions on the five factors also yielded a were insulted, they expressed more aggression against black than white victims. There were no other significant main or interaction effects on the aggression data. DEINDIVIDUATION AND INTERRACIAL AGGRESSION 69 Table 1 Factor Loadings on the Rotated Factors Factor Loading Factor 1 (Public Self-Awareness) Concerned with what experimenter thought of me Concerned with what victim thought of me Concerned with what other group members thought of me Felt self-conscious Factor 2 (Altered Experience) Emotions were different from normal Thinking was somewhat altered Felt aroused Factor 3 (Group Cohesiveness) Liked other group members Session was enjoyable Willing to volunteer for another study with same group Factor 4 (Time Distortion) Thoughts were concentrated on the moment Time seemed to go quickly Factor 5 (Responsibility) I had responsibility for harm-doing Responsibility was shared .78 .78 .71 .50 .69 .65 .59 .70 .65 .77 .69 .67 .83 .84 Note. These five factors accounted for 13%, 14%, 13%, 9%, and 10% of the variance, respectively. significant main effect, A = .801, F(5, 79) = 3.91, p< .004. Deindividuation and Aggression The correlation matrix among the original two factors of the internal state of deindividuation and shock intensity was decomposed into components corresponding to paths. This analysis attempted an exact replication of previous findings that there was a causal path from Altered Experience and Self-Awareness to shock intensity. The coefficients were .29 and -.20, respectively, confirming the previous structural model. There was no strong a priori theoretical model that postulated relationships among our three independent variables, the five components of an internal state of deinviduation, and the dependent measures of shock intensity and duration. Therefore, no path analysis was performed. However, a multiple regression analysis was performed using the five factors to predict aggression (sum of shock intensity and duration). All of the relationships were in the expected di- rection, but the only statistically significant predictor of aggression was Group Cohesiveness (standardized partial regression coefficient = .25, p < .05). Discussion Deindividuation The results of the present experiment replicated and extended previous findings (Prentice-Dunn & Rogers, 1980) that deindividuating situational cues produce an internal state of deindividuation that mediates aggression. We had suggested previously that different configurations of stimulus conditions might produce different factor structures of the deindividuation state. Although the present data are only suggestive, the Self-Awareness and Altered Experience components may be differentiated into more precise aspects of deindividuation. First, the original Self-Awareness factor may reduce to Public Self-Awareness, which focuses on reactions of others to oneself (Buss, 1980; Fenigstein et al., 1975). The self-awareness literature has demonstrated that reductions 70 RONALD W. ROGERS AND STEVEN PRENTICE-DUNN in self-awareness are associated with inNeither of the two alternative forms of a creased feelings of deindividuation (Ickes, Deindividuation X Race or DeindividuaLayden, & Barnes, 1978) and with increased tion X Insult interaction effect were obaggression (Scheier, Feningstein, & Buss, tained. It was surprising that deindividua1974). Diener, Lusk, DeFour, and Flax tion did not interact with the race variable. (1980) pointed out that self-awareness is not Donnerstein and Donnerstein (1976) found a unitary phenomenon, but may have several that race interacted with anonymity, potenfacets. tial retaliation, and censure, all of which Our original Altered Experience factor were manipulated in the individuation-deinmay be differentiated into three independent deividuation independent variable. These infactors. One factor contained only items as- teractions were produced by whites' fear of sessing altered thinking and altered feelings, high levels of retaliation from blacks. In the accompanied by increased arousal. Although present study, however, whites reported the we shall retain the label Altered Experience same level of expected counteraggression for this factor, it is interesting to note its from blacks (M = 6.3 on a 10-point scale) similarity to Fenigstein et al.'s (1975) and and whites (M =6.4).' Thus, the state of Buss' (1980) concept of private self-aware- deindividuation neither made group memness. Private self-awareness refers to the pro- bers hyperresponsive to immediate stimuli cess of being conscious of one's thoughts and nor oblivious to them. A similar finding had feelings, whereas the Altered Experience been obtained for high- versus low-aggresfactor refers to a product of that awareness. sive models (Prentice-Dunn & Rogers, 1980). Another interesting facet of the original The models had the same effect in the inAltered Experience factor may be found in dividuating and deindividuating conditions. the Time Distortion factor, which reflected One of the next steps in constructing a thea concentration on the "here and now" that ory of deindividuation will be to reexamine seemed to pass more rapidly than it actually and refine general propositions that a state did. Perhaps the most readily interpretable of deindividuation will render a person more new factor to emerge was Group Cohesive- (or less) responsive to environmental cues. ness, which was the best predictor of ag- An important conceptual and empirical task gressive behavior. Finally, the data sug- is to identify the specific classes of environgested that feelings of responsibility may be mental stimuli to which a state of deindividanother component of the subjective state of uation would make a group member more deindividuation. or less responsive. It is important to emphasize that these data only suggest the existence of these ad- Interracial Aggression ditional factors for several reasons: (a) the ratio of subjects to items was only 5:1; (b) The Race X Insult interaction effect (see there were a small number of items per fac- Figure 1) indicated that nonangered white tor (2-4); and (c) partially due to the small subjects aggressed more against same-race number of items, the mean reliability coef- victims than against different-race victims; ficient was a modest .60. Future research in contrast, angered white subjects agthat overcomes these problems is obviously gressed more against different-race victims required before we can confidently accept than same-race victims. The interaction efthese additional components of a deindivid- fect was stronger in the present study than uated state. Taken together, the available in Baron's (1979). A key difference between data indicate that the internal state of de- the two experiments was whether the agindividuation is composed of at least two fac- gressors were acting alone or in a group. tors, Self-Awareness and Altered Experi- Clearly, the interaction emerges more ence, which may be differentiated into more refined subcomponents that reliably distin1 guish individuating from deindividuating sitThe authors wish to thank Ed Donnerstein for suguations. gesting this interpretation. DEINDIVIDUATION AND INTERRACIAL AGGRESSION strongly within a group setting. As Miller and Dollard (1941) noted, "People in a crowd behave as they would otherwise, only more so" (p. 218). If whites were not insulted, they engaged in reverse discrimination (cf. Dutton, 1976) and behaved consistently with current norms. Similarly, noninsulted blacks act consistently with current norms for their race (Wilson & Rogers, 1975). When emotionally aroused, however, the whites' behavior was consistent with the older, traditional pattern of discrimination against blacks. When emotionally aroused, blacks also act consistently with their traditional pattern (Wilson & Rogers, 1975). This finding is but one specific manifestation of Sargent's (1948) thesis that anger can lead to several overt responses, one of which is regression to more primitive behaviors. When emotionally aroused, both blacks (Wilson & Rogers, 1975) and whites (present study) regress to chronologically earlier modes of appropriate racial behavior. This pattern of interracial aggression could aptly be labeled regressive racism. Although racism may be defined to include beliefs of inferiority of a different race, we prefer to adopt the definition of racism as simply the differential treatment of people based solely on their race (P. Katz, 1976). Regressive racism differs in at least two important ways from the two other major forms of individual racism. First, both the dominative and aversive racists (cf. Kovel, 1970) firmly believe in the inferiority of blacks; they never embraced the new egalitarian norms. Thus, they would not exhibit reverse discrimination. Second, the concepts of dominative and aversive racism have been applied almost exclusively to whites. Regressive racism, on the other hand, may be found in the members of any race. To a large extent, black and white Americans take a Janus-faced view of interracial encounters, one face looking forward but the other face focusing grimly on the past. Katz and his colleagues (e.g., Katz & Glass, 1979) could interpret the Race X Insult interaction effect as evidence of ambivalence-amplification. Although the data from our program of interracial aggression may 71 be consistent with either the regressive racism or ambivalence-amplification interpretations, there are three major problems with the latter analysis of our data. First, ambivalence-amplification assumes that an individual cannot cope with mixed feelings: that a different-race person's behavior that threatens our positive or negative attitudes must be resolved by enhancing behavior that will confirm one aspect of our feelings and suppress the other. There is no direct evidence for this assumption. It is more intuitively compelling to believe that people can act on a mixture of positive and negative feelings toward other people. Second, there is no evidence to support the proposed mediating process, whether it is labeled threat to self-esteem, guilt, distress over inequity, or dissonance. Several of Katz's studies (especially Katz, Glass, Lucido, & Farber, 1979) have tested the ambivalence-amplification hypothesis that ambivalence heightens guilt and that the ensuing behavior reduces guilt. Despite the great difficulties of measuring this mediator, Katz's studies have found no supporting evidence. Furthermore, the present study included an item assessing guilt feelings in the postexperiment questionnaire. The data analyses indicated that the independent variables did not differentially affect guilt, nor did the subjects' behavior reduce guilt. Third, it is very difficult for the ambivalence-amplification hypothesis to account for blacks' aggressive behavior (Wilson &Rogers, 1975). In contrast, the concept of regressive racism can account for the behavior of both whites and blacks. Additional support for and an extension of the concept of regressive racism may be found in the research from the Donnersteins' program. An interaction effect highly similar to the one displayed in Figure 1 has been reported between the race-of-victim variable and (a) anonymity-to-victim (Donnerstein et al., 1972), (b) threatened retaliation (Donnerstein et al., 1972), and (c) potential censure (Donnerstein & Donnerstein, 1973). Thus, in Figure 1, if the two points on the abcissa labeled "no-insult" and "insult" were to be relabeled (a) "nonanonymous" and "anonymous," (b) "retaliation" and "no-retaliation," or (c) "potential censure" 72 RONALD W. ROGERS AND STEVEN PRENTICE-DUNN and "no censure," the same interaction would be present. Crosby, Bromley, and Saxe (1980) interpreted the Donnersteins' findings as revealing that antiblack hostility is actually pervasive, but subtle and covert. The Donnersteins reported that whites were very hostile toward blacks in those conditions in which their negative behavior could not be detected or punished. These latter conditions (anonymity, no retaliation, no censure) are well-established disinhibitors of aggression that increase aggressive responding. The present study disclosed that a well-established instigator of aggression, anger arousal, produced the same effect as the disinhibitors. Perhaps regressive racism is revealed not only by emotional arousal, but also by a host of variables that disinhibit or instigate aggression. Moreover, regressive racism may be revealed whenever unprejudiced values are not fully internalized (one case of incomplete internalization occurs when the unprejudiced values are relatively new). Regressive racism may be a sufficiently inclusive concept to give coherence to a variety of interracial behavior. Interpreting the Donnersteins' data in terms of regression racism is speculative, and perhaps this extension of the concept goes too far. The concept of regressive racism, however, accounts quite well for anger-mediated interracial aggression. Both the blacks and whites in our studies and the whites in the Donnersteins' studies have been highly sensitive to the race of their victim. Sadly, neither whites nor blacks accorded the other race egalitarian treatment. Awaiting us, black and white together, is the fire next time; it is smoldering in the regressive racist. Reference Note 1. Rosenberg, J. Racial attitudes of undergraduate students. Unpublished manuscript, University of Alabama, 1980. References Baron, R. A. Human aggression. New York: Plenum Press, 1977. Baron, R. A. Effects of victim's pain cues, victim's race, and level of prior instigation upon physical aggression. Journal of Applied Social Psychology, 1979,9, 103114. Brigham, J., & Wrightsman, L. Contemporary issues in social psychology (4th ed.). Monterey, Calif.: Brooks/Cole, in press. Buss, A. H. Self-consciousness and social anxiety. San Francisco: Freeman, 1980. Campbell, A. White attitudes toward black people. Ann Arbor: University of Michigan, Institute for Social Research, 1971. Caplan, N. The new ghetto man: A review of recent empirical studies. Journal of Social Issues, 1970, 26, 59-73. Crosby, F., Bromley, S., & Saxe, L. Recent unobtrusive studies of black and white discrimination and prejudice: A literature review. Psychological Bulletin, 1980, 87, 546-563. Davis, E., & Fine, M. The effects of the findings of the U.S. National Advisory Commission on Civil Disorders: An experimental study of attitude change. Human Relations, 1975, 28, 209-227. Diener, E. Deindividuation: Causes and consequences. Social Behavior and Personality, 1977, 5, 143-155. Diener, E. Deindividuation, self-awareness, and disinhibition. Journal of Personality and Social Psychology, 1979, 37, 1160-1171. Diener, E. Deindividuation: The absence of self-awareness and self-regulation in group members. In P. Paulus (Ed.), The psychology of group influence. Hillsdale, N.J.: Erlbaum, 1980. Diener, E., Lusk, R., DeFour, D., & Flax, R. Deindividuation: Effects of group size, density, number of observers, and group member similarity on self-consciousness and disinhibited behavior. Journal of Personality and Social Psychology, 1980, 39, 449-459. Donnerstein, E., & Donnerstein, M. Variables in interracial aggression: Potential ingroup censure. Journal of Personality and Social Psychology, 1973,27,143150. Donnerstein, E., & Donnerstein, M. Research in the control of interracial aggression. In R. Geen, & E. O'Neal (Eds.), Perspectives on aggression. New York: Academic Press, 1976. Donnerstein, E., Donnerstein, M., Simon, S., & Ditrichs, R. Variables in interracial aggression: Anonymity, expected retaliation, and a riot. Journal of Personality and Social Psychology, 1972, 22, 236245. Dutton, D. G. Tokenism, reverse discrimination, and egalitarianism in interracial behavior. Journal of Social Issues, 1976, 32, 93-108. Fenigstein, A., Scheier, M. F., & Buss, A. H. Public and private self-consciousness: Assessment and theory. Journal of Consulting and Clinical Psychology, 1975, 43, 522-527. Festinger, L., Pepitone, A., & Newcomb, T. Some consequences of de-individuation in a group. Journal of Abnormal and Social Psychology, 1952,47,382-389. Griffin, B. Q., & Rogers, R. W. Reducing interracial aggression: Inhibiting effects of victim's suffering and power to retaliate. Journal of Psychology, 1977, 95, 151-157. DEINDIVIDUATION AND INTERRACIAL AGGRESSION Ickes, W., Layden, M. A., & Barnes, R. D. Objective self-awareness and individuation: An empirical link. Journal of Personality, 1978, 46, 146-161. Katz, I., & Glass, D. An ambivalence-amplification theory of behavior toward the stigmatized. In W. G. Austin & S. Worchel (Eds.), The social psychology of intergroup relations. Monterey, Calif.: Brooks/ Cole, 1979. Katz, I., Glass, D., Lucido, D., & Farber, J. Harmdoing and victim's racial or orthopedic stigma as determinants of helping behavior. Journal of Personality, 1979, 47, 340-364. Katz, P. Racism and social science: Towards a new commitment. In P. Katz (Ed.), Towards the elimination of racism. New York: Pergamon Press, 1976. Kovel, J. White racism: A psychological history. New York: Pantheon, 1970. Miller, N., & Dollard, J. Social learning and imitation. New Haven, Conn.: Yale University Press, 1941. Mills, J. A. A procedure for explaining experiments involving deception. Personality and Social Psychology Bulletin, 1976, 2, 3-13. Overall, J. E., Spiegel, D. K., & Cohen, J. Equivalence of orthogonal and nonorthogonal analysis of variance. Psychological Bulletin, 1975,82, 182-186. Prentice-Dunn, S., & Rogers, R. W. Effects of deindividuating situational cues and aggressive models on subjective deindividuation and aggression. Journal of 73 Personality and Social Psychology, 1980, 39, 104113. Ross, J, M., Vanneman, R. D., & Pettigrew, T. F. Patterns of support for George Wallace: Implications for racial change. Journal of Social Issues, 1976,32,6991. Rude, G. The crowd in history. New York: Wiley, 1964. Sargent, S. S. Reaction to frustration—A critique and hypothesis. Psychological Review, 1948,55,108-114. Scheier, M. F., Fenigstein, A., & Buss, A. H. Selfawareness and physical aggression. Journal of Experimental Social Psychology, 1974, 10, 264-273. Southern Commission on the Study of Lynching. Lynchings and what they mean. Atlanta, G.: Author, 1931. Taylor, D. G., Sheatsley, P. B., & Greeley, A. M. Attitudes toward racial integration. Scientific American, 1978, 238(6), 42-49. Wilson, L., & Rogers, R. W. The fire this time: Effects of race of target, insult, and potential retaliation on black aggression. Journal of Personality and Social Psychology, 1975, 32, 857-864. Zimbardo, P. G. The human choice: Individuation, reason, and order versus deindividuation, impulse, and chaos. In W. J. Arnold & D. Levine (Eds.), Nebraska Symposium on Motivation (Vol. 17). Lincoln: University of Nebraska Press, 1970. Received November 6, 1980 •
Purchase answer to see full attachment