Insidious Dangers of Benevolent Sexism: Consequences for Women's Performance

Abstract

Four experiments found benevolent sexism to be worse than hostile sexism for women's cognitive performance. Experiments 1-2 showed effects of paternalist benevolent sexism and ruled out explanations of perceived sexism, context pleasantness, and performance motivation. Experiment 3 showed effects of both paternalist and complementary gender differentiation components of benevolent sexism. Benevolent sexism per se (rather than the provision of unsolicited help involved in paternalism) worsened performance. Experiment 4 showed that impaired performance due to benevolent sexism was fully mediated by the mental intrusions women experienced about their sense of competence. Additionally, Experiment 4 showed that gender identification protected against hostile but not benevolent sexism. Despite the apparently positive and inoffensive tone of benevolent sexism, our research emphasizes its insidious dangers.

Full text

INTERPERSONAL RELATIONS AND GROUP PROCESSES
Insidious Dangers of Benevolent Sexism: Consequences for Women’s
Performance
Benoit Dardenne
University of Lie`ge
Muriel Dumont
University of Lie`ge and Catholic University of Louvain
Thierry Bollier
University of Lie`ge
Four experiments found benevolent sexism to be worse than hostile sexism for women’s cognitive
performance. Experiments 1–2 showed effects of paternalist benevolent sexism and ruled out explana-
tions of perceived sexism, context pleasantness, and performance motivation. Experiment 3 showed
effects of both paternalist and complementary gender differentiation components of benevolent sexism.
Benevolent sexism per se (rather than the provision of unsolicited help involved in paternalism) worsened
performance. Experiment 4 showed that impaired performance due to benevolent sexism was fully
mediated by the mental intrusions women experienced about their sense of competence. Additionally,
Experiment 4 showed that gender identification protected against hostile but not benevolent sexism.
Despite the apparently positive and inoffensive tone of benevolent sexism, our research emphasizes its
insidious dangers.
Keywords: benevolent and hostile sexism, discrimination, performance and working memory, ingroup
identification, sense of competence
Sexism is expressed in a variety of ways, some more subtle than
others and some more hostile than others. Examples of this attitude
are gender-related humor, sexist name-calling, sexual harassment,
and employment discrimination. Gender discrimination, like any
other type of discrimination, is less and less accepted in our
modern societies, and legislation has been designed to promote
gender equality. In Europe, for instance, Article 141 of the Euro-
pean Commission’s treaty states that any job classification system
that is used for determining pay must be based on the same criteria
for both men and women and be so written as to exclude any sex
discrimination. Nonetheless, even if obvious forms of sexism are
often socially condemned, more subtle forms of sexism are not.
Worse, they sometimes are promoted, as they may sound positive
in tone or even seem gallant or chivalrous. Consider, for instance,
a man helping a woman to carry luggage, paying for her meal at a
restaurant, or complimenting her caring abilities. One might ask
what is wrong with that. The problem is that people do not notice
that these behaviors may be threatening. These kinds of behavior
might, however, be expressions of paternalism or else might am-
bivalently flatter women while simultaneously implicitly suggest-
ing their inferiority. Paternalism suggests that men should take
responsibility for the welfare of women, who might not be able to
carry luggage by themselves, might not earn enough money to pay
at the restaurant, and should be remembered for their qualities
(warmth) rather than their weak points (competence). In this arti-
cle, we propose that paternalism as well as the other aspects of
benevolent sexism, by suggesting women’s lack of ability, are
devastating to women’s performance.
Sexism Versus Patronizing
Hostile and benevolent sexism often go hand in hand and result
in ambivalent attitudes toward women (Glick & Fiske, 1996,
2001). Hostile sexism is an obviously antagonistic attitude toward
women, who are often viewed as trying to control men through
feminist ideology or sexual seduction. Benevolent sexism, on the
other hand, is a more positive attitude (often paternalistic, but not
necessarily) toward women that appears favorable but is actually
sexist because it portrays women as warm but incompetent or
weak individuals in need of men’s protection and support. Sexism
Benoit Dardenne and Thierry Bollier, Department of Cognitive Sci-
ences, University of Lie`ge, Lie`ge, Belgium; Muriel Dumont, Department
of Cognitive Sciences, University of Lie`ge, and Catholic University of
Louvain, Belgium.
This research was supported by an incentive grant I-06/07 from the
University of Lie`ge, Belgium, to Benoit Dardenne. We are very grateful to
Susan T. Fiske for her valuable help.
Correspondence concerning this article should be addressed to Benoit
Dardenne, Department of Cognitive Sciences, University of Lie`ge, Boule-
vard du Rectorat 5, B-32, B-4000, Lie` ge, Belgium. E-mail:
b.dardenne@ulg.ac.be
Journal of Personality and Social Psychology, 2007, Vol. 93, No. 5, 764 –779
Copyright 2007 by the American Psychological Association 0022-3514/07/$12.00 DOI: 10.1037/0022-3514.93.5.764
764

thus prescribes how roles and relationships between men and
women are or should be, suggesting the superiority of men over
women. Glick and Fiske (1996) developed the Ambivalent Sexism
Inventory (ASI) to tap the ambivalence manifested through the
combination of these two aspects involved in sexism, and they
conceptually distinguished the two subcomponents.
Benevolent sexism overlaps with patronizing behavior. Patron-
izing is acting and treating others in a condescending manner,
thereby suggesting the patronizing person’s personal superiority or
the other person’s inferiority. When enacted by a man toward a
woman, patronizing thus resembles sexism because one person or
category of persons is thought to be superior to another.
In a recent study, Vescio, Gervais, Snyder, and Hoover (2005)
investigated the consequences that patronizing behavior of the
powerful has on women’s performance in stereotype-relevant do-
mains. They examined the simultaneously inequitable and patron-
izing behavior of powerful men toward their subordinates. After
women were praised for their work but assigned a devalued rather
than a valued position, their performance decreased, but men’s
performance increased on a subsequent masculine task (i.e., a
leadership task implying strong strategic, planning, and competi-
tive skills). The argument that the authors developed was that
patronizing behavior was perceived as unfair and elicited anger
among subordinates who received simultaneous praise and deval-
ued position assignments. However, men seemed to consider per-
formance as a way to fight the initial injustice, whereas women did
not. Indeed, anger in such a patronized condition was positively
correlated with men’s performance but was uncorrelated with
women’s performance. This was presumably the case because
women did not see performance of the masculine task as being a
possible way to eliminate anger in that setting (see Harmon-Jones,
Sigelman, Bohling, & Harmon-Jones, 2003).
Importantly, however, nothing could be concluded about the mech-
anisms leading women to perform worse under patronizing behavior,
except that anger did not cause the decrease. Rather, Vescio et al.
(2005) suggested that women might doubt their capacity to challenge
the initial unfair position assignment. Indeed, they had poor perfor-
mance expectations; they felt less confidence and personal control
than did men in such a stereotype-relevant domain, even before the
manipulation. Because they might not feel that they had the resources
to cope with patronizing behavior, women perhaps experienced a sort
of “learned helplessness,” which would be responsible for their per-
forming badly (Vescio et al., 2005). Unfortunately, such an alleged
process was not examined directly.
Patronizing, as operationalized by Vescio et al. (2005), is a form
of paternalist sexism involving both a positive side (praising) and
a negative side (devalued position assignment), the latter suggest-
ing women’s alleged incompetence. Ambivalence is as present in
patronizing behavior as it is in benevolent sexism. Indeed, benev-
olent sexism is a chivalrous attitude toward women that neverthe-
less is sexist by praising women on characteristics usually associ-
ated with subordinates and suggesting their dependence on men.
However, although women’s inferiority is implicitly suggested
through benevolent sexism and often not even noticed by women
themselves (Barreto & Ellemers, 2005), patronizing involves a
component of obvious discrimination, or at least obviously unfair
behavior, suggesting women’s alleged lack of ability. Hostile
sexism similarly explicitly communicates men’s negative view of
women. Patronizing is thus perhaps closer to hostile sexism in its
explicit discriminatory character. In our view, Vescio et al.’s
(2005) patronizing behavior might thus be conceived as Glick and
Fiske’s (1996, 2001) overall ambivalent sexism, which reveals
itself through both hostile and benevolent sexist attitudes.
However, research has shown that hostile and benevolent sex-
ism, though correlated, predict different values (Abrams, Viki,
Masser, & Bohner, 2003; Ford, Wentzel, & Lorion, 2001; Sakalli
& Glick, 2003; Viki & Abrams, 2003). Our aim in this article was
to determine the specific consequences that hostile and benevolent
sexism each have on the performance of women. Also, we exam-
ined the specific processes that are associated with each form of
sexism. That is, we considered anger and motivation to perform on
the one hand and helplessness or self-doubt on the other hand.
Moreover, in order to fully appreciate sexism’s impact on the
performance of women and to preclude the context of a masculine
domain to be the source of self-doubt, threat, or feelings of
helplessness, we considered the context of a feminine domain
instead, in which there are no stereotype-based lower expectancies
for women than for men.
Why Would Hostile and Benevolent Sexism Differ in
Their Consequences for Performance and Associated
Processes?
One possibility is that hostile sexism, because of the explicit
discrimination it suggests, would be identified as sexism, so hostile
sexism would elicit anger and increased motivation to perform. On
the other hand, benevolent sexism, being more subtle and sweet
and implicitly suggesting women’s lack of abilities, would not be
characterized as sexism and would not elicit as much motivation to
perform as hostile sexism would. Indeed, according to appraisal
theories of emotion, anger is most often elicited when people
appraise their situation as unfair and due to another person’s
agency (Frijda, Kuipers, & ter Schure, 1989; Scherer, 1988; Smith
& Ellsworth, 1985). Along similar lines, reactance theory (Brehm,
1966; Brehm & Weintraub, 1977) proposed that when a restriction
is seen as unfair, people get an unpleasant feeling called reactance.
This is an intense motivational state that results in people feeling
that they must do something to get around the restriction. Increased
motivation to perform and better performance would actually be
means to override the initial unfair expression of sexism. Apprais-
ing the situation as unfair—that is, identifying sexism as such—
would be a prerequisite for experiencing anger and increased
motivation to perform. Although we believed that benevolent
sexism would not be identified as sexism, women might neverthe-
less perceive the situation as unpleasant and react angrily to it.
Therefore, we doubted that hostile and benevolent sexism would
differ in their capacity to elicit a motivation to perform.
Alternatively, we propose another mechanism by which sexism
might affect women’s performance. A considerable body of re-
search has suggested that activating group stereotypes or traits
leads to behaviors that conform to these constructs. For instance,
activating the concept of professors or thinking of oneself as being
a professor, even if role-played, has led people to become contex-
tually smarter, whereas activating the concept of the elderly led
people to become forgetful (e.g., Bargh, 1997; Chen & Bargh,
1997; Dijksterhuis, Aarts, Bargh, & van Knippenberg, 2000;
Dijksterhuis & van Knippenberg, 1998; see Dijksterhuis & Bargh,
2001; Wheeler & Petty, 2001, for recent reviews). Implicitly
765
DANGERS OF BENEVOLENT SEXISM

suggesting that women lack abilities through benevolent sexism
should then similarly lead women to perform badly. On the other
hand, hostile sexism (which suggests that women are mean and
disliked by men but does not imply their lack of competence)
should not elicit poor performance.
Another cognitive process that is complementary rather than in-
compatible with the automatic behavior research just mentioned also
provides an interesting explanation. Because benevolent sexism im-
plicitly suggests women’s lack of competence but provides ambiva-
lent feelings due to the praise it simultaneously involves, such a subtle
form of sexism might lead women to doubt their abilities to perform
well. This assumes that although benevolent sexism would be subtle
enough not to be identified as sexism, women confronted with such
sexism would nevertheless experience benevolent comments as quite
unpleasant. The implicit suggestion that women lack ability would
then lead women to doubt their capacities and experience a decrease
in self-esteem and self-confidence as well as to feel preoccupied
during their tasks. This would definitely impair their performance. In
contrast, hostile sexism would generate much less intrusive thought,
because hostility is explicitly manifested and external. Indeed, con-
trary to benevolent sexism, hostile sexism leaves no ambiguity as to
the opinion of those who express it. Unpleasant feelings might thus
easily be attributed to the speaker’s sexism. In contrast, when inferi-
ority of women compared with men is suggested implicitly and
accompanied by praise, whatever happens in the situation cannot
easily be attributed to the person expressing benevolence. We thus
proposed that benevolent sexism triggers self-doubt among women.
Such doubts and other intrusive thoughts then disturb performance,
especially when tasks require working memory capacity.
Working memory is involved in many daily activities; for in-
stance, when people have to do several tasks simultaneously or
when they have to concentrate on a resource-consuming task.
Working memory focuses attention on temporarily activated in-
formation necessary for the task at hand (e.g., writing good argu-
ments) while inhibiting irrelevant information (e.g., one’s col-
leagues’ discussion about a movie; Engle, 2001; Engle, Tuholski,
Laughlin, & Conway, 1999). In a test situation, working memory
should hold the data necessary to solve the problem and, at the
same time, suppress task-irrelevant information. Schmader and
Johns (2003), for instance, found that the threat elicited by the
salience of a negative stereotype reduced women’s math perfor-
mance by diminishing working memory. We suggested that be-
nevolent sexism would trigger intrusive thoughts that would sim-
ilarly interfere with performance by reducing working memory
capacity. Indeed, one part of the available working memory re-
sources would be devoted to manage self-doubt and intrusive
thoughts, rather than be fully allocated to performing the task. In
sum, benevolent but not hostile sexism should increase “mental
intrusions,” such as unwanted thoughts related to self-doubt, anx-
iety, preoccupation, or threatened sense of competence (e.g.,
Beilock & Carr, 2005; Beilock, Kulp, Holt, & Carr, 2004), which
should impair women’s working memory capacity and decrease
their performance.
Variables Potentially Protecting Against Sexism
Sexism conveys attitudes about women. Therefore, gender iden-
tification might well moderate the effects of sexism on mental
intrusions and performance. Indeed, there are individual differ-
ences in the tendency to detect discrimination (e.g., Stangor et al.,
2003). Some women more strongly identify with their gender
group and have a better constructed and more positive image of
their ingroup than other women. The former may actually be better
able than the latter to detect and blame discrimination in a primary
appraisal. In a secondary appraisal, women who identify strongly
could then more easily disregard negative ideas about their gender
group than could women who identify weakly. For example,
Major, Quinton, and Schmader (2003; for a review, see Major,
McCoy, Kaiser, & Quinton, 2003) found that highly identified
women were particularly prone to detect discrimination when it
was not certain but relatively likely. Also, according to Benokraitis
(1997; see also Crocker & Major, 1989), subtle discrimination is
particularly effective simply because it is fundamentally ambigu-
ous (i.e., people have a hard time deciding whether this is sexism
or politeness). Thus, although benevolence may appear less am-
biguous for strongly than for weakly identified women, they may
experience uncertainty and mental intrusions because of the in-
competence implicitly suggested to characterize women. In con-
trast, when confronted with relatively clear instances of hostile
sexism, strongly identified women might be better able than those
who are weakly identified to prevent intrusive thoughts from
interfering with performance by blaming the obviously sexist
attitude.
Overview
In a series of four experiments, we proposed to test the specific
consequences of hostile and benevolent sexism on women’s per-
formance within a feminine domain. Two alternative processes
were considered. On one hand, hostile sexism should be more
easily identified as sexism compared with benevolent sexism.
Benevolent sexism might nevertheless be experienced as quite
unpleasant. Therefore, we believed that neither type of sexism
would lead to differentials in women’s motivation to perform.
However, we proposed that benevolent sexism— by implicitly
suggesting women’s lack of ability and because such thoughts
could not easily be externally attributed—would trigger mental
intrusions and self-doubt that would interfere with working mem-
ory capacity and impair performance. In contrast, because hostile
sexism would be easily identified as sexism, related statements
would be more easily blamed on the person expressing them and
fewer mental intrusions should be experienced. We thus propose
that a cognitive rather than a motivational process would be
responsible for the deterioration of women’s performance under
benevolent sexism.
In all four experiments, we used a job selection context in which
the position required typically feminine characteristics. We exper-
imentally manipulated the type of sexism expressed (hostile vs.
benevolent) and measured its impact on the performance of
women in tasks involving working memory. Experiment 1 tested
the hypothesis that benevolent sexism, being subtle, would indeed
not be perceived as sexism by female participants, whereas hostile
sexism would be. Although performance was hypothesized to be
affected by the type of sexism, we expected equal motivation to
perform in both conditions. The inclusion of a control condition
with no expressed sexism also allowed us to examine whether
hostile sexism increased performance, whether benevolent sexism
decreased performance, or whether both impacted performance.
766
DARDENNE, DUMONT, AND BOLLIER

Experiment 2 tested the hypothesis that, although not identified as
such, benevolent sexism would be perceived as equally unpleasant
as hostile sexism but would nevertheless lead to worse perfor-
mance. In Experiment 3, we focused on two different manipula-
tions of benevolent sexism in order to ascertain that impaired
performance was due to sexism per se rather than to some other
variables. Finally, Experiment 4 examined the role of thought
intrusions elicited by benevolent sexism in causing the deteriora-
tion of performance. This fourth study also investigated the po-
tentially protective role of gender identification on women’s reac-
tions to sexism.
Experiment 1
Experiment 1 was presented as a simulation of a job selection
interview. The type of sexism was manipulated through the in-
structions allegedly given by a recruiter being hostile, benevolent,
or nonsexist in content. Participants were asked to imagine that
they were job applicants and to complete tasks that would inform
hiring decisions. In addition to the effect of benevolent and hostile
sexism on performance, we investigated the extent to which both
kinds of sexism were perceived as being sexism and examined the
impact on the motivation to perform. Following Barreto and Elle-
mers (2005), we predicted that participants confronted with be-
nevolent sexism would fail to recognize it as a form of prejudice,
whereas hostile sexism would be easily detected as sexism. We
hypothesized, however, that both hostile and benevolent sexism
would trigger a high motivation to perform. Motivation to perform
could be seen as a kind of angry revenge (see e.g. Kaiser & Miller,
2001). Indeed, benevolent sexism, although not identified as sex-
ism, might be perceived by women also as unpleasant. However,
because unpleasantness could not be externally attributed, benev-
olent sexism would trigger doubts and diminish performance.
Method
Participants
Thirty-eight relatively uneducated French-speaking women (25
to 42 years old; 6 to 9 years of education) were recruited from a
government school for adults to participate in a study of job
selection. This school helps poorly educated, unemployed adult
women to acquire better knowledge and skills in order to increase
their employment chances. We reasoned that unemployed adult
women were probably confronted by sexism in their everyday life,
and therefore sexism manipulations would be most relevant to
them. Participants were randomly assigned to expressions of one
of three types of sexism (hostile vs. benevolent vs. none). The job
for which they supposedly applied was described as requiring
typically feminine qualities. That is, the job description matched
jobs for which these women would usually apply.
Materials and Procedure
Cover story. The experiment was presented as part of training
for job interviews. All instructions were presented in written
format. Participants were told that a chemical factory was offering
new jobs and was currently employing men. The recruiting pro-
cedure included performing a test that was said to be well-known
and frequently used for job recruitment. Testing was performed in
group sessions as is often the case in real-world settings. After
performing the task and answering some additional questions,
participants were fully debriefed and thanked for their participa-
tion. Special care was directed to communicating the point of the
study because some participants may have felt badly, especially in
cases of low performance.
Manipulation of sexism. Instructions said to come from the
recruiter expressed hostile, benevolent, or nonsexist attitudes. Hos-
tile and benevolent sexist comments were based on matching items
of the ASI (Glick & Fiske, 1996; see Dardenne, Delacollette,
Gre´goire, & Lecocq, 2006, for a French validation). That is, our
manipulations included manifestations of hostile or benevolent
sexism such as the ones tapped by the scale validating these
concepts. In other words, sexist cues in our experimental design
corresponded to explicit expressions of the corresponding attitudes
measured by the ASI. Specifically, hostile sexism was reflected
through the ideas that women belong to the “weaker sex,” that they
get offended too easily, and that they often exaggerate their situ-
ation at work to get more favors. Items such as these suggest men’s
superiority over women and the idea of women being difficult in
order to gain more relative power. All these could be perceived by
women as unfair considerations. Alternatively, benevolent sexism
was reflected through the ideas that women are nice but that they
need to be helped by men, implicitly suggesting then that they lack
ability. Such a component of benevolent sexism refers to pater-
nalism.
Specifically, in both sexism conditions, it was explained that a
new law on quotas obliged the industry to follow certain employ-
ment rules. In the hostile sexism condition, the recruiter’s instruc-
tions were as follows: “Industry is now restricted to employ a
given percentage of people of the weaker sex. I hope women here
won’t be offended, they sometimes get so easily upset! You’ll
work with men only, but don’t believe what those feminists are
saying on TV, they probably exaggerate women’s situation in
industry simply to get more favors!” In the benevolent sexism
condition, the recruiter’s instructions were as follows: “Industry is
now restricted to choose women instead of men in case of equal
performance. You’ll work with men only, but don’t worry, they
will cooperate and help you to get used to the job. They know that
the new employee could be a woman, and they agreed to give you
time and help.” Sexist attitudes might thus be attributed to both the
recruiter and the potential colleagues in both sexist conditions. In
the no-sexism condition, the recruiter’s instructions contained
nothing more than the job description. As part of the cover story,
the fact that the job included working with men was salient in all
conditions.
Job description. The job was never described in great detail.
Participants were informed that the job required communication
and social skills as well as work-team orientation. The abilities
mentioned as required for the job were “sensitive to clients’
needs,” “cooperative orientation,” “good social abilities,” and “at-
tentive to clients.” That is, the job was presented as requiring quite
typically feminine characteristics, such that participants would
consider it as relevant to themselves and matching their personal
skills.
Test said to be the basis for hiring decisions. The test con-
sisted of nine problem-solving items. Participants were presented
with a 4 ⫻ 4 square representing an ocean with each intersection
corresponding to a specific location identified by a letter (going
767
DANGERS OF BENEVOLENT SEXISM

from F to U). Each item identified the location of a ship by a letter,
and participants’ task was to estimate the shortest way for the ship
to travel from that location to another assigned location. Each
intersection between two letters was said to equal a distance of 2
km. Importantly, some rules also had to be taken into account with
respect to the force and direction of both wind and current as
indicated by either one or two arrows (going from left to right,
from top to down or the other way around). One arrow indicating
the same direction as the ship’s direction meant that the distance
was 1 km less than otherwise; one arrow indicating a direction
opposite to the ship’s direction meant that the distance was 1 km
more than otherwise. Two arrows indicated a force equivalent to 2
km. Arrows directed on the ship’s sides were not to be taken into
account. Problem-solving items were presented as follows: “The
ship is on location I and destination is P, wind is 11 and current
is 2. How far is the distance from I to P?” The correct answer for
that item was 8 km. Because of the substantial information that had
to be taken into account, substantial working memory capacity was
required to perform the task. The test comprised 9 items specifying
the starting and ending locations of the ship as well as the direction
and force of both wind and current. The solution to each item could
be correct, incorrect, or not provided. Correct answers ranged from
0 to 9. Ten minutes were provided to complete the test.
Dependent variables. The participants’ performance was eval-
uated by the number of absolute correct answers provided on the
test. The number of items attempted was also analyzed as a
separate dependent variable. Perceived sexism was measured right
after the manipulation (“Do you find that the introductory text was
sexist?”). Motivation to perform was added as a potentially con-
tributing variable (eight items; e.g., “I wanted to prove something
during the test,” “I gave my best during the test,” and “I would like
to know my score on the test”; ␣⫽.65).
Results
All dependent variables were analyzed using a one-way, three-
level analysis of variance (ANOVA) with sexism (hostile vs.
benevolent vs. none) as a between-subjects variable.
Perceived Sexism
As expected, the sexism expressed by the recruiter had a sig-
nificant effect on perceived sexism, F(2, 37) ⫽ 10.49, p ⬍ .001,
␩
2
⫽ .37 (see Table 1
). Participants confronted by a recruiter
manifesting hostile sexism perceived more sexism (M ⫽ 5.00) than
did those who faced benevolent sexism (M ⫽ 1.64), t(25) ⫽ 4.56,
p ⬍ .001, ␩
2
⫽ .45, or no sexism (M ⫽ 2.64), t(22) ⫽ 2.75, p ⬍
.02, ␩
2
⫽ .26. The latter two conditions did not differ significantly
from one another, t(23) ⫽ 1.36, p ⬎ .18.
Test Said to Be the Basis for Hiring Decisions
Performance. The sexism expressed by the recruiter had a
significant effect on performance, F(2, 37) ⫽ 12.37, p ⬍ .001,
␩
2
⫽ .41 (see Table 1). When confronted by benevolent sexism,
participants performed worse (M ⫽ 1.57) than when the recruiter
manifested hostile sexism (M ⫽ 6.31), t(25) ⫽ 4.66, p ⬍ .001,
␩
2
⫽ .47, or no sexism (M ⫽ 4.64), t(23) ⫽ 3.98, p ⬍ .001, ␩
2
⫽
.41. The last two conditions did not differ significantly from one
another, t(22) ⫽ 1.41, p ⬎ .17.
Number of items attempted. There were no significant differ-
ences in the number of problems attempted by participants between
the benevolent (M ⫽ 7.71), hostile (M ⫽ 8.31), or no-sexism condi-
tions (M ⫽ 7.91), F(2, 37) ⬍ 1, ns. Overall, participants attempted to
solve a large proportion of items (89% of items).
Motivation to perform. Motivation to perform was not affected
by the type of sexism expressed through the recruiter’s instruc-
tions, F(2, 37) ⫽ 1.60, p ⬎ .21. Globally, participants manifested
quite a positive motivation to perform (M ⫽ 4.44, SD ⫽ 1.14).
Mediation Analysis
To examine whether the effect of sexism on performance could
be mediated by women’s perceiving sexism, we performed four
regression analyses (Baron & Kenny, 1986). First, performance
was regressed on the type of sexism (hostile sexism ⫽⫺1;
benevolent sexism ⫽ 1) as a unique predictor. Replicating the
above ANOVA, analyses revealed significant sexism effects, ␤⫽
⫺.68, t(25) ⫽⫺4.66, p ⬍ .001. Second, perceived sexism was
also regressed on the type of sexism. Again, the sexism of the
recruiter had a significant effect on perceived sexism, ␤⫽⫺.67,
t(25) ⫽⫺4.56, p ⬍ .001. Third, the analysis with perceived
sexism as a predictor of performance was significant, ␤⫽.66,
t(25) ⫽ 4.35, p ⬍ .001. Finally, performance was regressed on
manipulated sexism and perceived sexism simultaneously. As can
be seen in Figure 1, perceived sexism no longer significantly
predicted performance, ␤⫽.36, t(25) ⫽ 1.91, p ⫽ .07; the effect
was now much smaller and marginal. Manipulated sexism, on the
other hand, was still a significant predictor of performance, ␤⫽
⫺.44, t(25) ⫽⫺2.33, p ⬍ .05. A Sobel test also indicated only a
marginally significant reduction in the predictive power of sexism
on performance (z ⫽⫺1.76, p ⬍ .08). Thus, if anything, perceived
sexism only partially mediated the effect of the sexism manipula-
tion on performance.
Discussion
Results of this first experiment demonstrated that the type of
sexism a recruiter expressed toward female job applicants did matter
and impacted their actual performance. As predicted, it was under
conditions that would seem the most benign—when benevolent sex-
ism was expressed—that outcomes were most detrimental to women.
Consistent with the fact that benevolent sexism is usually much
Table 1
Perceived Sexism, Performance, Number of Items Attempted,
and Motivation to Perform in Experiment 1
Variable
Type of sexism
Hostile Benevolent None
Perception of sexism 5.00
a
1.64
b
2.64
b
Performance 6.31
a
1.57
b
4.64
a
No. of items attempted 8.31
a
7.71
a
7.91
a
Motivation to perform 4.89
a
4.24
a
4.16
a
Note. Means with different subscripts differ significantly at p ⬍ .05.
768
DARDENNE, DUMONT, AND BOLLIER

subtler and more accepted than hostile sexism (e.g., Barreto & Elle-
mers, 2005), results indicated that our benevolent sexism condition
led women to qualify the context as no more sexist than when no
sexism was present in the recruiter’s instructions. Still, benevolent
sexism significantly diminished the performance of women compared
with a context in which no sexism was expressed. In contrast, hostile
sexism was clearly identified as being a form of prejudice but did not
impair performance. Rather, expressing sexism through its hostile
form led women’s performance to be good enough and similar to that
in the control condition.
One possible explanation for the difference in performance
between benevolent and hostile sexism conditions could be that
hostile sexism— because it was identified as sexism and per-
ceived as unfair—increased women’s motivation to perform,
which in turn preserved their performance. Results, however,
did not support such a revenge- or anger-based reaction. In fact,
the type of sexism expressed by the recruiter did not affect the
motivation to perform. In addition, perceiving hostile sexism as
sexism did not fully mediate the positive impact it had on
performance, compared with its benevolent counterpart. Al-
though Experiment 1 ruled out the possibility that hostility led
to a revenge reaction that would increase motivation to perform,
it remained nevertheless plausible that hostility led to perceiv-
ing the job itself as more motivating to women than when they
were confronted with benevolence. Experiment 2 addressed the
possibility that a revenge reaction to hostility would be directed
to the way the job itself was evaluated.
Although not identifying benevolent sexism as being sexism
was perfectly consistent with the literature, the results might seem
surprising. If not perceived as such, how could benevolent sexism
affect performance? Not identifying benevolent sexism as a form
of prejudice does not mean that women considered benevolent
sexist manifestations to be similar to nonsexist contexts. Indeed, it
would be reasonable to think that, because benevolent sexist com-
ments could indicate well-intentioned attitudes as well as unpleas-
ant sexism, women might have hesitated to qualify such attitudes
as being sexist. Nevertheless, if, as we proposed, benevolent
sexism implicitly suggests that women lacked abilities, women
should have considered it to be as awkward as hostile sexism.
Experiment 2 examined the perceived pleasantness of the context
as a variable that covaries with sexism and indicates women’s
ability to distinguish expressions of benevolent sexism from situ-
ations in which no sexism is expressed.
Finally, we proposed to generalize the findings of Experiment 1
on other criteria. Experiment 1 used a simulated job selection that
was presented in a written format to a population of poorly
educated, unemployed women. In Experiment 2, we focused on a
population of university female students, thus a more educated
sample than participants in Experiment 1, and again manipulated
the type of sexism. In order to further increase the ecological
validity of our design, the experimenter role-played a recruiter
behaving in a hostile, benevolent, or nonsexist manner.
Experiment 2
Method
Participants
Forty-five undergraduate females from the University of Lie`ge
volunteered to participate in a study on job selection (age ranged
from 18 to 23 years old). They were randomly assigned to three
conditions of sexism (hostile vs. benevolent vs. none). They par-
ticipated in group sessions of 7 to 12 participants. In order to
ensure gender salience, we allowed males to participate in the
study so that every group session was gender mixed. Data for the
(very few) male participants were not analyzed.
Materials and Procedure
Materials and procedure were essentially identical to those in
Experiment 1, except that all instructions and manipulations were
role-played. In order to evaluate women’s capacity to distinguish
benevolent sexism from no sexism on a variable other than per-
ception of sexism per se, we asked participants how pleasant they
perceived the context. We also measured participants’ motivation
to get the job with a special focus on evaluations of the job itself.
Participants were then fully debriefed with special attention to the
reasons leading to their possibly having performed badly and were
thanked for their participation.
Cover story. The experiment was presented as being part of
the male experimenter’s practicum on work and organizational
psychology during which he had to train as a recruiting officer.
Therefore, participants were asked to imagine that they were job
applicants. Apart from that point, the cover story, job description,
and test said to be the basis for hiring decisions were identical to
those in Experiment 1. After the main test, participants were asked
to answer a series of questions aimed at helping the experimenter
improve his skills as a recruiting officer.
Manipulation of sexism. The male recruiter expressed either
hostile or benevolent sexist comments (explicit manifestations of
attitudes), or he expressed no sexism. The comments were identi-
cal to the written instructions presented in Experiment 1.
Dependent variables. Performance was evaluated by the num-
ber of correct answers provided on the test. The number of items
attempted was also analyzed as a separate dependent variable. In
addition, we measured how pleasant the context was perceived to
be. This included evaluations of both the recruiter and the inter-
view (six items: “The interviewer was nice,” “The interviewer was
competent,” “The interviewer was sympathetic,” “The interview
was nice,” “The interview was too long,” and “The interview was
frustrating,” with the last two items being reversed-coded; ␣⫽
.78), and the participants’ motivation to get the job (two items: “I
was really motivated to get the job” and “I’ve totally played the
game to get the job,” r(43) ⫽ .61, p ⬍ .001). All answers were on
a 9-point Likert like scale (from 1 ⫽ not at all to 9 ⫽ extremely).
Perception of sexism
)°63.( **66. **76.-
Hostile (-1)/ Performance
Benevolent (1) sexism -.68** (-.44*)
Figure 1. Perceived sexism as a mediator of performance in response to
sexism (Experiment 1). Coefficients in parentheses represent parameter
estimates for a regression model containing both predictors. °p ⫽ .07.
*
p ⬍ .05.
**
p ⬍ .001.
769
DANGERS OF BENEVOLENT SEXISM

Results
All dependent variables were analyzed in a one-way ANOVA
with a three-level between-subjects variable, sexism (hostile vs.
benevolent vs. none).
Motivation to Get the Job and Perceived Pleasantness of
the Context
As can be seen in Table 2, motivation to get the job was not
affected by the recruiter’s expressed sexism, F(1, 42) ⬍ 1. Anal-
yses of perceived pleasantness of the context revealed an effect of
the sexism manipulation, F(1, 42) ⫽ 15.53, p ⬍ .001, ␩
2
⫽ .43.
More specifically, both hostile (M ⫽ 6.53) and benevolent (M ⫽
6.06) sexism conditions were moderately positive and did not
differ from one another, t(33) ⫽ 1.19, p ⬎ .24. Importantly,
participants perceived both sexist contexts as significantly less
positive than the nonsexist condition (M ⫽ 8.30), ts(24 and 27) ⬎
4.55, ps ⬍ .001, both ␩
2
⬎ .46.
Test Said to Be the Basis for Hiring Decisions
Performance. As can be seen from Table 2, the type of sexism
expressed by the recruiter had a significant impact on performance,
F(2, 42) ⫽ 6.03, p ⬍ .005, ␩
2
⫽ .22. Replicating Experiment 1,
participants confronted with a recruiter expressing benevolent sex-
ist attitudes performed worse than did participants confronted by
either a hostile sexist recruiter, t(33) ⫽ 2.12, p ⬍ .05, ␩
2
⫽ .12, or
a nonsexist recruiter, t(27) ⫽ 2.97, p ⬍ .01, ␩
2
⫽ .25. The last two
conditions were not significantly different from one another,
t(24) ⫽ 1.81, p ⬎ .08.
Number of items attempted. The sexism expressed by the
recruiter had no significant impact on the number of items attempted,
F(2, 42) ⬍ 1. Globally, almost all participants attempted to solve each
of the nine items within the time allotted for the test (see Table 2).
Discussion
Results of Experiment 2 were straightforward. We perfectly
replicated Experiment 1’s results on performance with a different
sample of participants and with a role-play instead of written
instructions. Moreover, hostility did not lead participants to be
more motivated to get the job than benevolent sexism or neutral
conditions. Confirming Experiment 1, performing well when con-
fronted with hostile sexism was not the result of a revenge reac-
tion. Furthermore, both hostile and benevolent sexism led partic-
ipants to view the context as more negative than in the absence of
sexism. Importantly, although Experiment 1 suggested that women
did not explicitly identify benevolent sexism as being sexism, they
nevertheless considered it equally as unpleasant as hostile sexism.
That is, even if women did not identify benevolent sexism as such,
they were not insensitive to its expression.
Confronting benevolent sexism resulted in women performing
worse than when confronted with hostile sexism. This result could
definitely not be attributed to hostile sexism increasing women’s
motivation to perform as a revenge- or anger-related mindset.
Indeed, although performance was better when hostile rather than
benevolent sexist attitudes were faced, motivation to get the job
was equivalent in all three conditions. Therefore, we proposed that
two key elements would be responsible for women’s decreased
performance: first, the implicit suggestion that women lack ability
communicated through benevolent sexism, and second, the diffi-
culty of making external attributions for these statements (i.e.,
blaming the recruiter).
Still, our specific manipulation of sexism was open to an alter-
native interpretation. In Experiments 1 and 2, we manipulated
benevolent sexism through instructions that started with the idea
that the firm was required to hire women (“choose women instead
of men in the case of equal performance”) and then further warned
them that they would receive help from men. Following the work
of Heilman and others, this statement of preferential selection
policies and offer of unsolicited help could by itself have had a
negative effect on women’s performance (e.g., Heilman & Alcott,
2001; Nadler & Fisher, 1986; Schneider, Major, Luhtanen, &
Crocker, 1996; Turner & Pratkanis, 1994). However, we doubt that
the mention of preferential selection policies could be responsible
for the decrease in performance. Indeed, both hostile and benev-
olent sexism conditions included such references, though only
benevolent sexism impaired performance. Nonetheless, expressing
benevolent sexist attitudes such as providing help in conjunction
with the salience of preferential selection policies could have
raised some levels of suspicion regarding the sincerity of the
apparently positive attitude displayed. Yet another explanation of
our results would be that unsolicited help rather than benevolent
sexism per se affected the performance of women. As a matter of
fact, our manipulation of benevolent sexism focused on its pater-
nalistic component and suggested that men would help women
who would be hired. In order to rule out these alternative expla-
nations, we again tested the hypothesis of a detrimental effect of
benevolent sexism on performance in Experiment 3, but using a
manipulation of benevolent sexism that did not contain any offer
of unsolicited help and instructions that did not mention any
preferential selection policies.
Another specific aspect of our material could also be seen as
problematic. Indeed, even though we presented the job as requiring
feminine qualities, the task said to be the basis for the hiring
decision might have been seen as rather masculine. Even if we
took great care not to imply any stereotypic expectations, our task
nevertheless required visual-spatial abilities with which women
might have felt uncomfortable. Selection procedures in real-world
settings often make use of psychometric tests, whatever the job
that is being applied for, assessing mental rotation, visual-spatial
abilities, logical deduction, or mathematic abilities. However, par-
ticipants might have perceived this task as having little face value
in tapping the qualities required for the job or might have per-
Table 2
Motivation to Get the Job, Pleasantness of the Context,
Performance, and Number of Items Attempted in Experiment 2
Variable
Type of sexism
Hostile Benevolent None
Motivation to get the job 6.16
a
5.94
a
6.85
a
Pleasantness of the context 6.53
a
6.06
a
8.30
b
Performance 7.50
a
5.63
b
8.60
a
No. of items attempted 8.94
a
8.95
a
9.00
a
Note. Means with different subscripts differ significantly at p ⬍ .01.
770
DARDENNE, DUMONT, AND BOLLIER

ceived the task in itself a threat. Because the task was the same in
all three conditions but led to low performance only when benev-
olent sexism was expressed, we doubt that this would be the case.
Indeed, why would women doubt their abilities to perform such a
task only when benevolent sexism was expressed, except if be-
nevolent sexism was suggesting their inferiority? Nevertheless,
one could argue that the ideas suggested by benevolent sexism
were relevant only when women were confronted with a task that
was ambiguous enough to be perceived as either stereotype-
relevant or -irrelevant. In order to rule out that hypothesis and
examine whether benevolent sexism per se had an impact on
performance, we used a performance task in Experiment 3 that
could not be seen as masculine in any of its characteristics and
would thus provide better face validity in regard to the qualities
required for the job.
Experiment 3
Experiment 3 was presented as a simulation of a job selection
interview. No mention of preferential selection was made in the
instructions, and all of them were presented in a written format.
The type of sexism was manipulated through instructions mani-
festing benevolent sexism including either the paternalistic or the
complementary gender differentiation component. Paternalism
consisted of providing unsolicited help. Complementary gender
differentiation consisted of magnifying and emphasizing differ-
ences between men and women in a way that favored women
without any mention of helping behavior. An additional group
received instructions that were not sexist in content. Participants
were asked to imagine that they were job applicants and they had
to complete a task on the basis of which hiring decisions would be
made. For the sake of generalization and to rule out potential
alternative explanations in terms of the nature of the task, we
included a performance task that could not be seen as masculine or
stereotype-relevant in any of its characteristics.
Method
Participants
Thirty women recruited on the campus of the University of
Lie`ge volunteered to participate in a study of job selection (age
ranged from 18 to 22 years old). They were randomly assigned to
written instructions conveyed by a recruiter who expressed one of
three types of sexism (benevolent sexism with help vs. benevolent
sexism without help vs. no sexism). The firm to which they suppos-
edly applied was described as requiring typical feminine qualities.
Materials and Procedure
Cover story. All instructions were presented in written format.
Participants were asked to imagine that they had been searching
for a job for several months and had applied for one that they
thought would be interesting. They had just received an invitation
to go to the interview. The firm proposing the job was known for
working with associates who were attentive to clients and sensitive
to their needs, provided collaboration, and manifested good social
skills (feminine qualities). Participants were told that the firm was
a chemical factory that offered new jobs and currently employed
men who would be the hired candidates’ colleagues. The recruiting
procedure was said to include performing a test that is well-known
and frequently used for job recruitment. After performing the task,
participants were fully debriefed with a special focus on the
reasons for possibly having performed badly and then thanked for
their participation.
Manipulation of sexism. The written instructions conveyed
attitudes expressing benevolent sexism with help (paternalist com-
ponent), benevolent sexism without help (complementary gender
differentiation component), or no sexism. Benevolent sexist com-
ments were based on matching items of either the paternalistic or
complementary gender differentiation subcomponents of the ASI
(Glick and Fiske, 1996; see Dardenne et al., 2006, for a French
validation). In the benevolent sexism with help condition, instruc-
tions were as follows: “You have to know that women who may be
hired will work with men only. This should not be a problem
because all of them agreed to cooperate and help new female
colleagues to get used to the job. Moreover, they are fully aware of
the importance of helping newly hired women, and all of them
agreed to take on the time to help you to adapt the best.” In the
benevolent sexism without help condition, instructions were as
follows: “You have to know that women who may be hired will
work with men only. This should not be a problem because they
are fully aware of the importance of hiring women in their firm.
Indeed, all of them think that the presence of women, who are
more cultured and sophisticated than men, will allow the firm to
benefit from their moral sense and refined taste, whereas these
aspects are often lacking where only men work.” Sexist attitudes
might thus be attributed to both the recruiter and the potential
colleagues in both sexist conditions. In the no-sexism condition,
instructions contained nothing more than the description of the
firm and cover story, including salience of working with men. In
neither case did the recruiter mention any law about quotas or
preferential selection in the hiring of women.
Test said to be the basis for hiring decisions. A modified
version of the Daneman and Carpenter (1980) Reading Span Test
(RST) was used. This test had been developed to measure working
memory. The test is known to be a good predictor of reading
comprehension and verbal reasoning ability (e.g., Daneman &
Carpenter, 1980; Daneman & Merikle, 1996). The use of the RST
as a measure of central executive functioning was based on the
idea that test performance reflects the total amount of attentional
resources that can be allocated to information storage and manip-
ulation (Just & Carpenter, 1992). Importantly, characteristics of
that test could not be perceived as being masculine in any way. On
each trial of the test, participants were asked to read a set of
sentences and to decide whether each sentence was grammatically
correct, while simultaneously maintaining the last word of each
sentence in their memory (one target word for each sentence). Set
sizes ranged from two to six sentences, after which target words
had to be recalled. Ten minutes were provided to complete the test.
The test was long enough (74 items) to ensure that it would not
likely be completed within that period of time.
Dependent variables. Performance was evaluated by the total
number of words correctly recalled across all trials. Because cog-
nitive resources devoted to both memory and grammatical deci-
sions might be somehow interdependent, memory performance
would also be evaluated with performance on the secondary task
being included as a covariate.
771
DANGERS OF BENEVOLENT SEXISM

Results
The participants’ performance on the total number of words
correctly recalled was analyzed by way of a one-way ANOVA
with a three-level between-subjects variable, sexism (benevolent
sexism with help vs. benevolent sexism without help vs. no sex-
ism). As can be seen in Table 3, the sexism expressed by the
recruiter had a significant effect on recall, F(2, 34) ⫽ 7.30, p ⬍
.002, ␩
2
⫽ .31. Least significant difference contrasts confirmed
that when confronting benevolent sexism with help ( p ⬍ .009) or
benevolent sexism without help ( p ⬍ .001), the participants per-
formed worse than when the recruiter manifested no sexism. The
two benevolent sexism conditions did not differ from one another,
p ⬎ .36.
When the performance on the secondary task was included as a
covariate, the type of sexism expressed by the recruiter still had a
significant effect on recall, F(2, 31) ⫽ 5.47, p ⬍ .009, ␩
2
⫽ .26.
Contrasts confirmed that when confronting benevolent sexism
with help or benevolent sexism without help, the participants
performed worse than when the recruiter demonstrated no sexism,
ps ⬍ .02. The two benevolent sexism conditions did not differ
from one another, p ⬎ .12.
Discussion
The main results of Experiments 1 and 2 were replicated with a
manipulation of benevolent sexism that did not contain any refer-
ences to unsolicited help or any mention of a preferential selection
of women over men. These results strongly demonstrated that
benevolent sexism by itself was a sufficient condition for under-
mining women’s performance. Experiments 1, 2, and 3 demon-
strated this effect on two different tasks, both involving working
memory capacity. Importantly, although the task used in Experi-
ments 1 and 2 could have been seen as discrepant with the job or
perceived as stereotype-relevant, these criticisms could not hold
with the RST of Experiment 3.
In addition, the RST used in Experiment 3 helped to further
illuminate the effects benevolent sexism had on performance. RST
was originally developed to assess “people’s ability to perform
active processing of a stimulus while simultaneously buffering
other information in working memory” (Whitney, Arnett, Driver,
& Budd, 2001, p. 2). Scores on the RST also predicted perfor-
mance on tasks that require inhibition of irrelevant thoughts and
information (Rosen & Engle, 1998). In our view, impaired scores
on the RST after confronting benevolent sexism suggested that
women suffered many interfering and irrelevant thoughts.
Experiments 1, 2, and 3 repeatedly demonstrated that benevo-
lent sexism had a negative effect on performance compared with
both hostile and nonsexist conditions. Moreover, we provided
evidence that the effect of benevolence was not caused by the
motivation to perform on the test, the motivation to get the job, or
the perception of sexism. Importantly, Experiment 3 also demon-
strated that women’s impaired performance was not due to the fact
that benevolent sexism included unsolicited help. Indeed, even
when benevolent sexism was manipulated through its complemen-
tary gender differentiation component, that is, even when abso-
lutely no helping attitudes were used, benevolence still under-
mined women’s performance. Finally, the same effect was found
even when no preferential selection policies were mentioned. It
was thus unlikely that women were suspicious regarding the sin-
cerity of the apparently positive attitude displayed because the
context did not provide any cue allowing them to do so.
Interestingly, when benevolent sexism was expressed to women,
they found it difficult to identify it as sexism while simultaneously
considering the situation unpleasant. This might indicate that they
faced much more ambiguity and doubt about how to identify the
situation, compared with when hostile sexism was expressed.
Indeed, when identifying (hostile) sexism in the recruiter’s dis-
course, participants knew what to think about his opinions, so his
remarks could more easily be discounted or externally attributed
(Crocker & Major, 1989). Certainly, when confronted with benev-
olent attitudes or with males’ willingness to help them, women
may have had trouble deciding whether they should attribute such
interactions to sexism. Importantly, doubts about how to appraise
benevolent sexist remarks would be raised only if such discourse
implicitly suggested that women somehow lack ability, together
with the praising aspects (helping or magnifying women) it in-
volved. Benevolent sexism promotes the idea that women lack
competence, and internal rather than external attributions for such
ideas are more likely. Such a subtle form of sexism might then lead
women to doubt their abilities to perform well, to experience a
decrease in their self-esteem and self-confidence, as well as to be
preoccupied during the task. Interfering thoughts such as these,
through taxing cognitive resources required for working memory
tasks, could then definitely have impaired performance. In con-
trast, hostile sexism likely generated much less intrusive thought
because hostility was explicitly manifested. The unpleasantness of
the situation can thus more easily be attributed to the recruiter’s
attitude (external attribution), leaving sufficient cognitive re-
sources to devote to the task at hand. Experiment 4 examined that
hypothesis.
Experiment 4
Experiment 4 was designed to examine the role of interfering
thoughts in explaining benevolent sexism’s deleterious effect on
women’s performance. On the specific tests we used, high perfor-
mance required working memory. Using the same procedure as in
Experiments 1 and 2, we included in this experiment measures of
preoccupation with the task, self-doubt, and performance self-esteem
(Heatherton & Polivy, 1991). First, we hypothesized that benevolent
sexism, compared with hostile sexism, would lead to the deterioration
of women’s performance. Second, we proposed that benevolent sex-
ism would lead to greater mental intrusions and that this would be
revealed through greater preoccupation with the task, greater doubts,
Table 3
Performance as a Function of Sexism in Experiment 3
Performance
Type of sexism
Benevolent
with help
Benevolent
without help None
Recall 27.40
a
22.30
a
38.60
b
Grammar/semantic 27.00
a
33.70
a
41.50
b
Note. Means with different subscripts differ significantly at p ⬍ .01.
772
DARDENNE, DUMONT, AND BOLLIER

and impaired performance self-esteem. Third, we hypothesized that
the level of mental intrusions elicited by benevolent sexism would
mediate its impact on women’s performance. Additionally, because
sexism tackles women as a group, gender identification may well
moderate the effects sexism had on performance and mental intru-
sions. Experiment 4 thus included measures of gender identification to
see whether it would moderate the impact of sexism on women’s
performance and mental intrusions.
Method
Participants
Forty-four low-educated women (aged 25 to 42 years old; 6 to
9 years of education) from the same government school as in
Experiment 1 volunteered to participate in a study on job selection.
They were randomly assigned to the expression of one of two
types of sexism (hostile vs. benevolent) communicated through
written instructions. As in Experiments 1–3, the job for which they
supposedly applied was described as requiring typically feminine
qualities.
Materials and Procedure
At the beginning of the experimental session, participants com-
pleted three items assessing the identification with their gender
group on a 9-point scale, with endpoints of 1 (do not agree) and 9
(agree totally): “I have many characteristics in common with other
women,” “I identify with the group of women,” and “Being a
woman is a very important reflection of who I am” (␣⫽.98).
Next, the cover story, material, and procedure were identical to
Experiments 1 and 2. In addition, preoccupation with the task,
self-doubts, and performance self-esteem were used as indicators
of mental intrusions. After performing the task, participants were
fully debriefed, with a special focus on the reasons for possibly
performing badly, and thanked for their participation.
Dependent variables. Performance and the number of items
attempted were examined, as in Experiments 1 and 2. Mental intru-
sions, as a potential mediating variable, were also measured. Specif-
ically, seven items assessed the level of preoccupation experienced
during the task (e.g., “During the task, I thought that my performance
would be poor,” “During the task, I felt my heart beating more
strongly,” “During the task, I wasn’t sure about the rules associated
with wind and current”; ␣⫽.97). The level of self-doubt experienced
after performing the task was measured with a series of 13 items (e.g.,
“Right now, I feel full of doubt,” “Right now, I feel frustrated by my
performance,” “Right now, I’m worried”; ␣⫽.98). Then the partic-
ipants completed seven items from Heatherton and Polivy’s (1991)
Performance Self-Esteem subscale (e.g., “Right now, I am having
trouble understanding what I’m reading,” “Right now, I am frustrated
or worried by my performance,” “Right now, I am confident in my
capacity”; ␣⫽.92). All items ranged from 1 (not at all)to9(totally).
Responses were coded such that agreement with items reflected
greater preoccupation, greater self-doubt, and lower performance self-
esteem.
1
Results
The data were analyzed using multiple regression, with sexism
(contrast-coded ⫺1 if hostile and ⫹1 if benevolent), identification
to the ingroup (centered; M ⫽ 5.87, SD ⫽ 2.26), and the interac-
tion between them as predictors (Brauer, 2002; Cohen, 1983; Judd,
2000).
Test Said to Be the Basis for Hiring Decisions
Performance. Replicating Experiments 1–3, sexism had a sig-
nificant effect on performance, ␤⫽⫺.64, p ⬍ .001. When
confronting benevolent sexism, the participants performed worse
(M ⫽ 2.35) than when facing hostile sexism (M ⫽ 4.62). Gender
identification was also related to the women’s performance, ␤⫽
.35, p ⬍ .005, such that highly identified women performed better
than low-identified women. These effects were qualified by a
significant interaction between sexism and identification, ␤⫽
⫺.28, p ⬍ .05. As can be seen in Figure 2, identification had no
effect on performance when women confronted a benevolent sexist
attitude, ␤⫽.16, p ⬎ .45. When facing hostile sexism, however,
women with stronger identification with the gender group per-
formed better, ␤⫽.58, p ⬍ .01.
Number of items attempted to be solved. The only significant
effect was that hostile sexism led participants to attempt to solve
more items than did benevolent sexism, ␤⫽⫺.46, p ⬍ .005.
Globally, hostile sexism led participants to give answers to a
greater number of items (M ⫽ 7.38) than did benevolent sexism
(M ⫽ 5.26). Neither the identification (␤⫽.14, p ⬎ .30) nor the
interaction emerged as significant (␤⫽⫺.16, p ⬎ .25).
Mental Intrusions
We examined the relationships among the three measures of
mental intrusions (preoccupation, self-doubt, and self-esteem),
which revealed very strong correlations, rs (44) ⬎ .93, ps ⬍ .001.
We thus aggregated them into a global measure of mental intru-
sions. In fact, a factor analysis indicated that all 27 items loaded on
a single factor, explaining 87% of the variance (␣⫽.97). The
mental intrusions regression revealed a significant effect of sex-
ism, ␤⫽.71, p ⬍ .001. As hypothesized, the participants reported
more mental intrusions when they had been confronted by benev-
olent (M ⫽ 6.63) rather than hostile sexism (M ⫽ 3.06). As can be
seen in Figure 3, the effect of identification was also significant, ␤
⫽⫺.29, p ⬍ .05, showing globally fewer mental intrusions for
highly identified women. The interaction was, however, not sig-
nificant, ␤⫽.13, p ⬎ .20.
2
Mediation Analysis
Our mediation hypothesis concerning the effects of hostile and
benevolent sexism on performance and number of items attempted
was tested in a regression analysis including both sexism (coded as
⫺1 if hostile and ⫹1 if benevolent) and mental intrusions as
1
The items are available from Benoit Dardenne.
2
Although the Performance Self-Esteem scale was referring to interfering
thoughts such as “I have the feeling that I’m not as smart as the others,” it
might be considered as conceptually distant from other intrusive thoughts.
Therefore, analyses were also conducted on the Mental Intrusions scale with-
out the specific subscale of Performance Self-Esteem and on performance
self-esteem separately. Analyses revealed exactly the same for both subscales
separately as for the scale when it was composed of the 27 items.
773
DANGERS OF BENEVOLENT SEXISM

predictors. According to Baron and Kenny (1986), three relation-
ships must be demonstrated in order to test mediation (see Figure
4), and we tested them all. First, we showed that our manipulation
of sexism predicted mental intrusions (␤⫽.68, p ⬍ .001), per-
formance (␤⫽⫺.61, p ⬍ .001), and number of items attempted (␤
⫽⫺.43, p ⬍ .005). Second, mental intrusions predicted perfor-
mance (␤⫽⫺.87, p ⬍ .001) as well as number of items attempted
(␤⫽⫺.61, p ⬍ .001). Third, as can be seen in Figure 4, the direct
effect of the recruiter’s sexism became negligible and even virtu-
ally null when the index of mental intrusions was entered into each
regression analysis, ␤⫽⫺.02, p ⬎ .80, for performance, and ␤⫽
⫺.06, p ⬎ .70, for number of items attempted. Sobel tests further
confirmed that the reductions in the direct effect of the recruiter’s
sexism on performance and number of items attempted were
significant (z ⫽⫺4.94 and ⫺2.95; p ⬍ .001 and .005, respec-
tively). Thus, the impact of our sexism manipulation on perfor-
mance and number of items attempted seemed fully mediated
through the influence it had on mental intrusions. More mental
intrusions (doubts, preoccupation with the task, and lowered self-
esteem) lowered performance and led participants to attempt to
solve fewer items. That is, through the preoccupations it activated,
benevolent sexism exerted its deleterious impact on women’s
performance.
To rule out other explanations for these relationships, we also
conducted the reverse mediation analyses with performance or
number of items attempted as mediators and mental intrusions as
the dependent variable. In each regression, the sexism manipula-
tion remained a significant predictor of mental intrusions when we
controlled for performance or number of items attempted (both
␤s ⬎ .24), ts(43) ⬎ 2.82, ps ⬍ .01.
3
Discussion
Experiment 4 provided additional support for the hypothesis that
benevolent sexism had a negative effect on performance compared
with a hostile form of sexism. Most importantly, Experiment 4
examined the mechanism through which benevolent sexism af-
fected performance. Indeed, benevolent sexism led women to
experience many mental intrusions, such as preoccupation with the
task, self-doubt, and decreased self-esteem. Such mental intrusions
3
Similar to analyses on mental intrusions, we conducted the mediation
analysis with the Mental Intrusions scale without the specific subscale of
Performance Self-Esteem and with performance self-esteem separately.
The mediation analyses revealed exactly the same for both subscales
separately as for the scale when it was composed of the 27 items.
0246810
Identification
1
2
3
4
5
6
7
8
Performance
Hostile sexism
Benevolent sexism
Figure 2. Relationship between performance and level of identification
according to the type of sexism (Experiment 4). The vertical bar represents
the mean level of identification.
0246810
Identification
1
2
3
4
5
6
7
8
9
Mental Intrusions
Benevolent sexism
Hostile sexism
Figure 3. Relationship between mental intrusions and level of identifi-
cation according to the type of sexism (Experiment 4). The vertical bar
represents the mean level of identification.
Mental intrusions
)**68.-( **78.- **86.
Hostile (-1) / Performance
Benevolent (1) sexism -.61** (-.02)
Mental intrusions
)**75.-( **16.- **86.
Hostile (-1) / Number of items
Benevolent (1) sexism -.43** (-.06) attempte
d
Figure 4. Mental intrusions as a mediator of performance (top) and
number of items attempted (bottom) in response to the type of sexism
(Experiment 4). Coefficients in parentheses represent parameter estimates
for a regression model containing both predictors.
**
p ⬍ .005.
774
DARDENNE, DUMONT, AND BOLLIER

in turn impaired their capacity to concentrate on the task and
weakened their working memory, slowing down their performance
on the task and damaging global performance (see also, e.g.,
Beilock & Carr, 2005; Beilock et al., 2004). In contrast, when
hostile sexism was expressed, women faced many fewer mental
intrusions and performance was preserved.
Although it was plausible that performing poorly led to preoc-
cupation, self-doubt, and decreased self-esteem, analyses revealed
that causality was not in that direction. It was through its impact on
mental intrusions that sexism influenced the women’s perfor-
mance. In fact, whatever the status of mental intrusions in the
analysis (mediator or outcome variable), sexism always signifi-
cantly predicted them. On the contrary, performance was no longer
predicted by type of sexism when it was considered as an outcome
variable and mental intrusions were entered as the mediating
variable.
It might be surprising that preoccupation, self-doubt, and per-
formance self-esteem items all loaded on a single factor. In addi-
tion, some might argue that mental intrusions such as these were
not “pure” and also included anxiety or worries. Indeed, doubts are
worries, and doubts include anxiety. Similarly, performance self-
esteem precisely taps some anxiety and worries about one’s per-
formance, competence, or abilities. Such doubts and preoccupa-
tions, being specific to performance or more diffuse, all reflected
thoughts that might have interfered with women’s working mem-
ory capacity and task performance.
Also, we showed that stronger gender group identification led
women to be much less prone to interfering thoughts than did
lower identification. The interaction between identification and
sexism did not emerge as statistically significant. Nevertheless, the
pattern of data displayed in Figure 3 was neatly consistent with the
proposal that benevolent sexism led to mental intrusions even for
highly identified women. That is, when benevolent sexism was
expressed, stronger identification did not reduce mental intrusions
enough (␤⫽⫺.28, p ⬎ .20), and performance decreased. When
sexism was expressed in a hostile manner, however, highly iden-
tified women seemed better able than low-identified women to
reduce mental intrusions (␤⫽⫺.47, p ⬍ .05) and to perform well.
It thus seemed that benevolent sexism impacted enough for both
low-identified and highly identified women to reach the threshold
of mental intrusions that was necessary to damage performance,
whereas hostile sexism only affected low-identified women suffi-
ciently to undermine a performance.
This is perfectly consistent if we consider that stronger identi-
fication helped to diminish negative ideas about the gender group.
However, the level of mental intrusions was low enough not to
disturb performance only for those women who strongly identified
with their gender group in the case where they faced hostile
sexism. Identification has often been shown to moderate the im-
pact of blatant discrimination, such that it led to attributing nega-
tive feedback to discrimination (e.g., Crocker, Voelkl, Testa, &
Major, 1991; Operario & Fiske, 2001) or contributing to self-
esteem protection (e.g., Major, Quinton, & Schmader, 2003; Mc-
Coy & Major, 2003). The positive effect of identification, how-
ever, might not last long. Correlational studies have shown that
chronic exposure to discrimination is indeed related to lower
well-being (see, e.g., Branscombe, Schmitt, & Harvey, 1999;
Schmitt, Branscombe, Kobrynowicz, & Owen, 2002). In the short
term, though, our results supported the idea that hostile sexism was
less damaging than benevolent sexism, which insidiously created a
basis for gender inequality.
General Discussion
In a series of four experiments, we demonstrated consistent
evidence that benevolent, more than hostile, sexism had a delete-
rious impact on women’s performance. Moreover, this was true
despite the fact that the context valued feminine qualities. Exper-
iments 1, 2, and 4 used a task that could be seen as rather
stereotype-relevant and discrepant from the qualities required for
the job offered, but this was not the case in Experiment 3. In all
four experiments, it was only when benevolent sexism was ex-
pressed that performance decreased. That is, we proposed that
women doubted their abilities to perform the task (regardless of
whether the task was discrepant with the job offered) only when
benevolent sexism was expressed because it suggested women’s
inferiority, and this idea could not easily be externally attributed.
Even though benevolent sexism was more positive in tone, appar-
ently benign, and not identified as prejudice, it was revealed to be
a powerful weapon in destroying women’s performance within
feminine domains.
Although benevolent sexism was not readily identified as sex-
ism (Experiment 1), both hostile and benevolent sexism were
experienced as somewhat unpleasant situations (Experiment 2),
and benevolent more than hostile sexism generated disturbing
mental intrusions (Experiment 4). Experiments 1 and 2 discarded
plausible alternative interpretations of our results in terms of
motivation to perform, and Experiment 3 ruled out the idea of a
simple impact of preferential selection or offer of unsolicited help
on the performance of women. In fact, benevolence expressed
through complementary gender differentiation alone (i.e., no men-
tion of unsolicited and unwanted help and of preferential selection)
was sufficient to decrease the performance of women. Sexism was
role-played in Experiment 2 but was conveyed through written
instructions in Experiments 1, 3, and 4. Female university students
participated in Experiments 2 and 3, whereas unemployed low-
educated women participated in both Experiments 1 and 4. Finally,
we used different tasks in Experiment 3 from those that were used
in Experiments 1, 2, and 4. Our results were thus replicated
through different paradigms, different instructions, and different
populations.
Taken together, our four experiments suggested that benevolent
sexism involved communication of detrimental thoughts. Under
the cover of inoffensive introductory remarks, benevolent sexism
created a mindset of preoccupation, self-doubt (including some
anxiety), and decreased self-esteem. Such mental intrusions then
interfered with the task to be performed and undermined perfor-
mance, despite the fact that the context valued feminine skills. On
the contrary, hostile sexism sounded and actually was more ag-
gressive. It was detected as prejudice and therefore left no ambi-
guity, thereby facilitating external attribution or blame; hostile
sexism did not impair the women’s performance.
Extending Vescio et al.’s (2005) work, we distinguished be-
tween the hostile and benevolent components of sexism that were
involved in patronizing behavior and showed their respective
impacts on both motivation to perform and interfering thoughts,
two processes that could have been thought to influence women’s
performance. In addition, we clearly identified mental intrusions as
775
DANGERS OF BENEVOLENT SEXISM

the process through which performance was damaged, and we shed
light on how gender identification partially protected against sexist
attitudes.
Clearly, benevolent statements conveyed some negative ideas
about women, who were suggested to be cooperative and warm but
in need of men’s help, and thus relatively incompetent. On the
other hand, hostile sexism suggested that women are uncoopera-
tive and cold, but independent, mean, and assertive. As argued by
Fiske and her colleagues (e.g., Fiske, Cuddy, Glick, & Xu, 2002),
many stereotypes are mixed in their evaluative content, but per-
ceptions of warmth and competence of social groups are often
inversely related to one another (Judd, James-Hawkins, Yzerbyt, &
Kashima, 2005). That is, focusing on a group’s warmth implicitly
communicated low competence inferences about that group. Once
activated, stereotypes may influence people’s automatic behavior
and nonconscious goal pursuit (see Dijksterhuis & Bargh, 2001;
Wheeler & Petty, 2001). By focusing on women’s warmth, benev-
olent sexism could thus subtly activate the idea of women’s low
competence and lead them to behave consistently. Importantly,
implicit activation of women’s alleged incompetence, in conjunc-
tion with the difficulty of making external attributions for the
statements, likely led benevolent sexism to be perceived by women
as unpleasant and to evoke intrusive thoughts and anxiety or
self-doubt about their competence.
In our experiments, we examined both performance and the
number of items attempted. In our view, performance was perhaps
the most relevant index of women’s actual competence and surely
the one having the greatest consequences in real life. In most
evaluative situations, it is the total number of correct answers that
counts. Employers usually look at the absolute score one receives
on selection tasks to decide whether to select a candidate. In all
four experiments, performance was impaired by benevolent more
than by hostile sexism. Even if taken into account, the number of
items attempted by participants had fewer implications for real-life
situations than did global performance. This measure nevertheless
revealed other aspects that could be affected when confronting
sexism. In Experiments 1 and 2, many trials were attempted, and
ceiling effects did not allow us to observe the impact of sexism on
that measure. However, Experiment 4 revealed that fewer trials
were attempted when women faced benevolent rather than hostile
sexism. This suggests that mental intrusions might (but not nec-
essarily) have affected both performance and the number of items
attempted by interfering with concentration and slowing down
performance on the required task. Indeed, the role of mental
intrusions on both these effects had been confirmed by our medi-
ation analyses.
Recent research on stigmatization has shown that the negative
effect of stigma on performance could be due to anxiety and
decrements in working memory (e.g., Schmader & Johns, 2003).
We suggest that benevolent sexism was a way of stigmatizing
women that interfered with performance by reducing working
memory capacity in contexts in which we would not expect
women to feel incompetent. By triggering numerous mental intru-
sions, such as self-doubt or preoccupation including anxiety and
reduced self-esteem, benevolent sexism lowered performance and
impaired cognitive capacity that would normally be devoted to the
task at hand. The content it activates and the difficulty of making
external attributions for the negative ideas suggested about women
make benevolent sexism a dangerous tool that might be used by
men to maintain power over women.
Osborne (2001); Spencer, Steele, and Quinn (1999); and Bos-
son, Haymovitz, and Pinel (2004), among a few others, have found
initial support for the idea that anxiety contributes to stereotype
threat (i.e., fear of confirming the negative stereotype associated
with the ingroup, resulting in stereotype confirmation). Close to
the concept of anxiety, arousal has also been found to play an
important role in the deleterious effects of stereotype threat (e.g.,
Ben-Zeev, Fein, & Inzlicht, 2005; Blascovich, Spencer, Quinn, &
Steele, 2001; O’Brien & Crandall, 2003). Along the same line,
Schmader and Johns (2003) found that stereotype threat reduced
women’s performance on a mathematical test by diminishing their
working memory capacity. Finally, Inzlicht, McKay, and Aronson
(2006) recently showed that stigma was ego depleting and that
trying to cope with it weakened the ability to control one’s behav-
ior in unrelated domains. When we paralleled our results with this
literature, it appeared that benevolent sexism acted as a stereotype
threat or a stigma within feminine domains and indeed had equiv-
alent damaging consequences on women’s performance. Also,
benevolent sexism induced some anxiety, mixed with doubts about
performance abilities. Although benevolent sexism might remind
us of stereotype threat, we doubt that hostile sexism matched the
classic control or nonthreatening conditions used in that literature.
Indeed, in contrast to research on stereotype threat showing that
highly identified individuals are more at risk of lowered perfor-
mance (e.g., Schmader, 2002), our results indicated that, if any-
thing, strongly identifying with the gender ingroup helped to
preserve performance in the face of hostile sexism.
Several theories stress the role of benevolent stereotypes and
“nice” ideologies in leading members of disadvantaged groups to
justify the status quo. Our experiments went further and suggested
that disadvantaged group members indeed behaviorally support the
basis of existing gender inequality and system justifications. All
these theories imply that, on some occasions, women are “accom-
plices” in their own subordination. Glick and Fiske (1996, 2001;
Glick et al., 2000) have constantly argued that benevolent sexism
and the “women are wonderful” effect documented by Eagly and
Mladinic (1989, 1993) serve to increase support for the system of
gender inequality. According to system justification theory (e.g.,
Jost & Banaji, 1994), ambivalent sexism theory (e.g., Glick &
Fiske, 2001), and social dominance theory (e.g., Sidanius & Pratto,
1999), stereotypes are legitimizing tools that help high-status
groups keep low-status groups in the state of subordination. When
stereotypes are behaviorally confirmed as in our experiments
(women performed poorly), reality rather than stereotypes surely
helps the legitimizing of status differentials. Following Jackman
(1994) and Cose (1993), dominant groups prefer to act benevo-
lently towards low-status groups, because the latter will more
easily agree on their inferior status (Haines & Jost, 2000; Rudman
& Heppen, 2003) and even support this hierarchical social system
(Jost & Kay, 2005). Considering the results of our experiments in
this regard, we confirm the pertinence and appropriateness of
dominant groups’ preferences.
Our results are also perfectly consistent with a study by Jost and
Kay (2005) recently demonstrating that women, but not men, are
more prone to support the general and diffuse system of inequality
after being exposed to benevolent stereotypes than in a comparably
favorable but nonstereotypical control condition. Hostile stereo-
776
DARDENNE, DUMONT, AND BOLLIER

types, on the contrary, did not lead women to increase system
justification.
Similarly, our results nicely fit Jackman’s (1994) velvet glove
theory as well as Benokraitis’s (1997) propositions about the
effectiveness of “sweet” discrimination in maintaining gender sta-
tus inequities. According to Karl Marx himself, it was soon rec-
ognized that legitimacy of dominance needed a state of “false
consciousness” among subordinates. Along with Jost and Banaji’s
(1994) theory of system justification, Glick and Fiske’s (2001)
theory of benevolent and hostile sexism, and Sidanius and Pratto’s
(1999) theory of social dominance, Jackman’s (1994, 2001) velvet
glove theory holds that subordinates are often accomplices in their
own subordination. Although the research on the psychology of
legitimacy has provided numerous empirical investigations of this
idea (see, e.g., Jost & Hunyady, 2002; Jost & Major, 2001), very
few studies have provided direct experimental evidence of the
“efficiency” of paternalistic and benevolent attitudes in maintain-
ing inequalities compared with more overt and hostile ones. By
providing evidence of their own incompetence, albeit fully con-
textual, women who confronted benevolent sexism in our experi-
ments offered legitimizing arguments to men who might be willing
to use them in justifying inequalities of the societal system.
In fact, although we do not wish to suggest that hostile sexism
would have no negative effects in the long term, in the short term,
exposure to explicit and hostile sexism was not that damaging. As
we showed, hostile sexism preserved performance. Moreover, as
we demonstrated, it was through mental intrusions that benevolent
sexism was detrimental to women. Informing them of the dangers
of accepting benevolent sexism may reduce uncertainties and help
preserve performance. This would undoubtedly facilitate women’s
countering the effects that dominant groups might be seeking when
using such an untrustworthy strategy.
To conclude, we provided strong and consistent evidence that
benevolent sexism is a very efficient tool that can push women into
providing the best justification for gender inequality. By “objec-
tively” performing poorly, women facing benevolent sexism could
indeed hardly contest their own (group) subordination. As outra-
geous as it is, hostile or blatant sexism may well paradoxically
eventually serve women’s cause. Further research is certainly
needed to ascertain the whole range of influences sexism has on
women’s behavior and their implications for intergroup relation-
ships. Although some influences may potentially be positive, we
do not recommend men to be overtly sexist, because this may be
damaging for women’s self-esteem in the long run. Rather, we
recommend that women and men not consider benevolent sexism
to be benign.
References
Abrams, D., Viki, G. T., Masser, B., & Bohner, G. (2003). Perceptions of
stranger and acquaintance rape: The role of benevolent and hostile
sexism in victim blame and rape proclivity. Journal of Personality and
Social Psychology, 84, 111–125.
Bargh, J. A. (1997). The automaticity of everyday life. In R. S. Wyer, Jr.
(Ed.), Advances in social cognition (Vol. 10, pp. 1– 61). Mahwah, NJ:
Erlbaum.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator distinction in
social psychological research: Conceptual, strategic, and statistical consid-
erations. Journal of Personality and Social Psychology, 51, 1173–1182.
Barreto, M., & Ellemers, N. (2005). The burden of benevolent sexism:
How it contributes to the maintenance of gender inequalities. European
Journal of Social Psychology, 35, 633–642.
Beilock, S. L., & Carr, T. H. (2005). When high-powered people fail:
Working memory and “choking under pressure” in math. Psychological
Science, 16, 101–105.
Beilock, S. L., Kulp, C. A., Holt, L. E., & Carr, T. H. (2004). More on the
fragility of performance: Choking under pressure in mathematical problem
solving. Journal of Experimental Psychology: General, 133, 584 – 600.
Benokraitis, N. V. (1997). Subtle sexism. Current practices and prospects
for change. Thousand Oaks, CA: Sage.
Ben-Zeev, T., Fein, S., & Inzlicht, M. (2005). Arousal and stereotype
threat. Journal of Experimental Social Psychology, 41, 174–181.
Blascovich, J., Spencer, S. J., Quinn, D., & Steele, C. (2001). African
Americans and high blood pressure: The role of stereotype threat.
Psychological Science, 12, 225–229.
Bosson, J. K., Haymovitz, E. L., & Pinel, E. C. (2004). When saying and
doing diverge: The effects of stereotype threat on self-reported versus
non-verbal anxiety. Journal of Experimental Social Psychology, 40,
247–255.
Branscombe, N. R., Schmitt, M. T., & Harvey, R. D. (1999). Perceiving
pervasive discrimination among African-Americans: Implications for
group identification and well-being. Journal of Personality and Social
Psychology, 77, 135–149.
Brauer, M. (2002). The analysis of continuous and categorical independent
variables: Alternatives to dichotomization. L’Anne´e Psychologique, 102,
449 – 484.
Brehm, J. (1966). A theory of psychological reactance. New York: Aca-
demic Press.
Brehm, J., & Weintraub, M. (1977). Physical barriers and psychological
reactance: 2-year-olds’ response to threats to freedom. Journal of Per-
sonality and Social Psychology, 35, 830–836.
Chen, M., & Bargh, J. A. (1997). Nonconscious behavioral confirmation
processes: The self-fulfilling nature of automatically activated stereo-
types. Journal of Experimental Social Psychology, 33, 541–560.
Cohen, J. (1983). The cost of dichotomization. Applied Psychological
Measurement, 7, 249–253.
Cose, E. (1993). Rage as a privileged class. New York: HarperCollins.
Crocker, J., & Major, B. (1989). Social stigma and self-esteem: The
self-protective properties of stigma. Psychological Review, 96, 608 –
630.
Crocker, J., Voelkl, K., Testa, M., & Major, B. (1991). Social stigma: The
affective consequences of attributional ambiguity. Journal of Personal-
ity and Social Psychology, 60, 218–228.
Daneman, M., & Carpenter, P. A. (1980). Individual differences in working
memory and reading. Journal of Verbal Learning and Verbal Behavior,
19, 450–466.
Daneman, M., & Merikle, P. M. (1996). Working memory and language
comprehension: A meta-analysis. Psychonomic Bulletin and Review, 3,
422– 433.
Dardenne, B., Delacollette, N., Gre´goire, C., & Lecocq, D. (2006). Latent
structure of the French validation of the Ambivalent Sexism Inventory:
L’Echelle de Sexisme Ambivalent. L’Anne´e Psychologique, 106, 235–264.
Dijksterhuis, A., Aarts, H., Bargh, J. A., & van Knippenberg, A. (2000).
Past contact, stereotype strength, and automatic behavior. Journal of
Experimental Social Psychology, 36, 531–544.
Dijksterhuis, A., & Bargh, J. A. (2001). The perception-behavior express-
way: Automatic effects of social perception on social behavior. In M. P.
Zanna (Ed.), Advances in Experimental Social Psychology (Vol. 33, pp.
1– 40). San Diego, CA: Academic Press.
Dijksterhuis, A., & van Knippenberg, A. (1998). The relation between
perception and behavior, or how to win a game of trivial pursuit. Journal
of Personality and Social Psychology, 74, 865–877.
Eagly, A. H., & Mladinic, A. (1989). Gender stereotypes and attitudes
777
DANGERS OF BENEVOLENT SEXISM

toward women and men. Personality and Social Psychology Bulletin, 15,
543–558.
Eagly, A. H., & Mladinic, A. (1993). Are people prejudiced against
women? Some answers from research an attitudes, gender stereotypes,
and judgment of competence. In W. Stroebe & M. Hewstone (Eds.),
European Review of Social Psychology (Vol. 5, pp. 1–35). New York:
Wiley.
Engle, R. W. (2001). What is working memory capacity? In H. L. Roediger
III & J. S. Nairne (Eds.), The nature of remembering: Essays in honor
of Robert G. Crowder (pp. 297–314). Washington, DC: American Psy-
chological Association.
Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. A. (1999).
Working memory, short-term memory, and general fluid intelligence: A
latent variable approach. Journal of Experimental Psychology: General,
128, 309–331.
Fiske, S. T., Cuddy, A. J. C., Glick, P., & Xu, J. (2002). A model of (often
mixed) stereotype content: Competence and warmth respectively follow
from perceived status and competition. Journal of Personality and
Social Psychology, 82, 878–902.
Ford, T. E., Wentzel, E. R., & Lorion, J. (2001). Effects of exposure to
sexist humor on perceptions of normative tolerance of sexism. European
Journal of Social Psychology, 31, 677–691.
Frijda, N. H., Kuipers, P., & ter Schure, E. (1989). Relations among
emotion, appraisal, and emotional action readiness. Journal of Person-
ality and Social Psychology, 57, 212–228.
Glick, P., & Fiske, S. T. (1996). The Ambivalent Sexism Inventory:
Differentiating hostile and benevolent sexism. Journal of Personality
and Social Psychology, 70, 491–512.
Glick, P., & Fiske, S. T. (2001). An ambivalent alliance: Hostile and
benevolent sexism as complementary justifications for gender inequal-
ity. American Psychologist, 56, 109–118.
Glick, P., Fiske, S. T., Mladinic, A., Saiz, J., Abrams, D., Masser, B., et al.
(2000). Beyond prejudice as simple antipathy: Hostile and benevolent
sexism across cultures. Journal of Personality and Social Psychology,
79, 763–775.
Haines, E. L., & Jost, J. T. (2000). Placating the powerless: Effects of
legitimate and illegitimate explanation on affect, memory, and stereo-
typing. Social Justice Research, 13, 219–236.
Harmon-Jones, E., Sigelman, J. D., Bohling, A., & Harmon-Jones, C. (2003).
Anger, coping, and frontal cortical activity: The effect of coping potential on
anger-induced left frontal activity. Cognition and Emotion, 17, 1–24.
Heatherton, T. F., & Polivy, J. (1991). Development and validation of a
scale measuring state self-esteem. Journal of Personality and Social
Psychology, 60, 895–910.
Heilman, M. E., & Alcott, V. B. (2001). What I think you think of me:
Women’s reactions to being viewed as beneficiaries of preferential
selection. Journal of Applied Psychology, 86, 574–582.
Inzlicht, M., McKay, L., & Aronson, J. (2006). Stigma as ego depletion.
Psychological Science, 17, 262–269.
Jackman, M. (1994). The velvet glove: Paternalism and conflict in gender,
class and race relations. Berkeley: University of California Press.
Jackman, M. (2001). License to kill: Violence and legitimacy in expro-
priative social relations. In J. T. Jost & B. Major (Eds.), The psychology
of legitimacy: Emerging perspectives on ideology, justice, and inter-
group relations. Cambridge, England: Cambridge University Press.
Jost, J. T., & Banaji, M. (1994). The role of stereotyping in system
justification and the production of false consciousness. British Journal of
Social Psychology, 22, 1–27.
Jost, J. T., & Hunyady, O. (2002). The psychology of system justification
and the palliative function of ideology. In W. Stroebe & M. Hewstone
(Eds.), European Review of Social Psychology (Vol. 13, pp. 111–153).
New York: Wiley.
Jost, J. T., & Kay, A. C. (2005). Exposure to benevolent sexism and
complementary gender stereotypes: Consequences for specific and dif-
fuse forms of system justification. Journal of Personality and Social
Psychology, 88, 498–509.
Jost, J. T., & Major, B. (2001). The psychology of legitimacy. Emerging
perspectives on ideology, justice, and intergroup relations. Cambridge,
England: Cambridge University Press.
Judd, C. M. (2000). Everyday data analysis in social psychology: Com-
parisons of linear models. In H. T. Reis & C. M. Judd (Eds.), Handbook
of research methods in social and personality psychology (pp. 370 –
392). Cambridge, England: Cambridge University Press.
Judd, C. M., James-Hawkins, L., Yzerbyt, V., & Kashima, Y. (2005).
Fundamental dimensions of social judgment: Understanding the rela-
tions between judgments of competence and warmth. Journal of Per-
sonality and Social Psychology, 89, 899–913.
Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehen-
sion: Individual differences in working memory. Psychological Review,
103, 761–772.
Kaiser, C. R., & Miller, C. T. (2001). Reacting to impending discrimina-
tion: Compensation for prejudice and attributions to discrimination.
Personality and Social Psychology Bulletin, 27, 1357–1367.
Major, B., McCoy, S. K., Kaiser, C. R., & Quinton, W. J. (2003). Prejudice
and self-esteem: A transactional model. In W. Stroebe & M. Hewstone
(Eds.), European Review of Social Psychology (Vol. 14, pp. 77–104).
New York: Wiley.
Major, B., Quinton, W. J., & Schmader, T. (2003). Attributions to discrim-
ination and self-esteem: Impact of group identification and situational
ambiguity. Journal of Experimental Social Psychology, 39, 220–231.
McCoy, S. K., & Major, B. (2003). Group identification moderates emo-
tional responses to perceived prejudice. Personality and Social Psychol-
ogy Bulletin, 29, 1005–1117.
Nadler, A., & Fisher, J. D. (1986). The role of threat to self-esteem and
perceived control in recipient reaction to help: Theory development
and empirical validation. In L. Berkowitz (Ed.), Advances in exper-
imental social psychology (Vol. 19, pp. 81–122). San Diego, CA:
Academic Press.
O’Brien, L. T., & Crandall, C. S. (2003). Stereotype threat and arousal:
Effects on women’s math performance. Personality and Social Psychol-
ogy Bulletin, 29, 782–789.
Operario, D., & Fiske, S. T. (2001). Ethnic identity moderates perceptions
of prejudice: Judgments of personal versus group discrimination and
subtle versus blatant bias. Personality and Social Psychology Bulletin,
27, 550–561.
Osborne, J. W. (2001). Testing stereotype threat: Does anxiety explain race
and sex differences in achievement? Contemporary Educational Psy-
chology, 26, 291–310.
Rosen, V. M., & Engle, R. W. (1998). Working memory capacity and
suppression. Journal of Memory and Language, 39, 418– 436.
Rudman, L. A., & Heppen, J. B. (2003). Implicit romantic fantasies and
women’s interest in personal power: A glass slipper effect? Personality
and Social Psychology Bulletin, 29, 1357–1370.
Sakalli, U., & Glick, P. (2003). Ambivalent sexism and attitudes toward
women who engage in premarital sex in Turkey. Journal of Sex Re-
search, 40, 296–302.
Scherer, K. R. (1988). Criteria for emotion-antecedent appraisal: A review.
In V. Hamilton, G. H. Bower, & N. H. Frijda (Eds.), Cognitive perspec-
tives on emotion and motivation (pp. 89 –126). Norwell, MA: Kluwer
Academic.
Schmader, T. (2002). Gender identification moderates stereotype threat
effects on women’s math performance. Journal of Experimental Social
Psychology, 38, 194–201.
Schmader, T., & Johns, M. (2003). Converging evidence that stereotype
threat reduces working memory. Journal of Personality and Social
Psychology, 84, 440–452.
778
DARDENNE, DUMONT, AND BOLLIER

Schmitt, M. T., Branscombe, N. R., Kobrynowicz, D., & Owen, S. (2002).
Perceiving discrimination against one’s gender group has different im-
plications for well-being in women and men. Personality and Social
Psychology Bulletin, 28, 197–210.
Schneider, M. E., Major, B., Luhtanen, R., & Crocker, J. (1996). When
help hurts: Social stigma and the cost of assumptive help. Personality
and Social Psychology Bulletin, 22, 201–209.
Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory
of social hierarchy and oppression. Cambridge, England: Cambridge
University Press.
Smith, C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal
in emotion. Journal of Personality and Social Psychology, 48, 813–
838.
Spencer, S., Steele, C. M., & Quinn, D. M. (1999). Stereotype threat and
women’s math performance. Journal of Experimental Social Psychol-
ogy, 35, 4–28.
Stangor, C., Swim, J. K., Sechrist, G. B., DeCoster, J., Van Allen, K. L.,
& Ottenbreit, A. (2003). Ask, answer, and announce: Three stages in
perceiving and responding to discrimination. In W. Stroebe & M. Hew-
stone (Eds.), European review of social psychology (Vol. 14, pp. 277–
311). New York: Wiley.
Turner, M. E., & Pratkanis, A. R. (1994). Affirmative action as help: A
review of recipient reactions to preferential selection and affirmative
action. Basic and Applied Social Psychology, 15, 43–69.
Vescio, T. K., Gervais, S. J., Snyder, M., & Hoover, A. (2005). Power
and the creation of patronizing environments: The stereotype-based behav-
iors of the powerful and their effects on female performance in masculine
domains. Journal of Personality and Social Psychology, 88, 658–672.
Viki, G. T., & Abrams, D. (2003). Infra-humanization: Ambivalent sexism
and the attribution of primary and secondary emotions to women.
Journal of Experimental Social Psychology, 39, 492–499.
Wheeler, S. C., & Petty, R. E. (2001). The effects of stereotype activation
on behavior: A review of possible mechanisms. Psychological Bulletin,
127, 797–826.
Whitney, P., Arnett, P. A., Driver, A., & Budd, D. (2001). Measuring
central executive functioning: What’s in a reading span? Brain and
Cognition, 45, 1–14.
Received September 27, 2005
Revision received March 14, 2007
Accepted March 22, 2007 䡲
New Editors Appointed, 2009 –2014
The Publications and Communications Board of the American Psychological Association an-
nounces the appointment of six new editors for 6-year terms beginning in 2009. As of January 1,
2008, manuscripts should be directed as follows:
● Journal of Applied Psychology (http://www.apa.org/journals/apl), Steve W. J. Kozlowski,
PhD, Department of Psychology, Michigan State University, East Lansing, MI 48824.
● Journal of Educational Psychology (http://www.apa.org/journals/edu), Arthur C. Graesser,
PhD, Department of Psychology, University of Memphis, 202 Psychology Building, Memphis,
TN 38152.
● Journal of Personality and Social Psychology: Interpersonal Relations and Group Processes
(http://www.apa.org/journals/psp), Jeffry A. Simpson, PhD, Department of Psychology,
University of Minnesota, 75 East River Road, N394 Elliott Hall, Minneapolis, MN 55455.
● Psychology of Addictive Behaviors (http://www.apa.org/journals/adb), Stephen A. Maisto,
PhD, Department of Psychology, Syracuse University, Syracuse, NY 13244.
● Behavioral Neuroscience (http://www.apa.org/journals/bne), Mark S. Blumberg, PhD, De-
partment of Psychology, University of Iowa, E11 Seashore Hall, Iowa City, IA 52242.
● Psychological Bulletin (http://www.apa.org/journals/bul), Stephen P. Hinshaw, PhD, Depart-
ment of Psychology, University of California, Tolman Hall #1650, Berkeley, CA 94720.
(Manuscripts will not be directed to Dr. Hinshaw until July 1, 2008, as Harris Cooper will
continue as editor until June 30, 2008.)
Electronic manuscript submission: As of January 1, 2008, manuscripts should be submitted
electronically via the journal’s Manuscript Submission Portal (see the website listed above with
each journal title).
Manuscript submission patterns make the precise date of completion of the 2008 volumes
uncertain. Current editors, Sheldon Zedeck, PhD, Karen R. Harris, EdD, John F. Dovidio, PhD,
Howard J. Shaffer, PhD, and John F. Disterhoft, PhD, will receive and consider manuscripts through
December 31, 2007. Harris Cooper, PhD, will continue to receive manuscripts until June 30, 2008.
Should 2008 volumes be completed before that date, manuscripts will be redirected to the new
editors for consideration in 2009 volumes.
779
DANGERS OF BENEVOLENT SEXISM