i am us


1 Copyright 2005 by the American Psychological Association Journal of Personality and Social Psychology 0022-3514/05/$12.00 DOI: 10.1037/0022-3514.89.4.566 2005, Vol. 89, No. 4, 566–582 “I Am Us”: Negative Stereotypes as Collective Threats Geoffrey L. Cohen and Julio Garcia Yale University Collective threat is the fear that an ingroup member’s behavior might reinforce a negative stereotype of one’s group. In a field study, self-reported collective threat was higher in stereotyped minorities than in Whites and was linked to lower self-esteem in both groups. In 3 experimental studies, a potentially poor performance by an ingroup member on a stereotype-relevant task proved threatening, as evidenced by lower self-esteem rd experiment. The latter study demonstrated the among minority students in 2 experiments and women in a 3 generality of collective threat. Collective threat also undermined academic performance and affected self- stereotyping, stereotype activation, and physical distancing from the ingroup member. Results further suggest that group identification plays a role in whether people use an avoidance or challenge strategy in coping with collective threat. Implications for theories of social identity and stigmatization are discussed. Keywords: social identity, stereotype threat, racial identification, stigma, academic achievement poorly and thus lending credence to the stereotype about the intellec- Because people derive both identity and self-worth from their tual inferiority of their racial or gender group (Spencer, Steele, & group memberships, their thoughts, feelings, and actions are influ- Quinn, 1999; Steele & Aronson, 1995; Steele, Spencer, & Aronson, enced by the objective outcomes of their group. That is, individual 2002; see also Aronson, 2002; Cohen & Steele, 2002; Cohen, Steele, psychology is affected by collective outcomes. People experience & Ross, 1999; Mendoza-Denton, Downey, Purdie, Davis, & Pietrzak, increases in self-esteem when fellow group members succeed, which 2002; Vorauer, Main, & O’Connell, 1998). In the present article, we may manifest in changes even at the hormonal level (Bernhardt, examine the situation where, although running no risk of personally Dabbs, Fielden, & Lutter, 1998). When other group members trans- lending support to a stereotype about their group, individuals are gress, individuals may experience guilt even if they personally had no concerned about the potentially stereotype-confirming acts of other involvement in these acts (Doosje, Branscombe, Spears, & Manstead, as it collective threat, members of their group. We call this concern 1999). The present article begins with the idea that just as people can issues from the collectively shared nature of social identities. We vicariously share in the objective outcomes of fellow group members, further suggest that in situations where one’s group is negatively they also may share in experiences that are more subtle and subjective stereotyped, an “I am us” mindset may arise out of the awareness that in nature. We explore this idea in the context of stigmatization based the way one is viewed and defined depends, in part, on the way that on race and gender. We suggest that the distress of stigmatization other group members are viewed and defined. need not arise from firsthand experience. Collective threat issues from the awareness that the poor perfor- Being personally victimized by overt discrimination can be threat- mance of a single individual in one’s group may be viewed through ening (Branscombe, Schmitt, & Harvey, 1999). Even subtler threats the lens of a stereotype and may be generalized into a negative based on one’s social identity may be distressing, such as the knowl- judgment of one’s group. Indeed, this apprehension may be well edge that one’s behavior could be used to reinforce a negative ste- founded, as people draw conclusions about entire groups based on the reotype about one’s group. For example, as research on stereotype behavior of individual group members (Henderson-King & Nisbett, threat has demonstrated, Black students completing an intelligence 1996). In the research presented here, we examine collective threat test, or women completing a math test, may worry about performing among students who face negative stereotypes about the intellectual ability of their group—that is, ethnic minority students in school generally and women in math in particular (Steele et al., 2002). They Geoffrey L. Cohen and Julio Garcia, Department of Psychology, Yale may feel threatened, we argue, when the possibility is made salient University. that a fellow group member may perform poorly in an academic Portions of this research were supported by a grant from the Nellie Mae context and thus reinforce a negative stereotype about their race or Education Foundation. We thank Mitchell Prinstein and his lab for helping to organize the data collection reported in the pilot study; Jiraorn Assarat, gender. Annie Hsu, Lillian Hsu, and Stephanie Gray for their invaluable assistance Why would an ingroup member’s potentially stereotype- conducting the experiments; Allison Master for her assistance in conduct- confirming behavior prove threatening? One reason involves the ing the pilot study and for comments on previous drafts; Daryl McAdoo for role of group membership as a source of self-definition (Tajfel & helping in the design of the stimuli used in Experiment 2; and Joshua Turner, 1986; see also Schmader, 2002; Walton & Cohen, 2003). Correll for feedback and assistance with statistical analyses. We also thank As research on social identity theory confirms, people use their Nancy Apfel, David Sherman, Steve Spencer, Eric Uhmann, and Greg groups as a basis of self-evaluation (Tajfel & Turner, 1986). Walton for feedback on earlier versions of this article. Because people want to see themselves in a positive light, they Correspondence concerning this article should be addressed to Geoffrey L. want to maintain a positive image of their group and are likely to Cohen, Department of Psychology, Yale University, 2 Hillhouse Avenue, P.O. feel threatened when their group could be viewed negatively Box 208205, New Haven, CT 06520-8205. E-mail: [email protected] 566

2 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 567 (Lewis & Sherman, 2003; Rubin & Hewstone, 1998; Schmader, On an exploratory basis, we also assessed whether the impact of collective threat varies with the personal importance of partici- 2002). Indeed, as research on the black sheep effect suggests, pants’ group identity. On the one hand, people who identify with people may disassociate from and even denigrate ingroup mem- their group may feel more threatened by the poor performance of bers whose behavior reflects negatively on their group (Lewis & an ingroup member, because they derive a greater sense of self- Sherman, 2003; Marques & Paez, 1994; see also Ellemers, Van identity from the positive representation of their group (Schmader, den Heuvel, De Gilder, Maass, & Bonvini, 2004). 2002). On the other hand, high group identification may act as a Regardless of the personal centrality of their group identity, source of social support and self-esteem that offsets the pain of people may view the stereotype-confirming behavior of an ingroup stigmatization (Branscombe et al., 1999; Wong, Eccles, & Samer- member as threatening because they are aware that they them- off, 2003). High group-identified individuals also have more mo- selves could be defined by others on the basis of their group tivation and ability both to reject negative representations of their (Cohen & Steele, 2002; Klein & Azzi, 2001; Vorauer et al., 1998). group (Doosje et al., 1999; Oyserman, Kemmelmeier, Fryberg, Knowing that an ingroup member might have confirmed the ste- Brosh, & Hart-Johnson, 2003) and to challenge its lower status in reotype in the minds of others, people may believe that they a hierarchy (Ellemers, Spears, & Doosje, 1997; Oyserman et al., personally will now be at a greater risk of being judged negatively. 2003; Spears et al., 1997; Tajfel & Turner, 1986)—tendencies that The concern that one could be viewed negatively because of one’s might buffer them against negative stereotypes of their group. group membership is threatening to self-worth (Branscombe et al., 1999; Ethier & Deaux, 1994; Mendoza-Denton et al., 2002; Vorauer et al., 1998), and it motivates the deployment of coping Pilot Study strategies among individuals low and high in group identification In a pilot study, we sought to determine whether students alike (see Klein & Azzi, 2001; Spears, Doosje, & Ellemers, 1997). belonging to intellectually stereotyped racial groups (i.e., Black More generally, it is threatening to believe that other people could and Latino students) report experiencing collective threat in school reject one on the basis of any personal characteristic (Leary, 1999). and whether the reported levels of threat are associated with lower This is especially true in a setting, such as school or work, where self-esteem. As past research finds that minority and majority people want to belong and where the judgments of others have members alike feel threatened by negative stereotypes of their large material and symbolic consequences (Steele et al., 2002). ́sert, Croizet, & Darcis, group (Aronson et al., 1999; Leyens, De We suggest that for many stereotyped students, the threat of 2000), we also hoped to ascertain whether the posited relationship other people confirming a negative stereotype about their group between collective threat and lower self-esteem holds regardless of may be at least as acute as the threat of personally doing so. This student race. may be the case for several reasons. First, people tend to think that We also examined, on an exploratory basis, whether collective their peers have less ability than they themselves have (Dunning, threat predicts a drop in students’ grade point average (GPA). Meyerowitz, & Holzberg, 1989; see also Dunning & Cohen, Believing that a fellow group member has reinforced a stereotype, 1992). Members of a stereotyped group may thus expect fellow people may worry that others now view that stereotype as more group members to be at a greater risk of performing poorly, and of valid than they did before. The belief that others endorse a stereo- substantiating a negative stereotype, than they personally are. type of one’s group as inferior can worsen performance in the Regardless of their own level of skill, moreover, such individuals domain of the alleged inferiority (Aronson, 2002; Steele et al., would realize that at least some people in their group will perform 2002). Accordingly, we expected collective threat to predict lower poorly. Consistent with this claim, Blacks express negative, ste- GPA for students targeted by a negative stereotype about their reotypical beliefs about other Blacks, sometimes even more so intellectual ability—Black and Latino students (Aronson, 2002). than Whites (Sniderman & Piazza, 1993). Another factor increas- We distributed surveys to a sample of 472 junior and senior ing collective threat arises from people’s tendency to believe that students attending a high school in a New England suburb. Of the they are less victimized by prejudice than are fellow group mem- participating students, 18% were Black, 3% were Latino, 75% bers (Crosby, 1984). Members of a stereotyped group may thus were White, and 5% were “other.” Students were classified as a acknowledge that the behavior of others in their group could be member of a stereotyped minority group if they were Black or 1 viewed in light of their race or gender but minimize the possibility Latino. Both Black and Latino students face a stereotype about that their own behavior could be so viewed. Moreover, because the intellectual inferiority of their race, and members of both there are more evaluative events involving other people than those groups perform worse academically than their nonstereotyped involving oneself, and because the performances of other people peers (Aronson, 2002). Furthermore, they did not differ on any are less controllable than are one’s own, situations of collective p  s 1.5,  s t .15). measure ( threat may prove especially stressful on account of their chronic and uncontrollable nature. 1 Consistent with the U.S. Census Bureau, “Black,” “Latino/Hispanic,” Our primary objective focused on determining whether collec- and “White” constitute racial categories of self-identification. These cate- tive threat constitutes a concern of negatively stereotyped students gories are sociopolitical constructs rather than biological ones. Also, the in an academic setting and whether it is, in fact, threatening. majority of students in the “other” ethnic category were Asian; all others Consistent with the research described above, our primary measure were a mix of non-Black and non-Latino ethnic groups. Although students of threat was self-esteem. Fears of confirming the stereotype in this group are members of an ethnic minority group, their academic oneself, and of being personally discriminated against, were either abilities are not negatively stereotyped; indeed, in some cases, they are statistically controlled (in our pilot study) or experimentally re- positively stereotyped (Shih, Pittinsky, & Ambady, 1999). Accordingly, we place these students in a separate racial category. moved (in Experiments 1–3).

3 COHEN AND GARCIA 568 varied with the type of racial threat. Collective threat and the threat was assessed with the item, “In school, I worry Collective threat of being stereotyped were rated as more worrisome than stereotype that people will draw conclusions about my racial group, based on (2, 444) F threat,  34.13, p  .01. Third, there was an interaction people in my race.” the performances of Stereotype threat other (2, 444) 15.22,  F between student race and type of racial threat, was assessed with the item, “In school, I worry that people will .01. The tendency for stereotyped minority students to report p  draw conclusions about my racial group based on my perfor- more racial threat than White students was most pronounced for was assessed threat of being stereotyped mances.” A more general collective threat and for the threat of being stereotyped. Neverthe- with the item, “In school, I worry that people will draw conclu- less, collective threat proved the highest rated racial concern sions about me, based on what they think about my racial group.” among minority students—significantly higher than both stereo- The latter two items were based on a questionnaire developed by type threat, .01, and the threat of being  p 71.17,  (1, 445) F  Steele et al. (2003). Responses were made on separate scales (1 3.87, F (1, 445)  stereotyped, p  .05. ). The order of the focal strongly disagree ;7  strongly agree To determine the relationship between collective threat and stereotype threat and collective threat items was counterbalanced; outcomes (i.e., self-esteem and GPA), we used regression, con- the item assessing threat of being stereotyped always came last. No trolling for stereotype threat, threat of being stereotyped, level of effect of order was found. The three measures of racial threat perceived racial discrimination (assessed with a two-item scale, proved to be correlated among stereotyped minority students (av- e.g., “How much discrimination do you suffer because of your .69), White students (average erage r  r  .75), and all other race?”), and racial identification. We measured racial identifica- students (average  .60). r tion with the Race Centrality subscale of the Multidimensional We wanted to assess whether stereotyped minority students, Inventory of Black Identity, supplemented with four items tapping compared with White students, reported higher levels of racial the importance of race to students’ social lives (e.g., “I feel the threat in general and of collective threat in particular. Analysis most comfortable with people in my racial group”; Sellers, Row- used participant race as a between-subjects independent variable ley, Chavous, Shelton, & Smith, 1997). We assessed self-esteem and type of racial threat (collective threat, stereotype threat, threat with Harter’s (1988) Global Self-Worth Scale for adolescents. of being supervised) as a repeated measure. Two orthogonal con- Because the measures used different scales, all independent and trast codes tested the effects of participant race, one for the dependent variables were first standardized. theoretically important contrast between stereotyped minority stu- As expected, higher collective threat predicted lower self- dents and White students and the other for the theoretically unim-  esteem, B .05. No other measure .19, t (413)  1.98, p  portant contrast between these two groups combined and all other p 1.6,  s t attained significance ( s .12). Furthermore, collective  students. threat predicted lower self-worth regardless of students’ race or Figure 1 displays the relevant means for the two focal ethnic level of racial identification. None of the two-way or three-way (1, 445)  groups. First, there was a main effect of student race, F interactions involving collective threat, student race, or racial p 103.34, .01. Stereotyped minority students reported higher  identification yielded a significant result ( t s  1). levels of racial threat than did White students. Second, responses 5 e etypeT eo St r a hr t eo tr Thr t etyped eo fB eing S a T c Coll t ehrt a e eiv 4 3 2 1 ents y r io u t te S W td u en n s hi M it St d Mean levels of racial threat as a function of students’ minority status: Pilot study. Scale ranges from Figure 1. 1( strongly disagree )to7( strongly agree ).

4 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 569 dents was collective threat, and it uniquely predicted lower self- We obtained academic records for all but 30 students; students esteem among all students. with missing data had either dropped out of school or moved. To Although the consequences of collective threat for self-esteem compute change in GPA, we regressed GPA at the end of the year did not appear to depend on racial identification, its consequences (after our assessment) on GPA at the beginning of the year (before for GPA did. Deriving identity from one’s racial group seemed to our assessment), using the unstandardized residuals as the out- buffer minority students against the negative impact of collective come. This outcome represents the difference between actual GPA threat on their GPA—a finding that is consistent with the protec- and expected GPA based on prior grades. It is analogous to a tive effect of racial identification on academic performance ob- change score. served by Wong et al. (2003; see also Oyserman et al., 2003; cf. Stereotyped minority students evidenced a decline in academic Schmader, 2002). Those who are high in ethnic identification may performance ( M .08), whereas White students showed a slight  be better able to mobilize social and psychological resources that  M .01. This widening in  .06), B  .23, gain ( t (398)  3.90 p ameliorate the threat of stigmatization (see Branscombe et al., the racial achievement gap arose from the worsening grades of 1999; Ethier & Deaux, 1994; Oyserman et al., 2003; Wong et al., minority students who reported both low levels of racial identifi- 2003). cation and high levels of collective threat. That is, although higher Encouraged by these correlational data, we conducted an exper- B collective threat predicted worse GPA overall,   .31, t (398) iment to assess the impact of collective threat on self-worth and on  p 3.10 .01, it also interacted with students’ race and level of  other important outcomes.   .11, t (398) B racial identification, .02. Thus,  p 2.32,  minority students exhibited a Collective Threat Racial Identifi- B  cation interaction, p  .01; White students t .26, (398) 3.08,  Experiment 1 1). Among minority students low in racial identifica-  t did not ( tion (one standard deviation below the mean), collective threat led Our objective in Experiment 1 again focused on examining .01. Among those 2.82, B to worse GPA,  .35, t (398)  p  whether the potentially stereotype-confirming behavior of an in- high in racial identification (one standard deviation above the group member constituted a threat. As in the pilot study, our p  mean), it did not ( 1.4, t  .18). The relevant means, in their primary measure of threat was self-esteem. Black college students original metric, are presented in Figure 2. observed a same-race peer in an intellectually evaluative situation In summary, our pilot study found that collective threat consti- modeled after that used in research on stereotype threat (Steele & tutes an openly reported concern on the part of minority students Aronson, 1995). Half of them were randomly assigned to a con- in an actual academic environment, one that they expressed with- dition designed to evoke collective threat. They saw another mem- out the imminent stressor of a test or assignment. Additionally, our ber of their ethnic group in a stereotype-threat situation—specifi- results suggest that people feel uniquely threatened by the potential cally, that person was preparing to complete a difficult and for acts of fellow group members to perpetuate a negative image of evaluative verbal ability test (Steele & Aronson, 1995). The re- their group. The highest rated racial threat among minority stu- maining Black students were assigned to a no-threat condition. 03 . y al c i al Low R d Ieni tfied 02 . Mno i i rties c l i aly a High R d Ieni tfied 01 . rties i Mno i 0 o High C oll t c eivT L ehrt e etiveThr t wC a o a e ll c - 1 0 . 0 - 2 . 0 -.3 Figure 2. Unstandardized residuals of end-of-year GPA regressed on beginning-of-year GPA (i.e., the difference between actual end-of-year GPA and expected end-of-year GPA based on prior grades) as a function of minority students’ level of racial identification and level of collective threat: Pilot study. Positive values reflect improvement in GPA; negative values reflect decline.

5 COHEN AND GARCIA 570 They saw the same person in a non-stereotype-threat situation, In summary, our primary prediction in Experiment 1 was that collective threat would lead to lower self-esteem—a conceptual specifically, preparing to complete verbal puzzles. In contrast to replication of the result predicted and obtained in the pilot study. participants in stereotype threat research, threatened participants in Additionally, we assessed outcomes linked to how people cope our study neither thought that their abilities were being evaluated with collective threat. nor expected to take a test themselves. We predicted that Black participants would have lower self-esteem in the threat condition, as only in this condition could the same-race peer confirm the Method stereotype about the intellectual inferiority of their race. Participants and Design No White students were assessed in this procedure, as research consistently finds that nonstigmatized students show higher self- Sixty-three Black undergraduates at Yale University (44 women, 19 evaluation even when exposed to an obviously poor intellectual men) participated in the study in exchange for $8. An official registrar’s list performance by a peer (Tesser, 1988). That is, because White specifying students’ ethnicity and contact information was used to identify students do not contend with a threatened social identity in this potential participants. They were randomly assigned either to a threat context, more individual-based social comparison processes condition or to a no-threat condition. One participant in each condition suspected that the study concerned responses to racial stereotypes; accord- predominate. ingly, their data were discarded prior to analyses. The results of the pilot study, which provided evidence for the role of collective threat in student GPA, gave rise to a secondary goal of Experiment 1, that is, exploring the consequences of Procedure collective threat for behavior. People cope with stressors by en- Students participated in the experiment individually. On arriving at the gaging in either a flight or a fight response (Cannon, 1932) or, to laboratory, they were greeted by an Asian American female experimenter draw on analogous terms from Blascovich and Tomaka (1996), a and told that a second participant would arrive momentarily. Approxi- threat or challenge response. In the context of collective threat, a mately 2 min later, a Black female student (a confederate) entered the flight response with respect to one’s social identity involves dis- waiting area, and the participant and the confederate were then escorted to tancing oneself from the threatening (i.e., stereotypical) qualities the laboratory room. For each participant, one of two experimenters (both Asian women) and one of three confederates (all Black women) were used. of one’s group (Pronin, Steele, & Ross, 2004; Spears et al., 1997; (There were no consistent effects involving either experimenter or confed- Steele & Aronson, 1995). It might involve disassociating oneself erate.) After thanking the participant for his or her participation, the from the same-race peer, as those who affiliate or come into close experimenter looked down at her clipboard and said, “It turns out that proximity with stigmatized individuals risk being denigrated them- we’re just about done with our main study. And we only need one more selves (Neuberg, Smith, Hoffman, & Russell, 1994). A flight person to wrap it up.” For that reason, she explained, only [confederate’s response might also lead to avoidance of intellectual challenges name] would complete the original study, because “her name appeared first where one would continue to contend with the alleged inferiority on the list.” The other person (i.e., the participant) would simply complete of one’s group (Aronson, 2002; Cohen et al., 1999; Mendoza- some “background questionnaires for a different study.” The participant and the confederate were each given an informed con- Denton et al., 2002). By contrast, a fight response entails asserting sent form to sign. The participant was then asked to wait while the solidarity with one’s group in the face of threat and embracing experimenter provided instructions to the confederate. The experimenter rather than eschewing non-negative qualities emblematic of one’s explained to the participant that these instructions were directed to the social identity (Branscombe et al., 1999; Klein & Azzi, 2001; “other subject” and that the instructions were thus irrelevant to the study Spears et al., 1997; Tajfel & Turner, 1986). A fight response might that the participant would complete. Because the instructions given to the also involve trying to refute the allegations directed at one’s group confederate were brief, their presentation in front of the participant ap- by confronting rather than avoiding intellectual challenges. peared to be a matter of convenience only. No participant (beyond the two Students completed a set of measures examining the conse- suspicious participants previously noted) questioned this aspect of the quences of collective threat for such fight-or-flight responses: (a) procedure during a thorough postexperimental interview. In the threat condition, the participant overheard the experimenter tell a stereotype distancing measure (Steele & Aronson, 1995; see also the confederate that the researchers were interested in the “various factors Klein & Azzi, 2001; Pronin et al., 2004; Spears et al., 1997), on that affect performance on problems that demand strong reading and verbal which participants indicate the applicability to themselves of traits abilities” and that “for the next 25 minutes, you’ll be working on a and interests stereotypical of their racial group (e.g., enjoying standardized test of verbal ability . . . identical in format to that of other basketball); (b) a physical proximity measure, that is, the distance standardized tests of verbal ability, like the SAT [Scholastic Aptitude that individuals choose to sit from a same-race peer who poten- Test].” The experimenter explained that the test was “quite difficult” in tially confirmed the stereotype; and (c) a situational readiness order to obtain an “accurate and reliable measure of your verbal abilities measure, in which participants indicate their willingness to expose and limitations.” These instructions were modeled after those used by Steele and Aronson (1995). Two procedural details reinforced the threat. themselves to a stereotype threat situation in which their own First, after the experimenter left the room momentarily, the confederate academic skills will be evaluated. On an exploratory basis, we also said, “I’m so bad at these standardized tests.” Second, after being given the included (d) a measure of stereotype activation. Although individ- test, the confederate asked the experimenter, “Will you be correcting this uals under the threat of a negative stereotype think about that at the end?” The experimenter replied, “Yes, your performance will be stereotype—increasing its cognitive accessibility—they are also evaluated” and then reiterated the importance of expending “your best motivated to suppress these thoughts. This pressure inhibits ste- effort” to help “us get the best possible evaluation of your verbal ability” reotype activation when mental resources are available (e.g., under (see Steele & Aronson, 1995). conditions of low cognitive load; Iserman, Spencer, Davies, & The procedure in the no-threat condition, again modeled after Steele and Aronson (1995), was identical to that of the threat condition with three Quinn, 2004).

6 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 571 exceptions. First, the experimenter did not mention ability and instead  (“I feel good about myself”). Responses were made on separate scales (1 stated that the researchers were interested in the “various factors involved ,7  strongly agree ). All items had loadings greater than strongly disagree in solving verbal puzzles” and that “for the next 25 minutes, you’ll be .50 on the first unrotated factor. After reverse coding where appropriate, we working on a set of verbal puzzles and games.” The puzzles were “quite   .75), with higher values signi- summed the items into a composite ( difficult,” the experimenter explained, because “we are interested in how fying more positive self-esteem. people solve challenging verbal puzzles.” Second, after the experimenter Stereotype distancing. This measure asked students to rate the extent to left the room momentarily, the confederate commented, “Verbal puzzles which they enjoy various activities, characterize themselves as having . . . I remember these.” Finally, after being given the test, the confederate various traits, and like various types of music and sports (Steele & asked, “So should I just be working on these verbal puzzles?” The exper- Aronson, 1995). Some of the activities and traits were associated with the imenter replied, “Yes, that’s what you’ll be doing here today” and reiter- stereotypic image of African Americans. In the activities category, the ated the importance of expending “your best effort” to help “us in analyz- stereotype-relevant items were “playing sports,” “socializing,” and “exer- ing the problem-solving process.” Manipulation checks administered at the cising.” In the traits category, the stereotype-relevant items were “aggres- end of the study confirmed that participants accurately overheard the sive,” “humorous,” “lazy,” “easygoing,” and “good-natured.” In the music instructions in each condition. category, the stereotype-relevant items were “rap,” “rhythm and blues,” In both conditions, participants were next told to bring their chair to a and “jazz.” In the sports category, the stereotype-relevant item was “bas- 2 nearby room. In this room, they completed the dependent measure ques- ketball.”  Responses were made on separate scales (1 ,7 not at all  tionnaires assessing state self-esteem, stereotype distancing, and racial extremely ). They were summed into a composite. stereotype activation. We also tested whether our instructions successfully Racial stereotype activation. Participants completed a word-fragment conveyed that participants’ own abilities were equally free from evaluation completion exercise (Steele & Aronson, 1995), in which 40 word frag- in the threat condition and in the no-threat condition. To do so, we ments were presented. Some of them (e.g., _ _ACE) could be completed measured how much participants felt that their verbal abilities were ex- ,FACE)or with a either with a stereotype-irrelevant word (e.g. posed to evaluative scrutiny. stereotype-relevant one (e.g. ,RACE).The potential race-relevant words Participants deposited the questionnaires in a drop-box upon completion included race, welfare, lazy, color, class, brother, black, bias, riot, soul, and then returned to the original room with their chair. At this time, the poor, and minority. The total number of stereotype-relevant words each participant saw the confederate seated and apparently finishing the study. participant generated constituted the measure of racial stereotype After the participant positioned his or her chair and was seated, the activation. experimenter (using a piece of tape) covertly marked the participant’s Perceived exposure to evaluative scrutiny. This measure asked partic- seating distance from the confederate. ipants to indicate the extent to which they felt that their verbal abilities In the presence of the participant, the experimenter asked whether the very much  not at all ,7 were being evaluated in the study (1 ).  confederate had any questions about the study. The confederate had none Racial identification. As in the pilot study, racial identification was and was subsequently thanked and excused. Next, the experimenter asked assessed. Because a long inventory risked alerting participants to our if the participant had any questions about the study (none had a substantive interest in race prior to the seating distance measure, a single-item scale question). The measure of readiness to enter a stereotype-threatening was used in this study. (For a discussion of the validity of single-item situation was then administered. The experimenter said that it was possible scales, see Robins, Hendin, & Trzesniewski, 2001.) This measure was that the researchers would run additional students through the same “test- administered after the other questionnaire measures. The scale asked par- taking study” that the other student had completed. But, the experimenter ticipants to indicate “How important is your racial background to you?” warned, the “test” in this future study would be “longer” and “more ,7  ). Prior to the experi- (1 extremely important  not at all important intense.” Should the participant wish to complete it, the experimenter ment, the scale’s reliability and validity were established. In a sample of 42 explained, the participant should leave his or her name and contact infor- undergraduates (16 Black, 26 White), the test–retest reliability (over 2–8 mation on a response form. To reduce social desirability pressures, the p .01. In another sample of 35  weeks) proved satisfactory,  (41) r .80, experimenter added that “some people are interested in doing this; some Black undergraduates, evidence of convergent validity was found in the people aren’t,” and that “if you don’t want to, that’s completely fine.” The form of a strong correlation with the Race Centrality subscale of the form provided two response options: “yes, please contact me” (with space for the participant’s name and contact information) and “no, I do not wish to participate.” Participants were instructed to seal their completed form in 2 These items had been selected on the basis of a prior investigation into a provided envelope after the experimenter departed. They were further the features associated with the stereotype of African Americans (Cohen & told to deposit the envelope in a drop-box, where a stack of envelopes Garcia, 2003). In this study, a questionnaire was administered to a sample ostensibly completed by previous participants was visible. Participants of convenience ( N  36). Respondents were asked to think about the were told that after they completed the form, they were free to leave. “cultural stereotype or image of African Americans” and were told that Upon exiting, participants were intercepted by the experimenter and “there are no objectively correct answers.” Additionally, they were assured debriefed. The participant and the confederate were reunited. The purpose both that the researchers were “not interested in your personal beliefs but of the study and the rationale for its deceptive elements were explained. in the content of the cultural stereotype or image . . . in both its negative and positive aspects” and that “there is no correlation between your Measures personal beliefs and your knowledge of the stereotype.” This method was similar to one used by Devine (1989). A list of miscellaneous activities and Self-esteem. This scale consisted of five items drawn from the State traits was then presented, and participants were asked to rate each one on Self-Esteem Scale (Heatherton & Polivy, 1991). Selected items tapped the very much fits does not fit stereotype at all )to7( a scale ranging from 1 ( psychological consequences of stigmatization (Steele et al., 2002), that is, with the stereotype ). Because it was important to identify characteristics doubts about one’s ability (Aronson & Inzlicht, 2004; Stangor, Carr, & that Black participants could be expected either to avoid or to affirm as a Kiang, 1998; Stone, Lynch, Sjomeling, & Darley, 1999) and about one’s function of a threat to their racial identity, the presented items encompassed social acceptance (Cohen & Steele, 2002; Mendoza-Denton et al., 2002). characteristics of negative, neutral, and positive valence (see also Biernat, The items were as follows: “I am confident in my abilities,” “I feel smart,” Vescio, & Green, 1996). To determine the items for subsequent use, we “I feel concerned about the impression I am making,” “I feel that others relied on the midpoint of our scale. Items with a mean rating greater than respect and admire me,” and “I am worried about what other people think 4 were classified as stereotype relevant. of me.” These were supplemented with a global measure of state self-worth

7 COHEN AND GARCIA 572 (34)  Multidimensional Inventory of Black Identity (Sellers et al., 1997), r Consistent with the results of the pilot study, State self-esteem.  p .01. Racial identification varied neither with gender nor with .79, state self-esteem was lower in the threat condition than in the ts  experimental condition ( 1). The measure was dichotomized because of F  (1, 57) p no-threat condition, 9.43,  .01. the small number of observations at the lower end of the scale that could Stereotype distancing. As expected, participants characterized be used for parameter estimation and because the fundamental trend in the themselves less stereotypically under threat than under no threat, data was more adequately captured by a simple binary split. Roughly half 4.18, (1, 57) F p  .05.  of the sample was categorized as “moderately racially identified” (6 or Racial stereotype activation. Threat led to inhibition (rather lower); the remaining participants (41%) were categorized as “highly than activation) of stereotype-relevant words relative to the no- racially identified.” threat condition, yielding a marginally significant effect, F (1, 3.70,  .059.  59) p Results Consistent with past research (Lewis & Seating distance. Sherman, 2003; Marques & Paez, 1994), participants distanced Data Analytic Strategy themselves from the specific person who had potentially discred- A one-way analysis of variance (ANOVA) was conducted, with ited their social identity. They sat farther from the Black peer in the experimental condition (threat vs. no threat) as the independent 5.87,  (1, 54) F threat condition than in the no-threat condition, variable. We used ANOVA so that main effects involving gender  p .02. could be included in the model where significant. To explore the Readiness to enter stereotype-threatening situation. Partici- effect of racial identification, we supplemented this analysis with pants had the opportunity to complete an exercise characterized as a2  2 ANOVA, using both experimental condition and partici- similar to, but harder than, the one undertaken by the same-race pants’ level of race identification as independent variables. For this peer. A chi-square test of independence—which assessed whether analysis, main effects and interactions involving gender were participants’ responses were associated with condition—yielded included in the model where significant. In no case did participant 2 no effect of condition,  N 1.  (1, 59)  gender interact with experimental condition. Because effects in- The threat manip- Perceived exposure to evaluative scrutiny. volving participant gender are tangential, they receive no further ulation had no impact on how much participants felt that their attention in Experiments 1 and 2. Because analyses involving  M verbal skills were being evaluated (no threat, 3.79; threat, racial identification are exploratory in nature, their presentation is  M 4.06; F  1). postponed to the end of the Results section. Degrees of freedom vary for different analyses due to the inclusion of significant main effects and interactions involving participant gender and due to Racial Identification: A Moderator? 3 missing values for some measures. As in the pilot study, students under collective threat had lower self-esteem regardless of their level of racial identification. The Main Effects of Collective Threat  Condition interaction was not significant Racial Identification F  1). Although highly racially identified minority students had ( Table 1 displays the means and standard deviations for each of (1, F more positive self-esteem than did their less identified peers, the dependent measures. 4.45, p  .04, both highly and moderately identified 55)  participants had lower self-esteem in the threat condition than in t  1.85, p  .07, and t (55)  the no-threat condition, (55) Table 1 p 2.51,  .02, respectively.  Dependent Measures as a Function of Threat Condition, Study 1 However, the two groups differed in how they coped with threat. Experimental Moderately identified participants avoided the stereotype in condition thought and action, whereas highly identified students did not. Racial identification interacted with threat for stereotype distanc- No threat Variable Fdf Threat (1, 52) (1, F .01; for stereotype activation,  ing, p F 12.52,  4.64,  .04; and for readiness to enter a stereotype-  p 57) State self-esteem 1, 57 9.43** 2 27.12 30.53 M  threatening situation,  N  59) (1, .01. Among  p 9.27,  SD 4.25 3.34 moderately identified students, threat (relative to no threat) led to Stereotype distancing 1, 57 4.18* M 61.75 58.00 SD 7.60 6.35 3 A few participants failed either to answer some questionnaire items or 1, 59 3.70† Stereotype activation to complete the situational readiness measure. In addition, a measure of M 2.00 2.57 seating distance was obtained for all but 4 participants; for the latter SD 1.20 1.12 5.87* 1, 54 Seating distance participants, either the experimenter had no opportunity to mark seating M 42.09 38.22 distance covertly or the participant moved his or her chair upon departure SD 4.76 5.43 and thus disturbed the placement of the mark. Finally, diagnostic proce- Readiness to enter dures revealed one outlier on the self-esteem measure (more than three stereotype-threat standard deviations from the condition mean). To prevent this outlier from 2 1 situation   1 exerting a disproportionate influence on significance tests, it was excluded Percentage assenting from analyses involving this measure. Doing otherwise—that is, using a to take test 78 69 nonparametric test or even retaining the anomalous observation in a para- p  .059. * p  .05. ** p  .01. † metric analysis—yields the same (i.e., statistically significant) result.

8 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 573 .02  p 2.47,  (52) t positive ones (e.g., enjoying basketball),  2.56, p  .02; to less less stereotypical self-ratings, (52) t (Biernat et al., 1996; Klein & Azzi, 2001). Their response pattern 2.82, .01; and to a decreased  p stereotype activation,  (57) t 2 echoes the findings of previous research regarding the ways in likelihood of assenting to take the test (55% vs. 93%),  N (1,  which high group-identified individuals cope with threats to their 34) .02. By contrast, among highly identified students,  p 5.69,  social identity, specifically by asserting their stereotypicality and  p 2.16,  (52) t threat led to more stereotypical self-ratings, .04; solidarity with the group (Spears et al., 1997; see also Branscombe  to no inhibition of the stereotype ( 1); and to a marginally t et al., 1999; Tajfel & Turner, 1986). increased likelihood of assenting to take the test (92% vs. 62%), 2 Under threat, moderately racially identified participants also   p 3.11,  N 25) .08. Although the interaction for seating (1,  proved less willing to expose themselves to an intellectually eval- distance was not significant ( F  1), only moderately identified uative situation. This response is analogous to what Steele and (1, F participants sat farther from the same-race peer under threat, disidentification and what social identity Aronson (1995) call p 3.98,  31) (1, F .055. Highly identified participants did not,  theorists call opting out (Lalonde & Silverman, 1994), wherein one  1.63, p  .22. 20) defensively devalues or altogether avoids the domain of threat. While this coping mechanism could be effective in some ways, it Discussion could discourage effort and challenge seeking that could otherwise benefit performance (Aronson, 2002; Cohen et al., 1999; Like the pilot study, Experiment 1 demonstrated that collective Mendoza-Denton et al., 2002). By contrast, highly identified stu- threat is linked to lower self-esteem. Black students had lower dents under threat were marginally more likely to expose them- self-esteem when they saw a same-race peer in an intellectually selves to an evaluative situation, in spite of the stereotype threat evaluative situation rather than an intellectually non-evaluative inherent in doing so and in spite of its characterization as more one. Although participants were not being personally evaluated in demanding than the situation faced by the same-race peer. These the threat condition—and did not feel that their abilities were participants might have viewed the situation as an opportunity to under evaluation more in this condition than in the no-threat defend the image of their group, by performing well and thus condition—they responded as though their personal worth were challenging the negative stereotype that may have been put into under assault. play by their fellow group member. Under collective threat, participants avoided the stereotype. Our results with respect to racial identification, while interest- They distanced themselves from the stereotypical image of their ing, are secondary. Our primary result is that the potentially social group (Pronin et al., 2004; Spears et al., 1997; Steele & stereotype-confirming behavior of ingroup members constitutes a Aronson, 1995). They dissociated from the person who had po- significant source of concern to stereotyped students in an aca- tentially discredited their group by sitting farther away (see also demic context and that this concern threatens self-worth even Marques & Paez, 1994). In addition, they inhibited thoughts about when people experience no evaluative threat themselves. Each of the stereotype (Iserman et al., 2004). the two previous studies supports this claim. As in the pilot study, collective threat led to lower self-worth regardless of students’ level of racial identification. Of course, this may have occurred because of the relatively high level of ethnic Experiment 2 group identification of all our minority participants. Nevertheless, it also seems possible that regardless of whether one personally A second experiment was conducted to replicate the effect of identifies with one’s group, one could still be threatened by the collective threat on self-esteem and to address issues not consid- prospect of being so identified by others (Cohen & Steele, 2002; ered in the previous two studies. One issue concerns whether Cohen et al., 1999; Klein & Azzi, 2001; Steele et al., 2002; shared group membership constitutes a necessary condition for the Vorauer et al., 1998). effect of collective threat on self-worth. We assert that the collec- As in the pilot study, an interesting—albeit exploratory—qual- tive threat requires the two individuals to have a common group ification emerged with respect to the influence of racial identifi- membership. Consequently, Black students should not experience cation on how people coped with collective threat. Less racially collective threat when aware that a White peer might perform identified students used a strategy we would characterize as social poorly on an intellectually evaluative task. Another issue concerns identity avoidance, that is, one consistent with a flight response. the role of the stereotype relevance of the task in collective threat. They withdrew from the stereotype both in thought and in action. We maintain that collective threat requires a potentially poor They described themselves less stereotypically, disassociated from performance by an ingroup member on a stereotype-relevant task. the same-race peer, and inhibited thoughts about the stereotype. Black students should not experience collective threat when they Their response echoes the way in which low group-identified are aware that a Black peer could do poorly on a stereotype- individuals have been found to cope with a threatened social irrelevant task. (Our no-threat condition in study 1 did not explic- identity. They distinguish themselves from their group (Spears et itly raise the possibility that the same-race peer might perform al., 1997) and disassociate from and even denigrate other group poorly.) members (Ellemers et al., 2004; Ethier & Deaux, 1994). By To address these two issues, Experiment 2 featured three exper- contrast, highly racially identified participants showed none of imental conditions: one collective threat condition and two no- these responses. Instead, they used a strategy we would character- threat control conditions. In the threat condition, Black students ize as that is, one consistent with a fight social identity affirmation, were led to believe that a same-race peer worked on a stereotype- response. Under threat, these participants rated themselves more relevant task, specifically, an intelligence test. In the first no-threat stereotypically. They did so, it is worth noting, not for negative control condition, Black students were led to believe that a same-  traits (e.g., aggressive; t 1) but selectively for neutral and race peer worked on a stereotype-irrelevant task—an art test. In the

9 COHEN AND GARCIA 574 IQ test.” The experimenter told the participant that after their partner second no-threat control condition, Black students were led to finished the test, the two of them would interact. Participants were in- believe that a White peer worked on an intelligence test. Once formed that prior to this interaction, they would complete some background again, our primary measure of threat was self-esteem. We pre- questionnaires. The procedure in the Jamal/art test condition was identical dicted that Black participants would have lower self-esteem in the to that in the Jamal/IQ test condition with one exception: The experimenter threat condition than in the two no-threat conditions. We also did not mention intellectual ability and instead said that the partner (“Ja- tested whether collective threat occurs, as our analysis predicts, in mal”) was completing a study that entailed “assessing a person’s drawing response to a potentially poor rather than strong performance on abilities” using “art puzzles.” the part of the ingroup member. To address this question, we asked Because the peer was apparently a fellow Yale student, he might be participants to estimate the ability of the peer. expected to perform well on the IQ test. Accordingly, we introduced the Continuing the exploratory aspect of our research, we took up possibility that he might perform poorly (and thereby confirm, in the Jamal/IQ test condition, the racial stereotype). To do so, we staged the three other objectives in Experiment 2. The first objective involved following scene. In all three conditions, the experimenter excused herself testing the robustness of collective threat. To do so, we made the from the participant’s lab room. She then knocked on the door of another collective threat in Experiment 2 less vivid and more remote than lab room where the other participant supposedly was working. Opening the it was in Experiment 1. In contrast to Experiment 1, participants in door, the experimenter called into the room (within earshot of the partic- Experiment 2 never met the same-race peer. They did not hear the ipant), “Hey [Jamal/Jeffrey], are you finishing up?” Inside the room, the test described as difficult. They overheard the peer express appre- experimenter then played a digital recording of the student’s response. (The hension about the test only once. Also in contrast to Experiment 1, recording had been engineered to sound authentic. A Black student’s voice participants were not made explicitly aware that the peer had had been used in the Jamal conditions, a White student’s voice in the entered a highly evaluative situation where explicit feedback iden- Jeffrey condition.) The response was: “Yeah, I guess so. To be honest with tifying his or her intellectual limitations would be given. Instead, you, I’m so bad at [these standardized tests/drawing].” The experimenter replied, “Oh yeah? Hmm . . . . Just finish up, and we’ll move on to the next participants were led to believe only that an intelligence test had section.” As noted previously, in the post-debriefing interview, 4 partici- been administered to a same-race peer. Would even this subtle pants (2 in the Jamal/IQ test condition and 1 in each of the remaining two threat undermine self-worth? conditions) said that they did not hear the recording (and were thus A second objective involved testing the effect of collective discarded prior to analysis). threat on test performance. Among less racially identified minority Participants then completed the questionnaires assessing state self- students, collective threat predicted lower GPA in the pilot study esteem, stereotype distancing, racial stereotype activation, and perceived and, in Experiment 1, caused more avoidance of a test. In light of exposure to evaluative scrutiny. After completing these scales, the measure these results, it seemed desirable to assess actual intellectual of readiness to enter a stereotype-threatening situation was administered in performance. the same manner described in Experiment 1. Finally, as in our previous studies, the role of racial identifica- Participants were then given up to 18 min to complete a verbal test. It is important to note that participants had no prior knowledge that they would tion in collective threat was explored. take this test. Thus, their responses to the previous measures could not be contaminated by the expectation of taking a test themselves. Next, partic- Method ipants completed the measures assessing racial identification and the per- ceived ability of the peer. Finally, participants were probed for suspicion, Participants and Design thoroughly debriefed about the true purpose of the study, and thanked for their participation. Participants were 63 Black undergraduates at Yale University. They took part in the study in exchange for either $8 or course credit. Paid participants were identified using a registrar’s list; students receiving course credit were Measures identified with a participant pool roster. Two students suspected that the purpose of the study concerned responses to racial stereotypes (one was in Unless otherwise noted below, all measures were identical to those used the Jamal/IQ test condition, the other in the Jeffrey/IQ test condition). in Experiment 1. No measure of seating distance was obtained, as this Another four students (distributed roughly equally across conditions) failed study did not include a confederate. to hear an important component of the experimental manipulation (noted We used the complete Performance and Social State Self- Self-esteem. below). Accordingly, data from these six students were excluded from Esteem subscales rather than only the five items selected from those scales analyses. The remaining sample comprised 57 participants (31 women, 26 in Experiment 1 (Heatherton & Polivy, 1991). These were supplemented men) randomly assigned to one of three experimental conditions: one with the same global state self-esteem item used in Experiment 1. All but collective threat condition and two no-threat control conditions. three items loaded on the first unrotated factor (with loadings  .40). The three items were excluded from the composite (including them does not alter the statistical significance of any reported result). After reverse coding Procedure  where appropriate, we summed the 12 items into a composite (  .79), Each participant was met by an Asian American female experimenter, with higher values signifying more positive self-esteem. One item asked participants who stated that the purpose of the study was to investigate “how people Perceived exposure to evaluative scrutiny. form impressions of one another on the basis of brief interactions.” The to indicate how much they thought their “intellectual abilities” were being same experimenter was used for all participants. She explained that the evaluated in the study; a second item asked how much they thought their ,7 participant’s partner was in the room across the hall. To signify the race of not at all  “general abilities” were being evaluated in the study (1  very much ). the partner, the experimenter referred to him as either Jamal (a common The verbal test consisted of 16 multiple-choice items Black name) or Jeffrey (a common White name). In the Jamal/IQ test and Verbal test. drawn from various practice verbal tests for the Graduate Record Exami- Jeffrey/IQ test conditions, the experimenter explained that the partner was nation (Educational Testing Service). Consistent with previous research completing another study that entailed “assessing a person’s intellectual (e.g., Shih et al., 1999; see also Steele & Aronson, 1995), our primary ability” and that the partner was “taking a shorter version of the standard

10 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 575 measure of performance was test accuracy—the number of problems 62.29) and the Jeffrey/IQ test condition (  63.12), B  .23, M correct divided by the number of problems attempted. Accuracy controls (50)  2.60, p  .02. The two no-threat conditions did not t for individual differences (e.g., with regard to knowledge of subject matter) t  1). differ ( that could affect the number of problems students attempt but not the As noted previously, the difference between the threat condition overall quality of their performance on the problems that they are able to and the two no-threat conditions should be largest among partic- undertake. ipants who had less confidence in the peer’s ability and who thus Participants estimated how much ability Perceived ability of the peer. had reason to believe that he would confirm, in the Jamal/IQ test the peer had on the task he was ostensibly completing (1  ,7 very little  condition, the negative stereotype. Although the peer had “said” ). a great deal that he was not good at the task, students in the study attended an The racial identification measure was expanded Racial identification. elite academic institution, and they would assume that any peer beyond the single-item scale used in Experiment 1. The relevant scale included the original item used in Experiment 1, the four-item Importance they encountered had the intellectual capacity to attend such a of Identity subscale of the Collective Self-Esteem Inventory (e.g., “The school. Also, as noted previously, both the evaluative nature of the racial/ethnic group I belong to is an important reflection of who I am”; situation and the peer’s expressed level of distress were down- Luhtanen & Crocker, 1992), and one item from the Race Centrality played in the threat condition used in this study. Perhaps as a result subscale of the Multidimensional Inventory of Black Identity (e.g., “I have of these factors, participants gave a high mean estimate of the a strong sense of belonging to people of African descent”; Sellers et al., same-race peer’s intellectual ability ( 0.87, on the  SD 5.09,  M 1997). Responses were made on appropriately labeled 7-point scales. All 7-point scale). Given this high estimate, the previous analysis six items had loadings greater than .50 on the first unrotated factor. provides only a weak test of our hypothesis. Accordingly, they were summed into a single composite ( .83), with   Indeed, the estimated ability of the peer interacted with threat higher values representing greater racial identification. Racial identifica-  p 3.76,  (43) t  .01. The B (i.e., the first contrast code), .38, M  26.28) and varied neither with experimen- tion was moderately high ( tal condition nor with participant gender ( t s  1). relevant means are displayed in Figure 3. Among participants who gave a low estimate of the peer’s ability (one standard deviation below the mean), self-esteem was lower in the threat condition Results t p 4.95,  (43)  .71, B than in the no-threat conditions,  Data Analytic Strategy .01. By contrast, among participants who gave a high estimate of the peer’s ability (one standard deviation above the mean), self- We used regression because our theoretically specified moder-  esteem was high regardless of condition ( 1). The results ts ator (i.e., the peer’s estimated ability) was continuous. Two or- suggest that collective threat requires uncertainty about the in- thogonal contrast codes tested the effects of experimental condi- group member’s ability. It is noteworthy that the minimal condi- tion, the first for the theoretically important contrast between the tion is uncertainty, not doubt. Participants defined as low along the  threat condition and the two no-threat conditions (Jamal/IQ test measure assessing the peer’s estimated ability fell not at the low 1) and the second for  1, Jeffrey/IQ test 2, Jamal/art test   end of the scale, but near its midpoint.  the contrast between the two no-threat conditions (Jamal/IQ test 1, Jeffrey/IQ test  0, Jamal/art test 1). We expected the  latter contrast to yield a null result. Stereotype Distancing We included main effects and two-way interactions involving No main effect of condition was found ( ts  1). However, threat participant gender where significant; the small sample size pre-  B again interacted with the estimated ability of the peer, .33, cluded testing three-way interactions. In no case did gender sig- .01. Among participants who gave a low t (46)  2.92, p  nificantly interact with threat (i.e., the first contrast code). Only estimate of the peer’s ability, self-ratings were less stereotypical in after testing the effects of the two contrast codes did we assess the threat condition ( M  50.39) than in the two no-threat condi- main effects and interactions involving the estimated ability of the M  62.77; for the Jeffrey/IQ test, tions (for the Jamal art/test, peer. As is appropriate, in testing a variable’s interaction with p M  .43, t (46)  2.64,   .01. By contrast, 56.32), B condition, we computed that variable’s interaction with each of the among participants who gave a high estimate of the peer’s ability, two contrast codes and retained both interaction terms if either one self-ratings were slightly more stereotypical in the threat condition proved significant. As expected, no interaction effects were found 64.00) than in the two no-threat conditions (  M ( 58.05 and  Ms s involving the second contrast code ( t .09).  s p 1.8,  t B  59.73, respectively), .07. While this  p 1.86,  (46) .24, Because the measures used different scales, we standardized all latter result should be regarded tentatively due to its marginal continuous independent and dependent variable measures before nature, it suggests a social identity affirmation response among computing regression coefficients. To supplement the presented participants who thought that the peer would refute the stereotype coefficients, we report means for the dependent measures in their rather than reinforce it. It is important to note that this affirmation original scale. Four participants declined to estimate the peer’s response occurred for neutral and positive stereotypical traits, B  ability (distributed roughly equally across conditions). Degrees of t 1).  t .03, not negative ones ( .28, (46)  2.25, p  freedom varied slightly because of missing values and because of the inclusion of significant main effects and interactions involving participant gender. Racial Stereotype Activation Stereotype activation was higher in the Jamal/IQ test condition State Self-Esteem 2.05) and the  M 2.44) than in the Jamal/art test condition ( (  M  (54) M  1.67), B  .19, t  Jeffrey/IQ test condition ( 2.13, p As expected, state self-esteem was lower in the Jamal/IQ test .04. The two no-threat conditions did not differ ( t  1.2, p  .26). M condition ( M  56.47) than in the Jamal/art test condition ( 

11 COHEN AND GARCIA 576 8 0 P y eve s l dH igh A erc i bi a Pr e e it eP erc i e eve s a Pr dL ow A bility 0 7 6 0 0 5 4 0 3 0 m ey Jf T l t l I Q t a J e s T I ,Q s fr , e et rt ,A T a a sa e Jm Figure 3. Mean levels of state self-esteem as a function of experimental condition and estimated ability of peer: Experiment 2. Higher values signify more positive self-esteem. only a trend among participants who gave a low estimate of the Although the threat condition appears to have increased the acces- peer’s ability, with those in the threat condition assenting to take sibility of the stereotype, some of this effect may be due to the 2  the test more often than those in the no-threat conditions, perceived presence of another Black person in the same study.  (1)  Consistent with this claim, mean activation in the Jamal/art test  3.26, p .08. Given the marginal nature of the effect, however, it condition fell halfway between the other conditions. should be viewed tentatively. Threat again interacted with the estimated ability of the peer,  (47) .28, 2.38, t B  p  .03. Among participants who gave a Perceived Exposure to Evaluative Scrutiny high estimate of the peer’s ability, activation was greater in the Participants reported that both their intellectual abilities and B  .41, t (47)  threat condition than in the no-threat conditions, their general abilities were being evaluated to the same extent in .01. By contrast, among participants who gave a low 3.15, p  the Jamal/IQ test condition (Ms  3.41, 4.39), the Jamal/art test t estimate of the peer’s ability, there was no effect of condition (  condition ( M s 3.61, 4.19), and the Jeffrey/IQ test condition  1). It seems possible that participants who worried that the ingroup s 1). This was the case regardless of the  t s (  3.40, 4.29; M member might perform poorly attempted to inhibit any activation 1).  t estimated ability of the peer ( that the joint presence of a Black peer and an intellectually eval- uative test would otherwise have elicited (Kunda, Davies, Adams, Racial Identification: A Moderator? & Spencer, 2002; Steele & Aronson, 1995). As in the previous two studies, collective threat had a greater Test Performance impact on participants who were low rather than high in racial identification. In contrast to our previous studies, however, this  M Test accuracy was lower in the Jamal/IQ test condition ( occurred only for our primary measure of threat, self-esteem. We .42) than in the Jamal/art test condition ( M  0.51) and the tested both the main effect of racial identification and its two-way  2.29, Jeffrey/IQ test condition ( M  0.57), B  .20, t (53) interactions with condition. (Because of limited sample size, we .03. For total number of problems solved correctly, the same  p excluded main effects and interactions involving the peer’s esti- significant difference between threat and no-threat conditions mated ability. It is important to note that the latter measure did not p ( .04), and nonsignificant difference between the two no-threat  p  correlate with racial identification, r s .58.) There was  s .16,  conditions ( .17), were found. The effect of condition did not p a main effect of racial identification. As in Experiment 1, partic- 1).  t vary with the estimated ability of the peer ( ipants had more positive self-esteem if they were high rather than .24,  (46)  2.07, p  B low in racial identification, .05. There t Readiness to Enter Stereotype-Threatening Situation .20,  B Threat interaction,  was also a Racial Identification  2.56, p  .02. Among less identified participants, self- t (46) Logistic regression identified no main effect of condition ( s p  M esteem was lower in the threat condition (  52.14) than in the .68). Threat again interacted with the estimated ability of the peer, 2 two no-threat conditions (Jamal/art test M  60.28; Jeffrey/IQ test   (1)  .05. However, follow-up analyses revealed 4.39,  p

12 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 577 themselves from the stereotypical image of their race, or otherwise B  .41, t (46)  3.76, p  .01. By contrast, M 66.22),  severing self-evaluation from the domains of threat (Major, Spen- among more identified participants, self-esteem was high both in cer, Schmader, Wolfe, & Crocker, 1998; Steele et al., 2002), M the threat condition ( 63.72) and in the two no-threat condi-  highly identified minorities may instead draw on their racial iden- M 66.08, 61.82, respectively).  tions ( s tity as a compensatory source of self-worth (see also Sherman & Cohen, 2002). Discussion However, the most important result of Experiment 2 is its replication of the negative effect of collective threat on self-esteem Like the two previous studies, Experiment 2 demonstrated that and its validation of the hypothesized parameters of this effect. An experiencing collective threat is linked to lower self-esteem. This important remaining question concerns the generality of the occurred even though—in contrast to Experiment 1—the threat phenomenon. was subtle rather than vivid. Students neither met nor saw the same-race peer. They received no explicit information that the peer had entered a difficult and highly evaluative situation aimed partly Experiment 3 at identifying limitations in intellectual ability. Additionally, as The studies presented thus far focus exclusively on ethnic mi- expected, the negative effect of threat on self-worth occurred only nority students. It is possible that factors unique to this group for participants who felt uncertain of the peer’s ability and who account for the results. To address this limitation, we conducted a thus had reason to think he might confirm the alleged intellectual conceptual replication of Experiment 1, this time focusing on inferiority of their race. women. Because women contend with negative stereotypes about Experiment 2 supported two other, theoretically derived condi- the quantitative ability of their gender group (Spencer et al., 1999), tions of collective threat. First, collective threat requires a shared they should experience collective threat if they are aware that a group membership. Black students did not experience lower self- female peer could perform poorly on a math test and thus lend esteem when the peer taking the IQ test was White rather than credence to the gender stereotype. Black. Second, collective threat requires the fellow group member to be at risk of confirming the stereotype. Black students did not Method suffer lower self-esteem when the Black peer took a test irrelevant to the stereotype. Participants and Design A secondary finding concerned the effect of collective threat on the way in which people oriented to their social identity. As in Thirty-two undergraduate women who were enrolled in either a math or an engineering course took part in the study in exchange for payment. We Experiment 1, Black students under collective threat distanced recruited math and engineering students on the basis of the assumption that themselves from the stereotypical qualities of their racial group people who care about a performance domain will want to maintain a (Steele & Aronson, 1995; see also Pronin et al., 2004; Spears et al., positive representation of their group in such a setting (Steele et al., 2002). 1997). Consistent with our theoretical analysis, they did so only if Participants were randomly assigned to one of two conditions. In the they felt uncertain of the same-race peer’s intellectual ability and collective threat condition, they observed another female student about to thus had grounds to believe that he might confirm the negative complete a test that had been described as diagnostic of math ability. In the stereotype. no-threat condition, they observed this student about to complete math Collective threat also undermined participants’ performance on puzzles that had been described as non-diagnostic of math ability. a standardized test that they were later, unexpectedly, asked to complete. This result offers some evidence that the association Procedure between collective threat and low GPA observed in our pilot study The procedure was almost identical to that of Experiment 1, with one issues, in part, from the causal impact of collective threat on important change. References to a “verbal test” in the threat condition were performance. changed to references to a “math ability test,” and references to “verbal A final secondary result concerns the moderating influence of puzzles” in the no-threat condition were changed to references to “math racial identification. Once again, less identified students proved puzzles.” Manipulation checks at the end of the study confirmed that more responsive to collective threat. Why this effect occurred for participants accurately overheard the instructions in each condition. One of self-esteem in the present study, and only for the other secondary three male experimenters conducted the study; one of seven female stu- measures in the two previous studies, is unclear. Differences in the dents played the confederate. intensity of threat may have played a role. When the threat lacks Each of the conceptual dependent variables assessed in Experiment 1 intensity, as in Experiment 2, racial identification may reduce the was assessed in Experiment 3. (Because of time constraints, the measure of likelihood of perceiving the situation as threatening to self-worth test performance used in Experiment 2 was dropped.) The measures of stereotype distancing and stereotype activation referred to stereotypes (see Ethier & Deaux, 1994). When the threat is more intense—on about women rather than about Black Americans. The stereotype distanc- account of either its chronic nature (as in the pilot study) or its ing measure used a validated scale developed by Pronin et al. (2004) that greater immediacy and plausibility (as in Experiment 1)—racial asks respondents to rate how much they characterize themselves as having identification may fail to protect self-worth from the knowledge various traits, enjoying various activities, and entertaining various future that one’s group has, in fact, been negatively represented. It may plans. Some of the traits (e.g., gossipy, emotional), activities (e.g., using instead act as a source of social support and self-esteem and thus makeup, flirting), and plans (e.g., thinking about leaving work to have facilitate recovery from threat (see Branscombe et al., 1999). In children) are associated with stereotypes of women who lack math ability. such cases, racial identification may ameliorate the normal esteem- As Pronin et al. (2004) recommends, participants rated both the self- protective adaptations to threat. Whereas low identified minorities descriptiveness of each item and its personal importance, using separate, may protect self-worth by disidentifying from school, distancing appropriately labeled 7-point scales. The scale contains 14 questions and

13 COHEN AND GARCIA 578 encompasses negative items (e.g., “gossipy”) and items more neutral or Table 2 positive in valence (e.g., “leaving work to raise children”). The gender Dependent Measures as a Function of Threat Condition, Study 3 activation measure again featured a word-fragment completion task. The 12 critical words were associated with gender and gender stereotypes (e.g., Experimental condition “lady,” “weak,” “pink”). Participants were also asked how much they thought their math abilities and their general abilities were being evaluated tdf No threat Threat Variable 7 in the study using two separate scales (1 ). very much   not at all, Gender identification was assessed with the item, “How important is your 2.71** State self-esteem  30 extremely not at all, 7  gender to you?” (1  ). Gender identification was M 24.75 29.69 1.3,  4.84) and did not vary with condition (  M moderately high ( t 6.30 3.65 SD .23). p  Math self-efficacy  3.02** 30 Two new, conceptually relevant measures were introduced in Experi- 4.81 5.94 M 1.17 ment 3. The first was math self-efficacy. A strong form of our argument is SD 0.93 29 Stereotype distancing 2.62* that an “I am us” mindset leads people to experience the threat of an M 48.80 58.81 ingroup member as if it were their own. A major consequence of stereotype 7.66 13.06 SD threat is a momentary drop in self-efficacy (Aronson & Inzlicht, 2004; Gender/stereotype Cohen et al., 1999; Stangor et al., 1998; Stone et al., 1999). A similar drop activation 2.34* 30 in self-efficacy may occur as a result of collective threat. To assess 6.75 5.31 M self-efficacy, we asked participants to indicate their level of agreement SD 1.35 2.05  strongly disagree, with the statement, “I am very good at math” (1  7 30 9.23** Seating distance ). The other new measure assessed imitation behavior. The strongly agree 11.00 9.69 M  goal to affiliate leads to mimicry (Lakin & Chartrand, 2003). To the extent SD 5.63 6.98 2 1  Imitation behavior 12.70**  that participants want to distance themselves from the same-gender peer Percentage taking under threat—as threatened Black participants did by sitting farther from cookie 13% 75% the same-race peer in Experiment 1—they may imitate her less. At the end Readiness to enter of the study, the experimenter presented the participant and confederate stereotype-threat with a plate of cookies. The confederate ate one. The dependent measure 2 1  situation   3.29† was whether participants ate one too. Seating distance from the confederate Percentage assenting was again measured. At the beginning of the study, the chair was posi- 75% to take test 50% tioned approximately 6 feet from the confederate. The dependent measure Note. For seating distance, negative numbers signify movement toward was the deviation (in inches) from this baseline. Negative values represent the same-gender peer, and positive numbers signify movement away from movement toward the confederate, positive values represent movement her. away.  .01. † p .05. **  p .07. * p  Results Participants had lower math efficacy in the Math self-efficacy. Data Analytic Strategy  threat condition than in the no-threat condition, 3.02, (30) t p  .01. tests. With t We conducted our analyses using a series of simple Like highly racially identified Black Stereotype distancing. one exception, no effects were found either for experimenter or for students in Experiment 2, women characterized themselves more confederate; accordingly, analyses collapse across these variables.  p 2.62, stereotypically under threat than under no threat, (29) t  The exception involved the measure of readiness to enter a .02. As found in the previous studies, this social identity affirma- stereotype-threatening situation, for which there was variation due tion response—when it occurred—involved endorsement of neu- to the experimenter. Accordingly, analysis of this measure uses a 2.87,  (29) t tral and positive qualities, .01, not negative ones  p logistic regression, with the experimenter effect controlled. One  ( t 1.3, p  .39). participant did not complete all the stereotype distancing items; Participants generated more Gender/stereotype activation. analysis of this measure thus involves one fewer degree of stereotype-relevant words in the threat condition than in the no- freedom. p  .03. threat condition, (30)  2.34, t Seating distance. As in Experiment 1, participants sat farther away from the confederate in the threat condition than in the Main Effects of Collective Threat no-threat condition, t (30)  9.23, p  .01. Imitation behavior. More participants took a cookie in the Table 2 displays the means and standard deviations for each of 2 no-threat condition than in the threat condition, the dependent measures.   N (1,  32) Replicating the result of the previous stud- State self-esteem. 12.70, p  .01. ies, we found that participants had lower state self-esteem in the The logis- Readiness to enter stereotype-threatening situation. t (30)  collective threat condition than in the no-threat condition, tic regression indicated that, compared with participants in the p 2.71, .01. (Variance was also greater in the threat condition   no-threat condition, participants in the threat condition were mar-  (1, 30) F than in the no-threat condition, .05. However,  p 4.09, ginally less likely to agree to complete the same task as the 2 a nonparametric Mann–Whitney U test—robust against violations  confederate,  3.29, .07. p  (1)  of equal variance—identified the same significant condition effect Participants re- Perceived exposure to evaluative scrutiny. ( .02.)  p 2.46,  z ported that both their math abilities and their general abilities were

14 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 579 s  being evaluated to the same extent in the threat condition ( M General Discussion s  2.25, 3.56) as in the no-threat condition ( M 1.94, 3.50; t s  1). When one belongs to a negatively stereotyped group, the aware- ness that how one is personally defined will be determined, in part, Gender Identification: A Moderator? by how one’s group is defined may give rise to an “I am us” mindset. As a consequence, not only do people worry that their Gender identification did not moderate the effect of collective own behavior could be used to lend credence to a negative stereo- ps  Fs .11). Consistent with the previous 2.6, all threat (all  type about their group (Steele et al., 2002), they also worry that the studies, however, any trends indicated that collective threat had a behavior of fellow group members could be used in this way. As larger effect on less identified women. The absence of a significant Lewin (1948) observed, “[S]ensitivity in regard to the conduct of interaction effect may be due to insufficient power. The sample other members of a group is but an expression of a fundamental size was roughly half of the ones used in each of the two previous fact of group life, namely, the interdependence of fate” (p. 190). experiments. While results with respect to gender identification are In our field study, minority students, unsurprisingly, reported thus tentative, collective threat had (as in the previous studies) a more concern than did White students that their poor performances clear negative impact on feelings of personal worth and associated could reflect on their race. However, they reported even greater outcomes. concern that the poor performances of fellow minorities could reflect on their race. This study also provided evidence that col- lective threat was indeed threatening. It predicted lower self- Discussion esteem even after controlling for self-reported stereotype threat, fear of being stereotyped, and perceived discrimination. Experiment 3 generalized the negative effect of collective threat Our experimental studies provide direct evidence that collective on self-esteem to women in math. As in Experiment 1, participants threat harms self-esteem. In Experiment 1, Black students had also distanced themselves from the ingroup member who had lower self-esteem when they observed a same-race peer under potentially reinforced the negative stereotype about their group by stereotype threat, that is, one who was about to complete a test sitting farther away. They were also less likely to indulge in the diagnostic of intellectual ability (Steele & Aronson, 1995). Exper- same snack as the same-gender peer. Experiment 3 further found iment 2 replicated this effect and supported two hypothesized that one of the cardinal symptoms of stereotype threat—low self- conditions of collective threat—that the two individuals share a efficacy (Aronson & Inzlicht, 2004; Cohen et al., 1999; Stangor et group identity and that one of them be confronted with a al., 1998; Stone et al., 1999)—also befell participants under col- stereotype-relevant situation rather than a stereotype-irrelevant lective threat. This was the case even though participants were one. Consistent with our conceptual analyses, Experiment 2 also aware that their math abilities were not being evaluated. Partici- indicated that a necessary condition of collective threat is uncer- pants seem to have adopted an “I am us” mindset. It was as if they tainty as to whether the ingroup member has the ability to do well had psychologically entered the ingroup member’s situation. and the accompanying supposition that he or she might confirm the With respect to how collective threat affected stereotypic stereotype. Experiment 3 generalized the effect to women in math. thoughts and self-characterizations, women responded in a manner The four studies strongly support our primary claim. Just as it is similar to that of highly racially identified Black students. They distressing to confront a stereotype-threatening situation oneself, it embraced the non-negative stereotypical aspects of their group is also distressing to observe a fellow group member face the same identity more under threat than under no threat. Additionally, situation. under threat, they showed no suppression of thoughts about their Collective threat occurred without our having introduced an group. Instead, they exhibited an increase in gender activation. explicit intergroup comparison (cf. Blanton, Crocker, & Miller, While women, on the whole, exhibited a social identity affir- 2000; Brewer & Weber, 1994). Participants were not led to believe mation response, only the most highly racially identified minority that the performances of ingroup members would be compared students did so. One explanation for this difference pertains to a with those of outgroup members. They received no feedback that difference between the stereotype targeted at women and the one the ingroup member had, in fact, performed poorly. Most impor- targeted at Black Americans. The stereotype targeted at women is tant, prior to completing the critical dependent measures, partici- specific (e.g., alleging low math ability), and it contains some pants in the present experiments neither undertook an intellectual positive components (e.g., women’s interpersonal and relational task nor expected to do so. This aspect of our experiments elimi- qualities are positively stereotyped; Glick & Fiske, 2001; Prentice nates confounds that could otherwise explain the effect of our & Carranza, 2002). By contrast, the stereotype targeted at Black manipulations on self-esteem, such as concern that one’s own Americans impugns their general intelligence (Devine, 1989; performance could confirm a negative stereotype (Steele et al., Steele et al., 2002) and even their humanity (Fredrickson, 2002). 2002) and discomfort over outperforming a fellow group member The risks of being viewed as a typical group member may thus be (Exline & Lobel, 1999). greater for Black Americans than for women. To express solidarity Seeing a fellow group member in a potentially stereotype- with one’s social identity may thus require a greater level of group threatening situation may prove, in some respects, more troubling identification for Black Americans than for women. This admit- than being in such a situation oneself. Collective threat is a chronic tedly speculative analysis calls for future research. What is clear in evaluative threat—as it involves all other individuals in one’s all three experiments is that collective threat not only harms group and evaluative situations in which one is not directly in- self-worth but changes people’s thoughts and orientation toward volved—rather than an acute evaluative threat involving only one’s own performance. Moreover, minority group members are their group.

15 COHEN AND GARCIA 580 ance of an intellectually evaluative test (Experiment 1), or expe- aware, as we all are, that some people do poorly on difficult intellectual tasks and that included among these are members of rienced a larger decrement in GPA (pilot study). These results are their group. Additionally, because people tend to believe that their consistent with past research showing that racial identification can own abilities are superior to those of their peers (Dunning et al., protect minority students either from the pain of stigmatization 1989), they may underestimate the possibility that they personally (Branscombe et al., 1999; Ethier & Deaux, 1994) or from its could do poorly on an intellectual task but exaggerate the possi- adverse academic consequences (Oyserman et al., 2003; Wong et bility that other people in their group could. People also tend to al., 2003). On the other hand, group identification has been found rationalize their own poor performances, thereby casting their to increase vulnerability to stereotype threat (Schmader, 2002). failures as immaterial to assessing their ability (Greenwald, 1980). Clearly, further inquiry is needed to identify the conditions under However, they do not interpret the poor performances of others as which group identification ameliorates and exacerbates threat. charitably as their own. As a result, people may expect the failures Our research raises an important ambiguity—whether people of fellow group members to be viewed as more reflective of a lack under collective threat feel concerned only with the image of their of ability, and thus as more validating of the negative stereotype, group or whether, in addition, they feel concerned with the impact than their own. Indeed, whereas we found that Black students’ that image has on the way they personally are viewed. Of course, self-esteem suffered as a result of exposure to a same-race peer people have many motives for maintaining a positive social iden- who might perform poorly on an intellectual test, other researchers tity, including collective and self-interests. have found that Black students’ self-esteem can prove surprisingly However, the entwined nature of these two concerns constitutes immune to their own poor performances on such tests (Major et al., a major point of our research. It is, we think, fundamental to the 1998). experience of stigmatization. This seems especially true for ethnic Beyond a negative impact on self-esteem, three other conse- minority students. Because the racial stereotype is widely known quences of collective threat were apparent. First, under some and because it impugns a universally valued trait, it would be circumstances, and for some people, collective threat can lead to difficult for many, if not most, ethnic minorities to feel that their lower achievement. It was associated with a drop in GPA (pilot personal reputations were ever fully independent of the reputation study) and with avoidance of an intellectual challenge (Experiment of their race. For good and for ill, our stake in our social identities 1) for less racially identified minority students. Collective threat makes the fates of fellow group members intimately linked to our also led to lower performance on a standardized test of verbal own. ability for minority students both low and high in racial identifi- cation (Experiment 2). Moreover, collective threat led to lower References self-efficacy (Experiment 3), which is a major cause of under- achievement (Bandura, 1997). A second consequence of collective Aronson, J. (2002). Stereotype threat: Contending and coping with unnerv- threat is dissociation from the ingroup member who had poten- Improving academic achievement: ing expectations. In J. Aronson (Ed.), tially reinforced the negative stereotype about one’s group—for (pp. 303–328). San Diego, Impact of psychological factors on education example, by sitting farther away from that person (Experiments 1 CA: Academic Press. and 3) or by declining to indulge in the same snack as the fellow Aronson, J., & Inzlicht, M. (2004). The ups and downs of attributional group member (Experiment 3). These results are consistent with ambiguity: Stereotype vulnerability and the academic self-knowledge of past research showing that people disassociate from and even 829– Psychological Science, 15, African American college students. denigrate those whose behavior could be seen as discrediting their 836. Aronson, J., Lustina, M. J., Good, C., Keough, K., Steele, C. M., & Brown, group (Lewis & Sherman, 2003; Marques & Paez, 1994). J. (1999). When White men can’t do math: Necessary and sufficient Third, collective threat gave rise to two distinct social identity factors in stereotype threat. Journal of Experimental Social Psychology, management strategies. One strategy involves downplaying the 29–46. 35, relevance of one’s social identity to the situation at hand. To do so, New York: Self-efficacy: The exercise of control. Bandura, A. (1997). people distance themselves from the stereotypical image of their Freeman. group (Pronin et al., 2004; Steele & Aronson, 1995) and may even Bernhardt, P. C., Dabbs, J. M., Fielden, J. A., & Lutter, C. D. (1998). mentally suppress the stereotype (Iserman et al., 2004). Such social Testosterone changes during vicarious experiences of winning and los- identity avoidance characterized Black students who expressed 59–62. Physiology & Behavior, 65, ing among fans at sporting events. less identification with their racial group (Experiment 1). A second Biernat, M., Vescio, T. K., & Green, M. L. (1996). Selective self- strategy involves affirming solidarity with one’s group in the face stereotyping. 1194– Journal of Personality and Social Psychology, 71, of collective threat. To do so, people embrace the non-negative 1209. Blanton, H., Crocker, J., & Miller, D. T. (2000). The effects of in-group characteristics emblematic of their group identity (Spears et al., versus out-group social comparison on self-esteem in the context of a 1997) and maintain and even increase thoughts about their group. Journal of Experimental Social Psychology, 36, negative stereotype. Such social identity affirmation characterized Black students who 519–530. expressed high identification with their racial group (Experiment Blascovich, J., & Tomaka, J. (1996). The biopsychosocial model of arousal 1) and women in general (Experiment 3). regulation. In L. Berkowitz (Ed.), Advances in experimental social One question we explored concerns the moderating influence of (Vol. 28, pp. 1–51). San Diego, CA: Academic Press. psychology group identification. Although its effect manifested on different Branscombe, N. R., Schmitt, M. T., & Harvey, R. D. (1999). Perceiving measures, the overall pattern was consistent. Minority students pervasive discrimination among African Americans: Implications for who expressed less identification with their group proved more group identification and well-being. Journal of Personality and Social responsive to collective threat. Under threat, they suffered a 135–149. Psychology, 77, greater loss of self-esteem (Experiment 2), showed more avoid- Brewer, M. B., & Weber, J. G. (1994). Self-evaluation effects of interper-

16 NEGATIVE STEREOTYPES, COLLECTIVE THREAT 581 dynamic time course of stereotype activation: Activation, dissipation, sonal versus intergroup social comparison. Journal of Personality and Social Psychology, 66, 268–275. and resurrection. Journal of Personality and Social Psychology, 82, New York: Norton. The wisdom of the body. Cannon, W. B. (1932). 283–299. Cohen, G. L., & Garcia, J. (2003). [Collective representations of African Lakin, J. L., & Chartrand, T. L. (2003). Using nonconscious behavioral Americans.] Unpublished raw data. Psychological Science, 14, mimicry to create affiliation and rapport. Cohen, G. L., & Steele, C. M. (2002). A barrier of mistrust: How negative 334–339. stereotypes affect cross-race mentoring. In J. Aronson (Ed.), Improving Lalonde, R. N., & Silverman, R. A. (1994). Behavioral preferences in academic achievement: Impact of psychological factors on education response to social injustice: The effects of group permeability and social (pp. 303–328). San Diego, CA: Academic Press. identity salience. Journal of Personality and Social Psychology, 66, Cohen, G. L., Steele, C. M., & Ross, L. D. (1999). The mentor’s dilemma: 78–85. Providing critical feedback across the racial divide. Personality and Current Directions in Leary, M. R. (1999). Making sense of self-esteem. Social Psychology Bulletin, 25, 1302–1318. 32–35. Psychological Science, 8, American Be- Crosby, F. (1984). The denial of personal discrimination. Resolving social conflicts. New York: Harper. Lewin, K. (1948). havioral Scientist, 27, 371–386. Lewis, A. C., & Sherman, S. J. (2003). Hiring you makes me look bad: Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and Orga- Social-identity based reversals of the ingroup favoritism effect. Journal of Personality and Social Psychology, controlled components. 262–276. nizational Behavior and Human Decision Processes, 90, 5–18. 56, Leyens, J., De ́sert, M., Croizet, J., & Darcis, C. (2000). Stereotype threat: Doosje, B., Branscombe, N. R., Spears, R., & Manstead, A. S. R. (1999). Are lower status and history of stigmatization preconditions of stereo- Journal Guilty by association: When one’s group has a negative history. Personality and Social Psychology Bulletin, 26, type threat? 1189–1199. 872–886. of Personality and Social Psychology, 75, Luhtanen, R., & Crocker, J. (1992). A collective self-esteem scale: Self- Dunning, D., & Cohen, G. L. (1992). Egocentric definitions of traits and evaluation of one’s social identity. Personality and Social Psychology abilities in social judgment. Journal of Personality and Social Psychol- Bulletin, 18, 302–318. 341–355. ogy, 63, Major, B., Spencer, S., Schmader, T., Wolfe, C., & Crocker, J. (1998). Dunning, D., Meyerowitz, J. A., & Holzberg, A. D. (1989). Ambiguity and Coping with negative stereotypes about intellectual performance: The self-evaluation: The role of idiosyncratic trait definitions in self-serving Personality and Social Psychology role of psychological disengagement. Journal of Personality and Social Psychology, assessments of ability. 34–50. Bulletin, 24, 57, 1082–1090. Marques, J. M., & Paez, D. (1994). The “black sheep effect”: Social Ellemers, N., Spears, R., & Doosje, B. (1997). Sticking together or falling categorization, rejection of ingroup deviates, and perception of group apart: In-group identification as a psychological determinant of group variability. In W. Stroebe & M. Hewstone (Eds.), European review of commitment versus individual mobility. Journal of Personality and (Vol. 5, pp. 37–68). West Sussex, England: Wiley. social psychology 617–626. Social Psychology, 72, Mendoza-Denton, R., Downey, G., Purdie, V. J., Davis, A., & Pietrzak, J. Ellemers, N., Van den Heuvel, H., De Gilder, D., Maass, A., & Bonvini, A. (2002). Sensitivity to status-based rejection: Implications for African (2004). The underrepresentation of women in science: Differential com- American students’ college experience. Journal of Personality and So- British Journal of Social Psychol- mitment or the Queen-bee syndrome? cial Psychology, 83, 896–918. ogy, 43, 315–338. Neuberg, S. L., Smith, D. M., Hoffman, J. C., & Russell, F. J. (1994). Ethier, K. A., & Deaux, K. (1994). Negotiating social identity when When we observe stigmatized and “normal” individuals interacting: contexts change: Maintaining identification and responding to threat. Personality and Social Psychology Bulletin, 20, Stigma by association. Journal of Personality and Social Psychology, 67, 243–251. 196–209. Exline, J. J., & Lobel, M. (1999). The perils of outperformance: Sensitivity Oyserman, D., Kemmelmeier, M., Fryberg, S., Brosh, H., & Hart-Johnson, about being the target of a threatening upward comparison. Psycholog- T. (2003). Racial–ethnic self-schemas. Social Psychology Quarterly, 66, ical Bulletin, 125, 307–337. 333–347. Racism: A short history. Fredrickson, G. M. (2002). Princeton, NJ: Prince- Prentice, D. A., & Carranza, E. (2002). What women and men should be, ton University Press. shouldn’t be, are allowed to be, and don’t have to be: The contents of Glick, P., & Fiske, S. T. (2001). An ambivalent alliance: Hostile and Psychology of Women Quarterly, 26, prescriptive gender stereotypes. benevolent sexism as complementary justifications for gender inequal- 269–281. 109–118. American Psychologist, 56, ity. Pronin, E., Steele, C. M., & Ross, L. (2004). Identity bifurcation in Greenwald, A. G. (1980). The totalitarian ego: Fabrication and revision of Journal of response to stereotype threat: Women and mathematics. personal history. American Psychologist, 35, 603–618. Experimental Social Psychology, 40, 152–168. Manual for the Self-Perception Profile for Adolescents. Harter, S. (1988). Robins, R. W., Hendin, H. M., & Trzesniewski, K. H. (2001). Measuring Denver, CO: University of Denver. global self-esteem: Construct validation of a single-item measure and the Heatherton, T. F., & Polivy, J. (1991). Development and validation of a Rosenberg Self-Esteem Scale. Personality and Social Psychology Bul- scale for measuring state self-esteem. Journal of Personality and Social letin, 27, 151–161. 895–910. Psychology, 60, Rubin, M., & Hewstone, M. (1998). Social identity theory’s self-esteem Henderson-King, E. I., & Nisbett, R. E. (1996). Anti-Black prejudice as a hypothesis: A review and some suggestions. Personality and Social function of exposure to the negative behavior of a single Black person. Psychology Review, 2, 40–62. 654–664. Journal of Personality and Social Psychology, 71, Schmader, T. (2002). Gender identification moderates stereotype threat The Iserman, E. C., Spencer, S. J., Davies, P. G., & Quinn, D. M. (2004). effects on women’s math performance. Journal of Experimental Social perils of avoiding negative thoughts: Thought suppression as a mediator 194–201. Psychology, 38, of stereotype threat. Manuscript submitted for publication. Sellers, R. M., Rowley, S. A. J., Chavous, T. M., Shelton, J. N., & Smith, Klein, O., & Azzi, A. E. (2001). The strategic confirmation of meta- M. A. (1997). Multidimensional Inventory of Black Identity: A prelim- stereotypes: How group members attempt to tailor an out-group’s rep- inary investigation of reliability and construct validity. Journal of Per- resentation of themselves. British Journal of Social Psychology, 40, sonality and Social Psychology, 73, 805–815. 279–293. Sherman, D. K., & Cohen, G. L. (2002). Accepting threatening informa- Kunda, Z., Davies, P. G., Adams, B. D., & Spencer, S. J. (2002). The

17 COHEN AND GARCIA 582 African American college achievement: A “wise” Schoem, D. (2003). tion: Self-affirmation and the reduction of defensive biases. Current intervention. Manuscript submitted for publication. Directions in Psychological Science, 11, 119–123. Stone, J., Lynch, C. I., Sjomeling, M., & Darley, J. M. (1999). Stereotype Shih, M., Pittinsky, T. L., & Ambady, N. (1999). Stereotype susceptibility: Journal of threat effects on Black and White athletic performance. Psychological Identity salience and shifts in quantitative performance. Personality and Social Psychology, 77, 1213–1227. 80–83. Science, 10, Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup Cambridge, MA: Sniderman, P. M., & Piazza, T. (1993). The scar of race. The psychology of behavior. In S. Worchel & W. G. Austin (Eds.), Harvard University Press. intergroup relations (pp. 7–24). Chicago: Nelson Hall. Spears, R., Doosje, B., & Ellemers, N. (1997). Self-stereotyping in the face Tesser, A. (1988). Toward a self-evaluation maintenance model of social of threats to group status and distinctiveness: The role of group identi- Advances in experimental social psy- behavior. In L. Berkowitz (Ed.), fication. Personality and Social Psychology Bulletin, 23, 538–553. chology (Vol. 21, pp. 181–227). Hillsdale, NJ: Erlbaum. Spencer, S. J., Steele, C. M., & Quinn, D. (1999). Stereotype threat and Vorauer, J. D., Main, K. J., & O’Connell, G. B. (1998). How do individuals women’s math performance. Journal of Experimental Social Psychol- expect to be viewed by members of lower status groups? Content and ogy, 35, 4–28. Journal of Personality and Social implications of meta-stereotypes. Stangor, C., Carr, C., & Kiang, L. (1998). Activating stereotypes under- 917–937. Psychology, 75, Journal of Personality and Social mines task performance expectations. Walton, G. M., & Cohen, G. L. (2003). Stereotype lift. Journal of Exper- 1191–1197. Psychology, 75, imental Social Psychology, 39, 456–467. Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual Wong, C. A., Eccles, J. S., & Sameroff, A. (2003). The influence of ethnic test performance of African Americans. Journal of Personality and discrimination and ethnic identification on African American adoles- Social Psychology, 69, 797–811. Journal of Personality, 71, cents’ school and socioemotional adjustment, Steele, C. M., Spencer, S. J., & Aronson, J. (2002). Contending with group 1197–1232. image: The psychology of stereotype and social identity threat. In M. P. Received July 8, 2004 Advances in experimental social psychology (Vol. 34, pp. Zanna (Ed.), Revision received March 24, 2005 379–440). San Diego, CA: Academic Press. Steele, C. M., Spencer, S. J., Nisbett, R., Hummel, M., Harber, K., &  Accepted April 18, 2005 New Editor Appointed, 2007–2012 The Publications and Communications (P&C) Board of the American Psychological Association announces the appointment of a new editor for a 6-year term beginning in 2007. As of January 1, 2006, manuscripts should be directed as follows: (www.apa.org/journals/emo.html), Elizabeth A. Phelps, PhD, Department of Psy- • Emotion chology, New York University, 6 Washington Place, Room 863, New York, NY 10003. . As of January 1, 2006, manuscripts should be submitted Electronic manuscript submission electronically via the journal’s Manuscript Submission Portal (see the Web site listed above). Authors who are unable to do so should correspond with the editor’s office about alternatives. Manuscript submission patterns make the precise date of completion of the 2006 volumes uncertain. The current editors, Richard J. Davidson, PhD, and Klaus R. Scherer, PhD, will receive and consider manuscripts through December 31, 2005. Should 2006 volumes be completed before that date, manuscripts will be redirected to the new editor for consideration in 2007 volume.

Related documents