Steele and Aronson (1995) introduced the term stereotype threat to refer to the situation faced when an individual’s behavior is likely to be perceived as conforming to a negative stereotype. They provided the first demonstration of stereotype threat in a study comparing the performance of African American and White college students on a set of verbal problems. Some of the students were told that the test was diagnostic of their verbal abilities whereas others were told that it was not, serving instead to simply provide the researchers with insight into verbal problem solving. A pattern of findings was obtained in which the African American students performed more poorly in the diagnostic condition compared to the white participants more generally and worse than those African Americans assigned to the non-diagnostic condition. When performance on the test was diagnostic and could confirm the stereotype that African Americans are less intelligent than white Americans performance was depressed, suggesting one basis for lower standardized test scores and academic performance among African Americans.
The stereotype threat phenomenon appears to be robust and one obtained in many populations and settings. In Steele and Aronson’s (1995) research, the domain was that of verbal intelligence, but the domain may potentially refer to many different competencies, attributes, or patterns of behavior that are implicated by group stereotypes. The majority of studies have chosen the domain of mathematics and tested stereotype threat in women (Ambady et al., 2004; Ben-Zeev, Fein & Inzlicht, 2005; Brown & Josephs, 1999; Brown & Pinel, 2003; Cadinu, Maass, Frigerio, Impagliazzo & Latinotti, 2003, study 1; Davies et al., 2002; Gresky, Eyck, Lord & McIntyrre, 2005; Inzlicht & Ben-Zeev, 2000; Inzlicht & Ben-Zeev, 2003; Johns, Schmader & Martens, 2005; Keller & Molix, 2008; Kiefer & Sekaquaptewa, 2007a; Kiefer & Sekaquaptewa, 2007b; Lesko & Corpus, 2006; Martens et al., 2006, study 1; Marx & Roman, 2002; Marx & Stapel, 2006a; Marx & Stapel, 2006b; Marx & Stapel, 2006c; Marx, Stapel & Muller, 2005; McGlone & Aronson, 2007; McIntyre, Paulson & Lord, 2003; McIntyre et al., 2005; O’Brien & Crandall, 2003; Oswald & Harvey, 2000; Pronin, Steele & Ross, 2004; Rosenthal & Crisp, 2006; Rydell, McConnell & Beilock, 2009; Schmader, 2002; Schmader & Johns, 2003; Schmader, Johns & Barquissau, 2004; Smith & White, 2002, study 1; Spencer, Steele & Quinn, 1999; Thoman, White, Yamawaki & Koishi, 2008). Mathematics has also been chosen as the domain in studying stereotype threat among Latino men and women (Gonzales et al., 2002), Asian-American women (Shih, Pittinsky & Ambady, 1999), white men (Aronson, Lustina, Good & Keough, 1999; Smith & White, 2002, study 2), and children ranging from elementary to high-school grade levels (Ambady, Shih, Kim & Pittinsky, 2001; Keller & Dauenheimer, 2003; Neuville & Croizet, 2007). Focusing on other skills for which women are thought to be disadvantaged, tests of visual-spatial abilities have also been employed (Martens et al., 2006, study 2; McGlone & Aronson, 2006).
Other studies have focused on educational performance (Aronson et al., 2002) or intellectual abilities in African Americans (Blascovich, Spencer, Quinn & Steele, 2001; Cadinu et al., 2003, study 2); Davis, Aronson & Salinas, 2006; Steele & Aronson, 1995 ) or Latinos (Schmader & Johns, 2003). Research has also examined stereotype threat among individuals with biracial identities (Shih, Bonam, Sanchez & Peck, 2007, study 3) and males whose academic or athletic identities were dominant in the setting (Yopyk & Prentice, 2005). However, not all studies have been focused on academic or intellectual performance. There have also been investigations of leadership opportunities (Davies, Spencer & Steele, 2005) and athletic ability (Stone & McWhinnie, 2008) in women, memory ability in older adults (Levy, 1996), childcare interactions of gay men (Bosson, Haymovitz & Pinel, 2004) and emotional sensitivity of men (Leyens, Désert, Croizet & Darcis, 2000; Marx & Stapel, 2006c; Marx & Stapel, 2006d).
The common finding among these studies is that a particular ability, trait, or pattern of behavior expected from a stereotype elicits confirmation of the stereotype from the group’s members.
A common theoretical explanation of this phenomenon has been in terms of Heider’s (1958) balance theory and theories of cognitive consistency (Nosek, Banaji & Greenwald, 2002; Rydel et al., 2009; Schmader, Johns & Forbes, 2008). According to these models, stereotype threat originates in cognitive inconsistency between the relations among three components—the domain, the self, and the stereotyped group. These three relations may be viewed as one’s self-concept with respect to a given domain, one’s social identity with respect to a given social group, and the implications of the group stereotype for performance in the domain. For example, if the domain of mathematics is one valued by a woman, who identifies with her gender, and the stereotype of women suggests poor mathematical ability, there will be a state of inconsistency between wanting to perform well in the domain and the stereotype suggesting that women will do poorly. Recognition by the individual of cognitive inconsistency among these relations results in a revised assessment of the situation or behavior changes (e.g., "the test was biased,") that restore a state of cognitive consistency. This process of recognizing cognitive inconsistency and trying to restore a balanced state is then thought to undermine performance due to interference and competition for cognitive resources in working memory by these inconsistent and arousing thoughts (Schmader & Johns, 2003; Schmader et al, 2008).
Domain identification refers to whether competency in the domain is meaningful and important to the individual's self-concept. It is basically a truism in stereotype threat research that the individual must be identified with the domain to experience stereotype threat. For this reason, studies have often pre-screened participants based on domain importance (Ben-Zeev et al., 2005; Brown & Pinel, 2003; Cheryan & Bodenhausen, 2000; Inzlicht & Ben-Zeev, 2003; Schmader et al., 2004), domain ability or achievement (Martens et al., 2006; Schmader, 2002; Schmader & Johns, 2003) or both importance and ability (Aronson et al., 1999; Davies et al., 2002; Marx & Roman, 2002; Spencer et al., 1999). A meta-analysis by Walton and Cohen (2003) obtained further support for the importance of domain identification in stereotype threat studies. They found that effects were significantly stronger, and about three times larger in magnitude, for those studies that pre-selected participants who were identified with the domain compared to those that were not. Additionally, a number of studies have obtained evidence for the moderating role of domain identification by examining its effect on stereotype threat directly, and finding threat primarily in individuals highly identified with it (Aronson et al., 1999; Cadinu et al., 2003; Gresky et al., 2005; Lesko & Corpus, 2006; Leyens et al., 2000; Pronin, Steele & Ross, 2004).
The perceived relation between the domain and the self is determined by both the person's identification with the domain and the diagnostic accuracy of an evaluative measure. Major and Schmader (1998) make this distinction in presenting two ways different ways in which the individual may react to negative academic experiences. Both facilitate self esteem and manifest themselves as reactions to stereotype threat. One of these is to devalue the domain, rendering it as unimportant to the self. This response may represent a temporary reaction to the situation, such as within the stereotype threat setting (Davies, Spencer, Quinn & Gerhardstein, 2002, study 3; Rosenthal and Crisp, 2006, study 1), or a more permanent, chronic form of disengagement from the domain. Whereas the former may not have long-term effects, the latter represents a much more destructive form of disengagement that could have life-changing ramifications. Alternatively, one could discount the relevance of feedback by denying that a particular test or experience truly represents an assessment of one’s ability (Lesko & Corpus, 2006). If it is not a valid measure of the ability in question then there is no real threat at hand.
Another aspect of the self thought to moderate stereotype threat is the degree of identification with the stereotyped group. Cognitive consistency theories predict that a higher degree of identification with a negatively stereotyped group should yield greater cognitive inconsistency or imbalance. A number of studies have indeed shown a direct relationship between stereotype threat effects and measures of group identification (Kiefer & Sekaquaptewa, 2007b; Schmader, 2002) and stigma consciousness (Brown & Pinel, 2003). However, the use of different measures for these constructs has sometimes failed to validate this relationship or found enhanced performance among more highly identified African Americans (Davis et al., 2006; Mendoza-Denton, Petrzak & Downey, 2008, study 3b).
One important factor to consider with respect to measures of either domain or group identification is the point at which such measures are obtained. Whereas greater identification should enhance stereotype threat effects, changing one’s perceptions of these relationships may serve to restore cognitive consistency and reduce stereotype threat. Thus, some studies have found that stereotype threat resulted in short-term increases in domain (Kiefer and Sekaquaptewa, 2007a) or group (Marx et al., 2005, study 2) identification. For example, Marx et al. found that women rated gender as more important to their identity and as a bigger part of themselves under stereotype threat conditions than a non-diagnostic control. They also had lower expectancies of success on a math test. These findings are not unexpected because heightened salience of a stereotyped group identity is a necessary condition for stereotype threat. By activating this identity the threat is elicited. Moreover, once elicited, the individual is thought to ruminate on the provocative cognitions (Schmader et al., 2008), further activating them in memory. If a restructuring of these cognitions is successful in reducing cognitive inconsistency, however, identification with the domain or group may then be reduced. Opposite relationships may result depending on when identification is assessed (see McFarland, Lev-Arey & Ziegert, 2003).
A demonstration of how highly identified individuals might modify their self identities in response to stereotype threat is provided by Pronin et al. (2004). In the first of several studies, they measured women’s self ratings on gender stereotypic traits and found that the women who had completed more quantitative subjects in school showed a pattern of self perception that they termed identity bifurcation. This phenomenon manifests itself in dissociation from those stereotypic traits that are related negatively to perceived success in the domain. Pronin et al. found not only that those women more highly identified with the domain rated themselves lower on these attributes, but in subsequent studies it was seen that manipulating stereotype threat resulted in situationally induced identity bifurcation. Thus, selectively differentiating oneself from the stereotyped group appears to provide one means of reducing stereotype threat and the cognitive inconsistency thought to underlie it.
Although stereotype threat was originally viewed in terms of how one’s behavior might make the “stereotype more plausible as a self-characterization in the eyes of others, and perhaps even in one's own eyes” (Steele & Aronson, 1995, p. 797, emphasis added), the effect occurs even when one’s behavior remains private (Inzlicht & Ben-Zeev, 2003; Wout, Danso, Jackson and Spencer, 2008). Assessment by others does not appear to be a necessary condition for stereotype threat and it may reflect a relatively private and personal reaction to the setting.
Despite the existence of stereotype threat as a private phenomenon, one can differentiate between threats to the individual and collective selves (Gaertner, Sedikides & Graetz, 1999; Gaertner, Sedikides, Vevea & Iuzzini, 2002). Wout, et al. (2008) argued that past research has often confounded these manipulations. They employed different experimental conditions in separating these threats. In a self-threat group, female participants took a math test that assessed their personal mathematical competency. These women graded their own tests and kept scores to themselves, making performance a private matter. A second, group-threat, condition was told the purpose of the test was to assess gender differences in math ability and that no personally identifying information was being collected, only the person’s gender. Thus, performance, although private regarding the self, had implications for a group with whom they identified. Although in the first of multiple studies, evidence of stereotype threat appeared only in the self-threat group, in a subsequent study, Wout el al. found evidence of depressed performance in the group-threat condition as well. However, this effect occurred only among participants who were highly identified with the group. This research suggests that more than one type of process might be operative in test settings and that the two need to be distinguished in future studies. Although self-threat was consistently found to depress performance, group-based threat required a particularly strong collective identity related to the social category.
The experience of stereotype threat is a subjective one and depends on how the individual interprets the situational context. In some cases, different interpretations may be possible. For instance, Stone, Lynch, Sjomeling and Darley (1999) were able to negatively impact the performance of white participants by telling them a golf game was a test of athletic ability whereas African Americans’ performances were impaired by telling them the same game was a test of sports intelligence. In this manner, the ambiguity of the behavioral context allowed for direct manipulation of the domain’s interpretation and its relevance to different racial stereotypes across experimental conditions, each playing to a different racial stereotype.
Other studies have not tried to modify the perceived domain per se, but rather, its perceived relationship to the stereotyped group. For example, Ben-Zeev et al. (2005) told participants in a no-threat condition that a standardized math test was chosen for the study specifically because prior research had found no differences in how men and women performed on the test. Similarly, a number of other studies have reduced threat, and enhanced performance, by describing the same test as culturally unbiased or lacking group differences (Blascovich et al., 2001; Brown & Pinel, 2003; O’Brien & Crandall, 2003; Spencer, Steele & Quinn, 1999). The assumed causes for such relationships also appear to be important. Thoman et al. (2008) found that attributing gender differences in quantitative test scores to an innate superiority of men reduced women’s performance compared to a condition where differences were attributed to difference in effort.
Once the domain has been assessed, the accessibility and applicability of a given stereotype may be strongly determined. For this reason, stereotype threat may arise even in the absence of specific invocation of the stereotype. Smith and White (2002, study 1), for instance, assessed the extent to which unfavorable stereotypes regarding the mathematical ability of white women (compared to men, study 1) and white men (compared to Asians, study 2) must be explicitly mentioned to invoke stereotype threat. Three groups were used in the study. In one case, no specific stereotype was mentioned and threat was assumed to be elicited implicitly. A second group read a seemingly credible article that argued for differences in mathematical abilities consistent with the stereotype and was told the test they would be taking had shown similar results. In this condition the stereotype was explicitly presented. Finally, a third group was treated the same as the explicit-stereotype condition but was told that the present test had not shown a pattern of differences consistent with the stereotype. This was the nullified-stereotype condition. The basic findings of these studies were that both implicit-stereotype and explicit-stereotype groups displayed poorer test performance than the nullified-stereotype group but comparable performance with respect to each other. In other words, the stereotype did not have to be explicitly mentioned to elicit stereotype threat, because it was, presumably, already accessible in the minds of the participants.
Even stronger evidence for the pervasive salience of stereotypic beliefs within these settings is provided by Walton and Cohen (2003). In their meta-analysis of stereotype lift (a bolstering rather than disruption of performance) and stereotype threat research, they found that effect sizes were much larger in those studies that negated the stereotype rather than simply leaving the domain-group relationship unspecified. This analysis further supports the idea that people may inherently perceive the relevance of applicable stereotypes based on contextual cues and their own membership in the stereotyped group. It also provides more compelling results than those provided by any single study. However, what many of these studies lack is the use of implicit measures of stereotype activation in addition to performance outcome measures. Knowing the strength and content of stereotype activation in these studies would greatly aid in their interpretation.
One reason stereotype threat may result even in the absence of strong contextual cues relating to the stereotype may be that classification in terms of highly salient physical characteristics, such as gender, skin color or age, occurs automatically, independently of one’s conscious goals (see Fiske & Neuberg, 1990). The composition of individuals within the setting alone can affect the salience of these characteristics (
A number of different manipulations have been tested by researchers as ways of reducing stereotype threat research. One is the presentation of positive role models or exemplars. Schmader et al. (2008) argue that positive exemplars, or portrayals of stereotype-disconfirming group members, may provide one means of changing the negative relationship between the group and the domain to one that is positive in valence. Studies show that presenting positive exemplars prior to assessment reduces the performance decrements associated with stereotype threat (Marx & Roman, 2002; McIntyre, Paulson & Lord, 2003; McIntyre et al., 2005). However, these effects have not always been exemplars related to the domain under study. In one study, for example, McIntyre et al. (2003, study 2) found that presenting descriptions of successful women in various fields other than mathematics resulted in better math performance than that found among women reading about successful corporations. Because these positive portrayals of successful women were not directly related to the domain in question, they do not appear to be interpretable in terms of modified beliefs regarding the stereotyped group’s competency.
Another strategy for reducing stereotype threat is self-affirmation. Martens et al. (2006, study 1) had participants in a self-affirmation condition rank order 11 characteristics or values that were personally important to them and then write about the highest ranked attribute. In a non-affirmation condition instructions were to choose the item ranked ninth and to write about its importance to other people rather than from their own personal perspective. Women in the affirmation condition showed relief from stereotype threat on a math test compared to those in the non-affirmation condition and compared to male participants.
Gresky et al. (2005) also manipulated the salience of particular aspects of participants' self-concepts. They found that constructing self-concept maps with many nodes by highly math-identified women bolstered test performance compared to those who constructed maps with few nodes or who did not construct self-concept maps. Similarly, Ambady et al. (2004) asked women undergraduates to answer personal questions about themselves, to list positive and negative characteristics of themselves, and to provide examples from their lives for each characteristic listed. The authors viewed this manipulation as fostering individuation and had half of the women perform this task. Additionally, in order to manipulate stereotype threat, the women were either subliminally primed with female word associates or neutral words beforehand. As predicted, the individuation task alleviated the performance impairment among women primed with the female words.
All of these interventions involved manipulations performed prior to the assessment task and presumably prevented the elicitation of stereotype threat. However, is it possible to actively suppress its expression after the fact? Unfortunately, this approach does not appear likely to succeed, at least without considerable effort. Active suppression of unwanted thoughts may further activate the very thoughts that one is attempting to suppress and lead to rebound effects (Macrae, Bodenhausen, Milne & Jetten, 1994; Wegner, 1994; Wegner & Erber, 1992; Wegner, Schneider, Carter & White, 1987). Furthermore, the thoughts used to suppress the stereotype’s application may adversely impact the individual’s ability to process other, non-stereotypic information (Bodenhausen & Macrae, 1996) and interfere with performance in the stereotype threat setting (Schmader et al., 2008). On the other hand, given repeated training at suppression strategies, it may be possible to achieve some success by eventual conversion of what began as conscious suppression of a stereotype into a relatively effortless automatic activity.
In contrast to the active suppression of interfering thoughts, the operation of inhibitory processes is likely to offer a less effortful, preconscious means of preventing the activation of stereotypes or other threat-inducing cognitions without commandeering working memory resources critical to performance (see Dijksterhuis & van Knippenberg, 1998 for a discussion of the distinction between category suppression and inhibition). Moskowitz, Gollwitzer and Schaal (1999) found evidence, for example, of the preconscious inhibition of stereotypes by chronic egalitarian goals. It is also possible to inhibit the activation of a stereotype or other category in memory by activating other competing categories (Dunn & Spellman, 2003; Hugenberg & Bodenhausen, 2004). For example, Macrae, Bodenhausen and Milne, (1995) had participants view a Chinese woman after subliminal priming of words related to either the female stereotype or others related to her Chinese ancestry. In a subsequent lexical decision task not only were reaction times to words related to the primed stereotype faster than neutral words but those related to the unprimed category were actually slower than the control, thereby demonstrating actual inhibition as well as facilitation of categorical representations in memory.
Through a change in the activation of alternate social identities from one having negative implications for performance in a setting to one having more positive implications, category inhibition may provide a basis for preventing stereotype threat. Indeed, a number of studies show that experimentally manipulated changes in the relative salience of alternate social identities moderate stereotype threat (Ambady et al., 2001; Gresky et al., 2005; McGlone & Aronson, 2006; Neuville & Croizet, 2007; Rydell, McConnell and Beilock, 2009; Shih et al., 1999; Yopyk & Prentice, 2005, study 1) and self-stereotyping (Sinclair, Hardin & Lowery, 2006; Sinclair & Lun, 2006).
One example is provided by Shih et al. (1999). They varied the salience of different stereotypes associated with Asian-American women by either priming the categories of ethnicity or gender. They found that the participants performed the worst on a math test when the gender stereotype was primed, best when the Asian stereotype was primed, and midway between these conditions for a control group. Activation of different stereotypes—one with negative implications for performance and one with positive implications—yielded opposite effects that were consistent with the stereotype. Thus, a shift in identity can reduce stereotype threat if that identity has opposite implications for performance.
This research suggests that interventions aimed at modifying the relative salience of different stereotypes and social identities may provide an effective means of preventing the emergence of stereotype threat. Further research into category inhibition as a basis for interventions in this area are warranted. Moreover, category activation and inhibition may ultimately underlie many of the past interventions proposed for stereotype threat.
For more information about stereotype threat visit ReducingStereotypeThreat.org.
No comments:
Post a Comment