Ambivalent Sexism Revisited

My how tempus fugits! It is close to 20 years since we began the collaboration that led to ambivalent sexism theory and its associated measure, both reviewed in our 1997 Psychology of Women Quarterly article, “Hostile and Benevolent Sexism: Measuring Ambivalent Sexist Attitudes Toward Women” (Glick & Fiske, 1997). Please find the original article at pwq.sagepub.com/content/21/1/119. In 1992, Peter, having recently survived a tenure review, was searching for a place to spend a year-long sabbatical. From his perspective, he had the good fortune (or, with the wisdom of hindsight, perhaps we can call it prescience?) to ask Susan whether it would be possible for him to visit for a year and collaborate. Neither of us can recall whether this initial contact occurred via e- mail (which, hard to fathom, was then a rather new thing!) or the regular post. What Peter does remember was a welcoming answer, which, as a faculty member at a tiny liberal arts college in the hinterland who possessed a then relatively meager publication record, seemed remarkably generous. (Embarrassingly the only prior contact the first author remembers was when Susan, as an action editor, had sent him an article to review. Unaccustomed to receiving such an honor, he breathlessly and profusely thanked her for choosing him as a reviewer. Alas, the thrill of being asked to review has since faded—but the message to fledgling reviewers remains—act like you have done it before!)

But Susan recalls another prior contact. She had come across Peter's earlier research on the gender and status dimensions of jobs (Glick, 1991) and occupational discrimination (Glick, Zion, & Nelson, 1988), which she had found extremely useful in expert testimony on gender biases in employment decisions. In those days, one often had to contact the author to obtain a reprint and that request may have constituted our first interaction, along with a subsequent bit of fan mail and gratitude from Susan to Peter for his useful and interesting work.

In any event, prior contact had been minimal; little did either of us know that we would soon forge such a productive, long-term collaboration and close friendship. But once Peter's sabbatical had been arranged, the planning for joint research began, and studying sexism was our most obvious link. While brainstorming, we noted that although sexism still seemed somehow more socially acceptable (or at least less politically incorrect) than racism, gender role attitudes had shifted so that the field lacked up-to-date indicators of sexist beliefs. Peter recalls Susan making the pivotal suggestion that we should create a contemporary sexism measure. The focus on ambivalence evolved partly from the insight that perceivers have an easier time justifying prejudices if they can affirm some subjectively positive (not just negative) beliefs about another group. We were also inspired by ambivalent racism work (e.g., Gaertner & Dovidio, 1986; Katz, Wackenhut, & Hass, 1986). But we soon realized that gender relations and, therefore, sexist attitudes, differ from race relations because men and women so often lead intimately intertwined lives, whereas Blacks and Whites typically experience much less contact.

Ambivalent sexism theory was incubated during a series of phone conversations, a scouting visit, and some pilot testing before Peter arrived at the University of Massachusetts at Amherst (then Susan's home institution) for the 1993–1994 academic year. We had already started work on our new, theory-based measure (a year was too short to start and complete such a project). Had we known that others had already begun constructing contemporary sexism scales (including Janet Swim, with whom Peter had gone to graduate school at the University of Minnesota, and Rupert Brown, whom Susan knew from trips to Europe), we might never have pursued the course we took. In this case, ignorance was an advantage. Noticing the same gap, we veered in a complementary direction to other measures that were simultaneously in development (e.g., Modern Sexism: Swim, Aiken, Hall, & Hunter, 1995; and Neosexism: Tougas, Brown, Beaton, & Joly, 1995). That several teams were working on new measures was no accident—each perceived a need. Psychology of Women Quarterly picked up on this zeitgeist—our article appeared in a special issue, edited by Irene Hanson Frieze and Maureen McHugh, on “measuring beliefs about appropriate roles for men and women” that showcased, compared, and contrasted both established and newer measures of gender- related attitudes (McHugh & Frieze, 1997).

What is So Special about Sexism?

Like the developers of the Modern Sexism and Neosexism scales, we took inspiration from contemporary racism research. In particular, racism theorists (e.g., Gaertner & Dovidio, 1986; McConahay, 1986) had focused on two key changes in racist attitudes in response to changing social norms. Specifically, racism had become (a) more subtle and (b) ambivalent. The first point certainly seemed to hold true for sexism. In U.S. samples, the instrument most frequently used to assess sexist attitudes, Janet Spence and Robert Helmreich's (1972) Attitudes Toward Women scale (AWS), had shown a steady decline in endorsement to the point that the variance on some items reflected differences in just how strongly people disagreed with them. These changes were documented by two articles in the same PWQ special issue as our piece, one by Spence and Hahn (1997) and the other by Twenge (1997; see also Twenge, 2011).

The AWS retained its usefulness in sexism research. It still predicted sex discrimination and its pioneering status made it uniquely situated to chart changes in American attitudes across decades during which unprecedented social change in women's roles had occurred. But, as noted above, a number of items on the AWS (e.g., “Sons in a family should be given more encouragement to go to college than daughters” or “It is ridiculous for a woman to run a locomotive and for a man to darn socks”) already seemed quaint. Although sexist attitudes were far from finished, the fact that the pioneering sexism scale developed in the early 1970s could seem “old-fashioned” by the early 1990s was astonishing. The field's measures could hardly keep pace with cultural developments, such as affirmative action and the backlash it evoked.

Although we had set out to construct a more subtle and contemporary scale, it was ambivalence, the other issue racism researchers had focused on, that became our persistent intellectual irritation, eventually producing a crucial pearl of wisdom. Theories about racist ambivalence were rooted in the “American dilemma,” which applied specifically to Whites’ treatment of Blacks. For example, aversive racism theory (Gaertner & Dovidio, 1986) proposed that well- meaning Whites now regularly tried to suppress and deny overlearned, negative stereotypes of Blacks. The Civil Rights movement had successfully created dissonance between American ideals and the undeniable history of brutal racism, slavery, and segregation in the United States. Sympathy for Blacks (motivated by egalitarian ideals and “white guilt”) created the “positive” pole of ambivalence toward Blacks. In short, racism theorists viewed racial ambivalence as a con- temporary phenomenon, a post-Civil-Rights-Era syndrome, in contrast to “old-fashioned racism,” which had been open, explicit, and unconflicted.

A similar analysis might seem to apply to sexism. Not only had gender role norms and attitudes clearly been changing, but by the late 1980s and early 1990s, Alice Eagly and Antonio Mladinic (Eagly & Mladinic, 1989, 1994) had demonstrated that American students (male as well as female) had more favorable stereotypes of women than men. Nobody was arguing that sexism (then defined primarily as hostility toward women) had disappeared, but the “women are wonderful effect” seemed to confirm that attitudes toward women had, like attitudes toward Blacks, transformed from hostile to ambivalent.

But was ambivalence toward women a recent phenomenon? Eagly and Steffen (1984) had already pointed out that positive attitudes toward women were rooted in women's role as nurturers. As these researchers noted, people think women are wonderful because of (not in spite of) traditional stereo- types about them. In contrast, theories of racism viewed traditional stereotypes as the primary cause of negative attitudes toward Blacks. Furthermore, the underpinnings of subjectively positive feelings toward Blacks (recognizing and sympathizing with their historical social disadvantages) did not seem to apply to men's attitudes toward women: “male guilt” over women's disadvantages did not seem to be that strong (if it existed at all).

To understand what differed about sexism, we focused on structural relationships between men and women. We started from the paradox that whereas gender relations (like so many other intergroup relations) traditionally involved dominance and subordination, members of the two groups also routinely have close romantic and familial relationships. In the work- place, men compete with women for resources and power; yet many of these same men (in part) devote their resources to providing for the women in their lives, primarily wives and daughters. This duality suggested that prejudice toward women did not fit the theoretical mold of prejudice as an unalloyed antipathy.

As an example, consider social distance measures of prejudice, which assume that a respondent's willingness to marry a person in another group indicates a person's lack of prejudice toward that group (Bogardus, 1927). A man's interest in marrying a woman does not necessarily suggest he rejects sexism. Indeed, gender-traditional men would likely be offended by the notion that they would marry any- one but a woman (showing an out-group, rather than an in- group, preference for their most intimate relationship). That a sexist might reject women at work yet fervently embrace them at home cuts to the core of how ambivalence toward women differs from racial ambivalence. Men are ambivalent about women because men genuinely love and like many women. Certainly, men's affection toward women did not ring as hollow as the White person's claim that “some of my best friends are Black.”

Developing and Testing the Theory and the Measure

The insights outlined above informed our approach to constructing a new sexism measure, the Ambivalent Sexism Inventory (ASI). We generated items based on our analysis of sexist ambivalence but also raided popular wisdom (e.g., books on “how to catch a man”) about gender relations. The subjectively positive scale items these conventional social beliefs helped us generate—similar to Eagly and Mladinic's (1989, 1994) “women are wonderful” findings—celebrated women's stereotypically traditional traits and roles, reinforcing our focus on men's and women's intimate interdependence as the source of the positive pole of ambivalence toward women. Overall, our subjectively positive items seemed just as archaic as any of the AWS's items. The only sense in which ours were more “subtle” was that they did not fit either the field's or popular society's definition of prejudice as a hostility. Despite their old-fashioned flavor, however, our subjectively positive items remained relevant in that they were endorsed by many respondents.

By contrast, when it came to items assessing hostile attitudes, we definitely confirmed that changing gender role norms required more subtle assessment of sexist attitudes. Our respondents had roundly dismissed items such as “Women are not as smart as men.” (One could practically hear them scoffing!) Recent social changes had indeed softened the edges of sexist hostility, or at least partially driven it underground. But we never doubted that sexist hostility had a long social history.

Our epiphany was that the subjectively positive side of sexist ambivalence had also been there all along—it was not a response to more egalitarian norms. Subjective benevolence toward women represented an old, not a new prejudice, and it probably had not changed much. Indeed, we informally noted that the benevolent items “sounded Victorian,” an observation that Stephanie Shields's (2007) historical analysis of complementarity in Victorian gender attitudes appears to confirm. These subjectively favorable attitudes did not require subtle assessment—unlike sexist hostility, such “positive” attitudes toward women elicited less social pressure toward suppression or change; instead of seeming pre- judicial, many people still viewed these attitudes as “nice” and “romantic.”

As the measure developed, so did the theory, each helping to bootstrap the other. We deepened our structural analysis of gender relations by more systematically considering the domains in which men and women compete (generating hostility) versus cooperate (generating paternalistic affection). The most crucial forms of interdependence (both competitive and cooperative) certainly had to include heterosexuality, gender roles, and male dominance in society. We specified hostile versions (sexuality as combat, competing gender roles, and women challenging male dominance) and benevolent versions (romantic intimacy, complementary gender roles, and women as cooperative subordinates) to include in the ASI's benevolent and hostile subscales.

In the discussion about what to call the measure, we briefly entertained the “Ambivalent Sexism scale” but rejected this label due to the unfortunate acronym it yielded; the ASI seemed safer (and perhaps more impressive than a mere “scale”). The subjectively negative (more conventionally “prejudiced”) subscale became the Hostile Sexism (HS) scale. Having already depleted our acronymic self-control, we could not resist the irony of naming the subjectively positive (more novel) subscale benevolent sexism (BS).

An Ambivalent Alliance

Our initial (and subsequent) research consistently showed that BS and HS correlated positively, often in the .4 to .5 range (Glick & Fiske, 1996, 1997). At first, this correlation seemed problematic, a potential threat to the validity of the scales. Prior theory and research on ambivalent attitudes suggested we should find either a negative or no correlation between subjectively positive and subjectively negative attitudes toward the same attitude object (i.e., women). The theoretical consensus characterized ambivalence as a mentally conflicted state, suggesting that people ought not simultaneously admit to or endorse both poles of their ambivalence (Thompson, Zanna, & Griffin, 1995). Rather, ambivalent individuals would, at any specific time, suppress one pole of their ambivalence while embracing and enacting the other, creating response polarization (Katz et al., 1986).

Contrary to the idea that one pole of ambivalence is suppressed when the other is activated, we thought that, in men's attitudes toward women, the very same men who said “can't live with them” were the ones who also said “can't live without them.” The correlation between “positive” and “negative” attitudes toward women, far from indicating that the subscales lacked validity, accurately reflected a key feature of sexist attitudes. Ambivalent sexists were not “men- tally conflicted”; rather, their subjectively positive and negative attitudes reflected complementary and mutually reinforcing ideologies. HS and BS were two sides of a sexist coin. And this double-sided coin (even if its hostile component had become more subtle) was at least as ancient as polarized stereotypes of the Madonna and Mary Magdalene. BS was the carrot aimed at enticing women to enact traditional roles and HS was the stick used to punish them when they resisted. One emphasizes reward and the other emphasizes punishment (hence their differing valences) but both work toward a common aim: maintaining a gender- traditional status quo.

Thus, benevolently sexist attitudes had not prevented men from behaving horribly toward women, including violent assault and murder. The protection and affection BS promises (and sometimes delivers) is readily withdrawn when women fail to conform to sexist expectations. But unlike most inter- group relations, in which “we” not only can but often prefer to live without “them,” heterosexual men truly could not imagine completely living without women. It is hardly a con- solation to female victims of male brutality that men's aim is not to eliminate all women. But it is nevertheless important to understand that men's hostility represents an attempt to control (rather than avoid completely or eliminate) women— intimidating them to keep them “in their place” and safely serving men's needs without challenging men's status, authority, and power (see Jackman, 1999, 2001). This helps to explain why women are often more at risk of violence from the men they love (i.e., with whom they are intimately interdependent) than from strangers.

The truth about sexism seems stranger than fiction, or at least stranger than any prior theory had suggested. BS does not represent sympathy for the underdog stemming from a contemporary sense of fair play. Rather it is a fundamentally antiegalitarian, gender-traditional attitude. Yet it is not only positively valenced, but intensely so, representing a genuine affection deeply rooted in a highly romanticized and intimate interdependence between the sexes. The label benevolent sexism represented our attempt to encapsulate the odd and jarring conjunction of what at first seemed inherently incompatible: subjective affection as a form of prejudice. This analysis, however, made perfect sense when considering BS as a paternalistic prejudice (see Jackman, 1994, for an incisive and comprehensive analysis of paternalism).

Subsequent Research and New Directions

Since our 1997 PWQ article, the ASI has been administered to tens of thousands of people around the world (Glick et al., 2000, 2004). When individuals represent the unit of analysis, correlations between HS and BS scores range from trivial to moderately positive, showing (as factor analyses consistently do) that HS and BS are distinct ideologies. But cross-cultural comparisons have confirmed that, as cultural ideologies, HS and BS go hand-in-hand: the correlation between national averages on HS and BS approaches .9. That is, if people in a society (on average) strongly endorse HS, they also strongly endorse BS and, conversely, if they reject one of these ideologies, they also (on the whole) reject the other. Supporting the idea that HS and BS represent complementary ideological supports for gender inequality, national averages on both scales predict national indices of structural gender inequality in power, resources, and outcomes (Glick et al., 2000, 2004).

In cross-cultural work, the scales have held up better than we could have hoped. The ASI's 2-factor structure (HS and BS) replicates in a diverse array of nations. Although cross-national comparisons hardly represent a time machine, the finding that the most gender-traditional cultures not only score high on HS but high on BS as well supports the notion that BS is not a contemporary, “politically correct” prejudice. We believe that the scales work well cross-culturally because ambivalent sexism theory and, consequently, the ASI have identified pervasive and fundamental aspects of gender relations—male dominance and heterosexual interdependence—that consistently foster, respectively, hostile and benevolent ideologies.

Research has also confirmed that BS presents unique problems for women, representing a barrier to equality not merely because of its complementary relationship to HS, but by producing its own insidious effects. Consider the following findings, all of which are unrelated to HS scores. Men who endorse BS are more likely to blame a female victim of an acquaintance rape if she has violated gender role expectations for feminine “purity” and chastity (Abrams, Viki, Masser, & Bohner, 2003). When women are led to expect BS in a workplace, they perform worse (Dardenne, Dumont, & Bollier, 2007). Women who endorse BS are more likely to accept a male romantic partner's (ostensibly protective) sexist restriction despite a potentially negative impact on their career aspirations (Moya, Glick, Expósito, De Lemus, & Hart, 2007), expect men to feel threatened and react violently in response to a wife's job promotion (Expósito, Herrera, Moya, & Glick, 2010), and prefer intimate male partners with high resources (Sibley & Overall, 2011). Finally, merely priming BS activates greater system justification among women, undermining their resistance to inequality (Jost & Kay, 2005); conversely, making sexist discrimination more salient to women can reduce their endorsement of benevolently sexist beliefs (Becker & Swim, 2011).

On the hostile side of the equation, Susan and her students have correlated HS with neuroimaging scans, showing that sexualized women are perceived more as “tools” than human beings (Cikara, Eberhardt, & Fiske, 2010). HS objectifies women, denying own human agency and reducing them to acted-on objects rather than autonomous actors. Peter, his colleagues, and students have also gone on to examine hostile reactions toward sexualized women within work contexts (Glick, Larsen, Johnson, & Branstiter, 2005) and toward apparently pregnant female job seekers (Hebl, King, Glick, Singletary, & Kazama, 2007).

Our structural analysis of sexism not only led to collaborative work using the ASI but also partially influenced our subsequent more general theory of prejudice, the Stereotype Content Model ([SCM]; Fiske, Cuddy, Glick, & Xu, 2002; Fiske, Cuddy, & Glick, 2007), developed along with Amy Cuddy. Although sexism represents a special case in the degree of intimate interdependence between the “in-group” and “out-group,” the SCM explores paternalism toward other groups viewed as low in status but cooperative with the rest of society (e.g., the elderly). Susan's student has pursued this kind of condescending pity toward older people, as well as what happens when they try to escape prescriptive age stereo- types (North & Fiske, under review).

The SCM also defined another type of ambivalence— envious prejudice toward groups viewed as high in status but competitive with the rest of society. This concept, in turn, has led us to rethink how HS, which, in its contemporary form tends to target powerful women, may best be characterized as an envious prejudice (whereas traditional sexist hostility, which viewed women as mentally inferior to men—was undoubtedly contemptuous). Additionally, Peter (Glick, 2002, 2008) has applied envious prejudice to scapegoating, challenging prior assumptions that scapegoating occurs mainly toward groups that are perceived as weak and vulnerable. His ideological model of scapegoating contends that prior envious stereotyping (which characterizes successful minorities as ill- intentioned and powerful) makes such groups likely targets of conspiracy theories when societies experience negative events (e.g., an economic meltdown). In a related vein, another of Susan's students has pursued how envious prejudice results in Schadenfreude, pleasure derived from the misfortune of others (Cikara & Fiske, in press).

Moving ahead, we look forward to continuing to see others make use of the ASI, so long as it remains a useful instrument for researching sexist attitudes. From a theoretical perspective, we hope that our analysis of the structural underpinnings of sexist attitudes inspires new ways of thinking about the complexities of gender relations, especially the perils of “positive” prejudice toward women, as well as the ambivalences involved in other prejudices.