Stereotype Suppression in High versus Low Stigma-conscious Women Experiencing Stereotype Threat

Activating the stereotype of women’s lesser ability in math during test taking causes women to underperform on math tests, a phenomenon known as stereotype threat. The present study examined how women high (HSC) versus low (LSC) in stigma consciousness differ in their experience of stereotype threat and their use of stereotype suppression. Given that stereotype threat effects are exacerbated for HSC women, HSC stereotype-threatened women were expected to underperform on a given math-task relative to their LSC counterparts. They were also expected to suppress the female stereotype to a greater extent prior to the task, and to suffer from greater postsuppressional rebound afterwards. Following a stereotype threat manipulation, in which only those in the threat condition were told that an upcoming task was indicative of math ability, participants completed a working memory test. As expected, HSC stereotypethreatened participants demonstrated test underperformance; however, they neither suppressed the female stereotype nor experienced post-suppressional rebound. The findings strongly suggest that HSC women are, in fact, more vulnerable to stereotype threat than are LSC women, who in turn appear to be protected against stereotype threat even when faced with explicit cues of prejudice. STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 3 Stereotype Suppression in High versus Low Stigma-Conscious Women Experiencing Stereotype Threat Despite making up almost half of the overall American workforce, women make up less than a quarter of the science, technology, engineering, and math (STEM) workforce (Beede, Julian, Langdon, McKittrick, Khan, & Doms, 2011). This underrepresentation has serious negative implications in our current job market, in which the high-growth sectors are becoming increasingly math and technology-oriented (Dickler, Lee, & Swiatek, 2011) and STEM jobs are among the most profitable (Langdon, McKittrick, Beede, Khan, & Doms, 2011) and gender-fair in terms of wage (Beede et al., 2011). Since STEM workers typically attain high-level degrees (Langdon et al., 2011), it is unsurprising that the gender gap found at the professional level is mirrored in postsecondary education. At the doctorate level, women are grossly underrepresented in mathematics and the physical sciences, with less than one-third of the doctorate degrees in these fields being awarded to women. Women make up an even smaller percentage of earned doctorates in engineering and computer science (National Science Fund, 2011). Shedding light on women’s underrepresentation in these quantitative fields is their underachievement at high-level math. A recent meta-analysis by Lindberg, Hyde, Petersen, and Linn (2010) suggests that the gender performance differential in pre-college math has dissipated over the past two decades, dispelling the stereotype that men are superior to women at math. However, despite being on par with men, female students are still underrepresented in university high-level math courses. Moreover, belief in the negative math-gender stereotype still thrives among students in these courses (Good, STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 4 Aronson, & Harder, 2008). Given that these courses act as gateways to careers in the quantitative fields (Good et al., 2008), women’s lack of support and achievement in these high-level math courses bar their further participation in mathematics and related areas. In order to make sense of the math-gender disparity within and beyond academia, some theorists have looked to broad sociocultural forces, such as discrimination, to explain why women have not yet achieved their potential in STEM fields (e.g., Ceci & Williams, 2009). On the contrary, other theorists, namely stereotype threat theorists, have looked to the pervasive math-gender stereotype as an explanation for women’s math underperformance (Steele & Aronson, 1995). Unlike broad sociocultural approaches, stereotype threat theory offers a strictly situational account for why women underperform in the math domain. The present study examines a potential individual differences factor that may impact a woman’s experience and response to stereotype threat. Stereotype Threat All invested math students feel pressure to perform well on math tests, but women bear an additional pressure that men do not. Unlike men, women must perform well in the face of the existing negative stereotype about their gender’s math ability. If a female student’s math performance is substandard, she runs the risk of confirming or being judged by the relevant math-gender stereotype. The added burden of having to contend with this stereotype may explain why women underperform at high-level math. This dilemma is known as stereotype threat (Steele & Aronson, 1995). Stereotype threat describes an immediate situational threat, or “threat in the air” (Steele, 1997), that is perceived when a test taker is aware that his or her performance may be evaluated with respect to the stereotype. According to stereotype threat theory, STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 5 perceiving this threat compromises the test taker’s ability, thereby causing underperformance. Spencer, Steele, and Quinn (1999) first demonstrated the negative effects of stereotype threat on women’s math performance using what has become the classic stereotype threat manipulation. By simply informing participants that a difficult math test demonstrates gender differences in performance, women in this “threat” condition underperformed relative to men, whereas women informed that the test was gender-fair performed no differently than men. The effectiveness of manipulating test presentation to induce stereotype threat speaks to the situational nature of stereotype threat; it arises from the performance situation itself. Given its situational nature, it follows that stereotype threat is not specific to a particular social group or task domain. A test taker is capable of experiencing stereotype threat insofar as he or she belongs to a group that is socially expected to underperform (Steele, 1997). The broad scope of stereotype threat has been supported by its numerous demonstrations in a variety of social groups and task domains, including Blacks (Steele & Aronson, 1995), Latinos (Gonzales, Blanton, & Williams, 2002; Schmader & Johns, 2003), and those of low socioeconomic status (Croizet & Claire, 1998) on intellectual tasks; Whites on tests of natural athletic ability (Stone, Lynch, Sjomeling, & Darley, 1999; Stone, 2002); and, of particular interest to the present paper, women on math tasks (Brown & Pinel, 2003; Carr & Steele, 2009; Davies, Spencer, Quinn, & Gerhardstein, 2002; Good, Aronson, & Harder, 2008; Kiefer & Sekaquaptewa, 2007; Logel, Iserman, Davies, Johns, Schamder & Martens, 2005; Quinn, & Spencer, 2009; Spencer, Steele & Quinn, 1999). Stereotype threat has even been demonstrated in social groups that are not traditionally stigmatized, such as White men. In a study by Aronson, Lustina, Good, STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 6 Keough, Steele, and Brown (1999), informing White male students that their math test performance would be compared to that of Asian male students, who are stereotypically superior at math, successfully provoked stereotype threat effects. Again, this finding reinforces the fact that, regardless of their past experience in the test domain, test takers can feel threatened by a stereotype when it is made relevant during the testing situation. Targets of stereotype threat. However, not all people who encounter stereotypethreatening testing conditions experience stereotype threat to the same extent. A person must satisfy several conditions in order to be susceptible to its negative effects. The most basic of these conditions is stereotype awareness. That is, people must be aware of the relevant stereotype in order to perceive the risks associated with stereotype-consistent performance. Although people must be aware of the stereotype, they do not need to subscribe to the stereotype, nor personally satisfy the stereotype in order to be threatened by it (Steele, 1997); people can still fear confirming a stereotype they know is untrue of themselves or in general. What is necessary, however, is an understanding that others in the testing situation may assess their performance according to the relevant stereotype. Additionally, people must identify with the stereotyped task domain in order to feel threatened by the task-relevant stereotype. Domain identification describes the perceived importance of a task domain to self-definition (Aronson et al., 1999). Since those who strongly identify with a task domain perceive task ability as important to their self-concept, it follows that a part of their self-regard is contingent on task performance. Therefore, domain identification is a necessary condition because it disposes people to the performance pressure characteristic of stereotype threat. The importance of domain identification has been supported by studies in which those who highly achieve within the STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 7 stereotyped task domain exhibit stereotype threat effects. Despite clearly refuting the relevant stereotype, high-achievers are not only capable of experiencing stereotype threat (Aronson et al., 1999; Good et al., 2007) but may be especially susceptible to its effects (Steele, 1997). This vulnerability may be due to the fact that high achievement within a task domain involves high identification with that domain. Although gender identification has also been recognized as a moderator of mathgender stereotype threat (Schmader, 2002), such that only women who sufficiently value being a woman are threatened by the female stereotype, this may depend on how the threat is framed. In a study by Wout, Danso, Jackson, and Spencer (2008), gender identification only moderated stereotype threat effects in a group-threat condition, in which test takers were told that their math test performance would inform the researchers about gender differences in math performance. Conversely, women in a self-threat condition, in which they risked confirming the relevant stereotype to only themselves, underperformed irrespective of their gender identification. Therefore, the extent of a woman’s gender identification may play a role only when the math-gender threat is oriented towards women as a whole, and not them individually (Brown & Pinel, 2003; Kiefer & Sekaquaptewa, 2007; Wout et al., 2008). Stereotype threat tasks. Just as there are necessary conditions for a person to experience stereotype threat, so are there necessary conditions for a task to be sensitive to its detrimental effects. In order for a test to demonstrate performance decrements, the stereotype must be applicable to the test itself (Wheeler & Petty, 2001). The necessity of stereotype applicability is supported by findings that math-gender stereotype threat STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 8 impedes performance only on math tests and not on verbal tests, for which women are not stereotyped (Logel, Walton, Spencer, Iserman, & Von Hippel, 2009). Furthermore, the test must be described as or assumed to be diagnostic of the stereotyped ability. Presenting a task as diagnostic induces stereotype threat because it creates the necessary evaluative context in which stereotypic assessment is possible. Diagnosticity of a task can also be implied by its difficulty. Stereotype threat is more likely to occur under difficult testing conditions because tests that require one’s highest effort are assumed to be diagnostic and therefore subject to stereotypic evaluation (Logel et al., 2009; Spencer et al., 1999). But, as demonstrated by the classic stereotype threat manipulation, these diagnostic conditions are threatening only if the relevant negative stereotype is made salient. Past research has shown that the manner of stereotype activation can vary widely, from the blatant to the subtle (Stone & McWheenie, 2008). Relatively blatant stereotype activation has included directly informing test takers that their social group underperforms on this test (e.g., Aronson et al., 1999) or implying it by stating that the test produces group performance differences (e.g., Brown & Pinel, 2003). More subtle stereotype activation operates through making the participant’s social group identity salient, which has been done in a number of ways, including completing a demographics form prior to testing (Steele & Aronson, 1995), being the only in-group member present (Inzlicht & Ben-Zeev, 2000; Sekaqueptewa & Thompson, 2003), being exposed to stereotypic group portrayals (Davies et al., 2002), and interacting with an outgroup member (Danso & Esses, 2001; Logel et al., 2009). Underlying processes. Negative performance following stereotype threat has been partially explained by several processes, all of which are initiated by the activation STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 9 of stereotypic constructs. It is important, however, to distinguish stereotype threat from another model that touches on stereotype activation: the ideomotor model. The ideomotor phenomenon describes the automatic tendency for behavior to follow suit with perceptual representations (Wheeler & Petty, 2001). That is, simply perceiving or thinking about an action or concept can prime a person to behave correspondingly. For example, people may unconsciously imitate the facial expressions or vocal patterns of a person they have recently seen (Dijksterhuis & Bargh, 2001). When applied to stereotype activation, the ideomotor model states that activating stereotypic constructs will result in stereotype-consistent behavior. Since these priming effects can occur without conscious awareness, stereotype threat and ideomotor theory differ in a critical way: the ideomotor mechanism does not rely on the test taker’s conscious feelings or motivational states to explain underperformance (Wheeler & Petty, 2001). Although this priming theory can explain phenomena caused by the activation of an otherstereotype, such as why activating the supermodel stereotype causes people to perform poorly on intellectual tasks (Dijksterhuis & van Knippenberg, 1998), or why activating the stereotype of the elderly causes people to walk more slowly (Bargh, Chen, & Burrows, 1996), it cannot explain why activating a negative self-relevant stereotype results in task underperformance. Davies, Spencer, Quinn, and Gerhardstein (2002) distinguished stereotype threat effects from ideomotor effects in a study in which men and women were exposed to commercials depicting the female stereotype prior to taking a math test. Although the female stereotype was activated in both men and women, as indicated by results of an Implicit Association Test (IAT), only women underperformed on the math test. Since the STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 10 ideomotor model purports that the activation of any stereotype, regardless of selfrelevance, results in stereotype-consistent behavior, this model cannot explain women’s underperformance. Instead, women’s emotional and cognitive responses should be taken into account, as they are under the stereotype threat model. According to the theory of stereotype threat, it is not stereotype activation in itself that causes underperformance, but the affective and cognitive responses to this stereotype activation. Affective responses, such as increased arousal (Ben-Zeev, Fein, & Inzlicht, 2005; O’Brien & Crandall, 2003) and heart rate (Croizet, Despres, Guizins, Huguet, Leyesns, & Meot, 2004), are collectively understood as a physiological stress response (Schmader, Johns, & Forbes, 2008). This stress response has been found to strain cognitive efficiency (Croizet et al., 2004). Further straining the test taker’s ability are the concerns and worries that accompany the stress response. Ruminating on the negative thoughts associated with the stereotype can co-opt necessary cognitive resources, namely working memory capacity (Beilock & Beilock, 2007). By absorbing a considerable amount of cognitive resources, either physiologically or cognitively, the negative thoughts activated during stereotype threat (Schmader & Johns, 2003; Schmader, Johns, & Forbes, 2008) leave behind only a limited amount of cognitive resources with which to complete the actual task. Thus, stereotype threat theory attributes task underperformance to the burdened cognitive capacity of the stereotype threatened. Long-term consequences of stereotype threat. Task underperformance is not the only consequence of stereotype threat. In the short run, stigmatized people who receive negative feedback on a stereotyped task may situationally “disengage” from the stereotyped task domain (Nussbaum & Steele, 2007), during which their self-esteem is STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 11 momentarily suspended from the performance outcomes of the stereotyped task (Major, Spencer, Schmader, Wolfe, & Crocker, 1998). Since situationally disengaged people can insulate themselves from negative feedback while still identifying with the task domain, this temporary disengagement may be an adaptive coping strategy (Nussbaum & Steele, 2007). However, repeated experiences of stereotype threat may eventually lead stigmatized people to disengage to the point that the domain is no longer a significant part of their self-concept or, in other words, to disidentify from the domain (Osborne, 1997; Steele, 1997). This psychological withdrawal from the stigmatized domain leads to underparticipation (Steele, 1997) as well as poor performance and motivation (Pinel, Warner, & Chua) within that domain. But certain stigmatized people are more predisposed to experience stereotype threat, and are therefore more at risk for domain disidentification, than are others (Brown & Lee, 2005; Pinel et al., 2005). One example of such an individual differences factor is stigma consciousness. Stigma Consciousness People differ in their experiences of stigmatization. Whereas some people do not allow a self-stigma to affect their experiences, others are constantly aware of, and therefore affected by, this self-stigma. The chronic awareness of one’s stigmatized status is known as stigma consciousness (Pinel, 1999). Stigma consciousness is not simply being aware of a self-stigma, as it is possible for a person to be aware of one’s stigmatized status and not be affected by it. Rather, stigma consciousness describes a fixation on one’s stigmatized status. By being self-conscious of their stigmatized status, people high in stigma consciousness interpret interactions and experiences from a perspective that emphasizes STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 12 their self-stigma (Pinel, 1999). As such, highly stigma conscious (HSC) people are constantly anticipating, and are therefore vigilant for cues of prejudice. These chronic prejudice expectations affect how people interpret interactions with out-group members. For example, relative to low stigma conscious (LSC) women, HSC women are more likely to anticipate negative interactions with men (Pinel, 2004) and, when actually faced with these negative interactions, are more likely to attribute them to discrimination (Pinel, 1999; Pinel, 2002). But the heightened vigilance of HSC women for prejudice may also operate on an unconscious level. In a study by Kaiser, Vick, and Major (2006), women’s chronic expectations for prejudice predicted their preconscious attention to sexist words during a Stroop task. Furthermore, women’s chronic prejudice expectations predicted preconscious attention only when the female stereotype was made salient during the testing situation, as it is during experiences of stereotype threat. Given that HSC women are more vigilant than LSC women for stereotype cues in threatening environments, it follows that they are more vulnerable to stereotype threat. Brown and Pinel (2003) demonstrated this vulnerability by informing both HSC and LSC women that an upcoming math test produced gender differences. However, even after this stereotype threat manipulation, only HSC (but not LSC) women underperformed as a result of stereotype activation. Brown and Pinel (2003), however, have suggested that using an extreme stereotype threat manipulation, in which the self-relevant stereotype is explicitly mentioned, may successfully induce stereotype threat in LSC women. By explicitly informing women that their group is expected to underperform, even women who are STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 13 otherwise unaware of the risks associated with stereotypic performance should feel burdened by the now salient stereotype. While this particular manipulation has not yet been used in stereotype threat research, Pinel (2004) successfully demonstrated that LSC women can be situationally manipulated to behave as HSC women if they are asked to reflect upon past stigmatization. When asked to focus on previous experiences of discrimination on the basis of their gender, women low in dispositional (trait) stigma consciousness adopted a situationally induced (state) stigma consciousness. As a consequence of this temporarily elevated stigma consciousness, women typically low in stigma consciousness adopted the same attributional style as women chronically high in stigma consciousness. Coping with stereotype threat. As previously mentioned, people disidentify from stigmatized domains if they continuously experience stereotype threat. The interplay between stigma consciousness, stereotype threat, and domain disidentification has been demonstrated within academic settings, as level of stigma consciousness has been found to predict the GPA (Brown & Lee, 2005) as well as the extent of psychological disengagement from the academic domain (Pinel et al., 2005) of academically stigmatized students (e.g., Blacks, Latinos) but not of non-stigmatized students (e.g., Whites, Asians). Thus, as a consequence of their predisposition to stereotype threat, HSC students are at greater risk than LSC students for academic disidentification. While some stereotype-threatened people cope with stereotype threat by psychologically removing themselves from stigmatized domains, others may rely on a more immediate coping strategy, one that occurs during test taking: stereotype suppression. STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 14 Stereotype Suppression Those who attempt stereotype suppression employ it as a means of clearing the mind of negative stereotypic thoughts, as they believe it will allow them to focus on the task at hand. However, due to the nature of the cognitive processes involved, thought suppression leads to unintended negative effects. One negative effect is the reduction of working memory capacity (Schmader & Johns, 2003). This reduction can be attributed to the high mental control required by stereotype suppression. As described by Wegner and Erber (1992), thought suppression is a taxing procedure involving two simultaneous cognitive processes: a controlled process responsible for self-distraction, and an automatic process responsible for detecting the undesired thought. At the onset of thought suppression, the controlled process engages in a “controlled distractor search,” through which thoughts unrelated to the target are sought out as potential distractors. Once an appropriate distractor is found, conscious attention shifts from the search to the selected distractor. But this distraction is only momentary, as it is eventually disrupted by the intrusion of the target. With the reappearance of the target thought, the controlled distractor search is reinitiated. Sustaining this cycle between searching and self-distracting is the automatic process. While the controlled process provides necessary distractors, the automatic process continually detects any signs of an emerging target through an “automatic target search.” In contrast to the controlled distracter search, this target search is continuous and beyond any conscious control or awareness. Paradoxically, the very intrusions that the controlled and automatic processes work to avoid are, in part, due to this automatic target search. That is, in order to detect STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 15 the target thought, the target thought must be kept activated, albeit unconsciously. Therefore, despite attempts to purge the mind of the target, thought suppression maintains it in trying to detect it. Moreover, the detection of a budding target thought inevitably draws conscious attention to it, consequently transforming an unconscious trace into a full-fledged intrusion. These searches, and the constant switching between them, are cognitively demanding. So, although thought suppression appears to free the mind of the burden of a salient stereotype, it limits cognitive capacity in itself (Schmader & Johns, 2003). Since thought suppression is already cognitively taxing, an additional cognitive load, such as that imposed by a time constraint or a concurrent task, undermines the success of thought suppression (Wegner & Erber, 1992). When deprived of necessary resources, the controlled process is incapable of finding distractors and the automatic process goes unchecked (Wegner, 1994). Consequently, the unceasing automatic target search repeatedly detects the target thought, resulting in widespread activation of the initially suppressed thought. Spencer demonstrates this suppression failure in a study in which placing stereotyped-threatened women under an additional cognitive load led to increased activation of stereotype-associated concepts (as cited in Schmader & Johns, 2003, p. 450). Similarly, ceasing a stereotype suppression attempt causes previously suppressed thoughts to become hyperaccessible (Wegner, Schneider, Carter, & White, 1987). This post-suppressional rebound has been demonstrated with the suppression of nonselfstereotypes as well as self-stereotypes. In a study by Macrae, Bodenhausen, Milne, and Jetten (1994), participants performed an impression formation task in which they were STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 16 provided a picture of a Black man and asked to write about a typical day in his life. Consistent with the rebound effect, the stories of participants who had been instructed to suppress the Black stereotype contained greater stereotypical content than did the stories of those who had not been instructed to suppress. Interestingly, similar results have been found without using explicitly instructing participants to thought suppress. In an experiment by Wyer, Sherman, and Stroessner (1998), participants performed the same impression formation task as that used by Macrae and colleagues (1994), only this time they were not instructed to suppress. Instead, participants were either told that the study was conducted by the organization, “African Americans for Intercultural Understanding,” or were given no background at all. Those told about the organization self-initiated stereotype suppression, as this information acted as a cue to rein in prejudiced responses. Consistent with previous research, the stories written by participants who spontaneously suppressed were greater in stereotypicality than were those of control participants. These situational cues have also prompted spontaneous thought suppression during experiences of stereotype threat. In one study by Logel, Walton, Spencer, and Iserman (2009), female engineering students who interacted with a sexist male colleague prior to an engineering test spontaneously suppressed the female stereotype prior to testing. This suppression was indicated by these women’s slower reaction times to stereotypic words on a lexical decision task taken before the test, relative to control participants’ reaction times. Logel, Isermen, Davies, Quinn, and Spencer (2009) replicated these results with female math students and, by using a lexical decision task STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 17 following the math test, also found evidence for the subsequent post-suppressional rebound effect as well. If stereotype suppression is a seemingly rational coping strategy, there is reason to expect that both HSC and LSC women will spontaneously suppress the female stereotype while experiencing stereotype threat. Although it is possible that HSC women are more prone to suppressing stereotypical thoughts than are LSC women, this is unlikely the case. Once people are led to experience stereotype threat, the relevant stereotype is activated in these people regardless of their chronic sensitivity to stereotypic constructs. Thus, assuming that the female stereotype is activated in both HSC and LSC stereotype threatened women, both groups should look to thought suppression as a means of improving test ability. However, this neither means that the stereotype will be activated to the same extent in HSC and LSC women, nor that their stereotype suppression will produce the same outcomes. Although there is no direct research that compares the spontaneous stereotype suppression of HSC and LSC women, there is some research that suggests that the female stereotype is more strongly activated in stereotype-threatened HSC women than in their LSC counterparts, and that, as a consequence, they will suffer greater post-suppressional rebound than will LSC women (e.g., Kiefer & Sekaquaptewa). In a study by Kiefer and Sekaquaptewa (2007), HSC women were found to possess strong gender-math stereotypic associations, as measured by an Implicit Associations Test. Moreover, these associations remained strongly activated during math testing even when the test was described as non-diagnostic, suggesting that the stereotype is chronically accessible for HSC women. In another study, in which both HSC and LSC women were randomly assigned to either suppress or not suppress the stereotype of STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 18 women’s poorer spatial skills while interacting with a male confederate, HSC women in the suppression condition displayed more stereotypic behavior (e.g., less dominant nonverbal behavior) in their interactions than did their LSC counterparts. These findings suggest that the unsuccessful stereotype suppression of HSC women may be due to their more accessible stereotypic constructs (Borton, Reiner, Vazquez, Ruddiman, & Anglin, 2011). Given that the female stereotype is chronically accessible for HSC women, their stereotype hyperaccessibility following directed suppression is to be expected. However, little is known about whether there are differences between HSC and LSC women in selfinitiated stereotype suppression, and whether HSC women will suffer greater costs during suppression due to the strongly activated female stereotype. The Present Study The present study examined how women high versus low in stigma consciousness differ in their experience and response to stereotype threat. First, I assessed whether LSC women would also experience stereotype threat when faced with an extreme math-related threat. Given past demonstrations that LSC women can be situationally manipulated to behave as HSC women, I expected that when the math-gender stereotype was explicitly activated during a testing situation, both HSC and LSC women would experience stereotype threat. Thus, I predicted that, regardless of stigma consciousness level, all women in the threat condition would underperform on a working memory capacity test relative to control participants. Second, I predicted all participants in the threat condition would spontaneously suppress the female stereotype prior to completing the working memory capacity test. This spontaneous stereotype suppression would be indicated by threat participants’ STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 19 slower reaction times to female stereotypic words relative to control participants on a lexical decision task prior to testing. Given that the female stereotype is more strongly activated for HSC women, I expected that HSC women to require greater suppression of the stereotype than would their LSC counterparts in order to focus on the test. Thus, I predicted that HSC women would avoid stereotypic constructs to a greater extent than would LSC women, and that they would therefore react more slowly than LSC women to stereotypic words. Moreover, I expected HSC women to require more effort to suppress the highly accessible stereotype, as indicated by self-report. Third, I expected that this additional effort would compromise the test performance of HSC women in the stereotype threat condition. I therefore predicted that these women would underperform on the working memory capacity task relative to their LSC counterparts. Furthermore, I expected this additional effort, when compounded with the cognitive demand of the test itself, to undermine the success of stereotype suppression. Due to their greater cognitive load, I expected HSC women to exhibit greater stereotype accessibility than their LSC counterparts once they ceased stereotype suppression. Therefore, my fourth hypothesis was that HSC women in the threat condition would react more quickly than LSC women in the threat condition to stereotypic words on a lexical decision task following the test. No such difference was expected in the control condition. Method Participants The participants were 60 female Hamilton College students who majored in mathematics (46.67%), biology (18.33%), economics (15%), biochemistry (10%), STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 20 chemistry (6.67%), or physics (3.33%). The participants ranged in age from 19 to 23 years (M=21.0, SD=.87), and were predominantly White (83.33%). Participants were recruited through e-mail and were compensated with either $10 or 1 extra credit point in a psychology course of their choice. Materials Math identification items. The math identification items used on the pre-study questionnaire were the same two statements used by Carr and Steele (2009) to assess math identification (“It is important to me that I do well at math” and “I am good at math”). Participants responded to each item along a 7-point Likert-type scale ranging from (strongly disagree) to 7 (strongly agree). Consistent with previous research, participants were required to score above the midpoint on each math identification item to qualify for the experiment (Aronson et al., 1999; Carr & Steele, 2009). Stigma Consciousness Questionnaire for Women (SCQ-W). The SCQ-W (Pinel, 1999) is a 10-item questionnaire that assesses the extent to which women are selfconscious of their stigmatized status (e.g., “Most men have a lot more sexist thoughts than they actually express,” “Stereotypes about women have not affected me personally” (reverse scored)). Participants responded to each item along a 7-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree), and 7 of the items were reverse scored. The mean of the 10 SCQ-W items served as a composite measure of stigma consciousness (Cronbach’s α = .81). Operation Span Task (OSPAN). The OSPAN (Turner & Engle, 1989) is a complex span task that measures working memory capacity. On Microsoft PowerPoint STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 21 software, participants were presented with a mathematical equation (e.g., 36/3 + 6 = 18) paired with a random word (e.g., bus). Each math-word pair was presented for 8 seconds, during which participants had to decide whether the math equation was correct as they rehearsed the word for later recall. These math-word pairs were presented in series that varied in length, ranging from 2 to 6 math-word pairs. At the end of each series, participants were asked to recall all of the words in the series in the order in which they were presented. Participants were given 3 practice sequences before completing 16 test sequences, for a total of 60 correct words. Score on the OSPAN test served as the primary dependent variable. Lexical decision task (LDT). The lexical decision task was used to assess the activation of the female stereotype by assessing the speed with which participants categorized serially-presented letter strings as either words or nonwords. Participants completed one lexical decision task before and one after the OSPAN; each lexical decision task had different words, and order was counterbalanced across participants. Each task included 10 stereotypic words (e.g., indecisive), 10 neutral words (e.g., wood), and 10 nonwords (e.g., stropline). Stereotypic and neutral words were the same as those used in previous studies (Carr & Steele, 2009; Logel et al., 2009). Words in different conditions were matched for length and frequency in the English language using norms established by Kucera and Francis (1967). Nonwords were created using a technique used by Lacruz and Folk (2004) in which real words were split into “heads” and “bodies” (e.g., stake = st + ake), and the heads and bodies of multiple words not used as either stereotypic or neutral letter strings were randomly recombined to produce nonwords. The words used to create the nonwords were randomly selected from an online word STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 22 generator, and nonwords were matched to stereotypic and neutral words in terms of length. Both lexical decision tasks were administered using PsyScope X software (Cohen, MacWhinney, Flatt, & Provost, 1993). Each letter string was presented in black 80 pt. Helvetica font at the center of a white screen for 1000 ms. Participants indicated whether the letter string was a word or a nonword by pressing either the ‘d’ key or the ‘k’ key, respectively. Following a 500 ms delay, the next letter string was signaled by a plus sign (‘+’) that remained on the screen for 500 ms before it was replaced by the next letter string. Stereotype suppression. To assess whether participants attempted stereotype suppression, participants answered the following questions: “Are you aware of the stereotype that women are not as good as men at math?”, “How important was it for you to do well on the working memory capacity test?”, and “While you were taking the test, did you try to avoid thoughts or feelings associated with the gender stereotype about math ability?” (yes/no). If participants reported suppressing the stereotype, they were also asked the following two questions, both assessed on 7-point rating scales: “How hard did you try to avoid thinking about the stereotype?” and, “To what extent did avoiding the stereotype interfere with your test taking ability?”. Procedure Participants who met the previously described criteria arrived at the lab individually, ostensibly for a study examining how math-oriented people process various types of information. Participants provided informed consent and were randomly assigned to the threat or control condition. STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 23 Those in the stereotype threat condition were told that the experimenter was particularly interested in gender differences in processing mathematical information. They were then told that the upcoming working memory capacity test strongly predicted math ability and future success in STEM fields, that men typically outperform women on the test, and that their results would be used to understand why this gender difference exists. In contrast, participants in the control condition were simply reminded that the experimenter was interested in the cognitive processing of math-oriented people. Control participants were then told that test performance was unimportant because the experimenter was only interested in observing what cognitive processes are used during the test, and that the test shows no gender difference in performance; there was no mention of STEM fields or math ability. Participants in both conditions were also told to give their best effort on the working memory capacity test, and that, in addition to the working memory capacity test, they would complete two verbal tasks, one before and one after the test. Following the stereotype threat manipulation, participants were led to a private room in which they sat in front of a Macintosh OS X version 10.5.8 laptop. All tests and questionnaires were completed under these conditions. Participants began the testing phase with the first lexical decision task, which was described as the first verbal task. Participants were instructed to complete the task as quickly and accurately as possible, and to alert the experimenter when they had finished. Participants then completed the OSPAN. Again, participants were instructed to tell the experimenter when they were done. After they were shown the same instructions as the first lexical decision task, participants completed the second, and final, lexical decision task, thus completing the STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 24 testing phase. Participants then completed the stereotype suppression and demographic items. Afterwards, they were thoroughly debriefed, compensated, and thanked for their participation. Results Manipulation Checks On the question assessing whether participants were aware of the math-gender stereotype, 2 out of the 60 participants indicated that they were unaware of the stereotype. Their data as well as the data of 4 other participants who were suspicious of the stereotype threat manipulation were deleted. Importance of doing well. On the item assessing participants’ ratings of the importance of doing well on the OSPAN test, there was a significant main effect for condition, t(53) = -2.30, p = .026 (β = −.30), such that participants in the control condition (M=6.04, SD=.79) rated the test as significantly more important than did those in the threat condition (M=5.42, SD=1.21). Neither the main effect of stigma consciousness nor the stigma consciousness x condition interaction was significant. Stereotype suppression ratings. Again, there was a significant main effect for condition, t(53) = -2.25, p = .029 (β = .30), such that those in the threat condition reported suppressing the stereotype to a greater extent during the OSPAN test than did those in the control condition. Neither the main effect of stigma consciousness nor the stigma consciousness x condition interaction was significant. In addition, there were no significant predictors of participants’ ratings of how effortful or disruptive stereotype suppression was. STEREOTYPE SUPPRESSION AND STIGMA CONSCIOUSNESS 25 Effects of Stereotype Threat on Working Memory Prior to entry in regression equations, stigma consciousness was centered around its mean (M=4.16, SD=.92). In all regressions, the centered stigma consciousness variable, the dummy-coded condition variable (0 = control, 1 = threat), and their interaction were entered as predictors. In the multiple regression analysis predicting OSPAN performance (Figure 1), there was a significant main effect for condition, t(53) = -2.03, p = .048 (β = -.26), such that, as expected, threat participants performed more poorly (M=16.62, SD=10.85) than did control participants (M=22.75, SD=12.46). However, this main effect was qualified by a significant interaction between condition and stigma consciousness, t(53) = -2.11, p = .04 (β = -.43). Consistent with my hypothesis, among participants in the threat condition, there was a significant negative relationship between stigma consciousness and OSPAN score (β = -.57, p=.003), whereas in the control condition, stigma consciousness was unrelated to OSPAN score (β = .08,