Running head: AROUSAL AND STEREOTYPE THREAT
What is stereotype threat? An investigation of the mediating role of arousal in threatening intellectual environments
Talia Ben-Zeev
San Francisco State University
Michael Inzlicht
New York University
Steven Fein
Williams College
Key Words: Sex differences in math performance, stereotype threat, arousal, misattribution
Word count: xxx words
Address of correspondence:
Talia Ben-Zeev
Psychology Department
Williams College
Williamstown, MA 01267
Abstract
The mere presence of even one male can undermine the problem-solving performance of highly skilled female students—an effect attributed to stereotype threat (Inzlicht & Ben-Zeev, 2000). In the current set of studies, we examine whether physiological arousal is a mediator of this effect. According to the Yerkes-Dodson (1908) theory of physiological arousal, performance is optimal at intermediate levels of arousal and decreases when arousal is either low or high, resulting in an inverted-U shaped function. We propose that stereotype threat may interfere with performance by leading to arousal that exceeds an optimal level. Participants in both experiments were asked to take a difficult math test with either two people of the same sex (same-sex group) or with two people of the opposite sex (minority group). In experiment 1, half of the women were given an external stimulus to which they could misattribute their arousal, and half were not. As predicted by the arousal hypothesis, stereotype-threat performance deficits were attenuated significantly when females were given an opportunity to misattribute their arousal. In experiment 2, participants were divided into high- and low- identifiers in mathematics. Results show a cross-over interaction between the identification and threat conditions, such that high-identifiers in the minority condition showed problem-solving deficits whereas low-identifiers in the minority condition showed surfeits in performance. Taken together, the results of both experiments support the arousal theory of stereotype threat.
What is stereotype threat? An investigation of the mediating role of arousal in threatening intellectual environments
The social context can prime particular aspects of one’s self, which, in turn, may hinder or facilitate intellectual performance. For example, Asian American females who were primed with being Asian performed better on a math test than Asian American females who were primed with being female (Shih, Pittinsky, & Ambady, 1999). The main theoretical construct, which has been used to explain how social context contributes to creating, perpetuating, or eliminating underperformance that has hindered stigmatized groups from realizing their intellectual potential, is stereotype threat (Steele, 1997). Stereotype threat occurs when targets of stereotypes alleging their inferiority in a relevant domain are reminded of the possibility of confirming these stereotypes resulting in performance deficits, such as women in math (Inzlicht & Ben-Zeev, 2000; Shih, Pittinsky, & Ambady, 1999; Spencer, Steele, & Quinn, 1999), European Americans in sports (Stone, Lynch, Sjomeling, & Darley, 1999), and African Americans, Latinos, and students of low socioeconomic status in a number of intellectual domains (e.g., Aronson & Salinas, 1997; Croizet & Claire, 1998; Steele & Aronson, 1995).
The research on threatening intellectual environments parallels earlier work on stereotype threat in documenting under-performance of the stereotyped. But how do threatening environments and stereotype threat work? Despite the enthusiasm that this line of research has generated, little is known about the specific mechanisms that underlie stereotype threat. Potential mediators that have been explored thus far, such as low self-efficacy, evaluation apprehension, and attentional distraction, have not been shown reliably to mediate the relationship between stereotype threat and performance (Spencer et al., 1999).
We contend that a factor that has not been examined sufficiently thus far but that has the potential to be an important mediator is physiological arousal (e.g., Steele & Aronson, 1995). According to the Yerkes-Dodson (1908) theory of physiological arousal, performance is optimal at intermediate levels of arousal and decreases when arousal is either low or high, resulting in an inverted-U shaped function. We propose that stereotype threat may interfere with performance by leading to arousal that exceeds an optimal level.
In one empirical investigation of the role of arousal in stereotype threat, Spencer et al. (1999) showed that although self-reported generalized arousal was related to both performance deficits and threat, it did not significantly reduce the direct relationship between threat and test performance. Nonetheless, these results do not rule out arousal as a mediator of stereotype threat. As Spencer et al. note, the numbers of participants in their study may have been too small to perform an adequate mediational analysis. In addition, because the measures of arousal were explicit, participants’ ratings of arousal may have been affected by demand characteristics.
In a more recent study, Osborne (2001) found correlational evidence consistent with the idea that race differences in general achievement and sex differences in math achievement may be at least partially mediated by arousal. Unfortunately, however, Osborne noted that this correlational study failed to determine whether stereotype threat actually existed during test administration, which implies that phenomena other than stereotype threat may have been responsible for the arousal effect. Thus, although suggestive, these results cannot confirm that arousal mediates threat.
The present research, then, was aimed at testing the arousal hypothesis of the stereotype threat phenomenon. Specifically, we tested the idea that threatening intellectual environments lead to increases in arousal or generalized anxiety, which in turn cause impairments in intellectual performance. If threatening environments lead to increased arousal, then (a) eliminating arousal should attenuate threat-induced performance deficits, and (b) individuals not self-invested in the threatened domain may actually benefit from threat. In both Studies 1 and 2 we examined the math test performance of female students placed in threatening intellectual environments. In study 1 we employed the misattribution paradigm to examine whether the effects of threat could be mitigated when highly able females were given a chance to attribute their arousal to a benign external source. In study 2 we sought to determine whether low math-identified females would actually experience surfeits in performance as a result of being aroused by a threatening intellectual environment.
Experiment 1
To avoid the pitfalls of
self-report, correlational
analysis, and some of the ecological validity challenges of directly
measuring arousal, we used a misattribution paradigm. This paradigm has been
used successfully in many studies over the years to assess the role of arousal
as a mediating variable without relying either on self-report or invasive
procedures (e.g., Olson, 1988; Savitsky, Medvec, Charlton, & Gilovich, 1998;
Zanna, & Cooper, 1974, 1976). The logic of using this paradigm in the
present research is as follows: If
arousal mediates threatening intellectual environments, then the deleterious
effects of threatening environments should be attenuated if threatened individuals
are led to attribute their arousal to a benign external source having nothing
to do with the relevant stereotype.
That is, to the extent that the effects of stereotype threat are
attenuated by the availability of this (erroneous) attribution for their
arousal, it would implicate the role of arousal in the stereotype threat
process. This is tested in Experiment 1.
Method
Participants
Research
on stereotype threat suggests that performance deficits will emerge primarily
among individuals who are invested in and identified with the threatened domain
(Aronson, Lustina, Good, Keough, Steele, & Brown, 1998; Steele, 1997). To select female students highly
identified with math, we had 128 Brown University undergraduates complete the
Mathematics Identification Questionnaire (MIQ; Brown, 2000) and indicate their
math SAT scores. From these, 37 students were selected on the basis of having
scored above the theoretical midpoint of the MIQ (M = 6.03, SD
=.87) and scored 600 or above on the math portion of the SAT (M =
696.76, SD = 56.82). These
37 highly identified female undergraduates participated in exchange for credit
toward a course requirement or for payment of $7.
Procedure
Two males and two females served as experimenters, so that half of the participants were greeted by a male experimenter, whereas the other half were greeted by a female experimenter. Instructions for the experiment were recorded on an audio tape and played for the participants once they were all seated. To minimize any expectancy effects, experimenters were only present in the lab when absolutely necessary (e.g., to pass out the math test).
Stereotype threat manipulation. Inzlicht and Ben-Zeev (2000)
showed that when placed in groups of three, female participants who were very
good math who
were highly identified with math performed more poorly with two
male partners than with two female partners. We used a similar manipulation of stereotype threat in the
present study. Female participants were randomly assigned to either same-sex (three
female students) or minority (one female student with two male confederates)
conditions. All participants
were informed that their performance on the test would be reported orally to
the other group members.
Misattribution manipulation. As part of a general cover story about examining performance on standardized tests, participants were told that one of the factors being studied was the effects of subliminal noise on test performance. All participants were seated in front of a large machine that was introduced as the subliminal noise generator. In order to illustrate how the audio generator worked, participants were exposed to a series of audible tones that increased in frequency, culminating in a subliminal tone of 20,000 Hz—a “silent tone” well beyond the range of human hearing. Participants were then informed that they would be exposed to this tone for the duration of the math test. Participants in the control group were told that the subliminal noise would have no discernible physical effects on them. In contrast, those in the misattribution condition were told that the noise was associated with a number of side effects, and that previous participants had noted an increase in arousal, nervousness, and heart rate. They were told not to be alarmed if they felt these side effects and were assured that the any such side effects would be temporary. Next, all participants were given the math test.
Math test. Participants had 20 minutes to complete the math test. The math test consisted of 20 difficult) multiple-choice items culled from the GRE test guide (Educational Testing Service, 1994). Each math item was, on average, answered correctly by only one third of test takers in past samples.
Manipulation
check. At the conclusion of
the test all participants completed a brief questionnaire. In addition to
filler items was a set of four items concerning how anxious or nervous the
subliminal noise made the participants feel. Using a 7-point Likert scale
ranging from 1 (strongly
disagree) to 7 (strongly agree),
participants responded to the following items: “I feel that the
subliminal noise made me feel nervous,” “The side-effects of the
subliminal noise did not make me feel jumpy,” “I was thinking about
the possible side-effects of the subliminal noise,” and “The
subliminal noise did not affect me in any way at all.” The four
individual measures of misattribution were combined into one measure (a = .70).
Results and Discussion
Manipulation Check
Participants’ responses on the manipulation check were subjected to a 2 X 2 ANCOVA, using self-reported SAT scores as the covariate, which revealed a main effect for the misattribution manipulation, F (1, 31) = 5.81, p < .0251. As predicted, participants in the misattribution condition were more likely to attribute nervousness and arousal to the subliminal noise (M = 2.59, SD = 1.11) than were the participants in the control group (M = 2.01, SD = 1.05).
Test Performance
If a threatening environment increases arousal that
hurts problem solving, then giving women the chance to misattribute their
arousal to a benign external source, should spare performance deficits that
would otherwise be triggered by stereotype threat. Specifically, women should
have performed worse on the math test if they were in the minority than if they
were in the presence of other female participants in the control but not in the
misattribution condition. As
can be seen in Figure 1, the results supported this prediction. The 2 X 2
ANCOVA on number of math problems answered correctly, with SAT scores as the
covariate, revealed a significant interaction between the manipulations of
sex-composition and misattribution, F (1, 31) = 6.06, p < .02.
Simple-effects analyses revealed that among women not given the misattribution
information, the typical stereotype threat effects emerged: Females in the
minority group did significantly worse (M = 6.34, SD = 2.43) than
did females in the same-sex group (M = 8.82, SD = 2.40), F
(1, 31) = 5.44, p < .03.
This constitutes a large effect size, Cohen’s d =
1.03. In contrast, there were no
significant differences between minority participants (M = 7.97, SD
= 2.33) and same-sex participants (M = 6.61, SD = 2.43) in the
misattribution condition, F (1, 31) = 1.44, ns. No other effects approached
significance.
----------------------------------------------------------
Insert Figure 1 here
----------------------------------------------------------
Arousal may have an important role as a mediator of stereotype threat. Women’s performance deficits in the presence of men were attenuated when women were given an opportunity to misattribute their arousal to an external source. When arousal was “removed,” by misattribution to a benign source, women were less likely to experience impairments in performance. Furthermore, far from exhibiting performance deficits, women who were in the minority actually showed a trend for performance surfeits in the misattribution condition. That is, they tended to do better on the math test than did minority females not given the opportunity to misattribute their arousal. Although this difference was not significant, it does suggest a trend that future research may revisit. By decreasing the heightened arousal experienced by women in the minority condition, the misattribution may have reduced their arousal closer to optimal levels (cf. Yerkes & Dodson, 1908). Women in the same-sex conditions should not have been as aroused, so it is conceivable that any reduction in arousal caused by misattribution may have dropped their arousal to sub-optimal levels, causing a slight decline
In Experiment 2, we examine arousal’s association with threatening environments further by examining whether individuals low in math identification can actually benefit from being placed in threatening intellectual environments.
Experiment 2
Stereotype threat predicts that performance deficits will only emerge in those individuals who are invested in and identified with the threatened domain (Steele, 1997). Stereotypes that allege low math ability can only pose a threat to individuals that actually care about having that ability or care about the social consequences of being seen as lacking in that ability (Aronson, Lustina, Good, Keough, Steele, & Brown, 1999); individuals who don’t identify with a domain would hardly be affected by group stereotypes that threaten it.
If arousal mediates stereotype threat, then, we would expect domain-invested females—those who approach a task with an already heightened level of arousal because of their high identification—to be debilitated by the extra-arousing properties of threat. Recall, performance is optimal at intermediate levels of arousal (Yerkes & Dodson, 1908). Thus, the extra pressure of stereotype threat may push the arousal of high identifiers beyond the optimal intermediate level, resulting in problem-solving deficits. In contrast, low-identified females—those who do not care as much about doing well in a domain that is not central to their self-concept—may initially approach a math task with lower level arousal. Under conditions of threat, however, arousal may rise to an intermediate level, resulting in optimal performance—they may actually experience benefits in performance (Aronson et al, 1999). In other words, if arousal mediates threat, we would expect low math identifying females to show a reversed pattern of problem-solving deficits from females who are high identifiers. This is tested in Experiment 2.
Method
Participants
Participants
were female undergraduate students at Williams College who were enrolled in an
introductory psychology class. At
the beginning of the semester, 161 potential participants filled out the
MIQ. Responses to the 13 items
were averaged into an index score.
Participant scores on this measure ranged from 2.00 to 8.46 (M =
5.95, SD = 1.14). Dividing
these students into three groups and taking the top and bottom third of
students on this measure produced a “high identified” group (n
= 31, M = 7.41, SD = 0.43) and a “low identified” (n
= 28, M = 4.60, SD = 0.46).
These 59 high and low identified students participated in exchange for
credit towards a course requirement or a payment of $6.00.
Procedure
As
in Experiment 1, female participants were randomly assigned to either the
same-sex or minority conditions.
The experiment began once all three participants arrived. Although their scores were not recorded
for this study, real male participants, as opposed to confederates, were
employed in the minority condition.
The female experimenter presented the same cover
story to all participants, who were told that the goal of the study was to
develop a new educational training program to improve performance on
standardized tests. Participants
were also informed that after the test they would orally compare their
performances with the other participants.
After allowing participants to listen and read the instructions, the
experimenter ushered them into individual cubicles and gave them twenty minutes
to complete the math test. After
the test, participants were asked to complete a demographic sheet that asked
them to report their mathematical SAT scores (SAT-M).
As in experiment 1, the experimenter was only
present in the room when absolutely necessary so as to minimize any effects
that may be due to their presence.
Two females served as experimenters.
Results and Discussion
We expected to find a cross-over interaction between
sex-composition and math identification, such that minority females would do
worse than same-sex females in the high identification group, but better than
same-sex participants in the low identification group. As can be seen in Figure
2, results support our hypothesis—low identified females actually
performed better on the math test when in the minority environment. The 2 X 2 ANCOVA on number correct,
using participants’ self-reported math SAT scores as the covariate,
revealed a significant interaction between sex-composition and math
identification, F (1, 54) = 7.44, p < .01. No other effects were significant or
close to being significant.
----------------------------------------------------------
Insert Figure 2 here
----------------------------------------------------------
Tests of simple effects demonstrate that high
math-identification participants got marginally fewer items correct when they
were in the minority condition (M = 9.71, SD = 3.07) than when
they were in the same-sex condition (M = 11.58, SD = 3.26), F
(1,54) = 2.82, p < .10.
Although this effect is only marginally significant, it still
constitutes a medium effect size, Cohen’s d = .59, suggesting that
with just a little more power the effect would have reached traditional levels
of significance. Low
math-identification participants, in contrast, actually showed the opposite
pattern of results: minority participants got significantly more items correct
(M = 9.77, SD = 3.37) than same-sex participants (M =
7.27, SD = 3.06), F (1,54) = 4.29, p < .05. This constitutes a medium effect size,
Cohen’s d = .78.
This experiment offers support for a number of
ideas. First, it provides
additional support for environmental effects on performance. When high-identified female
participants took a math test in an environment where males outnumbered them
they demonstrated the resultant problem-solving deficits.
Second, these results offer additional support to
the importance of domain identification in the experience of stereotype
threat. Specifically, these
results agree with Aronson et al.’s findings that participants suffered
math deficits under conditions of stereotype threat, but only if the
participants were classified as high in math identification (experiment 2;
Aronson et. al., 1999). These
findings corroborate Steele’s (1997) assertion that stereotype threat has
its greatest effect among the brightest, most confident, and most highly domain
identified stereotyped group members.
Third, these results support the prediction of the
Yerkes-Dodson theory of arousal and performance, which claims that performance
is best not at high or low, but at intermediate levels of arousal (Yerkes &
Dodson, 1908). It is thus possible
that stereotype threat represents a moderator of arousal. Research does indicate, after all, that
stereotype threat, and presumably threatening intellectual environments,
essentially motivate participants to disconfirm ability denigrating stereotypes
(Steele & Aronson, 1995).
Thus, it is possible that low identified individuals underperformed in
the same-sex group because arousal was too low for efficiency. In the minority condition, however,
when stereotype threat increased disconfirmatory motivation, arousal and thus
performance were at peak levels.
Conversely, when an increment of drive was added by looking at high
identifiers, total arousal in the minority condition may have become excessive
leading to performance deficits, whereas it may have regressed to optimal
levels in the same-sex condition leading to peak performance.
General Discussion
The findings support the prediction that cues in the environment, such as gender-composition, give rise to heightened arousal, which in turn, mediate stereotype threat. In experiment 1, performance deficits associated with stereotype threat were attenuated when females were given an opportunity to misattribute their arousal to an external source. When female participants in our study were outnumbered by males, they suffered impairments in performance. When they could misattribute their arousal to the subliminal noise, however, they exhibited no performance deficits. In Experiment 2, we found a cross-over interaction between identification with mathematics and threat. When female participants identified with the threatened math domain, they experienced performance decrements in the minority environment. When they were not identified with math, in contrast, they actually experienced problem-solving surfeits in the same minority environment. Taken together, both of these experiments suggest that arousal may have an important role as a mediator of threatening intellectual environments.
More generally, the results of the
present study converge with other findings, which imply that stereotype threat
has arousal properties. For
instance, recent research examining the role of stereotype threat in the high
incidence of high blood pressure among African Americans indicates that threatening intellectual environments
may be stressful (Blascovich, Spencer,
Quinn, & Steele, 2001). Black participants who experienced threat had blood
pressure that rose faster and remained higher than non-threatened Black
participants or White participants in any condition. Interestingly, although
blood pressure was affected by the threat manipulation, actual test scores were
not: Black participants scored as well as White participants in both threat and
non-threat conditions. On the face of it, it appears that the threat
manipulation was not successful; however, an alternative explanation that is in
line with the present results is possible. Perhaps the manipulation was indeed
successful, but participants failed to be impaired by it because they
attributed their arousal to being hooked up to the electrocardiograph.
Similarly,
Stone et al. (1999) found that White participants performed worse on an
athletic task when the stereotype about Black athletic superiority was made
salient. This effect, however, was attenuated once participants were asked to
monitor the effect of the lab space on their performance—giving them an
opportunity to misattribute their arousal. Although not central to either line
of research, these two sets of findings suggest that leading participants to
misattribute their arousal—by focusing on an electrocardiograph or the
effects of a lab space—may have reduced the negative effects of
stereotype threat on threatened individuals’ performance. This pattern of
results adds support to the idea that arousal may play a mediating role in
stereotype threat.
However, a number of important questions remain to be explored. First, through what process does arousal impair performance? One possibility, as suggested by Easterbrook’s (1959) theory of cue utilization, is that arousal can reduce the range of cues used in a task, which can ultimately result in attentional focus that is so narrow that task-relevant cues begin to be ignored. A challenge of ongoing research will be to empirically test this possibility. Second, is arousal actually necessary to produce threat? The current study demonstrates that arousal is associated with environmental performance deficits, but does not demonstrate that arousal must occur in order to produce impairments in performance (see Cooper, Zanna, & Taves, 1978, for a similar argument). It is possible that something associated with arousal—for example extra-task worry—and not arousal per se that leads to performance deficits. Another question concerns what factors other than arousal may mediate stereotype threat. Steele and Aronson (1995) have discussed the possible roles of distraction and lowered expectations, whereas Wheeler and his colleagues (Wheeler, Jarvis, & Petty, 2001; Wheeler & Petty, in press) suggest that automatic cognitive processes, such as ideomotor processes (cf. Bargh, Chen, & Burrows, 1996) may also play a role.
Arousal may be one of several mediators of stereotype threat. We turn to discus some other possible candidates next.
Other Possible
Mediators of Stereotype Threat
Other than arousal, there may be other mediators involved in producing stereotype threat. Steele and Aronson (1995) have discussed the possible implication of distraction and expectation. Depending on the situation and on the group, these mechanisms may be involved simultaneously or in alternation. Distraction theories claim that underperformance reflects some interference with the performer’s attentional processes (see Baumeister & Showers, 1986, for a review). Optimal performance requires attending to the task at hand and excluding other factors. Stereotype-induced underperformance, then, may reflect the distracting effects of shifting one’s attention to task-irrelevant cues, thereby ignoring important features of the task. People experiencing the burden of stereotype threat may become deluged with task-irrelevant worries (e.g. worries over the consequence of failing, concerns about how others will view their performance) and as a result cannot engage in the normal cognitive functions that allow them to achieve the most favorable performance (Wine, 1971). Threatened persons may also experience interfering self-consciousness. That is, they may focus their attention on the process of performance, which may paradoxically inhibit or interfere with the automatic and overlearned responses constituting that very performance (Baumeister, 1984). One possibility is that arousal activates one or more of these distracters thus leading to impairments in performance.
Expectation theories claims that underperformance reflects low performance expectations, which can moderate effort. Stangor, Carr and Kiang (1998), for example, suggest that the activation of negative stereotypes can lower task-performance expectations, which are strong enough to undermine initial task confidence. Lower expectations can undermine motivation and effort (Pyszczynski & Greenberg, 1983), which can in turn result in lower performance on all three indexes of achievement. Hence, negative stereotypes may lead to lower performance because they can undermine confidence, which, in turn, can lead to a withdrawal of effort.
In sum, stereotype threat may interfere with performance in a variety of ways: by causing an arousal that lowers performance, by diverting attention onto task-irrelevant worries, or by fostering lower performance expectations. Arousal appears to be involved in mediation, but how it is involved—be it through its association distracters or expectations—is still unknown.
In one ambitious study, for example, Spencer et al. (study 3; 1999) examined the contribution of all three of these mechanisms. As in previous studies, highly identified men and women were instructed to take a math test that was either described as showing gender differences in the past (threat condition) or not (no threat condition). Once all the participants read the test instructions, they filled out a questionnaire assessing generalized anxiety (e.g. “I am jittery”), distracting fears related to being evaluated—evaluation apprehension (e.g. “people will think I have less ability if I do not do well on this test”), and feelings of confidence and certainty—self-efficacy (e.g. “I am concerned about whether I have enough mathematical ability to do well on the test”). After filling out this questionnaire, all participants took the difficult math test.
Results replicated the basic stereotype threat effect: Women attained lower adjusted scores than men in the threat condition, but performed equally to men in the no threat condition. Again, when stereotypes were relevant to performance, women underperformed. Mediation of this effect was tested through a series of mediational analyses for each of the potential mechanisms (Baron & Kenny, 1986). The results, however, were unable to demonstrate that any of the hypothesized mechanisms strongly mediated stereotype threat. First, self-efficacy or task confidence was both unrelated to the stereotype threat manipulation and to test performance. Thus, confidence did not appear to cause stereotype threat. Similarly, evaluation apprehension—a potentially distracting worry—did not appear to mediate the relationship between stereotype threat and performance; although it was related to test performance, it was not related to the stereotype threat manipulation, nor did it reduce the direct relationship between the manipulation and performance. Finally, although generalized arousal was related to both performance and the threat manipulation, it did not significantly reduce the direct relationship between the threat manipulation and test performance. There was not strong evidence, then, that arousal mediated threat. Nonetheless, these results did not rule out arousal as mediator either (Spencer et al., 1999).
Aronson and his colleagues (1999) also investigated the cause of threat, but were unable to reach any definite conclusions. As in previous studies, they found that highly identified White students who were faced with the Asian math superiority stereotype attained fewer number correct than White students who were not made aware of the stereotype. The stereotyped participants also wondered more often what the experimenter would think of them in the threat condition than in the control condition. That is, threatened participants experienced evaluative worries that may have distracted their performance. However, when this evaluation apprehension was partialed out, the relationship between threat and performance remained unchanged. Therefore, worries about being negatively evaluated did not appear to cause threat-induced performance deficits.
Finally, Brown and Steele (1999, as reported in Marx, Brown, & Steele, 1999) directly explored the role of performance expectations. As in previous experiments, White and Black undergraduates took a difficult GRE test of verbal ability. Half of the participants were informed that the test was diagnostic of ability (threat condition), whereas the other half were told that it was diagnostic, but racially fair (no threat condition). Prior to taking the main GRE test, participants took either an easy or difficult verbal task that resulted in complete success or clear failure. The idea was that performance on this task would set either high or low performance expectations for the second task. Results demonstrated that Black participants performed worse than their White peers, even when they presumably had high performance expectations. Because high performance expectations could not overcome the hindering qualities of racial stereotypes, the authors concluded that stereotype threat was not mediated by expectations.
It appears, therefore that much work still needs to be done on the mechanisms underlying stereotype threat. The current study points to the mediational role of arousal. This mechanism, however, may work in interaction with other causal factors, such as, performance expectations and the associated withdrawal of effort, to create threatening intellectual environments.
Regardless of which mechanisms mediate stereotype threat, one may challenge performance decrements as stemming from stereotype threat itself, by positing other theoretical phenomena, such as the dumbing-down phenomenon.
Stereotype Threat vs. Dumbing
Down
Although stereotype threat is a relatively new hypothesis (Steele & Aronson, 1995), Irwin Katz and his colleagues had discussed a situational interpretation of intellectual performance deficits in stereotyped groups as early as 1963 (Katz & Greenbaum, 1963; Katz, Roberts, & Robinson, 1965). In a landmark study, Katz, Roberts, and Robinson (1965) found that manipulating the characterization of a test led to differential performances for Black participants, thus foreshadowing the stereotype threat phenomenon. Their interpretation of this effect, however, diverged from that of stereotype threat researchers. Katz et al. proposed that it reflected a dumbing-down process, which claims that targeted individuals consciously or unconsciously suppressed their own correct responses out of fear that they would be castigated for exceeding low performance levels prescribed by the ability-infringing stereotype (c.f. Komarovsky, 1946). It is thought that all members of stereotyped groups—high and low domain identified individuals—would be equally at risk of ability suppression because they all face the same social censure for violating stereotypes. Results from Experiment 2, however, indicate that minority-deficits are only produced in highly math-identified participants. Thus, although there is evidence suggesting that females may be socially sanctioned for excelling in “male” domains (Pfost & Fiore, 1990), and that they may voluntarily downplay their abilities in order to be admired by potential romantic partners (Jellison, Jackson-White, Bruder, & Martyna, 1975; Zanna & Pack, 1975), there is little evidence suggesting that minority-deficits are a product of ability suppression.
On the contrary, results from our current study indicate that performance deficits may be due to simultaneously valuing a domain and being threatened by stereotypes devaluing that domain. Although high identifying participants showed deficits relative to their same-sex peers, the reverse pattern occurred for the low identifiers—higher performance in the minority than in the same-sex conditions.
In sum, the current results suggest that arousal can play an important mediational role in producing the performance impairments witnessed in threatening intellectual environments. An implication of these findings is that interventions that can reduce or remove stereotype-linked arousal may help lift the burden of stereotype threat. “Wise schooling,” which makes threatening stereotypes less applicable to the behavior of stereotyped individuals (Steele, 1997), may operate in part by just such a removal of stereotype-linked arousal. Teaching students who are targets of stereotype threat to relax in the face of threat may also be a promising future direction. Some recent work by Fein (e.g., Fein, 1999; Fein & Maron, 1998) suggests that self-affirmation can buffer the self against the threatening implications of failure in a high-pressure situation, and so it too may moderate the pernicious link between threatening environments and performance deficits. In any case, as the evidence builds in support of the power and pervasiveness of stereotype threat, it becomes all the more important to identify the critical mechanisms that mediate its effects. The present findings point to arousal as one such mechanism.
References
Aronson, J., Lustina, M. J., Good, C., Keough, K., Steele, C. M., & Brown, J. When white men can’t do math: Necessary
and sufficient factors in stereotype threat. Journal of Experimental Social Psychology, 35, 29-46.
Aronson, J., & Salinas, M. F. (1997). Stereotype
Threat, Attributional Ambiguity, and Latino Underperformance. Unpublished manuscript, University of
Texas.
Bandura, A.
(1977). Self-efficacy:
Toward a unifying theory of behavioral change. Psychological Review, 84, 191-215.
Bargh, J. A., Chen, M., & Burrows, L. (1996).
Automaticity of social behavior: Direct effects of trait construct and
stereotype activation on action. Journal of Personality and Social
Psychology, 71Journal of Personality and Social Psychology, 71, 230-244.
Baumeister, R. F. (1984). Choking under pressure:
Self-consciousness and paradoxical effects of incentives on skillful
performance. Journal of Personality and Social Psychology, 46, 610-620.
Baumeister, R. F., & Showers, C. J. (1986). A
review of paradoxical performance effects: choking under pressure in sports and
mental tests. European Journal of Social Psychology, 16, 361-383.
Blascovich, J., Spencer, S. J., Quinn, D. M., &
Steele, C. M. (2001). African
Americans and high blood pressure: The role of stereotype threat. Psychological
Science, 12, 225-229
Brown,
R. P. (2000). The importance of
importance: Factor-analytic conclusions about domain identification.
Manuscript submitted for publication.
Cooper, J., Zanna, M. P., & Taves, P. A. (1978).
Arousal as a necessary condition for attitude change following induced
compliance. . Journal of
Personality and Social Psychology, 36, 1101-1106.
Croizet, J.
C. & Claire, T. (1998). Extending the concept of stereotype threat to social class:
the intellectual underperformance of students from low socioeconomic
backgrounds. Personality and
Social Psychology Bulletin, 24, 588-594.
Deaux, K., Major, B. (1987). Putting
gender into context: an interactive model of gender-related behavior. Psychological review, 94,
369-389.
Easterbrook, J. A. (1959). The effect of emotion on
cue utilization and the organization of behavior. Psychological Review, 66,
183-201.
Educational Testing Service. (1994). GRE:
Practicing to take the general test (9th Ed.). Princeton, NJ: Educational Testing
Service.
Fein, S. (1999, October). Prejudice
and performance: Dealing with self-image threats. Paper presented at the
Society of Experimental Social Psychology Preconference on the Self, St. Louis,
MO.
Fein, S., & Maron, B. A. (1998, August). Self-affirmation
and athletic and academic performance under pressure. Paper presented at
the 1998 Conference of the American Psychological Association, San Francisco,
CA.
Jellison, J. M., Jackson-White, R., Bruder, R. A. & Martina, W. (1975). Achievement behavior: A situational interpretation. Sex Roles, 4, 369-384.
Katz, I. & Greenbaum, C. (1963). Effects of anxiety, threat, and racial environment on task performance of Negro college students. Journal of Abnormal and Social Psychology, 66, 562-567.
Katz, I., Roberts, S. O., & Robinson, J. M. (1965). Effects of task difficulty, race of administrator, and instructions on digit-symbol performance of Negroes. Journal of Personality and Social Psychology, 2, 53-59.
Olson, J. M. (1988). Misattribution, preparatory
information, and speech anxiety. Journal of Personality and Social
Psychology, 54, 758-767.
Osborne, J. W. (2001). Testing stereotype threat: Does
anxiety explain race and sex differences in achievement? Contemporary
Educational Psychology, 26, 291-310.
Pfost, K. S., & Fiore, M. (1990). Pursuit of
nontraditional occupations: fear of success or fear of not being chosen? Sex
Roles, 23, 15-24.
Rosner, B. (1995). Fundamentals of Biostatistics
(4th edition). Belmont,
CA: Duxbury Press.
Savitsky, K., Medvec, V. H., Charlton, A. E., &
Gilovich, T. (1998). What, me worry? Arousal, misattribution and the effect of
temporal distance on confidence. Personality and Social Psychology Bulletin,
24, 529-536.
Shih,
M., Pittinsky, T. L., & Ambady, N. (1999). Stereotype susceptibility:
identity salience and shifts in quantitative performance. Psychological
Science, 10, 80-83.
Spencer,
S. J., Steele, C. M. & Quinn, D.
(1999). Stereotype threat
and women's math performance. Journal
of Experimental Social Psychology, 35, 4-28.
Stangor,
C., Carr, C., & Kiang, L.
(1998). Activating
stereotypes undermines task performance expectations. Journal of Personality and Social Psychology, 75,
1191-1197.
Steele,
C. M. (1997). A
threat in the air. How stereotypes
shape intellectual identity and performance. American Psychologist, 52, 613-629.
Steele,
C. M. & Aronson, J.
(1995). Stereotype threat
and the intellectual test performance of African Americans. Journal of Personality and Social
Psychology, 69, 797-811.
Stone,
J., Lynch, C. I., Sjomeling, M., & Darley, J. M. (1999). Stereotype threat
effects on Black and White athletic performance. Journal of Personality and
Social Psychology, 77, 1213-1227.
Wheeler,
S. C., Jarvis, W. B., & Petty, R. E. (2001). Think unto others: The
self-destructive impact of negative racial stereotypes. Journal of
Experimental Social Psychology, 37, 173-180.
Wheeler, S. C., & Petty, R. E. (in press). The
effects of stereotype activation on behavior: A review of possible mechanisms. Psychological
Bulletin.
Yerkes,
R. M., Dodson, J. D. (1908). The relationship of strength of stimulus to
rapidity of habit-formation. Journal of Comparative Neurology of Psychology,
18, 459-482.
Zanna, M. P. & Cooper, J. (1974). Dissonance and
the pill: An attribution approach to studying the arousal properties of
dissonance. Journal of
Personality and Social Psychology, 29, 703-709.
Zanna, M. P., & Cooper, J. (1976). Dissonance and
the attribution process. In J. H. Harvey, W. J. Ickes, & R. F. Kidd (Eds.),
New directions in attribution research (Vol. 1, pp. 199-217). Hillsdale,
NJ: Lawrence Erlbaum.
Zanna, M. P., & Pack, S. J. (1975). On the
self-fulfilling nature of apparent sex differences in behavior. Journal of
Experimental Social Psychology, 11, 583-591.
Authors Note
We
thank the following individuals for their insights and suggestions: Joshua
Aronson, Rachel Herz, Joachim Krueger, Daniel Newhall, Ken Ojugwo, Steve
Spencer, and Jack Wright.
Correspondence concerning this article should be addressed to Talia Ben-Zeev, Department of Psychology, 1600 Holloway Ave, San Francisco, CA 94132, email: tbenzeev@sfsu.edu, or to Michael Inzlicht at Mickeyi@lycos.com.
Footnote
1
As suggested by Rosner (1995), one case was excluded from the
analysis—a participant in the same-sex control group—because it was
identified as an outlier by the Extreme Studentized Deviate Statistic, ESD
(19) = 2.69, p < .05.
This statistic indicates that the particular case has an undue influence
and causes major changes to the distribution when it is included—the F
statistic for the main effect of the misattribution manipulation drops to .14 (p
<.72) from 5.81 (p < .025).
The excluded participant scored 5.75 on the manipulation check when the
mean and standard deviation are 2.01 and 1.05, respectively.
Figure Captions
Figure
1. Number of items correct,
corrected for SAT, as a function of sex-composition of group and
misattribution. Error bars
represent standard error.
Figure
2. Math performance, corrected
for SAT, as a function of sex-composition of group and math
identification. Error bars
represent standard error.
