Page 1
INTERPERSONAL RELATIONS AND GROUP PROCESSES
Gender Differences in Automatic In-Group Bias: Why Do Women Like
Women More Than Men Like Men?
Laurie A. Rudman
Rutgers University
Stephanie A. Goodwin
Purdue University
Four experiments confirmed that women’s automatic in-group bias is remarkably stronger than men’s
and investigated explanations for this sex difference, derived from potential sources of implicit attitudes
(L. A. Rudman, 2004). In Experiment 1, only women (not men) showed cognitive balance among
in-group bias, identity, and self-esteem (A. G. Greenwald et al., 2002), revealing that men lack a
mechanism that bolsters automatic own group preference. Experiments 2 and 3 found pro-female bias to
the extent that participants automatically favored their mothers over their fathers or associated male
gender with violence, suggesting that maternal bonding and male intimidation influence gender attitudes.
Experiment 4 showed that for sexually experienced men, the more positive their attitude was toward sex,
the more they implicitly favored women. In concert, the findings help to explain sex differences in
automatic in-group bias and underscore the uniqueness of gender for intergroup relations theorists.
Around the world and throughout history, men have enjoyed
higher status than women. They continue to possess enormous
advantages in terms of political power and economic resources as
well as a greater endowment of perceived competence, rationality,
and physical strength. As a result, men and women alike automat-
ically associate male gender with power (Rudman, Greenwald, &
McGhee, 2001), evaluate male authority figures more favorably
than female counterparts (Rudman & Kilianski, 2000), and more
readily misattribute status to unknown men than to unknown
women (Banaji & Greenwald, 1995). On the home front, only
women are expected to surrender their family name when they
wed; couples value male more than female progeny; and children
are more likely to be named after their fathers than their mothers,
irrespective of their sex (Jost, Pelham, & Carvallo, 2002).
In short, men are culturally valued more than women. For this
reason, one might expect men to show stronger automatic in-group
bias (i.e., own gender preference) than women. Indeed, consider-
able evidence suggests that people who belong to the most socially
valued groups strongly and automatically favor their own group.
For example, Whites, Christians, and Stanford University students
showed more automatic in-group bias compared with Blacks
(Nosek, Banaji, & Greenwald, 2002a), Jews (Rudman, Greenwald,
& McGhee, 2001), and California state college students (Jost et al.,
2002), respectively. In addition, Rudman, Feinberg, and Fairchild
(2002) directly measured status perceptions and found they pre-
dicted automatic (but not reported) in-group bias. For example,
Jews who perceived that Christians were higher in status than Jews
tended to associate Christians with positive attributes and Jews
with negative attributes. The same pattern emerged for Asians
compared with Whites, poor people compared with rich, and
heavyweight people compared with slim. Finally, high-status
group members routinely showed stronger in-group bias than did
low-status group members, irrespective of the basis for their status
(e.g., religious, ethnic, economic, or physical attractiveness), but
only at the implicit level. These findings cohere with system-
justification theory (SJT; Jost & Banaji, 1994), which argues that
dominants should show stronger in-group bias than minorities at
the nonconscious level. This is because only minorities are subject
to the tendency to devalue one’s group as a means of justifying the
status quo. In essence, minorities may implicitly adopt society’s
negative view of their group even when it conflicts with their
conscious beliefs, whereas dominants are less likely to experience
this evaluative conflict (Jost, Burgess, & Mosso, 2001).
However, gender groups are a proven exception to this rule,
because men are less likely than women to show automatic in-
group bias (i.e., own gender preference). Whereas women strongly
prefer female gender when response latency techniques are used,
men typically show neutral gender attitudes (i.e., nonsignificant
Laurie A. Rudman, Department of Psychology, Rutgers, The State
University of New Jersey; Stephanie A. Goodwin, Department of Psycho-
logical Sciences, Purdue University.
This research was partially supported by National Science Foundation
Grant BCS-0109997. Portions of this research were presented at the An-
nual Convention of the American Psychological Society, Atlanta, Georgia,
May 2003. We thank Amanda Diekman, Naomi Ellemers, Peter Glick, and
John Jost for their helpful comments.
Correspondence concerning this article should be addressed to Laurie A.
Rudman, Department of Psychology, Rutgers University, Tillett Hall, 53
Avenue E, Piscataway, NJ 08854-8040. E-mail: rudman@rci.rutgers.edu
Journal of Personality and Social Psychology, 2004, Vol. 87, No. 4, 494–509
Copyright 2004 by the American Psychological Association 0022-3514/04/$12.00 DOI: 10.1037/0022-3514.87.4.494
494
Page 2
preference for either gender; Nosek & Banaji, 2002; Richeson &
Ambady, 2001). This sex difference in automatic in-group bias is
provocative because it contradicts numerous SJT-derived findings
regarding other groups who differ with respect to cultural status.
Apparently, men and women resist using societal evaluations to
inform their own group attitudes—but why?
The Uniqueness of Gender Relations
For many reasons, gender represents a special case for inter-
group relations theorists. In contrast to other groups, men and
women are in closer contact throughout their lives, which may lead
to complex and often ambivalent evaluations of the opposite
gender (Fiske & Stevens, 1993). In addition, men are dependent on
women for biological reproduction, domestic labor, and child
raising, which gives women substantial dyadic power over men
(Glick & Fiske, 1996). Of course, women are also dependent on
men (e.g., for protection, economic stability, and social status).
Further, heterosexual romantic relationships are based on love and
physical intimacy, which can result in implicit idealization of the
opposite sex (Rudman & Heppen, 2003). Thus, there are many
factors that might be expected to inhibit out-group devaluation for
men and women alike.
However, this analysis suggests that women (as well as men)
should possess automatic gender attitudes that are relatively free of
bias, when in fact, women strongly prefer women (i.e., are implic-
itly sexist). Moreover, we cannot assume that men’s absence of
implicit sexism (vis-a `-vis evaluation) signals that they are more
egalitarian than women. For example, compared with women, men
are more likely to associate female gender with negative traits
(e.g., incompetence, weakness, and coldness; Richeson & Am-
bady, 2001; Rudman, Greenwald, & McGhee, 2001) and subordi-
nate rather than leadership roles (Rudman & Kilianski, 2000).
These findings add mystery to the phenomenon of men’s weaker
in-group bias. If men readily associate women with negative traits
and low-status roles, why would they not also evaluate women
unfavorably?
Explanations for Sex Differences in Automatic Gender
Attitudes
The present research examined five possible explanations for
sex differences in implicit gender attitudes. Each is proposed as an
answer to the question, “Why do women automatically like women
more than men automatically like men?” Our hypotheses took into
account potential sources for implicit evaluations. Implicit and
explicit attitudes may be distinguishable, in part, because they
stem from different causes. For example, compared with con-
trolled responses, automatic biases may be more influenced by
cultural status, cognitive balance, and emotional conditioning
(Rudman, 2004). Because a prominent source for implicit inter-
group attitudes—the cultural status of one’s group—cannot ex-
plain pro-female bias for women (or men’s absence of in-group
bias), we sought to uncover potential countervailing influences that
collectively might account for the special case of automatic gender
attitudes.
Cognitive Balance
Consistent with balance theory (Heider, 1958), implicit in-group
bias is often predicted by a confluence of self-esteem and in-group
identity (an observation that led to the unified theory; Greenwald
et al., 2002). Using the Implicit Association Test (IAT; Greenwald,
McGhee, & Schwartz, 1998) to assess self-esteem, identity, and
in-group bias, Greenwald et al. (2002) found a pattern that can be
described as “If I am good and I am X, then X is good” (where X
represents any social identity).1For example, Whites showed
pro-White bias to the extent that they liked themselves and iden-
tified as White. Across five experiments, self-reports showed no
evidence of cognitive balance, suggesting that this mechanism
influences implicit but not explicit associations.
More central to our concerns, Greenwald et al. (2002) also
found that women preferred women to the extent that they liked
themselves and were female identified. Greenwald et al. termed
this phenomenon a balanced gender identity, although attitudes and
self-esteem were similarly balanced (see Footnote 1). Men’s gen-
der identities were not examined, but it was assumed they would
echo women. This is because men and women alike showed
balance in IAT research involving ethnicity, ageism, gender ste-
reotypes (see also Rudman, Greenwald, & McGhee, 2001), and
academic attitudes (math vs. the arts; see also Nosek, Banaji, &
Greenwald, 2002b). Nonetheless, we suspected that sex differ-
ences in balanced gender identities might be revealed precisely
because men possess robust implicit self-esteem and masculine
identity (e.g., Greenwald et al., 2002; Greenwald & Farnham,
2000), but surprisingly weak in-group bias (Nosek & Banaji, 2002;
Richeson & Ambady, 2001). According to the unified theory,
people high on self-evaluation and own group identity ought to, on
average, favor their group. Men do not favor their gender group, so
their gender identities are likely not balanced. In short, the unified
theory would predict no sex differences in balanced gender iden-
tities, but past results have suggested otherwise.
Therefore, our first hypothesis was that cognitive balance would
only influence women’s (not men’s) implicit gender attitudes. If
so, an important caveat to the unified theory would be observed.
More important, sex differences in balanced gender identities
would help to explain men’s weaker in-group bias. Women’s
(typically high) self-esteem and gender identity promotes own
group preference because their implicit cognitions are balanced. If
this is not true for men, men will lack a psychological mechanism
that bolsters automatic in-group bias. That is, an absence of bal-
ance would indicate that men’s (typically high) implicit self-
esteem and gender identity is rendered moot vis-a `-vis their gender
attitudes. The expected result would be stronger automatic in-
group bias for women than men.
Gender Stereotypes
Our second hypothesis relied on a possible similar source for
automatic and controlled gender attitudes. People of both sexes
report greater liking for women than men, due, in part, to viewing
women as more communal (the “women are wonderful” effect;
1This description is suitable when predicting in-group bias, but it is
arbitrary from the standpoint of the unified theory and could just as easily
have been stated as “If I am good and X is good, then I am X,” or “If X is
good and I am X, then I am good.” Indeed, the signature of the unified
theory is that all three models are observed. That is, the theory is supported
when the constructs of self-evaluation, group identity, and group evalua-
tion are evenly balanced at the automatic level.
495
GENDER DIFFERENCES IN IN-GROUP BIAS
Page 3
Eagly & Mladinic, 1994; see also Eagly & Mladinic, 1989; Eagly,
Mladinic, & Otto, 1994). If this effect has an implicit analogy, then
women may be preferred to the extent they are associated with
communal stereotypic traits (e.g., warmth). Is it the case that
automatic gender attitudes and stereotypes are linked? In the only
investigation we are aware of, Rudman and Kilianski (2000)
measured gender attitudes using evaluative priming (Fazio, Jack-
son, Dunton, & Williams, 1995) and the communality–agency
stereotype using the IAT, and found a first-order correlation (con-
trolling for gender) of r(66) ? .27, p ? .05. That is, people who
associated female gender with communality and male gender with
agency also automatically evaluated women favorably. However,
because the priming task confounded gender with a high- and
low-status manipulation, and because the task may measure per-
ceptually based (rather than category-based) evaluation (Living-
ston & Brewer, 2002), Rudman and Kilianski did not provide clear
support for an implicit link between attitudes and stereotypes.
Developmental Events
The remaining hypotheses were derived from the possibility that
emotional conditioning influences automatic attitudes (Rudman,
2004). It has been theorized that implicit attitudes may be informed
by past experiences (Greenwald & Banaji, 1995), including devel-
opmental events (Wilson, Lindsey, & Schooler, 2000), more than
explicit attitudes. Our third hypothesis stems from the fact that
people are generally raised in a manner that leads to initially
stronger maternal (vs. paternal) bonding. Consistent with this
view, infants are more likely to mirror their mothers’ facial ex-
pressions compared with their fathers’ and female strangers’ (e.g.,
Kahana-Kalman & Walker-Andrews, 2001; Montague & Walker-
Andrews, 2002). Because mothers remain the dominant caregivers,
early maternal experiences might spill over into gender attitudes. If
so, implicit evaluations of the woman with whom people interact
first (and often) might extend to women in general. In essence,
pro-maternal evaluations might promote pro-female evaluations
for men and women alike, particularly at the automatic level. The
result would be sex differences in implicit in-group bias because
maternal gender matches own gender only for women (not men).
Male Threat Perceptions
Our fourth hypothesis stems from evidence that implicit atti-
tudes may be more sensitive to threat perceptions compared with
self-reports. For example, implicit (but not explicit) anti-Black
prejudice covaried with amygdala activation in Whites exposed to
photos of unfamiliar Blacks (Phelps et al., 2000). Because the
amygdala is associated with fear conditioning, the inference is that
implicit biases are linked to emotional learning. Similarly, Rud-
man, Ashmore, and Gary (2001) found reduced anti-Black implicit
and explicit prejudice in Whites who volunteered for diversity
education. It is noteworthy that reductions in implicit (but not
explicit) biases were predicted by Whites’ reduced fear of Blacks.
In tandem, these findings suggest that people may automatically
prefer women because they are less threatening than men. As a
group, men are more likely than women to commit murder, assault,
and other violent crimes (for a review, see Myers, 2002). Conse-
quently, men are perceived as more violent and aggressive than
women (e.g., Cicone & Ruble, 1978), and both genders possess an
automatic association between male gender and negative potency
(e.g., violence, destruction, rage; Rudman, Greenwald, & McGhee,
2001, Experiment 2). Thus, it is conceivable that implicit male
threat beliefs might inhibit automatic pro-male evaluation, for men
and women alike. By contrast, explicit gender attitudes might not
be as sensitive to emotional conditioning and, therefore, male
threat associations (Phelps et al., 2000; Rudman, Ashmore, &
Gary, 2001).
The Influence of Sexuality
Our fifth hypothesis focused on emotional conditioning through
sexual gratification. Compared with women, men report greater
liking for and interest in sex (Baumeister, 2000; Baumeister,
Catanese, & Vohs, 2001; Oliver & Hyde, 1993). If men associate
sex with women (e.g., through heterosexual encounters), then the
more they like sex, the more they should like women. Although
automatic attitudes toward sex have yet to be investigated, there is
some evidence that men’s implicit idealization of women has a
sexual component. Specifically, Rudman and Heppen (2003)
found that men showed strong evidence of implicit romantic
fantasies when they associated female partners with roles pertain-
ing to sex (e.g., sex kitten, sex goddess), but no evidence when
partner associations consisted of fairy-tale roles (e.g., Cinderella,
Sleeping Beauty). These findings tentatively point to the possibil-
ity that men’s gender attitudes may be linked to sexual gratifica-
tion. Because women’s fantasies consisted of associating male
partners with chivalry and heroism (e.g., Prince Charming, White
Knight), women’s sexualized fantasies were not examined. None-
theless, women should show the same linkage between liking for
the opposite gender and liking for sex. That is, sex differences in
in-group bias are likely not due to a difference in the proposed
correlation. Instead, men’s greater enthusiasm for sex may lead
them to favor the opposite gender more (i.e., show weaker in-
group bias) compared with women.
Overview of Research
Experiments 1–4 examined potential predictors of automatic
gender attitudes. In line with past research, we expected women to
show stronger in-group bias than men (Nosek & Banaji, 2002;
Richeson & Ambady, 2001). Our primary objective was to assess
support for five possible explanations for this phenomenon (sex
differences in balanced gender identities, gender stereotypes, de-
velopmental events, male threat perceptions, and attitudes toward
sex). The overarching aim was to discover whether other sources
of implicit attitudes, including cognitive balance and emotional
conditioning, might override cultural status when predicting auto-
matic gender attitudes.
To that end, we examined factors likely to orient (a) only
women toward own gender (cognitive balance; Experiment 1); (b)
both genders toward women (gender stereotypes, maternal evalu-
ations and male threat associations; Experiments 2 and 3); and (c)
men toward women (sexual gratification; Experiment 4) as a
means of understanding sex differences in automatic in-group bias.
The goal was to illuminate a variety of influences on implicit
attitudes that in concert might effectively counter the SJT-derived
prediction that men (a high-status group) ought to possess stronger
automatic in-group bias than women (a low-status group). In so
496
RUDMAN AND GOODWIN
Page 4
doing, we sought to underscore the uniqueness of gender for
intergroup relations theorists as well as to provide reasons why
prominent theories of implicit intergroup bias, including SJT (Jost
& Banaji, 1994) and the unified model (Greenwald et al., 2002)—
which are well supported using a variety of other groups—either
fail to apply or might asymmetrically apply to women, respec-
tively, when the domain is gender attitudes.
Pretesting Attributes for the Attitude IATs
Each experiment used the IAT to assess implicit constructs
because past research has supported its construct validity (for
reviews, see Banaji, 2001; Greenwald & Nosek, 2001) and its
flexibility (e.g., Greenwald et al., 2002; Rudman, Greenwald,
Mellott, & Schwartz, 1999). Flexibility was essential, given the
wide range of constructs assessed. However, a problem with past
gender attitude research is the use of traditional IAT stimuli to
represent pleasant and unpleasant constructs (Greenwald et al.,
1998). Because many of these constructs are gender linked (e.g.,
love, warmth, smile, war, brutal, assault), the attitude measure is
confounded with sex stereotypes. To avoid this problem, we pre-
tested the stimuli used in the present research (N ? 204; 123
women). On a scale ranging from 1 (very feminine) to 7 (very
masculine), we selected only attributes that were judged, on aver-
age, to be gender neutral (i.e., within the range of 3.78 to 4.23). If
men continued to show less automatic in-group bias than women
with this measure, we could be more confident that semantic
associations are not responsible for sex differences in automatic
gender attitudes.
Experiment 1
Experiment 1 assessed implicit gender attitudes, self-esteem,
and gender identity. Consistent with past research, no sex differ-
ences were expected for self-esteem and gender identity, but
women’s in-group bias should be stronger than men’s. In replica-
tion of Greenwald et al. (2002), women should show automatic
female preference to the extent that they possess high self-esteem
and feminine identity (i.e., balanced gender identities). If men’s
gender identities are imbalanced, then male preference should not
be predicted by a confluence of their self-esteem and masculine
identity. If sex differences in balanced gender identities are re-
vealed, sources other than cognitive balance will be indicated for
men’s gender attitudes. In short, both genders should possess the
tools that promote in-group bias when cognitions are balanced
(i.e., high self-esteem and gender identity), but countervailing
factors may inhibit men from using them to their own group
advantage.
Method
Participants
Volunteers (N ? 84; 43 men, 41 women) participated in exchange for
partial fulfillment of an Introductory Psychology course requirement. Of
these, 52 (62%) were European American, 11 (13%) were Asian American,
6 (7%) were Latino, and 15 (18%) reported another ethnic identification.
Implicit Materials
Following Greenwald et al. (2002), participants completed three IATs.
The gender attitude IAT uses synonyms for male and female as the target
categories (e.g., woman, lady, girl, gal, man, sir, boy, guy). The pleasant
attributes are good, happy, vacation, gift, sunshine, paradise, holiday, and
heaven. The unpleasant attributes are bad, awful, disease, trouble, pain,
failure, poison, and disaster. The self-esteem IAT uses these pleasant and
unpleasant attributes with the target categories me and not me (e.g., I, me,
my, mine, it, they, them, theirs). The gender identity IAT also uses the
target categories me and not me, but the attributes are male and female
synonyms (used as target categories in the gender attitude IAT). High
scores indicate more positive evaluation of in-group compared with out-
group members (gender attitude IAT), self compared with others (self-
esteem IAT), and greater association of self with own gender compared
with the opposite gender (gender identity IAT). Tasks were counterbal-
anced so that participants completed the in-group ? pleasant, me ?
pleasant, and me ? in-group blocks either first or second.2
Procedure
Participants were escorted to a private cubicle where they completed the
gender attitude, self-esteem, and gender identity IATs (in that order). They
then completed explicit counterparts to these measures. As in past research,
there was no evidence for a balanced identity using self-reports (Greenwald
et al., 2002). Therefore, these findings are not discussed.
Results and Discussion
Gender Differences
High scores on the gender attitude, self-esteem, and gender
identity IATs reflect preference for own gender, self (compared
with others), and identification with own gender, respectively.
Planned contrasts showed that women scored higher than men on
the gender attitude IAT, t(82) ? 5.26, p ? .001 (Ms ? 176 and 64,
SDs ? 109 and 117, respectively), resulting in a strong in-group
bias effect for women and a moderate one for men (ds ? 1.55 and
0.55, respectively). In a surprising departure from past research,
women also scored higher than men on the gender identity IAT,
t(82) ? 2.49, p ? .001 (Ms ? 218 and 143, SDs ? 146 and 139,
respectively). Nonetheless, both men and women showed strong
effect sizes on this measure (ds ? 1.02 and 1.53, respectively).
Finally, there were no sex differences on the self-esteem IAT,
t(82) ? 1.60, ns (Ms ? 167 and 212, SDs ? 121 and 135, ds ?
1.65 and 1.30, for women and men, respectively). Thus, men and
women strongly identified with their gender and showed robust
self-esteem, but women showed dramatically more in-group bias
than did men. Because we used a measure that untangled evalua-
tion from gender stereotypes, this sex difference was not likely due
to semantic associations.
Balanced Gender Identity Analyses
Variables were standardized and submitted to a series of
regressions (Greenwald et al., 2002), separately for women and
2In each experiment, we took advantage of recent improvements in the
IAT’s scoring methods (Greenwald, Nosek, & Banaji, 2003). Instead of
computing IAT effects from a limited set of trials, we averaged them from
all trials, a procedure that diminishes unwanted task counterbalancing
effects (i.e., the counterbalancing of blocks within the IAT can influence
the effect size, but is less likely to do so when the new scoring procedure
is used).
497
GENDER DIFFERENCES IN IN-GROUP BIAS
Page 5
men.3The analysis most central to the present research regressed
gender attitude on self-esteem and gender identity and their inter-
action (Model 1). However, according to the unified theory, results
should be similar for two other equations (see Footnote 1). These
concern the regression of self-esteem on gender attitude, gender
identity, and their interaction (Model 2), and the regression of
gender identity on gender attitude, self-esteem, and their interac-
tion (Model 3). In each case, the interaction term is entered in Step
1, and the main effects are entered in Step 2. This is because the
unified theory argues that only the interaction term should signif-
icantly account for variance in each criterion. That is, after ac-
counting for the interaction, main effects should not contribute
substantial variance. Therefore, the hallmark of a balanced gender
identity is that the change score for R2should only be significant
for the interaction term (entered in Step 1). Moreover, when variables
are skewed in a positive direction (as they are in the present analyses),
the interaction coefficient should be positive in each step.
Table 1 shows the results of the analyses, separately for women
(left side) and men (right side). As can be seen, women showed a
perfectly balanced gender identity. The positive interaction terms
were the sole predictors of the criterion in each model, replicating
Greenwald et al. (2002). By contrast, men showed no evidence of
a balanced gender identity. Men’s interaction terms were uni-
formly weak, as were the main effects.
To illustrate the form of women’s interaction results for Model
1, Figure 1 displays the regression lines predicting in-group bias
from self-esteem for women scoring 2 standard deviations above
and below the mean on the gender identity IAT. As can be seen,
female-identified women implicitly favored their gender to the
extent they possessed self-esteem. By contrast, women low on
female identity were less likely to favor women if they possessed
self-esteem. Thus, gender identity moderated the relationship between
gender attitude and self-esteem exactly as the unified theory predicts.
In sum, Experiment 1 confirmed our suspicion that men do not
possess a balanced gender identity. On average, participants
showed substantial implicit self-esteem and identified with their
gender, but their confluence predicted in-group bias only for
women. In essence, women can be characterized as “If I am good
and I am female, females are good,” whereas men can be charac-
terized as “Even if I am good and I am male, men are not
necessarily good.” This sex difference in cognitive balance sug-
gests that a mechanism that promotes female preference in women
does not similarly contribute to male preference for men. As a
result, Experiment 1 provides support for our first explanation for
why women show stronger automatic in-group bias compared with
men. However, it also suggests the need to determine factors that
do predict gender attitudes for men, given that cognitive balance
did not do so. Moreover, we sought to find additional factors
(beyond cognitive balance) that would help to account for the fact
that women show surprisingly strong automatic in-group bias,
despite their relatively low cultural status.
3The unified model is concerned with interaction terms, not bivariate
relations (reported here for completeness’ sake). Gender attitudes corre-
lated marginally with gender identity for women, r(39) ? .27, p ? .09, and
men, r(41) ? ?.28, p ? .08, but in opposite directions. Gender attitudes
were not related to self-esteem for women, r(39) ? .13, ns, or for men,
r(41) ? ?.01, ns. Self-esteem was weakly related to gender identity for
women, r(39) ? .30, p ? .06, and for men, r(41) ? .24, p ? .12.
Table 1
Balanced Identity Regression Analyses for Women and Men (Experiment 1)
Balanced identity model
Women (N ? 41)Men (N ? 43)
Step
?
t
? R2
p
Step
?
t
? R2
p
Model 1 (gender attitude)
Self-Esteem ? Gender Identity
Self-Esteem ? Gender Identity
Self-Esteem
Gender Identity
1
2
2
2
.42
.41
2.86**
2.30*
0.82
0.55
.17.0071
2
2
2
.14
.13
.11
0.88
0.76
0.54
1.75
.02 .38
?.13
.10 .02 .66
?.28.07 .22
Model 2 (self-esteem)
Gender Attitude ? Gender Identity
Gender Attitude ? Gender Identity
Gender Attitude
Gender Identity
1
2
2
2
.42
.73
2.70*
2.85**
1.18
0.51
.16.011
2
2
2
?.02
.05
.07
.27
0.10
0.30
0.45
1.59
.00 .92
?.23
?.11.08.12.06.28
Model 3 (gender identity)
Gender Attitude ? Self-Esteem
Gender Attitude ? Self-Esteem
Gender Attitude
Self-Esteem
1
2
2
2
.49
.41
.12
.08
3.51**
2.21*
0.80
0.43
.24.0011
2
2
2
.03
.07
0.20
0.46
1.77
1.64
.001.84
?.27
?.26 .02 .70.14.06
Note.
coefficients are shown. High scores on gender attitude, self-esteem, and gender identity reflect in-group bias, self-esteem, and identification with own
gender, respectively. A balanced identity is observed when the interaction term is positive and when it alone contributes to significant variance in the
criterion (Greenwald et al., 2002).
* p ? .05. ** p ? .01.
All constructs were measured using the Implicit Association Test. The criterion for each model is in parentheses. Standardized regression
498
RUDMAN AND GOODWIN
Page 6
Experiment 2
Experiment 2 examined two reasons why men and women alike
might automatically lean toward pro-female evaluation. First, if
implicit gender attitudes reflect people’s early experiences with a
maternal caregiver, then a preference for mother over father might
extend to a preference for female gender (Greenwald & Banaji,
1995; Wilson et al., 2000). In essence, people may be emotionally
conditioned to implicitly favor women as a result of maternal
nurturing. To bolster our interpretation of the causal direction of
this relationship, we included an explicit caregiver index with the
expectation that people raised primarily by their mothers would
also show stronger automatic pro-female bias. If emotional con-
ditioning influences automatic attitudes more than self-reports,
explicit gender attitudes should not be influenced by maternal
preference. Second, we examined whether people who automati-
cally associated female gender with communality (and male gen-
der with agency) would implicitly prefer women, in an extension
of past research (Eagly & Mladinic, 1994). We used the power–
warmth stereotype IAT because past research has yielded robust
effect sizes for men and women alike (e.g., Rudman, Greenwald,
& McGhee, 2001a). If gender stereotypes provide a source for both
implicit and explicit gender attitudes, then the predicted relation-
ship should be similar for the IAT and self-reports.
Method
Participants
One hundred fifteen (47 men, 68 women) volunteers participated in
exchange for partial fulfillment of an Introductory Psychology course
requirement. Of these, 63 (55%) were European American, 28 (24%) were
Asian American, 10 (9%) were Latino, and 14 (12%) reported another
ethnic identification.
Implicit Materials
Participants completed three IATs. The gender attitude IAT measures
associations between men and women (represented by male and female
names; e.g., Linda, David) and the pleasant or unpleasant attributes used in
Experiment 1. The parent attitude IAT measures associations between
mother and father (represented by synonyms; e.g., mom, dad, mother,
father) and the gender attitude IAT’s evaluative attributes. The gender
stereotype IAT measures associations between the gender attitude IAT’s
male and female names and warmth and power. The warmth attributes are
warmth, nurture, nice, love, caring, gentle, kind, and warm. The power
attributes are power, strong, confident, dominant, potent, command, assert,
and powerful. IAT blocks were counterbalanced such that participants
either completed the female ? pleasant, mother ? pleasant, and female ?
warmth blocks first or second.
IAT effect scores were computed such that positive scores indicated
more positive evaluation of in-group compared with out-group members
(gender attitude IAT), mother compared with father (parent attitude IAT),
and greater association of warmth with female gender and power with male
gender (gender stereotype IAT).
Explicit Materials
Gender and parent attitude indexes.
pants reported their feelings toward the target groups (men, women,
mother, father) on scales ranging from 1 (very cold) to 10 (very warm). To
serve as IAT counterparts, two difference scores were computed, such that
high scores reflected greater liking for in-group compared with out-group
On four thermometers, partici-
Figure 1.
Experiment 1’s women (N ? 41). All constructs were assessed with the Implicit Association Test (IAT). A high
score on the gender attitude IAT indicates pro-female bias. Regression lines were estimated using a millisecond
index and unstandardized regression coefficients. SD ? standard deviation.
Regression lines predicting gender attitudes as a function of self-esteem and gender identity for
499
GENDER DIFFERENCES IN IN-GROUP BIAS
Page 7
members (the gender attitude index), and mother compared with father (the
parent attitude index).
The caregiver index.
Participants responded to two forced-choice
items, “The primary caregiver when I was young was my (mother or
father)” and “The parent who most actively raised me was my (mother or
father).” The difference between responses was computed such that high
scores indicated a maternal primary caregiver. The two difference scores
were averaged to form the caregiver index, r(113) ? .68, p ? .001.
Gender stereotype index.
Participants separately reported how much
they associated women and men with the 14 warm and powerful attributes
used in the gender stereotype IAT, on scales ranging from 1 (not at all) to
7 (very much). Reliability coefficients for the four separate indexes were
adequate (all ?s ? .84). The difference between these ratings for the
warmth and powerful attributes formed the female stereotype and male
stereotype indexes, respectively. High scores indicate greater association of
women than men with warmth and greater association of men than women
with power, respectively.
Procedure
Participants were escorted to a private cubicle where they first com-
pleted the gender attitude, parent attitude, and gender stereotype IATs. The
latter two were presented in counterbalanced order, a procedural variable
that did not influence results. Participants then completed demographic
items (age, gender, ethnicity) and the explicit measures in the order
described above.
Results and Discussion
Gender Differences
IAT effects.
measures, separately for men and women. As in Experiment 1,
men showed less implicit in-group bias than women, t(113) ?
Table 2 (top half) shows the results of the implicit
4.61, p ? .001. In addition, women scored higher than men on the
parent attitude IAT, t(113) ? 4.10, p ? .001. By contrast, there
were no sex differences for implicit gender stereotypes, t(113) ?
1.00, ns.
Examining whether IAT effects were significantly different
from zero revealed an absence of in-group bias for men, t(46) ?
1.58, ns. Also, men’s slight preference for their mothers on the
parent attitude IAT proved to be nonsignificant, t(46) ? 1.36, ns.
By contrast, women showed strong preference for their own group
and for their same-sexed parent, both ts(67) ? 9.50, ps ? .001.
Finally, both men and women showed robust automatic gender
stereotyping, both ts ? 4.50, ps ? .001, in line with past research
(Rudman, Greenwald, & McGhee, 2001).
Explicit measures.
Table 2 (bottom half) shows results of the
explicit measures. As with the gender attitude IAT, men showed
less in-group bias than did women, t(113) ? 5.58, p ? .001. As in
the past, men reported liking women more than men (e.g., Eagly et
al., 1994). By comparison, there were no reliable sex differences
on the parent attitude or caregiver indexes, all ts(113) ? 1.70, ns.
On average, participants favored their mothers and reported a
maternal primary caregiver. Finally (and in contrast to the IAT),
men scored higher than women on both the female and male
stereotype measures, both ts(113) ? 2.65, ps ? .05.
Does Maternal Evaluation Predict Pro-Female Attitudes?
As a first step, we recoded gender attitude scores for men so that
high scores would reflect pro-female evaluation. As a result,
positive correlations were expected among the gender and parent
attitude measures. Table 3 shows the results. Due to the sex
differences found on several of our measures, we computed first-
order correlations, controlling for participant sex.
Table 2
Summary Statistics for Implicit and Explicit Measures (Experiment 2)
Measure
Men (N ? 47)Women (N ? 68)
M SDdM SDd
Implicit
Gender attitude IATa
Parent attitude IATb
Gender stereotype IAT
39a
33a
100a
172
167
150
0.24
0.23
0.77
184b
143b
114a
159
121
113
1.11
1.01
0.88
Explicit
Gender attitude indexa
Parent attitude indexb
Caregiver indexc
Female stereotype
Male stereotype
?1.49a
1.51a
1.17a
1.88a
1.82a
1.89
2.08
1.09
1.16
1.37
?0.70
0.59
1.21
1.69
1.48
0.75b
0.69a
1.29a
1.32b
0.53b
2.26
2.82
0.88
1.08
1.13
0.35
0.27
1.33
1.19
0.42
Note.
male stereotype measures reflect rating women as warmer than men or men as more powerful than women,
respectively. A high score on the gender stereotype Implicit Association Test (IAT) reflects both of these
stereotypic associations. IAT effects are reported in millisecond index. Effect sizes (Cohen’s d) were computed
using the pooled standard deviation term for each measure. By convention, small, medium, and large effect sizes
correspond to .20, .50, and .80, respectively (Cohen, 1988).
aHigh scores reflect in-group bias.
reflect a maternal primary caregiver.
Means not sharing a subscript differ at the p ? .05 level or smaller. High scores on explicit female and
bHigh scores reflect preference for mother over father.
cHigh scores
500
RUDMAN AND GOODWIN
Page 8
As can be seen in Table 3, the relationship between the gender
and parent attitude IATs was reliably positive.4In addition, the
positive relationship between the caregiver index and the gender
attitude IAT was significant. Thus, people who implicitly favored
their mothers or who reported being raised primarily by a maternal
caregiver also automatically favored women. These findings are
consistent with the hypothesis that early experiences with a ma-
ternal caregiver may bias implicit attitudes toward women in
general.5
Among the explicit measures, the gender and parent attitude
indexes were not significantly related. In fact, the IATs were more
strongly related than were their self-reported counterparts (z ?
2.08, p ? .05). Thus, parental evaluations, presumably formed
early in life, were more likely to influence automatic than self-
reported gender attitudes. In addition, people raised primarily by
their mothers reported liking their mothers more than their fathers
(but not women more than men). Finally, there was some conver-
gence among the implicit and explicit measures of gender attitude,
parental attitude, and gender stereotypes.
Do Gender Stereotypes Predict Pro-Female Attitudes?
Past research has found that people favor women if they also
perceive them as more communal than men (e.g., Eagly et al.,
1994). As can be seen in Table 3, we replicated this effect using
self-reports; the female stereotype index was related to the gender
attitude index, as was the male stereotype index. Thus, people who
viewed women as warmer (but also weaker) than men reported
female preference. However, there was no support for an implicit
analogy, because the relationship between automatic gender atti-
tudes and stereotypes was weakly negative. Thus, people who
associated women with warmth (and men with power) were not
more likely to automatically favor them over men.
In sum, Experiment 2 confirmed that men possess less in-group
bias compared with women and supports the hypothesis that emo-
tional conditioning, through developmental events, may influence
automatic gender attitudes in two ways. First, the expected positive
link between parent and gender attitudes was observed, but only
using the IAT. Second, people who were raised primarily by their
mothers also implicitly favored women. These findings are con-
sistent with our suggestion that maternal attitudes (learned early in
life) spill over into gender attitudes, but only at the automatic level.
Finally, Experiment 2 showed no linkage between the gender
stereotype and attitude IATs, although gender stereotypes were
related to gender attitudes using self-reports (Eagly & Mladinic,
1994). Thus, implicit and explicit methods yielded pro-female bias
that appeared to stem from different influences (maternal evalua-
tions for the IAT, gender beliefs for self-reports).
Experiment 3
Experiment 3 tested the hypothesis that both genders may au-
tomatically prefer women because men are more threatening (i.e.,
associated with violence and aggression). We included explicit
measures for comparison purposes. Because past research has
suggested that emotional conditioning may influence implicit more
than explicit attitudes (Phelps et al., 2000; Rudman, Ashmore, &
Gary, 2001), the proposed linkage between male threat perceptions
and gender attitude might best be observed when using the IAT.
Finally, we included Experiment 2’s power–warmth IAT to pro-
vide a second check on gender stereotypes as a predictor of
automatic in-group bias.
Method
Participants
Volunteers (N ? 65; 32 men, 33 women) participated in exchange for
partial fulfillment of an Introductory Psychology course requirement. Of
4For men, this relationship was r(45) ? .40, p ? .01; for women it was
r(66) ? .43, p ? .001.
5The observed relationship between gender and parental attitudes
should be particularly strong for people raised primarily by their mothers.
Because only 6 people (5%) reported being raised by their father, whereas
99 people (86%) reported being raised by their mother, there was insuffi-
cient power to examine this hypothesis.
Table 3
Correlations Among Implicit and Explicit Measures (Experiment 2)
Measure
Implicit measuresExplicit measures
Gender
attitudea
Parent
attitudeb
Gender
stereotype
Gender
attitudea
Parent
attitudea
Caregiver
indexc
Female
stereotype
Parent attitude IAT
Gender stereotype IAT
Gender attitude index
Parent attitude index
Caregiver index
Female stereotype
Male stereotype
.42**
?.13
.19*
.11
.20*
?.06
.07
—
?.10
.14
.25*
.03
?.01
?.03
—
.07
.04
—
.17
.04
.29**
.19*
—
?.18
.29*
.21*
.38**
.16
.06
—
?.01
?.10
—
.35*
Note.
stereotype measures reflect rating women as warmer than men or men as more powerful than women,
respectively. A high score on the gender stereotype Implicit Association Test (IAT) reflects both of these
stereotypic associations. Correlations were computed using transformed (logged) response latencies.
aHigh scores reflect preference for female over male gender.
father.
* p ? .05. ** p ? .01.
Correlations are first order, controlling for participant gender. High scores on explicit female and male
bHigh scores reflect preference for mother over
cHigh scores reflect a maternal primary caregiver.
501
GENDER DIFFERENCES IN IN-GROUP BIAS
Page 9
these, 41 (63%) were European American, 15 (23%) were Asian American,
and 9 (14%) reported another ethnic identification.
Implicit Materials
The gender attitude and gender stereotype IATs were adopted from
Experiment 2. Experiment 3 added the gender threat IAT, which uses the
same target constructs (male and female names), but requires categorizing
these with both threatening (e.g., violent, danger, hazardous) and safe (e.g.,
harmless, trustworthy, innocent) attributes. Participants were randomly
assigned to receive the pro-female task conditions (i.e., female ? pleasant,
female ? safety, female ? warmth) either first or second. High scores on
the gender attitude IAT reflected stronger in-group bias. High scores on the
gender stereotype and gender threat IATs reflected greater association of
women with warmth and safety, respectively, compared with men.
Explicit Materials
The gender attitude index was adopted from Experiment 2 and scored to
match its IAT counterpart. New to Experiment 3, the gender threat index asks
participants how much they associate men and women (separately) with the
gender–threat IAT stimuli on scales ranging from 1 (not at all) to 7 (very
much). The reliabilities for threatening adjectives were acceptable (both ?s ?
.85), but not for safe adjectives (both ?s ? .55). We therefore computed this
index as the difference between associating men versus women with threat.
Procedure
Volunteers completed the gender attitude, gender threat, and gender
stereotype IATs (in that order). The gender attitude IAT preceded the
others to avoid the possibility of unwanted context effects. The IATs
always preceded the self-report measures of gender attitude and gender
threat (administered in that order).
Results and Discussion
Gender Differences
IAT effects.
measures, separately for men and women. As in Experiments 1 and
Table 4 (top) shows the results of the implicit
2, women showed more implicit in-group bias than men, t(63) ?
3.95, p ? .001. In addition, women scored marginally higher than
men on the gender threat IAT, t(63) ? 1.98, p ? .05. Replicating
Experiment 2, there were no sex differences for implicit gender
stereotypes, t(63) ? 1.00, ns. With the exception of men’s gender
attitude IAT scores, all measures were significantly different from
zero for both men and women, all ts ? 2.06, ps ? .05.
Explicit measures.
Table 4 (bottom) shows results of the ex-
plicit measures. First, men reported a preference for women,
whereas women reported a preference for men. However, as with
the gender attitude IAT, men preferred the opposite gender more
than women did (i.e., men showed less in-group bias), t(63) ?
2.99, p ? .01. Second, although both sexes reported male threat
beliefs, men’s were stronger than women’s, t(63) ? 2.40, p ? .05.
Finally, as in Experiment 2, men scored higher than women on
both the female and male stereotype indexes, both ts(63) ? 2.03,
ps ? .05. All measures were significantly different from zero for
both men and women, all ts ? 2.17, ps ? .05.
Do Male Threat Perceptions Predict Gender Attitudes?
Our main hypothesis was that male threat perceptions would
predict gender attitudes using the IAT. If emotional conditioning is
less influential for controlled attitudes, self-reports might not echo
this effect. As a first step, we recoded gender attitude scores for
men so that high scores would reflect pro-female evaluation. Table
5 shows the correlations among Experiment 3’s variables. Because
of sex differences in some relationships, we present them sepa-
rately for men (top) and women (bottom) in Table 5.
As can be seen, IAT results yielded the expected positive
correlation between gender attitudes and male threat associations
for both men and women. As anticipated, the corresponding ex-
plicit attitude–threat correlations were weak (in fact, they were
weakly negative). This suggests that emotional learning is a stron-
Table 4
Summary Statistics for Implicit and Explicit Measures (Experiment 3)
Measure
Men (N ? 32)Women (N ? 33)
MSDdMSDd
Implicit
Gender attitude IATa
Gender threat IATb
Gender stereotype IAT
7a
205
183
136
0.03
0.43
1.17
189b
141b
140a
163
118
124
0.90
0.90
1.13
67a
146a
Explicit
Gender attitude indexa
Gender threat indexc
Female stereotype
Male stereotype
?2.25a
2.14a
2.05a
1.83a
2.31
1.57
1.34
1.54
?0.81
1.45
1.60
1.35
?0.70b
1.30b
1.05b
1.16b
1.85
1.25
1.02
1.06
?0.27
0.88
0.81
0.87
Note.
male stereotype measures reflect rating women as warmer than men, or men as more powerful than women,
respectively. A high score on the gender stereotype Implicit Association Test (IAT) reflects both of these
stereotypic associations. IAT effects are reported in a millisecond index. Effect sizes (Cohen’s d) were computed
using the pooled standard deviation term for each measure. By convention, small, medium, and large effect sizes
correspond to .20, .50, and .80, respectively (Cohen, 1988).
aHigh scores reflect in-group bias.
gender with safety).
Means not sharing a subscript differ at the p ? .05 level or smaller. High scores on explicit female and
bHigh scores reflect association of male gender with threat (and female
cHigh scores reflect association of male gender with threat.
502
RUDMAN AND GOODWIN
Page 10
ger predictor of implicit than explicit gender attitudes. A test for
differences between the implicit and explicit attitude–threat corre-
lations for men and women bolstered this interpretation (both zs ?
2.21, ps ? .05).
Replicating Experiment 2, Table 5 also shows that the associa-
tion between the gender attitude and stereotype IATs was negli-
gible for both genders. Finally, the gender threat and stereotype
IATs positively covaried, although especially for men. That is,
men who associated women with safety also linked them to
warmth (or alternatively, men who associated men with power also
linked them to threat). This finding was weaker but in the same
direction for women. Because the stereotype and threat IATs both
tap gendered associations, these linkages are not surprising.
In sum, Experiment 3’s focal finding was the observation that
male threat perceptions are linked to implicit pro-female evalua-
tion, for men and women alike. These results support the hypoth-
esis that women may be implicitly preferred, in part, because they
are less intimidating than men. Because this relationship was not
observed using self-reports, Experiment 3 extends Experiment 2’s
support for the hypothesis that emotional conditioning influences
automatic more than controlled evaluations (see also Phelps et al.,
2000; Rudman, Ashmore, & Gary, 2001).
Experiment 4
Sexual experiences may represent another opportunity for emo-
tional conditioning. Experiment 4 examined whether (heterosex-
ual) men and women might automatically prefer the opposite
gender to the extent they liked sex. If men implicitly like sex more
than women do, then this gender difference would help to explain
men’s weaker in-group bias. That is, if attitudes toward sex and
gender groups are linked, men’s greater enthusiasm for sex might
promote greater liking for the opposite sex compared with women.
To test this hypothesis, we administered a sexual attitude IAT in
addition to the gender attitude IAT. We also measured sexual
experience because our hypothesis depended on an established
association between sex and the opposite gender (i.e., through
sexual encounters), and this was likely variable for our young adult
participants. That is, the pattern for men could be characterized as
“If I associate women with sex and I like sex, then I like women.”
For sexually inexperienced participants, there might not be a link
between their gender and sexual attitudes. Thus, we expected
sexual experience to moderate the proposed link between sexual
and gender attitudes.
Because the IAT is a relative attitude index, the choice of what
to use as a contrast for sex in the IAT required careful consider-
ation. Past research on smokers’ attitudes showed that no smoking
was the best contrast for smoking (Swanson, Rudman, & Green-
wald, 2001). Smoking was represented with scenes that included
an ashtray and a cigarette; no-smoking scenes excluded these
stimuli. Similarly, Experiment 4 contrasted sex with no sex in the
sexual attitude IAT. To do so, we relied on photos of heterosexual
couples either engaged in sexual activity (e.g., kissing) or not. The
decision to use couples for each contrast controlled for several
confounds, including physical contact (asexual couples were
shown touching or hugging) and love and romance. In addition, it
was important that both men and women be represented in each
contrast, to control for pro-female bias. Finally, although identical
couples would have been ideal, this choice would have made
distinguishing between sexual and asexual couples too difficult,
given that both sets of photos portrayed (nongraphic) physical
contact. Therefore, we used different couples for each contrast,
matched on physical attractiveness (as judged by the authors, in
concert with graduate student assistance).
Method
Participants
One hundred fifteen (58 men, 57 women) volunteers participated in
exchange for partial fulfillment of an Introductory Psychology course
requirement. Of these, 56 (49%) were European American, 29 (25%) were
Asian American, 9 (8%) were Latino, 8 (7%) were African American, and
13 (11%) reported another ethnic identification. From the original sample
(N ? 124), 2 men were excluded for failing to follow instructions, and 7
participants were eliminated because they were not exclusively heterosex-
ual (4 men, 3 women).
Implicit Materials
The gender attitude IAT was adopted from Experiments 2 and 3.
Experiment 4 added the sexual attitude IAT, which uses the same pleasant
and unpleasant attributes but requires participants to categorize these with
sex or no sex. To represent sex, we chose photographs of five couples
engaged in nongraphic sexual activity (e.g., kissing or passionately em-
bracing). Two were outdoors (on a beach or leaning against a boulder), and
three were in indeterminate locations (the focus was on their faces and
upper torsos). Four of these couples were scantily clad (e.g., in swimsuits).
To represent no sex, we chose photographs of five couples who were
touching in a nonsexual way (e.g., hugging). Three couples were outdoors
(e.g., in a meadow) and two were indoors (e.g., posed at a restaurant table).
One couple was scantily clad (shown frolicking in swimsuits in the ocean).
Participants were randomly assigned to receive the pro-sex and pro-female
task conditions (i.e., sex ? pleasant, female ? pleasant) either first or
Table 5
Correlations Among Implicit and Explicit Measures for Men and
Women (Experiment 3)
Measure
Gender
attitude
IATa
Gender
threat
IATb
Gender
stereotype
IAT
Gender
attitude
indexa
Men’s correlations
Gender threat IAT
Gender stereotype IAT
Gender attitude index
Gender threat indexc
.36*
.05
.09
.13
—
.45**
?.21
.29
—
?.03
.14
—
?.20
Women’s correlations
Gender threat IAT
Gender stereotype IAT
Gender attitude index
Gender threat indexc
.43*
?.02
.13
?.27
—
.29
.02
.17
—
?.07
.19
—
?.14
Note.
(IAT) reflects associating women with warmth and men with power.
Correlations were computed using transformed (logged) response latencies.
aHigh scores reflect preference for female gender.
association of male gender with threat (and female gender with safety).
cHigh scores reflect association of male gender with threat.
* p ? .05.** p ? .01.
A high score on the gender stereotype Implicit Association Test
bHigh scores reflect
503
GENDER DIFFERENCES IN IN-GROUP BIAS
Page 11
second. A high score on the sexual attitude IAT reflects greater liking for
sex.
Explicit Materials
The gender attitude index was adopted from Experiments 2 and 3.
Participants also reported their attitudes toward sex, kissing, and making
love on 10-point thermometers (1 ? very cold; 10 ? very warm). They
similarly rated their evaluation of celibacy, virginity, and chaste dating
(without sex). The sex and no-sex items were averaged separately (both
?s ? .68), and the difference between the two measures formed the sexual
attitude index, on which high scores reflect greater liking for sex.
Participants responded to two items to assess their sexual experience (“I
consider myself to be sexually experienced” and “I have never had a sexual
partner”) on a scale ranging from 1 (strongly disagree) to 5 (strongly
agree). After recoding the second item, we averaged these to form the
sexual experience index, r(113) ? .58, p ? .001.
Procedure
Participants were recruited for the “Automatic Attitudes” project, de-
scribed generically (i.e., no mention was made of sex). They completed the
gender and sexual attitude IATs in counterbalanced order (a procedural
variable that did not influence results). To help participants distinguish sex
from no sex, they studied the photographs of couples immediately before
completing the sexual attitude IAT. This procedure was successful in
affording accurate categorization of sex versus no sex (mean error rate ?
4.47%, which was similar to the mean error rate for the gender attitude
IAT, 4.39%). They then completed the gender and sexual attitude indexes,
followed by the sexual experience measure. Finally, they reported their
gender, race, and sexual orientation.
Results and Discussion
Gender Differences
High scores on the gender and sexual attitude measures reflect
preference for own gender and sex, respectively. As can be seen in
Table 6, men showed the typical pattern of weaker in-group bias
compared with women on the gender attitude IAT, t(113) ? 7.41,
p ? .001. Men also showed more favorable evaluation of sex on
the sexual attitude IAT, t(113) ? 2.96, p ? .01, in support of the
measure’s known groups validity (e.g., Baumeister et al., 2001).
These sex differences were echoed by the explicit measures, both
ts(113) ? 2.82, ps ? .01. Finally, there were no gender differences
in reported sexual experience, t(113) ? 1.00, ns.6
A check on the correspondence of our implicit and explicit
sexual attitude measures yielded a modest positive correlation,
r(113) ? .25, p ? .01 (when controlling for participant sex; r ?
.17, p ? .08). The implicit and explicit gender attitude measures
were negligibly related, whether or not we controlled for partici-
pant sex (both rs ? .08, ns).
In sum, men showed weaker in-group bias and greater liking for
sex compared with women, irrespective of the measures used. Our
next set of analyses sought support for the hypothesized links
between gender attitudes, sexual attitudes, and sexual experience.
Do Sexual Attitudes Predict Gender Attitudes?
As a first step, we recoded gender attitude scores for men so that
high scores would reflect pro-female evaluation. As a result,
positive correlations were expected among men’s gender and
sexual attitude measures, particularly for men high on sexual
experience. By contrast, sexually experienced women might be
expected to show negative links between their gender and sexual
attitudes (i.e., they should prefer men to women if they like sex).
6A series of analyses examined whether Experiment 4’s measures
differed from zero for men and women. First, and in contrast to Experi-
ments 2 and 3, men’s implicit in-group bias was reliable, t(57) ? 2.53, p ?
.05. However, women’s in-group bias was much stronger, t(56) ? 12.42,
p ? .001. Second, men’s automatic preference for sex was reliable, t(57) ?
2.20, p ? .05, whereas women weakly favored no sex, t(56) ? 1.80, p ?
.08. Third, men reported significant preference for women, t(57) ? 3.51,
p ? .001, whereas women reported neutral gender attitudes, t(56) ? 1.00,
ns. Finally, men and women alike reported liking sex more than no sex,
both ts ? 4.63, ps ? .001.
Table 6
Summary Statistics for Implicit and Explicit Measures (Experiment 4)
Men (N ? 58)Women (N ? 57)
MSDdM SDd
Implicit
Gender attitude IATa
Sexual attitude IATb
50a
66a
148
197
0.31
0.33
247b
?44b
177
203
1.52
?0.22
Explicit
Gender attitude indexa
Sexual attitude indexb
Sexual experiencec
?1.67a
3.21a
3.28a
3.63
2.43
1.50
?0.51
1.38
2.17
?0.05b
1.36b
3.15a
2.89
2.21
1.51
?0.02
0.59
2.09
Note.
effects are reported in a millisecond index. Effect sizes (Cohen’s d) were computed using the pooled standard
deviation term for each measure. By convention, small, medium, and large effect sizes correspond to .20, .50,
and .80, respectively (Cohen’s 1988).
aHigh scores reflect in-group bias.
scores reflect greater sexual experience.
Means not sharing a subscript differ at the p ? .01 level or smaller. Implicit Association Test (IAT)
bHigh scores reflect preference for sex (compared with no sex).
cHigh
504
RUDMAN AND GOODWIN
Page 12
Thus, we expected Participant Sex ? Sexual Attitudes ? Sexual
Experience interactions when predicting gender attitudes. If emo-
tional conditioning in the form of sexual gratification is a more
important source for implicit attitudes, then the IAT should yield
more support for this expectation compared with self-reports.
Gender attitude IAT.
To test this hypothesis, IAT scores were
standardized in preparation for analyses that regressed gender
attitudes on participant gender, sexual attitudes, sexual experience,
and all interaction terms. This analysis revealed the expected
three-way interaction (? ? ?.25, p ? .01).7Among women,
simple effects showed a main effect for sexual attitude (? ? ?.30,
p ? .05), such that if they automatically liked sex, they also
favored men. No other effects for women approached significance.
By contrast, men showed the predicted Sexual Attitude ? Sexual
Experience interaction (? ? .59, p ? .001).
To illustrate this effect, Figure 2 displays the regression lines
predicting pro-female attitudes from sexual attitudes for men scor-
ing 2 standard deviations above and below the mean on the sexual
experience index. As expected, men high in sexual experience
showed positive correlation between their sexual and gender atti-
tudes. This is consistent with our prediction that men who associ-
ated women with sex would prefer them to men to the extent they
liked sex. Although we predicted that the relationship between sex
and gender attitudes would be weak among men low in sexual
experience, we instead found a strong negative correlation (i.e.,
men low on sexual experience preferred own gender to the extent
they liked sex). Possible reasons for this unexpected finding are
addressed in the General Discussion section. Here, we note that
men’s automatic in-group bias is indeed influenced by their liking
for sex and that this relationship is moderated by sexual experi-
ence, as predicted.
Gender attitude index.
A comparable analysis predicting ex-
plicit gender attitudes yielded only a significant Sexual Experi-
ence ? Participant Sex interaction (? ? ?.34, p ? .01). Follow-up
tests for men showed a positive relationship between sexual ex-
perience and pro-female bias, r(56) ? .31, p ? .05. Similarly,
sexual experience and pro-male bias was associated for women,
r(55) ? ?.29, p ? .05. No other effects were significant, including
the three-way interaction (?? ?.10, ns). Thus, explicit attitudes
toward sex were not linked to explicit gender attitudes. Instead,
people who were sexually experienced tended to prefer the oppo-
site sex.
In sum, Experiment 4’s focal finding was support for the pre-
diction that men who liked sex and engaged in sexual activity
would automatically favor women over men. Thus, to the extent
that men are sexually experienced, their greater interest in and
liking for sex may promote automatic preference for the out-group
7There was also a Sexual Experience ? Participant Sex interaction (? ?
.28, p ? .01). Follow-up tests showed that men high on sexual experience
also tended to possess pro-female bias, r(56) ? .27, p ? .05. By contrast,
women did not show a reliable link between sexual experience and gender
attitudes, r(55) ? .13, ns.
Figure 2.
for Experiment 4’s men (N ? 58). A high score on the sexual and gender attitude Implicit Association Tests
(IATs) indicate pro-female and pro-sex bias, respectively. Regression lines were estimated using a millisecond
index for the IATs and unstandardized regression coefficients. SD ? standard deviation.
Regression lines predicting gender attitudes as a function of sexual attitudes and sexual experience
505
GENDER DIFFERENCES IN IN-GROUP BIAS
Page 13
(women). In addition, we found that women who implicitly liked
sex tended to prefer men on the gender attitude IAT, although this
linkage was not moderated by their sexual experience. Finally,
using self-reports, both women and men high on sexual experience
expressed greater liking for the opposite gender. Although the data
suggest differences in the correlates of implicit and explicit gender
attitudes, they cohere in one important respect. Taken together,
they strongly suggest that gender attitudes have a sexual compo-
nent. Although our findings are novel and intriguing in this regard,
their importance for the present investigation centers on their
ability to explain, in part, why men show relatively weak in-group
bias. Because men are implicitly more enthusiastic about sex, their
dependence on women for sexual relations may lead them to
automatically favor the opposite sex.
General Discussion
Although men, historically and cross-culturally, are the domi-
nant sex, they possess remarkably weaker in-group bias than do
women. In four experiments, we found this sex difference per-
sisted using both implicit and explicit measures, despite using a
gender attitude IAT that was unconfounded with gender stereo-
types. Averaging IAT effect sizes for men and women across four
experiments revealed ds ? 0.28 and 1.27, respectively. (The
comparable effect sizes using self-reports were ?0.67 and 0.06,
respectively.) In each case, a high score reflects greater in-group
bias. Thus, we can claim with confidence that even when men are
responding automatically, their in-group bias is surprisingly frail
and that women’s in-group bias is particularly strong at the im-
plicit level (i.e., stronger than men’s by a factor of 4.5).
Because this sex difference in automatic in-group bias contra-
dicts SJT (Jost & Banaji, 1994), it undermines cultural status as a
prominent source for implicit attitudes (Rudman, 2004). Thus, we
sought to identify factors that would counteract, for gender groups,
the general tendency for high-status groups to show robust in-
group bias (e.g., Jost et al., 2002; Nosek et al., 2002a; Rudman et
al., 1999, 2002). Out of five hypotheses tested, we found support
for all but gender stereotypes. That is, implicit power–warmth
stereotypes and gender attitudes were not related, although Exper-
iment 2 replicated their previously observed linkage using self-
reports (e.g., Eagly et al., 1994). So what then does predict implicit
gender attitudes? That is, what factors, when taken together, might
account for sex differences in automatic in-group bias?
Why Do Women Like Women More Than Men Like Men?
Besides SJT, the most systematically investigated source for
implicit attitudes is cognitive balance. According to the unified
theory, people with high self-esteem and gender identity should
favor their own gender, but only at the automatic level (Greenwald
et al., 2002). Because both men and women possess robust implicit
self-esteem and gender identity, both groups should show strong
in-group bias. However, because men’s weak in-group bias con-
tradicts this prediction, Experiment 1 tested (and supported) our
suspicion that only women have a balanced gender identity. Thus,
women’s automatic in-group bias is stronger, in part, because they
alone possess a cognitive mechanism that promotes own group
preference. Because men supported the unified theory in research
unrelated to gender attitudes (e.g., Greenwald et al., 2002; Nosek
et al., 2002b; Rudman, Greenwald, & McGhee, 2001), Experiment
1’s finding of sex differences in balanced gender identities is all
the more remarkable.
In addition, recent evidence suggests that emotional condition-
ing influences implicit attitudes (Rudman, 2004). Consistent with
this hypothesis, Experiment 2 showed that people who implicitly
preferred their mothers also favored women in general. Of course,
the causal direction of this relationship is unknown, but we suspect
that because of early (even preverbal) attachment to maternal
caregivers, people’s mental machinery may be geared to automat-
ically favor the feminine sex. In support of this interpretation,
people who reported being raised primarily by their mothers also
showed stronger pro-female bias, and the sequence of events is
clear in this case. Thus, Experiment 2’s results suggest that simi-
larities in how children are raised (primarily by their mothers) can
lead to automatic maternal preference, and this preference extends
to liking women in general. Finally, the absence of this phenom-
enon using self-reported attitudes suggests that emotional condi-
tioning is more pertinent to implicit than explicit evaluations.
Similarly, Experiment 3 supported threat perceptions as a source
for implicit attitudes (see also Phelps et al., 2000; Rudman, Ash-
more, & Gary, 2001). As expected, compared with women, men
were more automatically linked to threatening attributes (e.g.,
violence and aggression; see also Rudman, Greenwald, &
McGhee, 2001, Experiment 2). More important, male threat asso-
ciations predicted automatic pro-female bias for both men and
women. Thus, men’s greater proclivity for violence and aggression
may bolster automatic preference for women, the less threatening
sex. By contrast, self-reported attitudes were not reliably associ-
ated with explicit male threat beliefs.
Experiment 4 showed the power of sex to predict heterosexuals’
gender attitudes. As expected, men reported greater liking for sex
than did women, echoing past research (e.g., Baumeister, 2000;
Oliver & Hyde, 1993). Unique to Experiment 4, this sex difference
was also shown using the IAT. Thus, men showed greater enthu-
siasm for sex, irrespective of measurement method. We suspected
that this enthusiasm might lead men to show pro-female bias,
provided they associated women with sex (i.e., were sexually
experienced). This hypothesis was not supported using self-
reported attitudes; instead, sexual experience was the sole predic-
tor of gender attitudes. For both men and women, the more sexual
encounters they had, the more they reported a preference for the
opposite sex (irrespective of their liking for sex). By contrast, the
sexual attitude IAT predicted automatic gender attitudes. First,
women implicitly preferred men if they also liked sex. Second,
men echoed this pattern, but with an important caveat—only if
they were high on sexual experience. Thus, men supported our
expectation that if they associated women with sex (through sexual
encounters that likely lead to emotional conditioning), they would
prefer women to the extent they liked sex. By contrast, men low on
sexual experience implicitly disliked women to the extent they
liked sex.
Although we did not expect the latter result, in hindsight it
supports emotional conditioning as a source for implicit attitudes.
A plausible interpretation is that men who like sex and have their
sexual needs fulfilled by women tend to automatically favor
women, whereas men who want but are deprived of sex may
implicitly resent women. That is, if we assume that sexual grati-
fication (or its absence) is an affective experience for men, the
506
RUDMAN AND GOODWIN
Page 14
pattern of results is not surprising. Moreover, this interpretation
coheres with Glick and Fiske’s (1996) argument that men may
express benevolence toward women because they depend on them
(e.g., for sexual relations). If women are not forthcoming, their
dyadic power over men may backfire, resulting in implicit sexism.
Finally, we also did not expect sex differences in the pattern of
relationships observed using the IAT. Why would sexual experi-
ence be less likely to play a role in women’s automatic gender
attitudes? If sexual gratification is a weaker emotional need for
women (Baumeister et al., 2001), then the amount of experience
may be less important to their identity and well-being than it is for
men. In essence, the emotionally conditioned link between atti-
tudes toward sex and the opposite gender may be sufficient for
heterosexual women, who might be described as “If I like sex, I
like men,” but for men, sexual gratification may be required to
strengthen the same link (i.e., “If I like sex and receive it from
women, I like women”). However, we caution that the converse
description for sexually inexperienced men (“If I like sex and don’t
receive it from women, I don’t like women”) assumes they are
heterosexual when, in fact, their very lack of experience may
prevent them from knowing their sexual orientation. If that were
the case, our subset of inexperienced men may have included an
unknowable number of incipient homosexuals, for whom liking for
sex would be expected to covary with implicit male preference.
Thus, the negative link between implicit sexual and gender atti-
tudes shown on the part of sexually inexperienced men may be due
to sexual frustration, homosexual inclinations, or some combina-
tion of both.
In concert, the findings suggest that for men and women alike
there are numerous factors that inhibit the generally observed
linkage between social status and automatic in-group bias (e.g.,
Jost et al., 2002; Rudman et al., 2002). Although men are inargu-
ably the dominant sex, they are raised by their mothers, threatened
by male violence, dependent on women for sex, and deprived of
cognitive balance vis-a `-vis gender attitudes. Similarly, women are
the subordinate sex, but they are raised by their mothers, intimi-
dated by male violence, less enthusiastic about sex, and possess a
cognitively balanced gender identity, which bolsters in-group bias
for the majority of women. Thus, factors that we have character-
ized as emotional conditioning (maternal nurturance, male threat
perceptions, and sexual gratification) can help to explain why
cultural status—which would predict stronger implicit in-group
bias for men than women—may be overridden in the case of
gender attitudes. In addition, on average, women’s generally high
self-esteem and feminine identity bolster their automatic liking for
women, whereas men’s liking for men is not a similar function of
their generally high self-esteem and masculine identity. The pic-
ture as a whole is one of women having several sources that
promote automatic in-group bias (cognitive balance, a match be-
tween maternal caregivers and own gender, and male threat be-
liefs) that are either not shared by men (cognitive balance) or that
hinder male preference (maternal nurturance, male threat beliefs).
By contrast, both genders share a source that promotes out-group
preference (sexual gratification), but this is somewhat weakened
for women, who are less enthusiastic about sex. These discrepan-
cies help to explain why women show dramatically greater liking
for women than men show for men at the automatic level.
Limitations and Future Directions
The research is limited by its use of only college-aged samples.
To date, no evidence exists to inform us as to whether children or
more mature adults would also show sex differences in automatic
in-group bias. Thus, future research should examine gender atti-
tudes in both younger and older samples. In particular, the results
of Experiments 2 and 4 might not generalize to older adults if the
influence of being nurtured by maternal caretakers is diluted as
people mature and if the meaning of sexual encounters changes
significantly over time.
The present findings are also limited by our use of the IAT as
the sole implicit measure. As previously noted, this decision was
made because of the method’s flexibility—a critical attribute,
given the broad array of constructs measured in our work. None-
theless, future investigations will be needed to determine the
generalizability of our findings to other response latency methods.
This is particularly important given the novelty of the relationships
examined (e.g., between parent and gender attitudes and between
sexual and gender attitudes).8
The present research suggests that implicit and explicit gender
attitudes can be distinguished, in part, by differences in their
underlying sources (Rudman, 2004). However, the research did not
compare the relative strength of various predictors (e.g., maternal
evaluations, male threat perceptions, and attitudes toward sex),
which might be tested in future research. Here, we speculate that
some sources might prove stronger than others, depending on the
context in which attitudes are assessed (for a review, see Blair,
2002). For example, exposing people to violent rap music en-
hanced the positive relationship between automatic negative Black
stereotypes and pro-White bias (Rudman & Lee, 2002). Similarly,
priming people to think about male criminals might bolster the link
between implicit male threat perceptions and gender attitudes.
There are avenues left unexplored in our efforts to predict
automatic gender attitudes. One intriguing possibility is that men
may resist own group preference for fear that it would signal latent
homosexuality (Schmitt & Branscombe, 2001). At the explicit
level, men may deliberately curb their in-group bias in order to
present themselves as “all male” (i.e., strictly heterosexual). Im-
plicitly, men are not likely to be able to use this strategy, but a
history of being taught to avoid the appearance of being gay could
lead to automatic own group aversion. Implicit associations are, at
bottom, well-learned mental routes connecting, in this case, gender
and evaluation. If we liken these routes to mental ruts, then any
experiences that diminish the groove linking men to positive
evaluation should help to account for men’s weaker in-group bias,
including the fact that men are socialized (early and often) to avoid
homosexual labeling. Perhaps the greatest obstacle to testing this
hypothesis is the need for a valid measure of implicit anxiety
toward homosexual labeling. Here, we merely point out that men
represent a unique case for intergroup relations theorists because
liking for own gender may cast doubt on their sexual identity,
8At the present time, this suggestion is made with respect to testing the
robustness of the gender attitude predictors. Because there is debate as to
whether implicit measures assess similar or very different aspects of
attitude objects (Fazio & Olson, 2003), it would be premature to sug-
gest that using other response latency tools would uncover additional
mechanisms.
507
GENDER DIFFERENCES IN IN-GROUP BIAS
Page 15
whereas women may freely show liking for women (including
displays of affection) without being subject to homosexual
labeling.
Another possibility, derived from social identity theory, is that
women, by virtue of their lower status, may share a stronger
collective bond compared with men (Ellemers, Spears, & Doosje,
1997). For example, the rejection–identification model posits that
societal discrimination can lead to a desire for group mobility that
in turn creates stronger affinity among women than men (Schmitt,
Branscombe, Kobrynowicz, & Owen, 2002). Future research
might find this perspective useful for predicting sex differences in
automatic in-group bias.
Finally, future work should examine gender attitudes in specific
contexts that might decrease pro-female bias by increasing the
salience of gender status differences (e.g., in performance set-
tings). In fact, Eagly and Karau (2002) argued that women are
viewed more positively than men as long as they remain in
traditional roles; as a result, the “women are wonderful” effect
might be better termed the “women are wonderful when” effect—
when they are not in charge (e.g., Rudman & Kilianski, 2000).
Nonetheless, our experiments suggest that in the absence of spe-
cific power manipulations, women strongly implicitly prefer their
own gender, whereas men do not.
Conclusion
The present research sought to underscore the special case of
gender for intergroup relations theorists. The fact that women
show stronger automatic in-group bias than men is provocative,
because it suggests a reversal of sexism at the implicit level.
Despite the suggestion (and growing evidence) that implicit atti-
tudes stem from societal evaluations (e.g., Jost & Banaji, 1994;
Jost et al., 2002; Rudman et al., 2002), men and women alike fail
to conform to status-driven expectations when they automatically
evaluate each other. Although we uncovered several factors that in
tandem help to account for this phenomenon, we have no doubt
that we left many stones unturned. However, we hope that our
work provides the impetus for future investigations of implicit
gender attitudes and spurs progress toward understanding gender
relations, which are uniquely important for human happiness.
References
Banaji, M. R. (2001). Implicit attitudes can be measured. In H. L. Roedi-
ger, J. S. Nairne, I. Neath, & A. M. Surprenant (Eds.), The nature of
remembering: Essays in honor of Robert G. Crowder (pp. 117–150).
Washington, DC: American Psychological Association.
Banaji, M. R., & Greenwald, A. G. (1995). Implicit gender stereotyping in
judgments of fame. Journal of Personality and Social Psychology, 68,
181–198.
Baumeister, R. F. (2000). Gender differences in erotic plasticity: The
female sex drive as socially flexible and responsive. Psychological
Bulletin, 126, 347–374.
Baumeister, R. F., Catanese, K. R., & Vohs, K. D. (2001). Is there a gender
difference in strength of sex drive? Theoretical views, conceptual dis-
tinctions, and a review of relevant evidence. Personality and Social
Psychology Review, 5, 242–273.
Blair, I. V. (2002). The malleability of automatic stereotypes and prejudice.
Personality and Social Psychology Review, 6, 242–261.
Cicone, M. V., & Ruble, D. N. (1978). Beliefs about men. Journal of Social
Issues, 34, 5–16.
Cohen, J. (1988). Statistical power for the behavioral sciences. Hillsdale,
NJ: Erlbaum.
Eagly, A. H., & Karau, S. J. (2002). Role congruity theory of prejudice
toward female leaders. Psychological Review, 109, 573–598.
Eagly, A. H., & Mladinic, A. (1989). Gender stereotypes and attitudes
toward women and men. Personality and Social Psychology Bulletin, 15,
543–558.
Eagly, A. H., & Mladinic, A. (1994). Are people prejudiced against
women? Some answers from research on attitudes, gender stereotypes,
and judgments of competence. In W. Stroebe & M. Hewstone (Eds.),
European review of social psychology (Vol. 5, pp. 1–35). New York:
Wiley.
Eagly, A. H., Mladinic, A., & Otto, S. (1994). Cognitive and affective
bases of attitudes toward social groups and social policies. Journal of
Experimental Social Psychology, 30, 113–137.
Ellemers, N., Spears, R., & Doosje, B. (1997). Sticking together or falling
apart: In-group identification as a psychological determinant of group
commitment versus individual mobility. Journal of Personality and
Social Psychology, 72, 617–626.
Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995).
Variability in automatic activation as an unobtrusive measure of racial
attitudes. A bona fide pipeline? Journal of Personality and Social
Psychology, 69, 1013–1027.
Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition
research: Their meaning and use. Annual Review of Psychology, 54,
297–327.
Fiske, S. T., & Stevens, L. E. (1993). What’s so special about sex? Gender
stereotyping and discrimination. In S. Oskamp & M. Costanzo (Eds.),
Gender issues in contemporary society (pp. 173–196). Newbury Park,
CA: Sage.
Glick, P., & Fiske, S. T. (1996). The Ambivalent Sexism Inventory:
Differentiating hostile and benevolent sexism. Journal of Personality
and Social Psychology, 70, 491–512.
Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition:
Attitudes, self-esteem, and stereotypes. Psychological Review, 102,
4–27.
Greenwald, A. G., Banaji, M. R., Rudman, L. A., Farnham, S. D., Nosek,
B. A., & Mellott, D. S. (2002). A unified theory of implicit attitudes,
stereotypes, self-esteem, and self-concept. Psychological Review, 109,
3–25.
Greenwald, A. G., & Farnham, S. E. (2000). Using the Implicit Association
Test to measure self-esteem and self-concept. Journal of Personality and
Social Psychology, 79, 1022–1038.
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring
individual differences in implicit cognition: The Implicit Association
Test. Journal of Personality and Social Psychology, 74, 1464–1480.
Greenwald, A. G., & Nosek, B. A. (2001). The health of the Implicit
Association Test at Age 3. Zeitschrift fu ¨r Experimentelle Psychologie,
48, 85–93.
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding
and using the Implicit Association Test: I. An improved scoring algo-
rithm. Journal of Personality and Social Psychology, 85, 197–216.
Heider, F. (1958). The psychology of interpersonal relations. New York:
Wiley.
Jost, J. T., & Banaji, M. R. (1994). The role of stereotyping in system-
justification and the production of false consciousness. British Journal of
Social Psychology, 33, 1–27.
Jost, J. T., Burgess, D., & Mosso, C. (2001). Conflicts of legitimation
among self, group, and system: The integrative potential of system
justification theory. In J. T. Jost & B. Major (Eds.), The psychology of
legitimacy: Emerging perspectives on ideology, justice, and intergroup
relations (pp. 363–388). New York: Cambridge University Press.
Jost, J. T., Pelham, B. W., & Carvallo, M. R. (2002). Non-conscious forms
of system justification: Implicit and behavioral preferences for higher
508
RUDMAN AND GOODWIN