How Sexy are Sexist Men? Women’s Perception of Male Response Profiles in the Ambivalent Sexism Inventory
In Studies 1 to 3, German female students (total N = 326) rated the likability and typicality of male targets: a nonsexist, a benevolent sexist, a hostile sexist, and (in Studies 2 and 3) an ambivalent sexist. When targets were presented as response profiles in the Ambivalent Sexism Inventory (Glick and Fiske 1996) (Studies 2 and 3), the benevolent sexist was rated to be most likable but least typical, whereas the ambivalent sexist was rated to be highly typical. Thus, women were aware of a link between benevolent and hostile sexism and approved of men’s benevolent sexism, especially when it was not paired with hostile sexism. Likability ratings were moderated by participants’ own benevolent sexism and feminist attitude.
KeywordsAmbivalent sexismAttractionFeminist attitudesGender relationsPerson perception
Relationships between men and women do not correspond well with classic notions of prejudice (e.g., Allport 1954), as cultural images of women have never been uniformly negative. Starting from this insight, ambivalent sexism theory (Glick and Fiske 1996) conceptualizes sexism as comprising two sets of beliefs with opposite evaluative implications, which have nonetheless been shown to often coexist in a perceiver’s mind. Whereas hostile sexism (HS) consists of negative beliefs and resentments against women (e.g., “Women are too easily offended”), benevolent sexism (BS) consists of beliefs that are subjectively positive in affective tone and tend to prescribe prosocial or intimacy-seeking behaviors (e.g., “In a disaster, women ought to be rescued before men”). To measure the two components of sexism, Glick and Fiske (1996) developed a self-report instrument, the Ambivalent Sexism Inventory (ASI). Numerous studies using the ASI showed that HS and BS are positively correlated both at the individual level (often around r = .40 or higher; e.g., Eckes 2001; Glick and Fiske 1996; Glick et al. 2000) and at the level of national averages for both men and women (r = .89; Glick et al. 2000). These findings support the notion that HS and BS form complementary parts of a sexist ideology.
Indeed, despite their positive appearance, benevolent sexist beliefs may contribute to the subordination of women in a subtle way. Research demonstrates that benevolent-sexist beliefs and acts may have particularly debilitating effects on women, e.g., by hampering women’s job performance (Dardenne et al. 2007), or predicting the extent to which students blamed the female victim of an acquaintance rape (Abrams et al. 2003). Detrimental effects of BS are suggested as well by a cross-cultural comparison: Across 19 countries, national levels of both HS and BS were inversely related to indices of gender equality (Glick et al. 2000). Several lines of research thus indicate that BS often produces damaging consequences, sometimes even more than HS (Dardenne et al. 2007).
As Glick and Fiske (2001) argued, women may be tempted to accept inequality in order to obtain the benefits that BS supposedly entails: “Benevolent sexism is disarming. ... It promises that men’s power will be used to women’s advantage, if only they can secure a high-status male protector” (p. 111). From this it follows that women may tend to like, and be attracted to, benevolent-sexist men. However, the exact cognitive processes underlying this attraction are not yet well understood. One possibility is that women are attracted to benevolent-sexist men without being aware that those men’s beliefs often go together with hostile-sexist attitudes. Another possibility is that women consciously accept a certain degree of hostility in men as long as it is counterbalanced by benevolence. And a final possibility is that women like particulary those benevolent-sexist men who do not also endorse hostile sexism. To study these possibilities, we assessed women’s perceptions of male targets differing in BS and HS, using a new methodology based on the ASI. Before describing our own research, we will discuss a previous study by Kilianski and Rudman (1998) which addressed the same topic.
Kilianski and Rudman (1998) hypothesized that women may fail to realize that benevolent and hostile sexist beliefs typically go together. They further argued that women, because of this failure to recognize BS and HS’ coexistence, may indirectly support HS in society by “wanting it both ways”, i.e. opposing HS but approving of BS. To examine this proposal, they presented female students with three male profiles, describing a hostile sexist, a benevolent sexist, and a nonsexist. For example, the hostile sexist was described as someone who “believes that women undervalue men and fail to appreciate everything that men do for them”, whereas the benevolent sexist was depicted as someone who “believes that women possess a naturally superior aesthetic sensibility which makes them better judges in matters of culture and taste” (pp. 349–350). When female students were asked to rate their impression of each profile on a scale from “strongly unfavorable” (0) to “strongly favorable” (6), they generally approved most strongly of the nonsexist (M = 4.36), but still expressed moderate liking for the benevolent sexist (M = 3.41), whereas they clearly disliked the hostile sexist (M = 1.03). When the students were later asked to judge the possibility that the hostile and benevolent profiles applied to the same person, they found it quite unlikely that this might be the case (M = 2.56 on a scale from 0, “very unlikely”, to 6, “very likely”). In addition, the more participants had preferred the benevolent sexist over the hostile sexist—a variable Kilianski and Rudman called “equivocal egalitarianism”—the lower they rated the likelihood that the two profiles may pertain to the same person (r = −.30). From these findings, Kilianski and Rudman concluded that women are not fully aware of the link between BS and HS, and that by responding positively to BS women may unwittingly support HS as well.
The Current Research
Our research follows up on the study by Kilianski and Rudman (1998). We will first point out two methodological ambiguities; then we will present new predictions and an extended, more balanced approach to the assessment of women’s perceptions of male sexism.
A first methodological problem that makes it difficult to interpret their findings is that Kilianski and Rudman (1998) did not present a fourth, ambivalent-sexist profile. If such a profile had been presented, participants could have been asked to rate how typical or realistic each of the four profiles was. If women were indeed unaware of a coexistence of hostile and benevolent sexist attitudes, they should have rated the purely hostile and purely benevolent profiles each as more typical than the ambivalent profile. Instead, Kilianski and Rudman only assessed participants’ ratings of the possibility that two distinct profiles, one described as hostile-sexist, the other as benevolent-sexist, “apply to the same person.” The two profiles, however, had been introduced explicitly as applying to different individuals, as designated by their initials. For a participant to respond that the two might actually be the same person would thus have meant to contradict previous experimental instructions, which had told her that she was judging two different men. Furthermore, just before making that “covariation estimate”, participants had indicated their approval of each profile. Especially those participants whose approval ratings for the benevolent and hostile profile greatly differed (i.e., the “equivocal egalitarians”) may have been motivated to answer consistently with their divergent approval ratings — they would thus be inclined to indicate that the two profiles are probably not describing the same person. Both experimental demand characteristics and a motivation to answer consistently may thus have contributed to the findings that the “covariation estimate” was rather low overall and was correlated negatively with the equivocal egalitarianism measure.
Another methodological problem is that Kilianski and Rudman’s (1998) hostile and benevolent profiles each contained information on only one of the two aspects of ambivalent sexism, thus allowing for different deductions about other characteristics of the target person: he could endorse the other aspect of ambivalent sexism, or he could reject the other aspect of sexism. This may have implications for ratings of the benevolent-sexist profile’s likability. If women indeed see BS and HS as covarying, then participants’ moderate likability ratings of the benevolent-sexist profile could be an underestimate of how they would rate a highly benevolent-sexist man if they were assured that he is not also a highly hostile-sexist man.
In our own Studies 2 and 3, we addressed these problems by correspondent changes to the methodology employed. First, we added a fourth male profile: that of an ambivalent sexist. Secondly, we provided information on the level of both BS and HS for each of the four profiles. We thus used a 2 (profile’s HS: low, high) x 2 (profile’s BS: low, high) within-subjects design, yielding profiles of a nonsexist (low HS, low BS), a benevolent sexist (low HS, high BS), a hostile sexist (high HS, low BS), and an ambivalent sexist (high HS, high BS). To keep the four profiles as parallel as possible, we did not use texts paraphrasing contents of the ASI, but instead presented men’s alleged answers to the ASI itself. Female participants thus saw ASI items along with the answers that each of four particular male respondents had ostensibly given. They were then asked to judge the four targets on scales assessing each target’s likability and typicality. With this extended methodology, participants’ awareness of whether benevolent and hostile sexist attitudes typically coexist in one person should be reflected in their relative typicality ratings: If participants believe that the two aspects of sexism are usually distinct, as hypothesized by Kilianski and Rudman (1998), then the ambivalent sexist should receive lower typicality ratings than both the benevolent sexist and the hostile sexist. If, however, participants believe that BS and HS usually go together, then the ambivalent sexist should receive higher typicality ratings than both the benevolent sexist and the hostile sexist. Finally, likability ratings for the benevolent sexist might be relatively higher than those found by Kilianski and Rudman because of the explicit pairing of high BS with low HS.
We set out to explore several hypotheses. In line with Kilianski and Rudman (1998), we assumed that women would generally disapprove of the openly negative prejudice against women that is expressed in men’s hostile sexism:
Women rate male profiles high in HS (i.e., the hostile sexist and the ambivalent sexist) as less likable than male profiles low in HS (i.e., the nonsexist and the benevolent sexist).
Furthermore, based on previous research on the perception of benevolent-sexist attitudes (Barreto and Ellemers 2005; Swim et al. 2005) we predicted that women would generally approve of men’s more benevolent prejudices toward women:
Women rate male profiles high in BS (i.e., the benevolent sexist and the ambivalent sexist) as more likable than male profiles low in BS (i.e., the nonsexist and the hostile sexist).
Taken together, the independent main effects of HS and BS implied by Hypotheses 1 and 2, respectively, should yield the highest likability ratings for the benevolent sexist and the lowest likability ratings for the hostile sexist, with both the nonsexist and the ambivalent sexist falling in between. Thus, when assured that a man’s BS is not accompanied by HS, women will rate his profile as highly attractive, even as compared to a nonsexist man. Kilianski and Rudman’s (1998) findings, by contrast, would suggest an interaction effect of the profiles’ hostility and benevolence, whereby the nonsexist profile, which is low in both hostility and benevolence, would be rated as the most likable.
With respect to the typicality ratings, we based our predictions on the fact that hostile and benevolent sexism are typically positively related (Glick et al. 2000). In daily life, women should therefore interact more frequently with ambivalent sexists than with either purely benevolent sexists or purely hostile sexists. The ambivalent sexist profile should thus appear highly realistic to women.
Women rate the ambivalent sexist profile as more typical than both the benevolent and the hostile sexist profile.
Again, an opposite prediction would follow from Kilianski and Rudman (1998). If women were unaware of the connection between hostile and benevolent sexism, then the profiles that are high in just one component of sexism (i.e., the benevolent and the hostile sexist) should be rated as more typical than the ambivalent sexist profile.
We further assumed that participants’ likability ratings would be influenced by their own level of BS and HS (see Kilianski and Rudman 1998). We predicted a special case of an attitude similarity effect: People generally like others who share their attitudes (e.g., Byrne 1971; Newcomb 1961). Furthermore, recent research dealing with women’s reactions to BS has shown that women high (vs. low) in BS were more likely to accept restrictions imposed on them by a man that were justified with benevolent sexist motives (Moya et al. 2007). In Studies 2 and 3 we therefore assessed participants’ own BS and HS and performed multiple regression analyses to predict likability differences among the profiles from these variables. We made the following predictions.
The lower women’s HS, the greater is their preference in likability ratings for low-HS male profiles over high-HS male profiles.
The higher women’s BS, the greater is their preference in likability ratings for high-BS male profiles over low-BS male profiles.
We further hypothesized that participants with high-feminist attitudes, who should be more sensitive to issues of gender equality (Morgan 1996), would recognize the harmful effects of both hostile sexim and benevolent sexism exhibited by men. Feminist women should therefore show an even greater preference for low-hostile sexist men, but less of a preference for high-benevolent sexist men. In Study 2, we assessed women’s feminist attitudes and used them as an additional predictor in regression analyses:
The higher women’s feminist attitudes, the greater is their preference in likability ratings for low-HS male profiles over high-HS male profiles.
The higher women’s feminist attitudes, the lesser is their preference in likability ratings for high-BS male profiles over low-BS male profiles.
In each study we also employed measures of equivocal egalitarianism, defined as the difference in likability ratings between the benevolent and hostile sexist profiles (see Kilianski and Rudman 1998). Specifically, we tested whether equivocal egalitarianism would predict typicality ratings of the ambivalent sexist profile. A negative correlation (Hypothesis 7) would conceptually replicate Kilianski and Rudman’s findings.
Finally, in Study 3 we assessed broader aspects of likability, including items assessing each targets’ short-term mate value and sexual attractiveness. This was done to explore whether men high in benevolent sexism might be seen not only as nice and likable, but also as “sexy” and of high mate value. The latter would be in line with traditional views of romance (e.g., the “knight in shining armor”) and with research showing that benevolent sexist attitudes are related to notions of paternalistic chivalry (Viki et al. 2003). For the mate value and sexual attractiveness items we entertained the same hypotheses as for likability in general.
Consideration of Cultural Differences between Germany and the USA
Before testing our hypotheses with the proposed new methodology, we examine if any differences in results this methodology might produce could be due to cultural differences between Germany and the USA. Based on relevant national indicators and previous sexism research, one should not expect large cultural differences. Germany and the USA are highly industrialized, Western societies, and both are doing relatively well in terms of the United Nations’ two main measures of gender equality. On the gender empowerment measure, which reflects women’s (relative to men’s) political and economic participation and power over resources, Germany’s most recent score is .852 (rank 8 of 108 nations surveyed), and the United States’ score is .769 (rank 18). On the gender development index, a gender-related version of the human development index reflecting economic wealth, life expectancy, and educational level, the two countries’ scores are almost identical at .937 (Germany: rank 21 of 157; USA: rank 19) (United Nations 2008).
In Glick et al.’s (2000) cross-cultural comparison of ASI scores in 19 countries, the German and US samples also showed highly similar results. Their factor structures were parallel, and the mean scores for HS and BS were similarly low, occupying neighboring ranks (15/16 and 13/14, respectively). Positive correlations between HS and BS at the individual level were also present in both countries, although they were somewhat lower in Germany (men: r = .25; women: r = .31) than in the USA (r = .44 for both men and women). If the magnitude of these correlations affects women’s perception of the likelihood that both BS and HS may be present in the same person, then the typicality of ambivalent sexism should be rated as being lower by German respondents than by US respondents. This would make it less likely that our Hypothesis 3 would be supported.
Overall, then, cultural differences in women’s perception of male sexism seemed unlikely. To provide more direct evidence for this prediction, we conducted an initial study in which we closely replicated the relevant parts of Kilianski and Rudman’s (1998) procedures with a German sample. If their results were due to aspects of the procedure, as discussed above, then we should find similar results in this direct replication study.
Participants and Procedure
One hundred and twelve female undergraduates at the University of Bielefeld (Germany) completed a questionnaire in a large group session during regular class hours. Participants’ age ranged from 19 to 41 years (M = 22.63, SD = 4.31). Where applicable, participants received partial course credit.
The questionnaire contained (1) the 22-item ASI (Glick and Fiske 1996; German version by Eckes and Six-Materna 1999) with its subscales HS (Cronbach's alpha=.87) and BS (alpha = .90); (2) German translations of the three male profiles used by Kilianski and Rudman (1998, pp. 349–350), each followed by a single-item favorability rating (0=strongly unfavorable to 6=strongly favorable); and (3) a single-item covariation estimate, where participants indicated the possibility that the hostile and benevolent profiles applied to the same person (0=very unlikely to 6=very likely).
Results and Discussion
The pattern of means for the favorability ratings closely replicated Kilianski and Rudman (1998). The nonsexist was evaluated most favorably (M = 4.42; SD = 1.16), followed by the benevolent sexist (M = 2.94; SD = 1.42) and the hostile sexist (M = 1.44; SD = 1.13). A repeated-measures ANOVA (MSE = 1.58) showed a significant effect of profile, F(2, 220) = 156.08, and paired-samples t-tests revealed that all pairwise comparisons were significant, all t > 8.00, all p < .001.
Participants’ ratings of the possibility that the hostile and benevolent profiles applied to the same person were low (M = 1.96; SD = 1.57) and significantly different from the scale midpoint of 3, one-sample t(111) = 7.00, p < .001. The mean obtained in our German sample was even significantly lower than that obtained by Kilianski and Rudman (M = 2.56; SD = 1.99), t(210) = 2.41, p < .02. Thus, participants generally indicated that the hostile and benevolent profiles were probably not describing the same person.
HS and BS scores were calculated by averaging across the respective subscale items. The mean for HS was 1.94 (SD = .86), that of BS was 2.36 (SD = .99). An analysis of partial correlations showed that participants’ HS scores (controlling for BS) predicted favorability ratings of the hostile profile, r(108) = .38, p < .001, but not of the benevolent profile, r(108) = .15, p > .11, whereas participants’ BS scores (controlling for HS) predicted favorability ratings of the benevolent profile, r = .35, p < .001, but not of the hostile profile, r(108) = −.16, p > .09. These partial correlations are similar in magnitude to those reported by Kilianski and Rudman (1998).
A continuous measure of equivocal egalitarianism was calculated by subtracting the favorability rating of the hostile profile from that of the benevolent profile. This measure was positive overall (M = 1.52; SD = 1.71). Equivocal egalitarianism was unrelated to participants’ HS (r = .07, p = .48) but positively related to participants’ BS (r = .35, p < .001); most importantly, replicating Kilianski and Rudman’s results, equivocal egalitarianism showed a moderate negative correlation with the covariation estimate (r = −.32, p = .001).
In sum, the relevant patterns of means and correlations in Study 1 were highly similar to those obtained by Kilianski and Rudman (1998) in the USA several years earlier. Ratings of the nonsexist were most favorable, those of the hostile sexist were least favorable, and those of the benevolent sexist fell in between. With respect to the covariation estimate, German undergraduates were even less likely than their US counterparts to rate the hostile and benevolent profiles as describing the same person. Replicating Kilianski and Rudman, lower covariation estimates went along with higher levels of equivocal egalitarianism.
Applying the same methodology in Germany thus yielded almost identical results. Given this high degree of correspondence between our data and those of Kilianski and Rudman (1998), it would be quite implausible to explain high typicality ratings of an ambivalent sexist profile (compared to purely benevolent or hostile profiles), and high likability ratings of a benevolent sexist profile (compared to all other profiles), which we predicted for our two subsequent studies, by cultural differences.
In our second study, we set out to test Hypotheses 1 to 7 as described in the Introduction. We applied the new methodology of presenting four different ASI response profiles and assessing each profile’s likability and typicality. To keep the information load manageable for participants, only ten ASI items (5 HS, 5 BS) were used to construct each target profile. The remaining 12 items (6 HS, 6 BS) were used to assess participants’ own HS and BS, respectively. By avoiding any overlap between the profiles and items that the participants answered, we tried to minimize potential effects of superficial similarity, which might result from participants’ rating as more likable those targets that gave similar answers as they had given themselves on certain items.
One hundred and eight female undergraduates at the University of Bielefeld (Germany) volunteered to participate. Participants’ age ranged from 19 to 41 years (M = 23.25, SD = 4.17). Each participant received a free cinema ticket or partial course credit.
Participants were seated at separate tables in a laboratory. They first completed a questionnaire containing a 12-item short form of the ASI and a 10-item scale measuring feminist attitudes. Then they received a sheet with four 10-item ASI profiles that had ostensibly been completed by four different men, and were asked to answer several questions pertaining to each of these men’s likability and typicality. Later, they indicated their age and subject of study. Then they were thoroughly debriefed.
ASI 12-item short form
To assess participants’ HS and BS, we generated two 12-item versions of the German ASI (Eckes and Six-Materna 1999). Each version contained six items assessing HS and 6 items assessing BS. Version A of the participant questionnaire contained items 2, 3, 6, 7, 8, 10, 11, 13, 15, 18, 19, and 20; version B contained items 1, 4, 5, 8, 9, 11, 12, 14, 16, 17, 21, and 22 (for item wording, see Appendix and Glick and Fiske 1996; note that items eight and 11 appeared in both versions). Each item was to be answered on a response scale from 1=do not agree at all to 6=completely agree. Internal consistencies were somewhat lower than the values reported for the full scale. In version A, Cronbach’s alpha was .70 and .55 for HS and BS, respectively; in version B, the respective values were .74 and .65.
To assess participants’ feminist attitudes, we used a German adaptation of an 11-item short form of the Liberal Feminist Attitude and Ideology Scale (LFAIS; see Morgan 1996, p. 380). Some US-specific item contents were adapted to the German context, and item eight was omitted, which yielded a ten-item scale. An item example was “A woman should have the same job opportunities as a man” (1=do not agree at all to 6=completely agree). The scale’s internal consistency was satisfactory (Cronbach’s alpha=.71).
ASI profiles of four male target persons were constructed using those ten ASI items (5 HS and 5 BS) that participants had not answered themselves, yielding two versions (A and B; see Appendix) that were randomly assigned. The four profiles were presented together on a large (29.7 cm × 42.0 cm) fold-out sheet, two at the top and two at the bottom. The ambivalent sexist, hostile sexist, nonsexist, and benevolent sexist profiles were labeled with different letters, here designated, for ease of presentation, as “A”, “H”, “N”, and “B”, respectively. To control for order effects, four versions were used according to a Latin square design (AHNB, HNBA, NBAH, and BAHN). At the top of each profile, a response scale with six options was shown: “−3=do not agree at all, −2=do not agree, −1=tend not to agree, + 1=tend to agree, + 2=agree, + 3=agree completely”. A handwritten number appeared in a space to the left of each item; each profile was prepared in a different handwriting to suggest that sheets had been photocopied from actual men’s questionnaire responses. Profile A showed agreement (all positive numbers) with both HS and BS items; profile H showed agreement with HS items but disagreement (all negative numbers) with BS items; profile B showed agreement with BS items but disagreement with HS items; and profile N showed disagreement with both HS and BS items. The profiles’ HS and BS means in versions A and B differed slightly (.2 scale points at most) because the alleged answers were based on previous responses of a male student sample such that high and low benevolence (or hostility) were represented by numbers that were one SD above or below the empirical item mean, respectively, rounded to the nearest integer. Participants were instructed to take a few minutes to form an impression of each target by imagining how this person would behave in everyday situations.
Participants then answered several items pertaining to each target. These were printed on a separate page for each target, with targets appearing in the same order as on the profile sheet. Participants were allowed to look at all four targets while answering.
Two items were used to check whether the target profiles’ benevolence and hostility, respectively, were perceived as intended (blanks were always filled with the letter denoting the profile): “How benevolent is person __’s attitude toward women?” (1=not at all benevolent to 6=very benevolent); and “How hostile is person __’s attitude toward women?” (1=not at all hostile to 6=very hostile).
Two items were used to assess the target’s likability: “How likable do you rate person __?” (1=very dislikable to 6=highly likable); and “Would you consider having someone like person __ as a romantic partner?” (1=certainly not to 6=certainly yes). The two items correlated highly across profiles (r = .56 to .83) and were averaged to form a likability index.
Three items were used to assess the target’s typicality: “What do you think, how often do men answer in a similar way as person __?” (1=very rarely to 6=very often); “How typical, do you think, are the answers given by person __?” (1=very atypical to 6=very typical); and “In my opinion, person ___ is ...” (1=definitely fictitious to 6=definitely real). The last item was presented on a separate page, after participants had been informed that some or all of the profiles might be fictitious. The three items showed moderate to high internal consistency across profiles (alpha=.57 to .77) and were averaged to form a typicality index.
To ascertain that the likability and typicality items measured separate constructs, we performed a factor analysis for each profile, entering the two likability and the three typicality items. We used maximum-likelihood extraction with promax rotation to allow for correlated factors (see Fabrigar et al. 1999). Results showed that the typicality items and the likability items consistently loaded on separate factors, which were not highly intercorrelated (r ranging from −.04 to + .33).
Results and Discussion
Participants’ ratings of the profiles’ benevolence and hostility, respectively, were analyzed with 2 (profile’s HS: high, low) x 2 (profile’s BS: high, low) repeated-measures ANOVAs. These revealed that, as intended, participants rated the high-hostile profiles as higher on hostility (M = 3.63) than the low-hostile profiles (M = 2.20), F(1, 106) = 246.12, p < .001, and rated the high-benevolent profiles as higher on benevolence (M = 4.56) than the low-benevolent profiles (M = 3.13), F(1, 107) = 188.09, p < .001. In addition, the high-hostile profiles were rated as lower in benevolence than the low-hostile profiles, F(1, 107) = 238.10, p < .001, and the high-benevolent profiles were rated as lower in hostility than the low-benevolent profiles, F(1, 106) = 124.30, p < .001. No interaction effects emerged, both p > .22.
Likability and Typicality Differences Among Profiles
Ratings of likability and typicality for four male target profiles (Study 2).
As predicted in Hypotheses 1 and 2, female students rated the benevolent sexist profile as the most likable (M = 4.62; SD = .95) and the hostile sexist profile as the least likable (M = 2.23; SD = .89); ratings for the nonsexist (M = 3.30; SD = 1.18) and ambivalent sexist profile (M = 3.53; SD = 1.05) fell in between (see Table 1). The ANOVA showed a strong main effect of the profiles’ hostility, F(1, 107) = 110.83, p < .001, supporting Hypothesis 1. The ANOVA also showed a strong main effect of the profiles’ benevolence, F(1, 107) = 123.22, p < .001, supporting Hypothesis 2. There was no interaction effect, F < 1. The female students thus rated as more favorable the profiles low (vs. high) in hostility and the profiles high (vs. low) in benevolence. All pairwise comparisons among means, except for the comparison between the nonsexist and the ambivalent sexist profile (p > .19), were highly significant (p < .001).
In line with Hypothesis 3, female students rated the ambivalent sexist profile as highly typical (M = 4.07; SD = .88), followed by the hostile sexist profile (M = 3.81; SD = .69) and the nonsexist profile (M = 3.13; SD = .84). The benevolent sexist profile was rated to be the least typical of actual men (M = 3.06; SD = .83). In the ANOVA, this pattern was reflected in a main effect of the profiles’ hostility, F(1,107) = 82.94, p < .001, and an interaction effect of the profiles’ hostility and benevolence, F(1,107) = 4.66, p < .04, but no main effect of the profiles’ benevolence, F(1,107) = 1.66, p > .20. Participants thus perceived profiles high in HS to be more typical of actual men than profiles low in HS. Furthermore, they rated the profile combining high hostility and high benevolence (the ambivalent sexist) as particularly typical for actual men. All pairwise comparisons, except for the comparison between the nonsexist and the benevolent sexist profile (p > .54), were significant (p < .02).
Effects of Participants’ Sexism and Feminist Attitudes
To test Hypotheses 4 to 6, which predicted moderating effects of participants' own BS and HS, as well as their feminist attitudes, on their relative preferences for low-HS and high-BS profiles, respectively, we conducted two separate multiple regression analyses. The dependent variable in the first analysis was participants’ relative preference for low-HS profiles over high-HS profiles; the dependent variable in the second analysis was participants’ relative preference for high-BS profiles over low-BS profiles. Because of the within-subjects variation of the profiles’ BS and HS, we could define each of these preferences as a difference score in likability ratings. Note that these scores represent the main effects of profiles’ HS and BS, respectively, that were found in the ANOVAs, such that the regression analyses test moderating effects of participants’ attitudes on these ANOVA findings. Participants’ HS (M = 3.12; SD = .72), BS (M = 3.56; SD = .79), and LFAIS scores (M = 4.53; SD = .51) were entered as concurrent predictor variables to control for their covariation and test their unique contributions. The correlation of HS and BS was r(105) = .37, p < .001. Feminist attitudes correlated negatively with both HS, r(105) = −.36, p < .001, and BS, r(105) = −.21, p = .03.
Results of multiple regression analyses predicting relative preferences in likability for low- (vs. high-) HS profiles and high- (vs. low-) BS profiles from participants’ hostile sexism, benevolent sexism, and feminist attitudes (Study 2).
Relative Preference for Low-HS
vs. High-HS Profiles
Relative Preference for High-BS
vs. Low-BS Profiles
Equivocal Egalitarianism and Perceived Typicality of the Ambivalent Sexist Profile
We computed a continuous measure of equivocal egalitarianism by subtracting the likability index for the hostile sexist from the likability index for the benevolent sexist (see Kilianski and Rudman 1998). Higher scores on this measure represent relatively greater amounts of equivocation. If, as proposed by Kilianski and Rudman and stated in Hypothesis 7, women with equivocal egalitarian attitudes fail to perceive a link between BS and HS, then scores on the equivocal egalitarianism measure should be negatively related to ratings of the ambivalent sexist profile’s typicality. A correlation analysis showed, however, that this was not the case: Equivocal egalitarianism was uncorrelated with the typicality index of the ambivalent sexist profile, r(106) = .10, p = .30. Thus, Hypothesis 7, which was based on findings by Kilianski and Rudman that had replicated in Study 1 using their original methodology, was not supported when women had the opportunity to rate an ambivalent sexist target directly.
Female students rated the purely benevolent sexist as the most likable of the ASI profiles and the purely hostile sexist as the least likable. The ambivalent sexist and the nonsexist were judged to be more likable than the hostile sexist but less likable than the benevolent sexist; they did not differ from each other in likability. Female participants’ likability ratings thus reflect independent effects of the profiles’ levels of HS and BS; as predicted in Hypotheses 1 and 2, male targets’ high BS and low HS each led to greater liking. Female students’ judgments of the profiles’ typicality were affected by the profiles’ level of HS, with high-HS profiles (H and A) being judged as more typical than low-HS profiles (N and B). Importantly, the typicality ratings also showed an interaction effect of the profiles’ HS and BS, such that the ambivalent sexist profile was judged to be most typical of men in general, whereas the purely benevolent profile was judged to be least typical. Hypothesis 3, which had predicted that the ambivalent sexist would be rated as more typical than both the benevolent and the hostile sexist, was clearly supported. Furthermore, contrary to Hypothesis 7, typicality ratings of the ambivalent sexist profile were unrelated to participants’ equivocal egalitarianism.
These results suggest conclusions that differ from those drawn by Kilianski and Rudman (1998). These authors had inferred that women may be unaware of the link between HS and BS. We argued, by contrast, that an unbiased test of women’s awareness of such a link would require a direct comparison between nonsexist, benevolent, hostile, and ambivalent profiles of men. In Study 2, where women had an opportunity to compare and rate all four profiles, they seemed to be well aware that the ambivalent sexist profile represents a typical male. Also, our female participants seemed to be aware of the fact that purely benevolent sexist men are quite atypical in contemporary society.
Nonetheless, our participants judged the benevolent sexist profile to be the most likable of all profiles, and the ambivalent sexist profile to be no less likable than the nonsexist profile. These findings suggest that women, in spite of their being aware of the connection between BS and HS, still fail to resist and criticize BS exhibited by men. On the contrary, when presented with a man who—atypically—exhibits high BS accompanied by low HS, they find this man more attractive than a nonsexist who exhibits neither high BS nor high HS. This is a remarkable finding which we will follow up on in Study 3, using a wider range of attraction measures.
Contrary to previous findings (Kilianski and Rudman 1998; the present Study 1) and to Hypothesis 4, participants’ own level of HS did not moderate their relative preference for low-HS profiles. Consistent with previous research and with Hypothesis 5, however, participants’ level of BS did moderate their relative preference for high-BS profiles. Women who were themselves high (vs. low) in BS thus preferred benevolent and ambivalent sexist male targets even more over nonsexist and hostile targets. Apparently, women’s own endorsement of benevolent sexist attitudes poses an obstacle against recognizing men’s benevolent sexist attitudes as harmful. That women’s BS had a stronger moderating influence on their evaluation of the respective attributes in men than did women’s HS may be due to the greater overall effect of profiles’ BS (vs. HS) on likability ratings.
The moderating effects of participants’ feminist attitudes on relative preferences for low-HS and high-BS profiles, respectively, only partly supported our predictions. In line with Hypothesis 6(a), high-feminist women showed an even greater relative preference for low-hostile men than did low-feminist women, but contrary to Hypothesis 6(b), high-feminist women were just as likely to prefer high-benevolent men over low-benevolent men as were low-feminist women. Holding feminist attitudes thus seems to intensify women’s rejection of hostile sexism, although it may contribute little to recognizing that men’s benevolent sexist attitudes can be just as harmful. However, the short LFAIS (Morgan 1996), which we used to measure feminism, mainly assesses attitudes that oppose blatant sexist prejudice and open discrimination; a desideratum for the future development of feminism scales would thus be to include items that also capture women’s awareness of more subtle aspects of sexism, as expressed in men’s benevolent sexist attitudes and behavior.
Although these initial results based on judgments of four ASI profiles are promising, we note a potential methodological limitation. In order to make agreement and disagreement with items unambiguously visible in the target profiles, the nonsexist profile featured a pattern of clear disagreement with all items. Participants might inadvertently have used this consistent disagreement as a heuristic cue in judging the nonsexist target, which may have reduced that target's perceived likability ("someone who disagrees with everything cannot be very likable").
To provide further tests of Hypotheses 1 to 7, as described in the Introduction, we designed a conceptual replication of Study 2. The methodology was slightly changed in two ways. First, in order to counteract any heuristic-based rejection of the nonsexist profile, all profiles now featured neutral filler items with which the target had ostensibly agreed. Secondly, the range of items measuring likability was extended; specifically, more items representing a man’s mate value and sexual attractiveness were included. These additions were exploratory, as we did not formulate different hypotheses for different kinds of likability items. Instead, we aimed at exploring whether increased liking for benevolent sexist men would generalize to perceptions of the target as a potential short-term mate or dating partner. This would complement work showing positive correlations between benevolent sexism and notions of traditional courtship behavior (Viki et al., 2003).
One hundred and six female students at the University of Bielefeld (Germany) volunteered to participate in a series of shorter studies that were ostensibly conducted together for reasons of economy. Participants’ age ranged from 19 to 47 years (M = 22.87, SD = 3.79). Each participant received a chocolate bar and was entered in a lottery for a 50-Euro gift voucher.
Participants were seated at separate tables in a laboratory. They first completed a 12-item short version of the ASI. Then they received a fold-out sheet with the four ASI profiles, and a questionnaire with items pertaining to each profiles’ likability, sexual attractiveness, and typicality. On a final sheet, participants indicated their age and subject of study, and were allowed to comment on the study. After completing all materials, participants were debriefed.
ASI 12-item short form
The short forms used to assess participants’ HS and BS were identical to those used in Study 1. Again, two versions were used; in version A, Cronbach’s alpha was .79 and .72 for HS and BS, respectively; in version B, the respective values were .67 and .66.
Each of the four male ASI profiles contained the same items as in Study 1 plus five filler items of neutral content (e.g., “The Internet offers the opportunity for women and men to communicate with people all over the world”), which were interspersed with the ASI items. The fillers always appeared in positions 2, 4, 5, 8, and 11, and the answers marked always indicated agreement (twice + 1, once + 2, and twice + 3). Again, four presentation orders of the profiles were used according to a Latin square design.
The same items as in Study 2 were used to check whether the target profiles’ benevolence and hostility, respectively, were perceived as intended.
Likability, sexual attraction, and typicality
Six items pertained to the target’s likability and sexual attraction: “How likable do you rate person __?” (1=very dislikable to 6=highly likable); “Would it be conceivable for you to have someone like person __ as a long-term romantic partner?” (1=definitely not conceivable to 6=well conceivable); “How good-looking do you think is person __?” (1=not good-looking at all to 6=very good-looking); “Would it be conceivable for you to have a one-night stand with person __?” (1=definitely not conceivable to 6=well conceivable); “Do you think person __ is a passionate man?” (1=not at all passionate to 6=very passionate); “How attractive do you rate person __?” (1=not at all attractive to 6=very attractive).
The three items pertaining to the target’s typicality were the same as in Study 2. A minor change was made to the response scale of the third item; it now read “In my opinion, person ___ is ...” (1=definitely not real to 6=definitely real).
Forming indices based on factor analysis
For each profile, we performed maximum-likelihood factor analyses with promax rotation on the likability, sexual attraction, and typicality items, allowing for correlated factors (see Fabrigar et al., 1999). These analyses showed that the three typicality items consistently loaded on a separate factor across the four profiles. They were thus combined to form an index of typicality (Cronbach’s alpha across profiles=.66 to .77).
The likability and sexual attraction items loaded on either two factors (for profiles A and B) or a single factor (for profiles N and H). For exploratory reasons, we forced a solution with two factors for all profiles and found that one factor was consistently defined by the items “likable” and “long-term partner”; as in Study 2, these items were thus averaged to form an index of likability (intercorrelation of items across profiles=.64 to .78). The other four items (“good-looking”, “one-night stand”, “passionate”, and “attractive”) showed substantial loadings on a separate factor; they were thus averaged to form an index that we called short-term sexual attraction (STSA; Cronbach’s alpha across profiles = .64 to .75). It should be noted, however, that the intercorrelations of the likability and STSA factors were high across profiles (r = .48 to .66).
Results and Discussion
Participants’ ratings of the profiles’ benevolence and hostility, respectively, were analyzed with 2 (profile’s HS: high, low) x 2 (profile’s BS: high, low) repeated-measures ANOVAs. As intended, participants rated the high-hostile profiles as higher on hostility (M = 3.57) than the low-hostile profiles (M = 2.32), F(1, 104) = 153.51, p < .001, and rated the high-benevolent profiles as higher on benevolence (M = 4.53) than the low-benevolent profiles (M = 3.08), F(1, 105) = 271.51, p < .001. In addition, the high-hostile profiles were rated as lower in benevolence than the low-hostile profiles, F(1, 105) = 114.62, p < .001, and the high-benevolent profiles were rated as lower in hostility than the low-benevolent profiles, F(1, 104) = 105.86, p < .001. No interaction effects emerged, both p > .27.
Likability, STSA, and Typicality Differences Among Profiles
Ratings of likability, short-term sexual attraction, and typicality for four male target profiles (Study 3).
Recall that Hypotheses 1 and 2 predicted higher likability ratings for low-HS and high-BS profiles, respectively. In line with these predictions, and replicating the findings of Study 2, female students rated the benevolent sexist profile as the most likable (M = 4.57; SD = .92) and the hostile sexist profile as the least likable (M = 2.33; SD = .85). Ratings for the nonsexist (M = 3.35; SD = 1.06) and ambivalent sexist profile (M = 3.67; SD = 1.07) again fell in between (see Table 3). The ANOVA yielded a strong main effect of the profiles’ hostility, F(1,105) = 102.99, p < .001, supporting Hypothesis 1. It also yielded a strong main effect of the profiles’ benevolence, F(1,105) = 150.91, p < .001, supporting Hypothesis 2. The interaction effect was not significant, F < 1. The students thus again rated as more likable the profiles low (vs. high) in hostility and high (vs. low) in benevolence. All pairwise comparisons among means were significant (all p < .04).
Short-Term Sexual Attraction
Although, not surprisingly, its means were generally lower, the STSA index showed a pattern very similar to that of likability. Female students rated the benevolent sexist profile as most sexually attractive (M = 3.52; SD = .85) and the hostile sexist profile as least sexually attractive (M = 2.50; SD = .78). Ratings for the nonsexist (M = 3.11; SD = .94) and ambivalent sexist profile (M = 3.21; SD = .88) fell in between (see Table 3). Main effects of both the profiles’ hostility, F(1,105) = 29.88, p < .001, and the profiles’ benevolence, F(1,105) = 45.00, p < .001, supported Hypotheses 1 and 2: The high-BS (vs. low-BS) profiles and the low-HS (vs. high-HS) profiles were rated as higher in sexual attractiveness (see Table 3). An interaction of profiles’ BS and profiles’ HS also emerged, reflecting that the hostile sexist profile (low BS / high HS) was rated as particularly unattractive as a short-term mate, F(1, 105) = 4.42, p = .04. Pairwise comparisons among means were significant (all p < .01) except for the comparison between the nonsexist and the ambivalent sexist (p > .41).
Recall that Hypothesis 3 predicted higher typicality ratings for the ambivalent-sexist profile than for both the benevolent-sexist and the hostile-sexist profiles. As in Study 2, the benevolent sexist profile was judged to be the least typical (M = 3.13; SD = .91). The nonsexist profile (M = 3.65; SD = .68), the hostile sexist profile (M = 3.83; SD = .77), and the ambivalent sexist profile (M = 3.68; SD = .80) were all rated higher in typicality than the benevolent sexist profile, but did not differ significantly from one another (see Table 3). Hypothesis 3 was thus only partially supported, as the female students rated the ambivalent profile as more typical than the benevolent profile, but not as more typical than the hostile profile. The ANOVA revealed main effects of both the profiles’ hostility, F(1,105) = 17.48, p < .001, and the profiles’ benevolence, F(1,105) = 14.21, p < .001, as well as an interaction effect of the two, F(1,105) = 7.14, p = .009. As in Study 2, then, female students found profiles high in HS to be more typical for actual men than profiles low in HS. Whereas they rated the two high-hostile profiles (H and A) as comparably high in typicality, they rated the profile combining low hostility and high benevolence (B) to be the least typical.
Effects of Participants’ Sexism
The means of participants’ HS and BS scores were 3.19 (SD = .81) and 3.78 (SD = .88), respectively, and their intercorrelation was r(104) = .37, p < .001. In multiple regression analyses, we tested if women’s own sexist attitudes influenced their relative preferences for low-HS (vs. high-HS) and for high-BS (vs. low-BS) in terms of both likability and STSA, as was predicted in Hypotheses 4 and 5. Analyses were analogous to those in Study 2, but only participants’ BS and HS were used as concurrent predictors.
Results of multiple regression analyses predicting relative preferences in likability and short-term sexual attraction for low- (vs. high-) HS profiles and high- (vs. low-) BS profiles from participants’ hostile sexism and benevolent sexism (Study 3).
Relative Preference for Low-HS
vs. High-HS Profiles
Relative Preference for High-BS
vs. Low-BS Profiles
Short-Term Sexual Attraction
Relative Preference for Low-HS
vs. High-HS Profiles
Relative Preference for High-BS
vs. Low-BS Profiles
The bottom part of Table 4 shows that there was a marginal effect of participants’ BS on their relative preferences in sexual attraction for low-hostile (vs. high-hostile) profiles (β = −.20, p = .055); this paralleled the effect on relative preferences in the likability ratings. No other moderating effects of participants’ sexist attitudes were found for the sexual attraction ratings, all p > .51.
Equivocal Egalitarianism and Perceptions of the Ambivalent Sexist Profile
We computed two continuous measures of equivocal egalitarianism, one based on the general likability index (as in Study 2) and one based on the STSA index, each time subtracting the score for the hostile sexist from the respective score for the benevolent sexist. Recall that Hypothesis 7 predicted a negative correlation of equivocal egalitarianism and typicality ratings for the ambivalent sexist profile. Correlation analyses showed that the two equivocal egalitarianism measures were highly intercorrelated, r(104) = .76, p < .001, but neither measure was related to the typicality index for the ambivalent sexist profile, r(104) = .13 and .06, respectively, both p > .19. Hypothesis 7 thus again received no support.
The inclusion of new items assessing aspects of physical attraction and sexual interest allowed us to differentiate two liking indices in Study 3. The first index (items “likable” and “long-term romantic partner”) was identical to the likability index used in Study 2 and combined general liking with long-term romantic interest. The second index (items “good-looking”, “attractive”, “one-night stand”, and “passionate”) was interpreted as reflecting the profiles’ short-term sexual attraction for participants. A caveat is in place regarding the conceptual distinctiveness of the two indices, because a one-dimensional interpretation of the six liking items would have been feasible. We still found it useful, for exploratory purposes, to distinguish between a more general, long-term liking measure and short-term sexual attraction.
For the likability index, the pattern observed in Study 3 showed remarkable consistency with the results of Study 2 (see Tables 1 and 3). Supporting Hypotheses 1 and 2, male targets’ low HS and high BS again each led to higher ratings of likability. Female students thus again rated the purely benevolent sexist as the most likable, and the purely hostile sexist as the least likable. Ratings for the nonsexist and ambivalent sexist again fell in between, but in Study 3 the ambivalent sexist was rated as significantly more likable than the nonsexist.
The sexual attraction ratings, although generally lower than the likability ratings, showed parallel influences of the profiles’ BS and HS, lending further support to Hypotheses 1 and 2. The way in which women see high-BS (and low-HS) men as “sexy” thus seems to be in terms of general likability and suitability as a long-term partner, as well as in terms of short-term sexual interest. Remarkably, again women perceived the benevolent sexist as more likable and as more sexually attractive than the nonsexist. Viewed from an evolutionary perspective (e.g., Buss and Schmitt 1993), one may conclude that displaying benevolent sexist behavior may have adaptive advantages for men in terms of both long-term and short-term mating strategies (see also Li and Kenrick 2006).
Regression analyses showed that women’s own sexist attitudes moderated the effects of the profiles’ BS and HS on likability and sexual attraction. Consistent with Study 2, it was participants’ BS that amplified a preference for males holding benevolent sexist attitudes. Somewhat surprisingly, there was a trend for high-BS women to be less bothered by men displaying a high level of hostile sexism.
Compared to Study 2, we note the relatively higher typicality ratings for the nonsexist profile. It is possible that this profile in particular was rendered more realistic by the inclusion of filler items featuring positive response values. We had suspected that the rather low evaluation of the nonsexist in Study 2 may have been caused by heuristic processing of the "negative" responses shown by this profile. However, although the inclusion of filler items did increase the nonsexist profile's perceived typicality, it did not change its relative evaluation compared to the other profiles. The benevolent sexist profile was still seen as clearly more likable, and now also as more sexually attractive, than the nonsexist profile. Furthermore, the ambivalent sexist profile, though less clearly than in Study 2, was again rated as high in typicality.
Taken together, our research conveys both good news and bad news. The good news is that women do not seem to lack awareness of the empirical link between BS and HS, contrary to what previous research using a different methodology may have suggested (cf. Kilianski and Rudman 1998). In two studies, female students rated ambivalent sexist male ASI profiles as highly typical of men in general, and rated purely benevolent sexist ASI profiles as the most atypical. The bad news is that in spite of this awareness, participants still approved of profiles expressing high benevolence at least as much as they approved of nonsexist profiles; and they judged a purely benevolent sexist profile even to be significantly more likable and sexually attractive than a nonsexist profile. Thus, when women are assured that a man’s benevolent sexism is not accompanied by hostile sexism, on average they find him highly attractive, even more attractive than a completely nonsexist man.
Is this pattern problematic? Does liking benevolent sexists but knowing they are rare birds strengthen the status quo of male dominance? When women say they like benevolent sexists, do they approve of those men who blamed victims of sexual violence in research by Abrams et al. (2003)? Do they support attitudes in men that lead to the kinds of remarks that hamper women's performance (Dardenne et al. 2007) or contribute to sexual harassment (Fiske and Glick 1995)? This appears plausible, but further research is needed to corroborate such conclusions. For example, one might conceptually replicate the present studies in a more realistic context in which men behave in a nonsexist, or purely benevolent, or ambivalent-sexist way (for a methodology that might be adapted for such use, see Siebler et al. 2008). If, under such circumstances, women still favor benevolent and ambivalent sexists over nonsexists, a clearer case could be made for their strengthening the status quo and supporting the subordination of women (cf. Jost and Kay 2005; Kilianski and Rudman 1998). Another realistic context in which it might be fruitful to pursue these ideas further is the study of human mating preferences and behavior (Buss and Schmitt, 1993). If a mere set of ASI responses can elicit large differences in the perceived sexual attractiveness of the person who allegedly gave these responses, how much more powerful must be men’s vivid displays of paternalistic “chivalrous” behavior in daily interaction?
We found moderating effects of women's own BS (Studies 2 and 3) on their approval of the target profiles' benevolence (and even on their tolerance for the target profiles’ hostile sexism). Study 2 also showed that feminist attitudes may increase the preference for low hostility in men. We had also predicted that high feminism would reduce women’s preference for benevolent sexist males. However, this prediction was not supported by our data. These findings may offer suggestions for potential interventions. The aim of educating women toward approving less of male BS might be reached by questioning their own benevolent sexist beliefs, whereas fostering more feminist attitudes may in itself not be sufficient. More optimistic findings in this regard, however, were reported by Branscombe and Deaux (1991): Women whose feminist attitudes had been made accessible by asking them to complete a feminism scale reported lower intentions than women in a control condition to act in a way that invites male benevolent sexist behavior (e.g., waiting for a man to open a door for them; p. 413).
It would further be useful to replicate our studies in other cultures. In the Introduction we addressed issues of culture and tentatively concluded that the USA and Germany were quite similar in terms of gender-related attitudes and societal development. This insight, and the present Study 1, strengthened our conviction that results produced with our new methodology would not be culture-specific. Yet neither Germany nor the USA are representative for the world at large, and it would be exciting to see our new method of assessing approval for others’ sexism applied to different research questions in different cultural contexts.
A potential limitation of our methodological approach is that asking participants to rate four target profiles in a fully within-subjects design may exaggerate the perception of differences among the profiles. It would thus be worthwile to conduct conceptual replications in which independent groups of participants rate one target profile each.
Finally, while the focus of the present investigation was on further developing the methodology to assess perceptions of sexism in others, future research should pay closer attention to conceptual questions. For example, might the fact that benevolent sexist men are viewed as rare contribute to their greater perceived likability? Although in our present data, the ratings of typicality and likability were generally unrelated, several lines of research suggest that perceived scarcity of an object or attribute may increase its perceived value (e.g., Cialdini 1993) or enhance the extremity of evaluations (Ditto and Jemmott 1989).
Using ASI profiles as stimulus materials offers a useful and well-controlled approach to studying women's perceptions of men's sexist attitudes (or, generally speaking, for studying any group’s perception of any target’s sexist attitudes). Using this approach to study female students' perceptions of male targets' likability, sexual attraction, and typicality yielded interesting new insights. Not only did female students perceive the coexistence of benevolent and hostile sexist beliefs in a single male individual as highly typical, they also approved more of males exhibiting high rather than low benevolent sexism. Most remarkably, when a male target’s high benevolent sexism was explicitly accompanied by low hostile sexism, women tended to like this purely benevolent-sexist man even more than a completely nonsexist man.
Preliminary reports of this research were given as invited presentations at Pontificia Universidad Católica de Chile, Santiago, Chile (September 2004), the University of Halle-Wittenberg, Germany (June 2006), the University of Basel, Switzerland (April 2007), the University of Magdeburg, Germany (May 2007), the Universities of Cape Town and Pretoria, South Africa (August 2007), and the University of Graz, Austria (March 2008). We thank Caroline Erdmann for her help with the data collection and analyses in Study 1, and the "Astoria" cinema, Bielefeld, for providing free tickets as incentives for participants in Study 2. Thanks also to Friederike Eyssel and Nina Vanselow, who provided helpful comments on a previous draft.