You are on page 1of 13

Research Article

Psychology of Women Quarterly


37(4) 507-518
She Loves Him, She Loves Him Not: ª The Author(s) 2013
Reprints and permission:
sagepub.com/journalsPermissions.nav
Attachment Style as a Predictor of DOI: 10.1177/0361684313497471
pwq.sagepub.com
Women’s Ambivalent Sexism Toward Men

Joshua Hart1, Peter Glick2, and Rachel E. Dinero3

Abstract
In two studies, we examined how romantic attachment style relates to women’s sexism toward men. Specifically, we applied
structural equation modeling and mediation analyses to the responses of over 500 self-reported heterosexual women. Study 1
included 229 women who answered questionnaires tapping attachment anxiety and avoidance, ambivalent sexism toward men,
romanticism, and interpersonal trust. We conducted Study 2 as a replication, changing questionnaire order to gauge the
robustness of results, using a new sample of 273 women. In general, women’s attachment anxiety predicted ambivalent sexism
(both benevolence and hostility) toward men, whereas women’s attachment avoidance predicted univalent hostility (and
lower benevolence) toward men. Romanticism mediated attachment style’s relationship to benevolence toward men, whereas
lower interpersonal trust mediated attachment’s relationship to hostility toward men. The results suggest that, for women (as
for men), sexist attitudes toward members of the other sex have roots in attachment style and associated worldviews. Better
understanding of women’s ambivalence toward men in romantic relationships may help to inform marital therapy.

Keywords
attachment behavior, sexism, sex role attitudes, romance, trust (social behavior), interpersonal attraction

Despite much research on sexism’s consequences, little is Women’s Sexism Toward Men
known about the personality factors that predispose indi-
Ambivalent sexism theory and research suggest that male
viduals to adopt sexist attitudes. This is especially true for
dominance combined with heterosexual interdependence
women’s sexism toward men, which receives much less
generates hostile and benevolent sexism between men and
attention than men’s sexism toward women, perhaps because
women (Glick & Fiske, 1996, 2001). Because members of the
powerful groups are not typically seen as suffering deleter-
other sex represent both potential romantic partners and out-
ious consequences from less powerful groups’ prejudices.
However, women’s sexism toward men plays a significant group members competing for social resources (e.g., status,
jobs), heterosexual men and women relate to one another
role in reinforcing gender stereotypes and perpetuating male
on both interpersonal and intergroup bases. Ambivalently
dominance (Glick et al., 2004), so researchers and others who
sexist men exhibit hostile sexism toward women viewed as
are committed to understanding and combating systems of
competitors (e.g., for jobs), but they direct their benevolent
gender inequality should be just as concerned with women’s
sexism toward women who fulfill intimacy goals (e.g., as
attitudes toward men as the other way around.
wives; see Glick & Fiske, 2001, for a review). Accordingly,
The present research focuses on personality predictors of
hostile sexism characterizes (some) women as manipulative,
women’s ambivalent sexism toward men by building on
the previous finding that men’s romantic attachment style untrustworthy adversaries, whereas benevolent sexism views
predicts sexism toward women (Hart, Hung, Glick, & Dinero,
2012). Might attachment style also help explain women’s
sexist attitudes toward men? Because attachment style
reflects general mental representations of close relationships 1
Department of Psychology, Union College, Schenectady, NY, USA
that, in turn, influence interpersonal and ideological orienta- 2
Department of Psychology, Lawrence University, Appleton, WI, USA
3
tions, we theorize that (as with men) heterosexual women’s Psychology Program, Cazenovia College, Cazenovia, NY, USA
attachment style predisposes sexist attitudes toward the other
Corresponding Author:
sex—with the two dimensions of attachment insecurity (anxi- Joshua Hart, Department of Psychology, Union College, 807 Union Street,
ety and avoidance) predicting distinct patterns of hostility and Schenectady, NY 12308, USA.
benevolence toward men (HM and BM). Email: hartj@union.edu
508 Psychology of Women Quarterly 37(4)

(some) women as special, indispensable, and morally pure the attachment system coalesces within the first year of life, it
beings whose companionship ‘‘completes’’ men. shapes, and is shaped by, experiences throughout life (Simp-
Similar to men’s sexism, women’s sexism toward the other son, Collins, Tran, & Haydon, 2007). Hence, by adulthood,
sex encompasses two overlapping facets: HM and BM (Glick people maintain elaborate, affectively tinged representations
& Fiske, 1999, 2001). HM expresses resentment of men’s or internal ‘‘working models’’ (Bowlby, 1982) of the self and
abuse of power, whereas BM expresses admiration for the others, based on prior experiences, especially in the context of
ways in which men’s power appears to benefit women within close relationships. These working models create expectations
traditional relationships, celebrating men as women’s ‘‘protec- about close relationships, but their influence extends well
tors and providers.’’ Like hostile and benevolent sexism beyond that domain to affect emotion and behavior regulation,
toward women, ambivalent beliefs about men maintain the sta- intergroup attitudes, sex, parenting, and psychopathology—to
tus quo in gender relations. For example, cross-cultural com- name just a few examples (see Mikulincer & Shaver, 2007,
parisons show that the greater the gender inequality in a for a comprehensive review of adult attachment theory and
society, the more the women endorse both HM and BM (Glick research).
et al., 2004). Additionally, whereas HM characterizes men as The attachment system’s internal working models,
arrogant, hypercompetitive, and callous, it also essentializes although unique to each individual, exhibit broad patterns
men as designed for dominance, viewing men’s power as an that reliably differentiate styles of attachment. Most notably,
unchangeable fact of life. Although HM does not correlate two insecurity dimensions distinguish how much people (a)
with gender inequality’s perceived legitimacy, it does predict yearn for intimacy, fear rejection and abandonment, and per-
gender inequality’s perceived stability (Glick & Whitehead, severate about relationships (attachment anxiety) and (b) fear
2010). Indeed, HM correlates positively with BM because both intimacy and pursue a strategy of self-reliance and emotional
presume that men have greater power. Sometimes this power is distancing from close relationship partners (attachment
used to harm or control women, as HM acknowledges, but it avoidance). People who are lower on both insecurity dimen-
can also be used to protect and provide for women, as BM sions are considered ‘‘secure.’’
celebrates. So, although HM is resentful toward men’s power, Attachment anxiety and avoidance reflect, in part, the
like hostile and benevolent sexism toward women, it correlates extent of individuals’ overall psychological security (i.e.,
positively with traditional values (Feather, 2004), negatively durability vs. insecurity or vulnerability; Hart, Shaver, &
with feminist beliefs (Anderson, Kanner, & Elsayegh, 2009), Goldenberg, 2005), which in turn shapes their views of them-
and positively with other sexist beliefs, including BM selves, their close relationship partners, and the world at
(Glick et al., 2004). Thus, despite important differences (dis- large. Indeed, people are likely to apply their attachment
cussed below), sexist attitudes toward women and men share working models to potential future partners (Brumbaugh &
complex (i.e., ambivalent) structures, cultural derivations, and Fraley, 2006) and to exhibit their attachment tendencies dur-
functions. ing interactions with strangers (e.g., support seeking or lack
But what accounts for individual differences, within thereof; Feeney, Cassidy, & Ramos-Marcuse, 2008). More-
cultures, in women’s endorsement of HM and BM? The evi- over, attachment style reflects general sensitivity to threats
dence described above suggests that HM and BM stem from to the self (i.e., insecurity vs. security) so that it influences
similar social–structural and ideological antecedents as intergroup attitudes, which frequently involve potential threat
hostile sexism and benevolent sexism, which underscores the and conflict. Specifically, priming secure attachment has
importance of understanding sexism toward men as part of a been found to reduce intergroup bias (Mikulincer & Shaver,
larger psychological and cultural context that supports gender 2001); in contrast, attachment anxiety relates to believing that
inequality. However, to our knowledge, no research has the world is a dangerous place (Weber & Federico, 2007) and
explored nonideological personality correlates of women’s attachment avoidance predicts social dominance orientation
HM and BM. Because gender relations involve intimate inter- (SDO), an adversarial intergroup ideology (Hart et al.,
personal as well as intergroup dimensions, we propose that 2012; Weber & Federico, 2007). In sum, theory and research
differences in women’s attachment styles predict distinct suggest that attachment style plays a prominent role in shap-
patterns of hostile and benevolent sexism toward men. ing individuals’ worldviews, including their interpersonal
and intergroup attitudes and ideologies.
Concerning men’s sexism toward women, Hart et al.
Attachment Insecurity and Sexism (2012) found that attachment avoidance predicts men’s uni-
Our research builds on Hart et al.’s (2012) theoretical and valent hostility toward women—that is, higher hostile sex-
empirical analysis of attachment styles and men’s sexism ism combined with lower benevolent sexism. This finding
toward women. They proposed that sexism derives in part from makes sense because attachment avoidance reflects working
the operations of the ‘‘attachment system,’’ the behavioral sys- models depicting others as undependable and intimacy as
tem thought to influence the maintenance of close interperso- threatening—a cynical disposition compatible with hostile
nal relationships throughout individuals’ lives (Bowlby, 1982; sexism’s characterization of women as manipulative and
Hazan & Shaver, 1994; Mikulincer & Shaver, 2003). Although untrustworthy but not with benevolent sexism’s idealization
Hart et al. 509

of romantic relationships. Accordingly, Hart et al. found First, due to the symmetries in heterosexual interdepen-
that the avoidance ! hostile sexism link was mediated by dence, we expected a similar pattern with respect to the
SDO, which is related to avoidant individuals’ tendency to fundamental relation between women’s attachment insecuri-
view the world as a cold, competitive jungle (Weber & ties and sexism toward men, as observed in Hart et al.’s
Federico, 2007). Further, the avoidance ! benevolent sex- (2012) research on men’s sexism. As with men, women’s
ism link was mediated by lower romanticism (i.e., cynical attachment anxiety should predict ambivalence toward the
beliefs about romantic love), reflecting avoidant individu- other sex—that is, endorsement of both HM (a cynical view
als’ discomfort with intimacy. of men as callous abusers of power) and BM (an idealized
Hart et al.’s (2012) research also revealed that men higher view of men as women’s protectors and providers). Further-
in attachment anxiety were prone to ambivalence toward more, women’s avoidance should predict univalent hostile
women, scoring higher on both benevolent sexism and hostile sexism toward men (higher HM and lower BM).
sexism than men lower in attachment anxiety. These findings Concerning mediators of the attachment ! sexism links,
are consistent with anxiously attached individuals’ simulta- we expected that women’s (like men’s) benevolence toward
neous craving for intimacy (often leading to overinvestment the other sex should be mediated by romanticism, an ideology
in relationships) and concern that others will not provide the that idealizes romantic interdependence (Sprecher & Metts,
intimacy and love they seek. Thus, anxious men idealize 1989). Individuals with anxious romantic attachment styles
women as potential partners (consistent with benevolent desperately crave ‘‘true love’’ and want to merge completely
sexism) but also worry that women will be uncaring (consis- with partners, whom they tend to idealize; but they simulta-
tent with hostile sexism). Accordingly, Hart et al. found that neously fear they will be rejected, leading to defensive beha-
the anxiety ! benevolent sexism link was mediated by viors (e.g., jealousy; Hazan & Shaver, 1987). Their intense
endorsement of romanticism, whereas the anxiety ! hostile desire for love may predispose anxious individuals toward
sexism link appeared to reflect anxious men’s generalization romanticism, which views ‘‘true love’’ as a powerful merging
of their relational concerns—that is, directing their occa- of two souls in a relationship ‘‘meant to be.’’ Among men,
sional rejection-based hostility toward women as a group. Hart et al. (2012) found that anxious individuals tended to
endorse romanticism, which accounted for their greater
acceptance of benevolent sexism. By contrast, avoidant men
tended to reject romanticism, which accounted for their lower
The Present Research endorsement of benevolent sexism. Because romanticism,
Given the parallels between sexism directed toward women like both benevolent sexism toward women and BM,
and sexism directed toward men and their complementary expresses views about intimate interdependence, in the pres-
roles in maintaining a system of gender inequality, we ent research we expected mediational effects with women
suggest that attachment insecurities will predict women’s that parallel men’s. Specifically, we predicted that anxious
sexist beliefs about men. In particular, intimate interdepen- women would endorse romanticism, and this linkage would
dence represents a symmetrical complementarity between men account for their acceptance of BM; in contrast, avoidant
and women—heterosexual members of each sex ‘‘need’’ each women would reject romanticism and, in turn, reject BM.
other—creating similarities between benevolent sexism and However, due to power asymmetries, we expected a key dif-
BM that seem unavoidably related to attachment (because both ference from the dynamics of men’s sexism toward women with
ideologies focus on intimate relationships between the sexes). respect to mediation of the links between attachment and HM.
Therefore, as with men’s benevolent sexism, women’s BM Hart et al.’s (2012) work showed that SDO (Sidanius & Pratto,
should be based partly on their emotional and ideological 1999) mediated the relationship between men’s avoidant attach-
proclivities toward potential romantic relationship partners and ment and hostile sexism, suggesting that avoidant men are prone
toward romantic relationships in general. to viewing women as members of an adversarial out-group.
However, the other structural factor undergirding sexism SDO represents a particularly attractive intergroup ideology for
makes women’s HM different from men’s hostile sexism in members of powerful groups because it justifies the current
at least one important way. In contrast to intimate interdepen- group hierarchy (i.e., that groups in power deserve their privi-
dence between the sexes, social power is inherently asymme- leges). By contrast, members of disadvantaged groups are less
trical, such that traditionally, men are higher and women are attracted to SDO, an ideology that justifies their poorer social
lower in power. Power’s asymmetry should create gender- outcomes as legitimate. Indeed, research shows a persistent gen-
specific dynamics because prejudice directed ‘‘upwards’’ der gap in SDO, with men endorsing it more strongly than
(i.e., from women toward men) qualitatively differs from women (Sidanius & Pratto, 1999). Further, as noted above,
prejudice directed ‘‘downwards’’ (Cuddy, Fiske, & Glick, unlike hostile sexism, HM does not function as a system-
2008). Below, we explicate both the expected similarities and legitimizing ideology (even though it relates to perceived sys-
differences in how attachment insecurities predispose tem stability); thus, HM (unlike hostile sexism toward women)
women’s sexism toward men, as compared to men’s sexism should not be functionally related to SDO (the system-justifying
toward women. antecedent to hostile sexism), at least among women. In other
510 Psychology of Women Quarterly 37(4)

words, the asymmetry in power between the sexes suggests that in the United States, who were recruited through Mturk.com
SDO would not mediate avoidant women’s hypothesized HM. (see Buhrmester, Kwang, & Gosling, 2011), and compensated
Although we included SDO in the current studies, we did with credit toward their Amazon.com accounts. We eliminated
not expect it to act as a mediator. We predicted instead that three individuals from data analysis whose responses were
‘‘interpersonal trust’’—people’s generalized expectancies obviously disingenuous (e.g., one participant responded with
about the trustworthiness of others (Rotter, 1967)—would all 4s and another with a 4-5-4-5 alternating pattern) leaving
mediate between women’s attachment avoidance and HM. 229 participants. Their ages ranged from 18 to 78 (Mdn ¼
We chose this construct for three reasons. First, trust (or mis- 30). Most were Caucasian (182, 79.5%), with the rest identify-
trust) seems likely to be part of avoidant individuals’ world- ing as Asian American/Pacific Islander (18, 7.9%), African
view because their working models depict close relationship American (17, 7.4%), Hispanic/Latina (6, 2.6%), Native
partners as distant and undependable; in fact, the original mea- American (5, 2.2%), Middle Eastern American (1, 0.4%),
sure of avoidance in adulthood emphasized difficulty trusting ‘‘other’’ (3, 1.3%), or ‘‘more than one’’ (8, 3.5%). (Percentages
romantic partners (Hazan & Shaver, 1987). If these working sum to more than 100 because participants were allowed to
models generalize beyond close relationships, avoidantly select multiple options.)
attached individuals would be lower on interpersonal trust.
Second, whereas hostile sexism is system justifying,
thereby naturally stemming from a system-justifying ideo- Materials and Procedure
logy like SDO (Jost, Banaji, & Nosek, 2004; Sibley, Participants first completed online questionnaires assessing,
Wilson, & Duckitt, 2007), HM expresses antipathy toward in order: attachment style, ambivalence toward men, roman-
men and their systemic power. Therefore, HM seems likely ticism, interpersonal trust, and SDO. We intentionally placed
to originate from a generally cynical and suspicious world- the hypothesized mediators last to reduce the chance that
view. If so, women’s avoidance should predict such cynical questionnaire order could spuriously contribute to the media-
views about others (i.e., lower trust), which, in turn, should tional relationships we predicted. For all measures, partici-
predict mistrust about men’s motives; that is, HM (which pants rated their agreement with items on a 1 (disagree
views men as seeking to control women). A third advantage strongly) to 7 (agree strongly) scale, and scores were aver-
to exploring trust as a mediator is that it might also mediate aged across all items (after transforming reverse-coded items
the relationship of women’s anxiety to HM (whereas SDO where applicable). Table 1 displays a coefficients for each of
is unrelated to anxiety, even among men). Attachment anxi- the scales. Finally, participants provided demographic infor-
ety involves concerns that others will not be responsive to mation (age, followed by ethnic identification and sexual
one’s needs or desires for intimacy. For anxious women, this orientation).
mistrust might, in turn, predict HM. The Experiences in Close Relationships (ECR) inventory
In sum, we hypothesized: (a) women’s attachment anxiety measured attachment anxiety and avoidance (Brennan, Clark,
would predict a tendency to endorse both HM and BM; (b) & Shaver, 1998). The ECR has been extensively validated
women’s attachment avoidance would predict a tendency to and used in hundreds of studies to measure adult attachment
endorse HM and reject BM; (c) the links from attachment inse- style, with typical a coefficients above .90 (Mikulincer &
curities to BM would be mediated by romanticism (with avoid- Shaver, 2007). It includes 36 items assessing attachment
ance leading to less romanticism and therefore lower BM, but anxiety (e.g., ‘‘I worry about being abandoned’’ and ‘‘I often
anxiety leading to higher romanticism and therefore higher wish that my partner’s feelings for me were as strong as my
BM); and (d) the links from both attachment insecurities to feelings for him/her’’) and avoidance (e.g., ‘‘I don’t feel com-
HM would be mediated by lower interpersonal trust. Because fortable opening up to romantic partners’’ and ‘‘I am nervous
we wanted to ensure that our results could not be attributed to a when partners get too close to me’’). Higher scores indicate
methodological artifact, such as the particular questionnaire greater anxiety or avoidance.
order we chose (i.e., such that completing one questionnaire A short version of the Ambivalence Toward Men Inven-
could prime concepts that influence responses on subsequent tory (AMI; Glick & Whitehead, 2010) measured BM and
questionnaires), we tested these hypotheses in two studies that HM. An example of the 6 items assessing HM is: ‘‘Even men
varied the order of questionnaire presentation. who claim to be sensitive to women’s rights really want a tra-
ditional relationship at home, with the woman performing
Study 1 most of the housekeeping and child care.’’ An example of one
of the six BM items is: ‘‘A woman will never be truly fulfilled
in life if she doesn’t have a committed, long-term relationship
Method with a man.’’ Higher scores indicate greater BM or HM. The
AMI has been validated in prior research, including a 16-
Participants nation study involving over 8,000 participants in which HM
After excluding 19 women who self-identified as either lesbian was shown to predict more negative and BM more positive
or ‘‘other,’’ participants were 232 heterosexual women located stereotypes of men (Glick et al., 2004). Glick et al. (2004)
Hart et al. 511

Table 1. Correlation Matrix for Both Studies’ Variables.

Study 1 Study 2
M (SD) M (SD) Anxiety Avoidance BM HM Romanticism Trust SDO

Anxiety 3.89 (1.23) 3.63 (1.23) (.94, .94) .30*** .17* .29*** .16* .18** .12
Avoidance 3.03 (1.21) 2.72 (1.17) .15* (.95, .96) .10 .18** .26*** .16* .02
BM 3.39 (1.25) 3.56 (1.19) .12 .15* (.75, .79) .43*** .46*** .05 .21**
HM 3.67 (1.17) 3.81 (1.14) .16* .21** .54*** (.79, .79) .20** .31*** .08
Romanticism 4.16 (0.93) 4.35 (0.98) .17** .31*** .41*** .18** (.86, .88) .10 .08
Trust 4.64 (0.77) 4.52 (0.83) .28*** .15* .25*** .36*** .06 (.72, .78) .08
SDO 2.63 (1.24) 2.54 (1.10) .04 .08 .24*** .21*** .07 .16* (.89, .86)
Note. BM ¼ benevolence toward men; HM ¼ hostility toward men; SDO ¼ social dominance orientation.
Study 1’s correlation coefficients are above the diagonal; Study 2’s are below the diagonal. The coefficient a for each measure is reported in parentheses along
the diagonal (Studies 1 and 2).
*p < .05. **p < .01. ***p < .001.

also found evidence for a consistent structure (with correlated of the scale used here has shown good internal consistency
but distinct HM and BM subfactors) across nations. Addi- (a ¼ .83; Weber & Federico, 2007). Higher scores indicate
tional detailed psychometric analyses of the AMI can be higher SDO.
found in Glick and Fiske (2001).
The 15-item Romantic Beliefs scale assessed romanti-
cism. Its reliability (a ¼ .81) and validity were established Results
in a large (N ¼ 730) sample of undergraduate students whose Table 1 displays the zero-order correlations among all study
responses to the romantic beliefs items correlated with their variables. Our main hypotheses—that anxiously attached
scores on related scales measuring aspects of love and women would exhibit ambivalence toward men (higher HM
romance (Sprecher & Metts, 1989). Items include: ‘‘Once I and BM), whereas avoidant women would show only hostile
experience ‘true love,’ I could never experience it again, to sexist attitudes toward men and reject benevolent ones
the same degree, with another person’’ and ‘‘The relationship (higher HM, lower BM)—are most stringently tested using
I will have with my ‘true love’ will be nearly perfect.’’ Higher partial correlations (e.g., the correlation between attachment
scores indicate higher romanticism. anxiety and BM controlling for attachment avoidance and
In the interest of brevity, we measured trust using the HM). As expected, attachment anxiety correlated positively
‘‘interpersonal exploitation’’ dimension of Rotter’s (1967) with HM, r(225) ¼ .18, p < .01, but its predicted correlation
Interpersonal Trust Scale (see Chun & Campbell, 1974). Rot- with BM did not attain significance, r(225) ¼ .11, p ¼ .10.
ter’s scale is one of the most frequently used trust measures Also as expected, avoidance correlated positively with HM,
(Colquitt, Scott, & LePine, 2007). The interpersonal exploita- r(225) ¼ .19, p < .01, and negatively with BM, r(225) ¼
tion dimension has been shown to have adequate internal con- .22, p < .01. Notably, SDO did not correlate with either
sistency (a ¼ .60; Chun & Campbell, 1974). It includes 8 avoidance or HM, consistent with our prediction that it would
items (e.g., ‘‘In dealing with strangers one is better off to not mediate the avoidance ! HM link.
be cautious until they have provided evidence that they are We next tested our complete model of the structural rela-
trustworthy’’ [reverse coded] and ‘‘Most people can be tions between attachment and sexism toward men, including
counted on to do what they say they will do’’). This dimen- the hypothesized mediating variables of romanticism and
sion seemed the most general compared with the other three interpersonal trust and using structural equation modeling
dimensions of interpersonal trust, whose items focus specifi- (SEM; see Figure 1). For the measurement model, we used
cally on political cynicism, hypocrisy, and reliable role per- the method described by Sibley, Wilson, and Duckitt
formance (e.g., ‘‘Most repairmen will not overcharge even (2007) to create three parcels for each latent variable in the
if they think you are ignorant of their specialty’’), although model. To create the parcels, items from each measure were
we think it is likely that using the entire scale would yield randomly divided into three groups, using equivalent numbers
similar results. Higher scores indicate greater interpersonal of reverse-scored items where possible. Items in each group
trust. were averaged to create three parcels (or observed variables)
The SDO scale included 8 items (Weber & Federico, to load onto each latent variable in the model. (See Table 2 for
2007), such as ‘‘some groups of people are simply inferior the factor loadings of the parcels that predict each latent
to others.’’ This widely used scale has been shown to relate variable.) The model fit the data adequately, w2 ¼ 205.96, df
to a broad range of hierarchical ideologies, including political ¼ 123, w2/df ¼ 1.67, root mean square of approximation
conservatism, just-world beliefs, and nationalism—in addi- (RMSEA) ¼ .05, comparative fit index (CFI) ¼ .97. Together,
tion to sexism (Sidanius & Pratto, 1999). The short version anxiety, avoidance, and trust explained 23% of the variance in
512 Psychology of Women Quarterly 37(4)

Figure 1. Structural equation model predicting women’s benevolence (BM) and hostility (HM) toward men from attachment anxiety and
avoidance for Studies 1 (top) and 2 (bottom). Dashed lines represent nonsignificant pathways.

HM, and anxiety, avoidance, and romanticism explained 27% Anxiety ! Romanticism ! BM
of the variance in BM. The SEM analysis supported all of the
Results supported our prediction that women’s romanticism
predicted links, with one exception: the avoidance ! trust
would mediate the positive anxiety ! BM link. Anxiety sig-
link. Although in the predicted direction, this relationship was
nificantly predicted the proposed mediator, romanticism, at
not significant (and we address this finding below).
.15 (p < .01), and in turn, the direct effect of romanticism on
Finally, we conducted direct tests of our mediational
BM was significant at .49 (p < .0001). The total effect of anxi-
hypotheses by following Preacher and Hayes’s (2004,
ety on BM was of .10 was not significant (p ¼ .10), and the
2008) recommendations. We ran mediation analyses using
direct link from anxiety to BM (.03) was also not significant
5,000 bootstrap samples and estimated bias-corrected and
(p ¼ .63). Despite the total effect of anxiety on BM being non-
accelerated 95% confidence intervals. When attachment
significant, the confidence interval for the indirect effect of
anxiety was the independent variable, we included avoidance
anxiety on BM via romanticism was .0222 to .1437, indicating
as a covariate (and vice versa); we also included both sexism
a significant test of mediation (because the confidence interval
dimensions in each analysis.
excludes zero).
Hart et al. 513

Table 2. Regression Weights (b) for the Parcels That Predict Each Discussion
Latent Variable.
Study 1 partially supported our central predictions that
Study 1 Study 2 women’s attachment anxiety would predict both hostility and
BM, whereas women’s attachment avoidance would predict
Parcel Parcel Parcel Parcel Parcel Parcel
Variables 1 2 3 1 2 3 HM combined with lower benevolence. Specifically, we
found support for all but one predicted relationship: The anxi-
Anxiety .90* .91* .92* .91* .91* .94* ety ! BM relationship, although significant as a first-order
Avoidance .95* .96* .95* .98* .96* .94* correlation, was reduced to nonsignificance using the more
BM .69* .78* .84* .73* .77* .80* stringent partial correlation (controlling for the BM–HM rela-
HM .77* .78* .66* .81* .79* .65*
Romanticism .70* .84* .87* .67* .85* .96*
tionship). Our results also supported the hypotheses that
Trust .73* .66* .78* .72* .75* .86* lower interpersonal trust would mediate the links between
attachment avoidance and HM, whereas romanticism would
Note. BM ¼ benevolence toward men; HM ¼ hostility toward men. mediate the links between attachment and BM. However,
*p < .001.
controlling for trust did not significantly reduce the total
effect of anxiety on HM, in contrast to our tentative predic-
Avoidance ! Romanticism ! BM tion that trust would mediate the anxiety ! HM link in
addition to the avoidance ! HM link. Also unexpectedly,
Results also supported the prediction that rejecting romanti- we found that women’s avoidance predicted HM due to
cism would mediate the (negative) avoidance ! BM link. covariance with anxiety, suggesting that a general insecurity
Avoidance significantly predicted the proposed mediator, factor could explain that avoidance ! HM link. These last
romanticism, at .28 (p < .0001), and in turn, the direct effect two unexpected findings, along with our concerns about pos-
of romanticism on BM was significant at .49 (p < .0001). The sible testing effects, led us to conduct a replication study.
total effect of avoidance on BM was significant at .22 (p <
.001), and the direct effect was not significant at .08 (p ¼
.20). The confidence interval for the indirect effect of avoid-
ance on BM via romanticism was .2181 to .0706, indicat- Study 2
ing a significant test of mediation. Study 2 replicated Study 1, but with one important difference:
We transposed the questionnaires’ presentation order. Specif-
Anxiety ! Trust ! HM ically, instead of completing the ECR first, followed by the
AMI, participants in Study 2 completed those questionnaires
We hypothesized that women’s lower interpersonal trust in the reverse order. Additionally, we swapped the order of
would mediate the anxiety ! HM link. Anxiety significantly the trust and romanticism questionnaires. Changing the ques-
predicted the proposed mediator, trust, at .09 (p < .05), and tionnaire order allows us to test the possibility that the order
in turn, the direct effect of trust on HM was significant at may have contributed to Study 1’s specific pattern of
.36 (p < .0001). The total effect of anxiety on HM was findings. For example, it seems possible that filling out the
significant at .16 (p < .01), and the direct effect was signifi- attachment questionnaires could have primed romantic rela-
cant at .13 (p < .05). However, the confidence interval for the tionship schemas that then influenced attitudes toward men
indirect effect of anxiety on HM via trust was .0037 to in the predicted direction. Hence, Study 2 tested whether our
.0722, which did not reach the threshold for a significant findings would replicate, as well as whether they would hold
test of mediation. up given a different questionnaire order.

Avoidance ! Trust ! HM Participants and Procedure


We had predicted that avoidance would influence HM via After excluding 26 women who self-identified as either les-
lower interpersonal trust. However, the significant zero- bian or ‘‘other,’’ participants were 273 women located in the
order correlation between avoidance and trust did not trans- United States recruited through Mturk.com. Three partici-
late into a significant path between avoidance and trust in the pants were removed because they had participated in Study
SEM. This suggested that covariance with another variable 1. The remaining participants’ ages ranged from 18 to 82
might account for the influence of avoidance on trust. Indeed, (Mdn ¼ 31). Most were Caucasian (203, 75.2%), with the rest
regression analyses revealed that the significant influence of identifying as African American (27, 10.0%), Hispanic/
avoidance on trust was eliminated (p ¼ .10) when controlling Latina (18, 6.7%), Asian American/Pacific Islander (15,
for attachment anxiety. Thus, we conclude that attachment 5.6%), Middle Eastern American (3, 1.1%), Native American
avoidance predicts lower interpersonal trust—and therefore (1, 0.4%), ‘‘other’’ (1, 0.4%), or ‘‘more than one’’ (6, 2.2%).
higher HM—in part due to common variance shared with (Percentages sum to more than 100 because participants were
attachment anxiety.1 allowed to select multiple options.) The materials and
514 Psychology of Women Quarterly 37(4)

procedure were identical to Study 1 except for the previously indirect effect of avoidance on BM via romanticism was
noted change in the questionnaire order. .1626 to .0518, indicating a significant test of mediation.

Anxiety ! Trust ! HM
Results
In Study 1, anxious women’s lower interpersonal trust did not
Table 1 displays the zero-order correlations between the significantly mediate the anxiety ! HM link, but in Study 2,
study variables. As in Study 1, the lack of a significant corre- it did. Anxiety significantly predicted the proposed mediator,
lation between SDO and avoidance suggests that SDO did not trust, at .15 (p < .0001), and in turn, the direct effect of trust
mediate the avoidance ! HM link (despite SDO correlating on HM was significant at .26 (p < .001). The total effect of
positively with HM in Study 2). Also as in Study 1, the partial anxiety on HM was not significant at .04 (p ¼ .36), and the
correlations supported our main predictions concerning the direct effect was also not significant at .003 (p ¼ .95). Not-
relation between attachment avoidance and both HM and withstanding the nonsignificant total effect of anxiety on
BM: r(266) ¼ .33, p < .001 and r(266) ¼ .32, p < .001, HM, the confidence interval for the indirect effect of anxiety
respectively. Different from Study 1, neither of the partial on HM via trust was .0140 to .0780, indicating a significant
correlations between attachment anxiety and HM or BM were test of mediation.
significant: r(266) ¼ .06, p ¼ .36 and r(266) ¼ .08, p ¼ .17,
respectively. The discrepancy between the zero-order and
partial correlations in Study 2 suggests that the association
Avoidance ! Trust ! HM
between attachment anxiety and each of the sexism dimen- Our hypothesis that avoidance would predict HM via lower
sions was due to shared variance between attachment anxiety interpersonal trust—which received qualified support in
and the other sexism dimension. Thus, it seems that, in Study Study 1—was better supported by Study 2’s results. Avoid-
2, women’s attachment anxiety predicted their HM in part ance significantly predicted the proposed mediator, trust, at
because of their BM—and vice versa. .11 (p < .05), and in turn, the direct effect of trust on HM
To test our full model, we again used SEM (see Table 2 was significant at .26 (p < .001). The total effect of avoidance
and Figure 1), applying the same parameters as in Study 1. on HM was significant at .28 (p < .0001), and the direct effect
The model again fit the data adequately, w2 ¼ 241.07, df ¼ was also significant at .25 (p < .0001). The confidence inter-
123, w2/df ¼ 1.96, RMSEA ¼ .06, CFI ¼ .97. Together, val for the indirect effect of avoidance on HM via trust was
anxiety, avoidance, and trust explained 13% of the variance .0070 to .0585, indicating a significant test of mediation.
in HM; anxiety, avoidance, and romanticism explained
12% of the variance in BM.2 We also conducted mediation
analyses according to the same procedures described in rela- Discussion and Comparison of the Two
tion to Study 1. Studies
Study 2 provided additional data with which to evaluate our
Anxiety ! Romanticism ! BM theoretical model. As in Study 1, the model was generally
Results replicated Study 1’s finding that romanticism signif- supported. The main exception was that neither the partial
icantly mediated the anxiety ! BM link. Anxiety signifi- correlations nor the total effects of attachment anxiety on
cantly predicted the proposed mediator, romanticism, at .15 HM and BM were significant in Study 2, despite significant
(p < .001), and in turn, the direct effect of romanticism on zero-order correlations. That the relation between attachment
BM was significant at .30 (p < .0001). The total effect of anxi- anxiety and sexism became attenuated when controlling for
ety on BM was not significant at .07 (p ¼ .17), and the direct other overlapping variables may suggest that the unique var-
effect was also not significant at .02 (p ¼ .67). Despite the iance in each sexism dimension explained by attachment
nonsignificant total effect, the confidence interval for the anxiety is relatively modest. However, even if anxiety
indirect effect of anxiety on BM via romanticism indicated predicts shared variance between HM and BM, the basic pre-
a significant test of mediation (.0171 to .0926). diction holds: Anxiously attached women evince greater
ambivalence (higher HM and higher BM) toward men.
Further, the mediation analyses (anxiety ! romanticism !
Avoidance ! Romanticism ! BM
BM and anxiety ! lower trust ! HM) were both significant
As in Study 1, rejecting romanticism significantly mediated the in Study 2. We note that according to current recommenda-
(negative) avoidance ! BM link. Avoidance significantly pre- tions concerning mediation analyses, the significant indirect
dicted the proposed mediator, romanticism, at .32 (p < .0001), effects trump the lack of significant total effects. For exam-
and in turn, the direct effect of romanticism on BM was signif- ple, Rucker, Preacher, Tormala, and Petty (2011) argue that:
icant at .30 (p < .0001). The total effect of avoidance on BM was
significant at .29 (p < .0001), and the direct effect was also Although there is value in testing the total effect of X on Y,
significant at .19 (p < .001). The confidence interval for the we propose that overemphasizing the X ! Y relationship
Hart et al. 515

before or after controlling for a mediator can lead to mislead- women, their lower interpersonal trust (Studies 1 and 2)
ing, or even false, conclusions in theory testing . . . research- accounted for their endorsement of HM (Study 2 only).
ers interested in understanding intervening effects in Importantly, the relations between the variables measured
proposed theoretical models should shift attention to testing in our studies are complex enough—anxiety and avoidance
the mediation effect itself and not constrain themselves by predict HM in the same manner but have opposite relations
placing undue emphasis on the significance of the X ! Y to BM—that they cannot be parsimoniously explained by
relationship. (p. 360) response biases, which should mitigate concerns about the
use of self-report methodology.
Moreover, Study 2 showed that the avoidance ! HM and Although caution is warranted regarding the findings that
BM relations found in Study 1’s partial correlation results were inconsistent across studies, taken together, these and
were robust and reliable, and Study 2 was also consistent with earlier findings (Hart et al., 2012) are consistent with the view
our theoretical model concerning the mediation of the avoid- that sexism may be partly derived from social beliefs and
ance to sexism links (avoidance ! lower romanticism ! attitudes that arise from attachment working models. Of
lower BM and avoidance ! lower trust ! HM). course, our use of correlational methodology restricts causal
Despite the overall consistency, there were two differ- inferences, but our findings agree with theoretical models
ences in mediational analyses across Studies 1 and 2. First, positing that just as sexism toward men and women share
the anxiety ! lower trust ! HM mediational path was sig- structural similarities, they share some antecedents. Broadly
nificant only in Study 2. It seems that the path may simply speaking, different patterns of sexism may be predisposed
be a weak one: All the constituent paths were significant in by different attachment insecurities: Individuals’ attachment
Study 1 but the test of mediation did not reach significance; anxiety and avoidance in adult romantic relationships lead
Study 2 used a larger sample and found significant mediation. them to develop worldviews that predispose them to adopt
Second, the avoidance ! lower trust ! HM path, significant or reject benevolent and/or hostile sexist attitudes toward
in Study 2, was nonsignificant in Study 1 when controlling members of the other sex.
for attachment anxiety. We conclude that covariance between However, important nuances distinguish between men’s
attachment anxiety and avoidance and between HM and BM and women’s attitudes toward the other sex. Whereas SDO
may sometimes complicate interpretation of the unique mediates the link between men’s attachment avoidance and
relations among them. hostile sexism toward women (Hart et al., 2012), it does not
mediate women’s hostile sexism toward men. We have sug-
gested that this difference reflects asymmetry in power, with
General Discussion men’s hostility toward women protecting men’s power and
We have presented a theoretical model which proposes privilege (as social dominance beliefs do), whereas women’s
attachment style as a potential personality antecedent to het- HM reflects lack of power relative to men. Instead, the pres-
erosexual women’s sexism toward men. We therefore ent study revealed that women’s HM is related partly to (lack
hypothesized that women’s attachment anxiety would predict of) interpersonal trust. Trust represents a more interpersonal
ambivalent sexism toward men (higher HM and BM), (as opposed to intergroup) attitudinal mediator, which was
whereas their avoidance would predict univalent hostile sex- not investigated in prior work on men’s sexism toward
ism (higher HM, lower BM). Results of the two studies were women (Hart et al., 2012). We can therefore only speculate
generally consistent with these hypotheses. In particular, the about trust’s potential role in men’s attitudes. However, given
links between avoidance and sexism were robust across both the competitive themes in the hostile sexism scale (e.g.,
studies. However, the links between attachment anxiety and ‘‘Feminists are seeking for women to have more power than
each sexism dimension were attenuated when controlling for men’’ or ‘‘Women are too easily offended’’; Glick & Fiske,
avoidance and/or the other sexism dimension, suggesting that 1996), interpersonal trust seems a likely mediator between
these links are comparatively weak. men’s attachment insecurity and hostile sexism.
The two studies also generally (but not completely) sup-
ported our model’s mediational predictions: Compared to
women lower in attachment anxiety, anxiously attached
Limitations
women tended to idealize romance (Studies 1 and 2), which The present study was limited in some ways. Most notably,
accounted for their endorsement of benevolently sexist atti- our correlational design precludes confident causal conclu-
tudes toward men (Studies 1 and 2), but they also tended to sions. Specifically, plausible alternative causal models may
generally distrust others (Studies 1 and 2), which accounted be compatible with our data; for example, attachment anxiety
for their endorsement of hostile sexist attitudes toward men and avoidance could have direct causal influence on sexist
(Study 2 only). By contrast, compared to women lower in ideologies, which in turn influence romanticism and interper-
attachment avoidance, avoidantly attached women tended sonal trust (flipping our proposed mediators and dependent
to reject romanticism (Studies 1 and 2), which accounted for variables). Indeed, we conducted exploratory analyses of
their rejection of BM (Studies 1 and 2); yet, like anxious alternative models testing sexism (HM and BM) as mediating
516 Psychology of Women Quarterly 37(4)

links from attachment to romanticism and trust, and we found possibility that the model we have described is moderated by
that they were significant in three of the four analyses for ethnicity or other demographic factors.
Study 1 and two of the four analyses for Study 2. Although
such bidirectional mediation is not uncommon, it underscores
the need for caution in drawing directional conclusions from Theoretical Implications
mediation analyses (we note that similar caution is warranted We hope the present research will prove useful to attachment
even in experimental studies employing mediation analysis; and other personality and social psychology theorists. Even
see Spencer, Zanna, & Fong, 2005). though causality is uncertain for our findings, they agree with
However, consistent with our preferred interpretation, a large literature depicting attachment style as a personality
prior research and theory suggest that stable attachment styles predictor of many psychological outcomes. That is, attach-
develop quite early in life (Bowlby, 1982), before children ment does not simply relate to individuals’ close relation-
are even aware that they belong to stable social groups (i.e., ships; rather, it broadly relates to patterns of thought,
before gender constancy develops; see Lutz & Ruble, feeling, and behavior across many domains over the life span.
1995). Furthermore, some experimental work shows that These ‘‘downstream’’ processes suggest a kind of coherence
priming attachment security changes ideology, including to psychological functioning in which seemingly disparate
values and intergroup attitudes (Mikulincer et al., 2003; processes actually operate in accord—for instance, as a result
Mikulincer & Shaver, 2001). Similarly, theorists propose that of a more fundamental motivational process. In an example
general intergroup ideologies precede rather than follow the from the present study, attachment anxiety, which reflects
development of attitudes toward specific groups. For exam- operations directed at maintaining feelings of security, relates
ple, in his dual process model of ideology and prejudice, to beliefs (i.e., romanticism) and, in turn, attitudes (i.e., BM)
Duckitt (2001) argues that general intergroup ideologies that reflect or resonate the underlying dynamic. In this way,
(e.g., SDO) precede and influence more specific intergroup women’s benevolent attitudes toward men could be construed
attitudes (e.g., racism or sexism). Further, a longitudinal as oriented toward augmenting a romantic worldview that func-
study by Sibley et al. (2007) supported this view when tions to counteract attachment insecurities. Such ‘‘motivated
predicting men’s hostile and benevolent sexism toward social cognition’’ (Kruglanski, 1996) appears to be ubiquitous
women. Specifically, after controlling for baseline levels of and may well represent a basic principle of psychological
all variables, general intergroup ideologies (right-wing functioning.
authoritarianism and SDO) predicted changes over time in In a similar vein, there may be other personality traits or
men’s sexist beliefs. Nevertheless, alternative models (e.g., processes beside attachment that contribute to sexism in
with reciprocal causal paths) remain plausible until they are analogous ways. If sexism is a general response to, or com-
disconfirmed, so longitudinal or experimental studies would pensation for, underlying insecurities, then perhaps other
be useful to help verify the causal model we have posited. forms of dispositional insecurity, such as lower self-esteem,
Another caveat that bears mentioning is that some of our intolerance of uncertainty, or death anxiety may also play a
results changed depending on which variables were included role (cf. Hart et al., 2005; Jost, Glaser, Kruglanski, & Sullo-
in analyses. For example, in Study 1, the avoidance ! trust way, 2003).
link became nonsignificant when anxiety was controlled.
Although this was not true in Study 2, and the test for media-
tion in Study 1 was significant despite the nonsignificant Practice Implications
effect of avoidance on trust, it is possible that a general For researchers interested in how sexist beliefs and attitudes
‘‘attachment insecurity’’ factor may account for much of the influence romantic relationship outcomes, our findings sug-
relation between avoidance and HM via lower interpersonal gest teasing out attachment style’s influence. Attachment
trust, with avoidance explaining little additional (unique) var- style likely influences relationship outcomes directly, but our
iance. Similarly, in Study 2 (but not in Study 1), attachment research suggests that it may also predict outcomes via
anxiety’s correlation with HM became nonsignificant when ideologies such as sexism, which has recently been found
controlling for BM, and vice versa. To our knowledge, there to predict romantic relationship dynamics (Overall, Sibley,
is no research that addresses the nature of the shared variance & Tan, 2011).
between anxiety and avoidance, or between HM and BM, Therapists, particularly marital counselors, may find the
respectively (research instead focuses on the unique individ- implications for romantic relationships to be of particular
ual properties of the two insecurity and sexism dimensions), interest. Attachment insecurities, of course, have long been
so it is difficult to interpret these findings. This may suggest known to predict conflict and instability in relationships. The
an area for additional research. current research adds insight into how attachment insecurity
Finally, it should be noted that we do not know the extent translates into more generalized sexist ideologies. Such ideol-
to which the present results can be generalized. For example, ogies can, in turn, influence the types of men to whom women
although we think our analysis is likely to apply across are attracted (Montañés, de Lemus, Moya, Bohner, & Megı́a,
cultures, additional research will be necessary to examine the 2013; Sibley & Overall, 2011) and desire for men to ‘‘take
Hart et al. 517

charge’’ in at least some aspects of their romantic relation- Bowlby, J. (1982). Attachment and loss: Attachment (Vol. 1, 2nd ed.).
ships (Sarlet, Dumont, Delacollete, & Dardenne, 2012). More New York, NY: Basic Books.
broadly, sexist ideologies can lead women to unconsciously Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report mea-
justify and tolerate inequalities (whether in relationships or surement of adult romantic attachment: An integrative overview.
life more generally; Connelly & Heesacker, 2012) that they In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and
might resist or seek to change if they more carefully reflected close relationships (pp. 46–76). New York, NY: Guilford Press.
on these beliefs. Therapists could explore, within the context Brumbaugh, C. C., & Fraley, R. C. (2006). Transference and attach-
of psychoeducational interventions, how clients’ attachment ment: How do attachment patterns get carried forward from one
insecurities and sexist beliefs might be linked in their past and relationship to the next? Personality and Social Psychology
current relationships and life choices. Bulletin, 32, 552–560.
Buhrmester, M., Kwang, T., & Gosling, S. D. (2011). Amazon’s
mechanical turk: A new source of inexpensive, yet high-
Conclusion
quality, data? Perspectives on Psychological Science, 6, 3–5.
To our knowledge, ours is the first research to closely exam- Chun, K., & Cambell, J. B. (1974). Dimensionality of the Rotter
ine personality correlates of women’s sexism toward men. In interpersonal trust scale. Psychological Reports, 35, 1059–1070.
general, research on sexism has neglected questions about Colquitt, J. A., Scott, B. A., & LePine, J. A. (2007). Trust, trust-
personality, instead focusing on ideological correlates of worthiness, and trust propensity: A meta-analytic test of their
sexism. But if ideology is partly a function of personality, unique relationships with risk taking and job performance.
it is important for researchers to explore the distal roots of Journal of Applied Psychology, 92, 909–927.
sexism and to piece together comprehensive models of its Connelly, K., & Heesacker, M. (2012). Why is benevolent sexism
antecedents. Not only will such lines of inquiry open up new appealing?: Associations with system justification and life satis-
possibilities for interventions aiming to reduce sexism, and to faction. Psychology of Women Quarterly, 36, 432–443.
understand and ameliorate related romantic relationship Cuddy, A. J. C., Fiske, S. T., & Glick, P. (2008). Warmth and
problems, but they will also help theorists understand more competence as universal dimensions of social perception: The
about the psychodynamic interplay between emotion and stereotype content model and the BIAS map. In M. P. Zanna
cognition as well as trace the processes leading from personal (Ed.), Advances in experimental social psychology (Vol. 40,
and interpersonal precursors to ideological outcomes. We pp. 61–150). Thousand Oaks, CA: Academic Press.
hope the present research will inspire additional investiga- Duckitt, J. (2001). A dual-process cognitive-motivational theory of
tions of the theoretical model we have articulated. ideology and prejudice. In M. P. Zanna (Ed.), Advances in
experimental social psychology (Vol. 33, pp. 41–113). San
Declaration of Conflicting Interests Diego, CA: Academic Press.
The author(s) declared no potential conflicts of interest with respect Feather, N. T. (2004). Value correlates of ambivalent attitudes
to the research, authorship, and/or publication of this article. toward gender relations. Personality and Social Psychology Bul-
letin, 30, 3–12.
Funding Feeney, B. C., Cassidy, J., & Ramos-Marcuse, F. (2008). The general-
The author(s) declared the following financial support for the ization of attachment representations to new social situations: Pre-
research, authorship, and/or publication of this article: This research dicting behavior during initial interactions with strangers. Journal
was supported in part by an internal faculty research grant from of Personality and Social Psychology, 95, 1481–1498.
Union College. Glick, P., & Fiske, S. T. (1996). The ambivalent sexism inventory:
Differentiating hostile and benevolent sexism. Journal of
Notes Personality and Social Psychology, 70, 491–512.
1. Although we did not predict interactions between anxiety and Glick, P., & Fiske, S. T. (2001). Ambivalent sexism. In M. P. Zanna
avoidance, exploratory analyses showed an interaction (p ¼ .03) (Ed.), Advances in experimental social psychology (Vol. 33, pp.
on hostility toward men (HM), such that secure attachment (lower 115–188). Thousand Oaks, CA: Academic Press.
anxiety and avoidance) predicted the lowest HM compared to the Glick, P., Lameiras, M., Fiske, S. T., Eckes, T., Masser, B., Volpato,
other attachment styles. This interaction did not replicate in Study C., . . . Wells, R. (2004). Bad but bold: Ambivalent attitudes
2, so we do not interpret it further. toward men predict gender inequality in 16 nations. Journal of
2. We speculate that the reduced variance explained in Study 2 Personality and Social Psychology, 86, 713–728.
compared to Study 1 may be attributable to our rearranged ques- Glick, P., & Whitehead, J. (2010). Hostility toward men and the
tionnaire order. perceived stability of male dominance. Social Psychology, 41,
177–185.
References Hart, J., Hung, J. A., Glick, P., & Dinero, R. E. (2012). He loves her,
Anderson, K. J., Kanner, M., & Elsayegh, N. (2009). Are feminists he loves her not: Attachment style as a personality antecedent to
man haters? Feminists’ and nonfeminists’ attitudes toward men. men’s ambivalent sexism. Personality and Social Psychology
Psychology of Women Quarterly, 33, 216–224. Bulletin, 38, 1495–1505.
518 Psychology of Women Quarterly 37(4)

Hart, J., Shaver, P. R., & Goldenberg, J. L. (2005). Attachment, Overall, N. C., Sibley, C. G., & Tan, R. (2011). The costs and
self-esteem, worldviews, and terror management: Evidence for benefits of sexism: Resistance to influence during relationship
a tripartite security system. Journal of Personality and Social conflict. Journal of Personality and Social Psychology, 101,
Psychology, 88, 999–1013. 271–290.
Hazan, C., & Shaver, P. R. (1987). Romantic love conceptualized as Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures
an attachment process. Journal of Personality and Social for estimating indirect effects in simple mediation models. Beha-
Psychology, 52, 511–524. vior Research Methods, Instruments, & Computers, 36, 717–731.
Hazan, C., & Shaver, P. R. (1994). Attachment as an organizational Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling
framework for research on close relationships. Psychological strategies for assessing and comparing indirect effects in multi-
Inquiry, 5, 1–22. ple mediator models. Behavior Research Methods, 40, 879–891.
Jost, J. T., Banaji, M. R., & Nosek, B. A. (2004). A decade of system Rotter, J. (1967). A new scale for the measurement of interpersonal
justification theory: Accumulated evidence of conscious and trust. Journal of Personality, 35, 651–665.
unconscious bolstering of the status quo. Political Psychology, Rucker, D. D., Preacher, K. J., Tormala, Z. L., & Petty, R. E. (2011).
25, 881–919. Mediation analysis in social psychology: Current practices and
Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003). new recommendations. Social and Personality Psychology
Political conservatism as motivated social cognition. Psychologi- Compass, 5, 359–371.
cal Bulletin, 129, 339–375. Sarlet, M., Dumont, M., Delacollette, N., & Dardenne, B. (2012).
Kruglanski, A. W. (1996). Motivated social cognition: Principles of Prescription of protective paternalism for men in romantic and
the interface. In E. T. Higgins & A. W. Kruglanski (Eds.), Social work contexts. Psychology of Women Quarterly, 36, 444–457.
psychology: Handbook of basic principles (pp. 493–520). New Sibley, C. G., & Overall, N. C. (2011). A dual process motivational
York, NY: Guilford Press. model of ambivalent sexism and gender differences in romantic
Lutz, S. E., & Ruble, D. N. (1995). Children and gender prejudice: partner preferences. Psychology of Women Quarterly, 35,
Context, motivation, and the development of gender concep- 303–317.
tions. Annals of Child Development, 10, 131–166. Sibley, C. G., Wilson, M. S., & Duckitt, J. (2007). Antecedents of
Mikulincer, M., Gillath, O., Sapir-Lavid, Y., Yaakobi, E., Arias, K., men’s hostile and benevolent sexism: The dual roles of social
Tal-Aloni, L., . . . Bor, G. (2003). Attachment theory and dominance orientation and ring-wing authoritarianism. Person-
concern for others’ welfare: Evidence that activation of the sense ality and Social Psychology Bulletin, 33, 160–172.
of secure base promotes endorsement of self-transcendence Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup
values. Basic and Applied Social Psychology, 25, 299–312. theory of social hierarchy and oppression. Cambridge, MA:
Mikulincer, M., & Shaver, P. R. (2001). Attachment theory and Cambridge University Press.
intergroup bias: Evidence that priming the secure base schema Simpson, J. A., Collins, W. A., Tran, S., & Haydon, K. C. (2007).
attenuates negative reactions to out-groups. Journal of Personal- Attachment and the experience and expression of emotions in
ity and Social Psychology, 81, 97–115. adult romantic relationships: A developmental perspective.
Mikulincer, M., & Shaver, P. R. (2003). The attachment behavioral Journal of Personality and Social Psychology, 92, 355–367.
system in adulthood: Activation, psychodynamics, and interper- Spencer, S. J., Zanna, M. P., & Fong, G. T. (2005). Establishing a
sonal processes. In M. P. Zanna (Ed.), Advances in experimental causal chain: Why experiments are often more effective than
social psychology (Vol. 35, pp. 53–152). New York, NY: Aca- mediational analyses in examining psychological processes.
demic Press. Journal of Personality and Social Psychology, 89, 845–851.
Mikulincer, M., & Shaver, P. R. (2007). Attachment in adulthood: Sprecher, S., & Metts, S. (1989). Development of the ‘Romantic
Structure, dynamics, and change. New York, NY: Guilford Press. Beliefs Scale’ and examination of the effects of gender and
Montañés, P., de Lemus, S., Moya, M., Bohner, G., & Megı́as, J. L. gender-role orientation. Journal of Social and Personal Rela-
(2013). How attractive are sexist intimates to adolescents? tionships, 6, 387–411.
The influence of sexist beliefs and relationship experience. Weber, C., & Federico, C. M. (2007). Interpersonal attachment and
Psychology of Women Quarterly, 37, 494–506. patterns of ideological belief. Political Psychology, 28, 289–416.
Copyright of Psychology of Women Quarterly is the property of Sage Publications Inc. and
its content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

You might also like