Vous êtes sur la page 1sur 9

Journal of

Experimental
Social Psychology
Journal of Experimental Social Psychology 40 (2004) 247–255
www.elsevier.com/locate/jesp

When saying and doing diverge: The effects of stereotype threat


on self-reported versus non-verbal anxietyq
Jennifer K. Bosson,a,* Ethan L. Haymovitz,b and Elizabeth C. Pinelc
a
Department of Psychology, 455 W. Lindsey, DAHT #705, The University of Oklahoma, Norman, OK 73019, USA
b
Department of Psychology, Vassar College, USA
c
Department of Psychology, The Pennsylvania State University, USA

Received 31 July 2002; revised 20 May 2003

Abstract

Although research has established that stigmatized individuals suffer impaired performance under stereotype threat conditions,
the anxiety presumed to mediate this effect has proven difficult to establish. In the current investigation, we explored whether non-
verbal measures would fare better than self-reports in capturing stereotype threat anxiety. Gay and heterosexual men interacted with
preschool children under stereotype threat or control conditions. As predicted, stereotype-threatened gay men demonstrated more
non-verbal anxiety, but not more self-reported anxiety, than non-threatened gays during these interactions. Furthermore, non-
verbal anxiety appeared to mediate the effects of stereotype threat on the quality of participantsÕ childcare skills. We discuss how
these findings advance stereotype threat research, and highlight their potential implications for gay childcare workers.
Ó 2003 Elsevier Inc. All rights reserved.

Keywords: Stereotype threat; Non-verbal anxiety; Gay males

Introduction effects—anxiety—has proven difficult to establish em-


pirically. We propose that researchers who failed to
Stereotype threat is the realization that oneÕs perfor- obtain evidence of the anxiety triggered by stereotype
mance on a particular task might confirm a negative threat did so because they searched for this evidence in
stereotype about oneÕs group (Steele & Aronson, 1995). peopleÕs self-reports, rather than in other, less control-
Once evoked, this concern may hinder the performance lable types of responses. Thus, the primary purpose of
of stereotyped individuals in various evaluative contexts. this study is to explore the effectiveness of self-reports
Although the outcomes of stereotype threat manipula- versus unobtrusive measures in capturing stereotype
tions are clear, the proposed mechanism driving these threat anxiety. Another purpose is to expand the extant
literature by testing stereotype threat effects in an as-yet-
unexplored group: gay men.
In their efforts to uncover stereotype threat anxiety,
q
We thank Donna Albert, Ryan Camire, Jenny Elek, Lesley most researchers utilize self-report scales, with mixed
Grapka, Jonathan Lytle, Agatha Maciejewski, Jocelyn Marlatt, results. Whereas findings from some studies suggest that
Danyale McCurdy, Lisa Paul, Courtney Pearson, Julia Rubin-Smith, self-reported anxiety is unrelated to performance among
Adam Scott, and Diya Uberoi for their assistance with data collection,
data entry, and videotape coding. We are especially grateful to Julie
stigmatized individuals (e.g., Aronson et al., 1999;
Reiss and the staff of the Wimpfheimer Nursery School for accom- Brown, Charnsangavej, Keough, Newman, & Rentfrow,
modating us, Tiffany Lightbourn for supervising various aspects of the 2000; Oswald & Harvey, 2001; Schmader, 2002; Steele &
project, and Bridget Murphy-Kelsey for sharing her knowledge of Aronson, 1995), other findings show that self-reported
behavioral coding. For their helpful comments on an earlier draft of anxiety accounts for a small but significant portion of
this paper, we thank Ryan Brown, April Phillips, Jennifer Prewitt,
Carolin Showers, and Virgil Zeigler-Hill.
the variance in the performance of stereotype-threatened
*
Corresponding author. Fax: +405-325-4737. individuals. For example, reports of anxiety made by
E-mail address: jbosson@ou.edu (J.K. Bosson). women (Osborne, 2001; Spencer, Steele, & Quinn, 1999),

0022-1031/$ - see front matter Ó 2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0022-1031(03)00099-4
248 J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255

Blacks, and Latinos (Osborne, 2001) partially mediated validity of this phenomenon, but also because it moves
the link between their stereotype-threatened status and stereotype threat research into the arena of concealable
their intellectual test performance. Still other findings stigmas (see also Croizet & Claire, 1998). There is reason
show that although stereotype threat increases partici- to expect that people with concealable versus visible
pantsÕ self-reported anxiety, this anxiety does not me- stigmas will experience and react to stereotype threat
diate the association between the threat and their somewhat differently. Because their stigma is not nec-
performance (Stone, Lynch, Sjomeling, & Darley, 1999). essarily apparent to observers, gay men may be rela-
In sum, empirical evidence for the anxiety-mediator tively unconcerned about the possibility of confirming a
hypothesis has been inconclusive. negative stereotype in othersÕ eyes (cf. Crocker, Major,
We see at least two possible reasons for the incon- & Steele, 1998). Alternatively, if concealing oneÕs stig-
sistency of these anxiety findings. First, stigmatized in- matized status is akin to keeping a secret, then gay men
dividuals may be particularly mindful of, and concerned may experience an ‘‘obsessive preoccupation’’ with their
with, the impressions they make on others (e.g., Frable, stereotyped status that actually heightens their vulner-
Blackstone, & Scherbaum, 1990; Pinel, 1999). For some ability to stereotype threat (Smart & Wegner, 1999).
individuals, this heightened awareness and concern may To test how stereotype threat affects gay men, we
constrain their self-presentational behaviors, compelling turned to the stereotype that gay men are dangerous to
them to convey a ‘‘cool, calm, and collected’’ image on young children. This belief appears to derive from two
self-report anxiety scales in an effort to appear invul- assumptions. First, many people incorrectly link ho-
nerable to stereotypes. Second, it is possible that ste- mosexuality with pedophilia (Herek, 2002). Second,
reotype threat processes are non-conscious (cf. Croizet many believe that gay men will corrupt children by ex-
& Claire, 1998; Oswald & Harvey, 2001), which could posing them to a ‘‘perverted’’ lifestyle, or even seduce
explain why they are difficult to tap with self-reports. them into lives of homosexuality (Herek, 1991). So
Regardless of why self-reported anxiety produces powerful is this stereotype, that the results of a 1996
mixed findings, we believe that a logical next step in the Gallup poll (cited in Neuberg, Smith, & Asher, 2000)
effort to capture stereotype threat anxiety involves the indicated large increases in AmericansÕ acceptance of
use of indirect measures. Such measures are ideal be- gays in most occupations—except as elementary school
cause respondents cannot easily control their responses teachers or clergy. Given the prevalence of the belief that
to them, and they may capture processes of which re- gays pose a threat to children, we wondered whether
spondents are unaware (Dovidio & Fazio, 1992). Sup- concern about being perceived as perverted would im-
porting this logic, Blascovich, Spencer, Quinn, and pair gay menÕs ability to interact comfortably and ef-
Steele (2001) observed increases in blood pressure fectively with children, as anecdotal and survey research
among stereotype-threatened Black men. Although not suggest that it might (Kissen, 1996; Olson, 1987).
necessarily a measure of anxiety, blood pressure and We exposed some gay men to a stereotype threat
other markers of physiological arousal may provide in- manipulation before they interacted with young children
sight into peopleÕs emotional states (e.g., Brownley, in a nursery school, and compared their self-reported
Hurwitz, & Schneiderman, 2000) while circumventing anxiety, non-verbal anxiety, and overall performance to
problems associated with self-reports. that of some non-threatened gay men. We assumed that
Another indirect measure involves observing peopleÕs stereotype-threatened gay men would exhibit more non-
non-verbal anxiety behaviors. Non-verbal behavior is verbal anxiety—but not necessarily more self-reported
thought to reflect peopleÕs spontaneous, genuine emo- anxiety—than non-threatened gays. Moreover, we ex-
tions (DePaulo, 1992), and should thus be somewhat pected non-verbal anxiety to mediate the relation be-
immune to self-presentation attempts. Observer-ratings tween stereotype threat and performance during the
of anxiety based on non-verbal behavior are often cor- interaction. For our comparison group, we used heter-
related with targetsÕ self-reported anxiety (e.g., Watson osexual men; we did not expect heterosexualsÕ anxiety or
& Clark, 1991), but under certain conditions—such as performance to be affected by the stereotype threat
when targets are motivated to conceal, or are unaware manipulation. Finally, we measured additional variables
of, their anxiety—non-verbal behavior may provide (familiarity with children, gay menÕs comfort being
more reliable information than self-reports (Ekman & ‘‘out’’) that might moderate our findings.
Friesen, 1969; Waxer, 1977). In the current study we
measure anxiety by way of peopleÕs non-verbal behav-
iors, and compare the effectiveness of this strategy with Method
that of a self-report scale.
As noted, an additional goal of this study was to Participants
explore how stereotype threat affects gay men. A focus
on gay menÕs reactions to stereotype threat is important Forty heterosexual, 22 gay, and 10 bisexual male
not only because it further establishes the external students at a small liberal arts college participated in
J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255 249

exchange for course credit or $10. Because stereotypes options: Heterosexual/straight, homosexual/gay, and
concerning gay and bisexual menÕs negative influences bisexual.
on children are similar (Mohr & Rochlen, 1999), we The remaining participants received the no-prime
combined the data from bisexual and gay participants version of the survey, which did not contain the sexual
into one group.1 We discarded responses from two orientation item or any other reference to sexual orien-
heterosexual and four gay participants because they re- tation. Both surveys also contained questions about the
ported strong suspicions about the experimental proce- participantÕs age, race, and familiarity with children.
dures, and one participant because he failed to complete Specifically, we asked ‘‘To what extent are you generally
a key questionnaire.2 This left 37 heterosexual and 28 comfortable around young (aged 4–6) children?’’ and
gay men, aged 18–22 years, in the final sample. ‘‘How much experience do you have interacting with
young children?’’ These two items were rated on 9-point
Procedure scales with endpoints of not at all/extremely and none/a
lot, respectively; because responses to these items were
Students in introductory psychology courses were correlated, r ¼ :78; p < :001, we averaged them to yield
recruited via the schoolÕs web-based participant pool. an index of familiarity with children.
Because the number of gay men in the participant pool When the participant finished the demographic survey,
was small, we recruited gay participants through an the experimenter took it from him, glanced briefly at his
email list for the gay student group, ads in the college responses, and said ‘‘Okay, letÕs get started.’’ This was
newspaper, and flyers posted around campus. Postings done so that participants in the prime condition would
stated that ‘‘Males [Gay and bisexual males] at least 18 believe it possible, but not definite, that the experimenter
years of age are wanted for a study exploring issues of was aware of their sexual orientation; we assumed that
college adjustment.’’ this state of uncertainty would heighten the impact of the
When participants arrived individually at the lab, an stereotype threat manipulation for gay participants.
experimenter greeted them and explained that the ori-
ginal study was cancelled because of problems with the Interaction task
materials. She then explained that she was working with Next, the experimenter led participants to the play-
a professor on an unrelated project exploring how col- room. The interaction task took place during ‘‘free
lege students interact with young children, and she in- play’’ time, during which the children engaged in un-
vited the participant to be in her study (after reassuring structured activities in a large playroom equipped with
him that it would take no longer than the original ex- books, dress-up clothes, games, and toys. Several adult
periment and that he would receive whatever credit or teachers and student workers were always present in the
money he had been promised). We included this de- room, but they typically did not intervene with the
ception to reduce the salience of sexual orientation for childrenÕs play behavior. Approximately 20 children,
gay participants, who were originally recruited specifi- aged 4–6 years old, were present in the playroom during
cally because they were gay. All but four (5.6%) students each experimental session. At the door to the playroom,
consented to the experimenterÕs request, and were sub- the experimenter instructed the participant to familiarize
sequently led to the campus nursery school where they himself with one child, or a group of children, and try to
learned that they would complete several questionnaires involve himself in the childÕs activity for 5 min. She then
and engage in a brief, videotaped interaction with some explained that the participant would be videotaped from
children.3 behind a one-way mirror, and directed him to remain in
the area in front of the mirror. No further instructions
Stereotype threat manipulation were given, because we wanted to observe participantsÕ
Participants first completed a demographic survey behavior in a relatively unstructured, naturalistic social
that contained the stereotype threat manipulation. Fol- situation. After giving instructions, the experimenter
lowing (Steele & Aronson, 1995, Study 4), we randomly exited and waited outside the playroom for 5 min; she
assigned approximately half of participants to receive then returned and led the participant to another room to
the prime version of the survey, which required them complete some final scales.
to circle their sexual orientation from the following
Self-reported reactions
The last set of scales began with a self-report measure
1
The removal of bisexualsÕ data from analyses did not change any of reactions to the interaction task. We measured two
of our patterns. variables that have been proposed to mediate the rela-
2
Inclusion of the suspicious participantsÕ data did not change any tionship between stereotype threat and performance:
of our patterns, although it did reduce statistical significance levels.
3
We do not know the sexual orientation of the students who
anxiety and evaluation apprehension. Using scales rang-
declined to participate because this variable was assessed during the ing from 1 (not at all) to 9 (extremely), participants in-
experimental sessions. dicated the extent to which they experienced five
250 J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255

different anxiety-related emotions during the task: Each rating pertained only to the 30 s that preceded it, and
anxious, worried, calm, pleasant, and relaxed. After judges considered both the intensity and the frequency of
reverse-scoring items so that higher scores indicated anxiety behaviors when making their ratings. Thus, a
more anxiety, we averaged across these items (a ¼ :82). rating of 5 could reflect either one or two extreme (but
Next, participants used the same scales to rate the extent brief) anxiety behaviors during a given 30-s segment, or a
to which they felt self-conscious, were concerned about series of more moderate anxiety behaviors lasting the
how they might appear to others, were able to be entire 30 s.
themselves (reverse-scored), and were able to act natural
during the interaction (reverse-scored). We averaged Coding childcare performance
these items to yield an index of evaluation apprehension
(a ¼ :80). A judge who was blind to all procedures and hy-
After providing these self-reports, participants com- potheses watched each videotaped interaction and rated
pleted a filler survey and then indicated their sexual participantsÕ childcare performance. A second judge
orientation by circling Heterosexual/Straight, Homo- rated a subset (N ¼ 25; 38%) of the videotaped sessions;
sexual/Gay, or Bisexual at the top of another question- interrater reliability on this subset was good (intraclass
naire. Instructions on this questionnaire explained that r ¼ :84), so we used only the first judgeÕs ratings. Judges
heterosexual participants were now finished with the who rated childcare performance did not also rate non-
study; gay participants were instructed to answer a series verbal anxiety, and vice versa.
of questions about how ‘‘out’’ they were. On scales an- We asked judges to imagine that they were evaluating
chored with 1 (not at all) and 9 (extremely), gay men potential childcare workers who had applied for a job at
rated ‘‘How ÔoutÕ are you, even to people who donÕt a daycare center. As part of the interview process, each
know you?’’ and ‘‘To what extent do you feel comfort- job applicant was videotaped interacting with the chil-
able being public about your sexual orientation?’’ Next, dren; the judgeÕs job was to watch each interaction in its
gay men estimated, on scales of 0–100%, the percentages entirety, and then rate the applicantÕs overall perfor-
of their close friends, immediate family, extended family, mance. Unlike the non-verbal anxiety ratings, which
and extended social network who knew they were gay. were based on the frequency and intensity of specific
We standardized responses to these six out-ness items, anxiety behaviors, overall performance ratings were in-
then averaged them to yield an overall out-ness index tended to capture a Gestalt-like impression of the par-
(a ¼ :89). After completing these measures, participants ticipantÕs childcare skills. Judges used scales ranging
were debriefed and thanked. from 1 (not at all) to 5 (extremely) to indicate the extent
to which the participant ‘‘has a natural flair for working
Coding non-verbal anxiety with children,’’ ‘‘seems to enjoy working with children,’’
‘‘has the potential to offer high-quality childcare,’’
A judge who was blind to all procedures and hy- and ‘‘appears to communicate well with children.’’
potheses watched each videotaped interaction and rated Finally, judges rated the quality of each participantÕs
participantsÕ non-verbal anxiety in the presence of the performance on a scale of 1 (did a poor job) to 4 (did
children. A second judge rated a subset (N ¼ 19; 29%) of a great job).
the videotaped sessions; because interrater reliability on
this subset was good (intraclass r ¼ :88), we felt confi- Scoring non-verbal anxiety and childcare performance
dent that the ratings made by the first judge represented
an accurate picture of participantsÕ anxiety. Judges were Although we instructed the judges who coded non-
instructed to ‘‘consider the extent to which the target verbal anxiety and childcare performance to attend to
appears uncomfortable in his surroundings,’’ and look different aspects of participantsÕ behavior, it is possible
for ‘‘any behaviors that communicate discomfort, anx- that these two sets of ratings actually reflect the same
iety, awkwardness, or a similar emotion.’’ These be- underlying construct (anxiety). Indeed, we suspect that
haviors included fidgeting, chewing on lip, playing with anxiety does factor into performance in some interper-
hair, biting nails, nervous smiling, stiff posture, and sonal contexts. To ensure that our anxiety and perfor-
averting eyes, among others. mance indices were not redundant, we submitted the 10
Starting when the participant first entered the play- non-verbal anxiety ratings and the five childcare per-
room, judges made one rating every 30 s for 5 min, for a formance ratings to a principal axis factor analysis with
total of 10 ratings. All ratings were made on scales of 1 (no varimax rotation. Two factors emerged with eigenvalues
discomfort evident) to 5 (extreme discomfort evident).4 greater than 2.0. The first factor captured the five per-
formance ratings (all factor loadings > .84), and ac-
4
We used ‘‘discomfort’’ in the endpoints of the rating scales, rather
counted for 48.3% of the total common variance. The
than the more specific ‘‘anxiety,’’ because we wanted coders to attend second factor captured eight of the non-verbal anxiety
to a broad array of anxiety-related behaviors. ratings (all factor loadings > .50), and accounted for
J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255 251

Table 1
Zero-order correlations and descriptive statistics
I II III IV V VI M SD N
I Familiarity with children — ).41 ).60 ).13 .03 ).11 5.72 1.92 65
II Self-reported anxiety — .72 .12 ).10 ).37 3.92 1.48 64
III Self-reported evaluation apprehension — .08 ).09 ).14 4.38 1.78 64
IV Non-verbal anxiety — ).47 ).01 2.10 0.55 63
V Childcare performance (standardized) — .06 0.03 0.90 63
VI Out-ness — )0.01 0.79 28
*
p < :10.
**
p < :01.

14.9% of the total common variance; the remaining two Similarly, in the model predicting evaluation apprehen-
anxiety ratings loaded equally strongly on both factors. sion, participants who were more familiar with children
To create final non-verbal anxiety scores, we deleted the reported lower evaluation apprehension, b ¼ :74; tð56Þ
two cross-loaded ratings, then averaged the remaining ¼ 3:34; p < :01, and no other effects reached significance,
ratings (a ¼ :88). To obtain final childcare performance ts < 1:08; ps > :28.
scores, we standardized the five performance ratings and
averaged them (a ¼ :95). Thus, although participantsÕ Non-verbal anxiety
non-verbal anxiety and childcare performance were
moderately correlated (see Table 1), the results of the We expected stereotype-threatened—as compared to
above factor analysis suggest that these scores reflect non-threatened—gay men to exhibit more non-verbal
distinct constructs. anxiety during the interaction. To test this, we regressed
participantsÕ non-verbal anxiety scores onto their sexual
orientation, stereotype threat condition, and the inter-
Results action term (familiarity with children was unrelated to
non-verbal anxiety, so we did not include it in this
Descriptive statistics and correlations among all model). As expected, the interaction was significant,
variables appear in Table 1. Because we used regression b ¼ :63; tð59Þ ¼ 2:91; p < :01 (see Fig. 1).6 To examine
analyses to test our mediational hypothesis, we con- whether the interaction pattern confirmed our expecta-
ducted the primary analyses with regressions to be tions, we compared the non-verbal anxiety of gay men in
consistent. the prime and no-prime conditions using contrast cod-
ing (Cohen & Cohen, 1983). Specifically, we assigned
Self-reported reactions codes of 1, )1, 0, and 0, to the gay/prime, gay/no-prime,
heterosexual/prime, and heterosexual/no-prime condi-
We did not expect threatened and non-threatened gay tions, respectively. We also assigned codes correspond-
men to report different levels of self-reported anxiety or ing to the remaining two orthogonal contrasts: 0, 0, 1,
evaluation apprehension. To test this prediction, we re- )1 for the heterosexual/prime vs. heterosexual/no-prime
gressed participantsÕ anxiety and evaluation apprehen- contrast, and 1, 1, )1, )1 for the sexual orientation main
sion scores separately onto their sexual orientation (coded effect.
as heterosexual ¼ 0, gay ¼ 1), their stereotype threat When we entered these contrast codes into a regression
condition (coded as no-prime ¼ 0, prime ¼ 1), and the analysis predicting non-verbal anxiety, our main contrast
sexual-orientation  stereotype-threat interaction term. comparing participants in the gay/prime vs. gay/no-prime
Moreover, because participantsÕ familiarity with children conditions was significant, b ¼ :29; tð59Þ ¼ 2:39; p ¼ :02.
was correlated with their scores on the self-report vari- The remaining contrasts revealed that heterosexuals in
ables (see Table 1), we also entered familiarity with chil- the prime condition appeared somewhat less anxious than
dren as a predictor in this model (after mean-centering it), heterosexuals in the no-prime condition, b ¼ :21; tð59Þ
as well as all interactions with this variable. In the model ¼ 1:70; p ¼ :09, and there was no main effect of sexual
predicting anxiety, familiarity with children was the only orientation, b ¼ :03; t < 1.
significant predictor, b ¼ :72; tð56Þ ¼ 2:87; p < :01,
and no other effects emerged, all ts < 1:62; ps > :11.5

5 6
Only 64 participantsÕ data were used in analyses on the self-report Only 63 participantsÕ data were used in analyses on non-verbal
measures because one participant did not rate his reactions to the anxiety and childcare performance because poor lighting conditions
interaction task. rendered two of the videotaped sessions unusable.
252 J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255

Fig. 1. Non-verbal anxiety as a function of sexual orientation and


stereotype threat condition. Fig. 2. Childcare performance (standardized) as a function of sexual
orientation and stereotype threat condition.

Childcare performance
performance, b ¼ :41; tð58Þ ¼ 3:41; p < :01, but the
To test whether stereotype-threatened gay men exhib- sexual-orientation  stereotype-threat interaction was
ited poorer childcare skills than non-threatened gays, we not, b ¼ :30; tð58Þ ¼ 1:42; p ¼ :16. Finally, the results
regressed the index of childcare performance onto sexual of a Sobel (1982) test indicated that the path between the
orientation, stereotype threat condition, and the interac- predictor and outcome variables was significantly re-
tion term. The interaction was significant, duced when the mediator was included in the model,
b ¼ :55; tð59Þ ¼ 2:59; p ¼ :01 (see Fig. 2). When we re- z ¼ 2:27; p ¼ :02. Thus, non-verbal anxiety appeared to
gressed childcare performance onto the same contrast mediate the relation between stereotype threat and
codes described previously, we found that gay men in the participantsÕ childcare skills during the interaction.
prime condition performed significantly more poorly Despite the apparent strength of these findings, the
than gays in the no-prime condition, b ¼ :26; possibility remains that our indices of non-verbal anxi-
tð59Þ ¼ 2:15; p ¼ :036. Also, heterosexuals in the prime ety and childcare performance were based on similar
and no-prime conditions did not perform significantly non-verbal behaviors, and thus, somewhat confounded.
differently, b ¼ :18; tð59Þ ¼ 1:48; p ¼ :14, and there was a We view this potential overlap between anxiety and
trend for gay men to perform better than heterosexual performance as endemic to the domain of study: in
men overall, b ¼ :21; tð59Þ ¼ 1:75; p ¼ :086. interpersonal contexts, anxiety and performance may go
hand-in-hand (e.g., Brodt & Zimbardo, 1981; Schlenker
Mediational analyses & Leary, 1982). Although the results of the previously-
reported factor analysis increase our confidence that our
We followed the steps identified by Baron and Kenny non-verbal anxiety and performance indices captured
(1986) to test whether non-verbal anxiety mediated the distinct constructs, we urge readers to keep the potential
relation between stereotype threat and childcare per- overlap in mind when interpreting the results of the
formance. First, results reported above established that mediational analyses.
the predictor variable (the sexual-orientation  stereo-
type-threat interaction) was significantly related to both Follow-up analyses
the mediator variable (non-verbal anxiety) and the
outcome variable (childcare performance). Next, we We conducted several additional analyses to answer
regressed childcare performance onto non-verbal anxi- teraction task was relatively unstructured, leaving open
ety, sexual orientation, stereotype threat condition, and the possibility that participants in the different conditions
the sexual-orientation  stereotype-threat interaction might have selected themselves into qualitatively different
term. Non-verbal anxiety was significantly related to social interactions. To explore this possibility, we asked a
J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255 253

na€ıve observer to watch each videotaped interaction threat produces anxiety. Thus, given the generalizability
and answer the following questions: (1) Did the par- of the stereotype threat effect, it seems likely that future
ticipant establish an ongoing (longer than 1 min) in- researchers would benefit from using unobtrusive indi-
teraction with any child? (2) With how many girls and ces of anxiety.
boys did he interact? (3) Did he read a book to the In addition to establishing that stereotype threat
children? (4) Did he engage in physical play (e.g., anxiety can be captured non-verbally, the current find-
roughhousing, dancing) with any children? and (5) ings extend the literature by focusing on gay men. To
What was the overall activity level in the room (on a our knowledge, no previous research has explored
scale of 1 ¼ very little to 5 ¼ a lot)? In separate anal- whether or how gay men experience stereotype threat,
yses, we regressed each of these variables onto par- despite the numerous negative stereotypes associated
ticipantsÕ sexual orientation, their stereotype threat with this group. Here, a simple reminder of their stig-
condition, and the interaction term. No significant matized identity caused gay men to become anxious and
effects emerged, all ts < 1:30; ps > :20. Thus, partici- awkward around children, suggesting that people with
pants appeared to select themselves into social inter- concealable stigmas are vulnerable to the effects of ste-
actions that were similar along several important reotype threat. Croizet and Claire (1998) made a similar
dimensions. point when they showed that students of low socioeco-
Finally, we explored the extent to which gay menÕs nomic status—also a concealable identity—experienced
out-ness moderated their non-verbal anxiety and child- stereotype-threat-induced performance decrements.
care performance during the interaction, to determine Taken together, these findings attest to the power of
whether our stereotype threat effect was limited to gay stereotype threat: Even people whose stereotyped status
men who are relatively uncomfortable revealing their is virtually invisible are vulnerable to concerns about
sexual orientation to others (i.e., those low in out-ness). confirming negative stereotypes in other peopleÕs eyes.
In separate analyses, we regressed gay menÕs non-verbal Aside from making these theoretical points, our
anxiety and childcare performance onto their out-ness, findings also bear important practical implications: If
stereotype threat condition, and the out-ness  stereo- stereotypes can impair gay menÕs interpersonal skills
type-threat interaction. In both models, a main effect of during brief interactions with children, they might sim-
threat emerged, such that gays in the prime condition ilarly compromise gay teachersÕ abilities in the class-
were more non-verbally anxious, b ¼ :47, and per- room. Concern about appearing ‘‘dangerous’’ or
formed more poorly, b ¼ :42, than gays in the no- ‘‘perverted’’ may place gay schoolteachers in a state of
prime condition, ts > 2:32; ps < :04. No other effects chronic preoccupation with their behaviors, appearance,
reached significance in either model, ts < 1:02; ps > :32. and speech in the classroom, and this preoccupation
It therefore appears that our effect was not driven may usurp teachersÕ already-strained mental and/or
merely by a subset of highly self-conscious gay men. physical resources. Indeed, many gay teachers complain
that their constant self-monitoring efforts leave them
feeling stiff and inauthentic with their students (Kissen,
Discussion 1996; Olson, 1987). As one teacher put it: ‘‘My biggest
disappointment is the inability to get really close to my
Mounting evidence indicates that targets of negative students. . ..IÕm afraid to hug them. . ..If there were no
stereotypes suffer performance decrements when they barriers or fears between us, I could be doing a better
are reminded of their stereotyped status. Although the job’’ (Trent, 1978, cited in Fassinger, 1993, p. 130).
performance-impairing outcomes of stereotype threat Additional research in a more structured daycare or
are clear, the underlying cause of these outcomes has school setting could explore the extent to which stereo-
evaded researchers. The current findings thus make an type threat anxiety affects real teacher–student interac-
important contribution to the literature by providing tions, as well as teachersÕ effectiveness. StudentsÕ
some of the first evidence that non-verbal anxiety me- evaluations of their teachers, or even studentsÕ perfor-
diates the effects of stereotype threat on gay menÕs per- mance on standardized tests, could serve as especially
formance during an interpersonal task. Specifically, gay useful indices of teachersÕ effectiveness—presumably,
men who were reminded of their stigmatized identity such indices would avoid the problem of potential
before interacting with young children exhibited poorer overlap that plagued our non-verbal anxiety and child-
childcare abilities than gays who were not reminded of care performance measures, and would thus strengthen
their identity, and this effect appeared to be driven by researchersÕ ability to draw conclusions about media-
the menÕs non-verbal anxiety. Despite non-verbal signs tion. We believe that such research is vital given the
of distress, however, threatened gay men did not self- finding that, when not threatened, gay men appeared to
report heightened anxiety or evaluation apprehension. possess a natural aptitude for childcare work. Although
Had we relied solely on self-reports, as others have done, not predicted, this trend is consistent with research
we would have obtained little evidence that stereotype suggesting that gay men may be particularly well-suited
254 J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255

to careers, such as preschool teacher, that require social sufficient factors in stereotype threat. Journal of Experimental
and artistic abilities (Chung & Harmon, 1994). Social Psychology, 35, 29–46.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator
By contrast, then, are heterosexual men poorly-suited variable distinction in social psychological research: Conceptual,
to the job of preschool teacher? At first glance, our strategic, and statistical considerations. Journal of Personality and
findings seem to raise this possibility: Heterosexuals Social Psychology, 51, 1173–1182.
appeared relatively anxious and unskilled while inter- Blascovich, J., Spencer, S. J., Quinn, D., & Steele, C. (2001). African
acting with children in the no-prime condition. On the Americans and high blood pressure: The role of stereotype threat.
Psychological Science, 12, 225–229.
other hand, affirming their heterosexuality prior to the Brodt, S. E., & Zimbardo, P. G. (1981). Modifying shyness-related
interaction substantially (though not statistically signif- social behavior through symptom misattribution. Journal of
icantly) alleviated heterosexualsÕ discomfort. Thus, the Personality and Social Psychology, 41, 437–449.
same manipulation that hurt gay menÕs childcare abili- Brown, R. P., Charnsangavej, T., Keough, K. A., Newman, M. L., &
ties appeared to boost heterosexual menÕs abilities—a Rentfrow, P. J. (2000). Putting the ‘‘affirm’’ into affirmative action:
Preferential selection and academic performance. Journal of
pattern that distinguishes our findings from other ste- Personality and Social Psychology, 79, 736–747.
reotype threat research. The typical pattern of findings, Brownley, K. A., Hurwitz, B. E., & Schneiderman, N. (2000).
as exemplified in Steele and Aronson (1995), shows a Cardiovascular psychophysiology. In J. T. Cacioppo, L. G.
large effect of the stereotype threat manipulation on the Tassinary, & G. G. Berntson (Eds.), Handbook of psychophysiology
performance of stigmatized individuals and little or no (2nd ed., pp. 224–264). New York: Cambridge University Press.
Chung, Y. B., & Harmon, L. W. (1994). The career interests and
effect of the threat manipulation on non-stigmatized aspirations of gay men: How sex-role orientation is related. Journal
participants. Why were heterosexuals in our study af- of Vocational Behavior, 45, 223–239.
fected by the threat manipulation? Perhaps heterosexual Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation
men in the no-prime condition were anxious because analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
they considered childcare a feminine activity, and Crocker, J., Major, B., & Steele, C. (1998). Social stigma. In D. T.
Gilbert, S. T Fiske, & G. Lindzey (Eds.), Handbook of social
therefore feared that they would do poorly, or that psychology (Vol. 2, (4th ed., pp. 504–553). Boston: McGraw-Hill.
observers would assume they were gay. Alternatively, Croizet, J. C., & Claire, T. (1998). Extending the concept of stereotype
heterosexuals in the prime condition may have felt re- threat to social class: The intellectual underperformance of students
laxed after indicating that they were straight because from low socioeconomic backgrounds. Personality and Social
they knew that predominant stereotypes link homosex- Psychology Bulletin, 24, 588–594.
DePaulo, B. M. (1992). Nonverbal behavior and self-presentation.
uality—not heterosexuality—with child-molestation (for Psychological Bulletin, 111, 203–243.
related ideas, see Shih, Pittinsky, & Ambady, 1999). Dovidio, J. F., & Fazio, R. (1992). New techniques for the direct and
Clearly, this puzzle merits additional scrutiny and we are indirect assessment of attitudes. In J. Tanur (Ed.), Questions about
conducting follow-up studies to replicate and explore questions: Meaning, memory, expression, and social interactions in
this effect. In the meantime, we note that the domain of surveys (pp. 204–237). New York: Russell Sage Foundation.
Ekman, P., & Friesen, W. V. (1969). Nonverbal leakage and clues to
childcare may provide a particularly interesting context deception. Psychiatry, 32, 88–105.
for research because it appears to activate performance- Fassinger, R. E. (1993). And gladly teach: Lesbian and gay issues in
impairing stereotypes for a variety of different social education. In L. Diamant (Ed.), Homosexual issues in the workplace
groups. (pp. 119–142). Washington, DC: Taylor & Francis.
Finally, although we chose to study stereotype threat Frable, D. E., Blackstone, T., & Scherbaum, C. (1990). Marginal and
mindful: Deviants in social interactions. Journal of Personality and
anxiety in an interpersonal context here, we believe that Social Psychology, 59, 140–149.
our recommendation to utilize indirect anxiety indices Herek, G. M. (1991). Myths about sexual orientation: A lawyerÕs guide
can be applied to research in other contexts as well. For to social science research. Law and Sexuality: A Review of Lesbian
example, if stereotype-threatened individuals are ob- and Gay Legal Issues, 1, 133–172.
served while taking a math or verbal exam, their non- Herek, G. M. (2002). Gender gaps in public opinion about lesbians
and gay men. Public Opinion Quarterly, 66, 40–66.
verbal behaviors can be coded for evidence of anxiety Kissen, R. M. (1996). The last closet: The real lives of lesbian and gay
that a self-report scale might not capture. Regardless of teachers. Portsmouth, NH: Heinemann.
the context chosen for investigation, the findings re- Mohr, J. J., & Rochlen, A. B. (1999). Measuring attitudes regarding
ported here highlight the theoretical and practical in- bisexuality in lesbian, gay male, and heterosexual populations.
sight that is gained by recognizing that, where stereotype Journal of Consulting Psychology, 46, 353–369.
Neuberg, S. L., Smith, D. M., & Asher, T. (2000). Why people
threat anxiety is concerned, people may feel one thing stigmatize: Toward a biocultural framework. In T. F. Heatherton,
but say another. R. E. Kleck, M. R. Hebl, & J. G. Hull (Eds.), The social psychology
of stigma. New York: The Guilford Press.
Olson, M. R. (1987). A study of gay and lesbian teachers. Journal of
Homosexuality, 13, 73–81.
Osborne, J. W. (2001). Testing stereotype threat: Does anxiety explain
References race and sex differences in achievement? Contemporary Educational
Psychology, 26, 291–310.
Aronson, J., Lustina, M. J., Good, C., Keough, K., Steele, C. M., & Oswald, D. L., & Harvey, R. D. (2001). Hostile environments,
Brown, J. (1999). When white men canÕt do math: Necessary and stereotype threat, and math performance among undergraduate
J.K. Bosson et al. / Journal of Experimental Social Psychology 40 (2004) 247–255 255

women. Current Psychology: Developmental, Learning, Personality, Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects
Social, 19, 338–356. in structural equations models. In S. Leinhart (Ed.), Sociological
Pinel, E. C. (1999). Stigma consciousness: The psychological legacy of methodology (pp. 290–312). San Francisco: Jossey-Bass.
social stereotypes. Journal of Personality and Social Psychology, 76, Spencer, S. J., Steele, C. M., & Quinn, D. M. (1999). Stereotype threat
114–128. and womenÕs math performance. Journal of Experimental Social
Schlenker, B. R., & Leary, M. R. (1982). Social anxiety and self- Psychology, 35, 4–28.
presentation: A conceptualization model. Psychological Bulletin, Steele, C. M., & Aronson, J. (1995). Stereotype threat and the
92, 641–669. intellectual test performance of African Americans. Journal of
Schmader, T. (2002). Gender identification moderates stereotype Personality and Social Psychology, 69, 797–811.
threat effects on womenÕs math performance. Journal of Experi- Stone, J., Lynch, C. I., Sjomeling, M., & Darley, J. M. (1999).
mental Social Psychology, 38, 194–201. Stereotype threat effects on Black and White athletic performance.
Shih, M., Pittinsky, T. L., & Ambady, N. (1999). Stereotype Journal of Personality and Social Psychology, 77, 1213–1227.
susceptibility: Identity salience and shifts in quantitative perfor- Watson, D., & Clark, L. A. (1991). Self- versus peer ratings of specific
mance. Psychological Science, 10, 80–83. emotional traits: Evidence of convergent and discriminant validity.
Smart, L., & Wegner, D. M. (1999). Covering up what canÕt be seen: Journal of Personality and Social Psychology, 60, 927–940.
Concealable stigma and mental control. Journal of Personality and Waxer, P. H. (1977). Nonverbal cues for anxiety: An examination of
Social Psychology, 77, 474–486. emotional leakage. Journal of Abnormal Psychology, 86, 306–314.

Vous aimerez peut-être aussi