Theories of arousal suggest that arousal should decrease perfor- mance on difficult tasks and increase performance on easy tasks. An experiment tested the hypothesis that the effects of stereotype threat on performance are due to heightened arousal. The authors hypothesized that telling participants that a math test they are about to take is known to have gender differences would cause stereotype threat in women but not in men. In the experi- ment, each participant took two tests—a difficult math test and an easy math test. Compared to women in a “no differences” con- dition, women in the “gender differences” condition scored better on the easy math test and worse on the difficult math test. Men’s performance was unaffected by the manipulation. These data are consistent with an arousal-based explanation of stereotype threat effects. Data were inconsistent with expectancy, evalua- tion apprehension, and persistence explanations of the stereotype threat phenomenon.

Keywords: stereotype threat; arousal; gender; math

When people are members of a group that can be neg- atively stereotyped in a social situation, they carry an extra burden: their performance might well be inter- preted in terms of the stereotype. Should the perfor- mance be consistent with the stereotype (e.g., a woman scoring poorly on a math test, an upper-class White man being clumsy on the dance floor), the behavior serves to confirm the stereotype in the eyes of the beholders; this phenomenon has been labeled stereotype threat (Spencer, Steele, & Quinn, 1999; Steele & Aronson, 1995). Stereotype threat has proven to be a robust phe- nomenon, applying to a diverse set of groups, including African Americans, White men, women, and people with low socioeconomic status and social class (Aronson et al., 1999; Leyens, Desert, Croizet, & Darcis, 2000; Quinn & Spencer, 2001; Spencer et al., 1999; Steele & Aronson,

1995; Walsh, Hickey, & Duffy, 1999. See Steele, Spencer, & Aronson, 2002, for a review.).

For a person to experience stereotype threat, she or he need only to have knowledge that some people hold a negative stereotype about their group. This knowledge can make a person worried about being viewed stereotypically, even when she or he does not endorse the stereotype. Although the psychological state of ste- reotype threat is an important issue (e.g., Aronson, Quinn, & Spencer, 1998; Croizet & Claire, 1998; Lee & Ottati, 1995; Oswald & Harvey, 2000-2001; Spencer et al., 1999; Steele & Aronson, 1995), much of the research has focused on the demonstration of the behavioral effects of stereotype threat and the range of conditions when it might occur (Aronson et al., 1999; Inzlicht & Ben-Zeev, 2000; Marx, Brown, & Steele, 1999; Quinn & Spencer, 2001; Stangor, Carr, & Kiang, 1998; Steele, 1997; Steele & Aronson, 1998; Stone, Lynch, Sjomeling, & Darley, 1999).

In a typical stereotype threat experiment, participants who are aware of being at risk for behaving consistently with a negative stereotype (e.g., being poor at math, ath- letics, academics, etc.) perform significantly more poorly than participants who do not feel that they are threatened by the stereotype. Removing stereotype threat can have a large effect; Spencer et al. (1999) erased gender differences in complex mathematics per- formance by describing their math test as one that did not produce gender differences, and Aronson et al. (1998) eliminated racial differences in math perfor- mance by presenting the test as “nondiagnostic.” In

782

Authors’ Note: Please address correspondence to Laurie T. O’Brien, Department of Psychology, University of California, Santa Barbara, CA 93106; e-mail: [email protected].

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/

these experiments, the removal of stereotype threat completely removed group differences, suggesting that many of the racial or gender group differences that appear to characterize groups are in fact the result of ste- reotype threat—an environmental challenge rather than an essential difference.

The fact that stereotype threat harms task perfor- mance on difficult tasks is well established (Aronson et al., 1999; Croizet & Claire, 1998; Spencer et al., 1999; Steele & Aronson, 1995). Although Steele and Aronson (1995) argue that stereotype threat is most likely to occur when the test is difficult, they acknowledge that in the case of African Americans and the stereotype of intel- lectual inferiority, the mere “act of taking a test pur- ported to measure intellectual ability may be enough to induce this threat” (p. 798). Thus, difficult tasks may not be an essential component to the experience of stereo- type threat.

Researchers have also argued that stereotype threat can occur even when the task is not difficult.

A test need not be difficult for stereotype threat to occur. Simply being in a situation where one can confirm a neg- ative stereotype about one’s group—women simply sit- ting down to the math test, for example, could be enough to cause this self-evaluative threat. (Spencer et al., 1999, p. 7)

Even when the task is easy, members of negatively stereo- typed groups are in a predicament not faced by others. “In situations where math skills are exposed to judg- ment—be it a formal test, classroom participation, or simply computing the waiter’s tip—women bear the ex- tra burden of having a stereotype that alleges a sex-based inability” (Spencer et al., 1999, p. 6).

Members of negatively stereotyped groups are fre- quently in situations where the difficulty of the task is unknown. Although it is certainly possible that threat is attenuated once people learn that the task they are to perform is easy, this knowledge may not completely negate the early effects of the threat (e.g., Zillman, 1972).

How does stereotype threat lead to decreased perfor- mance on difficult tasks? Steele and Aronson (1995) found that African American participants under condi- tions of stereotype threat showed increased activation of stereotypes about African Americans and more self- doubt than African Americans who were not under stereo- type threat and White Americans. Steele and Aronson argue that these findings are evidence of a disruptive evaluation apprehension in stereotype-threatened par- ticipants. Spencer et al. (1999) showed that stereotype threat increased anxiety, but they did not find conclusive evidence that anxiety, self-efficacy, or evaluation appre-

hension mediated the link between stereotype threat and performance in spite of the fact that anxiety, self-effi- cacy, and evaluation apprehension have been clearly linked to decreased performance in other literatures (e.g., Ashcraft, 2002).

Stereotype threat is not limited to low-status groups; Aronson et al. (1999) showed that the performance of White men on math was harmed when threatened with stereotypes of Asian superiority in math, and Leyens et al. (2000) showed that men experienced stereotype threat on a task that involved affective processing. Steele and Aronson (1995) suggested that stereotype threat might be mediated by low performance expectations (which could affect dominant and subordinate groups equally) but Brown and Steele (1999, cited in Marx et al., 1999) found that increasing performance expectations of African Americans did not remove the effects of ste- reotype threat.

The active ingredient in stereotype threat has not been definitively identified. Because of the importance of the phenomenon, and the complexity of social inter- actions between groups in conflict, we suggest that ste- reotype threat’s effect works through a variety of path- ways. Any explanation of the phenomenon should parsimoniously appeal to psychological processes that are common to women and men, Blacks, Whites, and Asians in the domain of mathematics, language, and sports—in short, it should be based in fundamental and common psychological processes. We suggest that one common and powerful psychological process that might contribute substantially to stereotype threat’s effects on performance is heightened arousal. High levels of arousal have been widely shown to interfere with com- plex and difficult tasks (e.g., Davis & Harvey, 1992; Markus, 1978; Spence, 1956; Zajonc, 1965, 1969) and operating under the fear and anxiety of confirming a negative stereotype, or being categorized as an exemplar of a negative stereotype, is sufficient to create arousal. High arousal should interfere with performance.

Does the arousal caused by stereotype threat impact perfor- mance? In our view, arousal is heightened activity, primar- ily in the sympathetic nervous system (SNS), that ener- gizes behavior. Arousal, through activation of the SNS, increases heart rate, systolic blood pressure, sweating, and causes the release of epinephrine. Arousal is non- specific and serves to energize behavior in a nondirective way. For this reason, arousal enhances the emission of dominant responses. When the task is complex or diffi- cult, arousal will lead to an increased number of inap- propriate responses, thereby creating a decrement in overall performance. On the other hand, when the task is simple, or well learned, arousal will lead to an increased number of appropriate responses thereby increasing overall performance (Bolles, 1967; Markus,

O’Brien, Crandall / STEREOTYPE THREAT AND AROUSAL 783

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/

1978; Spence, 1956; Weiner, 1985; Wright, Murray, Storey, & Williams, 1997; Zajonc, 1965; Zillman, 1972).

Much of the research on stereotype threat is consis- tent with an arousal explanation for stereotype threat effects. Most research involving stereotype threat has uti- lized difficult tasks (e.g., GRE verbal and math tests and GMAT math tests). This research has shown that people experiencing stereotype threat (and are presumably in a higher state of arousal) do more poorly on difficult tests (e.g., Aronson et al., 1999; Croizet & Claire, 1998; Spencer et al., 1999; Steele & Aronson, 1995).

A study by Blascovich, Spencer, Quinn, and Steele (2001) found that stereotype threat elevated blood pres- sure in American Blacks. Black and White Americans took an “intelligence test” while researchers measured their cardiovascular activity. To enhance stereotype threat, participants first watched a video of a White pro- fessor who presented the test as “an intelligence test developed at Stanford.” To reduce threat, a video with a Black professor presented the test as racially unbiased. Under conditions of stereotype threat, Blacks’ blood pressure rose more quickly and to a higher level than Whites, but under conditions of reduced threat, Blacks did not have higher blood pressure than Whites. These data suggest that stereotype threat creates arousal (see Wright et al., 1997, for a similar finding).

One important implication of an arousal theory of ste- reotype threat is that although performance on difficult tasks should be harmed by arousal, performance on easy or well-learned tasks should be enhanced. To date, only Spencer et al. (1999) have paid close attention to the relationship between task difficulty and stereotype threat. They gave male and female participants an extremely difficult test taken from the advanced GRE exam in mathematics and a moderately difficult test taken from the quantitative section of the GRE general exam. Whereas threat reduced women’s performance on the extremely difficult test, there was no difference between women in the two conditions on performance on the moderately difficult test. Unfortunately, this study does not offer a good test of the arousal hypothesis because it does not include a truly easy test. However, the results from Spencer et al.’s (1999) Study 1 are consis- tent with the arousal hypothesis; whereas arousal should increase performance on easy tasks and decrease perfor- mance on difficult tasks, it should not affect perfor- mance on tasks of moderate difficulty such as the test used in the Spencer et al. (1999) research.

Although there has not been a direct test of how ste- reotype threat affects performance on easy tasks, there is some research that is consistent with the arousal hypoth- esis that threat should lead to increased performance on easy tasks. Bielinski and Davidson (1998) analyzed math competency tests of more than 10,000 eighth-grade stu-

dents and more than 5,000 ninth-grade students; they found that men tended to perform better on the most difficult items, whereas women tended to perform better on the easiest items. This is the exact pattern of results an arousal explanation would predict for women under conditions of stereotype threat if one assumes that ste- reotype threat is “in the air” (Steele, 1997).

Overview and hypotheses. We suggest that a substantial active component of stereotype threat is increased arousal. Because arousal harms performance on diffi- cult tasks but improves performance on easy tasks, we predict that when compared to conditions of no stereo- type threat, people under stereotype threat will perform more poorly on difficult tasks but they will improve on easy tasks.

We conducted a study in which women were put in conditions of stereotype threat or were put in conditions that attenuated stereotype threat and then presented them both difficult and easy math problems. We pre- dicted that the stereotype of male superiority on math would create stereotype threat for women only, and for them only in the presence of purported gender differ- ences. Arousal from this threat would decrease math performance on difficult items and increase math per- formance on easy items for women. Because men are not at risk for confirming a negative stereotype about their math ability, we predicted that the stereotype threat manipulation would have no discernable effect on men’s math performance, regardless of test difficulty. Statis- tically, we predicted a three-way interaction between the stereotype threat manipulation, participant gender, and test difficulty.

Method

Participants. Participants in the experiment were 164 students (59 women, 105 men) enrolled in an introduc- tory psychology class who participated in exchange for partial course credit. The ethnic background of the par- ticipants included 7 African Americans, 6 Asian Ameri- cans, 4 Hispanics/Latinos, 2 Indians/Central Asians, 4 Native Americans, 137 Whites, and 4 people who indi- cated that they were of mixed ethnicity.1 Participants took part throughout two semesters in the spring and fall of 1999.

Procedure. After participants arrived at the laboratory and consented to participate in the study, the experi- menter handed participants a packet of materials that included tests of math ability. The cover sheet informed the participants that the researchers were evaluating a series of tests; the participants would be working on mathematical tests.

The participants read that previous research had sometimes “shown gender differences and sometimes

784 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/

shown no gender differences.” In the gender differences condition, we used the same manipulation adopted in previous studies (e.g., Spencer et al., 1999); participants were told, “The test you are about to take has been shown to produce gender differences.” Participants in the no differences conditions were told, “The test you are about to take has NOT been shown to produce gender differ- ences.” Spencer et al. (1999) argued that participants interpret information about gender differences in math tests to mean that men outperform women—informa- tion from debriefing suggested that our participants had the same interpretation.

Participants then filled out a brief questionnaire regarding their feelings prior to taking the tests. For the last 94 participants, we added a self-report measure of motivation and a five-item measure of evaluation appre- hension (taken from Spencer et al., 1999). After filling out this questionnaire, the experimenter told partici- pants how much time that they would have to complete the first test. One minute prior to the end of the test, the experimenter informed participants that they had 1 minute left to complete the test. When time was up, the experimenter asked the participants to stop working on the test.

The same procedure followed for the subsequent tests. All participants took three tests: a fairly easy test, a fairly difficult test, and a very simple math test to mea- sure effort and persistence (Biernat, 1989).2 The order of the tests was counterbalanced between participants. After completing the tests, participants were debriefed and dismissed.

Materials. Participants in the study completed three tests: an easy math test, a difficult math test, and a test to measure math persistence. The easy test consisted of three-digit multiplication problems; for example, multi- plying 143 by 495. Participants were not allowed to use calculators and were given 10 min to complete 20 problems.

The difficult test was 15 items from the quantitative SAT. Questions were given in a five-option, multiple- choice format and participants had 11 min to complete the questions. An example is as follows:

If x2 + y2 = 2xy, the x must equal:

A. –1 B. 0 C. 1 D. –y E. y

The correct answer is E. Three items had more than 90% of the sample answering correctly across condition; because the test was designed to be difficult, these items were deleted. Scores were calculated by giving partici- pants one point for each correct answer and correcting for guessing by subtracting one fifth of a point for each incorrect answer (Spencer et al., 1999).

The test to measure math persistence employs mental math. Participants were instructed to answer addition and subtraction problems in their head and write down the answer. Participants were allotted 8 min to complete 24 questions. The mental math test requires some con- centration and effort but is mathematically very easy, involving only addition and subtraction. Previous researchers (e.g., Biernat, 1989) have used the number of answers that participants get correct on this test as a measure of motivation, effort, and persistence.

Ninety-four participants also filled out measures of motivation and evaluation apprehension prior to com- pleting the tests. The motivation item asked participants how motivated they were to perform well on the test. Par- ticipants responded on a 100-point scale with 0 labeled not at all motivated and 100 labeled extremely motivated. The evaluation apprehension measure included five items taken from Spencer et al. (1999). Items include: “If I do poorly on this test, people will look down on me” and “If I don’t do well on this test, others may question my ability” (α = .75).

Results

Order effects. A mixed-model ANOVA with test-diffi- culty as a within-subjects factor (easy vs. hard) and order, sex, and condition as between-subjects factors found no significant main effects or interactions of test order on test performance. Order was deleted from further analyses.

Effects of threat manipulation and gender on performance. The main hypothesis test is the three-way interaction between gender, threat manipulation, and math diffi- culty. Characterizing a test as sensitive to gender differ- ences should increase women’s performance on an easy test, decrease women’s performance on a difficult test, and leave men unaffected by test difficulty. To make test scores comparable across test difficulty, we z transformed participants’ test scores by subtracting the overall sample mean from each participants’ score and dividing this amount by the sample standard deviation (see Johnson, 2000).3 The data were submitted to a mixed-model ANOVA; math difficulty (easy vs. hard) was a within- subjects factor and gender and threat manipulation (gender differences vs. no differences) were between- subjects factors. This analysis revealed the predicted three-way interaction between gender, threat, and test difficulty, F(1, 160) = 5.46, p < .05. The pattern of means is displayed in Figure 1; it supports the arousal hypothe- sis. Compared to the no threat conditions, stereotype threat lowered difficult math per formance but improved easy math performance among women.

To investigate the nature of the interaction, we car- ried out planned comparisons that offered a proper, direct test of the hypothesis. On the easy test, women in

O’Brien, Crandall / STEREOTYPE THREAT AND AROUSAL 785

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/

the gender differences condition performed better (M = +.20) than did women in the no differences condition (M = –.25), t(57) = 1.82, p < .05, one-tailed, d = .47. Men’s performance on the easy test did not vary as a function of condition; men in the gender differences condition (M = .05) did not perform significantly different from men in the no differences condition (M = .00), |t| < 1, ns. No other relevant pairwise comparisons on the easy test were significant.

On the difficult math test, women in gender differ- ences performed worse than women in no differences (Ms = –.64, –.17, respectively), t(57) = 1.79, p < .05, one- tailed, d = .46. Men showed no difference by condition, |t| < 1. These results mimic the standard stereotype threat result; removal of threat improves performance on diffi- cult tasks for the stereotyped group but does not touch the unstereotyped.

Men outperformed women in the gender differences condition on the difficult math test, t(160) = –3.81, p < .001, d = .82, and performed slightly better than women in no differences condition, t(80) = –1.85, p < .10, two- tailed, d = .39. The difference between men and women in the gender differences condition was more than twice the size of the difference between men and women in the no differences condition.

The ANOVA also revealed a significant main effect of sex, F(1, 160) = 6.69, p < .05, a Sex × Test Difficulty inter- action, F(1, 160) = 10.02, p < .01, and a Condition × Test Difficulty interaction, F(1, 160) = 7.61, p < .05. The pat- tern of these effects is subsumed by the three-way interac- tion, and these effects are visible in Figure 1.

Subsidiary Analyses

Problems attempted versus accuracy. The manipulation increased women’s scores on the easy math test and

decreased their scores on the difficult math test while leaving the men unaffected. Is it possible that the manip- ulation affected test scores simply because women under threat attempted greater problems on the easy test and fewer problems on the difficult test, or did the manipula- tion affect participants’ accuracy? We planned and con- ducted further analyses on the women’s data to deter- mine whether accuracy or the number of problems attempted contributed to differences in overall test scores. To create a measure of accuracy for each partici- pant, for the easy and the difficult tests we divided the number of problems correct by the number of problems attempted; this proportion score was transformed by tak- ing the arcsine of participants’ proportion score (Cohen & Cohen, 1975). To equate accuracy on both tests, the analyses employed z scores. Because only female partici- pants’ scores are examined in the subsidiary analyses, participants’ accuracy scores were transformed by sub- tracting the sample mean for women from each partici- pant’s accuracy score and dividing this amount by the sample standard deviation for women. Separate analyses performed on accuracy scores without the arcsign trans- formation and on accuracy scores transformed using the overall sample means and standard deviations yielded nearly identical results.

If arousal affects accuracy directly, rather than indi- rectly, through greater effort and persistence (measured by speed), then women in the gender differences condi- tion should perform more accurately on the easy test and less accurately on the difficult test, resulting in a two- way interaction between test difficulty and threat manip- ulation. The test of this hypothesis, a Test Difficulty × Threat Manipulation mixed-model ANOVA, yielded the predicted two-way interaction, F(1, 57) = 11.27, p < .01. As expected, when the test was easy, women in the gen- der differences (M = +.21) condition were more accurate than women in the no differences condition (M = –.19). When the test was difficult, however, women in the gen- der differences condition (M = –.27) were less accurate than women in the no differences condition (M = +.25). To further explore these results, we tested the signifi- cance of the differences between the conditions; when the test was difficult, the difference between the two con- ditions was significant, t(57) = 2.04, p < .05, one-tailed, and when the test was easy, the difference between the two conditions was nearly significant, t(57) = 1.58, p = .06, one-tailed. Threat directly affected accuracy.

To test whether improved performance might be due to enhanced effort due merely to greater numbers of problems attempted, we analyzed the number of prob- lems attempted using a Test Difficulty × Threat Manipu- lation mixed-model ANOVA. This analysis did not yield any significant effects, ps > .42. These results suggest that stereotype threat enhances accuracy on easy problems,

786 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

-1

-0.75

-0.5

-0.25

0.25

0.5

Women Gender

Difference

Women No Difference

Men Gender

Difference

Men No Difference

Easy Test Hard Test

T es

t P

er fo

rm an

Figure 1 Math performance of women and men on easy and difficult math problems by stereotype threat condition.

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/

harms accuracy on difficult problems, and that these dif- ferences are not a function of the number of problems attempted but rather on improved performance.

Problem difficulty and effect size. All the problems on the easy math test involved multiplication of three-digit numbers and were of comparable difficulty. The prob- lems on the difficult math test, however, were algebra problems of varying difficulty. The percentage of partici- pants answering a question correctly ranged from 28% (difficult problem) to 93% (relatively easy problem). Previous analyses have show that women in the no differ- ences condition outperform women in the gender dif- ferences condition on the difficult math test; the arousal hypothesis suggests that this difference should grow larger as problem difficulty increases. For men, any dif- ferences between conditions should be unrelated to problem difficulty.

To test this hypothesis, we examined “problem” as the unit of analysis for all 15 problems included in the alge- bra test. Problem difficulty was calculated by taking the arcsine of the sample mean for each problem. To aid in interpretation, scores were reverse-coded so that increasing numbers reflect increasing difficulty; a prob- lem difficulty score of “0” would indicate that all partici- pants answered a problem correctly. Next, the effect of the manipulation on each problem was calculated as the effect size “d” such that positive scores indicated partici- pants in the no differences condition were outperform- ing participants in the gender differences condition and negative scores indicated the reverse.

As predicted, the correlation between problem diffi- culty and effect size for women was significant and posi- tive, r = +.59, p < .03, whereas the correlation for men was nonsignificantly negative, r = –.03, p > .90. As Figure 2 shows, on the most difficult problems, female partici- pants in the no difference condition far outscored female participants in the gender differences. It is also worthwhile to note that consistent with arousal predic- tions, female participants in the gender differences con- dition outperformed female participants in the no dif- ferences condition on some of the easiest problems.4

Self-reports of motivation. Participants’ self-reports of motivation measured prior to beginning the math tests were submitted to a Sex × Manipulation ANOVA. This analysis yielded no significant effects of self-reported motivation on performance.

Ruling Out Nonarousal Explanations

Math persistence test. We tested whether the manipula- tion had an effect on math persistence. Five participants were deleted from this analysis for failure to follow direc- tions. A Sex × Condition ANOVA performed on math persistence test revealed no main effect of condition,

F(1, 155) = 1.08, ns. There was, however, a main effect of sex, F(1, 155) = 12.42, p < .01, with men correctly answer- ing more items (M = 19.1) than women (M = 15.8). The interaction between sex and condition was not signifi- cant, F < 1. These data suggest that sheer persistence does not explain the improvement on easy items for women in the no differences condition.

Evaluation apprehension. We submitted the evaluation apprehension scale to a Sex × Condition ANOVA. We found a main effect of sex, F(1, 90) = 5.29, p < .05, with women (M = 3.0) showing higher levels of evaluation apprehension than men (M = 2.5). There was no effect of condition and no interaction, both Fs < 1. Evaluation apprehension was uncorrelated with performance on the difficult test, performance on the easy test, and the measure of persistence.

DISCUSSION

Women under stereotype threat performed better on an easy math test and worse on a difficult math test than women who were not exposed to stereotype threat. These effects on women’s performance affected accu- racy (performance) but not the number of problems attempted (effort).

Research on arousal suggests that arousal enhances performance on easy tasks and harms performance on difficult tasks (Zajonc, 1965). Stereotype threat research has reliably demonstrated that the threat of confirming a negative stereotype harms performance on difficult tasks. We replicated this effect but also showed that ste- reotype threat can enhance performance on an easy task. We have argued that the possibility of confirming a nega- tive stereotype can cause arousal and that this arousal accounts for stereotype threat effects on performance. This hypothesis suggests that in addition to the well-

O’Brien, Crandall / STEREOTYPE THREAT AND AROUSAL 787

-0.05

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Problem Difficulty

P er

fo rm

an ce

i n

N o

D if

fe re

nc es

- P

er fo

rm an

ce i

n G

en de

r D

if fe

re nc

Females Males

- 1 SD + 1 SD

Figure 2 Predicting effect size (d) for men and women as a function of problem difficulty.

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/

established decrements in difficult tasks, stereotype threat should improve performance on easy tasks. The data supported the hypothesis. The present research provides empirical support for Steele and Aronson’s (1995) suggestion that arousal may be one causal mecha- nism for stereotype threat effects (Easterbrook, 1959).

How does stereotype threat lead to decreased perfor- mance on difficult tasks? Some have suggested that lower self-confidence and performance expectancies can be triggered by situationally relevant negative stereotypes (e.g., Stangor et al., 1998). Past stereotype threat researchers have been unable to effectively rule out the possibility that stereotype threat is caused by reduced expectancies. These data suggest that negative expectan- cies are not solely responsible for stereotype threat effects—negative expectancies should not elevate the performance of stereotype-threatened individuals on easy tasks; they should reduce them.

The current study also lends further evidence to the notion that stereotype threat effects are not mediated by evaluation apprehension. Evaluation apprehension was unaffected by the manipulation—instead, we found only that women had greater evaluation apprehension than men. In addition, evaluation apprehension was uncorrelated with performance in either condition. Consistent with research reported by Marx et al. (1999), the present research also suggests that the effects of ste- reotype threat on performance are not due to differ- ences in persistence; women’s math persistence did not vary by condition.

Recent theorizing on motivation might explain why women would perform poorly on difficult tasks and well on easier tasks. According to Brehm’s theory of motiva- tional intensity (Brehm, 1999; Brehm & Self, 1989), when a woman encounters a math task she perceives as manageable and expects success (such as multiplication problems), she will have approximately as much motiva- tion as needed to do well on the test. However, if the test is perceived as difficult enough that she is unwilling or unable to invest the effort required to succeed, then the motivation levels created by the stereotype threat manip- ulation will suffer sharp declines along with her perfor- mance. Thus, motivation is affected by the difficulty of the task. In the current study, participants’ self-reports of motivation were measured before participants had viewed the math test and when they had no knowledge of the test difficulty. Therefore, the present research does not offer a proper test of Brehm’s model—future research should measure participants’ motivation after learning of the task’s difficulty in order to examine a motivational explanation of stereotype threat effects on performance.

Stereotype researchers have raised the issue of whether participants must be consciously aware of nega-

tive stereotypes and their relevance to a performance sit- uation for stereotype threat effects to occur (Wheeler & Petty, 2001). For example, the results of Levy’s (1996) research with subliminal priming and older adults sug- gest that negative stereotypes can affect behavior with- out participants’ conscious awareness. Levy, Hausdorff, Hencke, and Wei (2000) showed that priming negative stereotype of aging can increase heart rate in older par- ticipants even when participants are unaware that nega- tive stereotypes have been primed. An arousal-based explanation for the effects of stereotype threat on behav- ior can parsimoniously account for circumstances under which people are aware of negative stereotypes and situa- tions in which stereotypes have been primed without people’s conscious awareness.

Early thinking on stereotype threat was groundbreak- ing because it suggested a situational explanation for gender differences in mathematics abilities and racial differences in standardized test scores—two findings that have frequently been chalked up to genetic or other biological bases. Stereotype threat research, including the current investigation, suggests that the situation mat- ters. Men and women may take math tests in the same room but they are not in the same situation.

In an atmosphere laden with stereotypes of incompe- tence, the present research suggests that people who are members of negatively stereotyped groups are likely to feel increased arousal. This arousal can interfere with people’s ability to succeed at difficult tasks. Teaching people skills to cope with arousal (e.g., relaxation tech- niques, planning) may prove an effective weapon against the detrimental effects of stereotype threat.

NOTES

1. Excluding participants who are not White had no effect on the pattern of results.

2. Originally, a fourth, very difficult test was included with the items from the math portion of the GMAT used by Spencer, Steele, and Quinn (1999). When preliminary data analyses were run on the first 71 participants, we discovered that participants of both genders in both conditions were scoring at a level no different from chance. Therefore, we eliminated this test from the procedure for the subsequent 94 par- ticipants. Spencer et al.’s participants were preselected for having above-average ability in math, whereas the participants in the current study were from general psychology classes. This fact most likely explains why Spencer et al. found differences between conditions using this test while we found all participants scoring near a chance level.

3. Analyses using untransformed test scores yielded a nearly identi- cal analysis. Transforming test scores for female participants using females’ sample mean and standard deviation and for male partici- pants using males’ sample mean and standard deviation also yielded a nearly identical analysis, but by definition, eliminated any main effect of sex. To compare men’s and women’s scores, we opted to transform test scores using the overall mean and standard deviation.

4. We also calculated the correlations using only the 12 problems that were retained in the final analysis of the test. The correlations were nearly identical, r = +.57 for women, r = –.04 for men. Finally, we also cal- culated separate problem difficulty scores for men and women. Again, this had little effect on the analyses; the correlation between effect size

788 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/

and problem difficulty for female participants was r = +.58, for men, r = –.03.

REFERENCES

Aronson, J., Lustina, M. J., Good, C., Keough, K., Steele, C. M., & Brown, J. (1999). When White men can’t do math: Necessary and sufficient factors in stereotype threat. Journal of Experimental Social Psychology, 35, 29-46.

Aronson, J., Quinn, D. M., & Spencer, S. J. (1998). Stereotype threat and the academic underperformance of minorities and women. In J. K. Swim & C. Stangor (Eds.), Prejudice: The target’s perspective (pp. 83-103). San Diego, CA: Academic Press.

Ashcraft, M. H. (2002). Math anxiety: Personal, educational, and cog- nitive consequences. Current Directions in Psychological Science, 11, 181-185.

Bielinski, J., & Davidson, M. L. (1998). Gender differences by item dif- ficulty interactions in multiple-choice mathematics items. Ameri- can Educational Research Journal, 35, 455-476.

Biernat, M. (1989). Motives and values to achieve: Different con- structs with different effects. Journal of Personality, 57, 69-95.

Blascovich, J., Spencer, S. J., Quinn, D., & Steele, C. (2001). African Americans and high blood pressure: The role of stereotype threat. Psychological Science, 12, 225-229.

Brehm, J. W. (1999). The intensity of emotion. Personality and Social Psychology Review, 3, 2-22.

Brehm, J. W., & Self, E. A. (1989). The intensity of motivation. In M. R. Rosenzweig & L. W. Porter (Eds.), Annual review of psychology (Vol. 40, pp. 109-131). Palo Alto, CA: Annual Reviews.

Bolles, R. C. (1967). Theory of motivation. New York: Harper & Row. Cohen, J., & Cohen, P. (1975). Applied multiple regression/correlation

analysis for the behavioral sciences. Hillsdale, NJ: Lawrence Erlbaum. Croizet, J. C., & Claire, T. (1998). Extending the concept of stereo-

type and threat to social class: The intellectual underperformance of students from low socioeconomic backgrounds. Personality and Social Psychology Bulletin, 24, 588-594.

Davis, M. H., & Harvey, J. C. (1992). Declines in major league batting performance as a function of game pressure: A drive theory analy- sis. Journal of Applied Social Psychology, 22, 714-735.

Easterbrook, J. A. (1959). The effect of emotion on cue utilization and the organization of behavior. Psychological Review, 66, 183-201.

Inzlicht, M., & Ben-Zeev, T. (2000). A threatening intellectual envi- ronment: Why females are susceptible to experiencing problem- solving deficits in the presence of males. Psychological Science, 11, 365-371.

Johnson, C. (2000, May). Stereotype threat in adolescents. Paper pre- sented at the annual meeting of the Midwestern Psychological Association, Chicago.

Lee, Y. T., & Ottati, V. (1995). Perceived in-group homogeneity as a function of group membership salience and stereotype threat. Personality and Social Psychology Bulletin, 21, 610-619.

Levy, B. (1996). Improving memory in old age through implicit self- stereotyping. Journal of Personality and Social Psychology, 71, 1092- 1107.

Levy, B. R., Hausdorff, J. M., Hencke, R., & Wei, J. Y. (2000). Reducing cardiovascular stress with positive self-stereotypes of aging. Jour- nals of Gerontology: Series B: Psychological Sciences and Social Sciences, 55B, 205-213.

Leyens, J. P., Desert, M., Croizet, J. C., & Darcis, C. (2000). Stereotype threat: Are lower status and history of stigmatization precondi-

tions of stereotype threat? Personality and Social Psychology Bulletin, 26, 1189-1199.

Markus, H. (1978). The effect of mere presence on social facilitation: An unobtrusive test. Journal of Experimental Social Psychology, 14, 389-397.

Marx, D. M., Brown, J. L., & Steele, C. M. (1999). Allport’s legacy and the situation press of stereotypes. Journal of Social Issues, 55, 491- 502.

Oswald, D. L., & Harvey, R. D. (2000-2001). Hostile environments, ste- reotype threat, and math performance among undergraduate women. Current Psychology: Development, Learning, Personality, Social, 19, 338-356.

Quinn, D. M., & Spencer, S. J. (2001). The interference of stereotype threat with women’s generation of mathematical problem-solving strategies. Journal of Social Issues, 57, 55-71.

Spence, K. W. (1956). Behavior theory and conditioning. New Haven, CT: Yale University Press.

Spencer, S. J., Steele, C. M., & Quinn, D. M. (1999). Stereotype threat and women’s math performance. Journal of Experimental Social Psy- chology, 35, 4-28.

Stangor, C., Carr, C., & Kiang, L. (1998). Activating stereotype under- mines task performance expectations. Journal of Personality and Social Psychology, 35, 4-28.

Steele, C. M. (1997). A threat in the air: How stereotype shape intel- lectual identity and performance. American Psychologist, 52, 613- 629.

Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellec- tual test performance of African Americans. Journal of Personality and Social Psychology, 69, 797-811.

Steele, C. M., & Aronson, J. (1998). Stereotype threat and the test per- formance of academically successful African Americans. In C. Jencks & M. Phillips (Eds.), The Black-White test score gap (pp. 401- 427). Washington, DC: Brookings Institution.

Steele, C. M., Spencer, S. J., & Aronson, J. (2002). Contending with group image: The psychology of stereotype and social identity threat. In M. P. Zanna (Ed.), Advances in experimental social psychol- ogy (Vol. 34, pp. 379-440). New York: Academic Press.

Stone, J., Lynch, C. I., Sjomeling, M., & Darley, J. M. (1999). Stereo- type threat effects on Black and White athletic performance. Jour- nal of Personality and Social Psychology, 77, 1213-1227.

Walsh, M., Hickey, C., & Duffy, J. (1999). Influence of item content and stereotype situation on gender differences in mathematical problem solving. Sex Roles, 41, 219-240.

Weiner, B. (1985). Human motivation. New York: Springer-Verlag. Wheeler, C. S., & Petty, R. E. (2001). The effects of stereotype activa-

tion on behavior: A review of possible mechanisms. Psychological Bulletin, 127, 797-826.

Wright, R. A., Murray, J. B., Storey, P. S., & Williams, B. J. (1997). Abil- ity analysis of gender relevance and sex differences in cardiovascu- lar response to behavioral challenge. Journal of Personality and Social Psychology, 73, 405-417.

Zajonc, R. B. (1965). Social facilitation. Science, 149, 269-274. Zajonc, R. B. (1969). Social enhancement and impairment of perfor-

mance in the cockroach. Journal of Personality and Social Psychology, 13, 83-92.

Zillman, D. (1972). Excitation transfer from physical exercise to sub- sequent aggressive behavior. Journal of Experimental Social Psychol- ogy, 8, 247-259.

Received October 15, 2002 Revision accepted November 7, 2002

O’Brien, Crandall / STEREOTYPE THREAT AND AROUSAL 789

at UH-Downtown on August 25, 2010psp.sagepub.comDownloaded from

http://psp.sagepub.com/